Rahul Mangharam

Rahul Mangharam received the 2016 US Presidential Early Career Award  (PECASE) from President Obama for his work on Cyber-Physical Systems. He  also received the 2016 Department of Energy’s CleanTech Prize (Regional),  the 2014 IEEE Benjamin Franklin Key Award, 2013 NSF CAREER Award, 2012 Intel Early Faculty Career Award and was selected by the National Academy of Engineering for the 2012 and 2017 Frontiers of Engineering. He is the Penn Director for Mobility21 – the $14MM Department of Transportation National University Transportation Center where the focus is on the design and deployment of safe autonomous systems. As such, he has a good understanding of the technical, regulatory and industry challenges to get safe AVs on the road.

Rahul is a Professor in the Dept. of Electrical & Systems Engineering and Dept. of Computer & Information Science at the University of  Pennsylvania. His interests are in cyber-physical systems which involves the tight coupling of communication, computation and control with physical systems. His current focus is on applications within medical devices, energy efficient buildings, automotive systems and industrial wireless control networks. He received his Ph.D. in Electrical & Computer Engineering from Carnegie Mellon University where he also received his MS and BS in 2007, 2002 and 2000 respectively.

A Driver’s License Test for Driverless Vehicles: AV software verification
and safety certification

Autonomous vehicles (AVs) have driven millions of miles on public roads, but even the simplest scenarios, such as a lane change maneuver, have not been certified for safety. As there is no systematic method to bound and minimize the risk of decisions made by the vehicle’s decision controller, the insurance liability of autonomous vehicles currently is entirely on the manufacturer. I will describe APEX, a tool for autonomous vehicle plan verification and execution across a variety of driving scenarios.  We will see the use of synthetic environments such as computer gaming to train and evaluate machine learning and decision control algorithms in future AVs.

I will demonstrate a toolchain that captures different AV driving scenarios (merge, highway driving, exit, roundabout, T-junction, etc.) and exhaustively testing to efficiently expose high-risk driving decisions. I will also show a lot of videos of AV crashes and how we can effectively use computer games such as Grand Theft Auto to efficiently capture high risk driving scenarios – example:


This talk will appeal to both a general audience and to experts in legal, machine learning, perception, planning, controls and manufacturing.