A Practical Method to Cover Evenly a Dynamic Region With a Swarm

Many applications require exploring or monitoring a region. This can be achieved by a sensor network, a large team of robots which can cover only a very small fraction each. When the region is convex, small, and static, it suffices to deploy the robots as a Centroidal Voronoi Tessellation (CVT). Instead, we consider that the area to cover is wide, not necessarily convex, and complex. Then, a smaller simple region is maneuvered and deformed to rake the full area. A few waypoints describing the region along time are provided to the robots. The goal is that the robots coordinate to dynamically deploy over this region evenly, near a CVT. Unfortunately, the distributed CVT computation algorithm converges too slowly for such exploration method to be practical. In this work, CVT computation is complemented with feedback and feedforward based control techniques, and dynamic consensus, to adjust the robot speeds so that they coordinate to cover the dynamic region. We demonstrate in simulation that the proposed method succeeds to achieve the goal of tracking the region, with the robots evenly deployed, while keeping the connectivity and avoiding collisions. We also compare the performance of the proposed method versus other alternatives.