CPS: Medium: Co-Simulation for Autonomous Vehicles Certification and Control (COSACC)

Autonomous driving, aided by machine learning (ML), has the potential to improve safety and efficiency while reducing road congestion. However, today's autonomous vehicles (AVs) struggle under arbitrary conditions due to the complexity and variety of scenarios encountered in practice. Meanwhile, increasingly advanced driver-assist systems with partial AV capabilities have been deployed without standardized testing and certification requirements, in particular with regard to computational aspects. This work aims to fill this gap. The objective of this work is twofold: (1) Create an environment supporting seamless transitions from in-lab digital twin testing over a hybrid co-simulation environment, part lab part physical, to fully deployed AV road testing. (2) Systematically identify minimum requirements, verification opportunities and limitations from which testing scenarios can be derived for lab, hybrid, and entirely physically deployed Avs, supporting the AV development cycle and certification. The project will bala the capabilities and limitations of verification within automotive control and its synergy to derive test scenarios for certification across development stages (digital twin, co-simulation, actual AV). The hypothesis is that vehicle verification and certification become synergistic in complementing each other and ideally need to be conducted only once (for verification) or at most twice (for testing) across the three stages. This project focuses on a subset of driver-assist systems with highly interacting subsystems. The work will investigate three levels of increasing complexity, ultimately requiring ML inference for object tracking, to test our hypothesis of synergy between verification and minimal testing for certification. Specifically, the project addresses research challenges of (1) real-time scheduling under hard and soft deadlines, including mixed-criticality scheduling and bounds on execution time for multicores and accelerators, (2) soundness of certification via verification at each development level and (3) both capabilities and limitations of validation with and without re-certification at each level. Expected results will advance Computer Science, Engineering and Cyber-Physical Systems (CPS) for low-level control systems while, within Civil Engineering, deriving test scenarios for certification from high-level requirements. It is this focus on the intersection between areas that facilitates the development of new methods for transitioning from digital controls to testing scenarios supported by our co-simulation approach with gradually increasing physical interaction. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.