Skip to content
Conference

Uncertainty-Aware Deep Kinematics and ESKF Fusion for Robust Vehicle Dead Reckoning

Aug 2026 · International Conference on Automation and Computing · pp. 1-10 · 0 citations · 22 references

Abstract

Achieving reliable vehicle localization in GNSS outage environments is a critical challenge. Traditional Dead Reckoning (DR) systems using low-cost Micro-Electro-Mechanical Systems (MEMS) Inertial Measurement Units (IMUs) inevitably suffer from rapid error accumulation. This paper proposes an uncertainty-aware deep kinematics and Error-State Kalman Filter (ESKF) fusion framework. A hybrid one-dimensional convolutional neural network and gated recurrent unit (1D-CNN-GRU) model, trained with a Gaussian Negative Log-Likelihood (GNLL) loss, predicts body-frame velocity and data-dependent variance from raw IMU data. In the ESKF update, the learned forward velocity is used as the neural longitudinal measurement, while the lateral channel is constrained by the vehicle non-holonomic assumption. The learned forward variance is mapped to the measurement noise covariance to adaptively regulate the Kalman gain. Evaluated on a 394.34 m KITTI GNSS outage stress case with complex cornering, the proposed method suppresses position drift. In this stress case, the method reduces the terminal error by 77.39% relative to the ESKF + NHC baseline and yields a terminal drift rate of 7.32%. These results indicate that, although the learned velocity uncertainty is not fully calibrated in this stress case, it can still serve as a relative measurement-quality cue for vehicle dead reckoning under the evaluated GNSS outage condition.

View source

Similar papers

Conference Sep 2026

A quality-aware robust extended Kalman filter for low-cost GNSS/IMU vehicle localization in urban canyons

Accurate vehicle localization in urban canyons using low-cost Global Navigation Satellite System (GNSS) and Inertial Measurement Unit (IMU) is challenging due to multipath and non-line-of-sight (NLOS) receptions. Conventional loosely coupled Extended Kalman Filters (EKFs) with fixed covariance often over-rely on degrad...

Bo-Qin Yang · 0 citations
Open access 2026

DRIFT-Net: Drive-State-Aware Weighted Interpolation for Error Forecasting and Trajectory Correction Network

DRIFT-Net is presented, a DRive-state-aware Weighted Interpolation for error Forecasting and Trajectory correction Network, to improve dead-reckoning-based vehicle localization and achieves improved performance compared with existing methods.

Jae-won Jeong, Ying Quan, Woo-Young Choi · 0 citations
Open access Sep 2026

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

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 p...

Ádám Ragasits, Katalin Kovács · 0 citations
Open access 2026

AMC-LIEKF: An Adaptive Maximum Correntropy Left-Invariant Extended Kalman Filter Method for Tightly Coupled Multi-UWB/IMU Integration

The error-state extended Kalman filter (ESKF) is widely used to fuse ultrawideband (UWB) and inertial measurement unit (IMU) data for robust localization in global navigation satellite system (GNSS)-denied environments. However, conventional ESKF-based fusion methods still face challenges such as geometric inconsistenc...

Cheng-Peng Yin, Xiang Yu, Rui-Chang Fan et al. · 0 citations
Conference Open access Aug 2026

Model-inertial deep fusion navigation and impact point uncertainty quantification for rocket sub-stages under prolonged GNSS outages

During atmospheric re-entry of rocket sub-stages, GNSS signals may be lost for extended periods due to plasma sheath and attitude occlusion. In such scenarios, the integrated navigation system degrades to pure inertial navigation, and position errors grow cubically over time, making conventional filtering and predictio...

Tao-Ran Xu, Jing Yang, Kang Wang et al. · 0 citations

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