Abstract

In this talk, we address the challenge of ensuring the safety of autonomous vehicles (AVs, also called ego actors) in real-world scenarios where AVs constantly interact with other actors. We introduce iPrism, which incorporates a new risk metric – the Safety-Threat Indicator (STI). Inspired by how experienced human drivers proactively mitigate hazardous situations, STI quantifies actor-related risks by measuring the changes in escape routes available to the ego actor. To actively mitigate the risk quantified by STI and avert accidents, iPrism employs a reinforcement learning-based Safety hazard Mitigation Controller (SMC) that learns and implements optimal risk mitigation policies.

Our evaluation of the success of the SMC is based on over 4800 safety-critical scenarios. The results show that

1. STI provides up to 4.9× longer lead time for mitigating accidents compared to widely used safety and planner-centric metrics,
2. SMC significantly (37% to 98%) reduces accidents as compared to a baseline Learning-by-Cheating (LBC) agent, and
3. in comparison with available state-of-the-art safety hazard mitigation agents, SMC prevents up to 72.7% of accidents that the evaluated agents are unable to avoid.

Speaker’s Profile

Dr Zbigniew Kalbarczyk is a Research Professor at the Department of Electrical and Computer Engineering and the Coordinated Science Laboratory of the University of Illinois at Urbana-Champaign. Dr Kalbarczyk research interests are in the design and validation of reliable and secure computing systems.

His current work explores:

1. emerging computing technologies, such as resource virtualization to provide redundancy and assure system resiliency to accidental errors and malicious attacks,
2. machine learning-based methods for early detection of security attacks, including defence against smart malware,
3. analysis of data on failures and security attacks in large computing systems to characterize system resiliency and guide development of methods for rapid diagnosis and runtime detection of problems, and
4. development of techniques for automated validation and benchmarking of dependable and secure computing systems using formal (e.g., model checking) and experimental methods (e.g., fault/attack injection).

Dr Kalbarczyk led the design and commercialization of

1. ARMOR high-availability software middleware to support resilient distributed applications and
2. NFTAPE software framework to support fault injection-based resiliency assessment.

Dr Kalbarczyk served as a program chair of Dependable Computing and Communication Symposium (DCCS), a track of the International Conference on Dependable Systems and Networks (DSN) 2007, and Program Co-Chair of Computer Performance and Dependability Symposium, a track of the DSN 2002. He was an Associate Editor of IEEE Transactions on Dependable and Secure Computing. He has published over 240 technical papers and is regularly invited to give tutorials and lectures on issues related to the design and assessment of complex computing systems. He is a member of the IEEE, the IEEE Computer Society, and IFIP Working Group 10.4 on Dependable Computing and Fault Tolerance.