2 RELATED WORK
2.1 Occupancy estimation
2.1.2 Occupancy estimation for evacuation planning
Occupancy estimation is useful in the case of evacuation of the occupants in the building.
When an emergency such as fire or disaster occurs in the building, occupants in the building stop
their activity and try to move to the exit to escape from the situation. During the evacuation of
occupants, the location information or the position distribution of occupants becomes important
to retrieve. This information can be utilized by first responder such as firefighters to realize the
situation in the building and deploy rescuing resources more efficiently. In [17], the author
proposed an approach with multiple models to estimate building occupancy during the
evacuation of occupants from the building. This approach is based on three different models: an
agent based model, which includes the detailed description about each individual occupant’s
velocity, behavior and trajectory; a coarse model, which represents the building structure and
traffic dynamics using a graph; a kinetic model which models the congested areas of the building
author uses Extended Kalman Filter to estimate the zone level occupancy predicted using the
coarse model and kinetic model. One important conclusion draw from the experiments by the
author is that the estimation result based on sensor combined with model is significantly better
than the estimation result based purely on sensor measurement. Another conclusion is that agent-
based model is not suitable for real time estimation due to the computational cost while the
coarse and kinetic model are more efficient and can be used in real-time estimation. The
framework of the method proposed in this work is similar to the framework of the method
proposed in this dissertation that is, given a set of sensors where various observation data can be
derived from and a set of models representing the dynamics of the occupants in the building, an
estimator is utilized to meld the model and observation to estimate the occupancy state. In [40]
the author proposed a simulation model for simulating emergency evacuation. The model is
featured with a GIS component and a C++ simulator. It uses the concept of micro-simulation to
simulate the detailed movements and positions of the moving vehicles. The positions of the
vehicle is modeled as locations which are road segments the vehicles can move into. A set of
locations becomes arc and form the road network. The vehicles can move only when its next
location has free capacity available. This paper gives an example on how micro-simulation can
be used in helping emergency evacuation planning. In [41], the authors propose to utilize a
wireless sensor network to detect hazard in the building and navigate occupants to avoid the
hazard. This approach is based on a navigation graph and a hazard spread graph of the building.
The navigation graph represents the time for the occupants to arrive at each node of the sensor
network and the hazard spread graph represents the time for the hazard spread to each of the
nodes so that the region covered by that node becomes hazardous. The safety of each scheduled
time of the current node. In [42], a model called SIMULEX is proposed to simulate occupants’
evacuation behaviors in large buildings. It first uses the program of DRAWPLAN to input the
floor plan of the building including rooms, furniture, walls and exits. It then uses program
GRIDFORM to create a distance map of the building. In the distance map, the floor plan of the
building is divided into blocks and each block is assigned with a value to indicate the shortest
distance from this block to the nearest exit of this block. SIMULEX uses three circles to
represent the shape of the human body, this makes it easier to calculate the distance between two
persons that are in close proximity and to calculate the overtaking of one person to another.
In[43], the author build an agent based model to simulate occupants evacuation in large public
building. This model is featured with two different sub models- the spatial environment model
and the agent decision model. The spatial environment model handles the structural information
of building environment, the layout of the building, the source of the fire and the dynamics of
toxic gas generated by the fire. The agent decision model handles the agents’ behaviors according to the environment and other agents. This includes the route planning, decision of the
speed and how to resolve confliction between agents when two agents intend to enter the same
area simultaneously. In[44], author build an evacuation model based on a graph with each vertex
representing a room, a segment of corridor or hall ways and each edge representing a pipeline
that occupant can transport on. The transportation of occupants on the edge of the graph is
determined by factors such as social affiliation, access visibility, tenable time, speed, flow rate
and distance between rooms.
Most of the methods listed in this section are specifically designed for the investigation of
occupancy dynamics under the scenario of emergency such as a fire or other hazard. Some of the
methods establish dynamic model for investigating the evacuation of occupants, some of the
methods utilize both models and data derived from sensor to estimate the occupancy then
provide the result of the estimation to first responders to make evacuation plan. Intuitively, the
method that utilizes both sensor data and simulation model such as the one proposed in [17]
yields best result for estimating the state of occupants in a building.