Driverless cars

At first glance, the cars look like they are performing an everyday manoeuvre. It is 12 November, and three cars are driving in a convoy on the motorway. The rear vehicle passes the middle vehicle and returns to the right lane. However, there is nothing normal about this manoeuvre on the A10 motorway. These are semi-self-driving cars, and TU Delft helped develop them.

The three cars (Toyota Priuses) are crammed full of technological gadgets. “They communicate with each other. The Prius doing the passing asks the other two vehicles to make room for it. A real driver carries out the actual manoeuvre, but this is still an important first step towards a self-driving car. We conducted this test in a real-life traffic situation rather than on a test track. This is new in itself, and as such we have provided a demonstration of the current capabilities of these systems,” explains prof. Bart van Arem, director of the Transport Institute. TU Delft will buy two Toyota Priuses this autumn which will be fitted with the new technology.

There is nothing new about the idea of self-driving cars, thanks, among others, to Google’s driverless car project and science fiction films. However, this year it will be revealed that the future is no longer science fiction. Right at this moment, all over the world, important tests and developments are taking place that will bring the semi-self-driving car another step closer. One of these is TU Delft’s experiment. At the same time in Japan, Mazda is testing how different types of cars and trams can communicate with each other (see box).

Car manufacturers are building more intelligent vehicles too. Take BMW. This autumn, the German car maker will be presenting the Traffic Jam Assistant. This technology ensures that a vehicle keeps a certain distance from the car in front of it. It will work at speeds of up to forty kilometres per hour. This will be particularly useful in traffic jams: the vehicle will increase speed and brake itself when it is needed.

The American car manufacturer Ford recently presented a system that uses three radars, ultrasonic sensors and a camera that monitors a 200 metre strip of road. The driver is given a warning if they are in danger of driving into anything, and if they fail to respond then the car will avoid the obstacle itself by braking or steering.

 

Wi-Fi network

Such systems are currently being applied and tested all over the globe by car manufacturers and universities. The cars are packed full of technology. “Take the radar, hidden behind the grill: the narrow beam can project more than 100 metres and detect the cars in front,” explains Van Arem.

In addition, the cars can communicate via a dedicated automotive Wi-Fi network. And there are cameras that monitor the road as well. “Among other things they can detect manoeuvres made by other cars. We use as many cameras as possible because they are much cheaper than a radar. Laser scanners are even better, but these too are very expensive.”

One technology that is very important in the development of intelligent cars is geographical positioning. It can be used to warn the driver it they are in the wrong lane, for example. This will require a kind of ‘super GPS’, because the current navigation systems do not have this degree of accuracy (the current margin of error is around 20 metres).

TU Delft is collaborating with Technolution, NXP and TomTom to build a new system. “We are using predictions of atmospheric conditions and satellite orbits to make corrections to the GPS signal and have achieved an accuracy of 50 centimetres. Next year we will be testing the new system on the A67 motorway, nearby the cities of Eindhoven and Venlo.”

Another critical factor is driver alertness. At first sight, you would think that the driver should be able to relax completely if their car does everything for them. However, they cannot sit back and read the newspaper or even fall asleep. Because what if an accident threatens and the system is defect, or there is a speck of dirt on the lens, or traffic is abnormally heavy? The driver may well then be required to intervene.

For this reason, the cameras in a self-driving car do not only monitor the road, but the driver too. “Using an infrared camera with eye tracking, we can monitor the driver’s alertness,” explains traffic psychologist Dr Raymond Hoogendoorn of TU Delft. “The role of the driver is changing. Instead of the controller of the vehicle they will become a kind of supervisor, comparable to an aircraft pilot.”

 

Psychology

At the global level, hundreds of millions are being spent on this technology, but of course the self-driving car is about more than just money and efficiency. The psychology is important too. “How will other road users respond to self-driving cars? We have come up with wonderful technologies, but we still know nothing about the psychology of the concept,” says Hammond.

Some studies suggest that if 10 percent of the cars were self-driving then there would be a 30 percent time saving in traffic jams. “But such studies are based on simulations! We have no idea whether this would also prove to be the case in practice. How will other drivers respond if they see a convoy of self-driving vehicles driving close behind each other? It could cause a distraction, in much the same way that traffic jams have a knock-on effect on traffic coming the other way due to curious on-lookers slowing down. We simply do not know. I believe we need to carry out more tests in driving simulators or on the road, so that we can find out how people actually behave when confronted with this new technology.”

It is also questionable whether people will be prepared to give up control in the first place. Will they accept their car being in command? “I think the success hinges on the user-friendliness of the system. And there has to be a clear advantage, such as less road accident victims, less fuel consumption and less traffic jams. And there will always be petrol heads who will not give up the wheel.”

Van Arem agrees. “But there is so much interest in self-driving cars that I doubt that many people will reject the concept. These days, journalists do not ask me if we need self-driving cars, but when they will be introduced.”

The professor of Transport and Planning sees a shift in thinking. “A few years ago, my colleagues said that lane keeping and adaptive cruise control could not be combined. They said it would be too dangerous because drivers would hardly have to be alert, because their cars would automatically use the right lane, and brake when needed. But this combination now exists, supplemented with a system that monitors the driver’s alertness.”

Reliability will also play an important role in the acceptance process, says Hoogendoorn. “If there are a couple of major incidents with self-driving cars and a car causes an accident or ends up in a field, then this will have a major impact on acceptance. This is why it is extremely important that we know precisely where a car is and where other road users are. This is particularly complex where it concerns pedestrians and cyclists. Cars are easy to detect with sensors, but pedestrians and cyclists are more difficult, among others because of their unpredictable behaviour. This is why I anticipate that the first systems will be mainly used only on motorways.”

 

Traffic jam assistant

In Japan they are currently studying how intelligent systems can save lives and prevent bumper-to-bumper accidents. Van Arem: “They are testing whether they can control the speed of acceleration externally. Traffic jams are particularly dangerous when the terrain is sloped. Traffic jams and bumper-to-bumper accidents could be prevented by controlling the speed of the cars. We are developing the rules for such a system.”

Alongside the psychology, another important aspect of such intelligent systems is the legislation. TU Delft is cooperating closely with RDW, the body responsible for registering motor vehicles and granting permits for imported vehicles. RDW was closely involved in the trial on the A10 motorway. Van Arem: “RDW will have to determine under which conditions self-driving cars will be permitted, which requirements they will need to meet and how to determine whether they comply.”

Van Arem and Hoogendoorn anticipate that changes will be introduced gradually over the coming years. “The first new introduction will be the traffic jam assistant. This system will take over the control of vehicles in traffic jams to ensure that traffic continues to flow. This will mean that traffic jams will be cleared sooner because there will be smaller gaps between vehicles,” explains Van Arem.

Parallel parking will also become an automated manoeuvre. “Parallel parking is difficult for humans, but simple for automated systems. The self-driving system will gradually assume more and more tasks, such as maximum acceleration, automatic braking and lane keeping.”

Van Arem emphasises that the transition will be gradual. “Car manufacturers claim that they will be producing the first self-driving cars in 2020. But they have not said how independent these cars will actually be.

Japan

A motorist crosses in front of a tram without paying attention, but is not hit because the tram stops automatically. This scenario is reality in the Japanese city of Hiroshima. Since October, cars and trams there are able to communicate with each other across 700 MHz radio waves. They inform each other of their direction of travel but also exchange information on the type of vehicle and its position. Trams can now detect a car making a turn or stalling in front of them. The tests are being performed by the car manufacturer Mazda, in collaboration with, among others, the University of Tokyo. The Mazda 6 being used in the experiment is also fitted with a warning system that detects pedestrians and road markings via sensors and cameras.

TU Delft and the self-driving car

About six years ago, the 3 universities of technology (3TU) agreed that Eindhoven University of Technology (TU/e) would focus on automotive while TU Delft would keep a low profile in this sector. And yet, TU Delft has now presented a semi-self-driving car. While at TU/e they are working on one as well. Is this a problem? “Not at all! In fact the technical universities actually cooperate on this technology. At TU/e they are focussing on the technology that goes into such cars, while we approach the subject as an issue of transport. For example, we are studying geographical positioning and the behaviour of traffic flows. We are also collaborating in the field of robotics. Of course there is some overlap between TU Delft and TU/e, but we are each approaching the intelligent car based on our own specialisations.”

Van Arem thinks it logical that both TU/e and TU Delft are involved in automotive. “High-tech cars combine so many disciplines, such as mechanical engineering, traffic engineering, robotics, ICT and electrical engineering. Almost all technical universities are involved in some way. Moreover, TU Delft has gained much recognition in this area thanks to the Nuna and Formula Zero. So it is not all strange that we are working on this concept.”