Reinforcement Learning for Robot Optimal Impedance Control: Theoretical Feasibility and Extensibility Analysis

Robot contact control in unstructured environments remains a challenging problem. Reinforcement learning (RL) has been proven effective and widely exploited in robot control. However, existing studies lack attention to the theoretical feasibility of RL in contact control and its extensibility to dif...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:IEEE transactions on industrial electronics (1982) s. 1 - 11
Hlavní autori: Wu, Han, Hu, Qinglei, Zheng, Jianying, Shao, Xiaodong, Liu, Yueyang, Li, Dongyu
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: IEEE 01.01.2025
Predmet:
ISSN:0278-0046, 1557-9948
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:Robot contact control in unstructured environments remains a challenging problem. Reinforcement learning (RL) has been proven effective and widely exploited in robot control. However, existing studies lack attention to the theoretical feasibility of RL in contact control and its extensibility to different environment models. To overcome these limitations, this article proposes an integral RL (IRL)-based optimal impedance control framework by means of perturbed system theory and incremental control technique. As a stepping stone, a perturbed system is built by incorporating the measurement error and the impedance controller into the linear environment dynamics. Then, a model-free IRL algorithm is designed to learn the optimal impedance control gain iteratively. The exponential stability of the nominal system and the uniform ultimate boundedness of the perturbed system are rigorously analyzed using the Lyapunov theory. The extended analysis based on incremental IRL reveals that the proposed algorithm can be directly utilized for nonlinear contact control without any modification. Finally, numerical simulations and hardware experiments validate the theoretical results.
ISSN:0278-0046
1557-9948
DOI:10.1109/TIE.2025.3607988