44 Chapter 5 Representing Pedestrian Behaviors The generic framework is based on the requirement that the dynamics of the interactions of individuals and many components in environment are primary determinants of the complexity behavior emerging in the system. Chapter 4 explained how the objects in simulation system relate to one another. In this chapter we put more emphasis on an individual Mouse, explaining how the agent (mouse) perceives its environment, how the agent (mouse) acts upon it, and how the agent (mouse) selects which action to perform. The latter part presents the computable forms and implemented models of general behaviors found in pedestrian movement. 5.1 Structure of an Agent “Mouse” Object Mouse in this context is an autonomous agent. An agent is a distinct object that has a task to perform and it is situated in an environment that it shares with other entities, perceiving and acting on these other entities. This Mouse is said to be autonomous because it is reactive, instinctively driven by stimulus. Each agent perceives its environment and has its own physical mobility to move about because of its internal structure. The internal structure of an "agent mouse" can be described in terms of three basic components: the external perception, the internal state, and the behavior production system. 5.1.1 External Perception Each agent has an ability to see objects and recognize stimuli in the environment. In this context, each agent's field of vision while walking is 360º. The perceptual system eliminates the objects and stimuli hidden behind the obstacle, as shown in Figure 5.1. Therefore, the non-shaded area in the figure determines the perceived scene and the perceived objects for an individual. The radius of perception has no limit because we assume that size of the environment is relatively small. Thus if there is no obstruction the individual can see through the whole space.45 This field of vision model is based on the concept of isovists (Benedikt, 1979) and Gibson's approach of visual perception (1979). An isovist, or viewshed, is the area in a spatial environment directly visible from a location within the space. Isovists are an intuitively attractive way of thinking about a spatial environment because they provide a description of space from inside, from the point of view of individuals as they perceive it, interact with it, and move through it. As such, isovists have particular relevance to architectural analysis. Benedikt introduced a set of analytic measurements of isovist properties to quantify space by considering the volume visible from a location. He then simplifies this representation by taking a horizontal slice, resulting in an isovist polygon that describes the visible area from a given location. Later he formulated an 'isovist field' by recording a single isovist property for all locations in a configuration by using contours to plot the way those features vary through space. Gibson argued that people see the environment while moving, not just in pauses between movements. That is, vision should be considered in terms of this broader awareness, which involves movements of the eyes, the head, and the body. As we take into account the fact that people experience space over time, looking in different directions, so that they are ultimately aware of their complete surroundings, we endow the agent with 360 degrees vision. However, in this simulation, the agents only have a short term memory, meaning they can only see and register the objects – obstacles and attractions – in the scene to the behavior production system but their abilities do not involve a high level of cognitive behavior. Thus they cannot construct a model or understanding of the environment based on their experience over time. Figure 5.1: Field of vision based on isovist concept46 5.1.2 Internal States Each agent has an internal state that informs its individual internal motivation. The internal state is affected by its external perception, received from the behavior production system. So, the behavior production system modifies the internal state as a result of perception, giving a signal or drive to an agent. The signal indicates any of those motivations needed to be satisfied so that the agent will be in a 'comfort zone'. The internal signal is sent back to the behavior production system for the action-selection process to determine an appropriate behavior. The internal state is defined by a set of variables that can take values between zero and three (Figure 5.2, 5.3). They represent state of hunger, and hurriedness. The degree of hunger increases over time, and can also be decreased if a proper behavior (eating) is executed. There is a significant different between the hunger scale and the hurriedness scale. A lower hunger value indicates a neutral state while a lower rush value becomes a higher value of relax, which may require a proper action to satisfy that need. Relax and rush are on the same scale because both relate to a period of time spent in the environment but in the opposite way. A 'relax' state allows an individual to spend more time, thus more activities, in the space but 'rush' degree is likely to discourage an individual's activity. The values will increase when an individual spends some time in space and they affect the speed (a rushed mouse will walk faster) and moving direction (a rushed mouse will use the shortest path). Figure 5.2: A scale of motivation degree of Hunger Left: a state of very hungry, Right: a neutral state Figure 5.3: A scale of motivation degree of hurriedness Left: high degree indicates rush, Right: low degree indicates relax47 Reflex Reactive Motivated 5.1.3 Behavior Production System Each agent Mouse has a set of behaviors (Table 5.1) registered in the behavior production system. It is a set of behaviors from which the system selects for an agent to perform the action, according to its internal state and the external perceptions at any time step. Wander is a default behavior for aimless agents and pass-through is a default behavior for purposive walkers. When there are no external signals – no station and moving obstacles – from the external perception system, and no signal from the internal state, agents use their default behaviors to drive themselves through the environment. These two behaviors are determined as reflex behaviors because this level of behavior