US algorithm helps autonomous vehicles reduce braking, saving 8% to 23% energy

Self-driving car technology has taken another step forward, thanks to a team from Clemson University, which recently concluded a three-year study into the energy-saving effects of self-driving cars.

(Image credit: Clemson University)

Ardalan Vahidi, a professor in the Department of Mechanical Engineering, said that he and his team created an algorithm that can help wirelessly connected self-driving cars predict the behavior of other vehicles to reduce braking operations. The less the vehicle brakes, the less energy is wasted, and the more energy-efficient the vehicle is. In the end, the research team found that the algorithm can save 8% to 23% of energy depending on the driving scenario.

As cars become more autonomous and more run on electricity rather than gasoline or diesel, Clemson engineers are also playing a key role in shaping the future of transportation.

The team involved in the study tested the algorithm on two separate self-driving cars, a gasoline-powered Mazda and an electric Nissan, both connected to the same wireless network so they could send and receive data such as speed and direction of travel. The two cars took turns driving on a closed track in southern Greenville County so that only one car was on the track at any given time.

The researchers used computer simulations to create "ghost" vehicles in front and behind the Mazda and Nissan vehicles, tricking them into thinking they were in traffic. This allowed the researchers to try more difficult driving scenarios because even if a collision occurred, it would be with the "ghost" vehicle and would not cause any damage or injury.

Some of the "ghost" vehicles are self-driving cars, while others are driven by computers simulating human drivers. Each test requires driving around the track seven times, and there are U-turns at both ends of the track, which will cause vehicle slowdowns and traffic jams. The test results found that Mazda and Nissan are more energy-efficient when following the self-driving "ghost" vehicles than following the "ghost" vehicles that imitate human driving.

The self-driving “ghost” car shared its intentions with the Mazda and Nissan cars a few seconds in advance, giving the ghost car and the real car a chance to coordinate braking. Simulated human drivers, like real human drivers, are unpredictable, giving the vehicles even less opportunity to cooperate.

"The Mazda and Nissan test vehicles were able to save 20 to 23 percent of energy when following a simulated autonomous vehicle connected to a wireless network, compared with only 8 to 12 percent when following a simulated human driver," the researchers said.

Although not part of the study, the team also found that this finding bodes well for commuters who don't want to wait in traffic, as braking often causes "phantom" traffic jams. However, allowing Mazda and Nissan to predict in advance what the "phantom" vehicles ahead will do could make traffic flow more smoothly and help alleviate stop-and-go traffic jams.