Most of us don’t like to wait at traffic lights for the light to turn green to pass them. Sometimes we try to bypass the signals and end up getting caught in the process, followed by paying a huge fine. Sometimes, this can also lead to accidents. It is true that waiting at the signal is very tiring and costs us in terms of wasted fuel and time, while harming the environment. Motorists and policy planners want a workable alternative to end this inconvenience, albeit a minor one. Researchers at the Massachusetts Institute of Technology (MIT) believe they have discovered one.
The team tried to find ways to ensure that motorists did not have to wait at an intersection for the signal to change colour. Instead, what if they get to the signal precisely at a time when it’s green. Although this is difficult to achieve for human drivers, it can be done consistently by an autonomous vehicle that uses artificial intelligence.
Using artificial intelligence, the vehicle’s speed can be set in such a way that it reaches the next signal in time to pass it without having to wait for the color to change to green.
In their study published on the prepress server arXivIn the study, researchers demonstrate a machine learning approach that can learn to control a fleet of self-driving vehicles in a way that keeps traffic flowing smoothly. Led by graduate student Vindula Jayawardana, the team of researchers says their approach reduces fuel consumption and emissions while improving the vehicle’s average speed.
“This is a really interesting place to step in. There is no better life because they were stuck at a crossroads,” senior author Cathy Wu was quoted as saying by the paper.
But there is another complication. The researchers want the system to learn technology that saves fuel while also reducing flight time. These goals may be incompatible. Wu says that while they want the car to move quickly to save travel time, they want it to slow down or not at all to reduce emissions. These competing scenarios can be very confusing for the learning factor.
As a result, researchers have devised an alternative solution known as reward shaping. They supplied the system domain with information that it could not learn on its own using reward formation. They punished the system in this scenario any time the car came to a complete stop, so it would learn to avoid doing so in the future.
They tested their control algorithm using a traffic simulation platform with a single intersection once they created it. As the cars approached the crossroads, their system didn’t cause any choppy traffic. When cars have to come to a full stop due to traffic in front of them getting stuck, this is known as stop-and-go movement.
More cars went through one green stage in the simulations, outperforming the model simulating human drivers. When compared to previous optimization strategies aimed at avoiding choppy traffic, their approach resulted in higher fuel savings and lower emissions.
If all the vehicles on the road were autonomous and connected to its system, they could reduce fuel consumption by 18 percent and CO2 emissions by 25 percent, while improving travel speed by 20 percent, they say. Even if only 2 percent of vehicles are self-driving, they can deliver at least 50 percent of total fuel and emissions reduction benefits.