Frequent traffic light stops drive up fuel consumption. In the age of automated driving, this could change. AI algorithms could optimise traffic lights so that autonomous vehicles hardly need to stop at all. //next has taken a look at how such an AI-controlled "green wave" could work.
A team of researchers at the Massachusetts Institute of Technology (MIT) has conducted a study on how machine learning can be used to control a fleet of autonomous vehicles so that they do not need to stop at traffic lights. In the simulations, this reduced fuel consumption and emissions while actually increasing the average vehicle speed.
This approach achieved the best results when the participating vehicles were all driving autonomously. But even if only 25 per cent of the cars use the MIT control algorithm, this already leads to significant fuel and emission benefits.
While humans may drive past a green light without giving it much thought, intersections can present billions of different scenarios depending on the number of lanes, how the signals operate, the number of vehicles and their speeds, the presence of pedestrians and cyclists, etc.
The team approached the problem using a model-free technique known as deep reinforcement learning. Reinforcement learning is a trial-and-error method where the control algorithm learns to make a sequence of decisions. It is rewarded when it finds a good sequence. With deep reinforcement learning, the algorithm leverages assumptions learned by a neural network to find shortcuts to good sequences, even if there are billions of possibilities.
How the process works in detail - and what obstacles the research team had to overcome and still has to overcome - you can read here on MIT News:
Text: Ingo Schenk