Robot Motion Planning in Dynamic Environments

This paper presents a method for computing the motions of a robot in dynamic environments, subject to the robot dynamics and its actuator constraints. This method is based on the concept of Velocity Obstacle, which defines the set of feasible robot velocities that would result in a collision between the robot and an obstacle moving at a given velocity. The avoidance maneuver at a specific time is thus computed by selecting robot's velocities out of that set. A trajectory consisting of a sequence of avoidance maneuvers at discrete time intervals is generated by a search of a tree of avoidance maneuvers. An exhaustive search computes near minimum-time trajectories, whereas a heuristic search generates feasible trajectories for on-line applications. These trajectories are compared to the optimal trajectory computed by a dynamic optimization that minimizes motion time, subject to robot dynamics, its actuator limits and the state inequality constraints due to the moving obstacles. This appr..