Improving Throughput of Mobile Robots in Narrow Aisles

Emergency brakes applied by mobile robots to avoid collision with humans often block the traffic in narrow hallways. The ability to smoothly navigate in such environments can enable the deployment of robots in shared spaces with humans such as hospitals, cafeterias and so on. The standard navigation stacks used by these robots only use spatial information of the environment while planning its motion. In this work, we propose a predictive approach for handling dynamic objects such as humans. The use of this temporal information enables a mobile robot to predict collisions early enough and avoid the use of emergency brakes. We validated our approach in a real-world set-up at a busy university hallway. Our experiments show that the proposed approach results in fewer stops compared to the standard navigation stack only using spatial information.