SALIENCE4CAV

Purpose and goal:

The purpose of the project was to improve the work processes around safety assurance of automated vehicles, with a special focus on a development process enabling continuous updates of products. Questions about safety for cooperative/collaborative vehicles, machine learning, and human-machine interaction were also included. The goals were to develop principles and methods that support the work of demonstrating that an automated vehicle is safe, and to advance the collective knowledge in the field through scientific publications based on our results.

Expected results and effects:

The results are principles and methods described in 15 open publications. For example, methods for precautionary driving, human/vehicle safety analysis, principles of safe minimal risk manoeuvres for both individual and cooperative/collaborative vehicles, and use of out-of-distribution detection as a safety mechanism for machine learning. In addition, we have worked with safety contracts for agile development.

Expected effects are improved and more efficient safety assurance that leads to fewer accidents and more accessible automated transport solutions. Approach and implementation:

The project was organized around four main topics: safety assurance, safety design, safety for machine learning, and variational safety. The method was analysis of relevant use cases and state-of-the-art, and identification of gaps. The research questions have been refined iteratively during the project.

Method development was mainly done at a theoretical level, in some cases supported by simulations. The results have been validated through scientific publications (peer-reviewed), presentations in relevant forums, and feedback from the project´s industrial partners.