Trajectory Tracking Control of A 2-Dof Robot Manipulator Using PSO-Optimized ANFIS Trained on Fuzzy-Pd Control Data

Trajectory tracking in robot manipulators is challenging because of nonlinear dynamics and interactions between joint motions. This study investigates a learning-based control framework for a two-degree-of-freedom (2-DOF) robot manipulator by integrating an expert-designed fuzzy proportional–derivative (Fuzzy-PD) controller, an adaptive neuro-fuzzy inference system (ANFIS), and particle swarm optimization (PSO). The Fuzzy-PD controller first provides a baseline tracking policy that incorporates expert knowledge into the control response. State data and the corresponding control signals generated by this controller are collected to construct a supervised learning dataset for a five-layer ANFIS architecture. Through this learning process, ANFIS approximates the nonlinear mapping between the selected inputs and the reference control actions, transferring the baseline controller’s behavior to a trainable neuro-fuzzy representation. PSO is incorporated into parameter optimization to explore the ANFIS search space and reduce dependence on the initialization and local search characteristics of conventional hybrid learning. The resulting controller is evaluated through numerical simulations of the 2-DOF manipulator performing trajectory tracking tasks. The reported responses indicate accurate reference tracking and rapid convergence of tracking errors under the simulated conditions. The framework offers a systematic approach to combining expert control knowledge, data-driven function approximation, and population-based optimization for nonlinear robotic control. Its practical value lies in reducing reliance on manual fuzzy-controller tuning while retaining a structured neuro-fuzzy representation of the control policy.