Optimal Control of Switched System Based on Neural Network Optimization

The optimal control problem of switched system is to find both the optimal control input and optimal switching signal and is a mixed integer problem. High computational burden in solving this problem is a major obstacle. To solve this problem, this paper presented hybrid neural network combining continuous neurons and discrete neurons and designed lyapunov function to guarantee the convergency of proposed hybrid neural network. This new solution method is more suitable to parallel implementation than the mathematical programming. Simulation results show that this approach can utilize fast converge property and the parallel computation ability of neural network and apply to real-time control.