Project: #95 Computer Aided Design for Safe Autonomous Vehicles Progress Report - Reporting Period Ending: March 30, 2020 Principal Investigator: Rahul Mangharam Status: Active Start Date: July 1, 2017 End Date: June 30, 2021 Research Type: Applied Grant Type: Research Grant Program: FAST Act - Mobility National (2016 - 2022) Grant Cycle: 2017 Mobility21 UTC Progress Report (Last Updated: Nov. 23, 2019, 2:59 p.m.) % Project Completed to Date: 70 % Grant Award Expended: 50 % Match Expended & Document: 50 USDOT Requirements Accomplishments Safety is the number one challenge for commercial mass deployment of Automated Vehicle technology. Industry players have been pushing their Autonomous Vehicle safety vision with frameworks in the form of “Robot Safety Laws”. Since all proposed safety frameworks are parametric models, understanding the parameter impact on the overall safety is crucial. The primary goal of this project is to develop a search engine for scalable testing of adverse driving instances. The search engine is developed as a test harness around autonomous vehicle simulation engines such as CARLA. We are able to test the robustness (i.e. how well are the safety design requirements met) for a variety of driving scenarios. In this effort we developed the robustness-guided testing methodology to establish the safety boundaries for AVs. We provide a system to develop safety benchmarks that help understand the impact of di?erent model parameters. Using our proposed test harness, we perform the Responsibility Sensitive Safety (RSS) parameter exploration, analyze the safe longitudinal distance across various traffic scenarios and explore the trade-off between conservative and assertive driving within the safety envelope. Impacts This work has been demonstrated at the Intel Autonomous Driving Community of Partners - a consortium of AV researchers from US, Europe, China and India. The work is being integrated as a core feature in the CARLA AV simulator. This work has been presented to the Inter-American Development Bank as a set of safety benchmarks to be used in Central America and South America as they develop models for rolling out AVs. This work is the subject of a planned AAA Summit on AV Safety and Liability in the tri-state region. This summit will be hosted on May 8, 2020 at the University of Pennsylvania campus in Philadelphia. This work has resulted in deployment partnership with StoneRidge Corporation, a $1B+ commercial truck automation company that is using digital mirrors to enhance safety of commercial vehicles. This work has been demonstrated at the NSF US-German Highly Automated Vehicles workshop in December 2019 in Washington DC. It is an example safety benchmark that is not based on a set of abstract scenarios but incorporates naturalistic driving behavior. This work is a subject of the upcoming Computing Research Association's Assured Autonomy Workshop, Feb 2020. It is also the primary theme of Mathematical Challenges and Opportunities for Autonomous Vehicles (Institute for Pure & Applied Mathematics, An NSF Math Institute at UCLA) in 2020. This work was the theme of the Penn PRECISE Center's "Safe AI" Industry Day that was held in October 2019 with over 170 industry participants in the University of Pennsylvania campus in Philadelphia. Other Mobility21 Next Generation Truck Freight Transportation Summit - https://precise.seas.upenn.edu/events/conference/20181024-moving-america-forward-next-generation-truck-freight-transport Safe AI Symposium - https://precise-industry-day.seas.upenn.edu/2019/ F1/10 Autonomous Racing (open-source software, simulators and hardware) - http://f1tenth.org/ Outcomes New Partners Intel Corporation StoneRidge International Corporation Inter-American Development Bank Columbia University Cyber-Physical Systems Week Embedded Systems Week IFAC World Congress Issues We are growing the number of research projects to 5 PIs in the coming year.