When a 49-year-old woman died on March 18, 2018, after getting struck by an autonomous Uber test vehicle in Tempe, Arizona, Bastien Beauchamp was inspired to create a product that would prevent more self-driving cars from hitting pedestrians and, in turn, save lives.
More than two years later, as part of UC Berkeley SkyDeck’s startup accelerator program, Beauchamp and his company, !Important, have developed technology that he calls “the digital version of a seat belt.”
The software uses smartphone location data and machine learning to calculate distances between pedestrians, drivers and self-driving cars and to alert all of them prior to a collision.
“Pedestrian fatalities are a significant problem,” said Beauchamp. “Our technology could save one million people over the next decade, if adopted.”
According to the World Health Organization, over 1.3 million people die from traffic collisions every year, and more than half of those who die are pedestrians, cyclists and motorcyclists.
While autonomous vehicles are statistically less likely to crash than human-driven ones, according to the American Automobile Association nearly 75% of Americans still feel unsafe on the road with self-driving cars, due, in part, to blind spots that currently exist in computer vision and object detection.
That’s where Beauchamp hopes !Important can help. The app will go out of beta mode and be available for download Sept. 1.
The technology works by analyzing the GPS coordinates of pedestrians, cyclists and drivers who have downloaded the app and then comparing it with their speeds and directions of travel. From there, an artificial intelligence (AI) algorithm processes the anonymous, encrypted data to decide if there’s a risk of a serious collision.
If there is, the app then sends an alert to the phones and vehicle systems involved. In some cases, with the right implementation, the app also triggers automatic brakes in both self-driving and people-driven vehicles to prevent potentially fatal collisions.
To confirm that a pedestrian is detected, the oncoming vehicle’s headlights flash, and the pedestrian’s phone vibrates.
This technology can counteract the unpredictability of pedestrians who may abruptly enter crosswalks or streets and cyclists who switch lanes without looking, Beauchamp said. The app could also generate crash data for both drivers and autonomous vehicles to utilize by mapping out areas where collisions with pedestrians have occurred in the past.
“There is no new hardware to install,” said Beauchamp, who is originally from Montreal, Canada. “It’s a software solution for smartphones and vehicles and can be installed very quickly with one update, and become universal.”
Beauchamp said Berkeley SkyDeck helped his company gain connections in the startup and AI industries. Since joining SkyDeck in fall 2019, Beauchamp as hired a team of 15 staff that include Berkeley students and faculty adviser and School of Information lecturer Uri Schonfeld, who heads the company’s AI research.
“With autonomous cars, there will be a transition period where the artificial intelligence needs to be trained right,” said Beauchamp, who has previously founded 10 startups. “But with our technology, we can save lives now, and SkyDeck continues to support our mission.”
The company has offices in Ann Arbor, Michigan — close to many auto companies — and downtown Berkeley. Beauchamp said he is in conversations with major cellphone carriers and car manufacturers to provide testing and validation of their technology. He hopes the app will soon be installed in vehicles and smartphones throughout the country.
That process, Beauchamp admits, may take some time.
“Our story might look like the original seatbelt in 1959,” he said. “It took many years for automakers to implement it and the government to regulate it. We are in this for the long run.”