VEH09
ROBUSTNESS OF ARCHITECTURES AND SYSTEMS
Challenges
The objective of the VEH09 project is to develop architectural and intelligence principles for the autonomous system as well as methods and justifications to ensure the personal safety and reliability of autonomous vehicle systems where the driver is no longer part of the vehicle control and decision loop.
This project involves a number of themes:
- Recommendations on the functional and system architectures used for localisation, mapping, perception, planning or decision-making, to optimise and qualify the autonomous steering system of a robot vehicle as highly dependable.
- Decision-making and intelligence modes of the autonomous cars, based on sound functional specifications, priority and redundancy strategies or artificial intelligence algorithms.
- The safety of people due to the autonomous nature of the vehicle to justify its use on open roads, knowing that this justification and the safety record cannot rely on existing standards.
- Implementation of test facilities to learn from actual situations in order to simulate and test demonstrators in critical situations and validate the corresponding solutions.
Research themes
- Operating safety of a level 4 autonomous vehicle in an urban environment
- Secure, non-vulnerable operating architecture
- Safety during trials
Project description
The robustness of the systems and architecture in terms of safety and reliability has to be demonstrated and tested to guarantee a high level of safety for road users.
Trajectory generation must focus on three concepts: the desired trajectory, a trajectory that optimises risks and an emergency or refuge trajectory.
The architecture of all these elements is built on the basis of operating safety studies, security requirements and the related potential degraded modes, and the operating performance of all the sensors and algorithm software.
In addition, it is necessary to monitor the vehicle and its systems as it drives through its environment to detect and learn real-life critical situations and study them for statistical risks, technical requirements and simulation purposes.
Future prospects
A robust system involves the identification of risk situations through the observation of a very high number of driving situations to establish the related probabilities and the methods required to ensure a safer trajectory. There may be different types of method: SWet HW real-time diagnosis methods, artificial intelligence methods (learning, neurone networks, etc.) for a high-performance decision module, performance qualification to ensure secure architecture, fault tolerance, and compatibility between technology and safety or vulnerability objectives.