AHEAD
An embedded, hierarchical, collaborative platform for autonomous driving
Last few years have seen the emergence of autonomous driving (AD for Autonomous Driving) following the successful implementation of advanced driver assistance systems (ADAS).
.However, managing such systems is becoming more complex due to the variety of challenges, including: 1) detection and perception through cameras and sensors, 2) communication with the environment, including decision-making systems and data transmission, 3) processing sensor information with sufficient accuracy and latency by designing new algorithms, 4) and finally executing and distributing algorithms with sufficient efficiency and QoS (Quality of Service) on new hardware platforms (HW for Hardware).
The processing and storage requirements for the next generation of DA will be significant, and will exceed the capacity of the vehicle's on-board hardware. A multi-tiered framework for AD tasks will be able to increase computing and storage power by offloading complex applications from constrained embedded platforms to the Fog and Cloud. The processing and memory capabilities of these higher levels of the hierarchy can provide an efficient means of implementing complex AD functionality.
Our goal is to propose a task distribution and scheduling framework in which decisions will be made at different levels of the framework based on time sensitivity, available resources, energy constraints, and accuracy requirements. In our approach, we propose a 3-tier architecture and execution policy to efficiently manage heterogeneous hardware (HW)/software (SW) resources for AD applications used in the Intelligent Transport System (ITS). The 3 levels are: the Edge level, which corresponds to the HW / SW resources in the vehicle, the Fog level, which corresponds to the HW / SW resources available in the road infrastructure, and finally the Cloud level, which provides servers and high-performance computing and storage capacities. The objectives of the AHEAD project are in line with many recent research fields, including Intelligent Transport Systems (ITS), Artificial Intelligence (AI), Embedded Systems and Cloud, Fog and Edge computing. It will be implemented in collaboration between the Université Polytechnique Hauts-de-France (UPHF) and other international partners.
Notre objectif est de proposer un framework distribution et d’ordonnancement des tâches dans lequel les décisions seront prises à différents niveaux du framework en fonction de la sensibilité au temps, des ressources disponibles, des contraintes énergétiques, et des exigences de précision. Dans notre approche, nous proposons une architecture à 3 niveaux ainsi qu'une politique d'exécution pour gérer efficacement les ressources matérielles (HW)/ logicielles(SW) hétérogènes pour les applications AD utilisées dans le système de transport intelligent (ITS pour Intelligent Transport System). Les 3 niveaux sont: le niveau Edge qui correspond aux ressources HW / SW dans le véhicule, le niveau Fog qui correspond aux ressources HW / SW disponibles dans l'infrastructure routière et enfin le niveau Cloud qui fournit des serveurs et des capacités de calcul et de stockage de haute performance. Les objectifs du projet AHEAD s'inscrivent dans le cadre de nombreux domaines de recherche récents, notamment les systèmes de transport intelligents (ITS), l'intelligence artificielle (AI pour Artificial Intelligence), les systèmes embarqués et le Cloud, le Fog et le Edge computing. Il sera mis en œuvre en collaboration entre l'Université Polytechnique Hauts-de-France (UPHF) et d'autres partenaires internationaux.
| Department(s) | Partner(s) | Overall amount |
|---|---|---|
| Informatics |
LAMIH (UMR CNRS 8201 / Université Polytechnique Hauts-de-France)
University of California Irvine (UCI), California, USA University of California Riverside (UCR) California USA IBM Thomas J. Watson Research Center, New-York, USA University of Bilkent (UB), Ankara, Turkey |
5 k€
|
| Main support | Rayout | Date(s) |
| CNRS |
International
|
2020 - 2021
|