UniMSF: A Unified Multi-Sensor Fusion Framework for Intelligent Transportation System Global Localization

TL;DR

UniMSF presents a scalable, factor-graph–based framework for robust ITS localization by unifying GNSS with diverse modalities such as IMU and 4D-radar. It introduces a modular triad: a front-end that builds modality-specific measurements, an outlier detection and online noise-estimation module, and a back-end that optimizes a sliding window of states with multiple factors, including GNSS pseudorange/TDCP, IMU, and 4D-radar velocity. The Radar-UniMSF case study demonstrates tight GNSS–4D-radar–IMU–TDCP integration, achieving significant accuracy gains in real-vehicle tests under challenging GNSS conditions, and showing the value of online pseudorange noise adaptation. The work advances ITS localization by delivering a plug-and-play, robust fusion framework with demonstrated improvements in accuracy and reliability, while outlining future directions such as incorporating deep learning for dynamic fusion weighting.

Abstract

Intelligent transportation systems (ITS) localization is of significant importance as it provides fundamental position and orientation for autonomous operations like intelligent vehicles. Integrating diverse and complementary sensors such as global navigation satellite system (GNSS) and 4D-radar can provide scalable and reliable global localization. Nevertheless, multi-sensor fusion encounters challenges including heterogeneity and time-varying uncertainty in measurements. Consequently, developing a reliable and unified multi-sensor framework remains challenging. In this paper, we introduce UniMSF, a comprehensive multi-sensor fusion localization framework for ITS, utilizing factor graphs. By integrating a multi-sensor fusion front-end, alongside outlier detection\&noise model estimation, and a factor graph optimization back-end, this framework accomplishes efficient fusion and ensures accurate localization for ITS. Specifically, in the multi-sensor fusion front-end module, we tackle the measurement heterogeneity among different modality sensors and establish effective measurement models. Reliable outlier detection and data-driven online noise estimation methods ensure that back-end optimization is immune to interference from outlier measurements. In addition, integrating multi-sensor observations via factor graph optimization offers the advantage of \enquote{plug and play}. Notably, our framework features high modularity and is seamlessly adapted to various sensor configurations. We demonstrate the effectiveness of the proposed framework through real vehicle tests by tightly integrating GNSS pseudorange and carrier phase information with IMU, and 4D-radar.

Figures (9)

• Figure 1: Overall structure of the proposed UniMSF, comprising three main modules: multi-sensor fusion front-end, outlier detection&noise model estimation, and factor graph optimization back-end.
• Figure 2: Illustration of factor graph for the proposed Radar-UniMSF. a) The construction process of TDCP measurements, using an example of 2 satellites observed over 4 continuous epochs. The solid lines represent the TDCPs between adjacent epochs that are utilized, while dashed lines indicate those across multiple epochs, which are optional. b) The factor graph framework of Radar-UniMSF.
• Figure 3: Experimental platform for data collection.
• Figure 4: Trajectory and experimental scenarios. (a) Tall building and brief sky occlusion region; (b) Long duration sky occlusion region.
• Figure 5: Trajectories of the three methods: IPT (yellow), Radar-VMSF (green) and Radar-UniMSF (red). Blue line indicates GT. The local trajectory in scenes (a) and (b) are also zoomed in.