A minimized zero mean entropy approach to networked control systems

A novel control method is proposed for networked control systems with nonlinear process, probably non-Gaussian process noise and time delays. The performance index of closed loop control system consists of entropy, mean value and control energy constraint. Two stochastic control methods for networked control systems are given under the same general frame. One method utilizes gradient optimal techniques to obtain an optimal control law, and the stability of closed loop systems is analyzed. The other method gives the optimal solution of control law directly. The implementation issues are emphasized. The sliding window techniques and non-parametric probability density function estimation techniques are employed to obtain the entropy of tracking error recursively. As a result, the controller can be on-line tuned. Finally, the methodology is illustrated by simulations.