A graph-based agent-oriented model for building occupancy simulation

A building occupancy simulation simulates the behavior of the occupants and assists in recognizing occupant’s behavior which can be used in energy consumption, smart assist as well as in emergency evacuation. Agent based models are widely used to represent the occupants, but suffer poorly due to increase in computation cost with the increase in agents. Graph based and cellular approaches are used to simulate large mass of occupants, but they treat the occupants as a homogeneous mass and thus, the emergent behavior due to individuality is lost in the model. This paper recognizes the limitations of the existing models that deals with the building occupancy simulation and aims to combine the benefits of agent and graph based modeling to come up with a rational and effective way of building occupancy simulation. This model departs from the existing ones through a difference in a way occupants are treated. The model sees them as entities having their own properties and decision making abilities rather than treating them as a group. This enables the model to handle evacuation systems in places with considerably larger amount of occupants like stadiums and concert arenas. As a case study, we conducted a simulation of emergency evacuation in a certain building with a selected type of agents using our model. The results demonstrate the feasibility of the proposed method, as well as the potential of simulating large mass of occupants within reasonable computational resources.