Path Planning for Maritime Drones in a Free Space Environment Based on Reinforcement Learning

Abstract

AbstractIn modern maritime reconnaissance tasks, realizing the adaptive path planning for the maritime drones has always been a challenging problem to overcome. Existing conventional methods for path planning focus on the mesh routing environment. However, the performance in a continuous free space environment may be unsatisfactory. Therefore, this paper proposes an improved Q learning method for the path planning of maritime drones on the surface of the ocean. This work discretes the flight environment of the maritime drones into a state space represented by several state variables, which greatly reduce the state dimension. During the training process, a state-aware guided learning strategy is proposed to accelerate the convergence. Numerical simulation experiments are performed to verify the generality, effectiveness and accuracy of the proposed algorithm with applications to different free space environments. The proposed learning algorithm is obviously superior compared with the traditional Q learning.KeywordsMaritime dronePath planningFree spaceQ learning