This work considers the problem of coverage of an initially unknown environment by a set of autonomous robots. A crucial aspect in multi-robot coverage involves robots sharing information about the regions they have already covered at certain intervals, so that multiple robots can avoid repeated coverage of the same area. However, sharing the coverage information between robots imposes considerable communication and computation overhead on each robot, which increases the robots' battery usage and overall coverage time. To address this problem, we explore a novel coverage technique where robots use an information compression algorithm before sharing their coverage maps with each other. Specifically, we use a polygonal approximation algorithm to represent any arbitrary region covered by a robot as a polygon with a fixed, small number of vertices. At certain intervals, each robot then sends this small set of vertices to other robots in its communication range as its covered area, and each receiving robot records this information in a local map of covered regions so that it can avoid repeat coverage. The coverage information in the map is then utilized by a technique called spanning tree coverage (STC) by each robot to perform area coverage. We have verified the performance of our algorithm on simulated Coroware Corobot robots within the Webots robot simulator with different sizes of environments and different types of obstacles in the environments, while modelling sensor noise from the robots' sensors. Our results show that using the polygonal compression technique is an effective way to considerably reduce data transfer between robots in a multi-robot team without sacrificing the performance and efficiency gains that communication provides to such a system.
展开▼
