UAV Coverage Path Planning of Multiple Disconnected Regions Based on Cooperative Optimization Algorithms

This article addresses the coverage path planning problem when an unmanned aerial vehicle (UAV) surveys an unknown site composed of multiple isolated areas. The problem is typically non-deterministic polynomial-time hard(NP-hard) and cannot be easily solved, especially when considering the scale of each area. By decomposing the problem into two cascaded subproblems—1) covering a specific polygon area; and 2) determining the optimal visiting order of different areas—an approximate solution can be found more efficiently. First, the target areas are approximated as convex polygons, and the coverage pattern is designed based on four control points. Then, the optimal visiting order is determined based on a state defined by area indices and control points. We propose two different optimization methods to solve this problem. The first method is a direct extension of the genetic algorithm, using a customized coding method. The second method is a reinforcement learning-based (RL-based) approach that solves the problem as a variant of the traveling salesman problem (TSP) through end-to-end policy training. The simulation results indicate that the proposed methods can provide solutions to the multiple-area coverage problem with competitive optimality and efficiency.