Dr Anders Sandberg at Oxford University and Research Director at think tank Eudoxa, shares how new technology in the near future will make the car become a smart robot. One that goes on autopilot and communicates with other cars, the personal calendar and Department of Transportation. And we are almost there. Not so long ago, Stockholm motorists installed a transponder on the windshield as a preparation for the experiment with congestion charging.
That modern cars are becoming more digital is nothing new, but the trend has far-reaching consequences which will mean that cars go from being tools to become robots. Here are seven possible effects of the technologies that are under development:
For the roadtrip, the wedding day or just to get away, one will be able to let the car select destination. The car knows what the driver and the passengers like, how they feel and what the occasion is. It also knows where you have been and not been. (A friend and his wife do this type of random trips each wedding day - though in a more analog approach.)
Paying for green light
Smart cars can communicate with the traffic system and negotiate the green light. (The technology is already in place for emergency vehicles and buses). Motorists who want to progress more quickly can pay a small sum and traffic light will prioritize them a little higher. Good source of revenue for the Road Administration and a reason to go after the busy business men's cars.
The same way that both water and land animals group for the task of protection and cooperate when hunting for food, smart cars can form not just columns (ie traffic jams), but self-organized flocks. During long trips on the highway one could have temporary neighbors. Children connect computer games against neighboring car's rear seat and chats. Romances between drivers may also arise across lane lines.
Road villas becomes the prank of 21th century
Just as human intellect, artificial intelligence can be fooled if it is fed with false impressions. With small effort it will be possible to fool smart cars so that they behave strangely. Strategically placed objects, with surfaces that reflect the car's radar, can get cars to become overly cautious, a particular route may attract cars that end up in seemingly inexplicable lanes.
The car can access the calendar in your computer and by itself find the most efficient order to visit different sites in. While the owner is asleep the car will evaluate itself and the calendar to decide whether a pharmacy visit or meeting with the colleague should be made first, and it closes the case by contacting the colleague’s calendar to propose a schedule change.
Depending on settings and that the car learns the driver's personality based on driving, a car can customize its personality to a high degree. Careful cars that avoid the risk of accidents, green cars carefully saving on emissions (and perhaps, if the owner wants it, subtly penalize non-environmentally conscious cars in progress by activating pane flushing just when you meet the environmentally unfriendly car), cars that try to avoid breaking traffic rules - or cars touching the limits (and other cars programmed to squeal if they observe something weird). Because a car's personality says a lot about the owner, and vice versa. Something insurance companies have taken on.
The collective car conscious
When a car's software learn something, it transmits the information to the vehicle manufacturer’s server, which can pass on knowledge to other cars. Not only new drivers updating the hardware in the car, but it is also conceivable that information, about how a car prevented or did not prevent a certain sort of accident, can get other cars to be safer.
There is no reason to believe that people are the best drivers. We are slow, has a wandering attention, can not interpret the signals from the car’s all sensors and, sometimes, we become stressed. In short, we are not primarily created to drive. That suggests that automation can be significantly better than people in more and more areas.
After the automatic gearbox, cruise control was probably the first step towards the robot-car. Adaptive cruise control is coming now, which makes the car detect cars in front of them with laser or radar, and slows down or accelerates in order to keep a suitable distance. Sensors can also warn or prepare the car for a crash.
The next step is communication between cars so they can jointly plan their behavior. One area studied is "platooning", where cars on motorways automatically coordinate their acceleration and form chains to save fuel and reduce noise. But to get these benefits, cars need to run close together, which requires precision and rapid response, as people hardly cope.
Networking cars can also act as mobile sensors - report road conditions, warn of obstacles and to alert rescue services if accidents happen.
In certain moments, such as parking, most people are probably willing to let the car steer itself. But it is likely that cars will be able to govern themselves in more and more areas.
The dream of the driverless car has lasted for a long time, but so far been held back by the difficulty to understand the car's nearest environment, navigation, control the car by self and to plan movements. Today, sensors and GPS are cheap and good enough to solve the first two problems, and digitization of car mechanics has made control manageable. What remains is the motion planning, a classic robotics problem where much research is ongoing.
Perhaps the strongest argument for driverless vehicles, are the results of the DARPA Grand Challenge, a competition for driverless vehicles sponsored in 2004, 2005 and 2007. 2004, no participant passed the goal line at all. 2005 all the cars, except one, beat last year's record, and five of the 23 vehicles passed the more difficult 212 km off-road route, which included tunnels and sharp turns. In 2007 the race was moved to an urban environment, where vehicles navigate a 96 km long stretch in which they were forced to follow traffic rules and deal with other traffic. The winner, Carnegie Mellon University’s “Tartan Racing” survived the journey in 4 hours and 10 minutes. Maybe not such a high speed, but the pace of the vehicles are impressive. Given that the U.S. military will continue to be very interested in finance and use autonomous vehicles, it is likely that we will see more efficient vehicles in the near future. Applications on civil cars are likely, especially as the EU and other players show more interest.
What is really important for the popularization of smart cars will be safety, how practical they are and the matter of responsibility. As with other technologies, the price of the intelligence will decrease and become negligible over time.
Whenever any form of safety equipment will be cheap enough, usually it also becomes mandatory. This means that it is perfectly conceivable that intelligent automation is not just an option but a standard.
The interface between man and machine is important. It is not enough that a car is smart, one must be able to spend time with it too.
The really difficult question is how much we will trust smart cars. People need a sense of having control of the situation. That is why many more are afraid to fly (no control) than to go by car (more control). Networking cars may well be seen as a threat to privacy - but who knows, maybe they also allow "social running" which instead encourages use. For example, the car proposes the trip objective, since it knows that other people with same interests as the driver is going just there.
The proliferation of autonomous vehicles can also be stopped by legislative difficulties, for example when questions of liability in case of an accident can not be resolved. Today we have little experience in dealing with autonomous machines responsibility, and even "normal" accidents could be due to automation. Although the system is safer than human drivers, big lawsuits could cost a lot of time, money and trust - the group of people rescued by the system is invisible and unaware of it, while those suspected to have been harmed by it are visible. Here it is important to have the ability to demonstrate increases in general safety, as it is likely that many states will protect their industry from litigation if it can demonstrate that autonomy provides security (and perhaps environmental) benefits.
All technology development takes time. We both over- and underestimate how quickly certain techniques can pass through. British analysts have made cautious predictions of various smart road infrastructures by year 2055 here.
I think they really underestimated how quickly our traffic will be robotized.
Dr. Anders Sandberg
James Martin research fellow Future of Humanity Institute, James Martin 21st Century School, Oxford University, and director of scientific research on think tank Eudoxa.