TIGRE: Topological graph based robotic exploration

Abstract

In this work we address the problem of autonomous robotic exploration and map building supported by high level information provided by online topological segmentation that incrementally generates an undirected connected graph of the environment. We formulate the exploration problem as the traversal of all the edges of the online graph, being the optimal solution an Eulerian path, if it exist. We propose an integrated approach, the TIGRE algorithm, that combines Graph-SLAM procedures with online Constrained Depth-First Search based graph traversal algorithms to efficiently explore the navigation area. Simulation results show that the online TIGRE algorithm provides similar error estimates for the robot poses and distance travelled than well-known off-line algorithms that use the full graph of the environment. Also, it outperforms other online and off-line algorithms in terms of estimated error.