NN controller of the constrained robot under unknown constraint

Abstract

In this paper, the problems faced in the constrained force control is studied (uncertainties in dynamic model and the unknown constraints). A neural network (NN) controller is proposed based on the derived dynamic model of robot in the task space. The feed-forward neural network is used to adaptively compensate for the uncertainties in the robot dynamics. Training signals are proposed for the feed-forward neural network controller. The NN weights are tuned on-line, with no off-line learning phase required. An on-line estimation algorithm is developed to estimate the local shape of the constraint surface by using measured data on the force and position of the end-effector. The suggested controller is simple in structure and can be implemented easily. Real-time Experiments are conducted using the five-bar robot to demonstrate the effectiveness of the proposed controller.