Researchers at MIT said in a new report that the energy needed to run computers on board a global fleet of autonomous vehicles could generate as many greenhouse gas emissions as all the data centers in the world today.
Data centers are widely known for their carbon footprint, the researchers said. Those centers currently account for about 0.3% of global greenhouse gas emissions, or about as much carbon as Argentina produces annually.
The MIT researchers built a statistical model that found that 1 billion autonomous vehicles, each driving for one hour per day with a computer consuming 840W, would consume enough energy to generate about the same amount of emissions as data centers currently do.
The researchers also found that in more than 90% of modeled scenarios, to keep autonomous vehicle emissions from exceeding current data center emissions, each vehicle would have to use less than 1.2 kW of power for computing, which would require more efficient hardware.
They said that in one scenario—namely, where 95% of the global fleet of vehicles is autonomous in 2050, where computational workloads double every three years, and where the world continues to decarbonize at the current rate —hardware efficiency would need to double faster than every 1.1 years to keep emissions in check.
Business-as-usual trends in decarbonization and current rates of hardware efficiency improvements are likely to fall short of constraining emissions from computing onboard autonomous vehicles, said one of the researchers.
As part of their study, the researches modeled advanced computing hardware and software that doesn’t exist yet.
To do that, they modeled the workload of a popular algorithm for autonomous vehicles, known as a “multitask deep neural network” because it can perform many tasks at once. They looked at how much energy this deep neural network would consume if it were processing many high-resolution inputs from many cameras with high frame rates, simultaneously.
For example, if an autonomous vehicle has 10 deep neural networks processing images from 10 cameras, and if that vehicle drives for one hour a day, it will make 21.6 million inferences each day. One billion vehicles would make 21.6 quadrillion inferences. By contrast, all of Facebook’s data centers worldwide make a few trillion inferences each day (1 quadrillion is 1,000 trillion).
“These vehicles could actually be using a ton of computer power,” one researcher said.
The MIT researchers modeled only computing — it did not take into account the energy consumed by vehicle sensors or the emissions generated during manufacturing.
