2023
Li, Tong; Li, Qingguo
A systematic review on load carriage assistive devices: Mechanism design and performance evaluation Journal Article
In: Mechanism and Machine Theory, vol. 180, pp. 105142, 2023, ISSN: 0094-114X.
Abstract | Links | BibTeX | Tags: Assistive device, Exoskeleton, Human power augmentation, Load carriage, Performance evaluation
@article{LI2023105142,
title = {A systematic review on load carriage assistive devices: Mechanism design and performance evaluation},
author = {Tong Li and Qingguo Li},
url = {https://www.sciencedirect.com/science/article/pii/S0094114X22003883},
doi = {https://doi.org/10.1016/j.mechmachtheory.2022.105142},
issn = {0094-114X},
year = {2023},
date = {2023-11-01},
urldate = {2023-01-01},
journal = {Mechanism and Machine Theory},
volume = {180},
pages = {105142},
abstract = {Manually carrying loads for a long distance and duration is highly physically demanding for humans but commonly required in outdoor activities. With the rapid growth and advancement of wearable robotics, many explorations have been conducted to design and evaluate robotic systems, aiming to enhance human capacity and/or endurance in load carriage tasks. In this paper, state-of-the-art robotic load carriage assistive devices are systematically reviewed regarding the mechanisms and their performances in experimental evaluation. Methods and strategies of assisting load carriage are analyzed and the existing devices are categorized into four classes: (1) reducing inertial force of the load, (2) transferring load weight to the ground, (3) assisting human joints, and (4) integration of transferring load weight and assisting human joints. The efficacy of different mechanism designs and assisting strategies are discussed from both engineering and biomechanical perspectives. Based on these analyses, challenges in the current studies and further efforts required in developing better assistive devices are discussed to provide insights for future studies.},
keywords = {Assistive device, Exoskeleton, Human power augmentation, Load carriage, Performance evaluation},
pubstate = {published},
tppubtype = {article}
}
Manually carrying loads for a long distance and duration is highly physically demanding for humans but commonly required in outdoor activities. With the rapid growth and advancement of wearable robotics, many explorations have been conducted to design and evaluate robotic systems, aiming to enhance human capacity and/or endurance in load carriage tasks. In this paper, state-of-the-art robotic load carriage assistive devices are systematically reviewed regarding the mechanisms and their performances in experimental evaluation. Methods and strategies of assisting load carriage are analyzed and the existing devices are categorized into four classes: (1) reducing inertial force of the load, (2) transferring load weight to the ground, (3) assisting human joints, and (4) integration of transferring load weight and assisting human joints. The efficacy of different mechanism designs and assisting strategies are discussed from both engineering and biomechanical perspectives. Based on these analyses, challenges in the current studies and further efforts required in developing better assistive devices are discussed to provide insights for future studies.
Woods, Joshua E.; Yang, Yuan-Sen; Lau, David T.; Erochko, Jeffrey
Towards the Use of Digital Image Correlation to Measure FRP-Concrete Debonding in Reinforced Concrete Structures Inproceedings
In: Benmokrane, Brahim; Mohamed, Khaled; Farghaly, Ahmed; Mohamed, Hamdy (Ed.): 8th International Conference on Advanced Composite Materials in Bridges and Structures, pp. 261–268, Springer International Publishing, Cham, 2023, ISBN: 978-3-031-09632-7.
Abstract | Links | BibTeX | Tags:
@inproceedings{10.1007/978-3-031-09632-7_30,
title = {Towards the Use of Digital Image Correlation to Measure FRP-Concrete Debonding in Reinforced Concrete Structures},
author = {Joshua E. Woods and Yuan-Sen Yang and David T. Lau and Jeffrey Erochko},
editor = {Brahim Benmokrane and Khaled Mohamed and Ahmed Farghaly and Hamdy Mohamed},
url = {https://link.springer.com/chapter/10.1007/978-3-031-09632-7_30#citeas},
isbn = {978-3-031-09632-7},
year = {2023},
date = {2023-09-27},
urldate = {2023-01-01},
booktitle = {8th International Conference on Advanced Composite Materials in Bridges and Structures},
pages = {261--268},
publisher = {Springer International Publishing},
address = {Cham},
abstract = {This paper presents a simple and cost-effective method to detect and measure debonding in reinforced concrete (RC) structures retrofitted with externally bonded fiber-reinforced polymer (FRP) sheets. Debonding is a phenomenon that occurs when the FRP sheet separates or delaminates from the concrete substrate. Digital image correlation has been used extensively in the past to measure displacement, strain, and crack fields over the surface of a structural element. In this study, a tool is developed in the freely-available image analysis software ImPro Stereo to allow for the measurement of the out-of-plane displacement field over the surface of a structural element. This tool is then applied to detect areas over the surface of a structural element where FRP-concrete debonding has occurred. Validation of the developed tool in ImPro Stereo in a controlled laboratory environment demonstrates that it provides a non-destructive, non-contact, and cost-effective solution to measure FRP-concrete debonding. The results show great potential for its application in structural condition assessment and structural health monitoring of civil structures.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper presents a simple and cost-effective method to detect and measure debonding in reinforced concrete (RC) structures retrofitted with externally bonded fiber-reinforced polymer (FRP) sheets. Debonding is a phenomenon that occurs when the FRP sheet separates or delaminates from the concrete substrate. Digital image correlation has been used extensively in the past to measure displacement, strain, and crack fields over the surface of a structural element. In this study, a tool is developed in the freely-available image analysis software ImPro Stereo to allow for the measurement of the out-of-plane displacement field over the surface of a structural element. This tool is then applied to detect areas over the surface of a structural element where FRP-concrete debonding has occurred. Validation of the developed tool in ImPro Stereo in a controlled laboratory environment demonstrates that it provides a non-destructive, non-contact, and cost-effective solution to measure FRP-concrete debonding. The results show great potential for its application in structural condition assessment and structural health monitoring of civil structures.
Chawla, Ishan; Pathak, P. M.; Notash, Leila; Samantaray, A. K.; Li, Qingguo; Sharma, U. K.
Inverse and Forward Kineto-Static Solution of a Large-Scale Cable-Driven Parallel Robot using Neural Networks Journal Article
In: Mechanism and Machine Theory, vol. 179, pp. 105107, 2023, ISSN: 0094-114X.
Abstract | Links | BibTeX | Tags: Cable-Driven Parallel Robot, Forward Kineto-Static, Inverse Kineto-Static, Neural network, Sagging Cables
@article{CHAWLA2023105107,
title = {Inverse and Forward Kineto-Static Solution of a Large-Scale Cable-Driven Parallel Robot using Neural Networks},
author = {Ishan Chawla and P. M. Pathak and Leila Notash and A. K. Samantaray and Qingguo Li and U. K. Sharma},
url = {https://www.sciencedirect.com/science/article/pii/S0094114X22003536},
doi = {https://doi.org/10.1016/j.mechmachtheory.2022.105107},
issn = {0094-114X},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {Mechanism and Machine Theory},
volume = {179},
pages = {105107},
abstract = {This kineto-static problem of cable-driven parallel robot (CDPR) considering cable mass and elasticity is a highly non-linear and computationally complex problem. Numerical methods are conventionally employed to solve this problem. However, numerical methods are highly dependent on initial values, iterative in nature, computationally unbounded, and have slow convergence speed, making them unreliable for real-time scenarios. Therefore, this paper proposes a neural network-based approach to obtain the inverse kineto-static (IKS) as well as the forward kineto-static (FKS) solution of CDPR with significant cable mass and elasticity. The simulation results are presented for two cases, i.e., for a redundantly constrained planar CDPR (RP-CDPR) and a minimally constrained spatial CDPR (MS-CDPR). The training time, dataset requirement and testing error is observed to be significantly lower for MS-CDPR in comparison to RP-CDPR. For both the cases, the proposed approach shows a significant improvement in the computational speed when compared to numerical methods for both IKS and FKS problem. Owing to the advantage of bounded and low computational time, the proposed neural network-based approach is recommended for use in real-time applications.},
keywords = {Cable-Driven Parallel Robot, Forward Kineto-Static, Inverse Kineto-Static, Neural network, Sagging Cables},
pubstate = {published},
tppubtype = {article}
}
This kineto-static problem of cable-driven parallel robot (CDPR) considering cable mass and elasticity is a highly non-linear and computationally complex problem. Numerical methods are conventionally employed to solve this problem. However, numerical methods are highly dependent on initial values, iterative in nature, computationally unbounded, and have slow convergence speed, making them unreliable for real-time scenarios. Therefore, this paper proposes a neural network-based approach to obtain the inverse kineto-static (IKS) as well as the forward kineto-static (FKS) solution of CDPR with significant cable mass and elasticity. The simulation results are presented for two cases, i.e., for a redundantly constrained planar CDPR (RP-CDPR) and a minimally constrained spatial CDPR (MS-CDPR). The training time, dataset requirement and testing error is observed to be significantly lower for MS-CDPR in comparison to RP-CDPR. For both the cases, the proposed approach shows a significant improvement in the computational speed when compared to numerical methods for both IKS and FKS problem. Owing to the advantage of bounded and low computational time, the proposed neural network-based approach is recommended for use in real-time applications.
2022
Roy, Shuvendu; Etemad, Ali
Impact of Labelled Set Selection and Supervision Policies on Semi-supervised Learning Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2211.14912,
title = {Impact of Labelled Set Selection and Supervision Policies on Semi-supervised Learning},
author = {Shuvendu Roy and Ali Etemad},
url = {https://arxiv.org/abs/2211.14912},
doi = {10.48550/ARXIV.2211.14912},
year = {2022},
date = {2022-11-27},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {In semi-supervised representation learning frameworks, when the number of labelled data is very scarce, the quality and representativeness of these samples become increasingly important. Existing literature on semi-supervised learning randomly sample a limited number of data points for labelling. All these labelled samples are then used along with the unlabelled data throughout the training process. In this work, we ask two important questions in this context: (1) does it matter which samples are selected for labelling? (2) does it matter how the labelled samples are used throughout the training process along with the unlabelled data? To answer the first question, we explore a number of unsupervised methods for selecting specific subsets of data to label (without prior knowledge of their labels), with the goal of maximizing representativeness w.r.t. the unlabelled set. Then, for our second line of inquiry, we define a variety of different label injection strategies in the training process. Extensive experiments on four popular datasets, CIFAR-10, CIFAR-100, SVHN, and STL-10, show that unsupervised selection of samples that are more representative of the entire data improves performance by up to ~2% over the existing semi-supervised frameworks such as MixMatch, ReMixMatch, FixMatch and others with random sample labelling. We show that this boost could even increase to 7.5% for very few-labelled scenarios. However, our study shows that gradually injecting the labels throughout the training procedure does not impact the performance considerably versus when all the existing labels are used throughout the entire training. },
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
In semi-supervised representation learning frameworks, when the number of labelled data is very scarce, the quality and representativeness of these samples become increasingly important. Existing literature on semi-supervised learning randomly sample a limited number of data points for labelling. All these labelled samples are then used along with the unlabelled data throughout the training process. In this work, we ask two important questions in this context: (1) does it matter which samples are selected for labelling? (2) does it matter how the labelled samples are used throughout the training process along with the unlabelled data? To answer the first question, we explore a number of unsupervised methods for selecting specific subsets of data to label (without prior knowledge of their labels), with the goal of maximizing representativeness w.r.t. the unlabelled set. Then, for our second line of inquiry, we define a variety of different label injection strategies in the training process. Extensive experiments on four popular datasets, CIFAR-10, CIFAR-100, SVHN, and STL-10, show that unsupervised selection of samples that are more representative of the entire data improves performance by up to ~2% over the existing semi-supervised frameworks such as MixMatch, ReMixMatch, FixMatch and others with random sample labelling. We show that this boost could even increase to 7.5% for very few-labelled scenarios. However, our study shows that gradually injecting the labels throughout the training procedure does not impact the performance considerably versus when all the existing labels are used throughout the entire training.
Sarkar, Pritam; Etemad, Ali
XKD: Cross-modal Knowledge Distillation with Domain Alignment for Video Representation Learning Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2211.13929,
title = {XKD: Cross-modal Knowledge Distillation with Domain Alignment for Video Representation Learning},
author = {Pritam Sarkar and Ali Etemad},
url = {https://arxiv.org/abs/2211.13929},
doi = {10.48550/ARXIV.2211.13929},
year = {2022},
date = {2022-11-25},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {We present XKD, a novel self-supervised framework to learn meaningful representations from unlabelled video clips. XKD is trained with two pseudo tasks. First, masked data reconstruction is performed to learn modality-specific representations. Next, self-supervised cross-modal knowledge distillation is performed between the two modalities through teacher-student setups to learn complementary information. To identify the most effective information to transfer and also to tackle the domain gap between audio and visual modalities which could hinder knowledge transfer, we introduce a domain alignment strategy for effective cross-modal distillation. Lastly, to develop a general-purpose solution capable of handling both audio and visual streams, a modality-agnostic variant of our proposed framework is introduced, which uses the same backbone for both audio and visual modalities. Our proposed cross-modal knowledge distillation improves linear evaluation top-1 accuracy of video action classification by 8.4% on UCF101, 8.1% on HMDB51, 13.8% on Kinetics-Sound, and 14.2% on Kinetics400. Additionally, our modality-agnostic variant shows promising results in developing a general-purpose network capable of handling different data streams. The code is released on the project website. },
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
We present XKD, a novel self-supervised framework to learn meaningful representations from unlabelled video clips. XKD is trained with two pseudo tasks. First, masked data reconstruction is performed to learn modality-specific representations. Next, self-supervised cross-modal knowledge distillation is performed between the two modalities through teacher-student setups to learn complementary information. To identify the most effective information to transfer and also to tackle the domain gap between audio and visual modalities which could hinder knowledge transfer, we introduce a domain alignment strategy for effective cross-modal distillation. Lastly, to develop a general-purpose solution capable of handling both audio and visual streams, a modality-agnostic variant of our proposed framework is introduced, which uses the same backbone for both audio and visual modalities. Our proposed cross-modal knowledge distillation improves linear evaluation top-1 accuracy of video action classification by 8.4% on UCF101, 8.1% on HMDB51, 13.8% on Kinetics-Sound, and 14.2% on Kinetics400. Additionally, our modality-agnostic variant shows promising results in developing a general-purpose network capable of handling different data streams. The code is released on the project website.
Zheng, Size; Li, Qingguo; Liu, Tao
Multi-phase optimisation model predicts manual lifting motions with less reliance on experiment-based posture data Journal Article
In: Ergonomics, vol. 0, no. 0, pp. 1-16, 2022, (PMID: 36398736).
Abstract | Links | BibTeX | Tags:
@article{doi:10.1080/00140139.2022.2150322,
title = {Multi-phase optimisation model predicts manual lifting motions with less reliance on experiment-based posture data},
author = {Size Zheng and Qingguo Li and Tao Liu},
url = {https://doi.org/10.1080/00140139.2022.2150322},
doi = {10.1080/00140139.2022.2150322},
year = {2022},
date = {2022-11-16},
urldate = {2022-01-01},
journal = {Ergonomics},
volume = {0},
number = {0},
pages = {1-16},
publisher = {Taylor & Francis},
abstract = {Optimisation-based predictive models are widely-used to explore the lifting strategies. Existing models incorporated empirical subject-specific posture constraints to improve the prediction accuracy. However, over-reliance on these constraints limits the application of predictive models. This paper proposed a multi-phase optimisation method (MPOM) for two-dimensional sagittally symmetric semi-squat lifting prediction, which decomposes the complete lifting task into three phases—the initial posture, the final posture, and the dynamic lifting phase. The first two phases are predicted with force- and stability-related strategies, and the last phase is predicted with a smoothing-related objective. Box-lifting motions of different box initial heights were collected for validation. The results show that MPOM has better or similar accuracy than the traditional single-phase optimisation (SPOM) of minimum muscular utilisation ratio, and MPOM reduces the reliance on experimental data. MPOM offers the opportunity to improve accuracy at the expense of efforts to determine appropriate weightings in the posture prediction phases. },
note = {PMID: 36398736},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Optimisation-based predictive models are widely-used to explore the lifting strategies. Existing models incorporated empirical subject-specific posture constraints to improve the prediction accuracy. However, over-reliance on these constraints limits the application of predictive models. This paper proposed a multi-phase optimisation method (MPOM) for two-dimensional sagittally symmetric semi-squat lifting prediction, which decomposes the complete lifting task into three phases—the initial posture, the final posture, and the dynamic lifting phase. The first two phases are predicted with force- and stability-related strategies, and the last phase is predicted with a smoothing-related objective. Box-lifting motions of different box initial heights were collected for validation. The results show that MPOM has better or similar accuracy than the traditional single-phase optimisation (SPOM) of minimum muscular utilisation ratio, and MPOM reduces the reliance on experimental data. MPOM offers the opportunity to improve accuracy at the expense of efforts to determine appropriate weightings in the posture prediction phases.
Cimolino, Gabriele; Graham, T. C. Nicholas; Levin, Laura; Wales, Michaelah; Wheeler, Michael
You Should Have Stayed Home: How to Captivate an Audience in VR Theatre Inproceedings
In: Extended Abstracts of the 2022 Annual Symposium on Computer-Human Interaction in Play, pp. 333–338, Association for Computing Machinery, Bremen, Germany, 2022, ISBN: 9781450392112.
Abstract | Links | BibTeX | Tags: Virtual Reality, VR Play, VR Theatre
@inproceedings{10.1145/3505270.3558383,
title = {You Should Have Stayed Home: How to Captivate an Audience in VR Theatre},
author = {Gabriele Cimolino and T. C. Nicholas Graham and Laura Levin and Michaelah Wales and Michael Wheeler},
url = {https://doi.org/10.1145/3505270.3558383},
doi = {10.1145/3505270.3558383},
isbn = {9781450392112},
year = {2022},
date = {2022-11-07},
urldate = {2022-01-01},
booktitle = {Extended Abstracts of the 2022 Annual Symposium on Computer-Human Interaction in Play},
pages = {333–338},
publisher = {Association for Computing Machinery},
address = {Bremen, Germany},
series = {CHI PLAY '22},
abstract = {Several recent productions have explored how theatrical performances can be performed in virtual reality. Their audiences move around the show’s virtual world, interact with its set, and listen to live performers. These virtual reality performances suggest a new form of digital play that utilizes the audience’s embodiment in the world and participation in its story to immerse them in it. In this paper, we describe how audiences interacted with the virtual reality play You Should Have Stayed Home. We contribute a design space for audio-visual entertainment media, use it to contextualize virtual reality theatre, and discuss how the show’s use of virtual reality technologies shaped how audiences play it.},
keywords = {Virtual Reality, VR Play, VR Theatre},
pubstate = {published},
tppubtype = {inproceedings}
}
Several recent productions have explored how theatrical performances can be performed in virtual reality. Their audiences move around the show’s virtual world, interact with its set, and listen to live performers. These virtual reality performances suggest a new form of digital play that utilizes the audience’s embodiment in the world and participation in its story to immerse them in it. In this paper, we describe how audiences interacted with the virtual reality play You Should Have Stayed Home. We contribute a design space for audio-visual entertainment media, use it to contextualize virtual reality theatre, and discuss how the show’s use of virtual reality technologies shaped how audiences play it.
Gillham, Jack; Hoult, Neil A.; Bentz, Evan C.
Field Test of a Shear Force Measurement Technique Using Fiber Optic Sensing under Variable Speed Truck Loading Journal Article
In: Journal of Bridge Engineering, vol. 27, no. 12, pp. 04022115, 2022.
Abstract | Links | BibTeX | Tags:
@article{doi:10.1061/(ASCE)BE.1943-5592.0001971,
title = {Field Test of a Shear Force Measurement Technique Using Fiber Optic Sensing under Variable Speed Truck Loading},
author = {Jack Gillham and Neil A. Hoult and Evan C. Bentz},
url = {https://ascelibrary.org/doi/abs/10.1061/%28ASCE%29BE.1943-5592.0001971},
doi = {10.1061/(ASCE)BE.1943-5592.0001971},
year = {2022},
date = {2022-11-01},
urldate = {2022-01-01},
journal = {Journal of Bridge Engineering},
volume = {27},
number = {12},
pages = {04022115},
abstract = {The measurement of reaction forces at bridge bearings would enable engineers tasked with maintaining bridges to detect potential damage to the bearing and bridge by detecting changes in the load distribution at the supports with time. Currently, measuring the load in the bearing requires sensors built into the bearing, which means that they are hard to repair when damaged and cannot be installed after the bridge is built (unless the bearings are replaced). A potential alternative is the use of distributed fiber optic sensors (DFOS) that could be used to measure curvature in the beams of a bridge, which can then be used to calculate the moment, shear, and ultimately reaction force due to live loading. To investigate this, a DFOS system was installed on a newly built steel girder bridge on a single beam near one of the piers. A series of load tests were undertaken using a truck with a known load and driving along the bridge directly over the instrumented beam at speeds ranging from pseudo-static up to 30 km/h. The maximum measured strain in the bridge beam was 15 microstrain, which was lower than can be measured with certain DFOS systems, and highlighted the need to select a system with appropriate accuracy and precision. The measured strains were used to calculate the beam shear at the pier as the truck moved across the bridge. These results were compared with a continuous beam and two grillage analyses, and it was found that, based on the continuous beam model, about 25% of the total truck load was being carried by the beam, which was lower than the code live load distribution factor suggested. The grillage models provided better estimates of load spreading but were still conservative and dependent on the choice of transverse stiffness.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The measurement of reaction forces at bridge bearings would enable engineers tasked with maintaining bridges to detect potential damage to the bearing and bridge by detecting changes in the load distribution at the supports with time. Currently, measuring the load in the bearing requires sensors built into the bearing, which means that they are hard to repair when damaged and cannot be installed after the bridge is built (unless the bearings are replaced). A potential alternative is the use of distributed fiber optic sensors (DFOS) that could be used to measure curvature in the beams of a bridge, which can then be used to calculate the moment, shear, and ultimately reaction force due to live loading. To investigate this, a DFOS system was installed on a newly built steel girder bridge on a single beam near one of the piers. A series of load tests were undertaken using a truck with a known load and driving along the bridge directly over the instrumented beam at speeds ranging from pseudo-static up to 30 km/h. The maximum measured strain in the bridge beam was 15 microstrain, which was lower than can be measured with certain DFOS systems, and highlighted the need to select a system with appropriate accuracy and precision. The measured strains were used to calculate the beam shear at the pier as the truck moved across the bridge. These results were compared with a continuous beam and two grillage analyses, and it was found that, based on the continuous beam model, about 25% of the total truck load was being carried by the beam, which was lower than the code live load distribution factor suggested. The grillage models provided better estimates of load spreading but were still conservative and dependent on the choice of transverse stiffness.
Zhang, Chuan; Zhao, Mingyang; Zhu, Liehuang; Zhang, Weiting; Wu, Tong; Ni, Jianbing
FRUIT: A Blockchain-based Efficient and Privacy-preserving Quality-aware Incentive Scheme Journal Article
In: IEEE Journal on Selected Areas in Communications, pp. 1-1, 2022, ISSN: 1558-0008.
Abstract | Links | BibTeX | Tags:
@article{9915368,
title = {FRUIT: A Blockchain-based Efficient and Privacy-preserving Quality-aware Incentive Scheme},
author = {Chuan Zhang and Mingyang Zhao and Liehuang Zhu and Weiting Zhang and Tong Wu and Jianbing Ni},
url = {https://ieeexplore.ieee.org/abstract/document/9915368},
doi = {10.1109/JSAC.2022.3213341},
issn = {1558-0008},
year = {2022},
date = {2022-11-01},
urldate = {2022-11-01},
journal = {IEEE Journal on Selected Areas in Communications},
pages = {1-1},
abstract = {Incentive plays an important role in knowledge discovery, as it impels users to provide high-quality knowledge. To promise incentive schemes with transparency, blockchain technology has been widely used in incentive schemes. Currently, privacy, reliability, streamlined processing, and quality awareness are major challenges in designing blockchain-based incentive schemes. In this paper, we design a blockchain-based eFficient and pRivacy-preserving qUality-aware IncenTive scheme called FRUIT. With well-designed smart contracts, FRUIT achieves privacy, reliability, streamlined processing, and quality awareness during the whole procedure. Specifically, we design a novel lightweight encryption method by combining matrix decomposition with proxy re-encryption and a privacy-preserving task allocation based on the polynomial fitting function and hash function. Then, we leverage our proposed lightweight encryption and task allocation to build an efficient and privacy-preserving knowledge discovery protocol in order to securely calculate the data quality and truthful knowledge. To promise user reliability in the incentive scheme, we utilize the Dirichlet distribution to realize the automatic reputation prediction based on the data quality by deploying the reputation management on the blockchain. Moreover, we also deploy the payment management on the blockchain, endowing the incentive scheme to reward participants based on the data quality automatically. Through a detailed security analysis, we demonstrate that data privacy and task privacy are well preserved during the whole process. Theoretical analysis and extensive experiments on real-world datasets demonstrate that FRUIT has acceptable efficiency and affordable performance in terms of computation cost, communication overhead, and gas consumption.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Incentive plays an important role in knowledge discovery, as it impels users to provide high-quality knowledge. To promise incentive schemes with transparency, blockchain technology has been widely used in incentive schemes. Currently, privacy, reliability, streamlined processing, and quality awareness are major challenges in designing blockchain-based incentive schemes. In this paper, we design a blockchain-based eFficient and pRivacy-preserving qUality-aware IncenTive scheme called FRUIT. With well-designed smart contracts, FRUIT achieves privacy, reliability, streamlined processing, and quality awareness during the whole procedure. Specifically, we design a novel lightweight encryption method by combining matrix decomposition with proxy re-encryption and a privacy-preserving task allocation based on the polynomial fitting function and hash function. Then, we leverage our proposed lightweight encryption and task allocation to build an efficient and privacy-preserving knowledge discovery protocol in order to securely calculate the data quality and truthful knowledge. To promise user reliability in the incentive scheme, we utilize the Dirichlet distribution to realize the automatic reputation prediction based on the data quality by deploying the reputation management on the blockchain. Moreover, we also deploy the payment management on the blockchain, endowing the incentive scheme to reward participants based on the data quality automatically. Through a detailed security analysis, we demonstrate that data privacy and task privacy are well preserved during the whole process. Theoretical analysis and extensive experiments on real-world datasets demonstrate that FRUIT has acceptable efficiency and affordable performance in terms of computation cost, communication overhead, and gas consumption.
Wu, Yangzheng; Javaheri, Alireza; Zand, Mohsen; Greenspan, Michael
Keypoint Cascade Voting for Point Cloud Based 6DoF Pose Estimation Miscellaneous
2022.
Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2210.08123,
title = {Keypoint Cascade Voting for Point Cloud Based 6DoF Pose Estimation},
author = {Yangzheng Wu and Alireza Javaheri and Mohsen Zand and Michael Greenspan},
url = {https://arxiv.org/abs/2210.08123},
doi = {10.48550/ARXIV.2210.08123},
year = {2022},
date = {2022-11-01},
urldate = {2022-01-01},
publisher = {arXiv},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
Li, Xianzhi; Aitken, Will; Zhu, Xiaodan; Thomas, Stephen W.
Learning Better Intent Representations for Financial Open Intent Classification Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computation and Language (cs.CL), Computational Finance (q-fin.CP), FOS: Computer and information sciences, FOS: Economics and business
@misc{https://doi.org/10.48550/arxiv.2210.14304,
title = {Learning Better Intent Representations for Financial Open Intent Classification},
author = {Xianzhi Li and Will Aitken and Xiaodan Zhu and Stephen W. Thomas},
url = {https://arxiv.org/abs/2210.14304},
doi = {10.48550/ARXIV.2210.14304},
year = {2022},
date = {2022-11-01},
urldate = {2022-11-01},
publisher = {arXiv},
abstract = {With the recent surge of NLP technologies in the financial domain, banks and other financial entities have adopted virtual agents (VA) to assist customers. A challenging problem for VAs in this domain is determining a user's reason or intent for contacting the VA, especially when the intent was unseen or open during the VA's training. One method for handling open intents is adaptive decision boundary (ADB) post-processing, which learns tight decision boundaries from intent representations to separate known and open intents. We propose incorporating two methods for supervised pre-training of intent representations: prefix-tuning and fine-tuning just the last layer of a large language model (LLM). With this proposal, our accuracy is 1.63% - 2.07% higher than the prior state-of-the-art ADB method for open intent classification on the banking77 benchmark amongst others. Notably, we only supplement the original ADB model with 0.1% additional trainable parameters. Ablation studies also determine that our method yields better results than full fine-tuning the entire model. We hypothesize that our findings could stimulate a new optimal method of downstream tuning that combines parameter efficient tuning modules with fine-tuning a subset of the base model's layers.},
keywords = {Computation and Language (cs.CL), Computational Finance (q-fin.CP), FOS: Computer and information sciences, FOS: Economics and business},
pubstate = {published},
tppubtype = {misc}
}
With the recent surge of NLP technologies in the financial domain, banks and other financial entities have adopted virtual agents (VA) to assist customers. A challenging problem for VAs in this domain is determining a user's reason or intent for contacting the VA, especially when the intent was unseen or open during the VA's training. One method for handling open intents is adaptive decision boundary (ADB) post-processing, which learns tight decision boundaries from intent representations to separate known and open intents. We propose incorporating two methods for supervised pre-training of intent representations: prefix-tuning and fine-tuning just the last layer of a large language model (LLM). With this proposal, our accuracy is 1.63% - 2.07% higher than the prior state-of-the-art ADB method for open intent classification on the banking77 benchmark amongst others. Notably, we only supplement the original ADB model with 0.1% additional trainable parameters. Ablation studies also determine that our method yields better results than full fine-tuning the entire model. We hypothesize that our findings could stimulate a new optimal method of downstream tuning that combines parameter efficient tuning modules with fine-tuning a subset of the base model's layers.
Li, Jonathan; Bhambhoria, Rohan; Zhu, Xiaodan
Parameter-Efficient Legal Domain Adaptation Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computation and Language (cs.CL), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2210.13712,
title = {Parameter-Efficient Legal Domain Adaptation},
author = {Jonathan Li and Rohan Bhambhoria and Xiaodan Zhu},
url = {https://arxiv.org/abs/2210.13712},
doi = {10.48550/ARXIV.2210.13712},
year = {2022},
date = {2022-11-01},
urldate = {2022-11-01},
publisher = {arXiv},
abstract = {Seeking legal advice is often expensive. Recent advancement in machine learning for solving complex problems can be leveraged to help make legal services more accessible to the public. However, real-life applications encounter significant challenges. State-of-the-art language models are growing increasingly large, making parameter-efficient learning increasingly important. Unfortunately, parameter-efficient methods perform poorly with small amounts of data, which are common in the legal domain (where data labelling costs are high). To address these challenges, we propose parameter-efficient legal domain adaptation, which uses vast unsupervised legal data from public legal forums to perform legal pre-training. This method exceeds or matches the fewshot performance of existing models such as LEGAL-BERT on various legal tasks while tuning only approximately 0.1% of model parameters. Additionally, we show that our method can achieve calibration comparable to existing methods across several tasks. To the best of our knowledge, this work is among the first to explore parameter-efficient methods of tuning language models toward the legal domain. },
keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
Seeking legal advice is often expensive. Recent advancement in machine learning for solving complex problems can be leveraged to help make legal services more accessible to the public. However, real-life applications encounter significant challenges. State-of-the-art language models are growing increasingly large, making parameter-efficient learning increasingly important. Unfortunately, parameter-efficient methods perform poorly with small amounts of data, which are common in the legal domain (where data labelling costs are high). To address these challenges, we propose parameter-efficient legal domain adaptation, which uses vast unsupervised legal data from public legal forums to perform legal pre-training. This method exceeds or matches the fewshot performance of existing models such as LEGAL-BERT on various legal tasks while tuning only approximately 0.1% of model parameters. Additionally, we show that our method can achieve calibration comparable to existing methods across several tasks. To the best of our knowledge, this work is among the first to explore parameter-efficient methods of tuning language models toward the legal domain.
Hungler, Paul; Chabot, Michael; Mirzoian, Lev; Moozeh, Kimia
Virtual Reality Broken Plant Tutorial for Capstone Design Journal Article
In: Proceedings of the Canadian Engineering Education Association (CEEA), 2022.
Abstract | Links | BibTeX | Tags:
@article{Hungler_Chabot_Mirzoian_Moozeh_2022,
title = {Virtual Reality Broken Plant Tutorial for Capstone Design},
author = {Paul Hungler and Michael Chabot and Lev Mirzoian and Kimia Moozeh},
url = {https://ojs.library.queensu.ca/index.php/PCEEA/article/view/15949},
year = {2022},
date = {2022-11-01},
urldate = {2022-11-01},
journal = {Proceedings of the Canadian Engineering Education Association (CEEA)},
abstract = {Providing student capstone teams with an opportunity to demonstrate their design competency and complete a team challenge inside a full-scale chemical processing plant provides a novel learning opportunity. A full-scale virtual reality (VR) chemical processing plant was built as an immersive learning application for the Chemical Engineering and Engineering Chemistry capstone design course at Queen’s University in Kingston. During their capstone design course, student teams are required to work on the design of an ampicillin processing facility and the VR application was designed to provide a high-fidelity representation of an operating plant, including a piping and instrumentation diagram (P&ID). To examine the impact of this new VR learning tool, a broken plant exercise was designed to examine the efficacy of the application. The student cohort was divided into two groups for the exercise; paper-based, and web VR based. Each group completed several surveys and a tutorial problem. Initial quantitative results comparing the survey responses from both groups will be presented.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Providing student capstone teams with an opportunity to demonstrate their design competency and complete a team challenge inside a full-scale chemical processing plant provides a novel learning opportunity. A full-scale virtual reality (VR) chemical processing plant was built as an immersive learning application for the Chemical Engineering and Engineering Chemistry capstone design course at Queen’s University in Kingston. During their capstone design course, student teams are required to work on the design of an ampicillin processing facility and the VR application was designed to provide a high-fidelity representation of an operating plant, including a piping and instrumentation diagram (P&ID). To examine the impact of this new VR learning tool, a broken plant exercise was designed to examine the efficacy of the application. The student cohort was divided into two groups for the exercise; paper-based, and web VR based. Each group completed several surveys and a tutorial problem. Initial quantitative results comparing the survey responses from both groups will be presented.
Hungler, Paul; Thurgood, Chris; Marinova, Mariya; White, Steve; Mirzoian, Lev; Thoms, Matthew; Law, Janice; Chabot, Michael; Moozeh, Kimia
Design and Development of an Open-Source Virtual Reality Chemical Processing Plant Journal Article
In: Proceedings of the Canadian Engineering Education Association (CEEA), 2022.
Abstract | Links | BibTeX | Tags:
@article{Hungler_Thurgood_Marinova_White_Mirzoian_Thoms_Law_Chabot_Moozeh_2022,
title = {Design and Development of an Open-Source Virtual Reality Chemical Processing Plant},
author = {Paul Hungler and Chris Thurgood and Mariya Marinova and Steve White and Lev Mirzoian and Matthew Thoms and Janice Law and Michael Chabot and Kimia Moozeh},
url = {https://ojs.library.queensu.ca/index.php/PCEEA/article/view/15939},
year = {2022},
date = {2022-11-01},
urldate = {2022-11-01},
journal = {Proceedings of the Canadian Engineering Education Association (CEEA)},
abstract = {It is challenging to provide students studying in chemical engineering, biotechnology and other related fields with an opportunity to tour and interact with a full-scale chemical processing plant. To address this challenge, an open-sourced virtual reality (VR) chemical processing plant was designed and built to provide students with an experiential learning opportunity. The VR plant is modelled after an ampicillin processing facility complete with a piping and instrumentation diagram (P&ID). The initial student experience inside the VR plant is a tour of the plant, various plant features and unit operations. The tour enables students to freely tour the plant but also engages them in a “Quest” style experience where they need to search for specific areas and components within the plant. An EngPad was designed to provide learners with a help tool to assist their navigation and strengthen their understanding during the VR experience. Experiential learning theory was used to guide the design of the VR application and take students through the four learning modes of concrete experience, reflective observation, abstract conceptualization, and active experimentation. A focus group provided feedback on the design and user interaction of the VR experience. This paper will outline how design features and enhancements were selected based on their connection to experiential learning theory.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
It is challenging to provide students studying in chemical engineering, biotechnology and other related fields with an opportunity to tour and interact with a full-scale chemical processing plant. To address this challenge, an open-sourced virtual reality (VR) chemical processing plant was designed and built to provide students with an experiential learning opportunity. The VR plant is modelled after an ampicillin processing facility complete with a piping and instrumentation diagram (P&ID). The initial student experience inside the VR plant is a tour of the plant, various plant features and unit operations. The tour enables students to freely tour the plant but also engages them in a “Quest” style experience where they need to search for specific areas and components within the plant. An EngPad was designed to provide learners with a help tool to assist their navigation and strengthen their understanding during the VR experience. Experiential learning theory was used to guide the design of the VR application and take students through the four learning modes of concrete experience, reflective observation, abstract conceptualization, and active experimentation. A focus group provided feedback on the design and user interaction of the VR experience. This paper will outline how design features and enhancements were selected based on their connection to experiential learning theory.
Ean, Richard; Rudie, Karen
A Universal Treatment of Architectures and Fusion Rules in Decentralized Discrete-Event Systems Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: electronic engineering, FOS: Electrical engineering, information engineering, Systems and Control (eess.SY)
@misc{https://doi.org/10.48550/arxiv.2210.16511,
title = {A Universal Treatment of Architectures and Fusion Rules in Decentralized Discrete-Event Systems},
author = {Richard Ean and Karen Rudie},
url = {https://arxiv.org/abs/2210.16511},
doi = {10.48550/ARXIV.2210.16511},
year = {2022},
date = {2022-10-29},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {This paper provides a universal treatment of architectures in decentralized discrete-event systems. Current approaches in discrete-event systems do not provide a direct way to compare existing architectures or proposed potentially novel architectures. Determining whether a new architecture is more general than an existing known architecture relies on producing examples ad hoc and on individual inspiration that puts the conditions for solvability in each architecture into some form that admits comparison. From these research efforts, a method has been extracted to yield a universal approach to decentralized discrete-event system architectures and their attendant fusion rules. This treatment provides an easy and direct way to compare the fusion rules -- and hence to compare the strength or generality of the corresponding architectures. },
keywords = {electronic engineering, FOS: Electrical engineering, information engineering, Systems and Control (eess.SY)},
pubstate = {published},
tppubtype = {misc}
}
This paper provides a universal treatment of architectures in decentralized discrete-event systems. Current approaches in discrete-event systems do not provide a direct way to compare existing architectures or proposed potentially novel architectures. Determining whether a new architecture is more general than an existing known architecture relies on producing examples ad hoc and on individual inspiration that puts the conditions for solvability in each architecture into some form that admits comparison. From these research efforts, a method has been extracted to yield a universal approach to decentralized discrete-event system architectures and their attendant fusion rules. This treatment provides an easy and direct way to compare the fusion rules -- and hence to compare the strength or generality of the corresponding architectures.
Jardine, Peter T.; Givigi, Sidney
Flocks, Mobs, and Figure Eights: Swarming as a Lemniscatic Arch Journal Article
In: IEEE Transactions on Network Science and Engineering, pp. 1-12, 2022, ISSN: 2327-4697.
Abstract | Links | BibTeX | Tags:
@article{9931405,
title = {Flocks, Mobs, and Figure Eights: Swarming as a Lemniscatic Arch},
author = {Peter T. Jardine and Sidney Givigi},
url = {https://ieeexplore.ieee.org/abstract/document/9931405},
doi = {10.1109/TNSE.2022.3217460},
issn = {2327-4697},
year = {2022},
date = {2022-10-27},
urldate = {2022-01-01},
journal = {IEEE Transactions on Network Science and Engineering},
pages = {1-12},
abstract = {Inspired by the natural mobbing behavior of birds, this work presents a novel, quasi-distributed swarming strategy called the Dynamic Lemniscatic Arch. It resolves the problem of producing globally-stable, evenly-spaced lemniscate (or, figure-eight) trajectories while relying on local interactions only. Such trajectories are advantageous in applications where energy consumption and mechanical strain must be minimized. Previous work in lemniscate curves has typically relied on predetermined trajectories, rather than on the emergent structure of the swarm. Furthermore, we enrich the traditional 2-dimensional lemniscate plane curve structure by forming an arch in the third dimension. This arch provides more consistent coverage in surveillance type tasks and, with minor variations in parameters, can be used to produce mobbing behavior. The technique relies on time-varying quaternion rotations linked to the positions of dynamically induced virtual agents. We provide a mathematical proof of stability, which demonstrates the swarm converges to the desired geometry. Simulations show that the strategy performs well with multiple agents and in numerous different configurations.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Inspired by the natural mobbing behavior of birds, this work presents a novel, quasi-distributed swarming strategy called the Dynamic Lemniscatic Arch. It resolves the problem of producing globally-stable, evenly-spaced lemniscate (or, figure-eight) trajectories while relying on local interactions only. Such trajectories are advantageous in applications where energy consumption and mechanical strain must be minimized. Previous work in lemniscate curves has typically relied on predetermined trajectories, rather than on the emergent structure of the swarm. Furthermore, we enrich the traditional 2-dimensional lemniscate plane curve structure by forming an arch in the third dimension. This arch provides more consistent coverage in surveillance type tasks and, with minor variations in parameters, can be used to produce mobbing behavior. The technique relies on time-varying quaternion rotations linked to the positions of dynamically induced virtual agents. We provide a mathematical proof of stability, which demonstrates the swarm converges to the desired geometry. Simulations show that the strategy performs well with multiple agents and in numerous different configurations.
Razi, Kamran; Mokogwu, Chiedu N.; Hashtrudi-Zaad, Keyvan
Coupled stability analysis of cascaded leader–follower teleoperation networks using Möbius transformation Journal Article
In: Mechatronics, vol. 88, pp. 102897, 2022, ISSN: 0957-4158.
Abstract | Links | BibTeX | Tags: Absolute stability, Cascaded networks, Möbius transformation, Passivity, Teleoperation, Time delay
@article{RAZI2022102897,
title = {Coupled stability analysis of cascaded leader–follower teleoperation networks using Möbius transformation},
author = {Kamran Razi and Chiedu N. Mokogwu and Keyvan Hashtrudi-Zaad},
url = {https://www.sciencedirect.com/science/article/pii/S0957415822001155},
doi = {https://doi.org/10.1016/j.mechatronics.2022.102897},
issn = {0957-4158},
year = {2022},
date = {2022-10-01},
urldate = {2022-10-01},
journal = {Mechatronics},
volume = {88},
pages = {102897},
abstract = {In this paper, we use the notion of Möbius transformation to develop a chain-wise methodology for the analysis of coupled stability in a cascade of leader–follower networks. The methodology tracks the transmission of the load passivity unit disk through the cascade to the user network. It allows for networks in the cascade that are neither passive nor absolutely stable. Due to its modularity, the proposed methodology is more suitable for dynamic distributed systems in which a network block, such as virtual coupling or delay in the cascade, is modified, added or taken out. We also introduce the notion of “bounded-Γ absolute stability” to verify the coupled stability for any prescribed range of loads, denoted by circles in the scattering domain. To demonstrate the effectiveness of the proposed methodologies, numerical and experimental results for a position–force control architecture under constant time delays are provided.},
keywords = {Absolute stability, Cascaded networks, Möbius transformation, Passivity, Teleoperation, Time delay},
pubstate = {published},
tppubtype = {article}
}
In this paper, we use the notion of Möbius transformation to develop a chain-wise methodology for the analysis of coupled stability in a cascade of leader–follower networks. The methodology tracks the transmission of the load passivity unit disk through the cascade to the user network. It allows for networks in the cascade that are neither passive nor absolutely stable. Due to its modularity, the proposed methodology is more suitable for dynamic distributed systems in which a network block, such as virtual coupling or delay in the cascade, is modified, added or taken out. We also introduce the notion of “bounded-Γ absolute stability” to verify the coupled stability for any prescribed range of loads, denoted by circles in the scattering domain. To demonstrate the effectiveness of the proposed methodologies, numerical and experimental results for a position–force control architecture under constant time delays are provided.
Jang, Jeonghyeon; Lee, Hoon; Kim, Il-Min; Lee, Inkyu
Deep Learning for Multi-User MIMO Systems: Joint Design of Pilot, Limited Feedback, and Precoding Miscellaneous
2022.
Links | BibTeX | Tags: electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Information Theory (cs.IT), Machine Learning (cs.LG), Signal Processing (eess.SP)
@misc{https://doi.org/10.48550/arxiv.2209.10332,
title = {Deep Learning for Multi-User MIMO Systems: Joint Design of Pilot, Limited Feedback, and Precoding},
author = {Jeonghyeon Jang and Hoon Lee and Il-Min Kim and Inkyu Lee},
url = {https://arxiv.org/abs/2209.10332},
doi = {10.48550/ARXIV.2209.10332},
year = {2022},
date = {2022-09-21},
urldate = {2022-09-21},
publisher = {arXiv},
keywords = {electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Information Theory (cs.IT), Machine Learning (cs.LG), Signal Processing (eess.SP)},
pubstate = {published},
tppubtype = {misc}
}
Yuan, Xiaoming; Chen, Jiahui; Zhang, Ning; Ni, Jianbing; Yu, Fei Richard; Leung, Victor C. M.
Digital Twin-Driven Vehicular Task Offloading and IRS Configuration in the Internet of Vehicles Journal Article
In: IEEE Transactions on Intelligent Transportation Systems, pp. 1-15, 2022, ISSN: 1558-0016.
Abstract | Links | BibTeX | Tags:
@article{9899364,
title = {Digital Twin-Driven Vehicular Task Offloading and IRS Configuration in the Internet of Vehicles},
author = {Xiaoming Yuan and Jiahui Chen and Ning Zhang and Jianbing Ni and Fei Richard Yu and Victor C. M. Leung},
url = {https://ieeexplore.ieee.org/abstract/document/9899364},
doi = {10.1109/TITS.2022.3204585},
issn = {1558-0016},
year = {2022},
date = {2022-09-21},
urldate = {2022-09-21},
journal = {IEEE Transactions on Intelligent Transportation Systems},
pages = {1-15},
abstract = {Digital Twin (DT) and Intelligent Reflective Surface (IRS), the most two promising technologies of 6G make the Internet of Vehicles (IoV) more adaptive. However, future autonomous driving needs powerful networking resources and high-quality wireless communications to guarantee the Quality of Service (QoS). Especially considering the time-varying physical operating environments of IoV, it is extremely urgent to improve resource utilization and wireless channel quality. In this work, we propose a Digital Twin-Driven Vehicular Task Offloading and IRS Configuration Framework (DTVIF) to efficiently monitor, learn, and manage the IoV. Specifically, we adopt Mobile Edge Computing (MEC) and IRS to provide augmented computing capacities for vehicles and improve transmission performance when vehicles communicate to MEC servers. DT is employed to achieve real-time data collection and digital representation of physical operating environments of IoV to better support decisions making. In order to reduce the overall delay and energy consumption of DTVIF, we propose a Two-Stage Optimization for Jointly Optimizing Task Offloading and IRS Configuration (TSJTI) algorithm based on Deep Reinforcement Learning (DRL) and Transfer Learning (TFL). In the first stage, we introduce Double Deep $Q$ -learning Networks (DDQN) to find the optimal offloading decision. In the second stage, based on the parameters learned from the first stage, we migrate the parameters from the first stage to find the optimal IRS configuration based on the Deep Deterministic Policy Gradient (DDPG) method. The simulations demonstrate that the proposed algorithm can effectively reduce the processing latency of task offloading and reduce the average energy consumption in DTVIF.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Digital Twin (DT) and Intelligent Reflective Surface (IRS), the most two promising technologies of 6G make the Internet of Vehicles (IoV) more adaptive. However, future autonomous driving needs powerful networking resources and high-quality wireless communications to guarantee the Quality of Service (QoS). Especially considering the time-varying physical operating environments of IoV, it is extremely urgent to improve resource utilization and wireless channel quality. In this work, we propose a Digital Twin-Driven Vehicular Task Offloading and IRS Configuration Framework (DTVIF) to efficiently monitor, learn, and manage the IoV. Specifically, we adopt Mobile Edge Computing (MEC) and IRS to provide augmented computing capacities for vehicles and improve transmission performance when vehicles communicate to MEC servers. DT is employed to achieve real-time data collection and digital representation of physical operating environments of IoV to better support decisions making. In order to reduce the overall delay and energy consumption of DTVIF, we propose a Two-Stage Optimization for Jointly Optimizing Task Offloading and IRS Configuration (TSJTI) algorithm based on Deep Reinforcement Learning (DRL) and Transfer Learning (TFL). In the first stage, we introduce Double Deep $Q$ -learning Networks (DDQN) to find the optimal offloading decision. In the second stage, based on the parameters learned from the first stage, we migrate the parameters from the first stage to find the optimal IRS configuration based on the Deep Deterministic Policy Gradient (DDPG) method. The simulations demonstrate that the proposed algorithm can effectively reduce the processing latency of task offloading and reduce the average energy consumption in DTVIF.
He, Yuanyuan; Tan, Xinyu; Ni, Jianbing; Yang, Laurence T.; Deng, Xianjun
Differentially Private Set Intersection for Asymmetrical ID Alignment Journal Article
In: IEEE Transactions on Information Forensics and Security, pp. 1-1, 2022, ISSN: 1556-6021.
Abstract | Links | BibTeX | Tags:
@article{9895280,
title = {Differentially Private Set Intersection for Asymmetrical ID Alignment},
author = {Yuanyuan He and Xinyu Tan and Jianbing Ni and Laurence T. Yang and Xianjun Deng},
url = {https://ieeexplore.ieee.org/abstract/document/9895280},
doi = {10.1109/TIFS.2022.3207911},
issn = {1556-6021},
year = {2022},
date = {2022-09-19},
urldate = {2022-01-01},
journal = {IEEE Transactions on Information Forensics and Security},
pages = {1-1},
abstract = {Private Set Intersection (PSI) is typically used to achieve ID alignment with protection of IDs in the preparation phase of Vertical Federated Learning (VFL). However, existing PSI approaches are limited to protecting IDs that are outside the intersection of participants, and most ignore the sensitivity of intersection for a weak party in an asymmetrical ID alignment. Since the set size of the strong party is much greater than the weak party’s in an asymmetrical federation, and the intersection usually accounts for a substantial part of the weak party set, the weak party’s sensitive sample IDs would be severely compromised through sharing the intersection. To address this issue, we propose Differentially private PSI Cardinality and PSI (DPSI-CA, DPSI) protocols, which protect the intersection cardinality and sensitive IDs inside the intersect ion for the weak party, respectively. First, DPSI-CA encodes IDs in binary notation, and combines them with the GM encryption, to perform the ID-matchmaking by executing bitwise plaintext XOR. Then, the encrypted matching results are independently perturbed using randomized responses to produce differentially private outputs for PSI-CA, and its unbiased estimate is added to remove the deviation brought by the randomization. Furthermore, DPSI fuses Pseudo-Random Function (PRF)-based zero sharing, garbled Bloom filter, and Oblivious PRF (OPRF)-based shares reconstruction, to successfully reconstruct the shares corresponding to sampled IDs in the intersection. Meanwhile, a randomized response is used to sample the inputs and perturb the outputs of the OPRF-based shares reconstruction, producing a randomly sampled intersection for the weak party and differentially private intersection for the strong party. Finally, the privacy analysis shows that our protocols provide differential privacy for the weak party’s sensitive sample IDs, and extensive experiment results illustrate the feasibility of the asymmetrical ID alignment involving millions of IDs.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Private Set Intersection (PSI) is typically used to achieve ID alignment with protection of IDs in the preparation phase of Vertical Federated Learning (VFL). However, existing PSI approaches are limited to protecting IDs that are outside the intersection of participants, and most ignore the sensitivity of intersection for a weak party in an asymmetrical ID alignment. Since the set size of the strong party is much greater than the weak party’s in an asymmetrical federation, and the intersection usually accounts for a substantial part of the weak party set, the weak party’s sensitive sample IDs would be severely compromised through sharing the intersection. To address this issue, we propose Differentially private PSI Cardinality and PSI (DPSI-CA, DPSI) protocols, which protect the intersection cardinality and sensitive IDs inside the intersect ion for the weak party, respectively. First, DPSI-CA encodes IDs in binary notation, and combines them with the GM encryption, to perform the ID-matchmaking by executing bitwise plaintext XOR. Then, the encrypted matching results are independently perturbed using randomized responses to produce differentially private outputs for PSI-CA, and its unbiased estimate is added to remove the deviation brought by the randomization. Furthermore, DPSI fuses Pseudo-Random Function (PRF)-based zero sharing, garbled Bloom filter, and Oblivious PRF (OPRF)-based shares reconstruction, to successfully reconstruct the shares corresponding to sampled IDs in the intersection. Meanwhile, a randomized response is used to sample the inputs and perturb the outputs of the OPRF-based shares reconstruction, producing a randomly sampled intersection for the weak party and differentially private intersection for the strong party. Finally, the privacy analysis shows that our protocols provide differential privacy for the weak party’s sensitive sample IDs, and extensive experiment results illustrate the feasibility of the asymmetrical ID alignment involving millions of IDs.
Kolahdouzi, Mojtaba; Sepas-Moghaddam, Alireza; Etemad, Ali
FaceTopoNet: Facial Expression Recognition using Face Topology Learning Miscellaneous
2022.
Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2209.06322,
title = {FaceTopoNet: Facial Expression Recognition using Face Topology Learning},
author = {Mojtaba Kolahdouzi and Alireza Sepas-Moghaddam and Ali Etemad},
url = {https://arxiv.org/abs/2209.06322},
doi = {10.48550/ARXIV.2209.06322},
year = {2022},
date = {2022-09-13},
urldate = {2022-01-01},
publisher = {arXiv},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
Park, Jeongeun; Yoon, Taerim; Hong, Jejoon; Yu, Youngjae; Pan, Matthew; Choi, Sungjoon
Active Visual Search in the Wild Miscellaneous
2022.
Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, Robotics (cs.RO)
@misc{https://doi.org/10.48550/arxiv.2209.08803,
title = {Active Visual Search in the Wild},
author = {Jeongeun Park and Taerim Yoon and Jejoon Hong and Youngjae Yu and Matthew Pan and Sungjoon Choi},
url = {https://arxiv.org/abs/2209.08803},
doi = {10.48550/ARXIV.2209.08803},
year = {2022},
date = {2022-09-09},
urldate = {2022-01-01},
publisher = {arXiv},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, Robotics (cs.RO)},
pubstate = {published},
tppubtype = {misc}
}
Hajian, Gelareh; Behinaein, Behnam; Etemad, Ali; Morin, Evelyn
Bagged tree ensemble modelling with feature selection for isometric EMG-based force estimation Journal Article
In: Biomedical Signal Processing and Control, vol. 78, pp. 104012, 2022, ISSN: 1746-8094.
Abstract | Links | BibTeX | Tags: Bagged Tree Ensemble (BTE), Force estimation, High-density (HD) recording, Isometric contraction, Surface electromyogram (EMG)
@article{HAJIAN2022104012,
title = {Bagged tree ensemble modelling with feature selection for isometric EMG-based force estimation},
author = {Gelareh Hajian and Behnam Behinaein and Ali Etemad and Evelyn Morin},
url = {https://www.sciencedirect.com/science/article/pii/S1746809422004876},
doi = {https://doi.org/10.1016/j.bspc.2022.104012},
issn = {1746-8094},
year = {2022},
date = {2022-09-01},
urldate = {2022-01-01},
journal = {Biomedical Signal Processing and Control},
volume = {78},
pages = {104012},
abstract = {EMG-based force estimation is crucial in applications, such as control of powered prosthetic and rehabilitation devices. Most previous studies focus on intra-subject force modelling. However, a generalized EMG-force estimation model, which is capable of estimating force across users, is needed for surgical and rehabilitation robotics. In this study, EMG signals are recorded from the long head and short head of the biceps brachii, and brachioradialis using 3 linear surface electrode arrays, during isometric elbow flexions, at different joint angles and forearm postures, while recording the induced force at the wrist. The recorded EMGs are pre-processed and segmented, and 336 time and frequency domain features are extracted from 21 EMG channels. We explore developing a model that can perform well across subjects, where Bagged Tree Ensemble (BTE) models are used to learn the non-linear, complex relationships between the EMG and force data. The BTE models are compared with several machine learning approaches. The BTE model performs best, giving an average normalized mean squared error (%NMSE) of 5.65±16.24%. To reduce the dimensionality of the feature space and improve force estimation performance, a novel feature selection technique, called modified sequential feature selection (MSFS) is implemented and compared to other commonly used feature selection methods. Results show that the MSFS algorithm outperformed other methods tested, significantly reducing the force estimation error. We also found that there was no effect of forearm posture and joint angle on force modelling accuracy, permitting the development of an isometric force estimation model generalized across participants and arm positions.},
keywords = {Bagged Tree Ensemble (BTE), Force estimation, High-density (HD) recording, Isometric contraction, Surface electromyogram (EMG)},
pubstate = {published},
tppubtype = {article}
}
EMG-based force estimation is crucial in applications, such as control of powered prosthetic and rehabilitation devices. Most previous studies focus on intra-subject force modelling. However, a generalized EMG-force estimation model, which is capable of estimating force across users, is needed for surgical and rehabilitation robotics. In this study, EMG signals are recorded from the long head and short head of the biceps brachii, and brachioradialis using 3 linear surface electrode arrays, during isometric elbow flexions, at different joint angles and forearm postures, while recording the induced force at the wrist. The recorded EMGs are pre-processed and segmented, and 336 time and frequency domain features are extracted from 21 EMG channels. We explore developing a model that can perform well across subjects, where Bagged Tree Ensemble (BTE) models are used to learn the non-linear, complex relationships between the EMG and force data. The BTE models are compared with several machine learning approaches. The BTE model performs best, giving an average normalized mean squared error (%NMSE) of 5.65±16.24%. To reduce the dimensionality of the feature space and improve force estimation performance, a novel feature selection technique, called modified sequential feature selection (MSFS) is implemented and compared to other commonly used feature selection methods. Results show that the MSFS algorithm outperformed other methods tested, significantly reducing the force estimation error. We also found that there was no effect of forearm posture and joint angle on force modelling accuracy, permitting the development of an isometric force estimation model generalized across participants and arm positions.
Etemad, CA) Ali (Kanata
SYSTEMS, METHODS, AND STORAGE MEDIA FOR GENERATING SYNTHESIZED DEPTH DATA
2022.
@{patent:20220292351,
title = {SYSTEMS, METHODS, AND STORAGE MEDIA FOR GENERATING SYNTHESIZED DEPTH DATA},
author = {CA) Ali (Kanata Etemad},
url = {https://www.freepatentsonline.com/y2022/0292351.html},
year = {2022},
date = {2022-09-01},
number = {20220292351},
keywords = {},
pubstate = {published},
tppubtype = {}
}
Antonyshyn, Luka; Silveira, Jefferson; Givigi, Sidney; Marshall, Joshua
Multiple Mobile Robot Task and Motion Planning: A Survey Journal Article
In: ACM Comput. Surv., 2022, ISSN: 0360-0300, (Just Accepted).
Abstract | Links | BibTeX | Tags: autonomous robotics, Autonomous Vehicles, cooperation, mobile robotics, motion planning, Task and motion planning, task planning
@article{10.1145/3564696,
title = {Multiple Mobile Robot Task and Motion Planning: A Survey},
author = {Luka Antonyshyn and Jefferson Silveira and Sidney Givigi and Joshua Marshall},
url = {https://doi.org/10.1145/3564696},
doi = {10.1145/3564696},
issn = {0360-0300},
year = {2022},
date = {2022-09-01},
urldate = {2022-09-01},
journal = {ACM Comput. Surv.},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
abstract = {With recent advances in mobile robotics, autonomous systems, and artificial intelligence, there is a growing expectation that robots are able to solve complex problems. Many of these problems require multiple robots working cooperatively in a multi-robot system. Complex tasks may also include the interconnection of task-level specifications with robot motion-level constraints. Many recent works in the literature use multiple mobile robots to solve these complex tasks by integrating task and motion planning. We survey recent contributions to the field of combined task and motion planning for multiple mobile robots by categorizing works based on their underlying problem representations, and identify possible directions for future research. We propose a taxonomy for task and motion planning based on system capabilities, applicable to multi-robot and single-robot systems.},
note = {Just Accepted},
keywords = {autonomous robotics, Autonomous Vehicles, cooperation, mobile robotics, motion planning, Task and motion planning, task planning},
pubstate = {published},
tppubtype = {article}
}
With recent advances in mobile robotics, autonomous systems, and artificial intelligence, there is a growing expectation that robots are able to solve complex problems. Many of these problems require multiple robots working cooperatively in a multi-robot system. Complex tasks may also include the interconnection of task-level specifications with robot motion-level constraints. Many recent works in the literature use multiple mobile robots to solve these complex tasks by integrating task and motion planning. We survey recent contributions to the field of combined task and motion planning for multiple mobile robots by categorizing works based on their underlying problem representations, and identify possible directions for future research. We propose a taxonomy for task and motion planning based on system capabilities, applicable to multi-robot and single-robot systems.
Antonyshyn, Luka; Silveira, Jefferson; Givigi, Sidney; Marshall, Joshua
Multiple Mobile Robot Task and Motion Planning: A Survey Journal Article
In: ACM Comput. Surv., 2022, ISSN: 0360-0300, (Just Accepted).
Abstract | Links | BibTeX | Tags: autonomous robotics, Autonomous Vehicles, cooperation, mobile robotics, motion planning, Task and motion planning, task planning
@article{10.1145/3564696b,
title = {Multiple Mobile Robot Task and Motion Planning: A Survey},
author = {Luka Antonyshyn and Jefferson Silveira and Sidney Givigi and Joshua Marshall},
url = {https://doi.org/10.1145/3564696},
doi = {10.1145/3564696},
issn = {0360-0300},
year = {2022},
date = {2022-09-01},
urldate = {2022-09-01},
journal = {ACM Comput. Surv.},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
abstract = {With recent advances in mobile robotics, autonomous systems, and artificial intelligence, there is a growing expectation that robots are able to solve complex problems. Many of these problems require multiple robots working cooperatively in a multi-robot system. Complex tasks may also include the interconnection of task-level specifications with robot motion-level constraints. Many recent works in the literature use multiple mobile robots to solve these complex tasks by integrating task and motion planning. We survey recent contributions to the field of combined task and motion planning for multiple mobile robots by categorizing works based on their underlying problem representations, and identify possible directions for future research. We propose a taxonomy for task and motion planning based on system capabilities, applicable to multi-robot and single-robot systems.},
note = {Just Accepted},
keywords = {autonomous robotics, Autonomous Vehicles, cooperation, mobile robotics, motion planning, Task and motion planning, task planning},
pubstate = {published},
tppubtype = {article}
}
With recent advances in mobile robotics, autonomous systems, and artificial intelligence, there is a growing expectation that robots are able to solve complex problems. Many of these problems require multiple robots working cooperatively in a multi-robot system. Complex tasks may also include the interconnection of task-level specifications with robot motion-level constraints. Many recent works in the literature use multiple mobile robots to solve these complex tasks by integrating task and motion planning. We survey recent contributions to the field of combined task and motion planning for multiple mobile robots by categorizing works based on their underlying problem representations, and identify possible directions for future research. We propose a taxonomy for task and motion planning based on system capabilities, applicable to multi-robot and single-robot systems.
Fernando, Heshan; Heykoop, Isabel; Woods, Joshua; Hoult, Neil
Development of an IoT Monitoring System for Bridge Bearing Movement using a MEMS Accelerometer-based Inclination Sensing Inproceedings
In: 2022 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE), pp. 314-318, 2022, ISSN: 2576-7046.
Abstract | Links | BibTeX | Tags:
@inproceedings{9918468,
title = {Development of an IoT Monitoring System for Bridge Bearing Movement using a MEMS Accelerometer-based Inclination Sensing},
author = {Heshan Fernando and Isabel Heykoop and Joshua Woods and Neil Hoult},
url = {https://ieeexplore.ieee.org/abstract/document/9918468},
doi = {10.1109/CCECE49351.2022.9918468},
issn = {2576-7046},
year = {2022},
date = {2022-09-01},
urldate = {2022-09-01},
booktitle = {2022 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)},
pages = {314-318},
abstract = {This paper details the system design, including hardware design and IoT architecture, for monitoring bridge bearing movement using MEMS accelerometer-based inclination sensing. The system has the potential to lower the cost and improve the monitoring of bridge bearings for degradation and failure. A prototype system is developed for deployment on the Waaban Crossing-a 1.2 km two-lane bridge that is currently under construction in Kingston, Ontario, Canada. An IoT system is designed to measure the movement of ten bearings on a single pier, and transmit the data wirelessly to an IoT Hub for monitoring and visualization. Methods for measurement and thermal calibration are presented with preliminary results.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper details the system design, including hardware design and IoT architecture, for monitoring bridge bearing movement using MEMS accelerometer-based inclination sensing. The system has the potential to lower the cost and improve the monitoring of bridge bearings for degradation and failure. A prototype system is developed for deployment on the Waaban Crossing-a 1.2 km two-lane bridge that is currently under construction in Kingston, Ontario, Canada. An IoT system is designed to measure the movement of ten bearings on a single pier, and transmit the data wirelessly to an IoT Hub for monitoring and visualization. Methods for measurement and thermal calibration are presented with preliminary results.
Greisman, Austin; Ivimey, Arjun; Connolly, Laura; Hashtrudi-Zaad, Keyvan
Power-Based Gravity Compensation for Flexible Joint Manipulators Inproceedings
In: 2022 International Symposium on Measurement and Control in Robotics (ISMCR), pp. 1-6, 2022.
Abstract | Links | BibTeX | Tags:
@inproceedings{9950579,
title = {Power-Based Gravity Compensation for Flexible Joint Manipulators},
author = {Austin Greisman and Arjun Ivimey and Laura Connolly and Keyvan Hashtrudi-Zaad},
doi = {10.1109/ISMCR56534.2022.9950579},
year = {2022},
date = {2022-09-01},
urldate = {2022-09-01},
booktitle = {2022 International Symposium on Measurement and Control in Robotics (ISMCR)},
pages = {1-6},
abstract = {In this paper, we propose an add-on dynamic gravity compensator for flexible joint manipulators. The compensator pumps sufficient power that matches the gravity power demand on the manipulator links, resulting in a reduced oscillatory response. The proposed power-based gravity (PG) Compensator uses moving window batch least squares to implement the power match. The performance of the proposed PG Compensator was validated when added to a base-line Proportional Derivative (PD) controller and two PD-based benchmark controllers. In every case, the addition of the compensator improves the system response for regulation operation, allowing for faster convergence to steady-state. Experimental testing was performed showing a 23% transient error improvement over the baseline controller without the power compensator.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we propose an add-on dynamic gravity compensator for flexible joint manipulators. The compensator pumps sufficient power that matches the gravity power demand on the manipulator links, resulting in a reduced oscillatory response. The proposed power-based gravity (PG) Compensator uses moving window batch least squares to implement the power match. The performance of the proposed PG Compensator was validated when added to a base-line Proportional Derivative (PD) controller and two PD-based benchmark controllers. In every case, the addition of the compensator improves the system response for regulation operation, allowing for faster convergence to steady-state. Experimental testing was performed showing a 23% transient error improvement over the baseline controller without the power compensator.
Taghanaki, Setareh Rahimi; Rainbow, Michael; Etemad, Ali
Self-Supervised Human Activity Recognition with Localized Time-Frequency Contrastive Representation Learning Miscellaneous
2022.
Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Machine Learning (cs.LG), Signal Processing (eess.SP)
@misc{https://doi.org/10.48550/arxiv.2209.00990,
title = {Self-Supervised Human Activity Recognition with Localized Time-Frequency Contrastive Representation Learning},
author = {Setareh Rahimi Taghanaki and Michael Rainbow and Ali Etemad},
url = {https://arxiv.org/abs/2209.00990},
doi = {10.48550/ARXIV.2209.00990},
year = {2022},
date = {2022-08-26},
urldate = {2022-01-01},
publisher = {arXiv},
keywords = {Computer Vision and Pattern Recognition (cs.CV), electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Machine Learning (cs.LG), Signal Processing (eess.SP)},
pubstate = {published},
tppubtype = {misc}
}
Liu, Dongxiao; Wu, Huaqing; Huang, Cheng; Ni, Jianbing; Shen, Xuemin
Blockchain-Based Credential Management for Anonymous Authentication in SAGVN Journal Article
In: IEEE Journal on Selected Areas in Communications, vol. 40, no. 10, pp. 3104-3116, 2022, ISSN: 1558-0008.
Abstract | Links | BibTeX | Tags:
@article{9854055,
title = {Blockchain-Based Credential Management for Anonymous Authentication in SAGVN},
author = {Dongxiao Liu and Huaqing Wu and Cheng Huang and Jianbing Ni and Xuemin Shen},
url = {https://ieeexplore.ieee.org/abstract/document/9854055
},
doi = {10.1109/JSAC.2022.3196091},
issn = {1558-0008},
year = {2022},
date = {2022-08-10},
urldate = {2022-08-10},
journal = {IEEE Journal on Selected Areas in Communications},
volume = {40},
number = {10},
pages = {3104-3116},
abstract = {In this paper, we propose a blockchain-based collaborative credential management scheme for anonymous authentication in space-air-ground integrated vehicular networks (SAGVN), named SAG-BC. First, we build a consortium blockchain among service providers and design a distributed system setup (DSS) scheme to securely generate public parameters for issuing credentials. Second, we design a collaborative credential issuance (CCI) scheme to generate a succinct and easy-to-manage subscription credential. The credential can be used by users to access different access points in SAGVN efficiently without revealing true identities from the authentication messages. With co-designs of zero-knowledge proofs and succinct on-chain commitments, SAG-BC provides efficient verifiability and incentives for credential management operations in SAGVN. By doing so, expensive on-chain storage and computational overheads are reduced in the DSS and CCI. Finally, we conduct a thorough security analysis to demonstrate that SAG-BC achieves security and verifiability for credential management in SAGVN. We set up a real-world blockchain network and conduct extensive experiments to show the feasibility and efficiency of SAG-BC.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we propose a blockchain-based collaborative credential management scheme for anonymous authentication in space-air-ground integrated vehicular networks (SAGVN), named SAG-BC. First, we build a consortium blockchain among service providers and design a distributed system setup (DSS) scheme to securely generate public parameters for issuing credentials. Second, we design a collaborative credential issuance (CCI) scheme to generate a succinct and easy-to-manage subscription credential. The credential can be used by users to access different access points in SAGVN efficiently without revealing true identities from the authentication messages. With co-designs of zero-knowledge proofs and succinct on-chain commitments, SAG-BC provides efficient verifiability and incentives for credential management operations in SAGVN. By doing so, expensive on-chain storage and computational overheads are reduced in the DSS and CCI. Finally, we conduct a thorough security analysis to demonstrate that SAG-BC achieves security and verifiability for credential management in SAGVN. We set up a real-world blockchain network and conduct extensive experiments to show the feasibility and efficiency of SAG-BC.
Moulton, Richard Hugh; Rudie, Karen
Online control of discrete-event systems: A survey Journal Article
In: Annual Reviews in Control, 2022, ISSN: 1367-5788.
Abstract | Links | BibTeX | Tags: Artificial Intelligence, Discrete-event systems, Limited lookahead, Online control, Supervisory control
@article{MOULTON2022,
title = {Online control of discrete-event systems: A survey},
author = {Richard Hugh Moulton and Karen Rudie},
url = {https://www.sciencedirect.com/science/article/pii/S1367578822000931},
doi = {https://doi.org/10.1016/j.arcontrol.2022.08.002},
issn = {1367-5788},
year = {2022},
date = {2022-08-06},
urldate = {2022-01-01},
journal = {Annual Reviews in Control},
abstract = {In control theory, as in other areas of engineering research, there is an inherent tension between the breadth of a technique’s applicability and its mathematical tractability. For the area of discrete-event systems (DES), this manifested itself in a theory of supervisory control that originally provided correct-by-construction guarantees for offline solutions to a restricted kind of deterministic process. Follow-on work extended the reach of these techniques to a number of new settings, notably the development of online control without sacrificing any of the original DES performance guarantees. The ability to enact online control opened the door to applying DES techniques to the adaptive control processes presented by modern technologies: processes with dynamic and time-varying natures, whose characteristics may be understood poorly or not at all. Although many works have built on the seminal work of online control in DES, we believe that these ideas have not reached their full potential due to the difficulty in translating them to adjacent fields. In this survey, we look back at 30 years of research concerning the online control of DES and closely related limited lookahead policies with an eye to making the works accessible to practitioners in the broader control theory and artificial intelligence communities. We conclude with some thoughts on future research directions for the further development and application of online DES control techniques to problems requiring intelligent control in our modern world.},
keywords = {Artificial Intelligence, Discrete-event systems, Limited lookahead, Online control, Supervisory control},
pubstate = {published},
tppubtype = {article}
}
In control theory, as in other areas of engineering research, there is an inherent tension between the breadth of a technique’s applicability and its mathematical tractability. For the area of discrete-event systems (DES), this manifested itself in a theory of supervisory control that originally provided correct-by-construction guarantees for offline solutions to a restricted kind of deterministic process. Follow-on work extended the reach of these techniques to a number of new settings, notably the development of online control without sacrificing any of the original DES performance guarantees. The ability to enact online control opened the door to applying DES techniques to the adaptive control processes presented by modern technologies: processes with dynamic and time-varying natures, whose characteristics may be understood poorly or not at all. Although many works have built on the seminal work of online control in DES, we believe that these ideas have not reached their full potential due to the difficulty in translating them to adjacent fields. In this survey, we look back at 30 years of research concerning the online control of DES and closely related limited lookahead policies with an eye to making the works accessible to practitioners in the broader control theory and artificial intelligence communities. We conclude with some thoughts on future research directions for the further development and application of online DES control techniques to problems requiring intelligent control in our modern world.
Roy, Shuvendu; Etemad, Ali
Analysis of Semi-Supervised Methods for Facial Expression Recognition Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2208.00544,
title = {Analysis of Semi-Supervised Methods for Facial Expression Recognition},
author = {Shuvendu Roy and Ali Etemad},
url = {https://arxiv.org/abs/2208.00544},
doi = {10.48550/ARXIV.2208.00544},
year = {2022},
date = {2022-07-31},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {Training deep neural networks for image recognition often requires large-scale human annotated data. To reduce the reliance of deep neural solutions on labeled data, state-of-the-art semi-supervised methods have been proposed in the literature. Nonetheless, the use of such semi-supervised methods has been quite rare in the field of facial expression recognition (FER). In this paper, we present a comprehensive study on recently proposed state-of-the-art semi-supervised learning methods in the context of FER. We conduct comparative study on eight semi-supervised learning methods, namely Pi-Model, Pseudo-label, Mean-Teacher, VAT, MixMatch, ReMixMatch, UDA, and FixMatch, on three FER datasets (FER13, RAF-DB, and AffectNet), when various amounts of labeled samples are used. We also compare the performance of these methods against fully-supervised training. Our study shows that when training existing semi-supervised methods on as little as 250 labeled samples per class can yield comparable performances to that of fully-supervised methods trained on the full labeled datasets. To facilitate further research in this area, we make our code publicly available at: this https URL},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
Training deep neural networks for image recognition often requires large-scale human annotated data. To reduce the reliance of deep neural solutions on labeled data, state-of-the-art semi-supervised methods have been proposed in the literature. Nonetheless, the use of such semi-supervised methods has been quite rare in the field of facial expression recognition (FER). In this paper, we present a comprehensive study on recently proposed state-of-the-art semi-supervised learning methods in the context of FER. We conduct comparative study on eight semi-supervised learning methods, namely Pi-Model, Pseudo-label, Mean-Teacher, VAT, MixMatch, ReMixMatch, UDA, and FixMatch, on three FER datasets (FER13, RAF-DB, and AffectNet), when various amounts of labeled samples are used. We also compare the performance of these methods against fully-supervised training. Our study shows that when training existing semi-supervised methods on as little as 250 labeled samples per class can yield comparable performances to that of fully-supervised methods trained on the full labeled datasets. To facilitate further research in this area, we make our code publicly available at: this https URL
Liu, Yuchen; Moore, Ian; Hoult, Neil
Steel Culvert Investigation from Field Testing to Design Equations Book Chapter
In: Pipelines 2022, pp. 58-65, 2022.
Abstract | Links | BibTeX | Tags:
@inbook{doi:10.1061/9780784484272.008,
title = {Steel Culvert Investigation from Field Testing to Design Equations},
author = {Yuchen Liu and Ian Moore and Neil Hoult},
url = {https://ascelibrary.org/doi/abs/10.1061/9780784484272.008},
doi = {10.1061/9780784484272.008},
year = {2022},
date = {2022-07-30},
urldate = {2022-07-30},
booktitle = {Pipelines 2022},
pages = {58-65},
abstract = {This paper summarizes a recently completed investigation of corrugated steel structures. The project commenced with field testing of two corrugated steel arch culverts in the city of Kingston, Canada, using optical fibers to provide detailed measurements to permit assessment of thrust and moment distributions under static and dynamic truck loading. The detailed responses provided new understanding of the structural behavior, including the critical nature of the front axle of the truck, when normal practice would focus on the tandem axles at the rear of the truck. The study also permitted assessment of the impacts of vehicle speed and the asphalt pavement. Next, improved orthotropic properties were developed considering the arc and tangent geometries of four common corrugation plate sizes, revealing the shortcomings of traditional sinusoidal approximations. Finite element modeling was then used to undertake an extensive parametric investigation to develop new design equations providing improved moment and thrust estimates for single and tandem axle load patterns, at a range of burial depths and corrugation geometries. The paper provides a summary of this research as well as provides insights gained from the investigation.},
keywords = {},
pubstate = {published},
tppubtype = {inbook}
}
This paper summarizes a recently completed investigation of corrugated steel structures. The project commenced with field testing of two corrugated steel arch culverts in the city of Kingston, Canada, using optical fibers to provide detailed measurements to permit assessment of thrust and moment distributions under static and dynamic truck loading. The detailed responses provided new understanding of the structural behavior, including the critical nature of the front axle of the truck, when normal practice would focus on the tandem axles at the rear of the truck. The study also permitted assessment of the impacts of vehicle speed and the asphalt pavement. Next, improved orthotropic properties were developed considering the arc and tangent geometries of four common corrugation plate sizes, revealing the shortcomings of traditional sinusoidal approximations. Finite element modeling was then used to undertake an extensive parametric investigation to develop new design equations providing improved moment and thrust estimates for single and tandem axle load patterns, at a range of burial depths and corrugation geometries. The paper provides a summary of this research as well as provides insights gained from the investigation.
Jardine, Peter Travis; Givigi, Sidney N.
Bimodal Dynamic Swarms Journal Article
In: IEEE Access, vol. 10, pp. 94487-94495, 2022, ISSN: 2169-3536.
Abstract | Links | BibTeX | Tags:
@article{9857917,
title = {Bimodal Dynamic Swarms},
author = {Peter Travis Jardine and Sidney N. Givigi},
doi = {10.1109/ACCESS.2022.3198993},
issn = {2169-3536},
year = {2022},
date = {2022-07-27},
urldate = {2022-01-01},
journal = {IEEE Access},
volume = {10},
pages = {94487-94495},
abstract = {This paper presents a bimodal control strategy for transitioning between different topologies of flocking and dynamic structures. The work is motivated by applications where agents must pursue and close in on a target prior to carrying out more sophisticated tasks. Specifically, it builds on previous work in consensus-based flocking to define a new type of dynamic target pursuit and capturing method called dynamic enspherement. The problem is formulated in terms of the canonical Reynolds rules and overlapping planes of encirclement using quaternions of arbitrary orientation. A new method for transitioning between flocking and enspherement is presented, the smoothness of which is demonstrated mathematically. The proposed approach is validated in a series of simulations on swarms of particles with double-integrator dynamics.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper presents a bimodal control strategy for transitioning between different topologies of flocking and dynamic structures. The work is motivated by applications where agents must pursue and close in on a target prior to carrying out more sophisticated tasks. Specifically, it builds on previous work in consensus-based flocking to define a new type of dynamic target pursuit and capturing method called dynamic enspherement. The problem is formulated in terms of the canonical Reynolds rules and overlapping planes of encirclement using quaternions of arbitrary orientation. A new method for transitioning between flocking and enspherement is presented, the smoothness of which is demonstrated mathematically. The proposed approach is validated in a series of simulations on swarms of particles with double-integrator dynamics.
Hajavi, Amirhossein; Etemad, Ali
Fine-grained Early Frequency Attention for Deep Speaker Recognition Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Audio and Speech Processing (eess.AS), electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Sound (cs.SD)
@misc{https://doi.org/10.48550/arxiv.2207.10006,
title = {Fine-grained Early Frequency Attention for Deep Speaker Recognition},
author = {Amirhossein Hajavi and Ali Etemad},
url = {https://arxiv.org/abs/2207.10006},
doi = {10.48550/ARXIV.2207.10006},
year = {2022},
date = {2022-07-20},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {Attention mechanisms have emerged as important tools that boost the performance of deep models by allowing them to focus on key parts of learned embeddings. However, current attention mechanisms used in speaker recognition tasks fail to consider fine-grained information items such as frequency bins in input spectral representations used by the deep networks. To address this issue, we propose the novel Fine-grained Early Frequency Attention (FEFA) for speaker recognition in-the-wild. Once integrated into a deep neural network, our proposed mechanism works by obtaining queries from early layers of the network and generating learnable weights to attend to information items as small as the frequency bins in the input spectral representations. To evaluate the performance of FEFA, we use several well-known deep models as backbone networks and integrate our attention module in their pipelines. The overall performance of these networks (with and without FEFA) are evaluated on the VoxCeleb1 dataset, where we observe considerable improvements when FEFA is used. },
keywords = {Audio and Speech Processing (eess.AS), electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Sound (cs.SD)},
pubstate = {published},
tppubtype = {misc}
}
Attention mechanisms have emerged as important tools that boost the performance of deep models by allowing them to focus on key parts of learned embeddings. However, current attention mechanisms used in speaker recognition tasks fail to consider fine-grained information items such as frequency bins in input spectral representations used by the deep networks. To address this issue, we propose the novel Fine-grained Early Frequency Attention (FEFA) for speaker recognition in-the-wild. Once integrated into a deep neural network, our proposed mechanism works by obtaining queries from early layers of the network and generating learnable weights to attend to information items as small as the frequency bins in the input spectral representations. To evaluate the performance of FEFA, we use several well-known deep models as backbone networks and integrate our attention module in their pipelines. The overall performance of these networks (with and without FEFA) are evaluated on the VoxCeleb1 dataset, where we observe considerable improvements when FEFA is used.
Farley, Andrew; Wang, Jie; Marshall, Joshua A.
How to pick a mobile robot simulator: A quantitative comparison of CoppeliaSim, Gazebo, MORSE and Webots with a focus on accuracy of motion Journal Article
In: Simulation Modelling Practice and Theory, vol. 120, pp. 102629, 2022, ISSN: 1569-190X.
Abstract | Links | BibTeX | Tags: Quantitative analysis, Robot Operating System (ROS), Robot simulation, Robot software
@article{FARLEY2022102629,
title = {How to pick a mobile robot simulator: A quantitative comparison of CoppeliaSim, Gazebo, MORSE and Webots with a focus on accuracy of motion},
author = {Andrew Farley and Jie Wang and Joshua A. Marshall},
url = {https://www.sciencedirect.com/science/article/pii/S1569190X22001046},
doi = {https://doi.org/10.1016/j.simpat.2022.102629},
issn = {1569-190X},
year = {2022},
date = {2022-07-18},
urldate = {2022-01-01},
journal = {Simulation Modelling Practice and Theory},
volume = {120},
pages = {102629},
abstract = {The number of available tools for dynamic simulation of robots has been growing rapidly in recent years. However, to the best of our knowledge, there are very few reported quantitative comparisons of the most widely-used robot simulation tools. This article attempts to partly fill this gap by providing quantitative and objective comparisons of four widely-used simulation packages for mobile robots. The comparisons reported here were conducted by obtaining data from a real Husky A200 mobile robot driving on mixed terrains as ground truth and by simulating a 3D mobile robot model in a developed identical simulation world of these terrains for each simulator. We then compared the simulation outputs with real, measured results by weighted metrics. Based on our experiments and selected metrics, we conclude that CoppeliaSim is currently the best performing simulator, although Gazebo is not far behind and is a good alternative.},
keywords = {Quantitative analysis, Robot Operating System (ROS), Robot simulation, Robot software},
pubstate = {published},
tppubtype = {article}
}
The number of available tools for dynamic simulation of robots has been growing rapidly in recent years. However, to the best of our knowledge, there are very few reported quantitative comparisons of the most widely-used robot simulation tools. This article attempts to partly fill this gap by providing quantitative and objective comparisons of four widely-used simulation packages for mobile robots. The comparisons reported here were conducted by obtaining data from a real Husky A200 mobile robot driving on mixed terrains as ground truth and by simulating a 3D mobile robot model in a developed identical simulation world of these terrains for each simulator. We then compared the simulation outputs with real, measured results by weighted metrics. Based on our experiments and selected metrics, we conclude that CoppeliaSim is currently the best performing simulator, although Gazebo is not far behind and is a good alternative.
Zand, Mohsen; Etemad, Ali; Greenspan, Michael
ObjectBox: From Centers to Boxes for Anchor-Free Object Detection Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2207.06985,
title = {ObjectBox: From Centers to Boxes for Anchor-Free Object Detection},
author = {Mohsen Zand and Ali Etemad and Michael Greenspan},
url = {https://arxiv.org/abs/2207.06985},
doi = {10.48550/ARXIV.2207.06985},
year = {2022},
date = {2022-07-14},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {We present ObjectBox, a novel single-stage anchor-free and highly generalizable object detection approach. As opposed to both existing anchor-based and anchor-free detectors, which are more biased toward specific object scales in their label assignments, we use only object center locations as positive samples and treat all objects equally in different feature levels regardless of the objects' sizes or shapes. Specifically, our label assignment strategy considers the object center locations as shape- and size-agnostic anchors in an anchor-free fashion, and allows learning to occur at all scales for every object. To support this, we define new regression targets as the distances from two corners of the center cell location to the four sides of the bounding box. Moreover, to handle scale-variant objects, we propose a tailored IoU loss to deal with boxes with different sizes. As a result, our proposed object detector does not need any dataset-dependent hyperparameters to be tuned across datasets. We evaluate our method on MS-COCO 2017 and PASCAL VOC 2012 datasets, and compare our results to state-of-the-art methods. We observe that ObjectBox performs favorably in comparison to prior works. Furthermore, we perform rigorous ablation experiments to evaluate different components of our method},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
We present ObjectBox, a novel single-stage anchor-free and highly generalizable object detection approach. As opposed to both existing anchor-based and anchor-free detectors, which are more biased toward specific object scales in their label assignments, we use only object center locations as positive samples and treat all objects equally in different feature levels regardless of the objects' sizes or shapes. Specifically, our label assignment strategy considers the object center locations as shape- and size-agnostic anchors in an anchor-free fashion, and allows learning to occur at all scales for every object. To support this, we define new regression targets as the distances from two corners of the center cell location to the four sides of the bounding box. Moreover, to handle scale-variant objects, we propose a tailored IoU loss to deal with boxes with different sizes. As a result, our proposed object detector does not need any dataset-dependent hyperparameters to be tuned across datasets. We evaluate our method on MS-COCO 2017 and PASCAL VOC 2012 datasets, and compare our results to state-of-the-art methods. We observe that ObjectBox performs favorably in comparison to prior works. Furthermore, we perform rigorous ablation experiments to evaluate different components of our method
Hajian, Gelareh; Morin, Evelyn; Etemad, Ali
Multimodal Estimation of Endpoint Force During Quasi-Dynamic and Dynamic Muscle Contractions Using Deep Learning Journal Article
In: IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1-11, 2022, ISSN: 1557-9662.
Abstract | Links | BibTeX | Tags:
@article{9825705,
title = {Multimodal Estimation of Endpoint Force During Quasi-Dynamic and Dynamic Muscle Contractions Using Deep Learning},
author = {Gelareh Hajian and Evelyn Morin and Ali Etemad},
url = {https://ieeexplore.ieee.org/document/9825705},
doi = {10.1109/TIM.2022.3189632},
issn = {1557-9662},
year = {2022},
date = {2022-07-11},
urldate = {2022-01-01},
journal = {IEEE Transactions on Instrumentation and Measurement},
volume = {71},
pages = {1-11},
abstract = {Accurate force/torque estimation is essential for applications such as powered exoskeletons, robotics, and rehabilitation. However, force/torque estimation under dynamic conditions is challenging due to changing joint angles, force levels, muscle lengths, and movement speeds. We propose a novel method to accurately model the generated force under isotonic, isokinetic (quasi-dynamic), and fully dynamic conditions. Our solution uses a deep multimodal convolutional neural network (CNN) to learn from multimodal electromyogram-inertial measurement unit (EMG-IMU) data and estimate the generated force for elbow flexion and extension, for both intra- and intersubject schemes. The proposed deep multimodal CNN extracts representations from EMG (in time and frequency domains) and IMU (in time domain) and aggregates them to obtain an effective embedding for force estimation. We describe a new dataset containing EMG, IMU, and output force data, collected under a number of different experimental conditions, and use this dataset to evaluate our proposed method. The results show the robustness of our approach in comparison to other baseline methods and those in the literature, in different experimental setups and validation schemes. The obtained $R^2$ values are 0.91 ± 0.034, 0.87 ± 0.041, and 0.81 ± 0.037 for the intrasubject and 0.81 ± 0.048, 0.64 ± 0.037, and 0.59 ± 0.042 for the intersubject scheme, during isotonic, isokinetic, and dynamic contractions, respectively. In addition, our results indicate that force estimation improves significantly when the kinematic information (IMU data) is included. Average improvements of 13.95%, 118.18%, and 50.0% (intrasubject) and 28.98%, 41.18%, and 137.93% (intersubject) for isotonic, isokinetic, and dynamic contractions, respectively, are achieved.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Accurate force/torque estimation is essential for applications such as powered exoskeletons, robotics, and rehabilitation. However, force/torque estimation under dynamic conditions is challenging due to changing joint angles, force levels, muscle lengths, and movement speeds. We propose a novel method to accurately model the generated force under isotonic, isokinetic (quasi-dynamic), and fully dynamic conditions. Our solution uses a deep multimodal convolutional neural network (CNN) to learn from multimodal electromyogram-inertial measurement unit (EMG-IMU) data and estimate the generated force for elbow flexion and extension, for both intra- and intersubject schemes. The proposed deep multimodal CNN extracts representations from EMG (in time and frequency domains) and IMU (in time domain) and aggregates them to obtain an effective embedding for force estimation. We describe a new dataset containing EMG, IMU, and output force data, collected under a number of different experimental conditions, and use this dataset to evaluate our proposed method. The results show the robustness of our approach in comparison to other baseline methods and those in the literature, in different experimental setups and validation schemes. The obtained $R^2$ values are 0.91 ± 0.034, 0.87 ± 0.041, and 0.81 ± 0.037 for the intrasubject and 0.81 ± 0.048, 0.64 ± 0.037, and 0.59 ± 0.042 for the intersubject scheme, during isotonic, isokinetic, and dynamic contractions, respectively. In addition, our results indicate that force estimation improves significantly when the kinematic information (IMU data) is included. Average improvements of 13.95%, 118.18%, and 50.0% (intrasubject) and 28.98%, 41.18%, and 137.93% (intersubject) for isotonic, isokinetic, and dynamic contractions, respectively, are achieved.
Li, Meng; Chen, Yifei; Zhu, Liehaung; Zhang, Zijian; Ni, Jianbing; Lal, Chhagan; Conti, Mauro
Astraea: Anonymous and Secure Auditing Based on Private Smart Contracts for Donation Systems Journal Article
In: IEEE Transactions on Dependable and Secure Computing, pp. 1-17, 2022, ISSN: 1941-0018.
Abstract | Links | BibTeX | Tags:
@article{9878045,
title = {Astraea: Anonymous and Secure Auditing Based on Private Smart Contracts for Donation Systems},
author = {Meng Li and Yifei Chen and Liehaung Zhu and Zijian Zhang and Jianbing Ni and Chhagan Lal and Mauro Conti},
url = {https://ieeexplore.ieee.org/abstract/document/9878045},
doi = {10.1109/TDSC.2022.3204287},
issn = {1941-0018},
year = {2022},
date = {2022-07-05},
urldate = {2022-01-01},
journal = {IEEE Transactions on Dependable and Secure Computing},
pages = {1-17},
abstract = {Many regions are in urgent need of facial masks for slowing down the spread of COVID-19. To fight the pandemic, people are contributing masks through donation systems. Most existing systems are built on a centralized architecture which is prone to the single point of failure and lack of transparency. Blockchain-based solutions neglect fundamental privacy concerns (donation privacy) and security attacks (collusion attack, stealing attack). Moreover, current auditing solutions are not designed to achieve donation privacy, thus not appropriate in our context. In this work, we design a decentralized, anonymous, and secure auditing framework Astraea based on private smart contracts for donation systems. Specifically, we integrate a Distribute Smart Contract (DiSC) with an SGX Enclave to distribute donations, prove the integrity of donation number (intention) and donation sum while preserving donation privacy. With DiSC, we design a Donation Smart Contract to refund deposits and defend against the stealing attack the collusion attack from malicious collector and transponder. We formally define and prove the privacy and security of Astraea by using security reduction. We build a prototype of Astraea to conduct extensive performance analysis. Experimental results demonstrate that Astraea is practically efficient in terms of both computation and communication.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Many regions are in urgent need of facial masks for slowing down the spread of COVID-19. To fight the pandemic, people are contributing masks through donation systems. Most existing systems are built on a centralized architecture which is prone to the single point of failure and lack of transparency. Blockchain-based solutions neglect fundamental privacy concerns (donation privacy) and security attacks (collusion attack, stealing attack). Moreover, current auditing solutions are not designed to achieve donation privacy, thus not appropriate in our context. In this work, we design a decentralized, anonymous, and secure auditing framework Astraea based on private smart contracts for donation systems. Specifically, we integrate a Distribute Smart Contract (DiSC) with an SGX Enclave to distribute donations, prove the integrity of donation number (intention) and donation sum while preserving donation privacy. With DiSC, we design a Donation Smart Contract to refund deposits and defend against the stealing attack the collusion attack from malicious collector and transponder. We formally define and prove the privacy and security of Astraea by using security reduction. We build a prototype of Astraea to conduct extensive performance analysis. Experimental results demonstrate that Astraea is practically efficient in terms of both computation and communication.
GILLHAM, Jack; BENTZ, Evan C.; HOULT, Neil A.
Measuring support reactions and damage detection in steel beams using distributed strain sensing Journal Article
In: Engineering Structures, vol. 262, pp. 114389, 2022, ISSN: 0141-0296.
Abstract | Links | BibTeX | Tags: Beams, Damage detection, Distributed fiber optic sensing, Support reactions
@article{GILLHAM2022114389,
title = {Measuring support reactions and damage detection in steel beams using distributed strain sensing},
author = {Jack GILLHAM and Evan C. BENTZ and Neil A. HOULT},
url = {https://www.sciencedirect.com/science/article/pii/S0141029622005041},
doi = {https://doi.org/10.1016/j.engstruct.2022.114389},
issn = {0141-0296},
year = {2022},
date = {2022-07-01},
urldate = {2022-01-01},
journal = {Engineering Structures},
volume = {262},
pages = {114389},
abstract = {While visual inspections are an important tool in the assessment of steel bridges, they cannot provide quantitative information about the bridge’s behavior. Distributed fiber optic sensors (DFOS) have the potential to provide added insight into the performance of bridges through the measurement of full strain profiles along a member as well as other key performance indicators that can be derived from these measurements. This study investigates the use of DFOS measurements to derive support reactions and detect local damage along the flexural members in a bridge. A series of load tests were undertaken to develop the method for support reaction estimation, starting with a simply supported beam, then continuous beams, and then finally a full model bridge. The beam test support reactions were evaluated against load cell measurements and variations in geometry were shown to have an impact on the measurements that could be accounted for through calibration. The model bridge reactions matched the total applied load and showed the expected distribution for the loading configuration applied. A second model bridge test was undertaken where the shear stress in the connection between the beams and the slab exceeded the capacity of the connection and localized damage occurred. Using the DFOS measurements, the neutral axis was calculated and used to detect this localized damage.},
keywords = {Beams, Damage detection, Distributed fiber optic sensing, Support reactions},
pubstate = {published},
tppubtype = {article}
}
While visual inspections are an important tool in the assessment of steel bridges, they cannot provide quantitative information about the bridge’s behavior. Distributed fiber optic sensors (DFOS) have the potential to provide added insight into the performance of bridges through the measurement of full strain profiles along a member as well as other key performance indicators that can be derived from these measurements. This study investigates the use of DFOS measurements to derive support reactions and detect local damage along the flexural members in a bridge. A series of load tests were undertaken to develop the method for support reaction estimation, starting with a simply supported beam, then continuous beams, and then finally a full model bridge. The beam test support reactions were evaluated against load cell measurements and variations in geometry were shown to have an impact on the measurements that could be accounted for through calibration. The model bridge reactions matched the total applied load and showed the expected distribution for the loading configuration applied. A second model bridge test was undertaken where the shear stress in the connection between the beams and the slab exceeded the capacity of the connection and localized damage occurred. Using the DFOS measurements, the neutral axis was calculated and used to detect this localized damage.
Pecly, Leonam; Hashtrudi-Zaad, Keyvan
Uncoupled Stability of Kinesthetic Haptic Systems Simulating Mass-Damper-Spring Environments with Complementary Filter Inproceedings
In: 2022 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), pp. 97-102, 2022, ISSN: 2159-6255.
Abstract | Links | BibTeX | Tags:
@inproceedings{9863311,
title = {Uncoupled Stability of Kinesthetic Haptic Systems Simulating Mass-Damper-Spring Environments with Complementary Filter},
author = {Leonam Pecly and Keyvan Hashtrudi-Zaad},
doi = {10.1109/AIM52237.2022.9863311},
issn = {2159-6255},
year = {2022},
date = {2022-07-01},
urldate = {2022-07-01},
booktitle = {2022 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM)},
pages = {97-102},
abstract = {Uncoupled stability, the condition by which the user is not in contact with the haptic device, is arguably a stringent stability condition for haptic simulation systems. Uncoupled stability of haptic systems simulating linear mass-spring or viscoelastic virtual environments have been analyzed. In this paper, we analytically and experimentally evaluate uncoupled stability for simulating mass-damper-spring virtual environments when only position or when position and velocity are available. In addition, the effect of using a linear combination of position and velocity in deriving acceleration estimate is also studied. Experimental results in a one degree-of-freedom device showed that the highest stiffness values are obtained when the acceleration is equally derived from position and velocity. This work will shed light on the interaction of the three dynamic components for virtual environment rendering.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Uncoupled stability, the condition by which the user is not in contact with the haptic device, is arguably a stringent stability condition for haptic simulation systems. Uncoupled stability of haptic systems simulating linear mass-spring or viscoelastic virtual environments have been analyzed. In this paper, we analytically and experimentally evaluate uncoupled stability for simulating mass-damper-spring virtual environments when only position or when position and velocity are available. In addition, the effect of using a linear combination of position and velocity in deriving acceleration estimate is also studied. Experimental results in a one degree-of-freedom device showed that the highest stiffness values are obtained when the acceleration is equally derived from position and velocity. This work will shed light on the interaction of the three dynamic components for virtual environment rendering.
Khazaeni, MA Yasaman (Needham
ÄRTIFICIAL INTELLIGENCE FOR LEARNING PATH RECOMMENDATIONS Bachelor Thesis
2022.
@bachelorthesis{patent:20220208018,
title = {ÄRTIFICIAL INTELLIGENCE FOR LEARNING PATH RECOMMENDATIONS},
author = {MA Yasaman (Needham Khazaeni},
url = {https://www.freepatentsonline.com/y2022/0208018.html},
year = {2022},
date = {2022-06-30},
urldate = {2022-06-01},
number = {20220208018},
keywords = {},
pubstate = {published},
tppubtype = {bachelorthesis}
}
Richard Ean, Karen Rudie
Epistemic interpretations of decentralized discrete-event system problems Journal Article
In: Discrete Event Dynamic Systems, 2022.
Abstract | Links | BibTeX | Tags:
@article{Ean2022b,
title = {Epistemic interpretations of decentralized discrete-event system problems},
author = {Richard Ean, Karen Rudie},
url = {https://doi.org/10.1007/s10626-022-00363-7},
doi = {10.1007/s10626-022-00363-7},
year = {2022},
date = {2022-06-22},
urldate = {2022-06-22},
journal = {Discrete Event Dynamic Systems},
abstract = {This paper presents epistemic characterizations to co-observability conditions in decentralized supervisory control of discrete-event systems. The logical characterizations provide more intuitive interpretations of the various co-observability conditions, and make immediately apparent the relations between the conditions. Closures under set union of some of the conditions are also discussed.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This paper presents epistemic characterizations to co-observability conditions in decentralized supervisory control of discrete-event systems. The logical characterizations provide more intuitive interpretations of the various co-observability conditions, and make immediately apparent the relations between the conditions. Closures under set union of some of the conditions are also discussed.
Mahmud, Zunayed; Hungler, Paul; Etemad, Ali
Multistream Gaze Estimation with Anatomical Eye Region Isolation by Synthetic to Real Transfer Learning Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, Machine Learning (cs.LG)
@misc{https://doi.org/10.48550/arxiv.2206.09256,
title = {Multistream Gaze Estimation with Anatomical Eye Region Isolation by Synthetic to Real Transfer Learning},
author = {Zunayed Mahmud and Paul Hungler and Ali Etemad},
url = {https://arxiv.org/abs/2206.09256},
doi = {10.48550/ARXIV.2206.09256},
year = {2022},
date = {2022-06-18},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {We propose a novel neural pipeline, MSGazeNet, that learns gaze representations by taking advantage of the eye anatomy information through a multistream framework. Our proposed solution comprises two components, first a network for isolating anatomical eye regions, and a second network for multistream gaze estimation. The eye region isolation is performed with a U-Net style network which we train using a synthetic dataset that contains eye region masks for the visible eyeball and the iris region. The synthetic dataset used in this stage is a new dataset consisting of 60,000 eye images, which we create using an eye-gaze simulator, UnityEyes. Successive to training, the eye region isolation network is then transferred to the real domain for generating masks for the real-world eye images. In order to successfully make the transfer, we exploit domain randomization in the training process, which allows for the synthetic images to benefit from a larger variance with the help of augmentations that resemble artifacts. The generated eye region masks along with the raw eye images are then used together as a multistream input to our gaze estimation network. We evaluate our framework on three benchmark gaze estimation datasets, MPIIGaze, Eyediap, and UTMultiview, where we set a new state-of-the-art on Eyediap and UTMultiview datasets by obtaining a performance gain of 7.57% and 1.85% respectively, while achieving competitive performance on MPIIGaze. We also study the robustness of our method with respect to the noise in the data and demonstrate that our model is less sensitive to noisy data. Lastly, we perform a variety of experiments including ablation studies to evaluate the contribution of different components and design choices in our solution. },
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, Machine Learning (cs.LG)},
pubstate = {published},
tppubtype = {misc}
}
We propose a novel neural pipeline, MSGazeNet, that learns gaze representations by taking advantage of the eye anatomy information through a multistream framework. Our proposed solution comprises two components, first a network for isolating anatomical eye regions, and a second network for multistream gaze estimation. The eye region isolation is performed with a U-Net style network which we train using a synthetic dataset that contains eye region masks for the visible eyeball and the iris region. The synthetic dataset used in this stage is a new dataset consisting of 60,000 eye images, which we create using an eye-gaze simulator, UnityEyes. Successive to training, the eye region isolation network is then transferred to the real domain for generating masks for the real-world eye images. In order to successfully make the transfer, we exploit domain randomization in the training process, which allows for the synthetic images to benefit from a larger variance with the help of augmentations that resemble artifacts. The generated eye region masks along with the raw eye images are then used together as a multistream input to our gaze estimation network. We evaluate our framework on three benchmark gaze estimation datasets, MPIIGaze, Eyediap, and UTMultiview, where we set a new state-of-the-art on Eyediap and UTMultiview datasets by obtaining a performance gain of 7.57% and 1.85% respectively, while achieving competitive performance on MPIIGaze. We also study the robustness of our method with respect to the noise in the data and demonstrate that our model is less sensitive to noisy data. Lastly, we perform a variety of experiments including ablation studies to evaluate the contribution of different components and design choices in our solution.
Mahmud, Zunayed; Hungler, Paul; Etemad, Ali
Multistream Gaze Estimation with Anatomical Eye Region Isolation by Synthetic to Real Transfer Learning Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, Machine Learning (cs.LG)
@misc{https://doi.org/10.48550/arxiv.2206.09256b,
title = {Multistream Gaze Estimation with Anatomical Eye Region Isolation by Synthetic to Real Transfer Learning},
author = {Zunayed Mahmud and Paul Hungler and Ali Etemad},
url = {https://arxiv.org/abs/2206.09256},
doi = {10.48550/ARXIV.2206.09256},
year = {2022},
date = {2022-06-18},
urldate = {2022-01-01},
publisher = {arXiv},
abstract = {We propose a novel neural pipeline, MSGazeNet, that learns gaze representations by taking advantage of the eye anatomy information through a multistream framework. Our proposed solution comprises two components, first a network for isolating anatomical eye regions, and a second network for multistream gaze estimation. The eye region isolation is performed with a U-Net style network which we train using a synthetic dataset that contains eye region masks for the visible eyeball and the iris region. The synthetic dataset used in this stage is a new dataset consisting of 60,000 eye images, which we create using an eye-gaze simulator, UnityEyes. Successive to training, the eye region isolation network is then transferred to the real domain for generating masks for the real-world eye images. In order to successfully make the transfer, we exploit domain randomization in the training process, which allows for the synthetic images to benefit from a larger variance with the help of augmentations that resemble artifacts. The generated eye region masks along with the raw eye images are then used together as a multistream input to our gaze estimation network. We evaluate our framework on three benchmark gaze estimation datasets, MPIIGaze, Eyediap, and UTMultiview, where we set a new state-of-the-art on Eyediap and UTMultiview datasets by obtaining a performance gain of 7.57% and 1.85% respectively, while achieving competitive performance on MPIIGaze. We also study the robustness of our method with respect to the noise in the data and demonstrate that our model is less sensitive to noisy data. Lastly, we perform a variety of experiments including ablation studies to evaluate the contribution of different components and design choices in our solution. },
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, Machine Learning (cs.LG)},
pubstate = {published},
tppubtype = {misc}
}
We propose a novel neural pipeline, MSGazeNet, that learns gaze representations by taking advantage of the eye anatomy information through a multistream framework. Our proposed solution comprises two components, first a network for isolating anatomical eye regions, and a second network for multistream gaze estimation. The eye region isolation is performed with a U-Net style network which we train using a synthetic dataset that contains eye region masks for the visible eyeball and the iris region. The synthetic dataset used in this stage is a new dataset consisting of 60,000 eye images, which we create using an eye-gaze simulator, UnityEyes. Successive to training, the eye region isolation network is then transferred to the real domain for generating masks for the real-world eye images. In order to successfully make the transfer, we exploit domain randomization in the training process, which allows for the synthetic images to benefit from a larger variance with the help of augmentations that resemble artifacts. The generated eye region masks along with the raw eye images are then used together as a multistream input to our gaze estimation network. We evaluate our framework on three benchmark gaze estimation datasets, MPIIGaze, Eyediap, and UTMultiview, where we set a new state-of-the-art on Eyediap and UTMultiview datasets by obtaining a performance gain of 7.57% and 1.85% respectively, while achieving competitive performance on MPIIGaze. We also study the robustness of our method with respect to the noise in the data and demonstrate that our model is less sensitive to noisy data. Lastly, we perform a variety of experiments including ablation studies to evaluate the contribution of different components and design choices in our solution.
Khaleghi, Leyla; Marshall, Joshua; Etemad, Ali
Learning Sequential Contexts using Transformer for 3D Hand Pose Estimation Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences
@misc{https://doi.org/10.48550/arxiv.2206.00171,
title = {Learning Sequential Contexts using Transformer for 3D Hand Pose Estimation},
author = {Leyla Khaleghi and Joshua Marshall and Ali Etemad},
url = {https://arxiv.org/abs/2206.00171},
doi = {10.48550/ARXIV.2206.00171},
year = {2022},
date = {2022-06-01},
urldate = {2022-06-01},
publisher = {arXiv},
abstract = {3D hand pose estimation (HPE) is the process of locating the joints of the hand in 3D from any visual input. HPE has recently received an increased amount of attention due to its key role in a variety of human-computer interaction applications. Recent HPE methods have demonstrated the advantages of employing videos or multi-view images, allowing for more robust HPE systems. Accordingly, in this study, we propose a new method to perform Sequential learning with Transformer for Hand Pose (SeTHPose) estimation. Our SeTHPose pipeline begins by extracting visual embeddings from individual hand images. We then use a transformer encoder to learn the sequential context along time or viewing angles and generate accurate 2D hand joint locations. Then, a graph convolutional neural network with a U-Net configuration is used to convert the 2D hand joint locations to 3D poses. Our experiments show that SeTHPose performs well on both hand sequence varieties, temporal and angular. Also, SeTHPose outperforms other methods in the field to achieve new state-of-the-art results on two public available sequential datasets, STB and MuViHand. },
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences},
pubstate = {published},
tppubtype = {misc}
}
3D hand pose estimation (HPE) is the process of locating the joints of the hand in 3D from any visual input. HPE has recently received an increased amount of attention due to its key role in a variety of human-computer interaction applications. Recent HPE methods have demonstrated the advantages of employing videos or multi-view images, allowing for more robust HPE systems. Accordingly, in this study, we propose a new method to perform Sequential learning with Transformer for Hand Pose (SeTHPose) estimation. Our SeTHPose pipeline begins by extracting visual embeddings from individual hand images. We then use a transformer encoder to learn the sequential context along time or viewing angles and generate accurate 2D hand joint locations. Then, a graph convolutional neural network with a U-Net configuration is used to convert the 2D hand joint locations to 3D poses. Our experiments show that SeTHPose performs well on both hand sequence varieties, temporal and angular. Also, SeTHPose outperforms other methods in the field to achieve new state-of-the-art results on two public available sequential datasets, STB and MuViHand.
Callanan, Ethan; Venezia, Rebecca De; Armstrong, Victoria; Paredes, Alison; Chakraborti, Tathagata; Muise, Christian
MACQ: A Holistic View of Model Acquisition Techniques Journal Article
In: 2022.
BibTeX | Tags:
@article{callananmacq,
title = {MACQ: A Holistic View of Model Acquisition Techniques},
author = {Ethan Callanan and Rebecca De Venezia and Victoria Armstrong and Alison Paredes and Tathagata Chakraborti and Christian Muise},
year = {2022},
date = {2022-06-01},
urldate = {2022-06-01},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Delamer, Jean-Alexis; Givigi, Sidney
Controller Based Interactive Proof to Ensure Trust in Multi-Agent Systems Journal Article
In: IEEE Transactions on Automatic Control, pp. 1-1, 2022, ISSN: 1558-2523.
Abstract | Links | BibTeX | Tags:
@article{9785901,
title = {Controller Based Interactive Proof to Ensure Trust in Multi-Agent Systems},
author = {Jean-Alexis Delamer and Sidney Givigi},
url = {https://ieeexplore.ieee.org/abstract/document/9785901},
doi = {10.1109/TAC.2022.3179217},
issn = {1558-2523},
year = {2022},
date = {2022-05-31},
urldate = {2022-01-01},
journal = {IEEE Transactions on Automatic Control},
pages = {1-1},
abstract = {When multiple agents need to perform tasks that need cooperation, communication is necessary for coordination and also to assess completion. In these cases, trust needs to be guaranteed so agents know if others are following a common strategy or if they are not cooperating. If communication is explicit, other agents, including agents not involved directly in the mission, can forge information and lead agents to believe that they are behaving according to the mission plan. Several methods have been proposed in the literature and in this paper, we propose a novel method based on interactive proofs in which trust between two agents is asserted by one agent demonstrating to the other that both follow the same mission plan. We formally represent the problem as a series of Markov Decision Processes and demonstrate that the formulation is complete and sound. Simulations of multiple agents in a maze problem demonstrate the behaviour of the system.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
When multiple agents need to perform tasks that need cooperation, communication is necessary for coordination and also to assess completion. In these cases, trust needs to be guaranteed so agents know if others are following a common strategy or if they are not cooperating. If communication is explicit, other agents, including agents not involved directly in the mission, can forge information and lead agents to believe that they are behaving according to the mission plan. Several methods have been proposed in the literature and in this paper, we propose a novel method based on interactive proofs in which trust between two agents is asserted by one agent demonstrating to the other that both follow the same mission plan. We formally represent the problem as a series of Markov Decision Processes and demonstrate that the formulation is complete and sound. Simulations of multiple agents in a maze problem demonstrate the behaviour of the system.
Soltanieh, Sahar; Etemad, Ali; Hashemi, Javad
Analysis of Augmentations for Contrastive ECG Representation Learning Miscellaneous
2022.
Abstract | Links | BibTeX | Tags: Artificial Intelligence (cs.AI), electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Machine Learning (cs.LG), Signal Processing (eess.SP)
@misc{https://doi.org/10.48550/arxiv.2206.07656,
title = {Analysis of Augmentations for Contrastive ECG Representation Learning},
author = {Sahar Soltanieh and Ali Etemad and Javad Hashemi},
url = {https://arxiv.org/abs/2206.07656},
doi = {10.48550/ARXIV.2206.07656},
year = {2022},
date = {2022-05-22},
urldate = {2022-05-22},
publisher = {arXiv},
abstract = {This paper systematically investigates the effectiveness of various augmentations for contrastive self-supervised learning of electrocardiogram (ECG) signals and identifies the best parameters. The baseline of our proposed self-supervised framework consists of two main parts: the contrastive learning and the downstream task. In the first stage, we train an encoder using a number of augmentations to extract generalizable ECG signal representations. We then freeze the encoder and finetune a few linear layers with different amounts of labelled data for downstream arrhythmia detection. We then experiment with various augmentations techniques and explore a range of parameters. Our experiments are done on PTB-XL, a large and publicly available 12-lead ECG dataset. The results show that applying augmentations in a specific range of complexities works better for self-supervised contrastive learning. For instance, when adding Gaussian noise, a sigma in the range of 0.1 to 0.2 achieves better results, while poor training occurs when the added noise is too small or too large (outside of the specified range). A similar trend is observed with other augmentations, demonstrating the importance of selecting the optimum level of difficulty for the added augmentations, as augmentations that are too simple will not result in effective training, while augmentations that are too difficult will also prevent the model from effective learning of generalized representations. Our work can influence future research on self-supervised contrastive learning on bio-signals and aid in selecting optimum parameters for different augmentations. },
keywords = {Artificial Intelligence (cs.AI), electronic engineering, FOS: Computer and information sciences, FOS: Electrical engineering, information engineering, Machine Learning (cs.LG), Signal Processing (eess.SP)},
pubstate = {published},
tppubtype = {misc}
}
This paper systematically investigates the effectiveness of various augmentations for contrastive self-supervised learning of electrocardiogram (ECG) signals and identifies the best parameters. The baseline of our proposed self-supervised framework consists of two main parts: the contrastive learning and the downstream task. In the first stage, we train an encoder using a number of augmentations to extract generalizable ECG signal representations. We then freeze the encoder and finetune a few linear layers with different amounts of labelled data for downstream arrhythmia detection. We then experiment with various augmentations techniques and explore a range of parameters. Our experiments are done on PTB-XL, a large and publicly available 12-lead ECG dataset. The results show that applying augmentations in a specific range of complexities works better for self-supervised contrastive learning. For instance, when adding Gaussian noise, a sigma in the range of 0.1 to 0.2 achieves better results, while poor training occurs when the added noise is too small or too large (outside of the specified range). A similar trend is observed with other augmentations, demonstrating the importance of selecting the optimum level of difficulty for the added augmentations, as augmentations that are too simple will not result in effective training, while augmentations that are too difficult will also prevent the model from effective learning of generalized representations. Our work can influence future research on self-supervised contrastive learning on bio-signals and aid in selecting optimum parameters for different augmentations.
Heaton, Jacqueline; Givigi, Sidney
A Deep CNN System for Classification of Emotions Using EEG Signals Inproceedings
In: 2022 IEEE International Systems Conference (SysCon), pp. 1-7, 2022, ISSN: 2472-9647.
Abstract | Links | BibTeX | Tags:
@inproceedings{9773832,
title = {A Deep CNN System for Classification of Emotions Using EEG Signals},
author = {Jacqueline Heaton and Sidney Givigi},
url = {https://ieeexplore.ieee.org/abstract/document/9773832},
doi = {10.1109/SysCon53536.2022.9773832},
issn = {2472-9647},
year = {2022},
date = {2022-05-16},
urldate = {2022-04-01},
booktitle = {2022 IEEE International Systems Conference (SysCon)},
pages = {1-7},
abstract = {Emotion classification has many applications in human-computer interaction, and is a necessary mode of communication for many different tasks where humans and robots must work together or in close quarters. When working with people who have trouble using verbal communication, or when it is unrealistic to expect verbal communication, robots must still be capable of taking the person’s emotions into account, whether through facial cues, body language, or other signals. Electroencephalograms are capable of capturing the signals of the brain, which can be processed and classified using various artificial intelligence architectures. In this paper, a deep convolutional neural network is applied to an emotion classification task, where it successfully learns to identify six second windows as one of four emotions: boredom, relaxation, horror, and humour. The neural network is applied to 14 individuals and a high accuracy of nearly 100% is achieved when the test data is chosen randomly from the dataset. A study is performed to find what conditions in the data are necessary for high classification accuracy. The emotion data was collected from subjects as they played four games of different genres, designed to evoke one emotion out of boredom, relaxation, humour, or fear, as assessed by the professional game critic services.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Emotion classification has many applications in human-computer interaction, and is a necessary mode of communication for many different tasks where humans and robots must work together or in close quarters. When working with people who have trouble using verbal communication, or when it is unrealistic to expect verbal communication, robots must still be capable of taking the person’s emotions into account, whether through facial cues, body language, or other signals. Electroencephalograms are capable of capturing the signals of the brain, which can be processed and classified using various artificial intelligence architectures. In this paper, a deep convolutional neural network is applied to an emotion classification task, where it successfully learns to identify six second windows as one of four emotions: boredom, relaxation, horror, and humour. The neural network is applied to 14 individuals and a high accuracy of nearly 100% is achieved when the test data is chosen randomly from the dataset. A study is performed to find what conditions in the data are necessary for high classification accuracy. The emotion data was collected from subjects as they played four games of different genres, designed to evoke one emotion out of boredom, relaxation, humour, or fear, as assessed by the professional game critic services.