iLoc: High-integrity Localization
for Automated Vehicles

Scope

Continuously and reliably estimating a vehicle's position in varying driving environments is essential for autonomous driving and safe operation. However, different from the aviation domain, dynamic and complex traffic environments make high-integrity localization very challenging in the vehicular domain. In our 1st iLoc workshop, we want to identify potential solutions to remedy the problems, such as uncertainties in both environmental perception and vehicle localization, vision-based deep learning models for integrity monitoring, and the development of standardization of integrity localization for automated vehicles.

A wide range of research on the above-mentioned topics from both academia and industry will be presented by scientists and experts. A mixture of presentations from invited speakers and paper authors is implemented in our full-day workshop. In order to guarantee a highly interactive and fruitful workshop, we recommend that all the authors of papers bring their posters to the workshop. All the participants can better discuss and communicate with your work during the one-hour poster session, or even during the lunch break.

Topics

The research questions of interests include but are not limited to:

• What are the leading factors for high-integrity localization for AVs, e.g., road geometry, vehicle dimension, sensor accuracy, environmental factors, vehicle motion, and other road users?

• What are the integrity measures for AVs, in longitudinal, lateral, vertical (e.g., multi-level interchange), pitch, roll, and heading angle dimensions?

• What are the developments of standardization in the vehicular domain for integrity performance?

• How to estimate the uncertainty and integrity risks in e.g., conventional and deep learning-based models for localization and autonomous driving?

• How to combine a vehicle kinematic model and road geometry to improve integrity estimation? Is the sequence of AV's maneuver tasks of driving from position A to B well defined, and how can a map (e.g., map as a sensor) be associated for its localization over its movement?

The following lists the potential but not exhausted research topics:

• Uncertainty propagation and updates while an AV drives in different environments.

• Map reference with its own integrity measure, including 2D digital map, high-definition map, 3D map/3D city model, even more precise survey map with both dynamic and static objects.

• Quantification and representation of the models' aleatoric (or statistical) and epistemic (or systematic) uncertainties for both environmental perception and localization.

• Uncertainty estimation of LiDAR point clouds registration and imagery data processing in e.g., probabilistic and deep learning-based models.

• State-of-the-art deep learning multi-modal data fusion for e.g., GNSS, LiDAR point clouds, images, 3D map localization information with integrity estimation.

Funding