Time-Varying Quadratic Programming by Zhang Neural Network Equipped with a Time-Varying Design Parameter γ(t)

摘要

In this paper, a recurrent neural network termed Zhang neural network (ZNN) with a time-varying design parameter γ(t) is developed and presented to solve time-varying quadratic programs subject to time-varying linear equalities. The updated design formula for the ZNN model possesses more generality because the design parameter considered is actually (e.g., in hardware implementation) time-varying, i.e., γ(t). The state vector of such a ZNN model with time-varying design parameter γ(t), can also globally exponentially converge to the theoretical optimal solution pair of the time-varying linear-equality-constrained quadratic program. To achieve superior convergence of the ZNN model, nonlinear activation functions are adopted as well, as compared with the linear-activation-function case. Simulation results substantiate the efficiency of such a ZNN model with a time-varying design parameter γ(t) aforementioned.