A new study by MIT researchers has found that the energy needed to run the powerful computers on autonomous vehicles could generate as many greenhouse gas emissions as data centers in the world today. Data centers currently account for 0.3% of global greenhouse gas emissions, the same amount of carbon produced by Argentina annually. The study aimed to examine the potential energy consumption and related carbon emissions if autonomous vehicles were widely adopted.

Modeling emissions

The researchers built a framework to study the operational emissions from computers on board a global fleet of electric autonomous vehicles. The model takes into account the number of vehicles, power of each computer, hours driven, and carbon intensity of the electricity powering each computer. The researchers also considered advanced computing hardware and software that does not exist yet and modeled the workload of a popular algorithm for autonomous vehicles, the multitask deep neural network.

Computing Workload

The research found that if an autonomous vehicle has 10 deep neural networks processing images from 10 cameras, it will make 21.6 million inferences each day. A global fleet of one billion vehicles would make 21.6 quadrillion inferences. The researchers were surprised by the quick workload accumulation of the algorithms. Autonomous vehicles could be used for moving goods as well as people, leading to a significant distribution of computing power along global supply chains.

Keeping Emissions in Check

To keep emissions under control, the researchers found that each autonomous vehicle must consume less than 1.2 kilowatts of energy for computing. This requires hardware efficiency to double faster than every 1.1 years. In over 90% of modeled scenarios, hardware efficiency must double faster to keep emissions under current data center emissions levels. The researchers noted that hardware efficiency must improve significantly for autonomous vehicles to have a smaller carbon footprint from the start.

Conclusion

The study highlights the potential for autonomous vehicles to become a significant source of greenhouse gas emissions. The researchers found that if business trends continue, it would not be enough to control emissions from autonomous vehicles. However, if efficiency improvements are made, autonomous vehicles could have a smaller carbon footprint from the start. The study serves as a warning for the industry to consider the environmental impact of autonomous vehicles and take necessary steps to minimize it.