Data assimilation in agent based simulation of smart environment

Agent-based simulation is useful for studying people's movement in smart environment. Existing agent-based simulations are typically used as offline tools that help system design. They are not dynamically data-driven because they do not utilize any real time sensor data of the environment. In this paper, we present a method that assimilates real time sensor data into an agent-based simulation model. The goal of data assimilation is to provide inference of people's occupancy information in the smart environment, and thus lead to more accurate simulation results. We use particle filters to carry out the data assimilation and present some experiment results, and discuss how to extend this work for more advanced data assimilation in agent-based simulation of smart environment.