arrow
Return

Reinforcement Learning-Based IT-2 Fuzzy Fractional-Order Sliding Mode Control for Leader–Follower Formation Tracking of Flexible-Joint Robots

delete2025-12-02
delete0
PRE
AI
G
G. Narayanan
S
Sangmoon Lee
S
Sangtae Ahn
DOI:10.1109/TFUZZ.2025.3639438delete
deleteOriginal
deleteShare
deleteSave
Abstract

Abstract

En 中文
This study presents a reinforcement learning-based interval type-2 (IT-2) fuzzy formation tracking control strategy for leader–follower flexible joint robots (FJRs), addressing challenges arising from nonlinearities, unmodeled dynamics, and unknown external disturbances. The proposed approach employs fractional-order sliding mode control (FOSMC) to guarantee finite-time convergence of the tracking error with rigorously proven stability. Unlike conventional backstepping methods, the IT-2 fuzzy FOSMC framework effectively addresses uncertainties and disturbances. By considering the memory effect of fractional calculus, a Hamilton–Jacobi–Bellman equation for the formation tracking dynamics is derived using an auxiliary system and equivalent transformation. A novel cost function based on the derivative of the formation tracking error is introduced, resulting in a discounted cost that accounts for the Laplacian matrix $\widehat{L}$ under a performance constraint. The optimal cost function and formation sliding mode control policy are shown to be gradually approximable via policy iteration. To achieve intelligent near-optimal control, a reinforcement learning algorithm based on an actor–critic neural network architecture is implemented, enabling all agents to compensate for system uncertainties and maintain formation tracking with the leader. Lyapunov stability analysis confirms that all follower FJRs can track the leader using actor–critic update laws for neural network weights, with distributed tracking error designs ensuring convergence to optimal values. Simulation results validate the proposed approach, demonstrating reduced chattering, improved formation tracking, and enhanced robustness.
Keywords:
Formation tracking
fractional-order sliding mode control (FOSMC)
interval type -2 (IT-2) fuzzy
leader–follower robots
reinforcement learning (RL)

Journal

IEEE Transactions on Fuzzy Systems cover
IEEE Transactions on Fuzzy Systems
IF:
11.9
Papers:
5.1K
Citations:
2.9W

Organization

K
kyungpook national university
Scholars:
842
Papers: 347
Citations: 1
Cited Papers

Cited Papers

Command filtered backstepping control of constrained flexible joint robotic manipulator
err2023-07-28
err0
errOAAI
errMohammad Mehdi Arefi; Navid Vafamand; Behrouz Homayoun; Mohammadreza Davoodi
errShare
errSave
Dynamic Observer-Based H-Infinity Consensus Control of Fractional-Order Multi-Agent Systems
err
IF0
err2025-01-01
err0
PREAI
errHe Li; Shuo Liu; Guanglei Meng; Quanyong Fan
errShare
errSave
researcher View more