Cofounder of Otto
VP Engineering at Nvidia
The Écoles des Mines and INRIA got interested (in self-driving vehicle technology) well before Google.
Claire Delaunay got into the world of robots and autonomous vehicles almost by chance. After completing her studies in Management Informatics, it was during a training course on the island of Réunion that she discovered a passion for robotics.
On the island, Claire was supposed to be working on tasks linked to her studies. She tells us: "I used to finish what I had to do as quickly as possible so I could work on the scientific aspects. I was able to work with the hardware that the vulcanologists were using to monitor the volcano. And I said to myself: 'This is what I really want to do'." Having decided to change direction and get into robotics, the young IT graduate did not beat about the bush. She reveals: "I built my own robot and went along to a specialist conference at my own expense to demonstrate it." Her bold approach paid off. Claire was offered a job at a startup called Intempora, a spinoff from the École des Mines (a prestigious Engineering college) in Paris. "At Intempora, they were already working on software for autonomous cars. The École des Mines and INRIA (the French Institute for Research in Computer Science and Automation) had got interested in the subject well before Google," she points out. Claire quickly realized that "the self-driving vehicle is technologically feasible but very difficult when it comes to deployment and finding a business model."
A twist of fate then led her to follow the man who had become her husband to the United States. She felt certain she would be able to obtain a work permit there but was unsure what job she would be doing. She explains: "That was back in 2009. Robotics wasn't as widely developed then." So she worked as a consultant, at the same time setting up a startup back in France. "And then, just for fun, I launched a project on Kickstarter, called Botiful." There again, things did not go quite as expected because "ARMINES, the institute in charge of monetizing technology on behalf of the École des Mines, refused to licence the product we had developed, and meanwhile the crowdfunding drive for Botiful was in full swing." Soon after, Claire was contacted by Google. That's when she first met Anthony Lewandowski, who was then working at Google Car. She recalls: "We realized that we had the same approach to the technology, the same way of thinking, the same – iterative – way of developing ideas, the same desire to move fast and face up to reality at a very early stage so as to find out whether the product actually worked."
Meanwhile Google was developing its robotics capability. The Internet giant started to buy up startups in the robotics field, one after the other, and set up a separate division, called Replicant, headed by Android founder Andy Rubin. "He left very quickly, without welding the different teams together, which didn't send out a very good signal to the industry", Claire remembers. So, three years after first meeting Anthony Lewandowski, she followed him to become one of the founders of Otto, alongside Lior Ron and Don Burnette. A dozen Google employees joined them.
The self-driving vehicle project
"One of the things that excited me most about Otto was that the business model was aligned with the technology", says Claire Delaunay. And the business model is of course key. "We developed a kit to enable a truck to run on the highway – exclusively on the highway – in autonomous mode", she explains. This is one way to avoid the risk of accidents that driverless cars face in cities, where, Claire points out: "There isn't any special infrastructure." The Otto co-founder is convinced that with this approach, autonomous trucks will become common place on the roads before self-driving cars. She explains that "it's perfectly possible for US state authorities to close off particular highway lanes during the night for the exclusive use of self-driving trucks.They would thus be sure to get traffic on those highways and could set up a special payment system. There would be a zero risk of accidents because cars wouldn't be allowed to run in those lanes."
Otto has also enjoyed favorable timing. Claire explains: "Autonomous vehicles are enjoying a media boost right now, it's a hot topic, everyone's talking about it. This has had two kinds of effect. Some people wanted to be part of it for fear of missing the boat, while others reckoned that the market is so significant that they had better invest in this field even though they hadn't grasped the full complexity of the business." In this situation, it hasn't been too hard to raise funds. "One day, a Venture Capital firm was so interested in what we were doing that they simply stuck a check for $10 million on the door of the building in Palo Alto where we were working", Claire tells us, pointing out, however: "But of course there are special conditions attached to the acceptance of that sort of amount, and that’s the real difficulty. Money doesn't come free of charge." Accepting the funds means allowing the investor to have a say in the business or even "giving him the power of life and death over your startup. Except when there's competition around and you're then in a better position to negotiate the terms and conditions."
On the highway to autonomous transportation
So with the funding assured, Claire Delaunay was able to concentrate on the many technical and logistical requirements of the project, with the pressure and heavy work-load that they imply. "The difficult thing with robotics is that you have software, electronics, hardware and they must all be integrated. You have to test, test and test again." Nor is it all that easy to drive an autonomous truck. Claire tells us: "We hired our very first driver throughCraigslist (a popular website where service providers can advertise), and he was quite surprised when he saw the engine and everything that we'd added in, but he soon became enthusiastic about the project." Another inconvenient aspect is the width of the vehicle. "The margin of error for staying inside the lane is 10 – 15 centimeters on each side so you need to have perfect control and do better than the human drivers who often cross the lane lines." These are considerations that might well worry other startups in this field, especially when you realize that Otto was taken over in summer 2016 by Uber, followed by a recent announcement that the company would henceforth focus exclusively on self-driving cars.
As Claire Delaunay and Anthony Lewandowski had intended, Otto’s smart truck was soon ready for a road-test in realistic conditions in the United States. Claire recalls: "When we made our 132-mile (212.433 km) road run from Colorado Springs to Fort Collins in October 2016, we were monitored for two and a half hours by the National Highway Traffic Safety Administration (NHTSA) and they concluded that the autonomous truck worked as well as, or even better than, a truck with a human driver." However, as this technology was designed exclusively for use on major highways, the basic idea was not to do away with drivers altogether but to allow them sufficient time to rest and sleep while on the go. This would appear to be a much more efficient approach for heavy goods vehicles (HGVs) than the current one, with its legal restrictions on the number of hours HGV drivers may spend behind the wheel without taking a break. Another goal of this technology is also to make motorways safer. Today an estimated 90% of all road accidents are due to human error. In fact, driving trucks remains one of the most dangerous jobs in the United States. According to the latest statistics, 786 HGV drivers died while doing their jobs in 2017, 17.3% more than in 2011.
If this is to happen, the urban infrastructure and the regulatory framework will be vital factors. Claire Delaunay underlines: "An autonomous vehicle is just a piece of technology. The real challenge is to work out how to integrate these vehicles into our economy and our way of life." For instance: "Let's suppose that everybody had a self-driving car. What are they going to do with them when they're not out on the road? Drive out of town to park them? And then how would they go and get them back? This is likely to lead to an increase in traffic and more traffic jams." To ensure successful adoption of autonomous vehicle technology, urban layout and road systems will need to be redesigned. The United States in particular, where basic infrastructure often leaves much to be desired, "has a lot to lose. We see that in China, when they test out self-driving trucks they get properly organized. They're not content to just close a few lanes during the night, they actually build new lanes."
Dedicated lanes for autonomous vehicles
So what will road mobility look like in ten years' time? Claire Delaunay predicts that "there'll be a mix of shuttle services, autonomous buses running very slowly that will transport merchandise on closed routes, and probably also trucks, provided that work is done on the infrastructure because a truck is a large vehicle, it goes fast and it takes time to stop." So does that mean that the US will concentrate on the main highways? Domino's Pizza and the self-service electric scooter provider Bird have each offered to finance road improvements, Domino's to fix potholes so that the pizzas arrive at their destination in perfect condition and Bird to pay for cycle lanes to help its customers get around more easily. On this basis, it is quite conceivable that "a freight company which earns 80% of its turnover from transporting goods on a given highway might be willing to rebuild the infrastructure in order to be able to deploy autonomous trucks and make costs savings", Claire argues.
And when can we expect to see flying cars? "In terms of the traffic environment, this would be much more straightforward than self-driving vehicles because there’s much less air traffic, no traffic lights… and no drunk drivers! The only problem you have with a flying car is the battery. As soon as we manage to alter the battery chemistry, the problem will be solved and we'll have flying cars in the air. Of course, there are also regulatory issues regarding air space, as with helicopters, but we have plenty of aeroplanes flying around nowadays and it works just fine. Very often the pilot handles the take-off and landing because those are always a bit tricky but all planes also have automatic pilot systems", Claire points out. Moreover, vertical take-off radically reduces the risks. "Any time you're running on the ground there's a danger because things are unpredictable. Anything can happen, an object or a living creature can appear in your path, or a tyre might burst. But when you make a vertical start, all you have to do is push a button and everything is already automated."
The futuristic city
And what about the more distant horizon – 20 or 30 years hence? Claire forecasts that "transport will be much more efficient, not necessarily due to self-driving cars but thanks to modular trains for instance – autonomous shuttle services between points with high public transit requirements, combined with solutions for the 'last mile'. For goods delivery we'll probably have autonomous shuttles that will be parked at a given location from such-and-such a time to such-and-such a time with all the packages inside and you'll just have to go up to the vehicle and identify yourself with your smartphone so as to pick up your order. It's perhaps hard to imagine what it will be like, but what it won't be like is everybody driving around in autonomous cars. It'll be much more complex and fragmented than that. It'll be a mix of different technologies. There'll probably be self-driving cars for the longer distances, running on the same principle as the trucks, on parallel highways, which might also be air corridors. And once we become more efficient with transport, that will spread the traffic out and open up possibilities for new ways of getting around."
To help these predictions become reality, Claire Delaunay hopes that the technology become accessible to as many people as possible. She underlines: "What I would love to see is a situation where technology doesn't belong to a single company and there's a way to ensure that inventions become open to the public. If someone manages to develop the safest autonomous vehicle technology, for example, it shouldn't become the property of a single company. Why not ensure that as many people as possible can benefit from it? That's why I joined Nvidia, I'm working there on developing a framework for robotics. The idea is not to produce the best robots in the world but to provide startups and companies that have good ideas with the right tools for developing and maintaining robotized systems so as to drive their business forward." Let's hope that the legislators and transport firms play their part as well so that, going forward, as many people as possible will be able to benefit from the advances being made in the field of mobility.