Terminal Sliding Mode Control—Neural Network for Arm Movement With Online Optimization Algorithm to Generate Optimal Joint Paths

ABSTRACT The research here focuses on regulating the motion of a three‐link human arm model in a plane with strong performance against disturbances from the external environment, unmodeled dynamics, and system uncertainty. A continuous terminal sliding mode controller (TSMC) is the adaptive and robust method for achieving this objective. Though the TSMC offers finite‐time convergence of the tracking error to zero, it does not eliminate the chattering effect of sliding mode control. In order to counter this limitation, a hybrid method is proposed where the TSMC is integrated with a boundary layer around the sliding surface and a recurrent neural network with one hidden layer. Apart from this integration, joint‐space paths cannot be uniquely predetermined due to the arm model's kinematic redundancy. An online routing algorithm is thus integrated with the hybrid method for generating optimal paths in real‐time during purposeful arm movement toward the target object. Simulation experiment results show that the proposed hybrid method with the integration of the online routing algorithm suppresses chattering considerably, offers accurate joint‐space trajectory tracking, and has a negligible error in following the prescribed workspace paths of the end‐effector.