Implicit Cooperation Strategies for Multi-robot Search of Unknown Areas

Although explicit coordination of team search may provide solid performance for small team sizes, it has been shown that such methods do not scale to larger teams due to limited communications bandwidth and computational requirements. In addition, methods that rely upon persistent, reliable network connections may have limited applicability to real-world search problems. In this work, we explore implicit cooperation enabled through sharing of search progress information. Prior research shows cooperation paradigms in which team members share a global task list result in interference and duplication of search when members choose the same search areas. Methods that only use local sensor information to identify search targets require fewer message exchanges and create less interference between robots than existing shared approaches. In addition, search progress and completion are more consistent due to the reduction in interference. Results based on simulations and physical experiments are presented that compare performance in terms of time-to-cover, consistency, and interference.