| Distributed robot teams have become ubiquitous. Robots are less expensive, more powerful, and easier to program. When distributed robot teams are used for surveillance, search and rescue missions, the distance between humans and danger can be increased. Robot-assisted search in an unknown environment is a critical component in all of these missions. There is little question that using a team of robots can improve the efficiency of a search task over using a single robot. However, design choices multiply when organizing a multirobot team to achieve a balance between flexibility, cost and coverage. Recently, proposed distributed coverage algorithms either share a global world representation that is used to determine subsequent search targets or coordinate explicitly via communicating specific task information. The resulting algorithms either rely too heavily on less than accurate information from team members or limit team size due to large communications overhead.; In this thesis, the repertoire of available search algorithms is expanded to include novel, biologically inspired approaches to coverage and coordination. Although many approaches implement footprint coverage (e.g., mine-sweeping), algorithmic changes expand search to use any arbitrary search sensor. Robots looking for chemicals or animals, can cover an unknown environment using a chemical or heat sensor with on-line planning. In addition, coordination using local search queues, rather than globally shared search queues is shown to improve search in terms of interference, communications loads and sensitivity to localization errors, three key factors in multirobot search. Both simulations and experiments are presented to support the efficacy of the novel approaches. |
