End-Effector Impedance Control of Robotic Arm Based on Enhanced Neural Network RBF-PID-PSO

Abstract

Abstract In the current calibration system for power transformers, In the current calibration systems for electrical power transformers, errors in the insertion and extraction of wiring terminals occur due to the reliance solely on camera-based positioning of the terminal's location by the robotic arm. This approach falls short in achieving precise assembly of various transformer terminals. Therefore, it is imperative to introduce active compliant control between the robotic arm's end-effector and the wiring terminal to enable smooth terminal assembly. This paper proposes a mechanical arm end-effector impedance control algorithm based on RBF-PID-PSO for terminal assembly. The algorithm employs a particle swarm optimization technique to adjust the weight matrix and center parameters within the Radial Basis Function (RBF) neural network. Simultaneously, a PID algorithm is introduced to optimize impedance control for end-effector forces. The impedance parameters obtained from the RBF neural network and the PID parameters are incorporated into the robotic arm model. This allows for the conversion of end-effector forces into position errors for impedance control, followed by the transformation of these errors into six joint angles through the robot's inverse kinematics. This ultimately achieves impedance control of end-effector forces. Comparative experiments and simulations conducted in Matlab demonstrate that this controller effectively reduces errors in force and torque control. Furthermore, it exhibits superior resistance to disturbances in force feedback signals, significantly improving compliant control performance.