Equi-RO: A 4D mmWave Radar Odometry via Equivariant Networks

2025-09-24

AI Summary

To address the degradation of LiDAR and camera performance under GPS-denied conditions in extreme weather, this paper proposes a 4D millimeter-wave radar odometry algorithm based on equivariant graph neural networks. Methodologically, we introduce Doppler velocity as both node and edge features in a graph structure for the first time; employ an equivariant-invariant feature disentanglement mechanism to enhance geometric consistency modeling of sparse radar signals; and jointly exploit range, azimuth, elevation, and velocity—four dimensions of information—for robust inter-frame registration. Experiments on both public and self-collected datasets demonstrate that our method achieves 10.7% and 20.0% improvements over state-of-the-art baselines in translation and rotation estimation accuracy, respectively. The proposed approach significantly enhances the robustness and precision of pose estimation under all-weather and low-texture conditions.

Technology Category

• Intelligent Robots: State Estimation
• Machine Learning: Graph-based Machine Learning
• Computer Vision: Motion & Tracking

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

Autonomous vehicles and robots rely on accurate odometry estimation in GPS-denied environments. While LiDARs and cameras struggle under extreme weather, 4D mmWave radar emerges as a robust alternative with all-weather operability and velocity measurement. In this paper, we introduce Equi-RO, an equivariant network-based framework for 4D radar odometry. Our algorithm pre-processes Doppler velocity into invariant node and edge features in the graph, and employs separate networks for equivariant and invariant feature processing. A graph-based architecture enhances feature aggregation in sparse radar data, improving inter-frame correspondence. Experiments on the open-source dataset and self-collected dataset show Equi-RO outperforms state-of-the-art algorithms in accuracy and robustness. Overall, our method achieves 10.7% and 20.0% relative improvements in translation and rotation accuracy, respectively, compared to the best baseline on the open-source dataset.

Problem

• Estimating accurate odometry for autonomous systems in GPS-denied environments
• Overcoming sensor limitations of LiDARs and cameras in extreme weather conditions
• Improving feature aggregation and correspondence in sparse 4D radar data

Innovation

• Uses equivariant networks for radar odometry
• Processes Doppler velocity into invariant features
• Employs graph-based architecture for sparse data