Before autonomous vehicles can fully be integrated into our roads, a massive amount of testing must take place. This is being done in a handful of ways, but perhaps the most effective is the approach by the Mcity team in Michigan.
A model city complete with traffic lights, fake pedestrians, street signs, buildings, even bike lanes and fire hydrants, Mcity opened for business in 2015. While UBER and Waymo (formerly the Google self-driving car project) continue to conduct long mileage testing around the globe, Mcity is tackling the most difficult autonomous driving situations head on at its 32-acre closed testing facility, such as how to handle blind turns, pedestrian crossings, and highway overpasses.
Since Mcity’s launch, the conversation around self-driving vehicles has changed in North America: Most accept that the revolution is coming, but a chorus of voices continues to call for increased testing and safety measures. Despite that fact that a few hundred autonomous vehicles have driven millions of miles with comparatively fewer accidents (3.2 accidents per million miles, vs. 4.2 per million miles for regular cars) policy makers and the general public have yet to be fully convinced.
The program’s director, Dr. Huei Peng, who leads a team backed by over 60 technology and industry partners, explained in a phone interview that Mcity is driven by improving road safety.
He believes that driving is a distraction from our lives and that autonomous cars will allow us to focus on other parts. According to Peng, roads will remain dangerous as long as humans are behind the wheel—think texting while driving, road rage, and drunk drivers.
Peng and his team are fixated on the numbers—especially the 1.3 million deaths on roads worldwide last year.
“See, the thing about humans is that we are good when we are good. Otherwise, we are deadly,” he said.
Peng and his team are currently testing Direct Short-Range Communication (DSRC) technology-equipped cars that talk to each other to navigate potentially hazardous situations. DSRC hugely advances the “line of sight” detection, creating a world where blind corners are no longer blind.
DSRC will also allow for vehicles to talk to traffic lights and future infrastructure (buildings, signs, etc.), and creates a safeguard when on-board detection fails (if the sun is creating glare on the camera, for example).
I asked Peng if he is worried about these vehicles’ “hackability.” He assured me that since the communication is close range, the odds of someone hacking the vehicle and taking on-board control or accessing data are quite low.
As with any new technology, significant testing must take place before the tech is handed over to the public. However, Peng says that the mileage thresholds set by various experts to test self-driving cars before putting them on the road are unreasonably high. Instead, Peng and the Mcity team use “accelerated testing,” which evaluates the most challenging driving situations such as adverse weather conditions, blind turns, low visibility, and reacting to vehicles that disobey the laws of the road (like blowing through a stop sign).
The team has broken down more than 25 million miles of driving data to identify these trouble areas into components that can be tested or simulated repeatedly. “In this way, just 1,000 miles of testing can yield the equivalent of 300,000 to 100 million miles of real-world driving,” Mcity says on its website.
As it stands, two great concerns remain in the push for automated vehicles: ensuring safe and responsible use, and encouraging widespread adoption. To achieve the urban utopia we all dream of, packed with driverless cars, we’ll need places like Mcity to keep pushing the boundaries of how we test them.