Adapting Probabilistic Roadmaps to Handle Uncertain Maps

Abstract— Randomized motion planning techniques are responsible for many of the recent successes in robot control. However, most motion planning algorithms assume perfect and complete knowledge of the environment. These algorithms can fail arbitrarily badly if there are errors in the model of the environment. In contrast, real world robot systems have succeeded by using explicit representations of model uncertainty in localization and mapping to compensate for sensor error. In this paper, we propose an extension of the Probabilistic Roadmap algorithm that allows us to compute motion plans that are robust to uncertain environment models. We show that the adapted PRM generates less collision-prone trajectories with fewer samples than the standard method.

• Type: Conference
• Year: 2006
• Where: ICRA
• Authors: Patrycja E. Missiuro, Nicholas Roy
• Keywords: ICRA 2006, Motion Planning, Motion Planning Algorithms, Motion Planning Techniques, Robotics