Computers that control self-driving cars are part of the automotive technological wave, however there is an environmental concern. These computers could be a trigger for increased global carbon emissions.
New research from MIT finds that if autonomous vehicles are widely adopted, then hardware efficiency will simultaneously need to advance rapidly to keep computing-related emissions in check.
This is based on a model that quantifies emissions generated by computers on fully autonomous vehicles. This finds that if self-driving cars are widely adopted, then their emissions will rival those generated by all the data centres in the world today.
In order to simply keep emissions at or below those levels would require hardware efficiency to improve more rapidly than its current pace. According to some projections, 95 percent of the global fleet of vehicles will be autonomous in 2050.
A further complication is that the computers driving autonomous vehicles today will not be the same systems when autonomous vehicles become commonplace. Typically computational workloads double every three years; this means that hardware efficiency would need to double faster than every 1.1 years to keep emissions under those levels.
The world’s data centres that house the physical computing infrastructure used for running applications account for about 0.3 percent of global greenhouse gas emissions (this is as much carbon as the country of Argentina produces annually).
To draw the parallel with autonomous vehicles, the researchers determined that 1 billion autonomous vehicles, each driving for one hour per day with a computer consuming 840 watts, would consume enough energy to generate about the same amount of emissions as data centres currently do.
By running various scenarios, the researchers determined that to keep autonomous vehicle emissions from zooming past current data centre emissions, each vehicle must use less than 1.2 kilowatts of power for computing.
The model developed to show this is a function of the number of vehicles in the global fleet, the power of each computer on each vehicle, the hours driven by each vehicle, and the carbon intensity of the electricity powering each computer.
The data set was very complex. For example, if an autonomous vehicle has 10 deep neural networks processing images from 10 cameras, and 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. To assess the vast data set, an algorithm called a multitask deep neural network was used.
The research appears in IEEE Micro and it is titled “Wheels: Emissions From Computing Onboard Autonomous Vehicles.”