Path planning for grasping operations using an adaptive PCA-based sampling method

The planning of collision-free paths for a hand-arm robotic system is a difficult issue due to the large number of degrees of freedom involved and the cluttered environment usually encountered near grasping configurations. To cope with this problem, this paper presents a novel importance sampling method based on the use of Principal Component Analysis (PCA) to enlarge the probability of finding collision-free samples in these difficult regions of the Configuration Space with low clearance. By using collision-free samples near the goal, PCA is periodically applied in order to obtain a sampling volume near the goal that better covers the free space, improving the efficiency of sampling-based path planning methods. The approach has been tested with success on a hand-arm robotic system composed of a four-finger anthropomorphic mechanical hand (17 joints with 13 independent degrees of freedom) and an industrial robot (6 independent degrees of freedom).