Purdue University leads research on autonomous transportation network
WEST LAFAYETTE, Ind. — In a future where fully autonomous vehicles rule the road, cars and trucks will move to a split-second torrent of data shared by computers that are on board, nearby and in the cloud. In this dynamic web of communication, any failure in the network is a threat to passengers. To ensure that cyber-physical systems — like a fully automated transportation system — can withstand breakdowns, hacking and unanticipated situations, the National Science Foundation (NSF) is funding a new center led by Purdue University to develop scientific mechanisms that enable resilience in the face of threats.
Researchers at Georgia Tech, the University of Southern California and the University of Wisconsin-Madison join Purdue in the $7 million grant awarded through the NSF’s Directorate for Computer and Information Science and Engineering, said Saurabh Bagchi, project lead, Purdue professor of electrical and computer engineering, and a member of the Purdue Institute for Physical Artificial Intelligence. Researchers will work with General Motors, Intel, Amazon Web Services and the Indiana Department of Transportation — each of which is represented on the center’s advisory board — with the goal of using technology development to help inform policy.
The grant, awarded by the NSF’s Division of Computer and Network Systems, will enable Bagchi and collaborators to focus on the problem of connected and autonomous systems (CATS). Through this project, the team expects to develop new computational approaches and solutions that ensure resilience in systems composed of interconnected hardware and software components, which individually may not be inherently highly reliable or secure. While the project will develop these tools around CATS, the foundational computer breakthroughs that will emerge from the project will be applicable to many large-scale cyber-physical systems.
At Purdue, Bagchi is joined by Somali Chaterji, an associate professor of agricultural and biological engineering; Tim Cason, a Distinguished Professor of Economics; Aravind Machiry, an assistant professor of electrical and computer engineering; Shreyas Sundaram, the Marie Gordon Professor of Electrical and Computer Engineering; and Carla Zoltowski, an associate professor of engineering practice.
The closest analog to autonomous transportation can be seen in advanced driver assistance systems like Tesla’s Autopilot and the “robotaxis” operated — with the aid of a remote driver when necessary — by a variety of companies in a few U.S. and Chinese cities. None are fully automated, but even these limited examples demonstrate the vulnerability of the communications web that governs these vehicles. In a widely circulated 2023 incident, for example, connectivity issues caused as many as 10 robotaxis in San Francisco to shut down, blocking traffic in a busy section of the city.
“Even when we’re looking only at one vehicle, it’s ripe for different kinds of failure and attack scenarios,” said Bagchi. “If the vehicle gets disconnected from the network, if the sensors are compromised because of an attack or they malfunction for one of many reasons, then the vehicle is flying blind, and bad things happen as a result of this.” In an article that recently appeared in Computer, a publication of the Institute of Electrical and Electronics Engineers, Bagchi and colleagues enumerated some other factors that protect network-based decision-making like ensuring built-in redundancy, physically protecting sensors in hazardous environments and even simply replacing default passwords on components in a given system.
A fully connected and autonomous transportation system — imagine the entire American road network functioning autonomously — would entail a web of communication with far greater complexity. Computation in this network would be shared among vehicles, roadside computers and the cloud. Each vehicle would be equipped with its own software/hardware stack and a vehicle, for example, might share weather, hazard and traffic conditions with other vehicles behind it on the road. Installing roadside “edge computers” is expensive, but each one enables light processing of local traffic where possible and acts as a fail-safe against disconnection from the network. Cloud computers, which can handle heavy processing, are cost-effective but strain network bandwidth. To top it off, all the new equipment must be installed around existing legacy infrastructure, which places additional constraints on design.
“It’s beastly expensive to stand this up from the ground up. And unless you get to safety and reliability, nobody is going to trust your CATS. It can have all the bells and whistles in the world, but unless you can give guarantees about the safety and security, nobody is going to use them,” Bagchi said.
Nevertheless, Bagchi said, having all vehicles connected and operating autonomously will improve safety. He estimates an 85% reduction in the current rate of 100 deaths per day according to the Insurance Institute for Highway Safety, and more than 14,000 collisions per day, according to the National Highway Traffic Safety Administration.
CATS would optimize automotive travel. At a GM test platform that the center may access in Michigan, cars skim past one another at a four-way intersection equipped with sensors rather than traffic lights, coordinating an efficient and flawless passage through the network without the need to stop.
The new center, dubbed CHORUS, will examine three broad classes of vulnerabilities: naturally occurring errors in hardware and software, malevolent actors, and unexpected interactions. Naturally occurring errors are baked into components as they are built. Malevolent actors deliberately seek to induce failure, like causing network congestion or shutting down components. Unexpected interactions cover scenarios where humans and systems interact in a manner that was not considered as the system was developed.
CHORUS, which will acquire two fully automated cars and conduct some tests at the GM facility, has carved the work into three “thrusts” in which researchers will model a CATS (including threats and how failures may ripple through the system), determine the most effective approaches to avoid failures, and finally test their resilience of their approach in the face of disruptions. By monitoring and detecting the failures, and then developing response and recovery tactics, the team hopes to learn how to maintain the core functionality of the system despite disruptions.
“By using this monitoring information, we run algorithms to very quickly decide what kinds of countermeasures to take so that the system can achieve its safety goals,” Bagchi said. “You’re not going to operate the system always at the optimal point. But you’re always going to operate the system in a way where it meets its safety and security and reliability guarantees.”
The NSF-funded CHORUS center is also a pillar of Purdue Computes, a strategic university initiative to further scale Purdue’s research and educational excellence.
About Purdue University
Purdue University is a public research institution demonstrating excellence at scale. Ranked among top 10 public universities and with two colleges in the top four in the United States, Purdue discovers and disseminates knowledge with a quality and at a scale second to none. More than 105,000 students study at Purdue across modalities and locations, including nearly 50,000 in person on the West Lafayette campus. Committed to affordability and accessibility, Purdue’s main campus has frozen tuition 13 years in a row. See how Purdue never stops in the persistent pursuit of the next giant leap — including its first comprehensive urban campus in Indianapolis, the Mitch Daniels School of Business, Purdue Computes and the One Health initiative — at https://www.purdue.edu/president/strategic-initiatives.
Media contact: Mary Martialay, mmartial@purdue.edu