The average human driver makes hundreds of sophisticated decisions while driving. If we see a child playing with a ball by the side of the road we’re aware that the child may suddenly dart into the road chasing after the ball. If we see someone texting as they walk towards a crosswalk, we know they might walk into the street without looking up. Good drivers pay attention to people who are distracted, they slow down depending on the situation. This knowledge is subtle, but also incredibly complex.

As we hear more and more about a future world filled with self-driving cars, the big question is – how do we code a driverless car to be able to predict human behavior as we humans do? Researchers at the University of Michigan Ann Arbor are working on a solution that may one day revolutionize how we get from point A to point B. They’re teaching algorithms that run self-driving cars to analyze pedestrian behavior in order to predict where a person will move next. So far, the software is showing signs of accuracy, but for just the next few seconds. Being able to accurately predict a person’s whole trajectory still remains a challenge.

This video was supported by the Alfred P. Sloan Foundation.