It’s a Monday morning in sunny Singapore, and you’re getting ready for work.
Before you leave the house, you pull out your mobile phone and hail for a vehicle. Minutes later, a one-seater driverless pod pulls up to your door and you hop in. You’re headed for the train station.
On the way there, dozens of similar pods are transporting people to and fro the train station and public facilities. Occasionally, a driverless cab whizzes by. What used to be car parks are now public gardens. Traffic jams are a distant memory.
At the train station, you hop off the pod and take the train to work. You alight at your designated stop, and another pod takes you to your workplace. You’ll never need to walk or cycle if you didn’t want to.
This is the vision for urban cityscapes that self-driving car startup NuTonomy is working toward.
I sat down with James Fu, director of technology, and Cody Kamin, manager of Singapore operations at NuTonomy to find out more.
Founded by Karl Iagnemma and Emilio Frazzoli in July 2015, NuTonomy is a spin-off from the Singapore-MIT Alliance for Research and Technology (SMART).
It is primarily a software company developing a full technology stack needed to power a driverless vehicle. This fits in with its larger mission to improve the transportation system in cities worldwide.
Although NuTonomy is based in Boston, two-thirds of its employees work in Singapore. The company has over 100 employees across both countries.
NuTonomy was acquired by automotive supplier Delphi in October 2017 for $450 million.
A new form of autonomy
NuTonomy is a portmanteau of the Greek letter nu (v) and and the word ‘autonomy’.
Fu explains that in robotics, an autonomous system is one that makes decisions on its own without external input, such as a self-driving car. Nu (v) is the symbol for velocity, and also sounds like the English word ‘new’.
Put together, NuTonomy represents a new form of autonomy that the company is working towards.
Mobility on demand
The NuTonomy team is developing what they call “mobility-on-demand”.
“This is not a car you own, this is a car that is part of a ride hailing network,” explains Kamin. “We want to shift away from this model of personal car ownership, and find ways to make the current transportation system more efficient.”
This ties in well with the Singapore government’s mission towards a car-lite society.
“Singapore doesn’t want to dedicate any more land to road,” says Kamin, “They want to figure out how to use our existing roads more efficiently to accommodate the growing population and economy.”
The team works closely with the government to turn this vision into reality.
“Hopefully in five years, you start to see autonomous vehicles operating across Singapore. Then in 10 to 20 years, there will be no need for drivers, and cities will start to look different,” says Kamin. NuTonomy wants to take its tech worldwide as well.
New towns like Punggol and Tengah will be designed with autonomous vehicles in mind, and might look different from current ones in several ways. For example, as car ownership decreases, there will be less need for car parks, and green spaces can take their place. Town areas will be developed around train stations to maximize use of the readily available public transport network.
“I think you’ll also have a lot more custom-built vehicles,” says Kamin. “Today, a lot of our vehicles look the same because people buy them to fit all their potential needs. If I want to go to the train station, I don’t need a five-seater to pick me up. I can go in a little pod. But let’s say we’re travelling with friends, or we want to go further. You’ll have a bigger shuttle to pick you up. Our vehicles can be customized to fit your specific need for a particular trip.”
Why autonomous vehicles?
Kamin believes that autonomous vehicles are a lot more efficient than taxi drivers.
“Taxi drivers make decisions to maximize their own revenue,” says Kamin. “For example, they will pick someone up in the morning and drive into the CBD. Even though there is demand outside the CBD, they don’t want to drive there because they’re not paid along the way. So they stay there until someone books a trip out, and a lot of time is wasted.”
This is in contrast to having a fleet of autonomous vehicles, where they think about how to make the fleet more efficient. “So sure, we’ll drive into the CBD. But we’ll also have a very good idea of where demand will crop up next, and send our car there,” says Kamin. “Simulations we have done show that this can serve 50 percent more trips.”
Reducing human error
While autonomous vehicles now have the option of a driver taking back manual control, Fu says that the goal is to develop a fully driverless system in the distant future.
“Most traffic accidents are due to human error,” says Fu. “Having autonomous vehicles will alleviate this issue. If we design a system that hands control back to drivers under critical situations, it creates a few problems. Firstly, how do you identify critical issues? Next, in that situation, can a human driver take over if he was not paying attention in the first place?”
As such, the team is designing a system that is fully autonomous without the need for any human intervention. “But in a situation where a car needs to fail, it should be designed to fail gracefully,” says Fu.
Fu says that NuTonomy’s greatest milestone to date is getting a car to move autonomously in a little over a year.
“We started from nothing in July 2015,” says Fu. “We had to build the software, get the cars, and start the hardware modifications.”
Within three months, the team built a functional system. By December that year, they purchased a car and started making hardware modifications “under the cloak of darkness in a public car park when the guards weren’t looking.”
In August 2016, they conducted the world’s first public pilot, making Singapore the first country in the world to have on-demand driverless taxis.
They conducted a full drive at One-North in early 2017 without any manual takeovers. The cars can now go to anywhere around One-North with a click of a button.
“Our biggest technical challenge is that road rules were not written with autonomous vehicles in mind,” says Fu.
For example, the rule to keep a safe distance from cyclists.
“What exactly is a safe distance? That’s left to the road user’s interpretation,” says Fu. “To program autonomous cars, we have to be very specific to know what that distance is.”
This is an even bigger challenge with road rules differing from city to city.
“20 to 30 years from now, we’re going to need an entirely different set of road rules,” says Kamin. “Something defined a lot more precisely, designed with autonomous vehicles in mind.”
AI systems also need to contend with chaos.
“One assumption people make is that everyone obeys road rules,” says Fu. “In reality, they don’t. You have people jaywalking. Cars make illegal U-turns. And to maintain safety, you have to break those rules too.”
“This is antithetical to how robotics have always operated,” says Kamin. “Robots follow rules. Now we have to train our robots to break rules. This is new in the robotics world.”
For this, the team is delving into intent prediction.
“The more we’re able to predict what other road users are going to do, the more smoothly and intelligently we’re able to navigate the environment, rather than being more cautious as we are today,” says Kamin.
How do they do this?
“That’s our secret sauce,” laughed Fu.
Another challenge is hiring more people, and keeping the team aligned and nimble as it grows.
“As we get bigger, there’s a natural tendency for things to fragment and get more chaotic,” says Kamin. “We have over 10 teams here, and as we got bigger, we were starting to march in 10 different directions, having sprints that started and stopped at different times. Breaking those individual cycles and bringing everyone back in line was a huge undertaking. But we see it paying off now. It’s going to continue being a challenge as we keep growing.”
For technical roles, the interview process starts with a phone screening, followed by two to three rounds of interviews with on-the-spot challenges. They keep the bar high, and look out for people who can hit the ground running.
Fu advises candidates to highlight not only the projects they’ve worked on in their day job, but the projects they do on their own.
“It speaks a lot about someone when they have their personal projects,” says Fu, “Whether it’s writing their own apps, tinkering with their cars, or even buying their own drones and making it fully autonomous. People here take a lot of initiative. When someone isn’t forced upon to do something but does it anyway, this person will fit in very naturally into NuTonomy. Everyone here is like that.”
“Before NuTonomy, everyone had their own pet projects. But this has become our pet project. And we’re all focused on this now,” says Fu.
Despite having no fixed working hours, team members sometimes lose track of time when solving interesting problems.
Kamin likens it with teenagers addicted to their computer games. “There have been many occasions where we had to tell people to stop working so late, to spend some time with their families,” says Kamin. “But they can’t. They feel very personally invested in the problem they’re solving, and they want to solve it. There’s also a sense of responsibility, when the problem they solve can help unlock another team’s progress.”
Likewise, Fu looks out for people who are intrigued by problems they’d like to solve, and are invested in solving them.
“We’re like the Avengers,” he says. “The Avengers have no working hours. They never rest.”
This post How driverless car startup NuTonomy is building the future of transport appeared first on Tech in Asia.