On the reinforcement learning extended state observer for a class of uncertain sampled-data control systems

In the last few decades,the extended state observer(ESO)[1]has been demonstrated to be an effective tool for dealing with uncertain control systems and many modi-fied ESOs have been proposed to get the desired estimation performance[2-4].Nevertheless,most of these methods as-sume that part of the model information is already known.Therefore,a data-driven tuning law for ESO without assum-ing model information for disturbances and noise seems to be a significant open issue.