DNOI-4DRO: Deep 4D Radar Odometry with Differentiable Neural-Optimization Iterations

2025-05-18

AI Summary

To address the challenge of high-precision odometry estimation from 4D radar point clouds, this paper proposes NeuRO-LOAM—a neural-optimization joint framework. Methodologically, it introduces a differentiable neural-optimization iterative operator that jointly embeds deep neural network–predicted motion flow and Gauss–Newton geometric optimization into end-to-end training; it also designs a dual-stream 4D radar backbone network that simultaneously encodes multi-scale geometric features and clustering-driven, class-aware features. Its key innovation lies in the first realization of differentiable, co-iterative coupling between motion flow prediction and pose optimization. Evaluated on the VoD and Snail-Radar datasets, NeuRO-LOAM significantly outperforms classical and state-of-the-art learning-based radar odometry methods. Notably, without map-based optimization, its performance matches that of LiDAR-based A-LOAM, demonstrating the feasibility of achieving high-accuracy, robust localization using 4D radar in complex environments.

Technology Category

• Intelligent Robots: State Estimation
• Search and Optimization: Learning to Search
• Computer Vision: Learning & Optimization for CV

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

A novel learning-optimization-combined 4D radar odometry model, named DNOI-4DRO, is proposed in this paper. The proposed model seamlessly integrates traditional geometric optimization with end-to-end neural network training, leveraging an innovative differentiable neural-optimization iteration operator. In this framework, point-wise motion flow is first estimated using a neural network, followed by the construction of a cost function based on the relationship between point motion and pose in 3D space. The radar pose is then refined using Gauss-Newton updates. Additionally, we design a dual-stream 4D radar backbone that integrates multi-scale geometric features and clustering-based class-aware features to enhance the representation of sparse 4D radar point clouds. Extensive experiments on the VoD and Snail-Radar datasets demonstrate the superior performance of our model, which outperforms recent classical and learning-based approaches. Notably, our method even achieves results comparable to A-LOAM with mapping optimization using LiDAR point clouds as input. Our models and code will be publicly released.

Problem

• Develops a deep learning-optimization hybrid 4D radar odometry model
• Enhances sparse 4D radar point cloud representation via dual-stream backbone
• Outperforms classical and learning-based methods in radar pose estimation

Innovation

• Combines neural networks with geometric optimization
• Uses differentiable neural-optimization iteration operator
• Integrates multi-scale and class-aware radar features