On paper, travel by car should be relatively efficient. Everyone goes the speed limit, uses their turn signal, maintains proper spacing, and follows the rest of the rules of the road. Unfortunately, humans are bad drivers. A single incorrect merge or tap on the breaks can cause massive backups as it propagates through the rest of the vehicles.
With a future of self-driving cars, these human tendencies should be taken out of the equation. Besides simple rules like green means go, stay in your lane, and don't hit the person in front of you, driverless cars use algorithms to get from point A to point B. Computer Scientists in Singapore's Nanyang Technological University have developed a new routing algorithm with the hopes of reducing these "spontaneous traffic jams."
These algorithms are part of the scientific field known as graph theory and are immensely complex for computers to calculate. Rather than try to calculate every possible path and situation the car could be in, the researches started by assuming that people would inevitably do dumb things. Their goal is to then minimize this traffic "breakdown" period by intelligently routing the overall traffic so that over the entire network, no significant breakdowns occur.
By absorbing these small traffic issues, they can create a smooth driving experience over the entire area. They were able to turn their model into a math problem and with machine learning, they were able to create this new algorithm. They estimate only about 10% of vehicles need to be equipped with it to beat traffic. Until it's widely available though, remember, you're not in traffic; you are traffic.
https://www.techspot.com/news/68675-scientists-new-vehicle-ai-algorithm-can-beat-traffic.html
