A New Parameter-Changing Integral ZNN Model With Nonlinear Activation for Solving Inequality Constraint Time-Varying Quadratic Programming

Abstract

The conventional zeroing neural network (ZNN) model faces significant challenges in handling time-varying noise, with its convergence speed being highly sensitive to initial conditions. In this paper, we propose a new parameter-changing integral ZNN model with nonlinear activation (NAPCIZNN) to effectively tackle time-varying quadratic programming problems with inequality constraints (IC-TVQP). By integrating a nonlinear activation function and dynamic parameter adjustment, the proposed NAPCIZNN model exhibits superior convergence speed and robust noise tolerance. We rigorously derive the theoretical upper bound for convergence time under noisy environments, providing a strong foundation for the model’s reliability. Comprehensive numerical simulations demonstrate that NAPCIZNN significantly outperforms traditional ZNN variants—including the original ZNN, nonlinear activated ZNN, integral ZNN, and piecewise variable parameter ZNN—in solving time-varying quadratic programming problems. Moreover, the practical application of the NAPCIZNN model in controlling the PUMA560 robotic manipulator showcases its robustness and precision in real-world scenarios. Empirical evidence from these applications validates the model’s exceptional capability in executing complex butterfly trajectory tracking controls with high accuracy and reliability.