Identification of decision rules in a human-controlled system: vehicles at a traffic intersection

The rules that govern decision making in systems controlled by humans are often simple to describe. However, deriving these rules from the actions of a group can be very difficult, making human behavior hard to predict. We develop an algorithm to determine the rules implemented by drivers at a traffic intersection by observing the trajectories of their cars. We apply such algorithm to a traffic intersection scenario reproduced in the Caltech multi-vehicle lab, with human subjects remotely driving kinematic robots. The results obtained on these data suggest that this kind of human behavior is to some extent predictable on our data set, and different subjects implement similar rules.