Towards Minimally-Intrusive Navigation in Densely-Populated Pedestrian Flow

Mobile robot navigation has achieved great success in last decades. However, traditional methods usually aim at minimizing trajectory length or time, the planned trajectories may hinder pedestrians from walking. To solve this problem, we propose two penalty terms called flow disturbance penalty (FDP) and individual disturbance penalty (IDP) which describe the disturbance of the vehicle in macro and micro levels, respectively. In order to conduct minimally-intrusive navigation, in the micro level navigation, we adopt sampling-based method to minimize not only distance to goal but also IDP. In the macro level navigation, A*-based algorithm has been adopted to minimize FDP while achieving the optimal trajectory length. In addition, we build a triangulation map which gives an abstract representation of the environment to improve real-time performance. Extensive experiments have been carried out to compare our method with the state-of-the-art methods. The results have validated that the proposed method not only can provide minimal disturbance, but also has better real-time performance.