| Approximate iterative algorithms based on reinforcement learning and adaptive dynamic programming can solve the model-dependent problem in the field of optimal preview control,thereby designing a data-based model-free controller construction algorithm,which has strong theoretical research significance and practical engineering significance.The core knowledge of Q-Learning algorithm,adaptive dynamic programming theory and optimal control theory is used to construct optimal preview controllers for multirate systems and discrete-time systems with different sampling periods.The specific research work is as follows:(1)The problem of constructing a model-free optimal preview controller for discrete-time systems is studied.Firstly,using the forward difference factor and the method of constructing augmented error system,an augmented expanded error system equation is constructed using system state,tracking error and previewable reference signal information;Secondly,using the core knowledge of Q-Learning algorithm and adaptive dynamic programming theory,the objective function of obtained augmented error system is constructed into Bellman equation,and it is converted into the form of Q-Function,so as to design a model-free controller construction algorithm;Finally,the auxiliary matrix and recursive least squares technique are used to design the implementation of proposed algorithm,and simulation experiments are used to prove the feasibility of theory and the superiority of algorithm.(2)The model-free optimal preview controller construction problem for a multirate system with a sampling interval of 2 is studied.Using discrete boosting technique,the multirate system with two sampling periods is upgraded to a single-sampling rate system,and then the forward difference factor and the construction method of augmented error system are used to obtain corresponding augmented error system equation;Based on the core knowledge of Q-Learning algorithm and adaptive dynamic programming theory,a model-free optimal preview controller design algorithm is designed for augmented error system;Matrix and recursive least squares techniques are used to design the implement of algorithm.Simulation experiments verify the feasibility of theory and the superiority of algorithm.(3)The model-free optimal preview controller construction problem for a multirate system with a sampling interval of N is studied.Compared with a multirate system with a sampling interval of 2,the system with a sampling interval of N is more general.Based on high-dimensional discrete boosting technology reconstructs the system into a system with a single sampling rate,and then converts it into an augmented error system equation,and finally designs the corresponding model-free optimal preview controller design algorithm,and designs the corresponding implementation method. |
