Computers in autonomous cars could become major drivers of CO2 emissions, a study warns

According to a new study, a global fleet of about a billion autonomous vehicles, each driving just one hour a day, can generate about the same amount of carbon emissions as data centers currently.

The study recently published in ieee micro, models the potential energy consumption and associated CO2 emissions when autonomous vehicles become widespread.

Data centers around the world, which house the physical computing infrastructure used to run applications, currently produce about 0.3 percent of global greenhouse gas emissions – about as much carbon as the country Argentina produces annually, according to scientists including those of the Massachusetts Institute of Technology USA said.

In the new study, they found that to prevent autonomous vehicle emissions from zooming past current data center emissions, each vehicle needs to consume less than 1.2 kilowatts of power for the calculation in over 90 percent of the modeled scenarios.

Researchers are calling for more efficient hardware to reduce emissions from computing operations in self-driving cars.

In one of the modeled scenarios, where 95 percent of the global vehicle fleet is autonomous by 2050, scientists say hardware efficiency would need to double faster than every 1.1 years to keep emissions below those levels.

“If we just keep the business-as-usual trends in decarbonization and the current rate of improvement in hardware efficiency, it doesn’t seem like it will be enough to limit emissions from computers onboard autonomous vehicles. This can become a huge problem,” study co-author Soumya Sudhakar said in a statement.

Scientists say it’s necessary to develop more efficient autonomous vehicles that have a lower carbon footprint from the start to clear the hurdle.

In the research, they developed a framework to study emissions from the operation of computers onboard a global fleet of fully autonomous electric vehicles.

Emissions were calculated based on a number of factors including the number of vehicles in the global fleet, the power of each computer in each vehicle, the hours driven by each vehicle and the carbon intensity of the electricity powering each computer.

The researchers also modeled the emissions based on the advanced computer hardware and software used in such vehicles.

As an example, they say if an autonomous vehicle has 10 deep neural network AI processing images from 10 cameras, and that vehicle drives an hour a day, it would draw 21.6 million conclusions every day.

For a billion vehicles, this would yield 21.6 quadrillion conclusions.

In comparison, all of Facebook’s data centers around the world make a few trillion inferences every day (1 quadrillion is 1,000 trillion), researchers explained.

“After seeing the results, that makes a lot of sense, but it’s not something that a lot of people have on their radar. These vehicles could actually be using a lot of computing power,” said Sertac Karaman, another author of the study.

You have a 360 degree view of the world. So while we have two eyes, they may have 20 eyes looking everywhere trying to make sense of all the things that are happening at the same time,” added Dr. Karaman added.

The numbers calculated in the study only account for the arithmetic operations in self-driving cars and do not take into account the energy consumption of the vehicle’s sensors or the emissions generated during manufacture, scientists say.

They say each autonomous vehicle needs to expend less than 1.2 kilowatts of energy on computing power to keep emissions from spiraling out of control.

According to researchers, in order to achieve this, the computer hardware must become more efficient much more quickly – efficiency is expected to double every 1.1 years.

“We hope people will consider emissions and carbon efficiency as important metrics to consider in their designs. The energy consumption of an autonomous vehicle is really crucial, not only for extending battery life but also for sustainability,” added Vivienne Sze, another co-author of the study.