Multi-UAVs Collaborative Search Scheme in Marine Environments using Deep Reinforcement Learning

Unmanned aerial vehicles (UAVs) have mobility in harsh environments and the flexibility to modify their flight altitude to acquire information in an adaptive manner, so these vehicles can be leveraged to solve the informative path planning (IPP) problem. As most of the research on IPP currently are on static targets and terrain detection tasks, in this paper, we focus on the search for dynamic targets on the sea surface, and study the issue of collaborative search with multiple granularities using a swarm of UAVs in marine environments. To capture the objects of interest (OOI), we design a binary classifier for UAVs to sense OOI based on a depth perception model of cameras. Afterwards, we propose a collaborative search scheme using the multi-agent deep deterministic policy gradient (MADDPG) approach to enable a swarm of UAVs to simultaneously maintain formation coordination while adapting to digital pheromone strength. The experimental results show that the proposed scheme can effectively detect targets on the sea surface in a multi-granular collaborative manner.