SIFT-Like Pose Tracking with LIDAR using Zero Odometry
Lists
Only you see this
Abstract
Navigating an unknown environment is a challenging task for a robot, especially when a large number of obstacles exist and the odometry lacks reliability. Pose tracking allows the robot to determine its location relative to its previous location. The ICP (iterative closest point) has been a powerful method for matching two point clouds and determining the transformation matrix between the maps. However, in a situation where odometry is not available and the robot moves far from its original location, the ICP fails to calculate the exact displacement. In this paper, we suggest a method that is able to match two different point clouds taken a long distance apart. Without using any odometry information, it only exploits the features of corner points containing information on the surroundings. The algorithm is fast enough to run in real time.
BibTeX
@article{kim2016sift,
author = {Jee-Soo Kim and Nojun Kwak},
title = {{SIFT-Like Pose Tracking with LIDAR using Zero Odometry}},
journal = {Journal of Institute of Control Robotics and Systems},
year = {2016},
doi = {10.5302/j.icros.2016.16.0149},
}Loading related content...