GNSS Anomaly-Tolerant Localization of Land Vehicles With ESKF and Residual-Based Detection–Identification–Adaptation

Anomaly-tolerant localization remains a critical challenge for autonomous vehicles, where reliable state estimation must be maintained despite global navigation satellite system (GNSS) degradation. This paper presents a robust global navigation satellite system/inertial measurement unit (IMU) fusion framework that integrates an error-state Kalman filter (ESKF) with a residual based detection–identification–adaptation (DIA) mechanism. Operating directly on GNSS observation dimensions, the framework selectively rejects or corrects inconsistent measurements using residual statistics and temporal consistency. The ESKF ensures stable propagation of position, velocity, attitude, and sensor bias, while the DIA module performs real-time residual evaluation, adaptive thresholding, and subset-based fusion to mitigate degraded GNSS inputs. The proposed method was evaluated through high-fidelity simulations and real-world road tests. Under typical GNSS signal degradation scenarios, including multipath propagation, partial obstruction, and sudden jumps, ESKF-DIA consistently delivers the highest trajectory accuracy and anomaly tolerance. Compared to the baseline ESKF method, the proposed method reduces the root mean square error by 20% to 35%, maintaining stable performance compared to other novel methods even with performance degradation. These results demonstrate that the framework provides a robust and practical solution for autonomous localization in complex environments with limited GNSS signals.