Multi-Robot Localization and Mapping Strategy: Utilizing Behavior Based Dynamic Tree Structure and Observer-explorer Routine

In this paper, we propose a simultaneous localization and map-building (SLAM) strategy to explore unknown environment. Multiple robots are deployed in unknown area and required to localize each other accurately while exploring. Instead of relying on active obstacle detector, we propose a method of exploration which uses only tactile sensors and inter-robot distance measurements. To avoid dead-reckoning and odometry errors, an observer-explorer based routine is adopted. A dynamic spanning tree structure is implemented for multi-robot coordination. Parent (observer) nodes monitor the distances of their children (explorer) nodes. In order to promote completeness of the map, the reconfigurable spanning tree structure favors unexplored area implicitly. An online behavior-based finite state machine drives the configuration of the structure. Our exploration scheme allows the differentiation between robots and obstacles or boundaries. When an obstacle or boundary is detected, it is recorded on a map that is shared with all other robots. The proposed strategy is simulated and results are presented.