A study by the Massachusetts Institute of Technology (MIT) examined the CO2 emissions in relation to the computing power of autonomous driving. This is enormous.
Autonomous driving requires high computing power, especially if it is operated globally and on a large scale. Already now crisps specially developed for this technology. A recent study from MIT (Massachusetts Institute of Technology) calculated the CO2 emissions of this computing power. The result: they will cause as many CO2 emissions as all data centers in the world today. These data centers currently account for about 0.3 percent of all emissions in the world and that is equivalent to the emissions of the country of Argentina.
The study assumes that one billion autonomous cars are powered by an 840-watt computer for one hour per day. Furthermore, the researchers calculated that the power consumption 90 percent of autonomous vehicles have to be less than 1.2 kilowatts in order not to overshoot the data center consumption.
If 95 percent of the vehicles in operation worldwide are autonomous around the year 2050, the computing capacity will double every three years. For this, the efficiency would have to be doubled every 1.1 years to keep the emissions under the values.
For the calculation, the researchers from MIT developed a model of autonomous electric cars including the power consumption of the computers on board. Research is also used, according to which the driving time in autonomous vehicles increases. On the other hand, optimal route and vehicle guidance could reduce the traffic load. And in the model, new computer techniques had to be anticipated. Therefore, an industry popular neural network algorithm has been used in various applications.
For the researchers themselves, it was surprising how steeply the consumption curve shot up. But an example makes it clear: An autonomous vehicle uses ten neural networks that analyze images from ten cameras. An hour’s drive makes 21.6 million conclusions a day. With a billion vehicles, it would be 21.6 trillion results and Facebook’s data centers are currently generating around a few trillion results. The calculation model of the electricity consumption of autonomous vehicles does not include the energy requirements of the sensors or the production.
According to the researchers, the solution could be a special driving algorithm with specific hardware that performs the necessary sensor and navigation calculations. However, vehicles currently have a lifespan of ten or 20 years before they are replaced.