Controller Design By Using Simultaneous Perturbation Stochastic Approximation With Changeable Sliding Window

Many searching problems use simultaneous perturbation stochastic approximation (SPSA) method so as to get optimal parameters nowadays. This paper proposes a model free control (MFC) method based on SPSA with changeable sliding window. The controller here is founded based on an improved SPSA algorithm whose sampling window is updated according to the variation of the system. As a kind of MFC method, the key advantage of this approach is that it can complete the controller design task by using the input and output data of the system instead of establishing a complex model for it. In order to improve the efficiency of the control method, a changeable sliding window is used here to estimate the parameters. After that convergence analysis for the improved SPSA method is also provided in this paper. Simulation results for hypersonic vehicle tracking problem demonstrate that compare with traditional SPSA algorithm this improved method can significantly improve the feasibility and efficiency.