On the Road to Autonomous Vehicles

By Marine Guyon, Delphine Hugo, Noamie Le Menn and Léo Roblin

Driving kills. The United States Department of Transportation reported that in the US, 37,461 persons died in motor vehicles crashes in 2016 . More strikingly, the National Motor Vehicles Crash Causation Survey (NMVCCS) found that between 2005 and 2007, 94% of traffic accidents could be attributed to drivers. In other words, drivers get in trouble because they make errors. It’s normal, that’s part of human nature.

But for some – Waymo, Apple, or Uber, to name all but few – this situation is no longer acceptable. They don’t think we should go down that path. They think it’s time to change. It’s time to “improve” driving by making it autonomous. Indeed we are told that overall such vehicles should save lives, reduce traffic jams and even free up time.

Truth be told, only the very last levels of automation could lead to such benefits. Yet, there are 5 levels of automation. At the first level, the so-called Advanced Driver Assistance System (ADAS) can from time to time assist the driver in some basic tasks such as accelerating or braking. At the next stage (level 2) the car can control both steering and accelerating/braking. That’s the case when it comes to letting the car park itself for instance. However, the driver still has to monitor the environment. Level 3 is all about letting the car drives. But the human has to be ready to take back control at any time needed. Level 4 is very similar to level 3 but here the car also monitors the environment, just like a human would. Nevertheless, humans need to pay attention because they could possibly take action. At the last level (5), humans are basically passengers. The vehicle is 100% autonomous.
For the time being, only the first and the second level of automation are allowed. For instance, you could find yourself driving next to a Tesla with a level 2 autonomy. Yet, the aim is obviously to reach the 5^th level. But a hard path might have to be done since autonomous vehicles bring some concerns.

Who is responsible?

First, the development of autonomous vehicles raises legal and ethical issues. For instance, who is responsible if an autonomous car run down a pedestrian, as happened in Arizona on March 2018 with an experimental Uber car? Who’s responsible in case of an accident? Is it the driver? (that wasn’t in control of the vehicle)? The car constructor? The software developer? The sensor manufacturer? Could it even be the State or the department’s responsibility? The debate is far from being closed but one thing is almost certain: there will be a liability shift to deal with, according to Coren Hinkle, law teacher at Grenoble Ecole de Management (GEM). Moreover, how to deal with insurance issues if the driver is no more responsible for his/her car?

Those questions are only a small sample of the ones triggered by the generalization of autonomous vehicles. In fact, the whole legal system has to be adapted in order to respond to all these new considerations. For instance, the 1968 Vienna Convention on traffic regulation impeded the expansion of higher degrees of autonomous vehicles. It was amended on March 2016 in order to authorize some type of autonomous cars on the road: the ones that abide by the United Nations rules on vehicles and that can be monitored or stopped by the driver. However, full autonomous vehicles (level 5) are not allowed yet.

Legal issues concern data privacy and the protection of personal data. In fact, autonomous cars are fully connected, and thanks to onboard sensors the vehicles have access to vehicle usage information (such as real time location, speed but also personal habits, schedule…) and private data (emails, private communications, phone information, online preferences). Thus, a single self-driving car could produce up to 100GB of data per second.  Yet according to a survey led by the Automated Driving Community, 48% of US consumers are unaware that connected cars can store their personal data. Therefore, the risk to lose control over your private data is real. Some provision are being implemented to regulate those issues. In the US, for example, the Security and Privacy in Your Car Study Act of 2017 calls for transparency concerning data collecting; the drivers should have the choice wether or not to accept the data collection and the information collected should not serve marketing or advertising purposes unless the drivers express their consent.

Some ethical issues, such as the “Trolley Problem”, are also raised by the public opinion. What if the car has to choose between two lethal options? Run over a group of children or plunge into a cliff and kill his occupants ? What would be the ethical rules used in the vehicle’ software to decide the outcome? A survey published in the magazine Nature has shown that ethical decisions vary according to the country so it seems hard even impossible to think of a universal code for this ethical question. A more important interrogation would be: can we let an artificial intelligence decide for that type of questions?

Should autonomous car be as human as possible?

Now, since the automatization is already present in some cities such as Lyon, which has developed autonomous buses, the cohabitation between human drivers and autonomous vehicles seems unavoidable. However, it will take time for humans to accept the idea of giving up the wheel and trust a machine. As a matter of fact, people generally focus on the negative aspects. Every accident reinforces their fear and reflects poorly on the automatization, such as the one involving an Uber autonomous car in Tempe (Arizona) which killed a pedestrian. Humans are also worried of becoming dependent of this innovation. Indeed, it’s difficult to say if we’ll still learn how to drive but “today cars are the target of numerous critics. […] Young people are not interested in having a car” according to Julien Mottin, automotive system director at Prophesee. This may be a way for the car industry to renew itself. That is why it’s important to start with low autonomy before progressively reaching the complete autonomy. But this raises another issue : the need to monitor drivers that may be less attentive so less operational to take back the wheel.  To deal with it, constructors are planning on installing monitors for the level 3 of automation.

To ease this transition and increase the efficiency : constructors intend to make the self-driving cars more humane. The goal is to reproduce and anticipate the human driving to make it safer and more comfortable for passengers. Researchers and companies are also looking at humans for ways to improve the automatization. For example, the company Prophesee decided to imitate the human vision to improve the cameras of autonomous cars. “With our cameras […] we have way more signals concerning small movements which are invisible to the naked eye and imperceptible for traditional cameras“ explained Julien Mottin. One of the biggest challenges of vehicles’ automation is to adapt them to humans’ reactions and behavior. Not only autonomous cars have to take into account the fact that human drivers react more slowly, but they also have to be aware of other road users and adapt to their location. It’s true that driving habits may be different and that’s why “We will have to socialize cars, to teach them to adapt their behavior according to the country they will be in”, according to Arnaud de La Fortelle, director of the robotic center of MINES ParisTech.

Thus, to facilitate the task of self-driving cars we may have to rethink our cities. In the future it seems that humans will share cars instead of owning them, and it means there will be less need of parking, less traffic jams and less pollution. The roads will particularly change : they will be narrower, there will be less signals and roads will be connected to facilitate the communication between vehicles (also called V2V). Moreover, the concept of platooning, a group of vehicles traveling very closely together safely at high speed, will be especially employed for freight transports. Globally, this automatization seems to be positive for our cities. However, all of te changes will require time and investments, and the question is wether they are achievable from an economic perspective.

How sustainable is the autonomous car market ?

The current car market isn’t as remarkable as it used to be. Although the sales are still significant, they tend to stagnate. It is not due to a financial problem but, as it has been said before, to a general indifference in cars : their image has been tarnished, and they are not a symbol of independence anymore. In this context, we could assume that the implementation of autonomous cars in the market would enable car companies to renew themselves, and it might be the main reason why the development of these vehicles is so anticipated by many car brands.

But what would be the impact of this new technology on the economy?

First, the current car market would change: the oligopoly which has existed for decades would be undermined by the arrival of the computer industry on the market. Indeed, car companies that aspire to enter the autonomous car market would have to work closely with high-tech businesses (which is already the case for some of them), and these new participants would definitively change the current market structure : new partnerships between car companies, startups and even universities (for research purposes) could be developed and create new opportunities for both cars and high-tech markets. All car brands would be forced to enter in this new technological aera, because the price of traditional vehicles would drop.

Then, it would be profitable for the State, because nowadays the lack of road security is extremely costly (€50 billion in France in 2016). Autonomous vehicles should drastically reduce the number of accidents, and savings could be redistributed. All the more that it would easily compensate the €1 billion loss from fines and radars.

Concerning employment, it is inevitable that numerous jobs would suffer from the development of autonomous vehicles (approximately 4 millions would be impacted in the US), but it would also create new ones. However, this last argument should be qualified, because these new professions would require specific knowledge that cannot be mastered by everyone.

Finally, the sales of autonomous vehicles would have a strong impact on customers: there could be an increase of shared mobility (and maybe no more real owners), and thus a new way of thinking about the car. It wouldn’t be a simple means of transport anymore, and vehicles could become a leisure / a work place. Autonomous vehicles could also enable car companies to reach new customer bases, because it would be an opportunity for people who were not able to drive (elderly, disabled people) to reach a certain form of autonomy.

In theory, this is what should happen in the near future. But in reality, the implementation of autonomous vehicles faces many obstacles : At a national level, the development of these cars on our roads would require huge investments, only possible with the State’s intervention.

Moreover, there are some major profitability issues for companies, because of the high cost of this technology and the lack of knowledge on the future demand. “The market feasibility is unknown”, according to Grégory Vanel, economy teacher at GEM, and thus no insurance company would accept to insure these cars. For Julien Mottin, “The business model for autonomous vehicles hasn’t been settled yet “.

There is little doubt that if all the vehicles were autonomous, roads would be safer. But, technological progress appart, roads and even the very idea of transportation would have to be rethought : That’s no easy thing to do. And all the more given that the economic sustainability of a new market has a way of being uncertain. Although the outlook of the autonomous vehicles market is far from being bleak, important questions are still unanswered. Amid those questions, what comes first to one’s mind are obviously the ethical issues.

Answering all of these questions will take time. Getting used to the answers will take money. Unsurprisingly we seem to lack both. As a result it seems that a long road has to be done before reaching the point when autonomous vehicles will be the norm.