Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-LiCa: A Motion and Targetless Multi LiDAR-to-LiDAR Calibration Framework

Dominik Kulmer, Ilir Tahiraj, Andrii Chumak, Markus Lienkamp

TL;DR

The paper addresses calibrating multiple LiDARs on a vehicle when FOV overlap is partial or absent, without relying on targets or external sensors. It introduces Multi-LiCa, a two-stage pipeline with coarse feature-based alignment (FPFH on voxelized clouds) and TEASER++ initialization, followed by fine GICP-based registration with a cascade merging strategy, plus LiDAR-to-ground calibration. Through evaluation on EDGAR and HeLiPR datasets, Multi-LiCa demonstrates robust accuracy, favorable runtimes, and superior performance in challenging configurations compared to a leading method CROON. The work enables practical, scalable multi-LiDAR fusion by eliminating manual targets and initial pose guesses, and is available in ROS 2 with open-source code. It also outlines future enhancements such as pose-graph optimization and broader data validation to further improve generalizability.

Abstract

Today's autonomous vehicles rely on a multitude of sensors to perceive their environment. To improve the perception or create redundancy, the sensor's alignment relative to each other must be known. With Multi-LiCa, we present a novel approach for the alignment, e.g. calibration. We present an automatic motion- and targetless approach for the extrinsic multi LiDAR-to-LiDAR calibration without the need for additional sensor modalities or an initial transformation input. We propose a two-step process with feature-based matching for the coarse alignment and a GICP-based fine registration in combination with a cost-based matching strategy. Our approach can be applied to any number of sensors and positions if there is a partial overlap between the field of view of single sensors. We show that our pipeline is better generalized to different sensor setups and scenarios and is on par or better in calibration accuracy than existing approaches. The presented framework is integrated in ROS 2 but can also be used as a standalone application. To build upon our work, our source code is available at: https://github.com/TUMFTM/Multi_LiCa.

Multi-LiCa: A Motion and Targetless Multi LiDAR-to-LiDAR Calibration Framework

TL;DR

The paper addresses calibrating multiple LiDARs on a vehicle when FOV overlap is partial or absent, without relying on targets or external sensors. It introduces Multi-LiCa, a two-stage pipeline with coarse feature-based alignment (FPFH on voxelized clouds) and TEASER++ initialization, followed by fine GICP-based registration with a cascade merging strategy, plus LiDAR-to-ground calibration. Through evaluation on EDGAR and HeLiPR datasets, Multi-LiCa demonstrates robust accuracy, favorable runtimes, and superior performance in challenging configurations compared to a leading method CROON. The work enables practical, scalable multi-LiDAR fusion by eliminating manual targets and initial pose guesses, and is available in ROS 2 with open-source code. It also outlines future enhancements such as pose-graph optimization and broader data validation to further improve generalizability.

Abstract

Today's autonomous vehicles rely on a multitude of sensors to perceive their environment. To improve the perception or create redundancy, the sensor's alignment relative to each other must be known. With Multi-LiCa, we present a novel approach for the alignment, e.g. calibration. We present an automatic motion- and targetless approach for the extrinsic multi LiDAR-to-LiDAR calibration without the need for additional sensor modalities or an initial transformation input. We propose a two-step process with feature-based matching for the coarse alignment and a GICP-based fine registration in combination with a cost-based matching strategy. Our approach can be applied to any number of sensors and positions if there is a partial overlap between the field of view of single sensors. We show that our pipeline is better generalized to different sensor setups and scenarios and is on par or better in calibration accuracy than existing approaches. The presented framework is integrated in ROS 2 but can also be used as a standalone application. To build upon our work, our source code is available at: https://github.com/TUMFTM/Multi_LiCa.
Paper Structure (15 sections, 1 equation, 9 figures, 1 algorithm)

This paper contains 15 sections, 1 equation, 9 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Evaluation of Existing Methods
  4. Methodology
  5. Coarse Alignment
  6. Fine Registration
  7. LiDAR-to-Ground Calibration
  8. Parameter Optimization
  9. Dataset Preparation
  10. Evaluation Metrics
  11. Results
  12. HeLiPR Dataset
  13. EDGAR Dataset
  14. Discussion
  15. Conclusion and Future Work

Figures (9)

  • Figure 1: EDGAR LiDAR setup from the front (top image) and back (bottom image) with a) mid-range Ouster OS-1 and b) long-range Seyond Falcon.
  • Figure 2: FOV of EDGAR LiDAR sensors.
  • Figure 3: EDGAR front (blue), left (green), and right (orange) LiDAR calibration result with OpenCalibs CROON.
  • Figure 4: Schematic representation of LiDAR sensors FOV and their overlap.
  • Figure 5: Samples of the HeLiPR dataset with LiDAR point clouds of Ouster (blue), Velodyne (black), Livox (orange) and Aeva (green).
  • ...and 4 more figures