Skip to content

Latest commit

 

History

History
19 lines (13 loc) · 2.49 KB

README.md

File metadata and controls

19 lines (13 loc) · 2.49 KB

srf_laser_odometry

Robust Planar Odometry Based on Symmetric Range Flow and Multi-Scan Alignment

❗ This repo is a fork. It only contains an mrpt-free SRF library meant to be consumed through the laser_odometry package (see laser_odometry_srf). :heavy_exclamation_mark:

This pkgs offers a fast and reliable estimation of 2D odometry based on planar laser scans.

SRF is the continuation of RF2O, a fast and precise method to estimate the planar motion of a lidar from consecutive range scans. SRF presents a dense method for estimating planar motion with a laser scanner. Starting from a symmetric representation of geometric consistency between scans, we derive a precise range flow constraint and express the motion of the scan observations as a function of the rigid motion of the scanner. In contrast to existing techniques, which align the incoming scan with either the previous one or the last selected keyscan, we propose a combined and efficient formulation to jointly align all these three scans at every iteration. This new formulation preserves the advantages of keyscan-based strategies but is more robust against suboptimal selection of keyscans and the presence of moving objects.

An extensive evaluation of this method is presented with simulated and real data in both static and dynamic environments Journal Article. Results show that our approach is one order of magnitude faster and significantly more accurate than existing methods in all the conducted experiments. With a runtime of about one~millisecond, it is suitable for those robotic applications that require planar odometry with low computational cost.

For a full description of the algorithm, please refer to:

  • M. Jaimez, J. Monroy, M. Lopez-Antequera, J. Gonzalez-Jimenez, Robust Planar Odometry based on Symmetric Range Flow and Multi-Scan Alignment. IEEE Transactions on Robotics, pp. 1623--1635, 2018 SRF paper

Dependencies

The code provided in this repo depends on the famous Mobile Robot Programming Toolkit MRPT. To avoid problems with versions, we used the defatult binary version (1.3.2-1) available in the official Ubuntu repository (sudo apt-get install libmrpt-dev mrpt-apps), and the ROS-pkg mrpt_bridge to convert data types between MRPT and ROS.