In order for robots to operate safely in human environments, they need to be able to react to a dynamically changing environment in real-time. This involves both fast processing of sensor data and fast motion planning based on the sensor data.

This work describes the implementation of a path planner for a 7-DOF robot arm, which is able to react to a dynamic environment very quickly. A SwissRanger SR3000 3D time-of- ight sensor is used to detect obstacles. This work uses the Probabilistic Roadmaps for Changing Environments algorithm introduced by Leven and Hutchinson. The algorithm generates a roadmap for an obstacle-free environment during a preprocessing stage. During the online planning stage blocked parts of the roadmap are invalidated and the roadmap is used for fast planning. Fast invalidation of blocked parts of the roadmap is accomplished through a mapping from occupied workspace grid cells to blocked roadmap parts. This workspace mapping is also generated during the preprocessing stage. It is shown that the implementation is able to plan within about 100 milliseconds while avoiding sensed obstacles, and, thus, is able to react to a dynamic environment. The path planner is implemented on the Care-O-bot 3, which is a home assistant robot developed at Fraunhofer IPA.