Dynamic Coverage Path Planning Algorithm for Multi - Stratospheric Airship Formation Based on Deep Reinforcement Learning

Autonomous aircraft path planning has demonstrated significant potential in various applications of the Internet of Aerial Things. However, static coverage solutions are inadequate for continuously moving aerial vehicles, such as fixed-wing aircraft and low-earth orbit satellites. Therefore, there is a pressing need for an algorithm capable of planning paths for dynamically moving aircraft. In this paper, we introduce a dynamic coverage path planning algorithm tailored for multi-airship formations that adhere to the dynamic constraints of stratospheric airships. The proposed framework leverages reinforcement learning to accumulate experience through exploration, storing it in an experience pool. This facilitates swift updates to the networks of each agent through semi-centralized exploration and centralized experience playback. Furthermore, the proposed algorithm assigns distinct rewards based on different task stages, enhancing the agent's suitability for area coverage studies.