A Self-Learnino Algorithm For The Parameter Setting Of Active Disturbance Rejection Controller And Its Application Research

Active Disturbance Rejection Control (ADRC) has been given more and moreattention due to its merits such as simple arithmetic, easy-realized, highly precise,strong robustness and so on. With the deep analysis to the structure of ADRC and itscomparison with classical PID by simulation, its better control performance has beenillustrated in this paper.Aimed at the problem of the difficult tuning of ADRC because of its manyparameters and the strong couple among them, a self-learning algorithm for theparameter setting is introduced in the paper, combined with the basic theory ofreinforcement learning, especially a special reinforcement learning algorithm. In thecertain interval, this algorithm can learn effectively a set of optimal parameters ofADRC through interaction with controlled plant. And, the algorithm is simulated for anopen-loop unstable system. The result of simulation shows the parameters of ADRClearned by the algorithm have a high control performance.In order to check its learning ability and feasibility of practical application, acorresponding learning control experiment, which is specific to a hardware platform, isdone on the hardware-in-the-loop simulation platform. The experimental result showsthis algorithm has online learning ability and practical application value.