Autonomous shuttle buses have been available for some years from manufacturers Navya, EasyMile and Local Motors. Despite their limited usability with only six to eight seats and driving at low speeds, these vehicles have been very popular with the general public. Their standard method of navigation relies on a fixed set of waypoints using RTK-GPS for point-to-point movement along a fixed route. Vehicles cannot deviate from their pre-programmed route, so the onboard safety officer has to manually drive the shuttle around any minor obstacle in its way. In this paper, we present an alternative navigation system for autonomous shuttle buses which combines lidar, IMU and GNSS navigation. Our ROS-based system is able to drive to any point on a university campus and will actively generate all drivable routes whenever it encounters stationary or movable obstacles that are not listed in its map. To aid software development, a simulation model of the shuttle bus with all of its sensors and a digital model of the university campus have been implemented in Carla.

The University of Western Australia: Thomas Bräunl, Kai Li Lim, Thomas Drage, Kieran Quirke-Brown, Zhihui Lai, Yuchen Du, Kyle Carvalho