Would you step into a bus that has no driver? It may sound futuristic, but it has in fact become a reality that is gaining traction internationally. EasyMile is one of the leading companies that is taking automation to new heights, designing software for driverless shuttles that are ideal for transporting passengers in various controlled environments, such as private campuses or industrial sites. With AVs, new challenges arise and create needs for new business models, job opportunities, and the technical skills to maintain safety standards.
We were curious to find out more. Where is this technology going to go? How safe can it be? Is automation the next frontier in transportation? To learn more about the challenges and opportunities facing EasyMile, we met with CEO Gilbert Gagnaire to discuss the possibilities he sees for automation in transportation.
How did the concept for EasyMile originate and what was the inspiration that got it going?
It was more of an opportunity than a strategy. Back in November 2013, I was doing some rock climbing in the Canary Island, I was a retired man for a few years already, and a friend of mine called me saying that I should have a look at a French company named Induct. The company was in serious financial difficulties and he asked me whether I would get involved in trying to help the company. I had a look and discovered that they were developing something conceptually similar to our EZ10. Ultimately I decided not to jump into the company and it has since been liquidated, but I found the concept interesting. It could be a game-changer for people who live in large cities, and I’ve spent my entire life in large cities, such as Singapore or Paris. So I just decided to make it happen.
What is your ultimate goal with EasyMile?
Well it’s hard to tell at this stage, and it really depends on how I interpret your question — is it EasyMile as a company or the ultimate goal for society?
I do not believe that one day we will be able to do door-to-door transportation in complex areas, in mixed traffic where AVs will be mixed with cars driven by human beings. It’s okay at the moment to make experimentations, but always with a safety driver. And you wouldn’t believe how hard this is to get rid of it — it’s incredibly difficult because without a safety driver, your entire system has to be fail-safe.
I’ll give you an example: if you drive on a highway and you’re in the left lane, you drive fast. Suddenly there’s a mechanical break down — what do you expect the vehicle to do? Stopping in the left lane is not an option — at least not a safe one. And in that case you expect human common sense to take the most appropriate decision. Can I cross the lane? Go towards the right and stop in the emergency lane? Is that possible in the context, knowing that something has broken down in the car? If there is no one on board, the engineer in charge of programming the system has to prepare all the decisions for any kind of scenario that potentially can occur, and it’s simply impossible because the combinations are too wide. So first-things-first, door-to-door without a safety driver is technically much harder than you would believe.
Second, would it be good for the society? I would say it depends on where. In a city like Los Angeles, which is very widespread, why not? In a city like Paris, which is pretty dense, it would be a massive traffic jam. If you just take the commuter train crossing Paris from East to West, it’s called RER A, this is the red lane, the capacity of this single commuter train is 50,000 passengers per hour and per direction in peak hours. So in the morning, in both directions, you have 50,000 passengers transiting using that train every hour. That sounds a little bit abstract, but now let’s imagine that every single passenger wants to shift their commute to a passenger car. On average there is 1.2 passengers per car in Europe. You would need a highway with 18 lanes in both directions to absorb that capacity. So one commuter train is the equivalent of a 36 lanes highway, 18 lanes in both directions. I mean in terms of space, can you imagine that in Paris? And guess what, there’s not just one commuter train. Plus the metro, plus the tramway, plus the buses. If everyone in Paris started using passenger cars, everyone would get stuck. So, depending on the context and the density of the city, if the city’s dense you need mass transit. Tokyo without the train, without the metro, it would be a nightmare. L.A. is okay, especially since there’s almost no public transport in L.A.
So there are really two aspects: technically speaking, I think it’s much harder than what people believe. And the door-to-door, without a safety driver, is not likely to happen for decades. And not only because of legal framework reasons, functional safety is impossible to enforce in complex environments.
Your software enables automation for various transportation platforms, such as urban, rural, private. Can you talk a little bit about the possibilities that you see for the technology you’re working on now? Like where do you see a need for autonomous vehicles?
There are needs everywhere, but there are only a few use cases that I believe can be addressed in the foreseeable future. And by foreseeable future, I mean three to five years. First, you want to get rid of the safety driver, which is a requirement, otherwise there is no possible business model… Why would you buy a very expensive autonomous vehicle if you still need a driver? That doesn’t work. So you need to get rid of it. To do that you need a reasonably well controlled environment. You also need to figure out how to have a positive business model. And you need a suitable legal framework.
Typically airports and industrial sites, such as power stations or car factories, are probably the ideal sites to start a real business case. The first reason is those environments are semi-controlled. It’s not accessible to the public — you cannot penetrate a power plant without a pass. Second, safety is already the norm. People are familiar with safety procedures: they use rules, they don’t play like kids, they know that what they do is dangerous. Third, in terms of economics, a car factory or a power station usually operate 24/7. If you look at the labor cost, you typically need four to five full-time drivers per vehicle to provide a 24/7 service. That means that the operational expenses are very high. So if you can get rid of the safety driver, you save a lot of money that you can invest into the vehicles and the fleet management solution. The last reason is that there is already an existing legal framework for robots operating at industrial sites. Automated guided vehicles have been around since probably 15 years, so the norms and the framework is already there—we know how to satisfy the vehicles, we know what we have to comply with, so we are not reinventing the wheel. I believe that the first use cases where AVs will pick up are likely to be airports and industrial sites or at least gated communities like retirement cities in the US—these kind of areas where it’s not easily accessible to the public. Because it’s simpler.
That makes sense. With the advancement of automation, how do you see the passenger economy developing as a result? With all of this new time that people are going to have — no longer driving themselves places — is there sort of a new economy developing out of that? Do you have any thoughts on that?
No, we have no thoughts about that — the main thought we have about economy is that for first and last line vehicles, like the EZ10, we should probably look at very different business models than the usual pay-per-use that is in force at the moment. It’s going to be very hard to force people to buy a ticket and to check that they have actually bought a ticket — and the cost of enforcement might exceed the money that you would collect with a ticket. So, we have to think about different ways to provide this mobility.
For example, imagine that you’re a property developer and you want to create a new residential area. If you build a traditional residential area or a business park, you will have to provide roads, road junctions, roundabouts, and car parks and very likely the proportion of land that will be dedicated to cars altogether will be close to 40%. It’s quite funny when you look at a satellite photo of a business park you realize that half of it is made of car parks. But, if you propose to the property developer to use that land to build more housing or more office space instead of car parks or roundabouts or blah blah blah, they will probably save a lot of money. And it’s not just money saving, but it’s that you will invest in something that produces income, such as office space. And using this saving, they might be able to finance the mobility using AVs on the site.
Another idea for a city area is that if we start implementing AVs in a district, very likely for free, the price of properties will go up. So the people who are living there will feel that they are richer and maybe they will be willing to pay a bit of money to keep this service running. These are just ideas to try to figure out a way to provide the mobility for free.
What have been some of your biggest obstacles so far?
The main one is that what we do is very difficult as long as you want to keep the safety level at the top. When you look again at the facts: in the US there are 35,000 people killed on the roads per year. The number is brutal. It’s awful and you want to get rid of it. The problem is that most of the time people don’t have enough perspective. And the perspective is that every year in the US 3,000 billion miles are traveled. And so, if you start dividing 35,000 dead people by 3,000 billion miles traveled, you realize that human drivers are incredibly safe. There’s 0.01 death toll per million of miles traveled — that’s not a lot! That’s way too much, right, but it’s an incredibly difficult target to match if you want to accomplish the same with a robot. And this is what we are talking about. It will never be acceptable for society that a robot is less safe than a human driver. That means that what we have to accomplish is a death toll of less than 0.01 per million of miles traveled — essentially, less than one fatal accident per billion of miles traveled. That’s the benchmark, and it’s incredibly difficult. That’s why, if you ask me whether the door-to-door will be feasible or when, I don’t know when it will be feasible… but not before decades. Because the benchmark of the human driver is already so high — and we have to do better.
It seems that another obstacle would be getting people to trust that a robot could get them from A to B. To trust in the technology that you would get into a vehicle that doesn’t have a driver—that’s a big leap of faith…
What we try to do is make the ride very smooth and comfortable. It’s very soft inside, you are not shaken at all, it rides very smoothly so as to make people believe that they are safe. But feeling safe is a question of perception that’s not necessarily the reality. When you drive at a hundred kilometers an hour on a highway following the car ahead with just two meters headway, you feel safe. Because nothing happens. Actually, you are not safe at all. If anything goes wrong with the car you are following, you are dead. It’s just that you don’t feel the danger. So we try our best to be good on both sides. We work hard on the functional safety to reach the benchmark: less than 1 dead people every billion of miles. And in parallel for user acceptance, we try to make the ride very soft, smooth, and comfortable to help people trust into the technology. But perception is not reality. We have to work on both sides.
Yes, of course. What has the reception been like so far? You have programs already happening in the United States and around the world…
So far on our side I would say that people are or seem to be happy. We have never experienced any incidents. But again, what we are doing at the moment is experimentations. It’s not a proper service. The day that people start relying on the service for their commuting, they will expect the shuttle to be on time, to be reliable, and they will want a real service. For the time being it’s more like a distraction…
It’s fun. So it’s quite easy to make them happy because they do not rely on it. If the shuttle is not there, they were not counting on it anyway. The day it becomes part of your daily life, I mean it’s like the train or it’s like the metro…
If it’s late, you’re furious.
Yeah, exactly. And we have not yet entered into this phase, because it’s not yet a public service.
So with automation, drivers will become non-essential and could lose their jobs. Are there new job opportunities you’ve seen arise with the technology?
Autonomous vehicles are not so autonomous—in the sense that the fleet has to be monitored. The vehicle does not decide by itself, ‘Oh okay, I need to fetch Mr. X.’ So there’s what we call a control and command center where all the missions are collected and then assigned. Of course it’s done automatically, but then in case of incident… for example a passenger has pressed the stop button. It’s like in an elevator, you press the call button you expect someone to talk to you, right? So the control center is there to manage any type of incident that can happen and deal with it. And then we also need people on the ground to do the cleaning of the vehicles and to check the sensors. In the morning you need to do a full check—make sure the vehicle is in good condition, otherwise safety would be compromised. If a sensor doesn’t work very well because there is mud on the glass, the safety is compromised. So there will be many more procedures to check the condition of the vehicle.
The purpose of these vehicles is not to get rid of bus drivers and metro drivers, it’s more to try to convince people who currently commute with their car to take a shuttle to the next train or metro station. So I don’t believe that this will destroy jobs in the public transport or in taxi companies, but rather that people will rely less on their private cars to go to work. That’s the main goal and it’s a win-win deal. Only car makers or oil refiners might complain because we will be using less gas.
Because they are electric…
Because they are electric and if you manage to pack 300 people in a metro, you consume far less than if everyone takes his own car. It’s that simple.
Where has EasyMile had the most support and success so far? Are there places where it’s been easier to implement?
To be honest, most countries in the world are quite friendly to this technology. We have customers in Japan, Singapore, China, Norway, Finland, Sweden, Germany, France, Dubai, and the US. We are all over in the US: Minnesota, Texas, Florida, California, and Colorado. There is a real appetite for this kind of technology. It’s just that most of the time people don’t understand the real level of complexity of this job when you want to do it without the safety driver, which is the ultimate goal.
Lastly, a very broad question about the distant future… Do you think that autonomous vehicles will be the future of transportation?
Honestly, I don’t know. I don’t know how far and how fast technology will be able to pick up in this direction. I know that there are very high hopes in the deep learning techniques and AI. The problem is, for the time being, that deep learning is based on scenarios that have been submitted to a machine to learn from those scenarios. But you can never cover the entire scenario that can occur in real life. And computers currently lack the human common sense — they have absolutely no common sense. They don’t even know that they don’t know. They just react based on the data sets that have been used to train them.
And in some ways we don’t want them to know more than that, right?
That’s right! So there’s this aspect: lack of common sense and also lack of interaction with the other, with users. When you drive your car, when you see a pedestrian on the walkway, within a split-second you know—you know what’s going to happen, you know whether the guy wants to cross the road or not. If yes, through eye contact, within a split of a second you have negotiated who will go first. That’s something that I can not imagine with a computer. This interaction through eye contact. First thing, even without the eye contact, you know the intention because you have common sense. You know that the guy is on the phone, you know that he is talking to his little girl, you know that he’s standing there because he’s waiting. And secondly, if he’s really wanting to cross the street, if you have eye contact, then within less than one second you decide what’s gonna happen. You cannot just have the world based on the rules. You could say okay, if the pedestrian is not on the zebra, I run over it. Is it socially acceptable? I don’t think so.
No, people are unpredictable.
Of course. So, for a massive pick-up of autonomous driving, either we need to get rid of pedestrians and cars driven by humans, or we need a breakthrough in the technology that we have no idea about at the moment. We just have no idea how to make it. I’m not pessimistic, it’s just that I’ve seen so many times in my career high expectations suddenly dropped. And AI is a very promising area, but you cannot do everything with AI.
Well, thank you!