Skip to main content
Open this photo in gallery:

Jose Rojas, left, and Faraz Khan, both software engineers, with a hacked, all-electric Kia Soul during Self Racing Cars, an annual open-track event for autonomous vehicles, in Willows, Calif., on March 23, 2019.AARON WOJACK/The New York Times News Service

Faraz Khan’s all-electric Kia Soul is his workaday car, but on a drizzly weekend last month, he and his teammate, Jose Rojas, were hoping it would drive them to a racing milestone.

So they squeezed on safety helmets, booted up their laptops and inched toward the west track of Thunderhill Raceway in rural Willows. Poised behind the wheel at the starting line, Khan stared into a laptop balanced on his partner’s knees. “In theory, this should work,” he said. “Although it might be an embarrassment.” The two men laughed off their nerves as Rojas initiated the car’s homegrown autonomous controls.

Khan and Rojas, both software engineers, have been using evenings and weekends since last fall to hack the Kia. They are among a loose-knit community of perhaps 50 independent hobbyist-engineers who – for fun and the sake of learning – rip open vehicle dashboards and splice essential wires to turn cars into robots.

“The bar is high. You need to get a car, and then you need to really mess with it,” Khan said with another laugh. “Most people aren’t comfortable with that.”

This 2.1-mile serpentine course usually rumbles with Porsches, Mazda roadsters and heaps of American muscle. In late March, however, the paddock hummed with mild-mannered hybrid and electric cars. Several dozen gear heads of a different stripe were here for the fourth instalment of Self Racing Cars – an annual open-track event for autonomous vehicles to race against the clock.

The auto industry is in the middle of its biggest technology shift in a century, with untold fortunes to be made – eventually. But for now, the promise of self-driving cars often outpaces the reality. This competition in Northern California suggests that it might be easier to let the cars run before they can walk. The racetrack strips away all of the distractions, variables and traffic of street driving and lets the rubber meet the road.

Self Racing Cars is the brainchild of Joshua Schachter, a Silicon Valley entrepreneur and an amateur race car driver. Schachter hasn’t hacked his personal vehicle but gets a kick out of turning everyday objects, like shopping carts, Etch-a-Sketch toys and water droppers, into improbable robots.

A month before the event, he explained that Self Racing Cars was an extension of his interests. “I want to see other people’s solutions to a problem,” he said. “And to learn about problems that I didn’t know existed.”

Khan and Rojas named their project OpenCaret, and they use laptops and Arduino boards – an open-source microcontroller platform – to trick the car into working as if it were being driven by a person. Khan mounted a $50 camera on the dashboard to provide vision.

Two weeks before the race, in an abandoned Sears parking lot near Khan’s home, they allowed me to “drive” from the passenger’s seat using a Microsoft Xbox controller. It was a video game come to life. Khan’s foot hovered over the brake pedal in case my thumb steered off course.

Schachter knows that automated driving is a dangerous challenge on real streets, even for the big names. His event boils it down: Who can log the fastest lap?

To transform the track into an engineering laboratory, Schachter enlisted the help of experienced racing officials. “When Joshua asked me to be a part of it, I didn’t hesitate,” said Terry Geiser, who over decades has overseen millions of miles of racing at Northern California’s famed motor sports venues, like Laguna Seca and Sonoma Raceway. He’s the track co-ordinator and safety engineer for Self Racing Cars.

“In 1950, you couldn’t buy a hot rod and take it to a drag strip,” said Geiser, a retired management consultant. “You had to build it yourself.”

He explained the camaraderie shared by old-school car guys – his generation favoured British roadsters in the 1960s – who compared notes about better ways to tune an engine or modify an exhaust system. “This event, with its hackers, tinkerers and engineers, is just like that for me,” he said.

On their first run, Khan’s Kia Soul barely made it through the first of Thunderhill’s twists and turns before the steering wheel swung all the way right, all the way left and back, in a loop. He pulled the plug and drove the Kia back to the paddock.

Geiser’s voice then boomed over the loudspeaker: “Humans, you’re up next.”

During long pauses in track activity, while coders squash bugs, the course is opened to vehicles driven by red-blooded drivers. Schachter invited me for a ride in his bright-orange 1992 racing-spec Mazda Miata. It’s entirely manual – no anti-lock brakes and definitely no Xbox controllers.

Schachter precisely manoeuvred around the course, flicking the gear-shifter and slamming the brakes before corners. The tail whipped around, and we were on the straightaway approaching 100 mph. How could software and sensors ever compete against a driver’s eyes, hands and firing synapses?

Three years ago, at the first Self Racing Cars event, Audi showed an autonomous sports-sedan prototype. Chris Urmson, a former chief technology officer for Google’s self-driving car company, which eventually became Waymo, was on hand. Other industry luminaries mingled.

By contrast, this year’s competition was more like a home-brew computer club. Only five teams, mostly small independent projects, competed.

A Tesla Model X joined in. It wasn’t an official entrant but belongs to Andrew Navarro, one of the half-dozen members of Team Soulless, another competitor. Navarro is a systems engineer and a former Army National Guard aviation officer. He now works at Polysync, a company in Portland, Oregon, that makes hardware and software to help companies digitally control steering, brake and throttle controls for development and testing. Polysync lent its silver Kia Niro hybrid to his team.

Navarro is a big fan of Tesla’s Autopilot technology. “As far as vehicles that have lane-keeping and adaptive cruise control, it’s the best around,” he said. But he’s quick to point out that a driver must monitor the Autopilot at all times.

For the first couple of turns, the Model X SUV wouldn’t let Navarro engage Autopilot. The vision system, assisted by radar and ultrasonics, is not designed for a track, so the vehicle couldn’t find the lane markings and other highway objects it expected to see. Autopilot finally engaged, but the vehicle got too close to the track’s edge and quit.

If Tesla’s Autopilot couldn’t make it, there wasn’t much hope for Khan and Rojas from OpenCaret, the students from Team Soulless, AEye (an artificial perception outfit) or Qibus (a three-week-old company creating remote vehicle-control systems).

But lo and behold, a red Lexus CT 200h hybrid hatchback fielded by Point One Navigation was zipping around the track on its own. (All teams must put an engineer behind the wheel ready to take over in a split second.)

Aaron Nathan, chief executive for Point One, feels the pain of his competitors. Three years ago, Point One attended as a new company with four people. Now it has 14 employees and is marketing its software to major automakers.

“In the beginning, we came just to meet people and talk shop,” Nathan said. “It was much more like a track day, so people were not reserved like at a trade show.”

Nathan and his co-founder, Bryan Galusha, both have advanced engineering degrees from Cornell. The university participated in the 2000s in the DARPA challenge, an autonomous-vehicle prize competition sponsored by the Defense Department.

“I’m very familiar with what it takes to build something that works at DARPA,” Nathan said. “I feel like Self Racing Cars is the closest thing we have to that now.”

Two years ago, the company managed a lap in 3 minutes 38 seconds. After sitting out last year, Point One returned last month with a performance of 2 minutes 6 seconds, an event record. That was only 20 seconds slower than Galusha’s fastest lap in the Lexus. The company’s navigation algorithms performed reliably for 10 nearly identical laps over the weekend.

Point One was the only team this year to complete an autonomous lap. But the advantage of its technology would disappear with multiple cars racing at once. Schachter is considering that proposition.

“The essence of racing is dealing with a changing situation,” he said. “I was thinking next year about throwing a tire into the track.”

This content appears as provided to The Globe by the originating wire service. It has not been edited by Globe staff.

Follow related authors and topics

Authors and topics you follow will be added to your personal news feed in Following.

Interact with The Globe