Towards Improving Confidence in Autonomous Vehicle Software: A Study on Traffic Sign Recognition Systems

Aslansefat, Koorosh; Kabir, Sohag; Abdullatif, Amr; Vasudevan Nair, Vinod; Papadopoulos, Yiannis

Abstract

The application of artificial intelligence (AI) and data-driven decision-making systems in autonomous vehicles is growing rapidly. As autonomous vehicles operate in dynamic environments, the risk that they can face an unknown observation is relatively high due to insufficient training data, distributional shift, or cyber-security attack. Thus, AI-based algorithms should make dependable decisions to improve their interpretation of the environment, lower the risk of autonomous driving, and avoid catastrophic accidents. This paper proposes an approach named SafeML II, which applies empirical cumulative distribution function (ECDF)-based statistical distance measures in a designed human-in-the-loop procedure to ensure the safety of machine learning-based classifiers in autonomous vehicle software. The approach is model-agnostic and it can cover various machine learning and deep learning classifiers. The German Traffic Sign Recognition Benchmark (GTSRB) is used to illustrate the capabilities of the proposed approach.

Citation

Aslansefat, K., Kabir, S., Abdullatif, A., Vasudevan Nair, V., & Papadopoulos, Y. (in press). Towards Improving Confidence in Autonomous Vehicle Software: A Study on Traffic Sign Recognition Systems. Computer,

• Journal Article Type: Article
• Acceptance Date: Apr 19, 2021
• Deposit Date: Apr 20, 2021
• Publicly Available Date: Oct 27, 2022
• Journal: Computer
• Print ISSN: 0018-9162
• Publisher: Institute of Electrical and Electronics Engineers
• Peer Reviewed: Peer Reviewed
• Keywords: Autonomous systems; Safety assurance; AI safety; Statistical distance measure; SafeML; Safe machine learning
• Public URL: https://hull-repository.worktribe.com/output/3757602