Advances in multi-agent polygonal area coverage

We address the problem of multi-agent repeated coverage of polygonal environments. The map of the environment and its static polygonal obstacles are given and each agent is equipped with a panoramic visual sensor with limited visibility range. Two graph-based methods are proposed to represent the environment based on the Visibility Graph and the Constrained Delaunay Triangulation. Using the two graph-based representations of the area, four coverage mechanisms are proposed and developed. The impact of the proposed graph-based representations are then investigated on the performance of the four variations of the coverage algorithm in terms of visiting frequency of points of the target area, total distance the agents traverse and quality of the workload distribution among the agents. The results show that optimizing each of the mentioned criteria requires a specific environment representation and choice of coverage algorithm. Therefore, more nuanced conclusions and guidelines are drawn from the results.