A Hyper-heuristic Algorithm Based on Q-Learning for 3D Drone Trajectory Planning

Trajectory planning problem for drone is crucial for safely executing tasks in complex environments. This study proposes a hyper-heuristic algorithm to solve the above problem. Firstly, the problem is formulated as a multi-constrained optimization problem, incorporating constraints on drone performance and environmental conditions. Secondly, a hyper-heuristic algorithm based on Q-learning (HHBQL) is introduced, which utilizes Q-learning (QL) as a high-level heuristic selection strategy and two heuristic algorithms as low-level optimizers to select suitable heuristic algorithms for drone trajectory planning. The low-level heuristics include particle swarm optimization (PSO) and quantum-behaved particle swarm optimization (QPSO). Subsequently, the generated paths are smoothed using cubic B-spline curves. Finally, simulation experiments are conducted in two environments comparing the proposed algorithm with three other algorithms. Results indicate that the proposed algorithm outperforms the other three algorithms in terms of optimization performance.