Sensor-based Online Planning of Time-optimized Paths in Dynamic Environments

Dynamic environments, in which robots and for example humans share a common workspace impose a variety of requirements on path planning algorithms, including real-time capabilities and collision tests based on sensor input. We propose a randomized-roadmap-based path planning algorithm that limits the number of collision tests and distance calculations to a volume achievable in realtime, while still being able to achieve high path clearance and statistical completeness given an unlimited number of planning cycles. It does so by exploiting the knowledge about static portions of the environment using a static, collisionchecked roadmap and by interleaving planning and execution. Image-based distance measurements are induced by the graph search algorithm and interpolated to allow planning of time-optimized paths on the roadmap with a reduced number of distance measurements.