GNSS Array-Based Multipath Detection Employing UKF on Manifolds

TL;DR

The paper tackles urban GNSS multipath by introducing a GNSS array-based multipath detector integrated with real-time attitude estimation via a UKF on manifolds. A RANSAC-inspired strategy reduces the combinatorial load of evaluating satellite subsets, enabling real-time discrimination of multipath-affected signals. The approach fuses GNSS with IMU data to refine both positioning and attitude while excluding compromised satellites, with performance demonstrated on KITTI-derived scenarios and static benchmarks. Results show substantial improvements in position and attitude accuracy, especially when many satellites are compromised, highlighting potential for robust urban navigation in safety-critical applications.

Abstract

Global Navigation Satellite Systems (GNSS) applications are often hindered by various sources of error, with multipath interference being one of the most challenging, particularly in urban environments. In this work, we build on previous research by implementing a GNSS array-based multipath detection algorithm, incorporating real-time attitude estimation for dynamic scenarios. The method fuses GNSS and IMU data using an Unscented Kalman Filter (UKF) on a manifold, enabling continuous attitude tracking. The proposed approach utilizes attitude information from satellite combinations to identify and exclude multipath-affected satellites, improving the accuracy of both positioning and attitude determination. To address computational challenges associated with evaluating large numbers of satellite combinations, we propose the use of the Random Sample Consensus (RANSAC) algorithm, which reduces the number of combinations assessed while maintaining high detection performance. Performance evaluations are conducted using trajectories and IMU readings from the KITTI dataset. GNSS observations are simulated based on ground truth positions and satellite ephemeris. The results demonstrate the effectiveness of the proposed approach in detecting satellites affected by multipath interference. Significant improvements in positioning accuracy are observed, particularly in scenarios where a large portion of the visible satellites are contaminated by severe multipath.

Figures (7)

• Figure 1: Signal Reception Scenarios for GNSS Receivers in Urban Environments
• Figure 2: The proposed antenna array configuration with 5 elements. Inter-element distances and baseline direction vectors are indicated.
• Figure 3: GNSS Array-IMU Fusion Framework for Multipath Detection and Positioning
• Figure 4: Classification performance comparison between DBSCAN (red) and proposed method (blue) across phase noise levels.
• Figure 5: Baseline MAE comparison. Proposed method outperforms DBSCAN in attitude estimation accuracy.