In this dissertation, agent-based model serves as a low abstraction level model in the
estimation framework. According to [27], agent-based model is referred as “a computational
method that enables a researcher to create, analyze, and experiment with models composed of
agents that interact within an environment.”[27]. The agent-based model has several features
which make it a useful and prominent tool in studying people’s movement behaviors in a building structure. First, the agent-based model has the feature of “ontological correspondence”[27]. An agent in the model can be directly corresponded to an entity of the real
world target system. In the case of the model of people’s movement in the building, each agent can represent an individual occupant moving inside the building. This makes it easier to
represent the whole dynamics of the occupants’ movement from each individual’s behavior. Second, the agent-based model can have “heterogeneous agent”[27]. This means each agent can
have its own preference of behavior or even have its own rule to perform actions. In the case of the model of people’s movement in the building, each agent update its position according to its own destination and velocity. The aggregation of individuals’ choices of behaviors form the overall dynamics of occupants. Third, the agent-based model can represent the environment [27]
where the agents perform their actions. In the case of the model of people’s movement in the
building, the environment is modeled with free spaces, obstacles and sensors and the agent are
designed to interact with these objects. A way point graph is also modeled according to the
structure of the building to assist the agent for navigation. Forth, the agent-based model can
model the interactions among agents [27]. Typically, the interactions among agents can be
modeled as a set of rules to define what messages or actions are passed from one agent to
another. This is particularly important in simulating occupants’ behaviors because the dynamic of occupants’ movements includes collective behaviors such as avoiding each other. Such behavior must be captured by the underlying simulation model in a convenient way. In the agent-
based model proposed in this dissertation, the interactions among agents, i.e. the avoidance
behavior, is implemented using a set of rules. By following these rules, the agents avoid each
other to avoid collision and resolve congestions.
Agent-based model is widely applied in studying people’s movements in different
environment. Pedestrians on a street or occupants in a building can be easily modeled as an
modeled. In [28], a typical agent-based model of pedestrian behavior is presented. This model is
featured with two categories of agents’ characteristics, the social-economic characteristics and behavioral characteristics. The social-economic characteristics of the agent include the individual
properties such as sex, income and are used to create a planned route passing through a sequence
of destinations on the map. The route is created using shortest path algorithm on a series of waypoints. It simulates each individual’s pre-defined plan based on their social-economic characteristics. The behavioral characteristics include agent’s attribute of activities such as speed, visual range and fixation. The speed defines how fast the agent can move in the
environment. The visual range defines the area an agent can observe using virtual sensor. And
the fixation defines how stick the agent will on its pre-defined route. The agents’ moving
behavior of this model is derived from the two categories of characteristics of the agents. In
another work described in [29], author developed an exosomatic visual architecture system for
guiding the natural movement of agents. It uses a grid graph to compute the visual range of an
agent. Based on the visual field of the agent, the agent uses the decision making process to
determine the movement of the agents. The two works presented above demonstrate two
fundamental essences for modeling pedestrians’ movement behavior, that is, the behaviors
originated from social-economic aspect such as navigation and behaviors originated from the
nature movement of people such as the rule-based decision process to avoid obstacles. The
agent-based modeling work in this dissertation borrows the ideas from these works. It models the
navigation behavior and avoidance separately in order to capture the basic dynamics of occupants’ movement. The details of the agent based model is described in chapter 3.
Except being used as tool in studying people’s movements, agent-based model also finds its application in social science. In social science, one of the major challenges that scientist may
face is that the experiment is difficult to conduct. Conducting an experiments on the research
subject requires certain treatment to particular social group. The outcome of social group being
treated is then compared with the social group that is not being treated to study the difference.
Unfortunately, in many cases, such treatments face ethnical issue and cannot be easily carried
out. Agent-based model provide a viable solution to this problem. Because the social entities are
modeled as agents and executed within the context of a computer program, ethnical issues are
naturally avoided. However, agent-based modeling and simulation in social science is different
from its application in natural science in a sense that it does not aim at predicting the future.
Rather, the purpose of application of agent-based modeling and simulation in social science is to
reproduce, understand social phenomena and to develop social theories. It is normally used as an
offline tool without incorporate real time data. This is the fundamental difference between
applications of agent-based modeling in social science and applications of agent-based modeling
in other disciplines. The application of agent-based model in social science is out of the
discussion of this dissertation. However, there exist similarity between the procedures of
building the agent-based model in social science and the agent-based model of people’s
movement behaviors in building.