insufficient for today's modern buildings due to their lack of flexibility. Evacuation
systems which are prepared in accordance with pre-defined evacuation scenarios
are not capable of routing according to the knowledge of what is inside the build-
ing during and after the occurrence. This may lead to direct people to the paths
which are closed or have gas leaks.
Emergency situations are not static events but rather dynamic and uncertain.
An ideal evacuation and routing system should be capable of taking into account
and evaluate the status of users or user groups and produce special evacuation
instructions according to these users or user groups. The stage in which people
spend most of time in case of emergency is not reacting or taking action but rather
the stage of realizing the event before starting to move. Uncertainty at the time
of the emergency and the lack of clear information about the incident are factors
in delaying the evacuation of the building. Therefore, a system that can provide
understandable and clear information to all users in real time and resolve concerns
of them will surely shorten the evacuation process. Such an ideal system is a smart
evacuation system that should avoid congestion by sharing people in different
paths or guide people to areas of risk (smoky and dangerous) to be taken in cases
of necessity without the need for the user to determine the route and allow them to
progress rapidly without hesitation.
To realize an ideal intelligent indoor evacuation system, a number of main
functionalities should be addressed. These functionalities are a spatial database for
the management of large spatial datasets, 3D GIS based routing engine centralized
in an appropriate host, mobile based navigation software for passing user related
data to the host and present routing instructions clearly to the user, an accurate 3D
indoor positioning system and a well organized wireless communication and sen-
sor network architectures inside building.
4.1 Preparing Data
Beginning from this section, we will focus on how we make routing process of
an indoor evacuation system gain intelligence. The aim of our proposed model is
to take into account the environmental and user specific variables in case of fire
occurring in a building and generate evacuation instructions needed till the user
reaches the exit in safety by predicting the usage risk of links on transportation
network of the building. The environmental and user specific variables affecting
fire response performance have been taken as input factors of neural network pro-
posed in this study to build our intelligent evacuation model. To create our univer-
sal data set we firstly have formed the risk levels for each factor and transformed
their values into [1-5] system (Table 1 ).
With meaningful combinations of 16 factors given above we had
14,817,008 records which constitute the universal set for our problem. For calcu-
lating the risk score of each record in universal set properly, we had weighted each
factor to provide priority order. We have used corresponding value of Fibonacci
series for each priority order as shown in Table 2 .