Research On The Data-Driven Control Strategies For Model-Unknown Systems Based On Lazy Learning

With the improvement of the complexity and the difficulty in modeling for the controlled objects, the design of the controller becomes more and more difficult. In view of this situation, data-driven control, which can design the controller only using the of controlled objects, was introduced into the control field.The research on trajectory tracking of model-unknown system based on Lazy Learning algorithm is proposed in the paper. Two kinds of data-driven control strategies based on Lazy Learning are presented. The theoretical analysis and simulation study for the trajectory tracking on a class of model-unknown system are conducted in the paper. The main work of this paper is as follows:Firstly, the Lazy Learning algorithm(LL), which has strong adaptability, is chosen to complete the online local modeling by comparing some advantages and disadvantages of the common data-drive control strategies.Secondly, the volume complexity of LL algorithm is decreased in this paper. A new database cluster strategy named Flag accordance rule is proposed. Next, a KFCM-Flag database cluster strategy is formed using the Flag accordance rule and Kernel-based fuzzy C-means in feature space(KFCM-F). The KFCM-Flag cluster strategy can simplify the instant local modeling process of LL. Then, the Center-Flag database hierarchical search strategy is put forward to choose the most suitable modeling information vectors with the k-vnn method. Moreover, a k-vnn-Flag data updating strategy is presented to ensure the new data updated in real time and prevent the database too redundant.Thirdly, a LL-GMVC controller is designed on the basis of improved LL algorithm and the generalized minimum variance control theory. What’s more, the simulation research on trajectory tracking control on a class of model-unknown system using LL-GMVC controller is carried out.Fourthly, a LL-PSO control method is proposed based on improved LL algorithm and PSO algorithm. Next, the objective function of trajectory tracking control is formed and the optimal control actions is gotten using LL-PSO. In addition, the simulation research on trajectory tracking control on a class of model-unknown system using LL-PSO controller is carried out. What’s more, several conclusions between LL-GMVC and LL-PSO are obtained by contrasting the two simulation results.Finally, the work in this paper is summarized. In addition, the future research is prospected.