1. Introduction
Many geospatial researchers have been interested in understanding and analyzing real-time human movements in micro-scale environments after the 9/11 attack in New York City 2001 and the 7/7 London bombing in 2005 (Cahan and Ball, 2002; Cutter et al., 2003; Kwan, 2003; Kwan and Lee, 2005; Pu and Zlatanova, 2005; Lee, 2007). Such disasters not only affect the immediate environment of multi-level structures in urban areas, but also impact upon their indoor space in ways that considerably reduce the speed of emergency response (Kwan and Lee, 2005). Reducing rescue time can have a significant impact on evacuation in disaster environments. However, most current GIS-based emergency management systems have been developed using 2D GIS to analyze geographical problems with 3D visualization systems to display 3D urban space. The systems have limitations in representing the internal structure of built-environments in 3D space and in analyzing human movements during emergency situations in indoor space.
The emergency response systems are one of the timecritical applications (TCA) which are decision supporting systems in emergency situations. As TCA, these GIS-based decision support systems require appropriate data management and efficient data acquisition and integration to assist decision makers whenever they need to make a decision in real-time. To develop real-time emergency response applications, these systems should incorporate dynamic geospatial data about emergency situations, including changes in evacuation status over time. This includes a navigable 3D GIS, a dynamic geospatial database, within-building positioning in real-time, analytical models to simulate possible trajectories of change and to formulate alternative decision scenarios, and a distributed information architecture (Pu and Zlatanova, 2005; Lee, 2007).
This paper seeks to further develop real-time emergency response systems. It focuses on proposing a Spatiotemporal Optimal Route Algorithm (SORA) based of a set of relevant temporal parameters. The algorithm is integrated with real-time monitoring systems in order to manage dynamic geospatial entities, including dynamic capacities and flow rates of hallways in multi-level structures. Temporal database entities are used to identify optimal routes and evacuation bottlenecks in buildings within the network in real-time emergency situations. This article first reviews current researches related to optimal routing methods for emergency responses in micro-scale environment. This is followed by a description of the proposed spatiotemporal optimal routing algorithm with detailed explanations and implementation. Finally the paper closes with conclusion and recommendations for future research.
2. GIS-based Emergency Response Systems
The analysis of human spatial behavior in urban environments has recently been extended from large regions at the metropolitan or urban scale to small areas (including the micro-space within buildings), and from the aggregate level to the individual level (Church and Marston, 2003; Kwan and Weber 2003; Lee, 2007). Suppose a fire occurs on a floor within a building. The situation will cause traffic bottlenecks and create danger zones within the building as a result of people bumping against each other or broken parts of the building (as obstructions) caused by fire and smoke. In this situation, an emergency response (ER) system should monitor the situation and guide evacuees or rescuers in real-time, because conditions inside the building can change suddenly over time. To find an optimal route, the system requires an indoor accessibility model that represents navigable spaces and their spatial relationships (Lee, 2007), as well as a Ubiquitous Sensor Network (USN) based monitoring system that keeps track of the changes in the physical environment and the movement patterns of people in the building - which are dynamic factors that affect the optimal evacuation route under emergency situations. As shown in Fig. 1, the architecture of a real-time ER system has three components: Data Management, Analysis, and Dissemination. The Analysis component contains two modules: a Context-based Query Module for accessing the spatial data, and a Context-based Spatial Analysis Module for analyzing the spatial data to clarity the emergency situations and to guide evacuees and rescuers to evacuate the buildings. The results of the analysis will be visualized on PDAs, mobile phones or others mobile devices through wireless communication.
Fig. 1.Architecture of a real-time emergency response system
Considerable developments and researches have sought to develop navigable three-dimensional (3D) data models and accessibility analyses to facilitate effective emergency response (e.g., Lee, 2007; Kwan and Ransberger, 2010; Chen et al., 2012). Kwan and Lee (2005) evaluated the speed of emergency response in an emergency situation in order to assess the potential benefit of a 3D GIS for improving the speed of response within micro-scale environments. Their results indicate that extending conventional 2D GIS to 3D GIS representations of the internal structures of high-rise buildings could significantly improve the overall speed of rescue operations. The findings have motivated some geospatial scientists to develop intelligent emergency evacuation systems for complex buildings using 3D GIS (e.g., Meijers et al., 2005). Lee (2007) developed 3D Navigable Data Model (3D NDM) based on the 3D Geometric Network Data Model to represent pedestrian movement behaviors within buildings or micro-scale environments. The model is integrated with Intelligent Transportation System (ITS) technologies, which is called an Intelligent Emergency Response (IER) system (Lee, 2007).
Currently most developed and implemented emergency response systems are based on predefined evacuation plans that consider mainly ideal situations without regard to previous occurrences or real-time events within buildings (Pu and Zlatanova, 2005; Lee, 2007). Because these systems are thus not intelligent emergency response (IER) systems, a real IER should incorporate dynamic geospatial data about the emergency situation, including changes in evacuation status over time (Lee, 2007). It needs to incorporate GIS capabilities and functionalities for representing the internal structure of micro-spatial environments (e.g., internal structure of buildings) and conducting GIS-based analyses to provide real-time navigation guidance both for reaching the disaster site and for negotiating within a complicated structure under emergency situations.
One challenging task in developing a real-time emergency response system is to compute an optimal route while taking into account dynamic geospatial factors in emergency situations. The classic problem of finding the shortest path over a network has remained an important research topic within many disciplines including transportation, geography, computer science and operations research (e.g., Dijkstra, 1959; Zhan and Noon, 1998; Delling et al., 2013; Azevedo et al., 2014). These research efforts have produced a number of shortest path algorithms. Yen (1971), for instance, first proposed a k shortest path algorithm to define a set of k-paths, which are required for specific purposes such as route guidance, road construction, and hazardous material transportation (Lim and Kim, 2005). In 3D space, Scott (1994) implemented a shortest path algorithm for an unindexed 3D voxel space using a cumulative distance cost approach. This approach produces a set of voxels such that each voxel contains an attribute about the cost of traveling to that voxel from a specific start point, assuming uniform friction of movement throughout the representation. The 3D shortest path algorithm moves through the “cost volume” along the steepest cost slope from target to origin using a 3 by 3 by 3 search kernel (Raper, 2000). Kirkby et al. (1997) implemented a modified version of the Dijkstra shortest path algorithm in a 3D GIS, in which the gradient over a 2.5D surface was added into the computation.
Some transportation researchers have introduced realistic conditions of time-varying flow and congestion within the transportation network for computing a dynamic version of the shortest path using the dynamic traffic assignment algorithm (Wu and Miller, 2002; Okabe and Kitamura, 1996; Zhang, 2013) or shortest path computation for time aggregated graph algorithm (SP-TAG Algorithm) (George et al., 2007). These algorithms, using spatiotemporal network database to deal with vehicle trip flows through a network in a link-by-link or node-by-node basis in successive time intervals, compute the shortest path in the given network for a given start time at the source node and the least travel time route over the entire time period. Even though the algorithms are implemented with a real world static network appended with a synthetically generated travel time series (George et al., 2007), they are limited in computing a real-time optimal route with temporal parameters to guide rescuers in emergency situations because the algorithms used historical spatiotemporal data saved in DB, but they did not use real-time data directly detected by real-time moving object tracking systems. Therefore, this paper focuses on developing a spatiotemporal optimal route algorithm for micro-scale emergency situations that is capable of using real-time data, which is one of the core functionalities of a GIS-based emergency response system.
3. Spatiotemporal Analysis for Emergency Responses in Indoor Space
To implement spatiotemporal analysis for emergency responses in indoor space, it needs to incorporate GIS capabilities and functionalities for 1) representing the internal structure of buildings integrated with dynamic geospatial databases and 2) conducting GIS-based analyses to provide real-time navigation guidance.
3.1 Modeling indoor space
To represent the internal structure of buildings in an ER system, two models are required. One is a geometry model to represent the geometry of the internal structure. The other is a topological model representing the connectivity relationships among 3D entities to be used to define real-time optimal routes. The geometry model represents internal structures of a building with several types of objects including room, wall, door, window, stairways, exit and so on. This model can be used for users (rescuer, evacuee, etc.) to understand interior of building and to visualize the indoor space in the system. However because most geometry models, including IFC models, CityGML LoD4 models, and KML models, don’t represent topological relationships among spatial objects, topology models are needed to conduct topological analyses on spatial-temporal geographical problems in order to define optimal routes within buildings.
In order to represent topological relationships among 3D objects, two methods are introduced, one is based on geometric primitive relations used in 3D Boundary Representations (B-rep) model, and the other is based on graph-based representations. The 3D B-rep model has some limitations when applied to navigation problems, which includes geometry computation problems, topological consistency problems and data volume problems (Lee, 2004). Because of the limitations, a network-based topological data model had been proposed.
In what follows, we describe a network-based topological data model based on the work of Lee (2001). We began with describing the Node-Relation Structure (NRS), and then outline a logical network data model and a geometric network data model, which can be used for optimal routing analyses. The NRS is a dual graph representing the topological relationships among 3D entities by Poincaré Duality (Kolbe et al., 2008). As illustrated in Fig. 2(b), connectivity relationships in the NRS are modeled as a network structure, the dual graph of 3D spatial units. A graph H = (V(H), E(H)) consists of two sets: a finite set V of elements called vertices and a finite set E of elements called edges, V(H) = {v1, ……, vn} and E(H) = {(vi, vj) / vi, vj ∈ V}. Each edge is identified with a pair of vertices (v1, v2), where v1, v2 ∈ V. The NRS is a logical data model, which is a pure graph representing the adjacency, connectivity and hierarchical relationships among the internal units (e.g. rooms and corridors) of a building. It does not represent the geometric properties (e.g., distance or size) of and among these units. In order to implement network-based analysis such as path finding, allocation, tracing and spatial interaction analysis in the NRS, the logical network model needs to be complemented by a geometric network model that accurately represents these geometric properties. For the transformation, node n6 in the NRS (graph H) representing a corridor within the building (Fig. 2(c)) is considered as a consolidated “Master_Node”. The Master_Node is a sub-graph representing a connectivity relationship among the compartmentalized zones of the corridor (Fig. 2(b)) generated to represent the relationships between a room and a hallway as one-to-one relations (one-to-many relations in the NRS). In other words, node n6 in the NRS (graph H) is converted into a linear feature (edge e1 in Fig. 2(c)) (Lee, 2004) in the GNM which is a sub-graph representing a two-dimensional shape such as a hallway. The GNM has been used to measure the relative accessibility of emergency response in order to evaluate the potential benefit of a 3D GIS for improving the speed of emergency response within microscale areas (Kwan and Lee, 2005), as well to develop an intelligent emergency evacuation system for complex buildings using 3D GIS (Meijers et al., 2005).
Fig. 2.Internal building models
In the study, we utilized the GNM to propose Accessibility Graph Model (AGM) not only to represent the connectivity relationships among navigable spaces in the building but also to represent the locations and characteristics of doors in topographic space, which is one of the important elements that determines optimal evacuation route under emergency situations (Pu and Zlatanova, 2005). Also, the AGM represents the connectivity relationships among sensor-covered zones in the building. The navigable space in the building is decomposed according to signal characteristics such as propagation and signal coverage areas. Sensor space is characterized by different localization techniques such as Wi-Fi, RFID or smoke detectors which differ in signal propagation or signal coverage.
In the AccessibilityGraph N, each node represents a space object such as rooms, hallways, exits or sensor covered zones. Edges link between nodes that represent connectivity relationships between two space objects. Fig. 3 shows the classes of the network model with a UML diagram. NodeN and EdgeN classes are associated with Network class. The diamond headed arrow pointing to the NetworkN indicates the composition relationship, and the composition cardinality is indicated in the diagram. The NodeN class has three subclasses: NodePt, DoorPt and EventPt. The NodePt class represents subspace of rooms and hallways. DoorPt represents doors, and EventPt class represents bottlenecks or disaster sites within a building. The relation between these subclasses and NodeN is an inheritance relationship and is represented by an arrow. The nodes in the model representing a sub-space refer to the sensors’ coverage (such as RFID Reader’s coverage, heat and smoke detectors’ coverage) inside the building (Lee, 2007).
Fig. 3.Unified modeling language object diagram of accessibility graph model
The edges (called EdgeN) connect either NodePts or DoorPts or EventPts. In this experimental implementation, the NodePt in the database contains the following attributes: an identifier (Node_ID), position data (x-y coordinates), a node type, a traffic cost (Node_Impedance), room use, occupancy data and a sensor identifier (Sensor_ID). The edge attribute consists of an identifier, a start node (Node_ID), an end node, a length (Length), traffic cost (Edge_Impedance), number of occupants, corridor capacity, and walking speed. The database schema for attribute data of class DoorPt and EventPt is as follows:
NodePt (Node_ID, x-y_Coord, Node_Type, Node_Impedance, Sensor_ID, RoomUse, Occupancy) EdgeN (Edge_ID, Length, Edge_Impedance, Flow_Rate, Occupancy_NO, Walk_Speed) DoorPt (DoorPt_ID, Occupancy, Flow_Rate, Node_Impedance) EventPt (EventPt_ID, Occupancy)
3.2 Spatiotemporal optimal route algorithm
In order to define and develop the analytical methods to be implemented in the planning/decision process, a spatiotemporal optimal routing process requires several important functionalities including geo-coding, map matching, traffic flow analysis, and shortest path analysis (Dane and Rizos, 1998; Miller and Shaw, 2001), as shown in Fig. 4. The location positioning event helps locate rescue personnel and disaster situations within the reference data (the network representation of a building). In other words, it translates a given room name or number (a postal address) into a geographic position, with x, y, (z) coordinates (generated by the geo-coding method), and vice versa (a reverse geo-coding method) within the building.
Fig. 4.Process flow diagram of spatiotemporal optimal routing
The map matching event assists in snapping the position of bottlenecks or disaster sites to the nearest location on a network segment when their reported or estimated location is outside the network (Xiong, 2000). The network connectivity analysis facilitates the defining of isolated networks or areas, which do not have an exit node connected to destination nodes due to blockage resulting from traffic congestion or blockage. The problem of finding isolated networks can be dealt with as a network partitioning problem, which partitions a network into separate subnetworks with a minimum number of connections between the resultant sub-networks (Cova and Church, 1997). The other function requires an optimal routing model to estimate the dynamic capacities and flow rates of hallways and stairwells (for EdgeN) per time period, as well as to identify traffic bottlenecks inside the building that are locations of congestion in the network during an evacuation event. The final event is an indoor navigation function utilized to identify feasible and safe routes within a multi-level structure and to provide navigation guidance for rescue personnel. This is a challenging task to accomplish in real time as an emergency situation unfolds since the real-time emergency response system must process evacuees’ current locations, while continuously updating occupancy movements and traffic flow impedances, in addition to identifying the optimal route from the disaster site to the safe location (Lee, 2007). Due to the limited scope of this study, this paper focuses on developing a Spatiotemporal Optimal Route Algorithm (SORA) applied in the internal structures of built-environments.
For calculating the optimal routes for rescuers, the optimal routing method needs to be integrated with USN-based tracking systems (Ubiquitous Sensor Network) in order to monitor dynamic geospatial entities, including dynamic capacities and flow rates of hallways per time period. The traffic impedances can be classified into two types: those associated with the physical environment and those associated with human behaviors. They are the dynamic factors that determine the optimal evacuation route under emergency situations (Pu and Zlatanova, 2005). Relevant environmental factors include damage status, toxicity status, power status and traffic capacities on/at halls, hallways, doors and stairs of the affected area. Human factors that affect people’s speed of movement include network flows, population density, age and gender, level of disability, terrain effects, and so on.
In graph theory, the shortest path problem is the problem of finding a path between two nodes such that the sum of the weights of its constituent edges is minimized. In this case, the vertices represent location and edges represent segments of road and are weighted by the time needed to traverse that segment. A weighted graph H = (V(H), E(H)) is represented by two sets: a finite set V of elements called vertices and a finite set E of elements called edges, V(H) = {v1, ……, vn} and E(H) = {(vi, vj) / vi, vj ∈ V}, and a real-valued weight function f: E → R). Each edge is identifi ed with a pair of vertices (v1, v2), where v1, v2 ∈ V. Formally, given a weighted graph, a start vertex v and a destination vertex u of V(H), find a shortest path P from v to a u of V(H) so that, Σf(p) is minimal among all paths connecting v to u where p∈P.
In order to calculate the spatiotemporal optimal path (Op, Eq. (1)) from a starting point s to a destination d in the indoor network for a building, the problems can be written in mathematical formula as follows:
The spatiotemporal optimal route Op (s, d) from a starting node s to a destination node d is determined by the combination of the optimal routes (Opt (v, u)) for a time lag (for example, m seconds). In other words, the optimal route Opt (v, u) is defined as the path from a node v to u at time t where Opt (v, u) ∈ Opt (v, d). The Opt (v, d) is re-calculated with a time lag and the real-valued weight function f: E → R is also re-computed by a time lag. The process will be stopped when u is reached to the destination node d. The Op (s, d) is defined as the union of the optimal route Opt (v, u).
Fig. 5 shows the Spatiotemporal Optimal Route Algorithm (SORA) process as a flowchart. The input data of the optimal route algorithm is an internal building model and the dynamic traffic impedances, which are measured using population density, capacities of hallway, stairways and doors, as well as damage status and toxicity status. After setting up entrances (starting points) and a target area, optimal routes are calculated for each entrance. The process is iterated every second until the optimal route is defined from an entrance to the target area. In the process, the dynamic parameters are updated from real-time USN-based monitoring systems by second. According to the updated parameters, the accessibility graph attribute data are updated every second. As a result, the optimal route for the rescuer can change swiftly over time according to the updated network model. The simplified pseudo code of the Spatiotemporal Optimal Route Algorithm (SORA), which is modified from the Dijkstra shortest path algorithm, is as follows:
Fig. 5.Flowchart of spatiotemporal optimal route algorithm
Spatiotemporal Optimal Route (ST-Route) Algorithm input: Network N, Number of Exit People E, Density of Smoke S output: Optimal Path from an exit to a danger area R
ST-Route(Network, start) while(every 1 sec) examine the Ei and Sk if (Ei>MaxNumber Ei || Si>Threshold Si) don't go Ri-1 ST-Route(Network, start) update and notify to rescuer until find dangerous people (Area R)
ST_Route(Network, start): for each node n in Network: dist[n] := infinity previous[n] := undefined for each edge e in Network: weight[e] := calculated dist[source] := 0 Q := the set of all nodes in Network while Q is not empty: u := node in Q with smallest dist[] remove u from Q for each neighbor v of u: alt := dist[u] + dist_between(u, v) * weight[e(u,v)] if alt < dist[v] dist[v] := alt previous[v] := u return previous[]
After completing the process for each entrance, one of the defined optimal routes is selected through evaluating the required time to navigate from an exit to the destination (Area R).
4. Experimental Implementations
To evaluate the proposed Spatiotemporal Optimal Route Algorithm (SORA), an experimental implementation of the spatiotemporal optimal routing system was undertaken. Components of the systems were embedded in a Java development environment. The study area used to test the proposed optimal route algorithm is 21C Building, located at the University of Seoul in South Korea. The geometry data for representing the internal structure of a building were CAD-based data. To represent the connectivity relationships among navigable spaces within the test area, an accessibility graph model (AccessibilityGraph N) was used. In the accessibility graph model, the node represents the sub-space of rooms in the geometry model, which is defined by sensors’ coverage (such as RFID Reader’s coverage, heat or smoke detectors’ coverage).
In order to implement the real-time emergency response system, we need to collect the locations of the evacuees using indoor positioning techniques, such as RFID techniques, to calculate the ‘Occupancy’ rate, which represents the distribution of the persons inside building during the emergency situation. However, because of the limited scope to evaluate the proposed algorithm, instead of using real-time moving object tracking systems, the results of an evacuation simulation system and a smoke simulator were used to identify the movement patterns of human beings in the emergency situation due to the limitations of equipped systems in the study area as seen in Fig. 6. Smoke simulation is performed by a Fire Dynamics Simulator (FDS), which is a simulator for visualizing how fire and smoke spread out when the origin of fire, exit and other factors are fixed. We estimated fire and smoke status and used them to calculate weights (for optimal routing process) per second. Evacuation simulation data was generated by using Particle Swarm Optimization (PSO) with the smoke dynamic simulation data generated by FDS, which could be used to evacuation simulation as an infl uential dynamic factor on evacuation. The temporal database contains the evacuation simulation data and the smoke simulation data created by the simulators per second.
Fig. 6.Data scheme of optimal routing system
Based on the proposed internal building models, the data were generated for the study building as shown in Fig. 7. The building model is a dual data model, which includes a geometry model to visualize the study area (Fig. 7(a)) and an accessibility graph model integrated with the temporal database to implement the spatiotemporal optimal routing analysis in the building.
Fig. 7.Internal building models of study building
Fig. 8 illustrates the map matching between the sensor coverage defined in the geometry model and the node in the accessibility graph model. In other words, the ‘Occupancy’ value at time t of NodePt object is assigned based on the number of persons being within the sub-space of the room defined by the smoke detector sensor’s coverage at time t.
Fig. 8.Map matching between the geometry model and accessibility graph model
The temporal database includes the results of the evacuation simulation and of the smoke simulation per second. The simulated data was used to implement the Optimal Routing Process for rescuers. Fig. 9 shows the results of the Fire Dynamic Simulator (FDS) and Fig. 10 depicts the results of the evacuation simulation. Estimated smoke densities were used to calculate traffic impedances (for optimal routing process) per second. Suppose there is a fire in the hallway in the third floor of the study building (marked ‘Fire’ in Fig. 9(a)), Fig. 9(b) shows the distribution of smoke density after 80 seconds. The results of the simulation were used to calculate the ‘Node_Impedance’ at time t (t increased by two seconds), as well to identify the dead zones, which are isolated zones with no feasible (or very difficult) escape routes and which evacuees should avoid in order to be able to evacuate the building safely. These results were used to run the evacuation simulator (called Na-ga-yo Simulator), for which the location of the dead zones defined from the smoke simulation was used an influential dynamic factor for identifying the optimal route. Fig. 10(a) illustrates the distribution of evacuees in the initial stage (about 400 evacuees), and Fig. 10(b) depicts the distribution of these evacuees after 60 seconds.
Fig. 9.Results of FDS smoke simulation
Fig. 10.Result of evacuation simulation
After 160 seconds, most evacuees escaped from the building, but about 40 persons couldn’t find exits and paths to get out of the building and were stuck in an isolated zone on the fourth floor, which was identified by the smoke simulation (where the density of smoke is over the critical value). We need to dispatch rescuers to guide the evacuees who were trapped in the isolated zone. The isolated zone (marked in Fig. 11(b)) is defined as a destination in order to define the optimal route in this study.
Fig. 11.Identifying an isolated zone
Suppose there is a fire in the hallway of the third floor of the 21C Building. The data detected by sensors is transmitted to the USN-based monitoring system, and an emergency call is sent to an emergency control center. The firefighters or rescuers will be dispatched immediately and arrive at the building after 60 seconds. To get to the disaster site (the isolated zone) in the building, the emergency control center needs to identify the disaster site and the optimal entrance and the path (Lee, 2007). Fig. 11 illustrates the defined isolated zone after 160 seconds. As soon as the rescue destination (bottlenecks or isolated zone) is identified, the real-time emergency response system needs to compute the optimal routes from exits to the destination for rescuers.
Fig. 12 shows the results of applying the SORA to the study area. A person is visualized as a dot, and scattered points represent the locations of evacuees. Exits are labeled with squares. Fig. 12 illustrates the results of the experimental implementation in four time slots. From the figures, the locations of evacuees changed and optimal route is also re-defined in each stage. In addition, the locations of evacuees and the smoke densities of sensor nodes also changed each time. Because the dynamic parameters (smoke density, speeds of evacuees, etc.) changed, the impedance of each link is updated. At the initial stage, a shortest path from the exits to the destination is defined as the route from Exit A to the destination. It is the same path at the stage after 60 seconds. After 102 seconds, the optimal route is re-defined as an alternative path from Exit B to the destination as seen in Fig. 12(d). After almost three minutes when all of persons evacuate the building, the rescuers go into the building through the route from Exit D to the destination where about 40 persons are isolated and need to be guided to safety areas outside of the building.
Fig. 12.Results of implementation
The results of the experiments are summarized in Fig. 13. It shows the travel times from each exit to the destination (labeled ‘Isolated Zone’ in Fig. 12) taking into account the delayed arriving times. The x-values represent rescuers’ delayed arriving times at each exit of the disaster building (entrances of the building) after firefighters have been dispatched, while the y-values represent travel times from each exit to the destination. The total travel time it takes to reach the destination with no delayed time is about one minute from Exit A and one and half minutes from other exits. In the experiment, there are two travel time peaks at 40 seconds and 110 seconds (rescuers’ delayed arriving times) because of unexpected traffic blocks inside the building. Further, the results suggest that in early stage it is better for rescuers to access the destination from Exit A, but after 2 minutes the travel time from Exit A is longer than the travel times from other exits. After 66 seconds, the optimal route is the path from Exit B to the destination as seen in Fig. 12 and the shortest travel time is about one and a half minute (see Fig. 13). This experiment demonstrates that because the optimal route may change due to changes in the dynamic traffic impedances, a real-time emergency response system that takes these changes into account can improve the overall speed of rescue operations. In addition, real-time information about the building’s feasible route disseminated to rescuers will help improve the speed and effectiveness of the evacuation process. This means that the rescuers can arrive earlier at isolated zones where some people are stuck because they couldn’t find exits and paths to get out of the building. This may have the effect of saving many lives in real-world emergency situations.
Fig. 13.Comparing analysis between delayed arriving time and travel times from each exit
5. Conclusions
This study shows that capabilities to take into account detailed dynamic geospatial data about emergency situations, including changes in evacuation status over time, are essential for emergency response systems. In order to deal with dynamic changes and uncertain emergency situations and to determine the most efficient way to evacuate buildings, a number of functionalities should be implemented in these systems - including GIS data handling, dynamic geospatial database, real-time data acquisition, and dynamic analytical models (e.g., fire dynamics models).
With the purpose of contributing to the development of real-time evacuation systems, this study presented an algorithm for finding a time-dependent optimal route in indoor space. This Spatiotemporal Optimal Route Algorithm (SORA) needs to be integrated with USN-based Tracking System in order to monitor dynamic geospatial entities, including the dynamic capacities and flow rates of hallways per time period. In this study, simulated monitoring data were used to identify the optimal route and building evacuation bottlenecks within the network in real-time emergency situations. The SORA is proposed to guide rescuers only, who need to know the best path for going from the building entrances to the disaster site (room) in the building. To detect the position of humans in a building at different time points, we use the results of an evacuation simulation system to identify the movement patterns of human beings in the emergency situation, instead of using real-time moving object tracking systems. The SORA was implemented with the simulated data for evaluating its effectiveness for performing 3D topological analysis. The results of the experimental implementation indicate that time-dependent optimal routing has potential to contribute in significant ways to developing real-time emergency response systems.
While focusing on proposing a SORA, there are several important issues for further research. First, to develop a real-time emergency response system, the implementation of the proposed SORA depends on the availability of accurate real-time data from diverse sources and the interoperability of various public agencies (Kwan and Lee, 2005; Goodchild, 2003). Second, the comprehensive indoor GIS data of the IER system raise serious considerations on data security, as the data for the system can be used by wrong persons, such as terrorists, for unexpected purposes. Database security is thus a very important component to implement and deploy in these systems. Third, the USN-based tracking system raises questions about protecting individual human rights and personal privacy (Armstrong, 2002; Kwan and Lee, 2005). Lastly, in order to develop real-time systems, the proposed algorithm needs to be applied to multi-levels indoor building, and the SORA needs to be improved for guiding evacuees also from the disaster site to safety areas.
References
- Armstrong, M.P. (2002), Geographic information technologies and their potentially erosive effects on personal privacy, Studies in the Social Sciences, Vol. 27, pp. 19-28.
- Azevedo, C.L., Cardoso, J.L., and Ben-Akiva, M. (2014), Vehicle tracking using the k-shortest paths algorithm and dual graphs, Transportation Research Procedia, Vol. 1, No.1, pp. 3-11. https://doi.org/10.1016/j.trpro.2014.07.002
- Cahan, B. and Ball, M. (2002), GIS at ground zero: spatial technologies bolsters World Trade Center response and recovery, GEO World, Vol. 15, pp. 26-29.
- Chen, X., Kwan, M.P., Li, Q., and Chen, J. (2012), A model for evacuation risk assessment with consideration of preand post-disaster factors, Computers, Environment and Urban Systems, Vol. 36, No. 3, pp. 207-217. https://doi.org/10.1016/j.compenvurbsys.2011.11.002
- Church, R.L. and Marston, J.R. (2003), Measuring accessibility for people with a disability, Geographycal Analysis, Vol. 35, No. 1, pp. 83-96. https://doi.org/10.1353/geo.2002.0029
- Cova, T.J. and Church, R.L. (1997), Modeling community evacuation vulnerability using GIS, International Journal of Geographic Information Science, Vol. 11, pp. 763-784. https://doi.org/10.1080/136588197242077
- Cutter, S.L., Richardson, D.B., and Wilbanks T.J. (2003), The Geographical Dimensions of Terrorism, Routledge, New York, pp. 111-116.
- Dane, C. and Rizos, C. (1998), Positioning Systems in Intelligent Transportation Systems, Artech House, Boston.
- Delling, D., Goldberg, A.V., Nowatzyk, A., and Werneck, R.F. (2013), Phast : hardware-accelerated shortest path trees, Journal of Parallel and Distributed Computing, Vol. 73, No. 7, pp. 940-952. https://doi.org/10.1016/j.jpdc.2012.02.007
- Dijkstra, E.W. (1959), A note on two problems in connection with graphs, Numerische Mathematik, Vol. 1, pp. 269-271. https://doi.org/10.1007/BF01386390
- George, B., Kim, S., and Shekher, S. (2007), Spatio-temporal network databases and routing algorithms: a summary of results, In: Papadias, D., Zhang, D., and Kollios, G. (eds), SSTD 2007, LNCS, Vol. 4605, Springer, Verlag, pp. 460-477.
- Goodchild, M.F. (2003), Geospatial data in emergencies, In: Cutter, S.L., Richardson, D.B., and Wilbanks, T.J. (eds), The Geographical Dimensions of Terrorisms, Routledge, New York, pp. 99-104.
- Kirkby, S., Pollitt, S., and Eklund, P. (1997), Implementing a shortest path algorithm in a 3D GIS environment, In: Kraak, M.J. and Moleanaar, M. (eds), Advances in GIS Research II (Proceedings of the 7th International Symposium on Spatial Data Handling), Taylor & Francis, London, pp. 437-448.
- Kwan, M.P. (2003), Intelligent emergency response systems, In: Cutter, S.L., Richardson, D.B., and Wilbanks, T.J. (eds), The Geographical Dimensions of Terrorism, Routledge, New York, pp. 111-116.
- Kwan, M.P. and Lee, J. (2005), Emergency response after 9/11: the potential of real-time 3D GIS for quick emergency response in micro-spatial environments, Computers, Environment and Urban Systems, Vol. 29, pp. 93-113. https://doi.org/10.1016/j.compenvurbsys.2003.08.002
- Kwan, M.P. and Ransberger, D. (2010), LiDAR assisted emergency response: detection of transport network obstructions caused by major disasters, Computers, Environment and Urban Systems, Vol. 34, pp. 179-188. https://doi.org/10.1016/j.compenvurbsys.2010.02.001
- Kwan, M.P. and Weber, J. (2003), Individual accessibility revisited: implications for geographical analysis in the twenty-first century, Geographical Analysis, Vol. 35, No. 4, pp. 341-353. https://doi.org/10.1111/j.1538-4632.2003.tb01119.x
- Kolbe, T.H., Becker, T., and Nagel, C. (2008), 1st Technical Report Discussion of Euclidean Space and Cellular Space and Proposal of an Integrated Indoor Spatial Data Model, Technische Universitat Berlin.
- Lee, J. (2001), A 3D Data Model for Representing Topological Relationships between Spatial Entities in Built Environments, Ph.D. dissertation, Department of Geography, The Ohio State University, Columbus, Ohio, USA.
- Lee, J. (2004), A spatial access oriented implementation of a topological data model for 3D urban entities, GeoInformatica, Vol. 8, No. 3, pp. 235-262.
- Lee, J. (2007), A three-dimensional navigable data model to support emergency response in micro-spatial builtenvironments, Annals of the Association of American Geographers, Vol. 97, No.3, pp. 512-529. https://doi.org/10.1111/j.1467-8306.2007.00561.x
- Lim, Y. and Kim, H. (2005), A shortest path for real road network based on path overlap, Journal of Eastern Asia Society for Transportation Studies, Vol. 6, pp. 1426-1438.
- Meijers, M., Zlatanova, S., and Pfeifer, N. (2005), 3D geoinformation indoors: structuring for evacuation, Proceedings of Next Generation 3D City Models, 21-22 June, Bonn, Germany, http://www.eurosdr.net/km_pub/no49/html/Citymodel/paperpresentation.htm (last date accessed : 2 April 2010).
- Miller, H. and Shaw, S.L. (2001), Geographic Information System for Transportation: Principles and Applications, Oxford University Press, New York.
- Okabe, A. and Kitamura, M. (1996), A computational method for market area analysis on a network, Geographical Analysis, Vol. 28, pp. 330-349.
- Pu, S. and Zlatanova, S. (2005), Evacuation route calculation of inner buildings, In: Van Oosterom, P.J.M., Zlatanova, S., and Fendel, E.M. (eds), Geo-information for Disaster Management, Springer Verlag, Heidelberg, pp. 1143-1161.
- Raper, J. (2000), Meltidimensional Geographic Information Science, Taylor & Francis, New York.
- Scott, M.S. (1994), The development of an optimal path algorithm in three dimensional raster space, Proceedings of GIS/LIS'94, pp. 687-696.
- Xiong, D. (2000), A three-stage computational approach to network matching, Transportation Research C, Vol. 8, pp. 13-36. https://doi.org/10.1016/S0968-090X(00)00006-1
- Wu, Y. and Miller, H. (2002), Computational tools for measuring space-time accessibility within transportation networks with dynamic flow, Journal of Transportation and Statistics, Vol. 4, No. 2/3, pp. 1-14.
- Yen, J.Y. (1971), Finding the K shortest loopless paths in a network, Management Sciences, Vol. 17, No. 11, pp. 712-716. https://doi.org/10.1287/mnsc.17.11.712
- Zhan, F.B. and Noon, C.E. (1998), Shortest paths algorithms: An evaluation using real road networks, Transportation Science, Vol. 32, pp. 65-73. https://doi.org/10.1287/trsc.32.1.65
- Zhang, S. (2013), One dynamic shortest path algorithm in a traffic network based on a genetic algorithm, Advances in Civil, Transportation and Environmental Engineering, Vol. 140, pp. 189-193.
Cited by
- Three-Dimensional Indoor Fire Evacuation Routing vol.9, pp.10, 2014, https://doi.org/10.3390/ijgi9100558
- Indoor Environments and Geographical Information Systems: A Systematic Literature Review vol.11, pp.4, 2014, https://doi.org/10.1177/21582440211050379