Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283193
Ni Zeng, Yiqiang Chen, Yang Gu, Dongdong Liu, Yunbing Xing
Sign Language Synthesis (SLS) is a domain-specific problem where multiple sign language words are stitched to generate a whole sentence in video, which serves to facilitate communications between the hearing-impaired people and healthy population. This paper presents a Variable Motion Frame Interpolation (VMFI) method for highly fluent SLS in scattered videos. Existing approaches for SLS mainly focus on mechanical virtual human technology, lacking high flexibility and natural effect. Also, the representative solutions to interpolate frames usually assume that the motion object moves at a constant speed which is not suitable for predicting the complex hand motion in frames of scattered sign language videos. To address the above issues, the proposed VMFI adopts acceleration to predict more accurate interpolated frames based on an end-to-end convolutional neural network. The framework of VMFI consists of variable optical flow estimation network and high-quality frame synthesis network that can approximate and fuse the intermediate optical flow to generate interpolated frames for synthesis. Experimental results on our realistic collected Chinese sign language dataset demonstrate that the proposed VMFI model achieves efficiency by performing better in PSNR (Peak Signal to Noise Ratio), SSIM (Structural Similarity) and MA (Motion Activity) and gets higher score in MOS (Mean Opinion Score) than other two representative methods.
手语合成(Sign Language Synthesis, SLS)是将多个手语单词拼接成一个完整的视频句子，以方便听障人群与健康人群之间的交流的领域问题。针对离散视频中高度流畅的SLS，提出了一种可变运动帧插值方法。现有的SLS方法主要集中在机械虚拟人技术上，缺乏高度的灵活性和自然效果。此外，典型的插值帧解通常假设运动对象以恒定速度运动，这不适用于预测分散的手语视频帧中的复杂手部运动。为了解决上述问题，本文提出的VMFI采用基于端到端卷积神经网络的加速来预测更准确的插值帧。VMFI框架由可变光流估计网络和高质量帧合成网络组成，高质量帧合成网络可以近似和融合中间光流，生成插值帧进行合成。在实际收集的中国手语数据集上的实验结果表明，所提出的VMFI模型在峰值信噪比(PSNR)、结构相似度(SSIM)和运动活跃度(MA)方面具有更好的性能，在平均意见得分(MOS)方面取得了比其他两种代表性方法更高的分数。
{"title":"Highly Fluent Sign Language Synthesis Based on Variable Motion Frame Interpolation","authors":"Ni Zeng, Yiqiang Chen, Yang Gu, Dongdong Liu, Yunbing Xing","doi":"10.1109/SMC42975.2020.9283193","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283193","url":null,"abstract":"Sign Language Synthesis (SLS) is a domain-specific problem where multiple sign language words are stitched to generate a whole sentence in video, which serves to facilitate communications between the hearing-impaired people and healthy population. This paper presents a Variable Motion Frame Interpolation (VMFI) method for highly fluent SLS in scattered videos. Existing approaches for SLS mainly focus on mechanical virtual human technology, lacking high flexibility and natural effect. Also, the representative solutions to interpolate frames usually assume that the motion object moves at a constant speed which is not suitable for predicting the complex hand motion in frames of scattered sign language videos. To address the above issues, the proposed VMFI adopts acceleration to predict more accurate interpolated frames based on an end-to-end convolutional neural network. The framework of VMFI consists of variable optical flow estimation network and high-quality frame synthesis network that can approximate and fuse the intermediate optical flow to generate interpolated frames for synthesis. Experimental results on our realistic collected Chinese sign language dataset demonstrate that the proposed VMFI model achieves efficiency by performing better in PSNR (Peak Signal to Noise Ratio), SSIM (Structural Similarity) and MA (Motion Activity) and gets higher score in MOS (Mean Opinion Score) than other two representative methods.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"7 1","pages":"1772-1777"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84382184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283150
Lingsheng Wang, L. Xu, P. Li, Siming Zha, Lei Chen
Alzheimer's disease (AD) is one of the most common neurodegenerative diseases. Understanding AD progression can empower the patients in taking proactive care. Mini Mental State Examination (MMSE) and AD Assessment Scale Cognitive subscale (ADAS-Cog) are two prevailing clinical measures designed to evaluate the AD progression. In this paper, we propose a weakly supervised Temporal Multitask Matrix Completion (TMMC) framework, which combines a novel transductive multitask feature selection scheme, to simultaneously predict AD progression measured by MMSE and ADAS-Cog, and identify related biomarkers trackable of AD progression. Specifically, by treating the prediction of cognitive scores at each time point as a regression task, we first formulate AD progression problem as a standard Multitask Matrix Completion (MMC) model. Secondly, considering the limited number of samples available in this study, we introduce a transductive feature selection scheme to jointly select the task-shared features for multiple time points and the task-specific features for different time points, and thus alleviate the over-fitting defect caused by Small-Sample-Size issue. Thirdly, aiming at the small change of cognitive scores between successive time points for a patient, we employ a temporal regularization scheme to capture the temporal smoothness of cognitive scores. Furthermore, we design an efficient optimization algorithm based on Alternative Minimization and Difference of Convex Programming techniques to solve the proposed TMMC framework. Finally, the extensive experiments performed on real-world Alzheimer’s disease dataset demonstrate the effectiveness of our TMMC framework.
{"title":"Modeling Disease Progression via Weakly Supervised Temporal Multitask Matrix Completion","authors":"Lingsheng Wang, L. Xu, P. Li, Siming Zha, Lei Chen","doi":"10.1109/SMC42975.2020.9283150","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283150","url":null,"abstract":"Alzheimer's disease (AD) is one of the most common neurodegenerative diseases. Understanding AD progression can empower the patients in taking proactive care. Mini Mental State Examination (MMSE) and AD Assessment Scale Cognitive subscale (ADAS-Cog) are two prevailing clinical measures designed to evaluate the AD progression. In this paper, we propose a weakly supervised Temporal Multitask Matrix Completion (TMMC) framework, which combines a novel transductive multitask feature selection scheme, to simultaneously predict AD progression measured by MMSE and ADAS-Cog, and identify related biomarkers trackable of AD progression. Specifically, by treating the prediction of cognitive scores at each time point as a regression task, we first formulate AD progression problem as a standard Multitask Matrix Completion (MMC) model. Secondly, considering the limited number of samples available in this study, we introduce a transductive feature selection scheme to jointly select the task-shared features for multiple time points and the task-specific features for different time points, and thus alleviate the over-fitting defect caused by Small-Sample-Size issue. Thirdly, aiming at the small change of cognitive scores between successive time points for a patient, we employ a temporal regularization scheme to capture the temporal smoothness of cognitive scores. Furthermore, we design an efficient optimization algorithm based on Alternative Minimization and Difference of Convex Programming techniques to solve the proposed TMMC framework. Finally, the extensive experiments performed on real-world Alzheimer’s disease dataset demonstrate the effectiveness of our TMMC framework.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"19 1","pages":"1141-1148"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84396549","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283475
Francisco B. de S. Rocha, Bruno Vicente Alves de Lima, R. Leal, Diego P. Rocha, Karoline de M. Farias, R. Rabêlo, A. M. Santana
This paper presents a structural recognition system using machine learning algorithms (Multilayer Perceptron, Support Vector Machine and Random Forest) and the environment information to analyzes the feasibility of the use of machine learning methods for the construction of topological maps. The proposed method combines the recognized information from a given scene with a topological graph to create a map. This map can be used to plan high-level tasks of robotic navigation. The topological nodes are used to store semantic information, such as the robot’s poses, sensor data and scene characteristics. The machine learning algorithms classification of the structural information as either rooms, corridors or doors obtained a satisfactory performance. The structural recognition provided by classification presents accuracy greater than 97% and topological maps built efficiently of classification.
{"title":"Machine Learning Applied to Topological Mapping for Structure Recognition","authors":"Francisco B. de S. Rocha, Bruno Vicente Alves de Lima, R. Leal, Diego P. Rocha, Karoline de M. Farias, R. Rabêlo, A. M. Santana","doi":"10.1109/SMC42975.2020.9283475","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283475","url":null,"abstract":"This paper presents a structural recognition system using machine learning algorithms (Multilayer Perceptron, Support Vector Machine and Random Forest) and the environment information to analyzes the feasibility of the use of machine learning methods for the construction of topological maps. The proposed method combines the recognized information from a given scene with a topological graph to create a map. This map can be used to plan high-level tasks of robotic navigation. The topological nodes are used to store semantic information, such as the robot’s poses, sensor data and scene characteristics. The machine learning algorithms classification of the structural information as either rooms, corridors or doors obtained a satisfactory performance. The structural recognition provided by classification presents accuracy greater than 97% and topological maps built efficiently of classification.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"31 1","pages":"1872-1877"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84507290","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283226
Shunsuke Nansai, Y. Ando, N. Kamamichi, H. Itoh
The purpose of this paper is to propose a new type of a kinetic chained walking robot capable of walking with only a single actuator, and is to design its trajectory tracking control system. Legged robots are able to move across irregular terrains, however, have an issue on energy efficiency compared with other morphology. A bio-inspired approach often provides effective solutions, for example, a lizard is able to mainly walk by utilizing only twisting its waist. To mimic this characteristic by robotics, a robot consisting of four-bar linkage mechanism is proposed. This idea improves simplification of its locomotion analysis. In this paper, two important kinematics characteristics are analyzed in order to propose locomotion ability and effectiveness of the robot. In particular, a turning angle and a stride distance are analysed. After that, a trajectory tracking control system is designed based on the PID control low. Ideas for the control system design in this paper are both to decide an bias of an input angle function as a input of the system and to set a control period on half period of the input angle function. Finally, effectiveness of the designed control system is verified via numerical simulations. A straight line and a circle trajectory are adopted for the verification. As the results, it is shown that the designed trajectory tracking control system is capable of tracking two different trajectory. In addition, it is also shown that the designed trajectory tracking control system satisfies the kinematics analysis results from the side of view of the kinematic of the robot.
{"title":"Kinematics Analysis and Tracking Control of Novel Single Actuated Lizard Type Robot","authors":"Shunsuke Nansai, Y. Ando, N. Kamamichi, H. Itoh","doi":"10.1109/SMC42975.2020.9283226","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283226","url":null,"abstract":"The purpose of this paper is to propose a new type of a kinetic chained walking robot capable of walking with only a single actuator, and is to design its trajectory tracking control system. Legged robots are able to move across irregular terrains, however, have an issue on energy efficiency compared with other morphology. A bio-inspired approach often provides effective solutions, for example, a lizard is able to mainly walk by utilizing only twisting its waist. To mimic this characteristic by robotics, a robot consisting of four-bar linkage mechanism is proposed. This idea improves simplification of its locomotion analysis. In this paper, two important kinematics characteristics are analyzed in order to propose locomotion ability and effectiveness of the robot. In particular, a turning angle and a stride distance are analysed. After that, a trajectory tracking control system is designed based on the PID control low. Ideas for the control system design in this paper are both to decide an bias of an input angle function as a input of the system and to set a control period on half period of the input angle function. Finally, effectiveness of the designed control system is verified via numerical simulations. A straight line and a circle trajectory are adopted for the verification. As the results, it is shown that the designed trajectory tracking control system is capable of tracking two different trajectory. In addition, it is also shown that the designed trajectory tracking control system satisfies the kinematics analysis results from the side of view of the kinematic of the robot.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"5 1","pages":"315-320"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85194058","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283007
Srikanth Amudala, Samr Ali, N. Bouguila
This paper presents a variational learning framework for the infinite generalized Gaussian mixture (IGGM) model. The generalized Gaussian distribution (GGD) has a proven capability in modeling complex multidimensional data due to the flexibility of its shape parameter. Infinite model addresses the model selection problem; i.e., determination of the number of clusters without recourse to the classical selection criteria such that the number of mixture components increases automatically to best model available data accordingly. We also incorporate feature selection to consider the features that are most appropriate in constructing an approximate model in terms of clustering accuracy. Experimental results on a medical application and image categorization show the effectiveness of the proposed algorithm.
{"title":"Variational Inference of Infinite Generalized Gaussian Mixture Models with Feature Selection","authors":"Srikanth Amudala, Samr Ali, N. Bouguila","doi":"10.1109/SMC42975.2020.9283007","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283007","url":null,"abstract":"This paper presents a variational learning framework for the infinite generalized Gaussian mixture (IGGM) model. The generalized Gaussian distribution (GGD) has a proven capability in modeling complex multidimensional data due to the flexibility of its shape parameter. Infinite model addresses the model selection problem; i.e., determination of the number of clusters without recourse to the classical selection criteria such that the number of mixture components increases automatically to best model available data accordingly. We also incorporate feature selection to consider the features that are most appropriate in constructing an approximate model in terms of clustering accuracy. Experimental results on a medical application and image categorization show the effectiveness of the proposed algorithm.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"19 1","pages":"120-127"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85218682","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9282860
Ryuto Tanaka, T. Akiduki, Hirotaka Takahashi
In recent years, driver’s delayed recognition has caused many traffic accidents. Cognitive workload decreases awareness and delays the driver’s attention on the surrounding environment. Conventionally, the degree of cognitive workload on a driver, namely, the driving workload, is estimated from the steering pattern of the steering wheel. Direct measurements of the hand motions operating the vehicle might more easily and accurately detect the small changes caused by driving workload than conventional methods. Therefore, we investigate the effect of cognitive workload on the steering operation and hand motions of drivers, and verify the applicability of our approach to driving-workload estimation. The hand motions refers to the behavior of the hands operating the steering wheel. From the acceleration of the hands, we derive an index of the driving workload. The proposed method was experimentally evaluated on seven participants performing a dual task. The estimation accuracy of the proposed method at least matched that of the conventional steering-entropy method.
{"title":"Detection of Driver Workload Using Wrist-Worn Wearable Sensors: A Feasibility Study","authors":"Ryuto Tanaka, T. Akiduki, Hirotaka Takahashi","doi":"10.1109/SMC42975.2020.9282860","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9282860","url":null,"abstract":"In recent years, driver’s delayed recognition has caused many traffic accidents. Cognitive workload decreases awareness and delays the driver’s attention on the surrounding environment. Conventionally, the degree of cognitive workload on a driver, namely, the driving workload, is estimated from the steering pattern of the steering wheel. Direct measurements of the hand motions operating the vehicle might more easily and accurately detect the small changes caused by driving workload than conventional methods. Therefore, we investigate the effect of cognitive workload on the steering operation and hand motions of drivers, and verify the applicability of our approach to driving-workload estimation. The hand motions refers to the behavior of the hands operating the steering wheel. From the acceleration of the hands, we derive an index of the driving workload. The proposed method was experimentally evaluated on seven participants performing a dual task. The estimation accuracy of the proposed method at least matched that of the conventional steering-entropy method.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"48 1","pages":"1723-1730"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85869064","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9282862
Qinlong Wang, Yang Yang, Teng Wan, S. Du
Point cloud registration a challenging task in situations with poor initial value and scenarios with limited geometric structure. In these cases, the correct correspondence between two point clouds is unknown and difficult to establish. To cope with this problem, the semantic of partial points is introduced in this paper. Firstly, the semantic information is used to find more reasonable correspondence, i.e. semantic point pairs. Secondly, we formulate a novel objective function to integrate the matching error of semantic point pairs as guidance of registration. Thirdly, a hyperparameter is applied to balance the confidence of semantic point pairs. At last, a novel algorithm under the ICP framework is presented to optimize the rigid transformation iteratively. The evaluation of KITTI data set reveals the robustness and accuracy of our method in the complex scenes mentioned above.
{"title":"Robust Point Set Registration Based on Semantic Information","authors":"Qinlong Wang, Yang Yang, Teng Wan, S. Du","doi":"10.1109/SMC42975.2020.9282862","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9282862","url":null,"abstract":"Point cloud registration a challenging task in situations with poor initial value and scenarios with limited geometric structure. In these cases, the correct correspondence between two point clouds is unknown and difficult to establish. To cope with this problem, the semantic of partial points is introduced in this paper. Firstly, the semantic information is used to find more reasonable correspondence, i.e. semantic point pairs. Secondly, we formulate a novel objective function to integrate the matching error of semantic point pairs as guidance of registration. Thirdly, a hyperparameter is applied to balance the confidence of semantic point pairs. At last, a novel algorithm under the ICP framework is presented to optimize the rigid transformation iteratively. The evaluation of KITTI data set reveals the robustness and accuracy of our method in the complex scenes mentioned above.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"57 1","pages":"2553-2558"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85949853","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283142
Yaniv Mordecai, Dan Zadok
Mission-critical communication (MCC) enables and supports operations by providing reliable connectivity and interoperability, facilitating operational continuity and allows mission-performers to focus on mission goals and objectives. Any communication technology used in isolation may fail. For instance, land mobile radio (LMR) networks may provide poor coverage to tactical "push-to-talk" radio devices within stone buildings, while cellular devices may not satisfy strict performance criteria (e.g. setup and response time). An alternative approach would be dynamic orchestration of such hybrid communication networks, which also involve Bluetooth, cellular, and cloud-based networking technologies. We propose an integrated approach that accounts for situation, location, and connectivity considerations to enhance MCC network availability, and mission-performers’ connectivity and readiness, by harnessing communication, location, and situational awareness in networking technologies, applications, and users. This framework provides a holistic cyber-physical perspective on the problem. Our approach is useful in various real-life applications for operational connectivity of first responders, e.g. when breaking into a scene of an emergency, in which LMR coverage is expected to deteriorate significantly.
{"title":"Co-Analysis of Connectivity, Location, and Situation in Mission-Critical Hybrid Communication Networks","authors":"Yaniv Mordecai, Dan Zadok","doi":"10.1109/SMC42975.2020.9283142","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283142","url":null,"abstract":"Mission-critical communication (MCC) enables and supports operations by providing reliable connectivity and interoperability, facilitating operational continuity and allows mission-performers to focus on mission goals and objectives. Any communication technology used in isolation may fail. For instance, land mobile radio (LMR) networks may provide poor coverage to tactical \"push-to-talk\" radio devices within stone buildings, while cellular devices may not satisfy strict performance criteria (e.g. setup and response time). An alternative approach would be dynamic orchestration of such hybrid communication networks, which also involve Bluetooth, cellular, and cloud-based networking technologies. We propose an integrated approach that accounts for situation, location, and connectivity considerations to enhance MCC network availability, and mission-performers’ connectivity and readiness, by harnessing communication, location, and situational awareness in networking technologies, applications, and users. This framework provides a holistic cyber-physical perspective on the problem. Our approach is useful in various real-life applications for operational connectivity of first responders, e.g. when breaking into a scene of an emergency, in which LMR coverage is expected to deteriorate significantly.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"19 1","pages":"706-712"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76845999","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283355
Bo Wu, Xiaoya Yang, Chuxiong Sun, Rui Wang, Xiaohui Hu, Yan Hu
How to achieve efficient cooperation among agents in partially observed environments remains an overarching problem in multi-agent reinforcement learning (MARL). Value function factorization learning is a promising way as it can efficiently address multi-agent credit assignment problem. However, existing value function factorization methods have been focusing on learning fully decentralized value functions, which are not effective for some complex tasks. To address this limitation, we propose a framework which enhances value function factorization by allowing communication during execution. Communication introduces extra information to help agents understand the complex environment and learn sophisticated factorization. Furthermore, the proposed mechanism of communication differs from existing methods since we additionally design a descriptive key along with the message. By the descriptive key, agents can dynamically measure the importance of different messages and achieve attentional communication. We evaluate our framework on a challenging set of StarCraft II micromanagement tasks, and show that it significantly outperforms existing value function factorization methods.
{"title":"Learning Effective Value Function Factorization via Attentional Communication","authors":"Bo Wu, Xiaoya Yang, Chuxiong Sun, Rui Wang, Xiaohui Hu, Yan Hu","doi":"10.1109/SMC42975.2020.9283355","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283355","url":null,"abstract":"How to achieve efficient cooperation among agents in partially observed environments remains an overarching problem in multi-agent reinforcement learning (MARL). Value function factorization learning is a promising way as it can efficiently address multi-agent credit assignment problem. However, existing value function factorization methods have been focusing on learning fully decentralized value functions, which are not effective for some complex tasks. To address this limitation, we propose a framework which enhances value function factorization by allowing communication during execution. Communication introduces extra information to help agents understand the complex environment and learn sophisticated factorization. Furthermore, the proposed mechanism of communication differs from existing methods since we additionally design a descriptive key along with the message. By the descriptive key, agents can dynamically measure the importance of different messages and achieve attentional communication. We evaluate our framework on a challenging set of StarCraft II micromanagement tasks, and show that it significantly outperforms existing value function factorization methods.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"46 1","pages":"629-634"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80791674","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-11DOI: 10.1109/SMC42975.2020.9283230
L. G. Crespo
This article proposes a framework for calibrating parametric models according to data subject to uncertainty. Data uncertainty might be caused by a poor metrology system, measurement noise, model-form uncertainty or by the inability to directly measure the inputs and/or outputs of interest. The formulations developed, called Forward Maximum Likelihood (FML) and Inverse Maximum Likelihood (IML), are applicable to datasets with and without uncertainty. The FML approach performs the calibration in the space of the model’s output thereby requiring repeated model simulations. Conversely, the IML approach leverages an ensemble of solutions to an inverse problem in order to perform the calibration in the space of the model’s parameters. The potential loss of performance incurred by the IML approach is often justified by a sizable reduction in computational cost. In addition, we use chance-constrained optimization to eliminate the effects of outliers on the calibrated model. This practice yields a model that increases the likelihood of most of the data in exchange for a reduction in the likelihood of a few of the worst-performing data points. Metrics for evaluating the benefits and risks of outlier elimination are also presented.
{"title":"Forward and Inverse Approaches to Model Calibration for Uncertain Data *","authors":"L. G. Crespo","doi":"10.1109/SMC42975.2020.9283230","DOIUrl":"https://doi.org/10.1109/SMC42975.2020.9283230","url":null,"abstract":"This article proposes a framework for calibrating parametric models according to data subject to uncertainty. Data uncertainty might be caused by a poor metrology system, measurement noise, model-form uncertainty or by the inability to directly measure the inputs and/or outputs of interest. The formulations developed, called Forward Maximum Likelihood (FML) and Inverse Maximum Likelihood (IML), are applicable to datasets with and without uncertainty. The FML approach performs the calibration in the space of the model’s output thereby requiring repeated model simulations. Conversely, the IML approach leverages an ensemble of solutions to an inverse problem in order to perform the calibration in the space of the model’s parameters. The potential loss of performance incurred by the IML approach is often justified by a sizable reduction in computational cost. In addition, we use chance-constrained optimization to eliminate the effects of outliers on the calibrated model. This practice yields a model that increases the likelihood of most of the data in exchange for a reduction in the likelihood of a few of the worst-performing data points. Metrics for evaluating the benefits and risks of outlier elimination are also presented.","PeriodicalId":6718,"journal":{"name":"2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","volume":"452 1","pages":"64-69"},"PeriodicalIF":0.0,"publicationDate":"2020-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85545718","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：