Fast local obstacle avoidance under kinematic and dynamic constraints for a mobile robot

This paper presents an efficient approach for reactive collision avoidance taking into account both vehicle dynamics and nonholonomic constraints of a mobile robot. Motion commands are generated by searching the space of actuating variables. Vehicle dynamics are considered by restricting the search space to values which are reachable within the next time step. The final selection among admissible configurations is done by an objective function which trades off speed, goal-directedness and remaining distance until an obstacle is hit when moving along the chosen path. The presented approach differs from previous ones in the selective use of precalculated lookup tables. These are the key to efficiency, and they especially allow the use of any-shaped robot contours. Furthermore, obstacle information from different sources can easily be considered without preprocessing. Extensive experiments on different robots have shown robust operation in dynamic and unprepared indoor environments with speed up to 1 m/s.