Multi-UAV Coverage Path Planning Based on Q-Learning

The objective of coverage path planning (CPP) is to guarantee that unmanned aerial vehicles (UAVs) can achieve complete coverage of the target region. In previous research works, the operation mode of UAVs was to be individually responsible for the coverage tasks of each sub-region. Nevertheless, in this study, multiple UAVs operate in a collaborative manner within the entire search area. This approach enables more efficient coverage and a more flexible accomplishment of the coverage mission. Given that the traditional methods for solving the UAV CPP problem often result in a high overall planning cost, this article puts forward multiple UAVs CPP algorithm based on Q-learning. To reduce the time required for UAVs to complete the coverage search task, a grid-based rotating region division algorithm is employed to minimize the area yet to be searched. By formulating the UAV CPP model, the path-planning problem is transformed into a multiobjective function optimization problem. The double-Q-learning algorithm is utilized to balance the global search and local exploitation of the algorithm. A total cost function, which takes into account both the distance cost and the turning cost, is adopted to iteratively optimize the path planning. The results of simulation experiments demonstrate that the paths planned by the proposed algorithm can enable multiple UAVs to fully cover the target region in various target areas at a lower total cost.