Close Range Positioning System using Wireless Networks. Abstract

Size: px

Start display at page:

Download "Close Range Positioning System using Wireless Networks. Abstract"

Philip Manning
6 years ago
Views:

1 Close Range Positioning System using Wireless Networks Timothy Tan Guang Rong 1, Ang Chee Wee 2 Abstract This study aims to investigate the accuracy and feasibility of using Radio Frequency (RF) wireless networking technology to determine the location of a person ( user ) or object through triangulation. With the implementation of theoretical constructs and accompanying software design, an experimental system was created. Our study indicates that implementing a real world system is feasible, but complex, as RF signal path can vary with different environmental conditions, resulting in fluctuation of received signal strengths. The experimentation results were gathered and analysis was performed to explain the differences between the computed and actual location data. Finally, the compensations and calibration required to design a more accurate system was determined. 1. Introduction In a military environment, an RF positioning system can be used to track mobile equipment in a base, or personnel in a battlefield. Such a system could be used independent of the Global Positioning System (GPS) which is currently deployed extensively in the military for positioning. By lowering the radio frequency used, the RF positioning system will be able to achieve a greater range, and thus function more effectively over a larger area. In an indoor environment, with many access points, such a system could be used to track users and to guide them (interactive guide) as they move within the building, using a database of the locations of access points in the building. This positioning technology will capitalize on the existing infrastructure of wireless APs that many companies have invested in to provide internet/intranet access. The overall strategy adopted in this study is described as follows. The user who is interested in determining his location would carry a wireless-capable laptop or PDA to measure the signal strength of at least three nearby Access Points (APs). These signal strengths are then used to estimate the distance between the user and each of the APs. With these estimated distances, the user s position would then be calculated using triangulation, a technique that combines the calculated distances from different APs to find the user s exact position. A program was developed in Java to accomplish the above. This program was designed to read from the signal strength log files, compute the position of the user, and graphically represent the position on the PDA/laptop screen. 2. Theory 2.1 What is wireless networking? IEEE is a standard protocol developed by the Institute of Electrical and Electronics Engineers (IEEE) to govern how wireless networking equipment should interoperate. The b standard is the most widely deployed for use in commercial intranet infrastructure, and was thus chosen for the project. It operates in the 2.4GHz frequency range with up to 11Mbps (Megabits per second) data rate. The 2.4GHz frequency range ( GHz) is split into 14 different channels, each 5MHz higher than the first. Each channel occupies a 22MHz bandwidth, and thus they overlap. Nonoverlapping channels are 5 channels apart from the adjacent ones, for example, channels 1, 6 and 11. IEEE b equipment are specified to operate in two possible modes: Infrastructure or Ad-hoc. In infrastructure mode, the wireless network consists of an AP that is connected to a wired local area network (LAN) infrastructure, and a group of wireless workstations connected to the LAN through it. This configuration is known as the Basic Service Set (BSS). Each wireless network is differentiated 1 SRP student, Victoria Junior College 2 Naval Logistics Department, Republic of Singapore Navy, MINDEF

2 by its Service Set IDentifier (SSID), as specified by the AP. In Ad-Hoc mode, the wireless workstations use the Independent Basic Service Set (IBSS) configuration, which allows them to communicate directly with one another if they are within radio range. A peer-to-peer network is formed this way. (3Com Corporation, Jupitermedia Corporation, 2002) Secondary wavefront Wavelet Receiving point Primary wavefront Figure 1: Wavelets forming a wavefront Figure 2: Components toward receiver 2.2 Wavelets By Huygens Principle, as electro-magnetic (EM) wave propagates in space, all points on a primary wavefront will form secondary wavelets that will then produce a new wavefront in the direction of propagation. This wavefront propagation can be represented using vectors, as in Figure 1. At each position on the wavefront, there is an EM component that points toward the receiver (see Figure 2). (Barry McLarnon, 1996) 2.3 Path Loss Calculation The maximum Effective Isotropic Radiated Power (EIRP) for an unlicensed antenna system transmitting in the GHz frequency range as specified by the Infocomm Development Authority of Singapore is 100mW. (Chua H.H. 2003) However, this transmit power is rarely used, as most antennae in b transceivers have a gain of at least 2dBi, and which would push the EIRP above the 100mW limit. Therefore the maximum transmit power used is usually 50mW. The transmitted RF signal s power dissipates with distance. In a path of free space, with no obstructions, the signal path loss in db can be calculated using the following derived equation: L p ( db) = lg r where r is the distance between transmitter and receiver (m) L p ( db) is the signal path loss in at 2.4GHz. (3Com Corporation, 2003) 2.4 Fresnel Zone The Fresnel zone is a defined space between two antennae that are communicating. Obstructions that protrude into this zone causes signal strength at the receiver to be reduced. The Fresnel zone radius will tell us how high both antennas have to be raised for optimal radio transmission. To determine the Fresnel zone radius (see Figure 3), the following is used: d1d where h is the radius 2 h = 17.3 f is the frequency in Gigahertz (GHz) f ( d1 + d 2 ) d 1, d 2 are in kilometres (km) (Cisco Systems Inc, 2000) A height clearance of 60% of the radius of the first Fresnel zone is sufficient for optimal radio transmission. Assuming that the maximum horizontal distance that will be used for this project is d2 h d1 Figure 3: Fresnel zone Page 2

3 30m, the minimum height clearance is determined to be 0.97m. Therefore, we ensured that there is at least 0.6m of height clearance from the floor when carrying out the experiments. 3. Materials To conduct the experiments, three Access Points (APs) and a suitable software driver that could extract data from the Network Interface Cards (NICs) of the users laptops had to be obtained. We attempted to look for existing facilities that had 3 or more APs already installed. However, in most of the sites accessible to us, there were at most 2 APs within a locale of usable signal strength. Moreover, it was difficult to obtain the exact positions of the APs, hence this option was ruled out. We had to acquire 3 APs and set them up in pre-defined locations to carry out the experiments. We tested various NICs with their proprietary drivers and software. While most of the software provided facilities to detect signal strength and display them on screen, most could not write these values into a file, or pass them as variables to an external software application. The Orinoco Gold NICs was eventually found to be suitable as its associated software had a logging facility that could be downloaded from the internet. (Agere Systems, Inc. 2003) 4. Methods 4.1 Software development Java was chosen as the development language as it incorporated an operating system-independent graphical user interface and more importantly, it supported multiple platforms and operating systems. This was essential as some of the drivers we found could only work on the Linux operating system, and the test machines were running Microsoft Windows XP. The Java application we developed aimed to display the position of the user graphically. It gave the user sufficient options to customize the output of the test run. The Graphical User Interface (GUI) is created using the JOptionPane class found in Java 2 SE SDK 1.3. (Sun Microsystems, Inc 2003) 4.2 Experiment procedures The signal strength of each AP gathered at every test location (in dbm) was logged into a plain text file by the Orinoco WaveLAN driver. The logging interval is one second. The data logging at each test location lasted for one minute, after which triangulation algorithm was called to process the data. The distance between the user and each AP was calculated from the best-fit line of signal strength against the log-distance graph obtained during calibration. The relationship between signal strength and distance can be represented as such: y = mx + c where y is the signal strength (in dbm) x is log (distance (m)) m, c are constants found during calibration The calculated distance of the wireless NIC from an AP will be the radius of a circle, centred on the AP. The circumference of the circle (representing all the possible locations of the user) is important in triangulation. In an ideal situation, all three circles should intersect at one point, giving the position of the wireless NIC. However, this situation is rarely achieved due to expected errors. Thus, the rules used to triangulate the position of the wireless NIC are as follows: a) When there are more than 3 APs present, only the 3 APs with the shortest calculated distances will be considered for calculation. b) Between 2 circles: Page 3

4 (i) If both circles intersect at two locations, the intersection point closer to the circumference of the third circle is chosen. The distance, between both APs is calculated using Pythagoras theorem. Connecting AP1, AP2 and any one of the intersections together, a triangle is obtained. In this case, AP2 is taken as the reference AP. (ii) If the circles intersection at one location, that point is used for further calculation. (iii) If the circles do not intersect, the point that is the shortest distance between their centres is chosen. (iv) If one circle is within another (no intersection), the outer circle is disregarded, as the AP with the stronger signal is deemed to be more accurate. 4.3 Suitable Locations The locations that were under consideration for the site of our experiments included an office, a park, a cemented pathway and an auditorium. After some trials, the only practical site was the auditorium. The other environments required a much more complex setup, as each of the three APs used had to be powered, and power sockets were not easily accessible in many sites. In a real-life system deployed outdoors however, this hindrance will be a non-issue as the APs would be powered by battery packs. 4.4 Calibration Calibration was carried out before readings were taken for the experiments. The calibration procedure involved placing an AP at a fixed location, whilst a laptop equipped with a wireless NIC is placed at varying distances from the AP. The average signal strength reading over 20 seconds for a particular distance was then taken and recorded. These readings were used to plot a graph of signal strength (in dbm) against the common logarithmic value of the distance. The equation of the best fit line was determined using regression techniques. The gradient of the line and the vertical axis intercept was noted and used for subsequent calculations. In this system, calibration has to be done for each location the system will be implemented in as the results from the calibration exercise is environment specific. However, it should be possible to generalise calibration results for other similar environments. christalite auditorium signal strength (dbm) lg (distance (m)) signal(dbm) calculated signal(dbm)/50mw Linear (signal(dbm)) y = x Linear (calculated signal(dbm)/50mw) y = -20x Figure 4: Calibration data from the Christalite Auditorium. The calibration data collected from the Christalite Auditorium (Figure 4) indicated that the signal strength fluctuated significantly. The data collected had a correct trend, and a regression line could thus be computed. Comparing the regression line (obtained from calibration data) with the theoretical results (using the path loss formula), there is a difference in the gradient of the two lines, indicating that signal loss at closer distances is disproportionately higher than that at further distances. The Page 4

5 gradient of the regression line is , and its vertical intercept is These calibration data is applied to all readings obtained in the experiment to convert signal strength to distance. 4.5 APs positioning at selected location (Christalite Auditorium) The chosen site of the experiment (Christalite Auditorium) had about 200 empty chairs dispersed. The APs were placed as far apart as possible in the auditorium, and they were set to use different channels to minimise interference between them. The wireless client loaded with our Java application was placed at designated spots within the room and signal strength readings were taken. The Java application was launched, the location of the APs and calibration data were entered, and the log files of each one-minute long signal strength data collection was fed in. The data collection was conducted at seven different locations in the auditorium. Using the signal strength information, our java program computed the coordinates at which it thinks we were at. This computed location is contrasted with the actual location to determine the effectiveness of the positioning method. 5. Results and Discussion Figure 5: Experimentation results (actual position vs. computed position) Figure 5 shows all the seven locations where data collection was done in the auditorium. In Figure 5, a unique symbol is used to represent each pair of computed and actual position. Grey and red is used to indicate the computed and actual positions respectively. Table 1 shows the accuracy of the computed position as compared to the actual position. The boundary refers to the imaginary triangle formed by lines joining the three APs. Location x = Distance between actual and computed position x<1m 1m<x<3m 3m<x<5m x>5m Within Boundary Table 1: Summary of results The computed position was more accurate for locations that were within the boundary, as in locations 1, 3 and 4. The computed positions for locations 2 and 5 were less accurate as they were at the boundary. All locations that were outside the boundary had very low accuracies (average error of 8m), Page 5

6 such as in locations 6 and 7. These locations were next to walls, outside the boundary. However, the computed results indicated them to be within the boundary. The greater errors experienced near the walls of the auditorium was likely due to the multipath effects that caused the received signal strength to be erratic. At locations close to any wall, there would be an obvious reflected path of sufficient signal strength to significantly affect the signal strength of the direct transmission path given destructive interference. At locations further away from the walls, the signal strength of the reflected path would not be significant enough to affect that of the direct path, even with destructive interference. This explains the higher accuracy of locations closer to the centre of the auditorium. (Button, Don. 1999, Howald, Rob. 2002) 6. Conclusion This result of this study indicates that a close range positioning system using wireless networks can indeed be constructed, provided that multipath effects are adequately compensated for. This compensation could be in the form of calibration locations that densely covers the operating area, so that multipath effects are incorporated into the calibration data. The alternative would be to limit the usage of such a system to a relatively open area without walls. An increase in the number of APs would ostensibly produce more accurate results. References 1. 3Com Corporation IEEE b Wireless LANs. USA, CA. 2. Jupitermedia Corporation, Understanding Ad Hoc Mode 3. Barry McLarnon, VHF/UHF/Microwave Radio Propagation: A Primer for Digital Experimenters. USA, AZ: Tucson Amateur Packet Radio. 4. Chua, H. H Infocomm Development Authority of Singapore. 5. 3Com Corporation Deploying Wireless LANs. USA, CA. 6. Cisco Systems, Inc Design Principles for Fixed Wireless Access Solutions. USA. 7. Agere Systems, Inc ORiNOCO WaveLAN Sun Microsystems, Inc The Source for Java Technology Button, Don Effect on Obstructions on RF Signal Propagation. EMS Technologies, Inc Howald, Rob Crossed Signals. USA: CMP Media LLC. Page 6

Link Budget Calculation

Link Budget Calculation Training materials for wireless trainers This 60 minute talk is about estimating wireless link performance by using link budget calculations. It also introduces the Radio Mobile