Guestblog by Roy Cobbenhagen and Lex Hoefsloot, Eindhoven University of Technology

The 2020’s will be an extremely interesting decade for the car, mobility and how we perceive the world around us. The progress of the sustainable energy generation and artificial intelligence will cause a revolution in both the energy system and the use of cars.

This two-part blogpost is a summary of a report by Roy Cobbenhagen and Lex Hoefsloot into the trends currently developing in the automotive sector and the energy sector. The focus in these blogposts is how these trends may evolve in the 2020-2030 decade.

The first part dealt with energy. Here it was argued that it is possible to switch to clean energy. We will also show what its challenges and its opportunities are. This was investigated together with the trends in personal mobility to paint a picture of how these two sectors will influence each other.

This second post will be about the growing intelligence in cars and how we will use cars. It will demonstrate the intricate relationship between autonomous vehicles, car sharing and ride sharing. It will also focus on the true challenges behind the introduction of the autonomous car.

Part 2: The Intelligent Car

Computing power (which leads to artificial intelligence) is going to make mobility easier and cheaper. On the one hand through ride and car sharing which are already taking advantage of ‘the internet of things’ to make using these services as easy as possible. On the other hand through autonomous cars which will gradually be able to fully take over the steering wheel. Combining car sharing and autonomous cars will give us the perfect form of transport: highly utilized (and therefore cheap) cars which will pull up right at your doorstep and during driving you can focus on doing work or have some social time with friends. This article will show how.

The connected car is enabling technology with no technological roadblocks

Accessing information about the car from a distance by connecting the car to the internet shows to be an enabling technology. This technology is well on its way and there seem no major problems in the future that will stop it. The only issue that needs be addressed and proves to be difficult is the security of the connection. This is a very important factor as it involves both safety and privacy. Even though this is a stressing subject, it seems not to pose any problems with the rollout of the connected car. We divided the possibilities of the use of connected cars into three sections: 1) personal, 2) industry and 3) cooperative driving. We will discuss the uses of the connected car in the coming sections about other technologies, as the connected car is an enabling technology.

The autonomous car is coming but automation will first act as a life saver

Looking at the five SAE levels of autonomy, we will soon reach level 4 (fully autonomous for most driving modes). Cars will do most of the driving themselves and the user will only occasionally intervene. There are many hurdles that need to be overcome before the autonomous car will reach level 5: the fully autonomous car (for all driving modes). We argue that this hurdles can be overcome and that there are significant advantages of level 5.

We split the autonomous technology into four phases: how the autonomous car senses, understands, decides and acts. ‘Acting’ is controlling the wheels, brakes and throttle of the car and this is not considered a hurdle and is overcome relatively easy. The ‘sensing’ phase requires better and cheaper sensors. This proves not be a hurdle and will improve significantly. The fusion of different types of sensors will then result in a better model of the world around the car and thus a better a job of ‘understanding’ the situation around the car. Connectivity of cars will help the sensors fuse as the cars can together generate a more accurate image of the world around because then the car is not solely dependent on what happens in the direct ‘line of sight’ of its sensors. This results in the fact that the car can ‘see around corners’. The connectivity will also lead to better cooperative driving as a group of autonomous cars can better interact with each other and the infrastructure.

The major difficulty lies in how the car ‘understands’ and ‘decides’.

The car will have to know what all human signals indicate (e.g. waving, frowning, nodding) and make predictions about human behavior. After it knows what the situation is, it must decide what to do. The number of scenarios coupled to these decisions grow exponentially with the increase of detail. This artificial intelligence (AI) is of extreme difficulty and can be considered the largest challenge in the technology of autonomous driving.

We argue that these will be overcome, but not shortly. Cars are already proving that machine learning in fleets of cars works (Tesla Autopilot[1]). The enormous amounts of situations that can be simulated virtually and physically (e.g. cameras in cars observe how drivers react to certain situations, from which the AI can learn) can provide the data that machine learning needs. This could be accelerated by sharing ‘the lessons learned’ between cars. Knowledge sharing between cars should not be compared to people teaching skills to other people, it is more like injecting knowledge. In some way, this network of cars can better be seen as one ‘brain in the cloud’ that knows how to act in certain situations. Larger car communities will therefore learn faster and provide a safer and better performing car for customers. Higher sales rates will then cause the biggest community to grow even bigger.

‘First man on the moon’-effect: the winner takes it all

This will cause a race (if it hasn’t already started!) between companies, countries and even continents to be the first to deploy a large number of (semi) autonomous cars. A race that will eventually lead to a level 5, fully autonomous car. The key to unlocking autonomous driving is a proper functioning AI system. But a highly intelligent AI system will not only heavily influence the automotive sector! Many industries will be disrupted by the advent of intelligent self-thinking machines. It is widely seen by experts that AI is going to be the most significant technology of the 21^st century. Whoever gets there first has a huge advantage in developing future technologies. We call this the ‘first man on the moon’-effect.

Non-technical roadblocks: ethical questions are difficult; the rest can be overcome.

With the increase of AI in the car we also must tackle ethical questions about which decisions the car must make. This is a difficult question and must be carefully dealt with. We believe that this is the largest non-technical roadblock. Must we collectively decide what a car must do? If the car has to choose between sparing the life of one person or another, which one must it choose? These are very hard questions and the main difficulty lies in the fact that we must agree on the answers and have the programmers program it. Or must the car act upon the preferences of its owner who was given a list of the options before using the car?

It’s a societal challenge.

Today, drivers are able to make those decisions independently and have to do so in a split second. Companies, other people and governments do not have to to bear the responsibility for these actions. But with autonomous cars, we all bear the responsibility and all of our choices will have (measurable!) consequences. It’s a chance for philosophers to get their hands dirty since philosophical ideas will have to be implemented and executed.

There are also questions about liability in accidents. We believe that there are plenty of viable options that solve the question of who is responsible in the case of an accident. One of promising (and interesting) options is to make the car a legal entity. In this way the owner is not responsible for its actions, nor do large automotive companies face legal claims by victims of accidents (which could be tackled with settlements by large teams of lawyers). If the car is a legal entity, then the insurance companies come into play. These can choose which brands of autonomous cars they ensure and which they refuse. This strengthens the burden of auto manufacturers to make safe autonomous cars. Whereas the users are left out of the guilt. This of course needs a lot of refinement, but it seems that this is a promising way.

In many places autonomous cars are not allowed on the road. It seems that whoever figures out autonomous cars first will have a major market advantage. This results in the fact that legislation will follow the technological process as the (union of) countries benefit(s) as well from the development of autonomous cars.

Ride sharing will provide cheap and easy transportation for all

Ridesharing is currently the cheapest form of non-private transportation (i.e. where you are not the driver) that is not subsidized by the government. Its community also shows exponential growth in both passengers and drivers. People speak of the ‘uberification’ of services. This means that there is a rise of applications that match demand with unused (social) capital. These applications have disrupted many industries by providing better (in quality and/or more personal) services than the existing applications because these services do not have the standard ‘business with employees’ structure. Uber (among others) is doing this in the transportation industry. The connected car is one of the reasons that this is possible as it can provide status on the location (and thus arrival time) of the car.

Car sharing would be possible today if there was enough investment

In our report you can find simulation results showing that car sharing applications are possible today even at low adoption levels. The prices are lower compared to owning a cheap and small ‘second car in the family’. More cars would be needed to attract regular users but this seems to be the problem with shared cars today: there is not enough investment made. If the number of cars would increase, the walking times to the cars will decrease tremendously and the overhead can be made smaller and costs can be reduced.

Not only cheaper mobility but also more luxurious cars for the same costs through car sharing

Shared cars are on average newer and better equipped than their non-shared counterparts because their lifetime is shorter (they drive more distance a day than privately owned cars) and a higher purchase price can be spread over a large group of users. The first mass adopted autonomous cars might very well be shared cars, because of their high cost in the early stages. This is another reason why shared cars will drive the development of the autonomous car as they have to be replaced by a newer version sooner than traditional cars.

The lack of a larger number of cars and the lack of a user-friendly system interface of ordering a car seems to be the primary reasons why car sharing services are not popular. We argue that ride sharing and autonomous cars will solve this problem.

Three trends accelerating each other.

Shared cars will be the first mass market for autonomous cars.

Since shared cars can be more expensive than non-shared cars while keeping the cost for the users equal, the mass market introduction of expensive autonomous car technology introduction will likely happen through car sharing. These shared fleets of autonomous cars will also help autonomous cars to learn from the situations they encounter on the road.

Ride sharing and fully autonomous cars will accelerate the introduction of car sharing.

Of course the apps that disrupt industries by uberification can also be disrupted themselves. This could be done by technology that provides the service cheaper and/or better than the human capital that the ‘uberificating’ apps provide. Ironically the uberification of transport will not continue in the way Uber does now! Because autonomous cars will be cheaper as a shared car than the Uber drivers are. This is for instance not the case for hotel chain Airbnb as there will always be empty rooms and that capital stays the same.

The major advantage of level 5 autonomy versus level 4 is that the driver is no longer needed. This would cut transportation costs dramatically as all the ridesharing services will become car sharing services. The trip costs go down rapidly because the only costs are now the ‘per kilometer costs’ which are incredibly low for BEVs and solar powered cars. This is also a major incentive for the entire logistics sector.

Another reason the transition to large scale autonomous car sharing will be done by ride sharing companies is because of the large user base it already has and its experience in coordinating complex systems of supply and demand. As stated before, a user-friendly system could prove to be one of the most important triggers for the success of car sharing. Ride sharing apps like the Uber app have exactly the same interface as an autonomous car sharing app would need. There is a certain activation capital needed to go to car sharing and the ideal candidate to provide this are the ride sharing companies as they have the interface as well as the user base.

Conclusion: clean, cheap and easy transportation for everyone

In the future we will have cheap sustainable electricity available that we can use in electric solar powered cars. This will lead to very low energy consuming cars at low costs. The usage of cars will also change as car sharing will emerge and have a dominant presence through autonomous vehicles and ridesharing. This will change public transportation as we know it.

Using autonomous shared cars will be as simple as pressing a button (like it’s an elevator) and probably even cheaper than travelling by bus, train or using your own car today.

This all will result in clean and cheap mobility available for everyone. For the people at the higher wealth level and the many hundreds of millions that will join this level in the coming years, but also for the people in poor countries as cheap mobility is what they need.

The consequences.

This clean and cheap mobility will most likely go hand in hand with an increase in the mobility demand. Will this lead to more congestion? A whole new study should be performed to see how our world will cope with an increased demand of mobility. Even though the mobility is clean and cheap, we will have an increase in the clean energy demand if we have more mobility. This is a strong incentive to keep improving on making the cars lighter and more aerodynamic.

[1] http://www.hybridcars.com/tesla-autopilot-already-improving-through-fleet-learning/).