Optimal Guidance for Reusable Launch Vehicle in Reentry Phase Based on Adaptive Dynamic Programming with Experience Replay

An optimal tracking guidance method for Reusable Launch Vehicles (RLV) in the reentry phase is proposed based on improved Adaptive Dynamic Programming (ADP) with experience replay (ER). An actor-critic ADP with novel network weight tuning algorithms is developed. By introducing the experience replay technique, the persistence of the excitation requirement can be assessed while updating the critic neural network. Therefore, the generalization performance is improved for the ADP-based controller. Simulation of the RLV guidance system under model uncertainty is conducted. Better performances in terms of smoothness and accuracy are achieved compared with noise expansion, demonstrating the effectiveness and advantages of the proposed method for the optimal trajectory-tracking guidance of RLV.