Inverse-Free Neurodynamic Approach With Self-Adaptive Gain for Time-Varying Quadratic Programming and Applications

Abstract

This letter proposes an inverse-free, noise-tolerant neurodynamic approach with a self-adaptive gain for solving time-varying quadratic programming problems (TVQPs). The proposed neurodynamic approach avoids inverting the coefficient matrix of TVQPs, resulting in lower computational complexity. It is demonstrated that the proposed approach ensures fixed-time convergence in noiseless conditions, and it achieves asymptotic convergence without requiring to anticipate the magnitudes of additive noises in noisy conditions. Additionally, the self-adaptive gain converges to a bounded constant rather than infinity in both noiseless and noisy scenarios. Simulation studies conducted on the redundant manipulator motion planning and ridge regression problem validate the effectiveness of the proposed approach.