Navigation strategy and path planning for autonomous road crossing by outdoor mobile robots

This paper describes the navigation strategy and path planning methods for autonomous road crossing by outdoor mobile robots in urban environments. Road-crossing is the part of outdoor robot navigation when a robot, while traveling along pedestrian sidewalks, approaches an intersection or a road crossing and needs to autonomously cross to reach the intended destination. In this work, the robot first autonomously travels along pedestrian sidewalks. For this, instead of using a pre-supplied navigational map, we endow the robot with the level of autonomy required such that it can spontaneously detect sidewalks and perceive a trajectory to navigate. While traveling along this trajectory, the robot performs real time detection of pedestrian push-button boxes. If a button box is detected, the robot deviates from the original trajectory and autonomously navigates to it. Next, using the detected button box as a landmark, the robot approaches the pedestrian crossing. It finally performs vision-based detection of the zebra crossings and generates the trajectory required to cross the road at a green signal. We demonstrate these methods are feasible by performing experiments with a custom built outdoor mobile robot using experimental conditions based on an actual pedestrian crossing in our university campus.