Applying social norms to high-fidelity pedestrian and traffic simulations

Smart cities are founded on complex interactions among architectural and urban designs, sensors, actuators and crowds of people with their devices. In this context, simulation becomes essential to study the effects of the technology and to understand how to improve its effectiveness on the social environment. The majority of the current pedestrian and traffic simulations adopt a bird-eye view and are driven by statistical models. While this is enough in many cases, e.g. to study traffic flow under common conditions assuming average cases, it is not appropriate when a higher level of fidelity is required. Simulated people need to show both a plausible behavior and mechanisms to coordinate with human participants in a natural way. Much of this coordination happens silently and is driven by social norms, that may vary according to culture and context. In this paper, we propose an approach to represent social norms in multi-agent systems that enables implicit coordination driven by observations of others' behaviors. This is applied specifically to the case of pedestrian movement. In order to allow for a more effective participation of humans in the simulation, our approach does not use central coordinators or coordination protocol, but rather each agent takes its own decision so to make more realistic interactions. A software architecture and initial experimental results are presented and discussed.