This paper proposes an Adaptive Fuzzy Sliding Mode Controller with Reinforcement Learning-Based Membership Function Optimization (RL-AFSMC-MFO) for robot manipulator trajectory tracking and presents a proposed hardware-validated online RL-based optimization of both membership function centres and widths within a Lyapunov-guaranteed AFSMC framework.
Abstract
This paper proposes an Adaptive Fuzzy Sliding Mode Controller with Reinforcement Learning-Based Membership Function Optimization (RL-AFSMC-MFO) for robot manipulator trajectory tracking. The dynamics and stability results are derived for the general <inline-formula> <tex-math notation="LaTeX">$n$ </tex-math></inline-formula>-degree-of-freedom (DOF) Euler-Lagrange formulation; hardware validation is conducted on three actively-controlled joints (J1–J3) of the 4-DOF Dobot MG400, with the fourth joint held fixed. The framework combines a Proportional-Integral Sliding Mode surface for disturbance rejection, a Mamdani fuzzy inference system for chattering elimination, and a lightweight reinforcement learning agent that adapts membership function centres and widths online—directly reducing the geometry-induced approximation error floor that gradient-only methods cannot address. A composite Lyapunov function establishes practical stability (uniform ultimate boundedness) with recovery of global asymptotic stability as the approximation error floor vanishes. Hardware validation across five scenarios confirms RMSE below 0.05 rad in all nine test conditions. Statistical repeatability of the closed-loop controller is established via repeated independent trials across three scenarios (<inline-formula> <tex-math notation="LaTeX">$N=5$ </tex-math></inline-formula> for the nominal case, <inline-formula> <tex-math notation="LaTeX">$N=3$ </tex-math></inline-formula> each for parametric and combined uncertainty). These trials yield closely consistent mean RMSE (0.0081–0.0084 rad) and coefficient of variation (13.4–13.8%) across all three, with a 100% pass rate. Zero-shot transfer to novel trajectories requires no retraining. The proposed method is deployed and validated on a <inline-formula> <tex-math notation="LaTeX">${\$}3$ </tex-math></inline-formula>,000 open-source ROS2/Python platform. Data-Driven Linear MPC (DLMPC) on the 7-DOF, torque-controlled Franka Panda is discussed as a conceptually complementary reference method rather than a head-to-head benchmark, given the substantial differences in platform cost, actuation interface, and degrees of freedom between the two studies. We present a proposed hardware-validated online RL-based optimization of both membership function centres and widths within a Lyapunov-guaranteed AFSMC framework. To our knowledge, the same framework is also proposed to be validated under simultaneous payload-plus-parametric uncertainty and under zero-shot cross-trajectory transfer, without retraining.
This paper develops an observer-based adaptive fuzzy fractional sliding mode control (AFFSMC) approach for trajectory tracking of a four-degree-of-freedom robotic manipulator subject to nonlinear coupling, parametric uncertainty, unmodeled friction, and external disturbances. The approach integrates three components wi...
To address the limitations posed by model uncertainties and the constraints of single control methods in high-precision trajectory tracking of manipulators, this paper proposes a novel dynamic coordination strategy integrating fractional-order sliding mode control (FOSMC) and fractional-order active disturbance rejecti...
Yi-Fang Wen, Ren-Zhong Wang, Ting-Wen Huang· Bulletin of the Polish Acade...· 0 citations
Traditional Proportional–Integral–Derivative (PID) controllers are the dominant solution for Unmanned Aerial Vehicle (UAV) attitude stabilization due to their simplicity and low computational cost. However, their fixed tuning parameters limit performance under nonlinear dynamics, rapidly varying flight conditions, and...
I. Laabousse, Y. El Bouzekri El Idrissi, Mohamed Haidoury et al.· Engineering, Technology &...· 0 citations
High-precision tracking for robotic manipulators is complicated by coupled nonlinear dynamics, uncertain parameters, external disturbances, and the bounded torque that every joint actuator can deliver. To meet these challenges simultaneously, this work develops a novel adaptive fixed-time control framework resting on t...