G-BEAM: Graph-Based Exploration And Mapping for Autonomous Vehicles

A novel solution to the problem of exploration and mapping of an unknown environment by an autonomous vehicle is presented. A hierarchical control system is adopted: a low-level reactive controller manages obstacle avoidance, and two high-level strategies are in charge of mapping and navigation tasks. The decision strategy implemented at the high-level is named G-BEAM, standing for 'Graph-Based Exploration And Mapping It builds a reachability graph used both as a trajectory planning tool and as a map. The reachability graph representation requires less storage resources with respect to a more traditional occupancy-map, and it can be directly exploited to compute the system's path towards a given target or unexplored locations. The latter are ranked according to the expected information gain that is realized when they are visited. Such information gain is then used in the cost function of the navigation strategy, which is based on a receding horizon concept. The controller has been successfully tested in various simulated environments. Comparison with other approaches in state of the art shows promising performance.