Motion planning for manipulators in dynamically changing environments using real-time mapping of free workspace

This paper introduces a method to efficiently compute global motion plans for robotic manipulators in dynamically changing environments. An offline computation step is used to construct a sparse roadmap to approximate the configuration space of the manipulator in an empty environment. When the robot is running, a representation of the environment to keep track of the robot's free workspace is maintained as sensor updates are received. The maintained representation of the free workspace is used in conjunction with the data computed offline to quickly compute good quality global motion plans.