Harmonious Sampling for Mobile Manipulation Planning

by Mincheul Kang, Donghyuk Kim, and Sung-Eui Yoon
Submitted to IEEE International Conference on Robotics and Automation (ICRA), 2019 (under review)

This figure shows samples generated from our harmonious sampler. A heat map shown in the bottom right shows a sample density in terms of the 2D floor projected from samples in the C-space consisting of the robot base and manipulator. Samples are distributed intensively near the goal configuration and at the narrow passage, resulting in efficient and effective exploration of solution paths.

Abstract

Mobile manipulation planning commonly adopts a decoupled approach that performs planning separately for the base and manipulator, to make each planning space low dimensional. While this approach is fast, it can generate suboptimal paths or cannot find solutions. Another approach, i.e., a coupled approach, is to jointly adjust the base and manipulator in a high dimensional space. This coupled approach addresses sub-optimality and incompleteness of the decoupled approach, but has not been widely used due to its excessive computational overhead. Given this trade-off space, we present a novel manipulation sampling method, harmonious sampling, to perform the coupled approach only in difficult regions, defined as manipulation regions, where we need to simultaneously maneuver the base and manipulator. For the rest regions, defined by base regions, we perform sampling mainly on the base, accelerating the planning process. Our method identifies such manipulation regions based on the reachability map generated from the target end-effector pose and narrow passage identified by generalized Voronoi diagram. We compare our method with the decoupled and coupled approaches in six different problems with different characteristics in the C-space. Our method shows meaningful improvements experimentally in terms of time to find an initial solution (up to 5 times faster) and final solution cost (up to 30% lower) over these techniques, especially in difficult scenes with narrow passage. This result is mainly attributed by our harmonious sampling, adaptively adjusting our sampling process between the coupled and decoupled spaces.

Video

Contents

Paper: PDF (1.91MB)

School of Computing
KAIST
Daehak-ro 291, Guseong-dong, Yuseong-gu, Daejeon, 305-701
South Korea
mincheul.kang (at) kaist.ac.kr