Skip to content
Open access

Adaptive Hybrid EKF-LSTM Framework with Magnitude-Bounded Fallback for Smartphone GNSS/IMU Sensor Fusion

Sep 2026 · Symposium on Intelligent Systems and Informatics · pp. 189-194 · 0 citations · 16 references

Abstract

Mobile positioning systems in urban environments face accuracy challenges due to non-linear sensor noise, multipath distortions, and signal loss. Traditional sensor fusion methods, such as the Extended Kalman Filter (EKF), rely on static parameters and struggle to adapt to dynamic environmental variations. This paper presents a hybrid, AI-augmented positioning framework that enhances classical state-space estimation with a neural time-series model. The proposed system integrates GNSS and IMU data within a loosely-coupled architecture. A Long Short-Term Memory (LSTM) network is trained on filtered motion characteristics to model the complex residual errors between the EKF estimates and ground-truth positions. The learned correction is applied additively to preserve physical model stability. Furthermore, a magnitude-based fallback mechanism is introduced to monitor neural corrections; if a prediction exceeds a deterministic threshold, it is rejected as an anomaly, and the system reverts to the baseline EKF state. The framework is evaluated using the Google Smartphone Decimeter Challenge dataset. Experimental results demonstrate that the hybrid architecture mitigates tracking divergence, achieving up to a 45.8% reduction in root-mean-squared positioning error relative to raw GNSS baselines.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.