Learning Control of a Robotic System Using Neural Networks

Abstract In this paper, deterministic learning control using neural networks (NNs) is presented for a robotic system with unknown system dynamics. The dynamics of the robotic system are represented by an n-link strict robotic manipulator. The adaptive NNs is employed as the first control strategy to approximate the unknown model of the system and adapt interactions between the robot and a patient. Deterministic learning control using learned knowledge from direct NNs with Radial Basis Functions (RBFs) is employed as the second control strategy to improve the system intelligence for energy conservation and reduce control tasks. Uniform ultimate boundedness (UUB) of the closed loop system is achieved under the condition of the Lyapunov's stability with full state feedback control. Extensive simulations are carried out to expound the efficacy of the proposed control strategies and the advancement of learning control.