H_? Tracking Control Research Based On Reinforcement Learning Technology

Artificial intelligence,as one of the three most advanced technologies in the 21 st century,is changing the world with its great influence.It involves a wide range of disciplines and is very challenging.Machine learning is a powerful driving force for the continuous development of artificial intelligence,and reinforcement learning,as a popular algorithm in machine learning,is also widely used in all walks of life.Using reinforcement learning technology to solve the problem of optimal control has attracted the attention of many scholars.Based on this topic,this paper has carried out in-depth research.In practical application,it is difficult to measure the state of the controlled system directly,and the cost of measurement is very high.Therefore,it is challenging for the system with unknown model parameters to learn the optimal controller from the state data.We propose a reinforcement learning Q-learning algorithm to solve the tracking control problem of linear discrete-time systems.At the same time,according to whether the interference term is considered in the system,two algorithms are proposed.The simulation results show that the system has better tracking effect when considering the interference term.Based on this,the main research work of this paper includes:1.For discrete-time unknown linear systems,an off-policy Q-learning algorithm is proposed to solve the LQT control problem.In order to solve this problem,this paper proposes a non strategic Q-learning method by constructing an augmented non minimum state space equation model transformation method,which does not require the system model parameters to be known,does not use the system state data,does not need to design a state observer,and only uses the input and output incremental data to make the system track the reference signal stably.2.The disturbance term is considered in the system,and the proposed algorithm is optimized.The ordinary optimal tracking control problem is transformed into the H _? tracking control problem.Compared with the general tracking control problem,the H _? tracking control problem has better anti-interference performance,which is verified in the simulation experiment.In view of the anti-interference and better tracking effect,this algorithm is used to simulate the outlet temperature of SC-1 ethylene cracking furnace.The results show that this algorithm can track the outlet temperature of ethylene cracking furnace well,which shows that this algorithm has a certain practical significance.3.The contribution of this paper is to consider the challenging problems brought by the unknown system model parameters and unmeasurable state to the implementation of state feedback control,and to avoid the computational complexity brought by the unknown system model parameters to the design of state observer and output feedback controller.The system can track the reference target stably and has a certain anti-jamming ability.