arXiv - CS - Systems and Control最新文献_第10页

Digital Twin-Driven Reinforcement Learning for Obstacle Avoidance in Robot Manipulators: A Self-Improving Online Training Framework 机器人机械手避障的数字双胞胎驱动强化学习：自我完善的在线培训框架

arXiv - CS - Systems and Control Pub Date : 2024-03-19 DOI: arxiv-2403.13090

Yuzhu Sun, Mien Van, Stephen McIlvanna, Nguyen Minh Nhat, Kabirat Olayemi, Jack Close, Seán McLoone

{"title":"Digital Twin-Driven Reinforcement Learning for Obstacle Avoidance in Robot Manipulators: A Self-Improving Online Training Framework","authors":"Yuzhu Sun, Mien Van, Stephen McIlvanna, Nguyen Minh Nhat, Kabirat Olayemi, Jack Close, Seán McLoone","doi":"arxiv-2403.13090","DOIUrl":"https://doi.org/arxiv-2403.13090","url":null,"abstract":"The evolution and growing automation of collaborative robots introduce more\u0000complexity and unpredictability to systems, highlighting the crucial need for\u0000robot's adaptability and flexibility to address the increasing complexities of\u0000their environment. In typical industrial production scenarios, robots are often\u0000required to be re-programmed when facing a more demanding task or even a few\u0000changes in workspace conditions. To increase productivity, efficiency and\u0000reduce human effort in the design process, this paper explores the potential of\u0000using digital twin combined with Reinforcement Learning (RL) to enable robots\u0000to generate self-improving collision-free trajectories in real time. The\u0000digital twin, acting as a virtual counterpart of the physical system, serves as\u0000a 'forward run' for monitoring, controlling, and optimizing the physical system\u0000in a safe and cost-effective manner. The physical system sends data to\u0000synchronize the digital system through the video feeds from cameras, which\u0000allows the virtual robot to update its observation and policy based on real\u0000scenarios. The bidirectional communication between digital and physical systems\u0000provides a promising platform for hardware-in-the-loop RL training through\u0000trial and error until the robot successfully adapts to its new environment. The\u0000proposed online training framework is demonstrated on the Unfactory Xarm5\u0000collaborative robot, where the robot end-effector aims to reach the target\u0000position while avoiding obstacles. The experiment suggest that proposed\u0000framework is capable of performing policy online training, and that there\u0000remains significant room for improvement.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"26 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140196685","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}

引用次数: 0

Current-Based Impedance Control for Interacting with Mobile Manipulators 基于电流的阻抗控制与移动机械手互动

arXiv - CS - Systems and Control Pub Date : 2024-03-19 DOI: arxiv-2403.13079

Jelmer de Wolde, Luzia Knoedler, Gianluca Garofalo, Javier Alonso-Mora

{"title":"Current-Based Impedance Control for Interacting with Mobile Manipulators","authors":"Jelmer de Wolde, Luzia Knoedler, Gianluca Garofalo, Javier Alonso-Mora","doi":"arxiv-2403.13079","DOIUrl":"https://doi.org/arxiv-2403.13079","url":null,"abstract":"As robots shift from industrial to human-centered spaces, adopting mobile\u0000manipulators, which expand workspace capabilities, becomes crucial. In these\u0000settings, seamless interaction with humans necessitates compliant control. Two\u0000common methods for safe interaction, admittance, and impedance control, require\u0000force or torque sensors, often absent in lower-cost or lightweight robots. This\u0000paper presents an adaption of impedance control that can be used on\u0000current-controlled robots without the use of force or torque sensors and its\u0000application for compliant control of a mobile manipulator. A calibration method\u0000is designed that enables estimation of the actuators' current/torque ratios and\u0000frictions, used by the adapted impedance controller, and that can handle model\u0000errors. The calibration method and the performance of the designed controller\u0000are experimentally validated using the Kinova GEN3 Lite arm. Results show that\u0000the calibration method is consistent and that the designed controller for the\u0000arm is compliant while also being able to track targets with five-millimeter\u0000precision when no interaction is present. Additionally, this paper presents two\u0000operational modes for interacting with the mobile manipulator: one for guiding\u0000the robot around the workspace through interacting with the arm and another for\u0000executing a tracking task, both maintaining compliance to external forces.\u0000These operational modes were tested in real-world experiments, affirming their\u0000practical applicability and effectiveness.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"151 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140197026","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}

引用次数: 0

Most Likely Sequence Generation for $n$-Grams, Transformers, HMMs, and Markov Chains, by Using Rollout Algorithms 利用滚动算法为 $n$ 格、变换器、HMM 和马尔可夫链生成最可能序列

arXiv - CS - Systems and Control Pub Date : 2024-03-19 DOI: arxiv-2403.15465

Yuchao Li, Dimitri Bertsekas

{"title":"Most Likely Sequence Generation for $n$-Grams, Transformers, HMMs, and Markov Chains, by Using Rollout Algorithms","authors":"Yuchao Li, Dimitri Bertsekas","doi":"arxiv-2403.15465","DOIUrl":"https://doi.org/arxiv-2403.15465","url":null,"abstract":"In this paper we consider a transformer with an $n$-gram structure, such as\u0000the one underlying ChatGPT. The transformer provides next word probabilities,\u0000which can be used to generate word sequences. We consider methods for computing\u0000word sequences that are highly likely, based on these probabilities. Computing\u0000the optimal (i.e., most likely) word sequence starting with a given initial\u0000state is an intractable problem, so we propose methods to compute highly likely\u0000sequences of $N$ words in time that is a low order polynomial in $N$ and in the\u0000vocabulary size of the $n$-gram. These methods are based on the rollout\u0000approach from approximate dynamic programming, a form of single policy\u0000iteration, which can improve the performance of any given heuristic policy. In\u0000our case we use a greedy heuristic that generates as next word one that has the\u0000highest probability. We show with analysis, examples, and computational\u0000experimentation that our methods are capable of generating highly likely\u0000sequences with a modest increase in computation over the greedy heuristic.\u0000While our analysis and experiments are focused on Markov chains of the type\u0000arising in transformer and ChatGPT-like models, our methods apply to general\u0000finite-state Markov chains, and related inference applications of Hidden Markov\u0000Models (HMM), where Viterbi decoding is used extensively.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"2016 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140297572","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}

引用次数: 0

Hybrid Unsupervised Learning Strategy for Monitoring Industrial Batch Processes 用于监控工业批量生产过程的混合无监督学习策略

arXiv - CS - Systems and Control Pub Date : 2024-03-19 DOI: arxiv-2403.13032

Christian W. Frey

引用次数: 0

Autonomous Underground Freight Transport Systems -- The Future of Urban Logistics? 自主地下货运系统--城市物流的未来？

arXiv - CS - Systems and Control Pub Date : 2024-03-13 DOI: arxiv-2403.08841

Lasse Bienzeisler, Torben Lelke, Bernhard Friedrich

引用次数: 0

Adaptive Gain Scheduling using Reinforcement Learning for Quadcopter Control 利用强化学习进行四旋翼飞行器控制的自适应增益调度

arXiv - CS - Systems and Control Pub Date : 2024-03-12 DOI: arxiv-2403.07216

Mike Timmerman, Aryan Patel, Tim Reinhart

引用次数: 0

Characterizing Flow Complexity in Transportation Networks using Graph Homology 利用图同源性表征交通网络中的流量复杂性

arXiv - CS - Systems and Control Pub Date : 2024-03-09 DOI: arxiv-2403.05749

Shashank A Deshpande, Hamsa Balakrishnan

引用次数: 0

Secure and Energy-efficient Unmanned Aerial Vehicle-enabled Visible Light Communication via A Multi-objective Optimization Approach 通过多目标优化方法实现安全节能的无人机可见光通信

arXiv - CS - Systems and Control Pub Date : 2024-03-03 DOI: arxiv-2403.15410

Lingling Liu, Aimin Wang, Jing Wu, Jiao Lu, Jiahui Li, Geng Sun

{"title":"Secure and Energy-efficient Unmanned Aerial Vehicle-enabled Visible Light Communication via A Multi-objective Optimization Approach","authors":"Lingling Liu, Aimin Wang, Jing Wu, Jiao Lu, Jiahui Li, Geng Sun","doi":"arxiv-2403.15410","DOIUrl":"https://doi.org/arxiv-2403.15410","url":null,"abstract":"In this research, a unique approach to provide communication service for\u0000terrestrial receivers via using unmanned aerial vehicle-enabled visible light\u0000communication is investigated. Specifically, we take into account a unmanned\u0000aerial vehicle-enabled visible light communication scenario with multiplex\u0000transmitters, multiplex receivers, and a single eavesdropper, each of which is\u0000equipped with a single photodetector. Then, a unmanned aerial vehicle\u0000deployment multi-objective optimization problem is formulated to simultaneously\u0000make the optical power received by receiving surface more uniform, minimize the\u0000amount of information collected by a eavesdropper, and minimize the energy\u0000consumption of unmanned aerial vehicles, while the locations and transmission\u0000power of unmanned aerial vehicles are simultaneously optimized under certain\u0000constraints. Since the formulated unmanned aerial vehicle deployment\u0000multi-objective optimization problem is complex and nonlinear, it is\u0000challenging to be tackled by using conventional methods. For the purpose of\u0000solving the problem, a multi-objective evolutionary algorithm based on\u0000decomposition with chaos initiation and crossover mutation is proposed.\u0000Simulation outcomes show that the proposed approach is superior to other\u0000approaches, and is efficient at improving the security and energy efficiency of\u0000visible light communication system.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"119 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140303357","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}

引用次数: 0

Homeostatic motion planning with innate physics knowledge 利用与生俱来的物理知识进行恒定运动规划

arXiv - CS - Systems and Control Pub Date : 2024-02-23 DOI: arxiv-2402.15384

Giulia Lafratta, Bernd Porr, Christopher Chandler, Alice Miller

引用次数: 0

Spatiotemporal Observer Design for Predictive Learning of High-Dimensional Data 高维数据预测学习的时空观测器设计

arXiv - CS - Systems and Control Pub Date : 2024-02-23 DOI: arxiv-2402.15284

Tongyi Liang, Han-Xiong Li

引用次数: 0