Q-Learning Dynamic Path Planning for an HCV Avoiding Unknown Threatened Area

Abstract

For a hypersonic cruise vehicle (HCV) avoiding unknown threats from ground or space, dynamic path planning is an essential issue to fly safely in near space. Q-learning is the most typical one of reinforcement learning methods that has been effectively used for robotic path planning. However, few works is reported for an HCV with hypersonic flying characteristics considered. In this paper, a Q-learning based dynamic path planning method is presented. In the method, the vehicle is assumed as a mass point with time-invariant altitude and velocity for simplicity and the flight zone of the vehicle is modeled as an average grid map, where an ε-greedy strategy is designed to establish the Q-learning algorithm. A simulation case has illustrated the effectiveness of the given method and shows that the method is suitable for the cruise phase path planning of an HCV.