Urban navigation faces persistent challenges: signal obstruction, multipath, and non-line-of-sight (NLOS) reception severely degrade Global Navigation Satellite System (GNSS) performance. Multi-sensor fusion—particularly GNSS/Inertial Navigation System (INS) integration—combines short-term inertial accuracy with long-term GNSS stability, while visual measurements add motion constraints that suppress inertial drift. However, reliable cycle-slip detection and rapid ambiguity convergence remain difficult. Pseudorange-aided combinations are vulnerable to code noise and multipath; missing frequencies prevent valid multi-frequency combinations; and frequent interruptions break the continuity required for between-epoch differencing. Moreover, ambiguity information converged before blockage is often discarded when signals return. Because of these challenges, in-depth research into cross-epoch ambiguity preservation and rectification is urgently needed.
Researchers from Wuhan University, China, published (DOI: 10.1186/s43020-026-00227-7) the study in Satellite Navigation on 09 November 2026. The team proposed a Cross-Epoch Ambiguity Tracking (CEAT) framework within a Multi-State Constraint Kalman Filter (MSCKF)-based GNSS/INS/Vision tightly coupled system, preserving converged ambiguities across discontinuous carrier-phase arcs for cycle-slip detection and ambiguity rectification. The study evaluated the method through simulated cycle-slip tests and real-world urban and campus datasets, comparing its performance against representative cycle-slip detection methods.
The CEAT workflow manages ambiguities through a globally updated ambiguity pool, handling four tracking cases: new tracking, continuous tracking, Cycle Slip Rectification (CSR) tracking, and failure tracking. When new GNSS observations arrive, pseudorange and carrier-phase residuals are constructed using stored ambiguities and propagated poses. Median-centered residual analysis mitigates receiver-clock and propagated-position errors; the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm separates cycle-slip-free observations from affected ones; and weighted least-squares estimation, followed by LAMBDA-based integer cycle-slip estimation with ratio-test validation, rectifies affected ambiguities. A pseudorange-aided recovery strategy provides reasonable initialization when reliable rectification fails. Experiments on urban and campus trajectories used a vehicle-borne platform with a Septentrio mosaic-X5 mini receiver, ADIS-16470 Micro-Electro-Mechanical Systems (MEMS) IMU, and FLIR stereo cameras, with ground truth from a tactical-grade SPAN-ISA-100C IMU. CEAT achieved 3D root mean square errors (RMSEs) of 0.26 m and 0.20 m for fixed solutions on the urban and campus trajectories, respectively, with corresponding 3D positioning availabilities of 90.0% and 96.7%, outperforming three representative cycle-slip detection methods (HMW-GF, GB-WL, and EDM). Ablation studies confirmed the necessity of the ambiguity pool, CSR, residual clustering, weighted least-squares refinement, and visual-inertial priors.
The authors said the key insight is that ambiguity information converged before signal blockage should not be discarded. "By preserving and reusing these ambiguities across discontinuous carrier-phase arcs, CEAT enables rapid cycle-slip detection and ambiguity rectification after signal reacquisition," they said. They noted that the method maintains ambiguity continuity and improves float-ambiguity precision, as reflected in lower Ambiguity Dilution of Precision (ADOP) values. "Accurate visual-inertial priors are essential for constructing reliable carrier-phase residuals," they added, emphasizing that this prior information supports both cycle-slip detection and ambiguity rectification in GNSS-challenged environments.
CEAT offers practical value for autonomous driving, intelligent transportation, and location-based services in dense urban areas where GNSS signals are frequently obstructed. Its compatibility with Factor Graph Optimization (FGO) frameworks—demonstrated by integrating CEAT into FGO-SAW and FGO-TAR—suggests broad applicability across estimator architectures. The method reduces repeated ambiguity reinitialization, a persistent bottleneck in high-precision GNSS/INS/Vision integration. Future work will explore deep-learning-based methods to improve predicted-position accuracy, enhancing residual clustering robustness and pseudorange-aided recovery. The authors acknowledge limitations: performance degrades when a large proportion of carrier-phase observations are simultaneously affected by cycle slips, and prolonged outages cause visual-inertial error accumulation.
###
References
DOI
10.1186/s43020-026-00227-7
Original Source URL
https://doi.org/10.1186/s43020-026-00227-7
Funding information
National Natural Science Foundation of China (Grant 42374016); Key Program of the National Natural Science Foundation of China (Grant 42330105); Major Program (JD) of Hubei Province (2023BAA025); Open Fund of Hubei Luojia Laboratory (Grant No. 260100023)
About Satellite Navigation
Satellite Navigation (ISSN: 2662-1363; ISSN: 2662-9291) Satellite Navigation is the official journal of the Aerospace Information Research Institute. The journal aims to report innovative ideas, new results, and progress in the theories, techniques, and applications of satellite navigation. The journal welcomes original articles, reviews and commentaries.