How close are we really to the emergence of robility?

Chris Urmson, who helped this technology to take the first steps on Google before setting up its own company Aurora , says that 50 years may pass before their ubiquitous appearance

Chris Urmson, former head of the Robo Mobile project at Google, at a conference in Sun Valley, Idaho, 2017.

Cars that drive themselves - in fact, without the supervision of a person in the passenger seat - will appear soon. Not tomorrow, not next year. But Chris Urmson, CEO of Aurora, a company that manufactures software for robots for automakers, says that in the next 5-10 years, Americans will begin to meet robomobili on the roads of many large and small cities across the country. He says that everywhere they will appear only in 30-50 years.

Before establishing Aurora, Urmson was in charge of the engineering company that designed the robot. In 2007, when he was a professor at Carnegie Mellon University, he was a member of one of the very first groups that created the robot at the DARPA Grand Challenge, where people tried to make robobiles capable of crossing the Mojave Desert. This week, the company announced a new partnership with Chrysler, concluded to study robo technology, as well as a new round of investments from Hyundai.

Recently journalist April Glazer interviewed Urmson for an If Then podcast . A transcript of their conversation, edited for clarity, is given below. They talked about the obstacles to the massive introduction of robots, whether it is safe to test the technology in real conditions, and why the cone would be an ideal form for a robility car.
')
April Glazer : Let's start with the news this week, and perhaps this will help explain what Aurora is, because it is a bit different from other robo mobile development projects. You work with Chrysler to study technology, but you do not create robobom yourself. Can you explain what is happening there, so that we can understand what you are doing?

Chris Urmson : Our goal is to spread the benefits of robotic technology safely, quickly and widely, and we reach it by working with other companies. At the beginning of work, we decided “let's do what we are good at,” and we think that this occupation is the creation of a driver. Therefore, we work with several companies to bring our technology to the market with their technology. In this case, we signed an agreement with Fiat Chrysler to bring this technology to the commercial vehicle market.

But you don't make cars?

No, we do not make cars. Fiat Chrysler is doing this, right? They do it very well. They have been doing this for 100 years, they understand what is needed for this, and they do an excellent job.

Looking back when you started working on this technology 15 years ago, what inspired you to create robots? Was it the idea that the machines are bad and we need to re-invent them somehow, or did you just get caught up with this cool technology? Why robomobili?

Personally, I worked in a project with NASA as one of the team members, we were in the Atacama Desert in Chile, in this amazing place, and the robot I helped to do was moving at a slow pedestrian speed, 15 centimeters per second, it was an experiment and therefore it did not work periodically. My dissertation leader arrived and informed me that the DARPA Grand Challenge was being held, where I had to drive through the desert at a speed of 80 km per hour, and I thought that this sounds cool and a very interesting task. In the end, it sucked me up, and here I am, about 15-16 years later.

That is, it was a very cool idea, and then you began to realize that it has a use.

Yes it is. In those days, the competition was sponsored by the Ministry of Defense, and we thought about how we could save the young people working in supplies, at that time it was mainly about Iraq and Afghanistan. Then, while we were preparing first for those competitions, and then for city ​​races , I began to realize that this technology has amazing applications on the road. Every year, 40,000 Americans die on the roads, and 1.3 million people around the world, with the majority of these accidents, 95%, due to human error, so the technology that we develop will eventually reduce this number to zero This is an amazing opportunity.

Then, if you think about the number of people unable to move like you, or me, have they lost the right to drive a car, or have drunk, and should not drive, or have sight problems, or something else is limited in possibilities, giving these people access is also an amazing opportunity.

I hope that this does not worsen the road situation, but we will return to this issue. As I understand it, cars will use computer vision and AI to drive and avoid obstacles, but what about navigation. I hope they will not rely on Google Maps. How will navigation work?

Navigation, I think, will work the way it does today when you get into the car and decide where you want to go. The mapping technology will be slightly different from Google Maps, since these maps are designed for people, and our map uses a vehicle and other things are important to it.

Will we have to create new maps for each city where such cars will work?

I suppose so. There is one layer designed for people, where it will be shown where the nearest grocery is located, where you are heading so you can understand it. And the cars themselves will use a different map on which those things that are important for driving will be recorded, with a very high resolution. Where are the traffic lights, where are the lanes, who is inferior to whom at this intersection.

I read that you think that robomobils will slowly take to the streets in 10 years, but before their widespread use it will take another 30-50 years, well, or a very long time. Why do you think such delays are expected in this case?

I think in the next five years we will see a slight development of the situation. Several hundred or several thousand cars will appear. This, in the Silicon Valley jargon, is the moment of transition from zero to one, that is, proof of the technology's performance, understanding how customers will want to use it, self-beliefs — that is, society — that the technology is safe enough, that we trust cars on the road, and this will be the first phase. Having passed this phase, we will come to their conclusion on the market on a commercial scale. For us, as a company, these partnerships are very important, we are working to integrate this technology into cars, to reduce costs, and to start producing and using them the way we use cars today.

Apparently, huge investments from other players, and, probably, even from the state will be required. You know that cars have become so common because the state has invested in the development of the road network. What investments will be required from cities to work in concert with roboMobiles?

I think that one of the most interesting moments is that we will not need huge investments from the state in the short term. When cars began to appear, we did not sharply build a bunch of roads, hoping that cars would appear on them, or even just asphalt roads. We had paths, trodden by people, horses, carts, bicycles, and then, when the car came, we realized that it was inconvenient to drive through the mud, and if there were less dirt on the roads, it would be easier for us to drive between cities, which led to the appearance of paved roads, and eventually to the system of intercity highways.

I think the same will happen with automatic cars. In fact, it will be easy and simple to drive around the working roads, on which it is easy for people to drive, so now all we need is to make them a little better for people. And then, when this technology really begins to unfold actively, we can think about what we have learned, how to make the roads safer, gradually more efficient, and I think that local authorities, state authorities and federal authorities will invest in these parts of the infrastructure. .

I recently imagined that robomobilam will need some kind of fence to exclude pedestrians and cyclists. And you say that this infrastructure is not required at the very beginning?

I think so, and, to be precise, this is one of the ways to solve this problem. If we wanted to spend a lot of money on infrastructure, this would speed up the emergence of a certain class of robots. They would not need to be so smart, because, yes, if you put fences around them, limit the presence of people and other cars, the problem is greatly alleviated. Such systems are already today. If you look at Heathrow Airport in London, there is a system of transporting people from the terminal to the parking lot, which just drives along a road fenced by a concrete barrier.

Our approach, and I have been thinking about this for quite a long time, that it would be too much to ask the cities to invest huge money in building special roads for vehicles that are not yet existing. In fact, we need to take the technology and adapt it to the way we live and work today, so that it works on the existing roads, because otherwise this technology will not appear.

This will not happen, and some people say that, in general, cars harm our lives, because of them there are roads cutting through cities, increasing segregation, preventing us from living at the expense of local and renewable resources. What do you think about the robo and this problem? Will they continue to isolate and scatter us?

I hope not. This is definitely not what I would like to see on my gravestone. But if you think about it, my view on things is much more optimistic. The tremendous amount of time people spend in cars in today's cities actually requires close study. I heard statistics on which in San Francisco people spend 30% of their time searching for parking. I heard about more disturbing statistics that 80% of traffic in Paris is made up of people looking for parking.

So imagine that you have a robotic car that takes you to the right place, you jump out of it, and he goes further, picks up another person, and brings him where he wants. And we suddenly get rid of a huge cluster of cars in the city. If you look at the plan of today's cities, then 30-40% of the space is allocated to the roads and parking. Therefore, if you have automated cars, as a service for transporting people, be it private or public transport, you do not need to allocate as much space for parking. This place can be used both for parking and for housing, for residential and business construction. There is a great opportunity to take the center of our city [San Francisco / approx. trans.], where is full of these urine-stinking concrete monsters, and turn it into something more interesting.

It turns out that not everyone will necessarily buy their own mobile when they begin to appear in large quantities.

For the centers of large cities, I definitely think that this technology will be a platform for sharing the machines in which people will sit down and then get out of them. This is a more convenient option for buses or taxis.

So, as a result, management programs will have to make decisions about whose safety to put in the first place, people in the car or people outside the car, and I understand that this is a continuum that may vary depending on the context. What do you think about that?

This is a public question, and I think that in the end some norms will emerge that society will work out. I think this is a bit different from owning your own car. If I buy a car, I want it to protect me. If these cars work as public transport, within the framework of a shared road network, I think we need to think seriously about other people on the road, since they will not necessarily use this machine.

My personal opinion is that we basically need to think about vulnerable road users, cyclists and pedestrians, put them in priority, and only then about other cars traveling nearby, since they are in hard protective shells, and I think that if think in such a way that this problem can somehow be made out.

Indeed, people in cars are protected by more people outside cars. How to balance the need to test these machines in the real world with the real danger of this test? Phoenix [the city where a pedestrian was crushed to death by one of the Uber unmanned vehicles] was such a place to test, but it scares me that we need to do this, but at the same time these tests will take place where we live.

I think the whole thing is in the necessary restrictions and process. We develop technology and test it on the roads. This is a necessary part of training, as with people who first get a student's driver's license [in the USA / approx. transl.], and then go on the road, and they impose restrictions. The same is true for this technology. This is the only way to collect the data we need in order to be sure that the technology we create will work, and for us it is a matter of creating well-trained and competent drivers. Our drivers observe for many weeks, then work in the passenger seat, studying technology, and then sit in the driver's seat.

In our case, they really are employees of the company. I think this helps ensure that they are more interested in results than contract workers who others can use. So this is a matter of education, and then a culture, where safety is put in the first place. It is no coincidence that when I mentioned the company's mission, to extend the benefits of robotic technology safely, quickly and widely, security comes first.

Is there any particular type of machine, or form of machine, best suited for robomobiles? The smaller the better, or vice versa, is it better to use vans?

The driver we create will work with all of them. How can I drive my car, or rent a car and drive it, I think this is a very important feature of the system we are creating. Yes, I think that there is some kind of optimum robotic car, it probably looks like a cone so that you can see everything that happens around it.

This is what I would like to know. Like in an ideal world, would it be a transparent cone?

Well, if you already create a car that will only drive itself, and I think this is not a particularly interesting car, then yes, it will probably look like a cone so that you can see the situation right next to it, but in practice our The company does not create technology for the sake of technology. We create technology that can help and serve people, and for us it should be a car with a comfortable cabin where people can rest or work along the road, or it will be a car with several seats so that people can ride together creating a more useful public system. transport, or it can be a delivery van, or an electric wagon transporting goods between warehouses. So it's all about the purpose of the machine.

This question may be obvious or annoying, but it bothers me a lot. Where does this obsession with robots come from? Is it possible to make improved buses, or trains, or other transport carrying many people? Machines are in many ways quite uncomfortable.

Yes, the question is reasonable, how do we transport people safer and more efficiently? Buses are good, trains are good when there is some concentration in the places of departure and arrival, but it turns out that people usually do not live that way. We are moving from our home to the office, and exactly one person wants to make this route. From my area to my office, there may be half a dozen people who want to go there, and I think the whole thing is finding the right platform, the right machine that optimizes the number of trips so that it is efficient.

When you have a bus for 50 people, and at 2 am there is only a driver and one passenger, it is terribly inefficient. And one of the problems of route planning is where this route is located, where the bus should go. The process is quite complicated. It is also suboptimal for almost everyone, but if we had automatic cars for 4-5 people who can go from one point to another, pick up these people, and deliver somewhere where they will not be far from, it will be quite an interesting transportation option, and with this technology, I think it can be implemented effectively in terms of cost.

You've been doing this for so long. What still fascinates you?

I think almost everything. Yes, as you say, I have been doing this for almost 15 years. It was incredible to watch technological developments. I am, in fact, an engineer, I adore all this, and seeing what our team can do today, comparing it with what we could do then, one wonders how incredible it is. We are digging further, exploring the possibilities of this technology, and transportation is a thing fundamental to everything. The way we move, how we transport goods and food that we consume, how they get to us.

The opportunity to touch this, to improve the infrastructure in our country and in the world in a fundamental way, is very interesting. And when I manage to talk with people who are affected by this technology, to transfer them from theoretical understanding to an idea of ​​how this can change their lives, I consider these points to be special, and it seems to me that we have an incredible opportunity, but also an incredible responsibility to people working in this field in promoting this technology in a safe and thoughtful way.