Wednesday, March 14, 2018 - 9:00am to 10:00am
KEC 1007

Speaker Information

Houssam Abbas
Postdoctoral Fellow
Department of Electrical and Systems Engineering
University of Pennsylvania

Abstract

Although autonomous vehicles have driven several millions of miles, even the most basic maneuvers such as a lane change are prone to systemic accidents. As blackbox algorithms provide decision control for life-critical systems they pose a significant liability, and require new foundations for testing and verification for safety and performance. In my research, I seek to answer the question: “What is a Driver’s License Test for Driverless Vehicles?”. More generally, how can we provide safety guarantees for autonomous systems that incorporate perception, planning and control sub-systems across a range of environments? In this talk, I will describe two efforts for the verification and testing for such complex autonomous Cyber-Physical Systems.

I first describe Robustness-Guided Verification (RGV), a formal technique for verifying hybrid dynamical systems with spatial, temporal and reactive safety properties captured as temporal logic specifications. I apply RGV to autonomous vehicle (AV) models with idealized perception pipelines to quantitatively verify the AV decision controller across a range of driving scenarios. I then discuss the challenges of incorporating perception in the verification effort, and present a test harness for autonomous vehicles that allows the testing of black-box code, including perception pipelines, in a photorealistic simulation environment. This opens the way to exciting research directions in real-time verification, multi-vehicle control, and computer vision. Design automation, as embodied in this research, will supply crucial evidence and safety benchmarks to regulatory bodies like NHTSA, the FAA and FDA as society navigates its way in a driverless world.

Speaker Bio

Houssam is a postdoctoral fellow in the Department of Electrical and Systems Engineering at the University of Pennsylvania. His research interests are in the verification, control and conformance testing of autonomous Cyber-Physical Systems. Current research includes the verification of life-supporting medical devices, the verification and control of autonomous vehicles with a view towards certifying such systems, and anytime computation and control. Houssam holds a PhD in Electrical Engineering from Arizona State University.