As if idling in a line of cars at a red light forever wasn’t bothersome enough, vehicles emit greenhouse gasses while they’re stopped in traffic. Not only that, exposure to excess vehicle emissions while idling in traffic can be a major health risk.
What if drivers could time their trips so well that they never hit a red light and just kept going through greens? This happens to most of us only once in a blue moon, but researchers think that autonomous cars can use machine learning to more commonly avoid major traffic.
A new study from the Massachusetts Institute of Technology demonstrated that several autonomous cars controlled by a machine-learning program could pass through an intersection and help traffic flow more smoothly.
The research team, lead author Vindula Jayawardana and senior author Cathy Wu, found that their simulations showed a reduction in fuel consumption and emissions while also improving average vehicle speed. The system works best if all cars are autonomous, but even if only a quarter of cars on the road are autonomous it still substantially reduces fuel consumption and emissions.
“This is a really interesting place to intervene. No one’s life is better because they were stuck at an intersection. With a lot of other climate change interventions, there is a quality-of-life difference that is expected, so there is a barrier to entry there. Here, the barrier is much lower,” says Wu.
Mathematical models for solving traffic problems usually focus on one ideal intersection, but traffic around a city is very complicated with billions of different possible scenarios. Wu and Jayawardana used a model-free technique called deep reinforcement learning to allow their algorithm to make a series of decisions and is rewarded by good decisions, so it learns as it goes. But they also wanted to reduce fuel consumption while optimizing travel time.
“To reduce travel time, we want the car to go fast, but to reduce emissions, we want the car to slow down or not move at all. Those competing rewards can be very confusing to the learning agent,” Wu said.
For this, they used a technique called reward shaping. This helps by, for example, penalizing the system when it comes to a full stop so it doesn’t do that again and finds a way to drive without stopping.
Further research is needed, but early simulations show a great reduction in traffic and emissions with the new system. Wu and Jayawardana’s system might create autonomous vehicle guidance which not only optimizes our traffic and reduces emissions, but helps with our public health, even if it’s just by reducing road rage.