Circuit Implementation of Fixed-Time Zeroing Neural Network for Time-Varying Equality Constrained Quadratic Programming

This paper proposes a novel circuit framework of zeroing neural network for time-varying equality constrained quadratic programming problems (TEQPPs). It is proved that the designed circuit can not only parallel solve TEQPPs in a fixed time, but also cost less hardware resources to be implemented by virtue of its simple structure. Rigorous analysis derives the convergence time upper bound of this novel circuit framework in noiseless and bounded noise polluted condition respectively. Moreover, this circuit can also avoid calculating the pseudoinverse of coefficient matrix when solving TEQPPs. Several circuit experiments are simulated to validate those conclusions.