People who worry about the safety of self-driving cars might either be reassured or concerned by an approach that Jingang Yi, professor of mechanical and aerospace engineering, is proposing: add race car driving skills and techniques to the brains of autonomous vehicles.
“Professional drivers perform extreme maneuvers to avoid crashes with other vehicles, or to avoid running off the road,” Yi said. “But typical drivers out on the streets and highways don’t know how to do them.”
When it comes to autonomous driving control and accident avoidance, Yi notes that the consensus among designers is to keep the car stable. But race car drivers often sacrifice stability on purpose, such as steering the car into a skid, to get out of dangerous situations.
Gaining Agility by Sacrificing Stability
“When we looked at race car driver data provided by Ford, we found these drivers gained agility by temporarily sacrificing the stability of their car, Yi said. “We want to quantify this and characterize this kind of driving style.”
Yi has a grant from the National Science Foundation to tackle the challenge of translating these human skills into control system designs. But he assures us not to expect our self-driving cars to routinely burn rubber to slide into a tight parallel parking slot.
“Maybe 99 percent of the time we will be in the stability region,” he said. “We would invoke these performance driving techniques when they are the only way to avoid an accident.”
One area where Yi wants to fortify vehicle stability is in motorcycles.
Improving Motorcycle Stability
“These ‘single-track’ vehicles are inherently unstable platforms,” he said. “Our work on physical control systems is not aimed at making motorcycles self-driving, but to help drivers interact with this inherently unstable platform.”
When driving or riding on these vehicles, Yi notes that body movement influences the vehicle. So he wants to understand how balancing tasks relate to driving tasks.
“These two control tasks connect with each other and also compete with each other. We are trying to reveal how humans achieve that and translate that understanding into control systems for single-track vehicles.”
Beyond vehicle stability, Yi is looking at using autonomous vehicles as a control strategy to improve traffic flow and capacity in a “real world” setting with a mix of autonomous and manually driven vehicles.
“One way we manage highway traffic is to use ramp metering to avoid congestion at interchanges,” he said. “But if we can control autonomous vehicles in a distributed way, a connected way, we can try to influence the manual drivers.” So Yi is working on traffic modeling with a mix of traffic to manage flow, improve efficiency and increase capacity.