Visual servoing with velocity observer and neural compensation

Abstract

The normal visual servoing of robot has two drawbacks: it needs joint velocity sensors, and cannot guarantee zero steady state error. We make two modifications to overcome these problems. Sliding-mode observer is applied to estimate the joint velocities, and a RBF neural network is used to compensate gravity and friction. Based on Lyapunov and input-to-state stability analysis, we prove the stability of visual servoing system with observer and RBF neural networks.