Optimal tracking controllers with Off-policy Reinforcement Learning Algorithm in Quadrotor

Abstract

In this study, the optimal tracking control problem for the quadrotor which is a highly coupling system with completely unknown dynamics is addressed based on data by introducing the reinforcement learning (RL) technique. The proposed Off-policy RL algorithm does not need any knowledge of quadrotor model. By collecting data, which is the states of quadrotor system then using an actor-critic networks (NNs) to solve the optimal tracking trajectory problem. Finally, simulation results are provided to illustrate the effectiveness of proposed method.