Joint Space Based Force Sensorless Bilateral Control with BP Neural Network Gravity Compensation for 6-PSS Parallel Actuator

Abstract

Bilateral control systems without force sensors are widely used in human system interaction. In order to improve the accuracy of force estimation, an active gravity compensation based on BP neural network is proposed, and based on this, a bilateral control framework based on disturbance observer and reaction force observer for 6-PSS parallel actuator is established. Compared with the Newton-Euler method to establish a dynamic model for gravity compensation, this method does not require real-time forward kinematics solutions, thereby avoiding complex numerical calculations and non-convergence. In addition, the proposed method improves the accuracy of force estimation and the transparency of the system. Experiments are conducted using 6-PSS parallel actuators in an experimental setup to demonstrate the effectiveness of the proposed method.