Collaborative Path Planning of Multiple Carrier-based Aircraft Based on Multi-agent Reinforcement Learning

Path planning of carrier-based aircraft is of great significance to improve the scheduling efficiency on the aircraft carrier deck. However, it is not easy to find the optimal paths for multiple carrier-based aircraft since the environment of carrier deck is highly dynamic and complex. To overcome this issue, we propose a collaborative path planning model based on multi-agent reinforcement learning. The collaborative path planning for multiple carrier-based aircraft is modeled as a multi-agent reinforcement learning problem, and we build a model based on the state and action space of the carrier-based aircraft. Then we train the model in the simulated gird environment of USS Ford. Finally, the performance of the proposed model is evaluated by experiments under three fixed scenarios and ten random scenarios, and results are shown in the form of simulation visualization. The experimental results show that compared with RRT-Star algorithm, PSO algorithm and deep reinforcement learning DQN model, the proposed model has lower response time, higher completion rate and shorter average path length.