Noise-Resistant Adaptive Gain Recurrent Neural Network for Visual Tracking of Redundant Flexible Endoscope Robot With Time-Varying State Variable Constraints

Abstract

Redundant systems are highly favored due to their superior performance. However, rapid real-time solutions, state variable constraints, and noise-resistance of redundant systems are still challenging issues to be solved. Therefore, a novel noise-resistant adaptive gain zeroing neural network (NRAG-ZNN) is designed to handle time-varying redundant system problems with state variable constraints, and it is successfully applied to redundant flexible endoscope robot (RFER) visual tracking. First, a tracking control scheme with tracking error, joint position, and velocity constraints is designed based on RFER's motion vision coupling model, whose inequality constraints are transformed into equality constraints by introducing a nonlinear complementary problem function. Then, a novel NRAG-ZNN is designed to handle the above time-varying redundant control scheme with state variable constraints. Next, the fast convergence of the designed NRAG-ZNN under noise interference and noise-free is rigorously proved by using the Lyapunov theory. Finally, the universality and practicability of the proposed methods are checked by using the arbitrary time-varying redundancy numerical test example and RFER. The experimental results indicate that the designed methods can effectively address the time-varying redundancy problems with state variable constraints, and have the ability to resist noise interference and better practicability. Compared with the existing works, the designed methods have higher convergence precision, faster convergence velocity, and stronger noise-resistant interference ability.