A collision-avoiding mechanism based on a theory of mind

We develop a collision-avoiding mechanism for a system of individual agents (pedestrians) that move in a crowd trying to reach their different goal points. The agents avoid collisions on the basis of a model of the other agents' behavior, a "theory of mind," which is realized at different levels through an iterative process (the first level, or level 0, corresponds to ignoring the other agents' behavior, level 1 to assuming that the other agents will ignore each other, and so on). The model is conceived in order to perform an evolutionary simulation of some basic parameters that determine the agent's sensorial, cognitive and behavioral system (the perception of the agent's own size, the attraction to the goal, the radius and angle of the field of view and the level of the theory of mind). In this preliminary work we present our model, show that it reproduces some of the simplest organized behaviors of a system of pedestrians, and focus on some features of the theory of mind, as the difference between odd and even levels.