Exploration Transform: A stable exploring algorithm for robots in rescue environments

Autonomous robots in rescue environments have to fulfill several task at the same time: They have to localize themselves, build maps, detect victims and decide where to go next for further exploration. In this contribution we present an approach that provides a robust solution for the exploration task: Based on the knowledge of the environment that the robot has already acquired, the algorithm calculates a path to the next interesting "frontier". Our comprehensive approach takes into account the distance to the next frontier and the difficulty of the path for the robot. Those difficulties can result from narrow passages but also from wide, open spaces where the sensors cannot detect any landmark. For the native exploration task, the algorithm is fed with occupancy grids. For the search task, it can also process maps that encode additional information, e. g. places that have not been searched by other sensors yet. The Exploration Transform was successfully implemented on our mobile system Robbie and was used during the RoboCup German Open 2007 and the RoboCup World Championship 2007. Using this approach, our Team "resko" achieved the "Best in Class Autonomy Award" in the Rescue Robot League in both competitions.