Real-time navigation for a personal mobility in an environment with pedestrians

This paper describes a navigation system in a dynamic environment for a two-wheeled inverted pendulum mobile robot, PMR. Our system is organized by localization, detection and tracking of pedestrians, and trajectory planner. The localization is robust to effects of moving obstacles and pitching movements of the robot, and the trajectory planner creates a path with a certain smoothness considering movements of pedestrians. In addition, the planner introduces strategies to avoid pedestrians to be friendly to pedestrians around the robot. Besides, our system can run on two laptop PCs in real time. Finally, we show experimental results as well.