Driverless cars could reach wide-scale adoption more quickly if human and artificial intelligence are combined in autonomous vehicles in a system likened to air traffic control, say researchers.
“Today’s autonomous vehicles can drive relatively well in typical settings, but they fail in exceptional situations—and it’s those situations that are the most dangerous”
A team at the University of Michigan’s Transportation Research Institute (UMTRI) is working on a project that draws on a technique called instantaneous crowdsourcing. The project is focused on providing real-time remote backup for autonomous vehicles facing a disengagement – a situation where the car’s systems may need human help.
“Today’s autonomous vehicles can drive relatively well in typical settings, but they fail in exceptional situations—and it’s those situations that are the most dangerous,” said project leader Walter Lasecki, an assistant professor of computer science and engineering.
“Designing autonomous systems that can handle those exceptional situations could take decades, and in the meantime, we’re going to need something to fill the gap.”
The instantaneous crowdsourcing approach aims to provide a human response to potential disengagements in just a few milliseconds – avoiding the need for a vehicle to pull over and wait for a remote human operator to take control or advise on next steps.
The UMTRI team’s approach would use a pre-emptive system, identifying the likelihood of a potential disengagement by a driverless vehicle.
If this likelihood exceeded a pre-set threshold, the vehicle’s software would then share its data with a remote control centre, allowing the system to analyse the data, generate several possible scenarios and show them to human supervisors.
These supervisors would then respond to the simulations and these responses would be sent back to the vehicle, allowing it to choose from a library of human-generated responses. The whole process would take less than five seconds, according to researchers.
Robert Hampshire, a research professor at UMTRI and U-M’s Ford School of Public Policy, said this approach would be far cheaper than requiring a human driver in every vehicle.
“There were 3.2 trillion miles driven in the US last year, and the best autonomous vehicles averaged one disengagement every 5,000 miles,” said Hampshire.
“We estimate that if all those miles were automated, you’d need around 50,000 to 100,000 employees, distributed city by city. A network like that could operate as a subscription service, or it could be a government entity, similar to today’s air traffic control system.”