Opportunities of Self Supervised Learning for GNSS: Evaluation of a Deep Learning-Enhanced PVT Algorithm

Thomas Barbero; Bertrand Ekambi

cs.ROAug 26, 2026

Opportunities of Self Supervised Learning for GNSS: Evaluation of a Deep Learning-Enhanced PVT Algorithm

Authors: Thomas Barbero, Bertrand Ekambi

Organizations: Abbia GNSS Technologies, Toulouse, France

Abstract

This work proposes a Deep Learning Enhanced PVT algorithm to mitigate multipath interference in dense urban areas. A supervised objective jointly predicts range corrections and uncertainty, while a JEPA-based self-supervised pretraining stage improves representation quality. The algorithm is evaluated over diverse driving scenarios, substantially improving PVT accuracy, particularly for unseen harsh urban conditions. These results highlight the potential of unlabelled GNSS data to improve generalization performance.

Figures & tables

Panel of Figure 1 — Figure 1: Methodology overview.

Contextual Features		MP Indicators
Name	Inclusion Range	Name	Inclusion Range
Azimuth	]0,360]	C/N 0 Rover	$>5$
Elevation	]0,90]	Rover/Base C/N 0 Ratio	N/A
Satellite ID & Constellation	N/A	Range-Rate Consistency	] $-10$ ,10[ $\backslash$ {0}
GPS Time	N/A	Doppler Residuals	] $-10$ ,10[ $\backslash$ {0}
Velocity (ENU)	N/A	SD Code Residuals	] $-150$ ,150[ $\backslash$ {0}

Table 1: Summary of the extracted features and their respective inclusion range.

	Base Learning Rate	Weight Decay	Warmup	Epochs	Batch Size
Pretraining	1e-4	0.04	30%	50	1000
Finetuning	3e-4	0.01	10%	30	500
SPRVO	3e-4	0.01	70%	80	500

Table 2: Summary of the different training phases hyper-parameters.

	ID	Slight OOD	Heavy OOD
Baseline	4.98	9.16	28.76
SPRVO	2.85	7.28	21.26
SSL+ FNT	2.64	7.72	18.96

Table 3: Average of the 50 th and 95 th percentiles of 3D Positioning Error per method and dataset.

Panel of Figure 2 — Figure 2: Range/Prediction Level analysis.

Panel of Figure 3 — Figure 3: Average attention score per relative sky plot bin for the layers of (a) Encoder and (b) Decoder.

Pretraining Data \| Finetuning Data	ID	Slight OOD	Heavy OOD
Basis \| Basis	2.55	9.00	20.06
Basis \| Mixed	2.57	9.16	18.67
Mixed \| Basis	2.55	8.59	19.70
Mixed \| Mixed	2.68	8.82	18.71

Table 4: Average of the 50 th and 95 th percentiles of 3D Positioning Error per pretraining and finetuning data combination.

A Hybrid Neural-Assisted Unscented Kalman Filter for Unmanned Ground Vehicle Navigation
Mar 12, 2026Gal Versano, Itzik KleinRobot Navigation Unscented Kalman Filtering
Autonomous Navigation and Sensor Fusion Lab, Hatter Department of Marine Technologies, Charney School of Marine Sciences, University of Haifa, Israel.
Development of Vision-Language Model-based GNSS Spoofing Detection for Autonomous Vehicle Navigation
Jul 27, 2026Mohammed Aldeen, Muhammad Sami Irfan, Sagar Dasgupta +3VLMs for Autonomous Driving Autonomous Driving
School of Computing, Clemson University, Clemson, SC, USA · Department of Civil, Construction & Environmental Engineering, The University of Alabama, Tuscaloosa, AL, USA · Glenn Department of Civil Engineering, Clemson University, Clemson, SC, USA
Faster Training, Fewer Labels: Self-Supervised Pretraining for Fine-Grained BEV Segmentation
Feb 20, 2026Daniel Busch, Christian Bohn, Thomas Kurbiel +3Bird's-Eye-View Segmentation Autonomous Driving Perception
University of Wuppertal · APTIV

Mar 12, 2026J/K move · Enter open · S save

A Hybrid Neural-Assisted Unscented Kalman Filter for Unmanned Ground Vehicle Navigation

Gal Versano, Itzik Klein

Autonomous Navigation and Sensor Fusion Lab, Hatter Department of Marine Technologies, Charney School of Marine Sciences, University of Haifa, Israel.

Robot Navigation Unscented Kalman Filtering Covariance Estimation Multimodal Sensor Fusion

Open paper AskPDF arXiv

Modern autonomous navigation for unmanned ground vehicles relies on different estimators to fuse inertial sensors and GNSS measurements. However, the constant noise covariance matrices often struggle to account for dynamic real-world conditions. In this work we propose a hybrid estimation framework that bridges classical state estimation foundations with modern deep learning approaches. Instead of altering the fundamental unscented Kalman filter equations, a dedicated deep neural network is developed to predict the process and measurement noise uncertainty directly from raw inertial and GNSS measurements. We present a sim2real approach, with training performed only on simulative data. In this manner, we offer perfect ground truth data and relieves the burden of extensive data recordings. To evaluate our proposed approach and examine its generalization capabilities, we employed a 160-minutes test set from three datasets each with different types of vehicles (off-road vehicle, passenger car, and mobile robot), inertial sensors, road surface, and environmental conditions. We demonstrate across the three datasets a position improvement of

12.7\%

compared to the adaptive model-based approach. Thus, offering a scalable and a more robust solution for unmanned ground vehicles navigation tasks.

Abstract

Explore similar work

A Hybrid Neural-Assisted Unscented Kalman Filter for Unmanned Ground Vehicle Navigation

Development of Vision-Language Model-based GNSS Spoofing Detection for Autonomous Vehicle Navigation

Faster Training, Fewer Labels: Self-Supervised Pretraining for Fine-Grained BEV Segmentation