Knowledge-base topological exploration for mobile robots

This paper introduces novel graph exploration based on reasoning procedure. The proposed approach does not need any markers and only a similarity measures of the places and routes are used for decisions and loop closing. The edge selection process is driven by information gain, which is computed for each edge based on the probability of the loop closure and consecutive merges performed if the loop closure take place. The edge with high probability of closing the loop and high number of edge consecutively merged and therefore eliminated from the map is preferred. A loop-closing procedure used in the exploration algorithm utilizes the information about the environment structure. A priori knowledge of the environment properties is incorporated into the reasoning procedure as logic rules. Proposed exploration algorithm were experimentally verified in simulator and with real robot in indoor environment.