A Self-learning Terrain-following Method for Aircrafts

This paper addresses the decision problem for terrain-following flights (TFFs) of aircrafts in unknown environments, where global trajectory planning methods cannot be applied directly. In such cases, the uncertainties and complexities of the terrain are needed to be dealt with especially. In order to realize the automatic generation of a terrain-following (TF) controller, a self-learning TF method is proposed in this paper. The self-learning TF method is based on a MDP model and utilizes reinforcement learning algorithm to optimize the maneuver strategies of the aircraft. The TF controller is generated off-line, and then it's applied and tested in some unknown environments. Simulation results illustrate the effectiveness of this method.