Navigation in GNSS-Denied Environments Using MEMS-Grade Sensors and Geophysical Anomalies: A Particle Filter Approach

James Brodovsky, Philip Dames

Peer Reviewed

Abstract:	This work further explores previous investigations into geophysical-aided navigation using MEMS-grade sensors by implementing a Rao-Blackwellized particle filter (RBPF) architecture for integrating gravity and magnetic anomaly measurements with strapdown inertial navigation. Building on prior work demonstrating feasibility with an Unscented Kalman Filter (UKF), this particle filter approach explicitly handles the non-Gaussian, potentially multimodal posterior distributions that arise from map-matching operations in geophysical navigation. The RBPF implementation partitions the navigation state into nonlinear horizontal position states represented by particles and linear states (altitude, velocity, attitude, and sensor biases) estimated by Kalman filters conditioned on each particle hypothesis. This architecture naturally accommodates the highly nonlinear measurement model relating magnetometer and gravimeter readings to position via interpolation through geophysical anomaly maps. Using the same MEMS-Nav dataset of smart-phone-grade sensor data and open-source geophysical maps (IGPP gravity anomaly and WDMAM magnetic anomaly), the particle filter demonstrates substantially improved performance relative to the UKF approach under systematic GNSS degradation scenarios. Results across 21 trajectories show that the particle filter achieves consistent and dramatic reduction in INS drift using magnetic and gravity anomaly aiding, with all trajectories demonstrating improvement and a mean RMSE reduction of approximately 1,566 km compared to a GNSS-degraded RBPF baseline. Across the 21 trajectories with corresponding GNSS-degraded UKF runs, the mean geophysically-aided RBPF RMSE improvement relative to the UKF baseline ranges from 0.2–1.4 km depending on anomaly modality. The simple Gaussian measurement model employed in this work provides a foundation for future enhancements incorporating physics-informed models from recent airborne magnetic navigation literature. This research confirms that Rao-Blackwell particle filter architectures offer a viable alternative to Kalman filter variants for low-cost geophysical navigation on MEMS-grade platforms.
Published in:	Proceedings of the ION 2026 Pacific PNT Meeting April 13 - 16, 2026 Hilton Waikiki Beach Honolulu, Hawaii
Pages:	399 - 410
Cite this article:	Brodovsky, James, Dames, Philip, "Navigation in GNSS-Denied Environments Using MEMS-Grade Sensors and Geophysical Anomalies: A Particle Filter Approach," Proceedings of the ION 2026 Pacific PNT Meeting, Honolulu, Hawaii, April 2026, pp. 399-410. https://doi.org/10.33012/2026.20618
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In