Int. J. Mobile Network Design and Innovation, Vol. 4, No.

2, 2012

109

First responder communication in urban environments Kamill Panitzek*, Immanuel Schweizer, Tobias Bnning, Gero Seipel and Max Mhlhuser
Telecooperation Lab, Technische Universitt Darmstadt, Hochschulstr. 10, D-64289, Darmstadt, Germany E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] *Corresponding author
Abstract: Communication is crucial for first responders. Crisis management is nearly impossible without good means of communication. Unfortunately the communication technology used by first responders today does not scale well. Also most of the given infrastructure, such as cell towers, might be destroyed. In recent research ad-hoc and peer-to-peer-based communication has been proposed to solve the problem of resilient communication. Most mobile devices are equipped with wireless transceivers that make them suitable to participate in ad-hoc networks. But node density might be too small for a connected topology. In this paper, we therefore discuss the implications of an emergency switch for privately owned wireless routers. Wireless routers can transition to an emergency mode to create a supportive wireless mesh network. To analyse if such a network would be beneficial and give a resilient topology real data from wireless routers in a city is gathered. We calculate the locations of these routers from GPS traces and the resulting topologies are analysed investigating suitability and resiliency issues. Keywords: wireless mesh network; urban area; private wireless routers; first response; resiliency; analysis; mobile network design. Reference to this paper should be made as follows: Panitzek, K., Schweizer, I., Bnning, T., Seipel, G. and Mhlhuser, M. (2012) First responder communication in urban environments, Int. J. Mobile Network Design and Innovation, Vol. 4, No. 2, pp.109118. Biographical notes: Kamill Panitzek received his Masters degree in Computer Science from Technische Universitt Darmstadt in 2010. He is currently a PhD student and works as a Research Associate at the Telecooperation Lab, Technische Universitt Darmstadt. His research interests lie in the areas of P2P-based service delivery platforms, mobile ad-hoc networks, civil security, and QoS issues related to P2P-based systems. Immanuel Schweizer received his Masters degree in Computer Science from Technische Universitt Darmstadt in 2009. Since then, he has joined the Telecooperatiom Group of Professor Mhlhuser as a PhD student working on sensor networks in urban environments. Besides sensor networks, he has a research interest in P2P and ad-hoc networks as well as data mining. Tobias Bnning received his Bachelor degree in Computer Science from Technische Universitt Darmstadt in 2010. He is currently a Master of Science student in the field of internet- and web-based systems and works on his master thesis about service-oriented architectures for mobile ad-hoc networks in cooperation with the Telecooperation Lab, Technische Universitt Darmstadt. Gero Seipel received his Bachelor degree in Computer Science from Technische Universitt Darmstadt. He worked before on several projects provided by the Telecooperation Lab at Technische Universitt Darmstadt. He is currently a Master of Science student and works on his Master thesis. His research interests lie in the areas of wireless (sensor) networks, roles and communities and clustering. Max Mhlhuser received his PhD from Universitt Karlsruhe, in 1986. He has over 25 years of experience in research and teaching in areas related to ubiquitous computing, networks and distributed systems, e-learning, and privacy and trust. He held permanent or visiting professorships at the Universities of Kaiserslautern, Karlsruhe, Linz, Darmstadt, Montral,

Copyright 2012 Inderscience Enterprises Ltd.

110

K. Panitzek et al.
Sophia Antipolis (EURECOM), and San Diego (UCSD). Since 2000, he is the Head of the Telecooperation Lab, Technische Universitt Darmstadt working in three research fields: middleware and infrastructures, novel multimodal interaction techniques, and human protection in ubiquitous computing (privacy, trust and civil security).

Introduction
1 2 3

Our contribution is threefold: discuss the implications of an emergency switch for private wireless routers estimate positions of wireless routers using lateration methods on wardriving data analyse graph properties to investigate the suitability as resilient communication topology.

In 2011, we have been reporting on the idea of using publicly available wireless routers to supplement the communication infrastructure during first response missions in urban areas (Panitzek et al., 2011). This would only work in city centres as most routers are privately owned in other areas. Therefore, we now analyse the possibility of using privately owned routers as communication backbone. Communication between first responders, citizens and command and control personnel is viable to the success of any first response mission. In recent research ad-hoc and peer-to-peer communication has been proposed to improve communication (Bradler et al., 2008a, 2009). Using decentralised communication architectures will improve resilience as link and node failures can be mitigated by the network. This is true only if enough nodes are available as a fallback requiring minimum node density. In disaster response this is given close to the incident, the so called incident site and treatment area (Aschenbruck et al., 2009), as most first responders will be located there. It is also given close to stationary facilities like the command and control centre or hospitals, the so called hospital zone. But anywhere else in between, also called transport zone, the node density might be too low leaving the whole network vulnerable. A support infrastructure is needed delivering a high node density throughout the entire disaster area. It should also be readily available and working with available wireless standards. We have made the point in 2011 that wireless routers fulfil those requirements, but routers without encryption, usable out of the box, are scarce. Even though the resulting topologies build with just those routers might be suitable in city centres a higher density is needed elsewhere. Now, we build upon that analysis by additionally investigating privately owned routers in a predefined city area. We assume that at some point in the future an emergency switch was to be implemented that would make all existing communication resources available to first responders (Section 2). This leads us to investigate the density of all routers in city areas and how they would be suitable to build a communication topology. We extended a given Android wardriving app to be able to gather our own dataset of all available routers in our hometown Darmstadt, Germany (Section 3). A lateration method is used to estimate the positions of the wireless routers from given fingerprints. Using the unit disc graph model (Clark et al., 1990) for wireless topologies we are able to generate ad-hoc mesh networks with different parameters. On the resulting networks we evaluated important graph theoretical properties such as average path length and resilience against node failures in Section 4.

Emergency mode for wireless routers

The amount of households equipped with internet connection reaches almost 100% in developed countries today. Many of those households use wireless routers to connect multiple devices to the internet and to each other. In apartment buildings this leads to a multitude of available wireless routers and a very high density especially in urban areas. Users can recognise this density if they search for available wireless devices at home. The routers found during the scanning process belong to neighbours in the same building or in houses on the opposite side of the road. This infrastructure is very dense and should be available to first responders during missions in such urban environments. This is why we propose an emergency switch for wireless routers. The idea is to create a wireless mesh network (Bruno et al., 2005; Raniwala and Chiueh, 2005) on top of the given privately owned wireless routers. This infrastructure then acts as the backbone in case of a disaster. It can fill the communication gap between the incident site and the command centre. The emergency switch is meant to disable the security protocols and allow public access to the wireless router. It can only be part of the network if it is open to all users and offers its resources. Abuse of such a network must be avoided at all cost. Therefore this network should be isolated from the citizens home network to protect peoples privacy. This goal could be easily accomplished as it is already today possible to install a home network and a guest network in parallel to grant internet access to visitors. Such functionality is included in the AVM Fritz!Box 7390 (http://www.avm.de/en/Produkte/FRITZBox/FRITZ_Box_F on_WLAN_7390/index.php), for instance. A similar guest concept is used in the project mobile access of the city of Aachen, the university RWTH Aachen, and industrial partners. This project provides free mobile internet connectivity to participants of the project. Every citizen may participate to the project by sharing his or her wireless router to increase the network coverage in the city. The project also provides services for city visitors, like 3D

First responder communication in urban environments city sightseeing flight or localisation. Also in this project the data traffic of mobile users is tunnelled to their own homes to protect owners of the routers from illegal usage and law infringements. This ensures that all internet usage is bound to the right users identities. In addition mobile users are protected by encrypting the wireless communication between user and the router. To realise the proposed emergency switch, similar measures must be taken to protect both the citizens and the first responders during their mission. The emergency switch would enable an open guest mode that on the one hand protects peoples privacy and on the other hand makes the existing communication resources available to first responders. Legal aspects must be considered and could be solved by identifying individual users of the emergency network similar to the mobile access project.

111 we use Android devices, since they have wireless network adapters and GPS sensors integrated. Also, there exist many wardriving apps for Android like WiGLE, G-MoN (http://www.wardriving-forum.de/wiki/g-mon), or wardrive (https://github.com/raffaeleragni/android-wardrive) that allow for scanning for wireless LAN networks in the vicinity and store the data in a database. Since additional modifications to such an app must be made, we chose to use the open source app wardrive. In the original code the app scans for wireless networks in the vicinity every three seconds and stores the current time, the current GPS position, and readings from the wireless networks adapter into a local database. If a network already exists in the database the information will be replaced with new readings if the GPS coordinates differ more than a specific value and the signal strength of the new data point is higher than the old one. But using this method the locations of routers will end up on the streets we move through during our measurements and not on the actual locations of these routers. Therefore, we changed the application code to continuously log the information into the database, thus creating also multiple entries about individual networks. Using this data we can then use lateration methods to estimate the actual router location. In addition we also implemented the ability to measure in periods smaller than three seconds. Our modified version of the wardrive app was installed on a Motorola Xoom Tablet. With this device we walked in slow walking speed through the predefined area of our hometown as depicted in Figure 1. This area includes the city centre and is of roughly 0.47 km in size. It is determined by the streets Hgelstrae, Kirchstrae, Holzstrae, Zeughausstrae, Bleichstrae, Kasinostrae, and Neckarstrae. First, we measured in intervals of one second. But this stressed the hardware too much which is why we returned to the predefined value of three seconds between measurements and reduced our walking speed instead.

Methodology

Our goal is to analyse the wireless infrastructure in the city centre of our hometown Darmstadt, Germany. But before we can analyse any networks that could be created in the case of an emergency we need to determine the locations of wireless routers in our hometown. Since the exact locations are not stored in any database and therefore cannot be accessed by us, we tried to determine the locations of all routers ourselves. For this purpose we first measured signal strengths of wireless networks in a specified area of our hometown and then calculated the positions of the routers using a lateration method.

3.1 Measuring positions of wireless routers

There exist several ways to learn about existing wireless routers in cities. Projects like WiGLE (http://www.wigle.net) store data about wireless networks by users driving with their cars or walking through the city while scanning for wireless networks. This procedure is referred to as wardriving and several applications exist for different kinds of hardware and operating systems. But the main problem with such data collections is that most of these data points are located on main roads or even highways. This is due to the fact that many users participating in wardriving projects drive along the same main roads or highways when driving to work, for instance. Once inside the city most of the users will go different directions therefore the data density decreases with the size of the corresponding road. Another problem is that in such projects routers are placed on the streets instead of their real locations, because no calculations are done while wardriving. Our idea was to improve data quality by calculating the real locations of wireless routers using lateration methods on the gathered data. To find out where wireless routers are resided throughout the city, we first have to scan for wireless networks at multiple locations in the city. For this purpose

3.2 Lateration
Once we collected all data from our measurements in the city area, we can use lateration to estimate locations of the wireless routers. With trilateration (Savvides et al., 2001) exactly three data points are needed for the calculation. But since we collected a huge amount of data points we use the multilateration approach where all measured points to one wireless router are used for the calculation, thus improving the estimation. The method is depicted in Figure 2 and starts by choosing two arbitrary measurement points Pt and Pu with intersecting circles of the radii rt and ru calculated from the measured signal strength. A signal strength s is converted into a distance r by using a formula extracted from the findings of Faria (2005) and Rappaport (1996):
r = 1/ 3(10( s 10) / 35 1)

112
Figure 1

K. Panitzek et al.
Area of measurement

Source: Pictures of maps in this work taken from http://maps.google.com

We also calculate the points in which the circles intersect and denote them Qt,u,1 and Qt,u,2. Then a second pair of arbitrary points Pv and Pw are picked with intersecting circles of radii rv and rw, respectively. For these two points we also calculate the points in which the circles intersect and denote them Qv,w,1 and Qv,w,2. Now we pick those two intersection points that are closest to each other and calculate the point in between, being the centroid [cf., Figure 2(a)]. To incorporate all other points of the measurements we pick the next two arbitrary points Px and Py with intersecting circles and calculate the intersection points Qx,y,1 and Qx,y,2 as well. Using the point closer to the old centroid we calculate the new centroid being in the middle between both points [cf., Figure 2(b)]. We iteratively proceed with all pairs of points with intersecting circles, thus further correcting the estimation of the wireless routers location.
Figure 2 Multilateration

Evaluation

Qt,u,1 Pt rt centroid Qv,w,1 Pv rv Qv,w,2

(a)

Qx,y,1 rx Pu ry Py Qx,y,2 intersection nearest intersection new centroid old centroid Px

Qt,u,2

ru

Pw rw

(b)

After applying the multilateration approach to our measurements we receive an estimated map of all wireless routers in the city centre of our hometown Darmstadt. This map is depicted in Figure 3. In our measurements we collected 32,983 points covering an area of about 467,500 m in size. In these measurements we identified 1,971 unique BSSIDs (unique wireless routers). Out of these 1,971 routers 212 had no encryption at all (public routers) and are represented with grey markers on the map below (cf. Figure 3). Of the remaining encrypted routers, 129 were encrypted using WEP encryption and are represented using white markers, 240 and 421 used WPA and WPA2 encryption, respectively. We also found that 951 routers provided both, WPA and WPA2 encryption in parallel. All WPA and WPA2 encrypted routers are represented by the black markers on the map below. Eighteen routers had an unknown encryption or were ad-hoc access points most probably from people surfing in cafes with their laptops or tablets using WiFi-tethering on the mobile phone. Now that we have this map of our hometown we can use this information to virtually construct mesh networks and analyse properties of these networks. For this purpose we use different communication ranges. We always connect two routers if they are in distance of d metres with d {10, 20, 30, 40, 50, 60}, therefore, creating six different graphs. We also distinguish between encrypted (private) and non-encrypted (public) networks to compare the results to our approach proposed in 2011 (Panitzek et al., 2011). In general we believe a communication range of d = 30 metres to be very realistic in an urban environment. Figure 4 shows how these mesh networks would look like when constructed with d = 30 metres or d = 60 metres using all routers or

First responder communication in urban environments public routers only. In the following evaluation we will show that in the city centre of our hometown we have enough wireless routers to create a resilient emergency mode mesh network. To analyse different graph theoretical properties of our constructed city mesh networks we used the graph-theoretic network analyser (GTNA) (Schiller et al., 2010), a Java-based framework that allows for graph-theoretic analysis of arbitrary network topologies. The tool has the ability to either import snapshots from network simulators or to generate popular network topologies using topology generators. Therefore, a large set of topology generators and commonly used graph metrics are already provided by GTNA. However, the framework is easily extendible through a well-defined plug-in interface. GTNA also allows for a graphical visualisation of the analysed graph metrics using the build-in plotting module.

113 These nodes or smallest clusters are integrated in the number of clusters, too. The maximum cluster size is then the number of nodes forming the biggest cluster in the entire network. The shortest path length distribution is defined as the distribution over the shortest paths between all pairs of nodes in the network.
Figure 3 Map of Darmstadt with estimated positions of wireless routers

4.1 Metrics
For our analysis and to compare the networks constructed with different communication ranges we use the graph metrics node degree, number of clusters, maximum cluster size, and shortest path length distribution. The node degree is defined as the number of links or edges that are connected to one node. Since the constructed graphs are undirected we do not distinguish between in- and out-degree. A cluster is a set of interconnected nodes. If a node has a node degree of zero the node is isolated, thus creating a cluster of size 1.
Figure 4 Constructed mesh networks in Darmstadt, (a) all routers, d = 30 m (b) all routers, d = 60 m (c) public routers only, d = 30 m (d) public routers only, d = 60 m

(a)

(b)

(c)

(d)

114

K. Panitzek et al. a node has almost 36 neighbours in average. As we believe 30 metres of communication range in an urban area being very realistically we can see that these networks already have only 2% isolated nodes and about ten neighbouring nodes in average with a maximum of 28 neighbours. This draws such a network very suitable for an emergency routing mode.

4.2 Degree distribution

As expected, with increasing communication range d the average node degree as well as the maximum node degree of the analysed graphs increase continuously. Also, the number of nodes having a node degree of zero decreases with an increasing communication range. With a range of d = 10 metres about one third of all nodes are isolated. With a range of d = 20 metres less than 7% of all nodes are isolated. However, with a range of d = 30 metres only 2% of all nodes are isolated and with d = 60 metres no node is isolated at all. The distribution of node degrees is depicted in Figure 5. With an increasing communication range d the distribution of the node degree gets further stretched out. This means that the resulting network becomes even more interconnected, since more other routers can be reached from one router.
Table 1 d Avg. deg. Max. deg. Edges CPL Diameter Network properties for different ranges d 10 m 1.324 8 1,010 1.204 4 20 m 4.74 16 3,619 1.864 8 30 m 40 m 50 m 60 m 10.313 17.301 25.802 35.58 28 42 54 68 7,874 13,209 19,700 27,165 2.43 3.427 4.426 4.698 10 17 20 19

4.3 Network fragmentation

To analyse capabilities of the network to resist failures of most important nodes we look at the clusters of the networks. We look at the number of clusters and how this number behaves when removing the most important nodes, the so called hubs. These hubs are nodes with the highest node degree. We successively remove these hubs and count the number of clusters. This tells us in how many fragments the network will partition if the most interconnected nodes in the network fail. As shown in Figure 6 assuming a range of ten metres the network generally results in a large number of clusters. In the initial state without removing any nodes the network is already partitioned into almost 800 clusters. But with d = 20 metres range the number of clusters already decreases drastically to less than 200 clusters. A range of d = 30 metres decreases the number of clusters even further resulting in about 50 clusters in the initial state. When we now remove hubs we see an increase in the number of clusters almost immediately in the networks generated with a range of d = 10 and d = 20 metres. But with a range of d = 30 metres the network is so highly interconnected that even removing 16% of the most interconnected nodes does not change the number of clusters. Assuming a higher range amplifies this effect even more.

As presented in Table 1 we can observe the average node degree in the different networks. The average node degree is calculated by calculating the average value over all node degrees in the network. As we can see, in the network constructed with a range of d = 10 metres the average node degree is 1.324. This roughly means in average every node has just one neighbour, a node it is connected to. In networks with a range of d = 30 metres one node has about ten neighbours in average and with a range of d = 60 metres
Figure 5 Degree distribution
0.35

0.3

10m all routers 20m all routers 30m all routers 60m all routers

0.25

0.2

P(d)
0.15 0.1 0.05

0 0 10 20 30 Degree d 40 50 60 70

First responder communication in urban environments

Figure 6 Number of clusters when removing most interconnected nodes
900 800 700 600 500 400 300 200 100 0 0 10 20 30 40 50 60 % of removed Nodes 70 80 90 100 10m all routers 20m all routers 30m all routers 60m all routers

115

Figure 7

Maximum cluster size when removing most interconnected nodes

1600 1400 1200
Max Cluster Size

Number of Clusters

10m all routers 20m all routers 30m all routers 60m all routers

1000 800 600 400 200 0 0 10 20 30 40 50 60 % of removed Nodes 70 80 90 100

When removing these hubs the question arises what happens with the biggest cluster in the network? Since the biggest cluster is the sub-network with the highest amount of nodes we assume that these nodes also have higher node degrees than in the remaining clusters of the network. Also, the node with the highest node degree might most probably be in the biggest cluster. So we now look at the size of the biggest cluster in the network when successively removing hubs (cf., Figure 7). After removing 10% of these most interconnected nodes of the network with range d = 20 metres we see the first drop in the maximum cluster size. This means the biggest cluster has been partitioned. With a communication range of d = 30 metres we can remove up to 22% of most interconnected nodes until partitioning the

biggest cluster. With a range of d = 60 metres this partitioning happens after removing 62% of hubs. The partitioning happens because a so called bridge node was removed. A bridge node connects two sub-clusters with each other. This means that if a bridge node gets removed, also other nodes are disconnected from the network, because they were only connected through this particular node. Identifying such bridge nodes is of great importance to improve resiliency of the network and to protect the network against attacks. In our previous work we developed a decentralised mechanism to identify such bridge nodes (Bradler et al., 2011). In the field, once such critical nodes are identified, we could place additional infrastructure to assist network topology in that place and to prevent network partitioning.

116
Figure 8

K. Panitzek et al.
Shortest path length distribution
0.9 0.8 0.7 0.6 0.5 10m all routers 20m all routers 30m all routers 60m all routers

P(l)
0.4 0.3 0.2 0.1 0 0 2 4 6 8 10 12 Shortest Path Length l 14 16 18 20

4.4 Routing performance

The shortest path length distribution shows how the shortest paths are distributed over all nodes in the network. As the shortest path length is calculated by considering all pairs of nodes in the network and calculating the shortest paths between them, it is of great benefit to routing performance if most pairs of nodes have small shortest paths. Figure 8 presents our findings of this analysis. Here it seems that a network constructed with a range of d = 10 metres has the best routing performance because more than 80% of all node pairs have a shortest path length of one hop. As presented in Table 1 the characteristic path length (the average shortest path length) is 1.204 hops and the diameter (longest shortest path length) is only four hops. But when recalling the number of clusters (cf., Figure 6) such a network has almost 800 isolated clusters. This means that most nodes cannot even be reached because a path is missing. With a communication range of d = 30 metres 33% of node pairs have a shortest path of one hop but only 50 clusters exist in total and the biggest cluster consists of over 71% of all nodes. The characteristic path length in this network is 2.43 hops and the diameter is ten hops. With increasing communication range the diameter and the characteristic path length are almost twice as high as in networks with d = 30 metres. This can be seen as a drop in routing performance since the number of reachable nodes does not increase as much as the average path lengths. It seems as if a range of d = 30 metres also is a good tradeoff between routing performance and network resiliency as described before.

4.5 Public routers only

In our previous work (Panitzek et al., 2011) we proposed to use public routers only to create a mesh network in case of an emergency. As we identified more than 200 open networks in the city centre of our hometown during our measurements, we now also compare the performance of networks constructed by using only public routers to the previous ones. Considering the node degree, networks with public routers only have about 30% of isolated nodes when constructed with a range of d = 30 metres and about 5% isolated nodes when constructed with a range of d = 60 metres. When including all nodes only 2% of all nodes are isolated in networks with a range of d = 30 metres. When looking at the size of the biggest cluster and comparing both networks with each other the discrepancy between these networks becomes obvious (cf., Figure 9). In public networks with d = 30 metres less than 20% of all nodes are connected to the biggest cluster whereas more than 71% of all nodes are connected to the biggest cluster when also including private routers into the network creation process. When looking at the shortest path length distribution we can see that about 60% of all node pairs are connected with only one hop in both networks with range d = 30 and d = 60 metres considering only public routers. The longest path is five hops in both cases due to the low density of the nodes. And because less than 20% of all nodes can be reached the networks with public routers only have very bad routing performance. This clearly shows private routers will drastically improve not only network performance but also resiliency in an emergency mode mesh network.

First responder communication in urban environments

Figure 9 Comparison public routers only to all routers
1600 1400 1200 30m public routers 60m public routers 30m all routers 60m all routers

117

Max Cluster Size

1000 800 600 400 200 0 0 10 20 30 40 50 60 % of removed Nodes 70 80 90 100

Conclusions

References
Aschenbruck, N., Gerhards-Padilla, E. and Martini, P. (2009) Modeling mobility in disaster area scenarios, Performance Evaluation, Vol. 66, No. 12, pp.773790. Bradler, D., Aitenbichler, E., Schiller, B. and Liebau, N. (2009) Towards a distributed crisis response communication system, Proceedings of the 6th International ISCRAM Conference. Bradler, D., Kangasharju, J. and Mhlhuser, M. (2008a) Evaluation of peer-to-peer overlays for first response, Sixth Annual IEEE International Conference on Pervasive Computing and Communications (PerCom), pp.463467. Bradler, D., Schweizer, I., Panitzek, K. and Mhlhuser, M. (2008b) First response communication sandbox, Proceedings of the 11th Communications and Networking Simulation Symposium. Bradler, D., Krumov, L., Muhlhauser, M. and Kangasharju, J. (2011) BridgeFinder: finding communication bottlenecks in distributed environments, The International Conference on Information Networking 2011 (ICOIN2011). Bruno, R., Conti, M. and Gregori, E. (2005) Mesh networks: commodity multihop ad hoc networks, IEEE Communications Magazine, Vol. 43, No. 3, pp.123131. Clark, B.N., Colbourn, C.J. and Johnson, D.S. (1990) Unit disk graphs, Discrete Mathematics, Vol. 86, Nos. 13, pp.165177. Faria, D.B. (2005) Modeling Signal Attenuation in IEEE 802.11 Wireless LANs, Vol. 1. Computer Science Department, Stanford University. Johnson, D., Maltz, D. and Broch, J. (2001) DSR: the dynamic source routing protocol for multi-hop wireless ad hoc networks, Ad Hoc Networking, edited by C.E. Perkins, Chapter 5, pp.139172, Addison Wesley. Panitzek, K., Bradler, D., Schweizer, I. and Mhlhuser, M. (2011) City mesh resilient first responder communication, Proceedings of the 8th International ISCRAM Conference.

In this paper we discussed the importance and the implications of an emergency switch for private wireless routers in urban environments. This switch should enable an emergency mode of operation creating a supportive wireless mesh network to assist ad-hoc communication in first response missions. For this purpose we changed an open source wardriving app for Android and gathered real data of wireless routers in our hometown and used lateration methods to estimate their actual locations. Using graph theoretical analysis on constructed networks with different parameters we found that with a communication range of 30 metres a mesh network could be easily constructed in urban areas like our hometown. The resulting networks showed to be resilient to node failures. In comparison to networks based on public routers only we highlighted the drastic improvement gained with an emergency mode switch inside privately owned routers. In our future work we plan to evaluate these networks through different ad-hoc routing protocols, such as ad-hoc on demand distance vector routing (AODV) (Perkins and Royer, 1999) and dynamic source routing (DSR) (Johnson et al., 2001) using simulations. Furthermore, we plan to use an energy model to evaluate how battery life performs under these conditions. We also plan to integrate these networks in our first response communication sandbox (Bradler et al., 2008b) to evaluate the benefit of such a supportive infrastructure for first response missions.

Acknowledgements
This work has been partially funded by the DFG research unit 733 QuaP2P.

118

K. Panitzek et al.
Savvides, A., Han, C-C. and Strivastava, M.B. (2001) Dynamic fine-grained localization in ad-hoc networks of sensors, Proceedings of the 7th Annual International Conference on Mobile Computing and Networking MobiCom 01, pp.166179. Schiller, B., Bradler, D., Schweizer, I., Mhlhuser, M. and Strufe, T. (2010) GTNA: a framework for the graph-theoretic network analysis, Proceedings of the 2010 Spring Simulation Multiconference on SpringSim 10.

Perkins, C.E. and Royer, E.M. (1999) Ad-hoc on-demand distance vector routing, Proceedings WMCSA 99, Second IEEE Workshop on Mobile Computing Systems and Applications, pp.90100. Raniwala, A. and Chiueh, T. (2005) Architecture and algorithms for an IEEE 802.11-based multi-channel wireless mesh network, Proceedings IEEE 24th Annual Joint Conference of the IEEE Computer and Communications Societies, Vol. 3, pp.22232234. Rappaport, T.S. (1996) Wireless Communications: Principles and Practice, Vol. 2, Prentice Hall, PTR, New Jersey.