A title so unbelievable it almost sounds like a clickbait
Over the past years cars have become increasingly more automated. Twenty years ago a driver would have to (almost) solemnly control the car, but recently Artificial Intelligence (AI) has been taking over more and more tasks of the driver. Many modern cars have convenient AI tools such as lane assist, automated parking or automatic braking systems. Various car producers (Toyota, Tesla, BMW, etc.) are working hard to develop a completely autonomous car and are suggesting that within five years these cars will become reality. However, there has been some concern regarding this development. Most concerns are with regard to the safety of the AI systems or the AI capability of driving the car, however, some concerns are also with regard to the energy cost of the AI that is controlling these cars. In a recent article in “het Financieele Dagblad” Prof. Dr. Max Welling discusses intelligence per kilowatt-hour and gives the example of what we will eventually be paying to drive around in autonomous cars. However, this statement is a bit out of context as autonomous cars have various positive side-effects, eventually making us not worse off money/energy-wise.
Let’s start of by first addressing the energy (and thus money) consumption of the AI in autonomous cars at the moment. Welling claims that with current technologies autonomous cars use at least around 2000 Watts extra to drive around. Humans take around 20 Watts to control a car, so indeed this number seems to be out of proportion. To put this in perspective, the writer calculates that with a billion cars on the road globally, all travelling roughly an hour a day, we find ourselves with a total cost of 200 billion euros per year to be driven around autonomously. Yes, this does seem worrying, but I am writing this article to take away your worries! So let’s start by putting this 200 billion euros price tag in perspective.
According to the National Motor Vehicle Crash Causation Survey Report carried out by the National Highway Traffic Safety Administration 90% of the crashes in the United States are caused by human errors. Therefore, if we were to completely automate the American traffic fleet we could theoretically get rid of these crashes. Bertoncello and Wee conducted a follow-up research calculating how much money potentially could be saved for the U.S. economy if these crashes were prevented. Their calculation came to the absurd number of 190 billion U.S. dollars per year. Even if we were to correct for currency and for example assume that autonomous cars cannot prevent 100% (let’s say it takes care of 75%) of the crashes caused by humans, we will still already have covered far more than half (115 billion euros) of the costs of autonomous driving AND we haven’t even looked outside the U.S.!
So, the money matter is already non-existent, but one might still argue that we will be using a lot more energy to drive around in our autonomous cars. Before we start talking about the energy efficiency increases of an autonomous car it would be useful to know how many Watts a car consumes when in use. Although this value depends heavily on your speed and your acceleration/braking pattern, averages can be found using data gathered by the Dutch Government and Statistics Netherlands (Centraal Bureau voor de Statistiek, CBS). Together they have (Dutch) data on the total energy consumption of the traffic, travelling times and the amount of cars. Combining this data we can find that a typical Dutch passenger car (we are only considering passenger cars as they are the main energy consumer in traffic and will be the first to be automated) uses roughly 55000 Watts, which is a good average applicable in many locations around the world. We can already note that an increase of 2000 Watts will only increase this amount by 3.6%, so if we were to find an increase in energy efficiency of 3.6%, due to automation, the problem of the increased energy consumption will already be dealt with. So let’s dive into the various advantages of autonomous cars.
First of all, as computers have faster reaction speeds than humans, cars will be able to follow much closer to each other than what humans will be ever be able to safely do. This tailgating is known as platooning and the positive effect of this is a reduction in aerodynamic drag experienced by the tailgating car. Research has shown that the second car (in a group of two sedans) experiences a drag reduction of around 25%, while the leading car experiences a small increase in drag. Combining the two values, the average decrease in drag is 10% per car (at 30 m/s).1 It is important to realize that drag is not the only energy consumer of a moving car, however on a highway it is by far the key energy consumer of a car (for more information about car energy consumption, I suggest Mackay). Also note that now the effect has split over two cars, but if more cars are available for platooning the decrease in drag can be significantly higher than 10% per car. But let’s assume a worst-case scenario where there are on average only two cars available on the highway. As we use less energy during platooning we can allow ourselves to use extra energy when we are not platooning. Making a short calculation we conclude that for every second of platooning 2.778 seconds can be driven without additional energy consumption.
Another interesting energy efficiency increase is the nearly perfect driving style of an autonomous car. Researchers developed an energy management system in combination with a driving cycle optimization algorithm, which was able to increase fuel efficiency up to 56%-86% for fossil-fuel cars. For plug-in hybrid cars it was even able to boost energy efficiency up to 115%.2 However, other researchers claimed to have only found decreases in fuel efficiency in the range of 30%-50%.3 As we are talking about fuel efficiency a 100% increase in fuel efficiency means a 50% decrease in energy consumption. Therefore, even if the researchers with the lower estimates turn out to be more accurate our additional energy consumption of the AI is more than compensated.
A critical note towards this blog post is that a possible increase in car use, due to automation, has not be taken into consideration. However, I do not believe this to be problematic. Even if actual efficiency increases in fuel consumption of autonomous cars will be according to the lowest scientific estimates, we will still have a decrease in energy consumption of around 12% (even without including energy efficiency increases due to platooning), which will act like a buffer for the growth in use. Furthermore, an increase in car usage will most likely mean a decrease in alternative transportation methods, thus decreasing energy consumption in these sectors. Lastly, it is speculated that with the coming of the autonomous car, the sharing economy, which has been growing over the past years, will also grow strongly within the car sector. Therefore, increasing for example the amount of carpooling while simultaneously reducing the total amount of cars needed for transportation, thus limiting actual additional energy consumption from increased usage.
In conclusion we can say that a transition to autonomous cars will indeed save us money and decrease energy consumption within the transportation sector. Thus, after all, the title was not a clickbait.
1 Simulation of the aerodynamic interaction of two generic sedans moving very closely (2011). Zhu et al.
2 Forward power-train energy management modeling for assessing benefits of integrating predictive traffic data into plug-in-hybrid electric vehicles (2012). He et al.
3 Help or hindrance? The travel, energy and carbon impacts of highly automated vehicles (2016). Wadud et al.