Abstract

We present a Reinforcement Learning-based Robust Nonlinear Model PredictiveControl (RL-RNMPC) framework for controlling nonlinear systems in the presenceof disturbances and uncertainties. An approximate Robust Nonlinear ModelPredictive Control (RNMPC) of low computational complexity is used in which thestate trajectory uncertainty is modelled via ellipsoids. Reinforcement Learningis then used in order to handle the ellipsoidal approximation and improve theclosed-loop performance of the scheme by adjusting the MPC parametersgenerating the ellipsoids. The approach is tested on a simulated Wheeled MobileRobot (WMR) tracking a desired trajectory while avoiding static obstacles.

 

Quick Read (beta)