{"title":"Data-Driven Inverse Reinforcement Learning for Heterogeneous Optimal Robust Formation Control.","authors":"Fatemeh Mahdavi Golmisheh, Saeed Shamaghdari","doi":"10.1109/TCYB.2025.3546563","DOIUrl":null,"url":null,"abstract":"<p><p>This article presents novel data-driven inverse reinforcement learning (IRL) algorithms to optimally address heterogeneous formation control problems in the presence of disturbances. We propose expert-estimator-learner multiagent systems (MASs) as independent systems with similar interaction graphs. First, a model-based IRL algorithm is introduced for the estimator MAS to determine its optimal control and reward functions. Using the estimator IRL algorithm results, a robust algorithm for model-free IRL is presented to reconstruct the learner MAS's optimal control and reward functions without knowing the learners' dynamics. Therefore, estimator MAS aims to estimate experts' desired formation and learner MAS wants to track the estimators' trajectories optimally. As a final step, data-driven implementations of these proposed IRL algorithms are presented. Consequently, this research contributes to identifying unknown reward functions and optimal controls by conducting demonstrations. Our analysis shows that the stability and convergence of MASs are thoroughly ensured. The effectiveness of the given algorithms is demonstrated via simulation results.</p>","PeriodicalId":13112,"journal":{"name":"IEEE Transactions on Cybernetics","volume":"PP ","pages":""},"PeriodicalIF":9.4000,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Cybernetics","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1109/TCYB.2025.3546563","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"AUTOMATION & CONTROL SYSTEMS","Score":null,"Total":0}
引用次数: 0
Abstract
This article presents novel data-driven inverse reinforcement learning (IRL) algorithms to optimally address heterogeneous formation control problems in the presence of disturbances. We propose expert-estimator-learner multiagent systems (MASs) as independent systems with similar interaction graphs. First, a model-based IRL algorithm is introduced for the estimator MAS to determine its optimal control and reward functions. Using the estimator IRL algorithm results, a robust algorithm for model-free IRL is presented to reconstruct the learner MAS's optimal control and reward functions without knowing the learners' dynamics. Therefore, estimator MAS aims to estimate experts' desired formation and learner MAS wants to track the estimators' trajectories optimally. As a final step, data-driven implementations of these proposed IRL algorithms are presented. Consequently, this research contributes to identifying unknown reward functions and optimal controls by conducting demonstrations. Our analysis shows that the stability and convergence of MASs are thoroughly ensured. The effectiveness of the given algorithms is demonstrated via simulation results.
期刊介绍:
The scope of the IEEE Transactions on Cybernetics includes computational approaches to the field of cybernetics. Specifically, the transactions welcomes papers on communication and control across machines or machine, human, and organizations. The scope includes such areas as computational intelligence, computer vision, neural networks, genetic algorithms, machine learning, fuzzy systems, cognitive systems, decision making, and robotics, to the extent that they contribute to the theme of cybernetics or demonstrate an application of cybernetics principles.