An adaptive Q-learning based particle swarm optimization for multi-UAV path planning

In recent times, the path planning of unmanned aerial vehicles (UAVs) in 3D complex flight environments has become a hot topic in the field of UAV technology. Path planning is a crucial process that involves determining the trajectory of the UAV from the point of origin to its destination. However, a number of algorithms proposed for this task have been proven inefficient in this 3D space. In response, this paper proposes the use of an adaptive Q-Learning based particle swarm optimization to tackle the problem. This algorithm introduces the Q-Learning algorithm and designs four states and actions for each particle. Based on the accumulated experience from reinforcement learning, the particles can choose the appropriate action in different states. To evaluate the performance of the AQLPSO algorithm, extensive simulation experiments were conducted. These experiments involved comparing the AQLPSO algorithm with existing algorithms such as PSO, PSO-SA, and RMPSO. The results of the simulations demonstrated that the AQLPSO algorithm outperformed these algorithms in terms of multiple performance metrics. It effectively solved the UAV path planning problem in 3D complex flight environments by reducing the likelihood of falling into local optima, improving efficiency, and achieving faster convergence towards the global optimal solution.