Hybrid agent-based and graph-based modeling for building occupancy simulation

Abstract

Various models have been developed to simulate the dynamics of building occupancy, among which the two popular types of models are agent-based models and graph-based models. Agent-based models have the advantage of handling microscopic behavior and heterogeneity while graph-based models have the advantage of computation efficiency when simulating a large number of occupants. In previous work, we developed a graph-based agent-oriented model that incorporates some agent properties into a graph-based model to simulate a large number of occupants in an efficient way. Built on previous work, this paper presents a hybrid agent-based and graph-based modeling approach that supports simulation using agent-based models in certain areas of building occupancy (e.g., intersection areas where occupants have a lot of interactions) and graph-based models in other areas. We describe the structure of the hybrid model and show experimental results that demonstrate different features of the model for building occupancy simulations.