/cdn.vox-cdn.com/uploads/chorus_image/image/63712651/IMG_5918.0.jpeg)
NuTonomy, an autonomous tech spinoff from the Massachusetts Institute of Technology, has plans to deploy "thousands" of fully self-driving taxis all over Singapore by 2019, according to company COO Doug Parker.
Already, the company has started to test its vehicle in a one-block radius in Singapore's business district, called 1 North. With the blessing of the government, nuTonomy and other autonomous tech companies will be able to expand to a nearly four-mile radius as early as this summer.
But there's still a lot of work to be done. For one, Parker told Recode during a demo in Singapore, the team is working on making the trip more comfortable for riders. During a brief test drive of the Mitsubishi i-MiEV outfitted with nuTonomy software, there were indeed a number of abrupt stops and some clumsy maneuvering.
Clumsy as it may have been, the car — which is still very much in the research and development phase — was navigating a more complex environment than other autonomous car companies typically test in. Many carmakers, in fact, have yet to even begin testing autonomous vehicles around pedestrians.
During this test drive, there were people; there was construction; there was even a fairly busy intersection.
Being able to understand traffic lights, navigate to a destination and not just detect obstacles but figure out when and how to pass them is no small feat for an autonomous vehicle. Often, that clumsiness was simply a result of the vehicle being overly careful and leaving considerable space between it and the object it was skirting.
Determining when it's safe to overtake a stopped car is a significant challenge for autonomous cars. Many of today's semi and fully autonomous systems, which depend largely on vehicles around them to determine how fast or slowly to drive, would wait patiently behind the car in front of it until it moved. But nuTonomy cars use formal logic, Parker said.
"Essentially, we establish a hierarchy of rules and break the least important," he said. "For example, one rule is 'maintain speed.' Another is 'stay in lane.' We violate the 'stay in lane' rule because maintaining speed is more important."
Grand Theft Autonomous?
At one point during the test drive, the car passed another car that was stopped on the left side of the road. To do that, it veered all the way to the right, then abruptly turned left to overtake the stopped car. Had it been a human driver taking a road test, the maneuver would have resulted in an automatic failure.
Though Parker joked that he was always telling his engineers they could reduce the "buffer space" the car gives other objects or people on the road, that buffer is a necessary safety net when deploying an autonomous car in a complex and unstructured environment.
Parker said the company is actively speaking to a number of automakers about potential partnerships and would even consider eventually licensing its technology to original equipment manufacturers. But for now, the company is focused on meeting its 2019 goal.
"Singapore is very progressive and has set technical milestones to allow cars to drive without safety drivers," he said. "There is still lots of work to do to achieve those milestones. So our focus is on technology."
This article originally appeared on Recode.net.