DREAM - DRiving aids by E-GNSS Al and Machine learning

Leveraging DREAM DRiving aids by E-GNSS Al and Machine learning on Galileo unique differentiators, like the High Accuracy service and OSNMA, and on innovative Artificial Intelligence/Machine Learning techniques, DREAM is aimed at designing, producing and demonstrating, in an operational environment, an integrated solution to improve positioning technologies and environment perception systems supporting the development of driver assistance systems for buses, up to TRL7. The target application of this device is "Driving aids for buses" although it can be extended to other public transports or private vehicles.

Although the current use of GNSS in Public Transport is relatively limited when compared to other, more developed sectors, driving aid applications have experienced continuous growth over the past years due to their undoubted benefit in transport safety but also on traffic and emissions reduction, thanks to the avoidance of accidents and collisions.

The growing trend to include GNSS solutions fused with other sensors, like cameras or LIDAR based on AI/ML techniques, is not only driven by users demand but also by the European regulation, supporting and mandating the development of these type of solutions.

DREAM aims at developing innovative artificial intelligence and machine learning methods improve positioning to technologies and environment perception systems supporting the development of driver assistance systems for buses.

The final goal is to provide high accuracy and reliable information that could support ADAS systems like Red Light Violation Warning, Curve Speed Warning, Collision Avoidance, and Wrong Way driving.

To achieve this goal, AI/ML techniques for GNSS, inertial and visual/LIDAR sensors are combined in a very well-balanced way, so that, the advantages of all these sensors are exploited and their limitations mitigated, resulting in an accurate, highly available and resilient solution.

Challenge and technical solution

DREAM aims to address the stringent requirements of ADAS systems in challenging urban environments by developing advanced Al solutions. It focuses on enhancing GNSS accuracy and resilience through leveraging Galileo differentiators and Al techniques to detect spoofing, multipath and NLOS situations and ensuring correct ambiguity resolution. For improved localization in GNSS denied scenarios, Al-driven methods will be applied for IMU calibration and denoising, as well as LiDAR/Visual SLAM (enhance with Al based moving objects removal). It will also feature 3D bounding boxes for object detection and geo-referenced maps supporting LiDAR/Visual localization and improving situational awareness.