Smart traffic is more than self-driving cars. One engineer is tackling these transportation systems with an early career award grant from the National Science Foundation (NSF).
Changing lanes, upcoming merges, blind spots — wouldn’t it be great to broadcast to other vehicles what your own car is doing and where? Well, the technology does exist, and the data-rich messages in vehicle-to-vehicle communication are less salty than the ones exchanged in human languages. But the technology needs more development before it’s ready for rush hour.
Kuilin Zhang explained the challenge is forecasting traffic, which is inherently uncertain. Zhang, assistant professor of civil and environmental engineering and affiliated assistant professor of computer science at Michigan Technological University, is a recipient of an NSF CAREER Award to improve automated driving decisions using predictive, real-time feedback within and between vehicles. The project totals $500,000 over a five-year span and puts some of Michigan Tech’s prime mobility testing facilities to use.
Self-driving Cars
Single-vehicle controls and sensors — or the tech that makes a vehicle self-driving — has been the primary focus of autonomy research to date. However, Zhang knows as a civil engineer that to understand traffic uncertainty, researchers can’t stop at examining a single car. As a computer scientist, he knows that the key is shared data.
A vehicle may be autonomous, defined at different levels by SAE International, but safe and efficient traffic requires it to be connected, meaning the vehicle communicates with the driver, other vehicles and infrastructure like traffic signals, signs or bridges.
“I want to understand how the real-time data from other vehicles can be utilized for automatic driving and routing decisions in the moment,” Zhang said. “In my vision of the future, we have more predictable, more robust, safer transportation systems — and it’s based on being connected and the data we can gather.”
The data fuels forecasts that swiftly analyze and predict changes in and around a vehicle. To make that happen, Zhang’s CAREER Award reflects the interdisciplinary demands of creating autonomous and connected systems: uniting principles from computer science, traffic science, operations research, control theory and machine learning, among others. The layers of decision making get complicated fast, which is why Zhang is using multiple tools to study feedback systems.
Routing Games and Predictive Modeling
A self-driving car is made as much of algorithms as steel and aluminum. To make a driving decision, like turn the steering wheel or hit the brakes, a vehicle needs good equipment and good data. To make a routing decision, like take this exit or change lanes, a vehicle needs everything it used for driving decisions plus an additional level of awareness of its surroundings and other vehicles.
That’s where communication comes in: connected as vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I) or vehicle-to-everything (V2X). Lab simulations using model predictive control (MPC) models and congestion games, called n-person dynamic routing games that crunch the numbers on the rippling effects of decisions made within a scenario, illuminate what a predictive, real-time V2V system is capable of.
Of course, road testing is crucial. To that end, Zhang will work with the APS LABS and their fleet of connected and automated vehicles, fully outfitted with a range of autonomous and connected vehicle equipment, and the five V2I traffic signals installed in Houghton, Michigan, by the Michigan Department of Transportation last fall. Additionally, Zhang will be collaborating with the company HERE Technologies, which specializes in high-definition (HD) maps and traffic data for automotive and traffic technology.
“Through this collaboration we will provide Professor Zhang HERE HD Live Map samples for simulation and road testing of connected and automated vehicles,” said Xin Chen, direcor of HERE Technologies. He explains that similar to HD maps, V2X improves road safety and efficiency; combining the two is what makes Zhang’s project unique. “I have known Professor Zhang for many years and his award is a testimony of his achievement, and his project will demonstrate the critical role of HD maps in connected and automated driving.”
Education
NSF CAREER Awards emphasize both research and teaching. To Zhang, the two are as connected as the vehicles he studies.
“Michigan is the leading state studying autonomous and connected vehicles,” Zhang says. “Here at Michigan Tech, we’re training future traffic engineers and, to prepare them for our roads, we need a new approach to studying traffic.”
Zhang works with students across many levels, from graduate students refining congestion games and undergraduates testing out the models on an app specially developed for Zhang’s classes, to high school students taking a summer internship in the lab to learn the basics of self-driving cars.
“The department’s vision is to provide unique educational and research opportunities for students by faculty that promote fundamental understanding of our disciplines while providing a strong foundation to adapt to future change and improve the condition of humanity,” said Audra Morse, chair of the Department of Civil and Environmental Engineering. “Zhang’s research in connected and automated transportation systems in smart cities does just that.”
Much like the research itself, Zhang’s teaching style spans disciplines and tools. To create a future where vehicles forecast the movements and changes in traffic, engineers — and future engineers — model the complexity of how autonomous and connected rubber hits the road.
