Recurrent neural network based second order sliding mode control of redundant robot manipulators

Abstract

In this paper, the second-order sliding-mode control approach integrated with the recurrence neural network (RNN) is applied to the redundant robot manipulator. In order to design the joints trajectories, first, the performance index is defined as the sum of squares of the final trace tracking error. Then, regarding the constraint of joints physical condition, a constrained optimization problem is obtained. Using projection theory, the dynamic model of RNN is extracted. The simplicity of RNN structure and its high convergence analysis are of the main features. To design second-order sliding-mode control law, first, the proper sliding surface is defined. Then, using the modified sliding condition, the control law is extracted. The controller momently applied the last data of the RNN as the reference input. Simulations are applied to the 7-DOF PA-10 robot manipulator.