Lidar-only Odometry based on Multiple Scan-to-Scan Alignments over a Moving Window

2025-03-27

AI Summary

To address the susceptibility of pose estimation to cumulative drift in lidar-only SLAM, this paper proposes a multi-scan alignment-driven sliding-window pose graph optimization framework. Instead of conventional frame-to-frame sequential registration, our method performs independent ICP alignments among multiple point clouds within a sliding window to construct geometrically consistent pose graph constraints, jointly optimizing current and historical poses; it further enables joint refinement of historical scans without redundant registration. Our key innovation lies in the first deep integration of multi-scan geometric consistency modeling with incremental pose graph optimization—achieving significantly improved long-term robustness while maintaining real-time performance. Evaluated on KITTI, MulRan, and a custom vehicle-mounted dataset, the method achieves state-of-the-art accuracy, markedly suppresses long-term drift, and reduces computational overhead by approximately 30%.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Multi-modal Vision
• Search and Optimization: Learning to Search

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
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Lidar-only odometry considers the pose estimation of a mobile robot based on the accumulation of motion increments extracted from consecutive lidar scans. Many existing approaches to the problem use a scan-to-map registration, which neglects the accumulation of errors within the maintained map due to drift. Other methods use a refinement step that jointly optimizes the local map on a feature basis. We propose a solution that avoids this by using multiple independent scan-to-scan Iterative Closest Points (ICP) registrations to previous scans in order to derive constraints for a pose graph. The optimization of the pose graph then not only yields an accurate estimate for the latest pose, but also enables the refinement of previous scans in the optimization window. By avoiding the need to recompute the scan-to-scan alignments, the computational load is minimized. Extensive evaluation on the public KITTI and MulRan datasets as well as on a custom automotive lidar dataset is carried out. Results show that the proposed approach achieves state-of-the-art estimation accuracy, while alleviating the mentioned issues.

Problem

• Estimates robot pose using multiple scan-to-scan alignments to reduce drift errors
• Avoids map-based registration to minimize computational load and error accumulation
• Achieves high accuracy by optimizing pose graph with previous scan refinements

Innovation

• Uses multiple scan-to-scan ICP registrations
• Optimizes pose graph for accurate estimation
• Minimizes computational load by avoiding recomputations