Predefined-Time Convergent Neural Solution to Cyclical Motion Planning of Redundant Robots Under Physical Constraints

Abstract

In industrial activities, robot arms are usually requested to perform repetitive tasks. As a special recurrent neural network, zeroing neural network (ZNN) has been successfully applied to cyclical motion planning (CMP) of redundant robots. However, most of the existing ZNN models for CMP have not taken into account physical limits of robot arms. Unfortunately, if physical limits are violated, the robot will fail to complete the task and even be damaged in practice. To overcome this problem, this article proposes a novel ZNN model for CMP of physically constrained redundant robots. The ZNN model is endowed with predefined-time convergence by using various designed activation functions. Rigorous theoretical analysis is conducted to prove the predefined-time convergence of the accelerated ZNN model. Numerical experiments are then comparatively performed, showing that the ZNN model delivers superior convergence and it is applicable to CMP of redundant robots with physical constraints considered.