Simulating Human Activities for Synthetic Inputs to Sensor Systems

Abstract

We are developing human activity simulations that could be used to test distributed video sensor networks. Our ultimate goals are to build statistical models of pedestrian density and flows at a number of urban locations and to correlate those flows with population movement and density models represented in a spatiotemporal modeling system. In order to create known populace flows, we have built a virtual populace simulation system, called CAROSA, which permits the authoring of functional crowds of people going about role-, context-, and schedule-dependent activities. The capabilities and authoring tools for these functional crowd simulations are described with the intention of readily creating ground truth data for distributed sensor system design and evaluation.