Using GIS to Map the WiFi Signal Strength: University of Nebraska Lincoln, City Campus Final Project Report Submitted by: Yanfu Zhou / Student Submitted to: Dr.Nam Date: Dec 10, 2012
Contents Abstract... 1 Introduction... 2 Research goals... 3 Related studies... 4 Case Study 1: Mapping a Wireless Internet Network... 7 Case Study 2: WiFi Map Project in Long Island University... 8 Method and analysis... 10 Conclusions and Discussions... 15 Sources... 16 Acknowledgement... 16
Abstract The popularity of laptops, mobile devices among people leads to the needs for good quality of WiFi in public places. In planning field, serving public is one of our research goals. This study was thought out in order to approach the goal. In my study, I first discussed the necessity of making a map showing WiFi signal strength, and then I reviewed some references and did an interview in order to demonstrate the complexity of mapping the WiFi signal strength. I also studied two existing examples, they tried to map WiFi signal by different ways and both of them have some weaknesses in their research. Also in my study, a WiFi signal strength map finally has been made, although it is not considered to be accurate. 1
Introduction The increasing of mobile device, such as ipad, Samsung Tablet, and Google Chrome Laptop, has promoted the application of WiFi technology. WiFi has become very popular in public places such as parks, campus, airports, coffee shop and restaurant now. Providing free WiFi can attract more guests and preserve their visiting in business. This is very obvious when we walk into a free WiFi coffee shop or restaurant. While providing free WiFi seems a good way for enhance more sales in businesses, but based on my research, promoting visiting by providing free WiFi sometimes is not a good idea, when too many people get crowd in coffee shop. That is because water body will obstruct the signal. While people may get crowd in coffee shops, bars, or restaurants easily, but in our campus, people only get crowd in classrooms and sometimes get crowd in an openspace occasionally. Currently, WiFi signal has covered most buildings and areas in our campus. But outside the buildings the WiFi signal in some places turned to unstable sometimes. During the time when there is a big activity hold in campus, such as go big red, visitors get crowd in campus and the needs for WiFi rising, but the signal actually dropping. In addition, based on my daily experience in our campus, there is also a strong need among students and faculties for covering the openspace by good quality of WiFi signals in our campus. When the weather is nice, some people prefer to use their laptops outside the building (See Fig.1). However, there is no evidence showed that we must add more tables or chairs in our openspace, because whether people love to sitting on the chairs or sitting on the grass, people have varied opinions. And also improving WiFi quality in our campus does not mean it can enhance and promote the usage of openspace. But I believe, because most young people can not live without internet today. It will be a good try for us to attract people to the openspace by placing more WiFi access points (AP) in our campus, when there is a nice weather. And some inactivated openspace may become activated in our campus. In short, it is helpful to consider the WiFi factors when we try to change an inactivated openspace into a successful openspace in the planning fields. And mapping the WiFi signal not only can provide convenience for faculties, students, and visitors, but also can provide helpful information for the network administrators to make better investment decisions for WiFi antenna placement on the Campus [liu.edu 2011]. It is necessary for us to have a map showing the WiFi signal strength in our campus. 2
Fig 1 Lovers are using WiFi in our campus. Research goals Based on my interview with Mr. Michael Davison, who is working in the information services in our campus, I learned many factors can bring impact to the WiFi signal strength measurement. Actually it is very hard to map WiFi signal very precisely in the real world. Measurement tools, location of the access point (AP), buildings, trees, moisture, and a large group of moving people, and even seasons, they all have impact to WiFi signal mapping. I will demonstrate these factors with as many details as I can in my report. While mapping WiFi signal is hard, but it does not mean impossible. So based on my research, my goal is not to produce a very precise WiFi map for serving public. I considered my goal is to provide helpful information to the information services for wireless management in our campus. Although my map is not precise, it is better than there is nothing. WiFi technology is an old technology for proximity 20 years, and it has been applied so widely in our campus now, but there is still not a detailed WiFi signal map has been made for the public (See Fig.2). 3
Fig 2 Existing WiFi signal map lacks of details. Related studies In my research, the most weakness is that I only use my PC to measure the WiFi signal strength and I cannot figure out which access point I have connected to. In our campus, smart phones (iphone) cannot get WiFi signal when the signal strength is lower than -65 dbm. And ipad cannot get WiFi signal when the signal strength is lower than -70 dbm. Because different devices have different requirements for the signal strength, the results produced by my measurement can not be applied to other mobile devices. In short, if we use different platforms to measure the signal, we will get different results. In addition, even if we use the same platform to measure the signal strength in our campus, we still cannot get same results, because we may not on the same channel or we may not connected to the same AP (Different AP use different IEEE 802.11 standard), the results also will be impacted by the transmitters, not only the clients. Also different devices have different Watts. They can have impact to my result too. For example, when the battery power on my computer drop down, the CPU will assign more power to the graphics, and that means my wireless card will get little power. Also I have been told that most of the AP outside our buildings is using IEEE 802.11g at a 15.4 Watts and their data rate is low, it meets the minimum requirement for calling from a phone, but for laptop or ipad, it does not mean they will have a good link quality even if the readings (scanning) are high. Although sometimes the measurement tools can cheat us, we should not be so sad that we cannot get trustful results, we can still figure out the range of WiFi signal strength in different places in our campus, but that need as many measurement as 4
we can, it is a timeless measurement and the precision of the signal strength will be approached gradually. Just as what I guess based on my results, trees also have impact to the WiFi signal, because leaves contain almost 90% water. Besides, Ghazanfari s research also showed that water has some impact to WiFi signal. He found that WiFi signal not only decreases with the distance become larger, but also decreases with the increasing of water content in soil [Ghazanfari 2012] (See Fig.3). And actually it is not the leaves or soils obstruct the signal. It is the water body obstructs the signal. For instance, human beings, concrete structures, plants and soils, they all contain some water in them. And also this is why seasons can also have impact to the signal, because the air moisture varies different in different seasons. Because so many factors have impact to WiFi signal, and some of the factors change all the time. It is hard for us to get a stable measurement value even if in a short time. But if we measured the same point thousands of time by the exactly same device, the values will show us a range, and the range can be calculated to a constant, such as mean. If we get those works done, our map result will be precisely, but still cannot reflect how frequency the signal change caused by the environmental factors. While it will give us a good suggestion, where the best place might be for us to using WiFi in our campus. Fig 3 Ghazanfari s research showed that WiFi signal not only decreases with the distance become larger, but also decreases with the increasing of water content in soil [Ghazanfari 2012]. In common sense based on our life experience, WiFi signal strength decreases as distance between the AP and user become larger, and that is true [Faria 2005] (See Fig.4). Beal s research also showed that the 5
RSS (received signal strength) decreases as distance between device and AP increases [Beal 2003] (See Fig.5). And Beal s research also showed us that the WiFi signal regression looks more like curve correlation than linear correlation. It is not an absolutely linear regression, but how does its regression look like? I will demonstrate it in Case Study part in my report. Fig 4 the chart showed in Faria s research [Faria 2005]. Fig 5 the chart showed in Beal s research [Beal 2003]. As the distance between an AP and a user increases, the received signal power will have short and long distance fluctuations, referred to as multipath fading and shadow fading, respectively [Mantilla 2010] (See Fig.6). Shadow fading can be caused by obstructions in our environment and we cannot estimate the impact of shadow fading by doing calculations precisely. Since it cannot be simply estimated only by calculations, we found another way to estimate it. We can measure its value in the real world. But that needs a very huge data collection, and the precision of our estimation depends on how many data we collected. Unfortunately, the more the data we collected, the better our results will be. Fig 6 Multipath fading and shadow fading were well explained in Mantilla s research [Mantilla 2010]. 6
Case Study 1: Mapping a Wireless Internet Network Before I started my project, I have checked some information from the Internet and I found that Matte Baker and his team have mapped WiFi signal at the Centre of Geographic Sciences, in Lawrencetown, Nova Scotia, Canada in 2006. They have also formed a final report of their project and they tried to map as accurately as possible the following items: (1) Visibility of a wireless network in Lawrencetown (2) Strength of the wireless signal over distance (3) A 3D visualization of the visibility (4) A 3D visualization of signal strength over distance. The aim of Baker s research is to incorporate technical data involving the broadcast of wireless Internet signals into ArcMap and ArcScene to test the GIS Mapping capabilities of a wireless Internet network. In their report, they said that there have been countless studies being done all over North America by Cities, Municipalities and Towns to determine the economic viability of a wireless network. And as preliminary studies, and from a business point of view, these studies are effective. But as these studies move further into implementing this technology, people want a better idea of what this network will look like in reality. And this is where GIS and 3D Modeling, and subsequent graphic presentation of this network information will be most useful. Baker and his team used the viewshed function in the ArcMap Spatial Analyst extension in their study. And they thought this function provides great detail into the behavior of the wireless signal. The advantage that I discovered in their research is they considered the elevation or height information into their calculation. They listed five key elements to the visibility tool, which include: (1) Spot Elevation of Transmitter (2) Transmitter and Receiver Height (3) Horizontal Scan Angles (4) Vertical Scan Angles (5) Beginning and Endpoint of Horizontal Signal. They adopted LIDAR All Hits DEM to calculate the signal visibility in their project. The basic ideas of calculating the signal strength is very simple, it can be demonstrated in a equation as below: Received Signal Strength (db) = Transmitted Power (dam) Path Losses (db) And they use this model to calculate the Path Loss: Path Loss = 20 log 10 f + 20 log 10 d 28 Where: f - frequency in MHz, d - distance in meters 7
(*Note: If we have a perfect, clear line of sight, then the path loss is predictable based on the distance and frequency [seattlewireless.net 2012]) And here are their results (See Fig.7) Fig 7 the final result showed in Baker s research. The weakness in their studies is that they have not considered some unpredictable environmental factors (Such as shadow fading caused by occasional obstructions) into their analysis. The unpredictable factors cannot be estimated by formulas, but this does not mean we cannot decrease its impact to our analysis results. Actually, we can approach the real value of these impacts that caused by unpredictable factors indirectly by manually checking some random points in our campus by many times. Case Study 2: WiFi Map Project in Long Island University The WiFi Map Project was taken in Long Island University (LIU) from 2010 to 2011. The main objective of this project is creating maps showing the availability and strength of WiFi signal across LIU campus. There are three reasons that for why they want to map the WiFi signal. First, WiFi antennas a large investment made piecemeal over time. Second, WiFi antenna locations documented but not mapped. Third, WiFi signal intensity is unknown and generally available for mapping on campus. The data collection tool used in their project was iphone/ipad with ios5.0 operation system. They developed an app to collect WiFi signal strength and by using their app, they can send the signal strength value with coordinate to their database. They also have a team in charge of finding antennas in LIU campus (See Fig.8). They have collected 2400 points in their campus by using their app. They estimated that the average precision of the GPS on iphone is 5 meters (See Fig.9). And the accuracy mean in football field is 19 meters, and its stand deviation is 1.7 meters. The accuracy mean in building 8
area is 37 meters and its stand deviation is 14 meters. Their results showed that the accuracy of iphone GPS is better in football field than in building area (See Fig.10). Fig 8 WiFi antennas in LIU campus, red point means antennas that are inside buildings, yellow point means antennas that are outside buildings. Fig 9 iphone GPS precision has been estimated. Fig 10 Accuracy and Precision. In their research, the WiFi signal was only collected by Apple ios5.0 platform, so there is one weakness in their research, because different devices may response differently to the IEEE 802.11 a/b/n/g. Not all the devices need a high rate frequency (RF). But this does not mean their results cannot be trustful, their results can be more trustful to users who have similar devices. However, we should notice that even if the users have exactly the same device like them, their results still cannot be 100% trustful, because there are countless environmental factors that can also have impact to there measurement. Some environmental factors may have margin effects, but some may not, it depends on different situations. While I was very happy to see they considered some factors in their project. They listed four conditions that may have big impact to their results in some situation: (1) Weather (2) Humidity (3) Traffic/Timeframe (4) iphone vs ipad. 9
Another weakness in their research is that they have not considered signal regression problem into their data analysis. They just did buffer by using their antenna location data. And I think they have not figured out within what distance the device can have signal, but just one step out of the distance, there will be no signal. However, if we want to figure out this point, we need more discussions about signal regression modeling and do more calculation to find out this point. The last weakness in their research is that they have not considered terrain obstruction impact into their study. In other words, they have not done the signal visibility analysis. But this requires a high resolution DEM data. While although there are three weaknesses in their research, I think they have tried their best to map the WiFi signal strength in LIU campus (See Fig.11). Fig 11 IDW model map result and Kriging model map result. Method and analysis While in my research, there is nothing can be considered creative, I just did the similar things as what LIU did. There is a little different in our apps. After Apple released ios6.0, we cannot get WiFi signal strength on apple mobile device now, because Apple has banned the usage of their private API in their development toolkits (SDK). This can be only done on a Jailbreak apple mobile device that has ios5.0 system now. So in order to measure the signal strength value, I used my Linux PC in my project (See Fig.12). Because the information services only have an AP location documented table that cannot be shared with me for security awareness, and there is not any AP location database has been built in our campus before, I cannot do the signal regression simulation analysis in my GIS program, and also the 3D signal visibility analysis, which based on the previous one. In order to approach my goal as best as I can, I decided to use the building footprints to do the analysis. I assumed that all the APs in our campus are located in the buildings, so the signal will drop if you go far away the buildings. However, after my interview with Davison, I found my assumption is totally wrong. He told me that there are some 10
antennas out side the buildings in our campus, and he showed me some examples (See Fig.13). While because he cannot tell me where all those antennas exactly are, I still cannot do the signal regression simulation analysis. I still use the building footprints to do the analysis, because it is better than there is nothing. Before I started my analysis, I also downloaded LIDAR DEM data from the databank, but it only contain ground hits information, so even if we get the antenna location data, we still cannot do the 3D signal visibility analysis. I exported building footprints from the car parking map that published on our university website. Then I used the Photoshop to do some graphic works before I imported the building footprints into the R2V. R2V is an advanced raster to vector conversion software for automated map digitizing, GIS data capture and CAD conversion applications. It will automatically digitalize raster image to shape file (See Fig.14). Then I did distance analysis in GIS program by using the building footprints layer (See Fig.15). Fig 12 Parameters showed on Linux PC. Fig 13 Cisco 1310 APs running 802.11G at 15.4 Watts were found on the Roof of love library. 11
Fig 14 Digitalize Raster Image to Shape file. Fig 15 the Buffer Result (Building Footprints). I finally collected 281 points in our campus (See Fig.16). And for most of the points, I collected them in sunny weather during Thanksgiving, so I think there are no big differences in humidity. And during Thanksgiving, there are few people in our campus. I used the IDW model (inverse distance weighting) and Kriging model to interpolating the collected signal strength value (See Fig.17). The results of IDW and Kriging are very similar. But actually, they are totally different calculation processions. First, Kriging uses information about the spatial structure of the data to predict the value of an unsampled location, while IDW only uses the distance to do prediction, which is a simple mathematical formula. Second, Kriging uses all sample points to construct the semivariogram, but it only uses points within a certain distance of the unsampled location to develop the equations to compute a value. Because Kriging examines data to obtain spatial autocorrelation rather than assigning a universal distance power value (IDW), it provides a more reliable statistic result. With Kriging model, we can produce a surface that smoother than IDW model. In IDW model, placing a higher power (more emphasis) on the nearest points will cause less smooth map surface. Currently, Kriging model is more popular than IDW model when we need to produce temperature or precipitation maps. Fig 16 281 random points were measured. Fig 17 Interpolation (IDW Model). The last step in my analysis is using Map algebra function in GIS program to plus the layers together. And here are my results (See Fig.18 and Fig.19). 12
Fig 18 The WiFi Signal Strength Map (IDW). 13
Fig 19 The WiFi Signal Strength Map (Kriging). 14
Conclusions and Discussions As a planning student, my research goal is to serving the public. While during my research I discovered among public, there is a need that people want to know where the best place might be for them using their laptops, smart devices in our campus. And it is very obviously that existing WiFi signal map cannot meet this need. It is necessary for us to produce a detailed and reliable WiFi signal map for the public in our campus. In order to produce a good quality WiFi signal map, mapping WiFi signal strength precisely has became my primary task. It needs a huge and long time data collection work, but it is still possible to produce a precise WiFi signal strength map in our campus. By using different platforms or devices, we can collect the data gradually, but we must be sure that the environmental factors are minimized to the smallest before we get started. And if we can choose several most popular platforms or devices as our measurement tools, the results also will be more reliable to the public. When using the GIS program to analysis WiFi signal, it is better for us, GIS Technicians, to learn some basic wireless knowledge, because this is helpful for improving the WiFi signal GIS model in analysis. 15
Sources Daniel B. Faria. Modeling Signal Attenuation in IEEE 802.11Wireless LANs. Google scholar: 2005 E. E. Mantilla, C. R. Reyes, B. G. Ramos. IEEE 802.11 b and g WLAN Propagation Model using Power Density Measurements at ESPOL. World Academy of Science, Engineering and Technology, 2010: 46 Ehsan Ghazanfari, Suk-Un Yoon, Liang Cheng, Muhannad Suleiman, Sibel Pamukcu. Wireless Signal Networks for Subsurface Modeling and Geo-Event Characterization. Google scholar: 2012 James R. Beal Jr. CONTEXTUAL GEOLOCATION: A SPECIALIZED APPLICATION FOR IMPROVING INDOOR LOCATION AWARENESS IN WIRELESS LOCAL AREA NETWORKS. Google scholar: 2003 WiFi Map Project http://gis.liu.edu/ iphone to Google Fusion Tables WiFi Mapping Project http://geo.objectgraph.com/2011/01/14/iphone-to-google-fusion-tables-wifi-mapping-project/ Mapping a Wireless Internet Network - Using ArcGIS 9.1 http://www.docstoc.com/docs/54726671/mapping-a-wireless-internet-network-using-arcgis-91 Seattle Wireless - Path Loss http://www.seattlewireless.net/pathloss Acknowledgement The following people contributed ideas, knowledge, and resources to this project: Michael Davison UN-L Domain Manager / Wireless Network Supervisor Communications and Operations - Networking University of Nebraska Lincoln Kristin Sorensen Graduate Teaching Assistant School of Natural Resources 16