Abstract:
This thesis is concerned with autonomous exploration with single and multirobot systems. In particular, the robots are assumed to be endowed with three-dimensional laser sensors. The exploration strategies are based on bubble space representation that has been previously proposed to represent nodes in topological maps. First, the exploration of an environment by a single robot is considered. There are two aspects to this problem: terrain mapping and determining where to go. Terrain mapping aims to infer the environmental surface shape - as this certainly would a ect the robot in determining where to go. For this, a novel approach based on bubble space representation is proposed and experimentally evaluated. For explorative navigation, the movement direction should be such that it should point the robot to unexplored territory while being accessible. A novel approach is proposed where the generation and recognition of nodes and their associated edges are achieved simultaneously with graph exploration in a topological map based on bubble space. The validity of these approaches are demonstrated by simulations and real-time experimental results.Next, the explorative navigation strategy is extended to multirobot exploration. In this case, the robots are assumed to be communicating with each other and determine their movement directions using the bubble surface information as well as their relative position information. Experimental results with real data show that the robots are able to explore unknown territories without much overlapping.
