Co-RaL: Complementary Radar-Leg Odometry with 4-DoF Optimization and Rolling Contact

TL;DR

This work tackles robust ego-motion estimation for legged robots operating in challenging environments by fusing mmWave radar with leg odometry. It introduces a $4$-DoF radar velocity factor with decoupled RANSAC to significantly reduce vertical drift and a rolling-contact-aware preintegrated leg odometry factor to mitigate contact-induced biases. The proposed radar-leg odometry demonstrates superior z-axis accuracy and resilience to vision failures on real-world datasets, including stairs and uneven terrain, and is validated against multiple baselines. The authors also release a comprehensive dataset to support further research in radar-leg SLAM for robotics.

Abstract

Robust and accurate localization in challenging environments is becoming crucial for SLAM. In this paper, we propose a unique sensor configuration for precise and robust odometry by integrating chip radar and a legged robot. Specifically, we introduce a tightly coupled radar-leg odometry algorithm for complementary drift correction. Adopting the 4-DoF optimization and decoupled RANSAC to mmWave chip radar significantly enhances radar odometry beyond the existing method, especially z-directional even when using a single radar. For the leg odometry, we employ rolling contact modeling-aided forward kinematics, accommodating scenarios with the potential possibility of contact drift and radar failure. We evaluate our method by comparing it with other chip radar odometry algorithms using real-world datasets with diverse environments while the datasets will be released for the robotics community. https://github.com/SangwooJung98/Co-RaL-Dataset

Figures (7)

• Figure 1: Example of radar drift scene that includes large dynamic objects and leg contact drift scene including frequent contact impact and vibration. Large dynamic objects are shown in the yellow box. Both scenes are recorded using the intel realsense D435i attached to the SPOT sensor system while acquiring Stair sequence. Odometry of the proposed method (Ours) and comparison group are included in top-down and side views.
• Figure 2: Pipeline of the proposed system with 4-DoF radar velocity factor and rolling contact preintegrated leg odometry factor.
• Figure 3: Overview of the rolling contact on legged robots with round feet. Rolling motion on the contact frame occurs as the robot moves.
• Figure 4: System configuration. LiDAR and camera are attached only for baseline trajectory generation and video recording.
• Figure 5: Trajectory result of each sequence's top-down and side views. Color matching of methods is the same for every graph. The red, green, and blue dotted boxes and their enlarged figures are for \ref{['study1']}, \ref{['study2']}, and \ref{['study3']}, respectively.