Reinforcement learning-based multi-strategy cuckoo search algorithm for 3D UAV path planning

Abstract

Unmanned aerial vehicles are applied extensively in various fields due to their advantages of low-cost, high-maneuverability, and easy-operation. However, the path planning problem of unmanned aerial vehicles, which directly determines the flight safety and efficiency, still remains challenging when building and optimizing the path model. To further study the path planning problem, we firstly construct it as a constrained optimization problem. The objective function considers the costs of path length and threat, and the constraints involve the collision and turning angle. Additionally, we employ the theory of B-Spline curve to represent the planned paths to facilitate the optimization of established model. Then, aiming at the poor searchability and slow convergence speed of current optimization methods, we propose a reinforcement learning-based multi-strategy cuckoo search algorithm. Specifically, we establish an innovative reinforcement learning-based multi-strategy mechanism and a reinforced switch parameter based on the theory of reinforcement learning. To verify the effectiveness of the proposed algorithm, extensive experiments are carried out on the CEC’17 benchmark test and different three-dimensional path planning problems. Detailed statistical analysis of the experimental results confirm the superiority of our proposed algorithm to the other well-established algorithms.