Assisted GPS System Design. Group 815

Transcription

1 Assisted GPS System Design Group 815 May 20, 2009

2

3 Abstract This report describes the design of an Assisted GPS system that holds an exchange of information with an assisting server and computes its position after receiving the necessary data. Along the following chapters we present our algorithm for solving the coarse time error and computing the pseudoranges. The report also presents a short description of the equipment used and an explanation of the way in which the communication works. ii

4

5 Preface The increasing number of mobile facilities has led to new openings in the market of Location Based Services (LBS) when nowadays applications like Turn by turn navigation, context aware notifications and advertising, and E-911/E-112 services are becoming more and more common. Assisted GPS has a crucial role in these applications, providing a fast and precise location for the LBS user. Andrei Marinescu Dragoş Cătălin iv

6

7 Acknowledgements We would like to express our appreciation for all the guidance, patience, and help provided throughout the project along the two semesters that have passed to our supervisor Prof. Dr. Kai Borre and to Dr. Darius Plaušinaitis. vi

8

9 Contents 1 Introduction 1 2 Background A-GPS Advantages Standard versus Assisted GPS Receivers Typical Acquisition Scheme, Cold Start Typical Acquisition Scheme, Warm Start Typical Acquisition Scheme, Hot Start Assistance Information MS-Assisted and MS-Based GPS Frequency Assistance Time Assistance for Code Phase Problem Analysis Positioning Possible Solutions Performance Comparison Project Outline Algorithm Description Design and Algorithms viii

10 CONTENTS Decoding the TOW Solving the Coarse Time Error Program Description Description of Ephemeris Data The Equipment The Base Station The Receiver The Radio Link The Mobile Station - The Rover Measurements and Results 38 7 Conclusion and Outlook 42 A Work Description 43 A.1 Project Journal A.2 Time Planning and Scheduling B The Source Code 46 B.1 The Computation Part B.1.1 Coord.m B.1.2 Test.m B.1.3 Find Preamble.m B.1.4 Acquisition.m B.1.5 Select file.m B.1.6 Recpo ls.m B.1.7 Extract site.m B.2 The Transmission Part B.2.1 Transmit.m ix

11 CONTENTS B.2.2 Receive.m Bibliography 60 x

12

13 List of Figures 1.1 The GPS Constellation [ A-GPS System [ A Complete A-GPS System [googlepages.com/ngps] The Cell ID Method Assisted-GPS [ Assistance Flowchart Preamble decoding [NAVSTAR, 1995] Navigation Data [kom.aau.dk] The azimuth and elevation angle [searchcio-midmarket.com] Flowchart of coarse time solution The SV signal strength The equipment Dimension Plot Location Plot Location Zoomed Plot A.1 Project timeflow xii

14

15 Chapter 1 Introduction Although the Global Positioning System (GPS) was designed for military applications, nowadays the number of civilian applications have by far surpassed the military ones. Due to the latest developments in cost, size, and performance, the civilian receivers have become ubiquitous, being present on the market in everyday objects like cars, PDAs, mobile phones, E-911 / E-112 services, and services that need precise positioning like land surveying, mining, and navigation (on land, water, and air). Because of the latest advances in technology there is an increasing demand for GPS systems integrated in Figure 1.1: The GPS Constellation [ mobile phones. Another decisive factor that affects the integration process are governmental regulations for emergency services. Crucial elements in the daily use of applications are power consumption and speed, so our project presents an approach to the implementation of a GPS based system while paying attention to these previously mentioned factors. The goal of this project is to try to address and correct the most important drawbacks of standard GPS by designing a system in which the receiver s acquisition function is aided by information received from the network. This is 1

16 CHAPTER 1. INTRODUCTION known as an Assisted GPS (A-GPS) receiver. An A-GPS receiver is basically a GPS receiver operating within a mobile telephone network that receives some information from the network assistance server. This information helps the device track and utilize weak signals and reduce the time needed for a first fix (time needed to obtain the first position). Assisted GPS is a way of improving the overall performance of standard GPS receivers by providing the required information through an alternative communication channel, information that otherwise the GPS receiver would have to decode from the satellites. The availability of such information helps increase the receiver s sensitivity which allows it to track weak signals such as those received inside buildings that otherwise would not be available. In our implementation we are using a Software Defined Radio (SDR) receiver, in which we modified the way the acquisition is done and some of the ways it computes the necessary information in order to make it faster to process and to use shorter signal samples. The rest of the document is organized as follows: in Chapter 2 we make a short description of the A-GPS system for a better understanding of our work, Chapter 3 presents the problem definition and the possible approaches to solving the problem, in Chapter 4 we describe our solutions to the problem along with the algorithm that we used, in Chapter 5 we present the equipment used and the way we implemented the transmission, while in the last parts, Chapter 6 and 7, we show the results we obtained and the measurements done, and finally we draw a conclusion based on the work and the results obtained. We have also attached in the Appendix our code together with the project journal and time planning. 2

17 Chapter 2 Background Assisted GPS describes a system that uses a terrestrial RF network to improve the performance of Global Positioning System (GPS) receivers by providing information about the satellite constellation directly to the GPS receiver. 2.1 A-GPS Advantages With A-GPS, a wireless network sends information directly to the GPS receiver, which allows the device to quickly locate the four (or more) satellites needed and process the data contained in their signals. The A-GPS information includes identification of the visible satellites. Because the receiver is now searching only for specific signals, the amount of time it takes for a GPS receiver to obtain its first location or time-to-first-fix (TTFF) is reduced from minutes to seconds. Assistance is also provided to the GPS receiver by sending the ephemeris data for each satellite so that this data doesn t have to be decoded from the GPS signals. The receiver must still obtain signals from at least four satellites to determine the time it took each signal to arrive at the receiver; however, it does not have to decode all the subframes. Assisted GPS effectively increases the sensitivity of the receiver so that it is able to obtain and demodulate the satellite signals in areas where unassisted GPS could not. Furthermore, since the ephemeris data is already provided to the receiver, it can determine position more quickly than if unassisted, even in clear view of the sky. 3

18 CHAPTER 2. BACKGROUND It is important to note that these advantages will be seen primarily under circumstances when the device is in an unfriendly RF environment. The most obvious situation is when the device is first powered on, when there is no valid ephemeris data on the GPS receiver, so the positions of the satellites in the sky are unknown. In this circumstance, the assistance information enables the receiver to obtain a fix more quickly than an unassisted device and in some cases, obtain a position fix where an unassisted device could not obtain one at all. If a GPS receiver has been functioning and has been demodulating the satellite signals prior to entering an unfriendly RF environment, the assistance data is initially unnecessary and offers no advantage. Figure 2.1: A-GPS System [ However, if the receiver remains in this unfriendly RF environment for a period of time, the satellites viewable over its position will change. In addition, the ephemeris data of each satellite will also change, as corrections are made to its orbit on a regular basis. For these reasons, ephemeris data becomes stale and needs to be updated on the GPS receiver. Regular updates of ephemeris data to the receiver enable the device to continue to operate in conditions where an unassisted device would cease to operate. 4

19 CHAPTER 2. BACKGROUND Standard versus Assisted GPS Receivers Based on the information that a standard receiver is provided with at the moment it is switched on, we could find three different types of starts. Cold start A GPS receiver that has just been recently switched on, or due to other reasons, has no a priori knowledge (of the frequency of the signals or of its location) is considered to perform a cold-start. The receiver will have to search a large range of frequencies caused by the Doppler effect of satellite motion and receiver oscillator offset, but also due to a small contribution from receiver velocity. The satellite motion causes the receiver to search 8.4 KHz of unknown frequencies. The receiver motion causes a small Doppler effect of up to 1.5 Hz for each 1 km/h of receiver speed. There will be an additional 1.5 KHz of unknown frequency offset for each 1 ppm of unknown receiver oscillator offset. Therefore, the total range of frequencies that has to be searched is between 10 and 25 KHz. A typical receiver searches this range in bins of 500 Hz each, thus 20 to 40 bins to search. The receiver performing cold-start has also to perform a search of all possible code phase bins. The total code phase search space is 1023 chips. A traditional GPS receiver will have two correlators per channel (with a correlator spacing of half a chip), and will thus be able to search one chip at a time. Therefore, the receiver will have to search a two-dimensional space of about up to ; search that requires at least 20 s (the integration time is 1 ms, so it takes one second to search all 1023 possible phases). Once the correlation peak has been found for each satellite, the search is over, but the receiver cannot compute position until it has decoded the TOW and ephemeris data. The data is transmitted every 30 s, so the expected decode-time is 30 s, if no data bits are lost. The total TTFF is therefore at a minimum, approximately 1 minute, provided that the signal isn t blocked and it doesn t fade during this time, or else data bit errors may occur and the receiver will have to wait another 30 s for the data to be retransmitted. 5

20 CHAPTER 2. BACKGROUND Warm Start In this case, a more common one, the GPS receiver has some idea of the a priori position (from the last known position stored in the memory); it will also have a rough idea of time and of the reference frequency. The receiver will be able to calculate the approximate satellite positions and velocities from the almanac data stored in the memory). In this case the number of frequency bins can be dramatically reduced, but it will still be necessary to decode the TOW and ephemeris data, in order to compute the position, so the TTFF will still be greater than 30 s. Hot Start In this situation the a priori knowledge is very good. It can be the case where the receiver has already decoded the ephemeris for all visible satellites and computed position, then was turned off, and then was turned on a few minutes later. As in a warm start, the receiver would have a reduced number of frequency bins to search. Also, if the real time clock were good enough (precise to more than 1 ms) then the receiver could reduce the code phase search space too, so the acquisition time could be reduced to less than 1 s. And since time is known accurately the TOW doesn t have to be decoded. In this case TTFF can be less than 1 s. The assistance data that is provided to an A-GPS receiver can not only replace the broadcast ephemeris but can also reduce the frequency and code phase search space. To reduce the frequency search space it is necessary to have a rough a priori position, rough a priori time, and satellite orbits; information required for computing the Doppler frequencies. To reduce the second dimension in the search space, the code phase - search space, it is necessary to have a good a priori position (better than ±150 km since in 1 ms the signal travels 300 km) and a good a priori time (better than 1 ms since the code phase can be up to 1 ms). The position is usually known to within a few kilometers (from the location of a cell tower), however time is often not known to better than 1 ms, in which case it s called coarse time, while if it s known to better than 1 ms, the assistance is called fine-time assistance. 6

21 CHAPTER 2. BACKGROUND Typical Acquisition Scheme, Cold Start Firstly, an acceptable frequency bin spacing must be chosen, then search each frequency bin along the code phase axis so that all possible phases are searched in that bin. It is common to search the most likely frequency bin first, and then move outward from there if no signal is found. For a cold start the frequency bin centered at zero would be searched first, then the frequency bin centered at +500 Hz, then the bin at -500 Hz, then the bin at Hz and so on until a signal is found or until the entire search space has been exhausted Typical Acquisition Scheme, Warm Start When a receiver is on, tracking and computing position, it is normal to compute the TCXO (Temperature Compensated Crystal Oscillator) offset and store this value in NVRAM (Non-Volatile RAM). Similarly, a recent position is maintained in NVRAM. Therefore, when the receiver is turned off, and then turned on again some time later, the last TCXO offset and the last known position would be known. A real-time clock is normal to maintain approximate time, and an almanac is normally available in ROM or NVRAM. At a warm start the receiver can use the last known position and approximate time and almanac to compute the expected satellite positions and apparent Doppler values. These values, plus the last known TCXO offset, can be used as the starting point for the frequency/code search. The receiver will still need to decode TOW and ephemeris before it can compute the position. Expected minimum TTFF for a warm start is thus close to 30 s Typical Acquisition Scheme, Hot Start For a hot start things are better. The frequency offset is expected to be well known when the receiver is off for a short amount of time. Furthermore, if the receiver time was maintained to sub-millisecond accuracy while the receiver was off, then the code phase might be known to some fraction of a millisecond before the hot start, and it would not be necessary to search all the code phases. Also, if the ephemeris is already available, and the TOW is already known, then the position can be computed without further information. Therefore, the expected minimum TTFF for a hot start can be as low as one second. 7

22 CHAPTER 2. BACKGROUND Assistance Information Acquisition of signals would be easier if the following information are known: Frequency offset Accurate time Code phase Receiver position Ideally, an assisted cold start would then look like a hot start. The receiver would not have to search many frequency bins, and if the time is accurate enough it would not have to search many code phases either. Once the signal is found, the receiver doesn t have to decode ephemeris data because these would be provided by the assistance data. Figure 2.2: A Complete A-GPS System [googlepages.com/ngps] 8

23 CHAPTER 2. BACKGROUND 2.2 MS-Assisted and MS-Based GPS In A-GPS there are two major approaches, which are known as MS- Assisted and MS-Based GPS. MS is short from Mobile Station, which means the terminal device, the GPS receiver. In MS-Assisted GPS the position is computed at the server and the GPS receiver just has to acquire the signals and send the measurements to the server. In MS-Based GPS the position is computed by the receiver itself. This two different types lead to different properties in each case. For the MS-Assisted case, as long as it doesn t compute the position itself, it doesn t need necessarily satellite orbit data (almanac and ephemeris). The data is stored only at the server which directly computes the assistance data and sends it to the terminal Frequency Assistance The information received by an A-GPS device when turned on is similar with that of a standard GPS performing a hot start. This way the A-GPS device will have knowledge of recent position, time and receiver oscillator offset, along with valid ephemeris for the satellites in view. The information is afterwards used to compute the expected satellite Doppler frequencies and reduce the frequency search space. This implies the same advantages as the ones of a hot start. MS-Based Case For a MS-Based receiver the assistance data comprises time, reference frequency position, and/or ephemeris. With this information the expected Doppler frequency is computed for each satellite. Out of the three components of Doppler frequency (relative satellite motion, receiver motion, and receiver oscillator offset) only the receiver motion is unassisted, but since this component is usually small it can be neglected. The four components of assistance are: Date and Time Delivered as GPS week and seconds of the week. For the purpose of frequency assistance the accuracy needs to be good only to a few seconds. Reference Frequency Local oscillator in mobile phones calibrated by the signal received from the cell tower. The tower frequency is typically 9

24 CHAPTER 2. BACKGROUND known to ±100 ppb. A priori Position Position assistance usually derived from a data base of cell tower positions (mobile phone case). The accuracy of this position is about 3 km, but depends on area (urban - higher precision / rural - lower precision) Almanac and/or Ephemeris Almanac or ephemeris used for computing satellite position. Both can be provided in the MS-Based assistance data. The approximately correct time and position is needed to calculate the relative satellite motion (the expected Doppler frequency). MS-Assisted Case For a MS-Assisted receiver the data comprises reference time, reference frequency, and satellite Doppler. The expected Doppler is computed at the server using the approximate position of the GPS receiver, which is typically the same position that would be used in the MS-Based case. The expected Doppler is provided to the receiver at a reference time. The local oscillator is calibrated in the same way as in the MS-Based case. Afterwards the receiver uses this information to computed the expected frequency for each satellite Time Assistance for Code Phase Time assistance is necessary for the purpose of providing an a priori estimate of code phase. The code phase is a function used for the receiver s position, when the receiver clock generates a local correlator delay. The complete range of possible GPS C/A code phases is 1 ms (before the code repeats itself). In conclusion if the receiver time is not known to a precision of better than 1 ms we cannot provide an a priori estimate of code phase. In this case the A-GPS receiver has to search all possible code phases in each frequency bin, just like a non-assisted receiver. If the receiver time is known to better than 1 ms then the assisted position accuracy becomes appropriate for code phase search. If the assisted position accuracy is worse than 150 km the code phase becomes ambiguous to ±150 km, which summed is 300 km, that is 1 ms (at speed of light) of code phase. If the time assistance is better than 1 ms of accuracy, it is called fine-time assistance. Otherwise it is called coarse-time assistance. Currently the only network providing fine time assistance is the CDMA network. 10

25 CHAPTER 2. BACKGROUND Fine-time Assistance in MS-Based Case In the MS-Based receiver case the data for assisting the code phase search comprises: Fine-time Position Almanac and/or Ephemeris The provided data (fine-time, position, and almanac/ephemeris) is used by the receiver to compute the expected code phase for each satellite. Finetime is also used to synchronize the receiver clock to GPS time. Fine-time Assistance in MS-Assisted Case In the MS-Assisted receiver case the assistance data comprises: Fine-time Expected satellite pseudorange and rate The pseudorange is computed at a server, using the approximate position of the GPS receiver. The expected satellite pseudorange and rate is provided to the GPS receiver at some reference time. Fine-time assistance provides a very fast TTFF at increased sensitivity, within tens of milliseconds. Coarse time assistance on the other hand requires more time to track the GPS signals due to Doppler offset. Though with coarse time TTFF can be reduced to 1 s, many other details have to be taken into consideration. It also implies decreased sensitivity to obtain such a rapid TTFF. 11

26 Chapter 3 Problem Analysis 3.1 Positioning A few decades ago a constellation of satellites was launched by the U.S. Department of Defense to provide global coverage of objects on Earth, and this is nowadays the way each vehicle can navigate throughout the urban jungle and around the globe. A small device does all the job without any assistance from the user else then providing energy to power on and compute its position from the satellites and five other ground stations placed strategically around the globe. Although there have been a lot of improvements in the past decades, there are still a lot of problems to be addressed. For a standard GPS, certain conditions have to be fulfilled in order to obtain reasonable performance, namely a strong continuous reception of signal from satellites, meaning that obstruction and multipath are to be avoided as much as possible during tracking and data acquisition. We also have to take into consideration other factors like power consumption, TTFF (time needed to compute the first position), and accuracy of positioning. In urban areas, where everything is surrounded by tall buildings, it is easy for the signal to be reflected, or not to be tracked at all, making it difficult to meet the criteria mentioned above. Therefore new solutions were needed to solve some of the problems. The evolution of technology and the continuous minimizing of the circuitry and devices made it possible for more and more services to be integrated within diverse accessories, without rising the prices of the devices too much. Navigation is one of the parts that were integrated along this line of evolution. 12

27 CHAPTER 3. PROBLEM ANALYSIS 3.2 Possible Solutions The ever growing mobile phone market calls for more and more services in order to lure the clients. One of the new branches in this field requires the integration of navigation devices with mobile phones. One of the aspects that contributed to the fast development of positioning in handheld devices was the FCC reglementations. The U.S. Federal Communication Commission (FCC) made it mandatory that wireless emergency 911 callers must be located with an accuracy of 125 m or better. To meet the FCC location requirement, wireless network operators can either use a network-based location or a handset-based location. Most networkbased caller location systems employ either the Time-Difference Of Arrival (TDOA) approach or the Angle Of Arrival (AOA) approach to determine the caller s location. Cell Identification - Cell ID The simplest method to position a mobile terminal is to use Cell ID. It is time effective and has a very low power consumption. Figure 3.1: The Cell ID Method Cell ID positioning considers the location of the base station to be the location of the caller. The accuracy of this method can be as good as a few hundred meters in urban areas, but as poor as 32 km in suburban areas and rural zones. The accuracy depends on the known range of the particular network base station serving the handset at the time of positioning. An improved method, Enhanced Cell ID, uses more cells in an urban area and locates the user by intersecting the cells. The method works only in 13

28 CHAPTER 3. PROBLEM ANALYSIS areas of overlapping cells. Angle of Arrival Measurement - AoA AoA uses phased-array antennas to compute the angles at which the signal arrives at the base stations. A minimum of 2 sites is required to compute the caller s location with this method. Unfortunately there are some practical problems associated with the AoA measurement technique. Firstly, the accuracy of direction measurement is directly proportional to the size of the antenna array involved, so direction finding needs antenna arrays of reasonable size at the base stations. Secondly, due to the multipath phenomena, it is not always possible for the network to distinguish between the signals directly from the terminal and the ones reflected by surrounding buildings. The latter problem is a general one for positioning and is not specific to the AoA problem. Time of Arrival Measurement - ToA This method exploits the time of arrival information to locate the mobile position based on the intersection of the distance circles. Obviously three time measurements are required to determine a unique position by triangulation similar to the GPS principle where a circle becomes a sphere in space and an additional measurement is required due to receiver-clock bias. This method needs precise timing (microsecond precision) in order to provide accurate measurements and synchronized clocks. Otherwise, a 1 s timing error could lead to a 300 m position error. Therefore, the ToA method is only used in the techniques based on GPS. Time difference of Arrival Measurement - TDOA TDOA works by accurately computing the time difference of arrival of a signal emitted from the object to three or more receivers. To mitigate the problem of estimating the absolute time used for the radio wave to reach the mobile terminal from a base station TDOA method can be employed which determines the mobile terminal position based on the time difference measurement rather than the absolute time measurement. It should be pointed out that the accuracy is affected by cochannel interference, blockage, and multipath. 14

29 CHAPTER 3. PROBLEM ANALYSIS Assisted GPS The disadvantage of handset-based technologies, not being able to track GPS signals inside buildings, can be overcome using A-GPS technology through which a GPS receiver (embedded in mobile phone) operating within a cellular network receives some information from the network assistance server which helps the receiver to track and utilize weaker signals. The biggest disadvantage of the methods presented so far is represented by the fact that the methods provide only plane positioning with a resolution that doesn t exceed a few hundred meters. As the performance of the satellite-based GPS receiver is improving rapidly and the receiver size and price keep decreasing, it becomes convenient and economical to implement an assisted GPS solution for mobile terminal positioning, which require software and hardware modification to both the mobile and the network. Generally speaking, it is difficult to employ a traditional autonomous GPS receiver in mobile terminals for the following reasons: TTFF relatively long due to long acquisition time of the navigation message ranging from 30 s to few minutes The GPS receiver is incapable of detecting weak signals that result from indoor and urban canyon operations Power consumption of the GPS receiver is relatively high per fix Figure 3.2: Assisted-GPS [ 15

30 CHAPTER 3. PROBLEM ANALYSIS To deal with these problems the A-GPS method can be employed. The reduction in acquisition time and power consumption is due to the fact that the Doppler versus code phase uncertainty space is much smaller than that in a conventional GPS receiver. This allows for a rapid search speed and for a much narrower signal search bandwidth, which enhances sensitivity. Once the embedded GPS receiver acquires the available satellite signals, the pseudorange measurements can be delivered to the server for position calculation or can be used directly by the terminal to compute its position. The disadvantage of the A-GPS is that, in addition to the requirement of a GPS reference network and additional location determination units in the network, the mobile terminal must be equipped with, at a minimum, a GPS antenna and RF (radio frequency) downconvertor circuit. Also, it must make provisions for some forms of digital processing software or dedicated hardware. These facts inevitably lead to increased terminal cost. 3.3 Performance Comparison In general, it can be stated that ToA, TDoA, and AoA show much better results for accuracy and consistency than Cell ID. However, they provide only a moderate yield, because lateration involves at least three base stations and LMUs respectively to be in range, which is not the case everywhere, especially not in rural areas. Method Accuracy Consistency Yield Rural Suburban Urban Cell ID >10km 2-10km m Poor Good TDoA m 40-50m 40-50m Average Average A-GPS 5-10m 5-30m 5-50m Good Good Table 3.1: LBS performance evaluation A-GPS provides the best performance characteristics of all positioning methods. Compared to stand-alone GPS, accuracy is significantly improved due to the consideration of D-GPS correction data during the computation of position fixes. The best accuracy is achieved in rural areas, while in urban areas A-GPS suffers from shadowing effects near huge buildings, but much less than conventional GPS. The yield of A-GPS is quite acceptable, because acquisition assistance data delivered by the network helps to improve the sensitivity of the GPS receiver inside the terminal. 16

31 CHAPTER 3. PROBLEM ANALYSIS Least but not last, a decisive factor that places A-GPS in front of the other positioning systems is the advantage of 3D positioning. A-GPS provides the 3 rd dimension when computing the position, unlike the other systems that offer plane positioning. This ability can be useful in off-ground cases, like tall building or in airborne vehicles. We have chosen the A-GPS positioning approach for our project because of these mentioned facts and also because we believe it is the most promising solution for the future, as the Global Navigation Satellite System advances further and further with the introduction of new satellites through the GALILEO and GLONASS. In our project we tried to implement a system in which an A-GPS receiver communicates with the assistance server. When GPS is designed to cooperate with the mobile network, the network assists the terminal s GPS receiver to improve the performance in several matters. These performance improvements will: minimize the terminal s search window and speed up the receiver s start-up and acquisition times; increase the sensitivity of the receiver; indoor positioning (low SNR situations) is obtained easier through assistance from the mobile network increase the lifetime of the device s battery, due to rapid start-up time There are two methods of network-assisted GPS, MS-Based and MS-Assisted. In the MS-Based method the computation of the terminal s position is done in the mobile device, using GPS measurements received from the mobile network. The data may include reference time for GPS, satellite IDs, and/or satellite position information. The MS-Based method maintains a full GPS receiver functionality in the MS device, and the position computation is carried out by the MS, thus allowing stand-alone position fixes. In the MS-Assisted method, computation of the GPS position is done by the server, with the terminal providing information on GPS satellite pseudoranges, and other information. The computation can take place at the server, or may be carried out in a separate unit, which communicates with the server. For timing assistance the server uses the approximate MS location found by other methods (e.g. Cell ID), together with satellite position data from 17

32 CHAPTER 3. PROBLEM ANALYSIS the GPS reference receiver network to provide the MS with information to assist in predicting the time of arrival (the code phas) of the satellite signals. The information is either directly in the form of calculated times of arrival (MS-Assisted method), or MS location and satellite position data allowing the MS to calculate times of arrival (MS-Based). 3.4 Project Outline During the last semester we have worked on the project Data Acquisition in Assisted GPS, [?], dealing mostly with the simulation of data transmissions between a server and a mobile station and the theoretical part of the implementation. The goal for this semester was to carry on the work we started in the first semester and try to make a physical implementation of the system while developing the A-GPS algorithm further in order to make it work on the hardware with the expected data provided from the assistance server. More precisely, in the project we implemented a system in which assistance information is sent from a server to a MS through a radio link in order to speed up the position computation. The system works by having a BS that broadcasts assistance data to all device in the area covered, and with MSs usingthe data received to increase the sensitivity of the device and the TTFF. We use a Topcon receiver connected to the antenna on the roof of AAU to acquire data from satellites in real time. The receiver is connected to a computer that collects the data acquired by the receiver, and after processing it, and removing any redundant information, it transmits the remaining information through a radio link to another computer that has a SDR receiver attached to it. The purpose of our project is to make the SDR receiver compute the position faster in the aided case then in the standard situation. The time needed for the receiver to compute the position will be decreased due to the fact that the BS will broadcast ephemeris data, the a priori position, the integer part of the pseudoranges, resulting in a more narrow frequency uncertainty space and less information to compute. The key element of the A-GPS system developed by us is obtaining the code phase. While the time needed to acquire the satellites signal will be significantly shortened after receiving the Doppler information, the precision of the tracking will also be increased. Thus the task of detecting the code 18

33 CHAPTER 3. PROBLEM ANALYSIS phase is eased for the receiver. After we obtain the code phase all that is left to do is to recompose the pseudorange, using the assistance data that provides the integer ms pseudorange and the sub-integer ms distance provided by the code phase, and obtain a final position applying the algorithms developed in the program. The steps that the algorithm follows are presentedd in the flowchart

34 CHAPTER 3. PROBLEM ANALYSIS Figure 3.3: Assistance Flowchart 20

35 Chapter 4 Algorithm Description In the attempt to reduce the time needed for a first fix computation we followed two approaches in our project. In both cases our solution was developed based on the SDR receiver designed by Prof. Dr. Kai Borre and Dr. Darius Plaušinaitis, program that requires 37 s of signal to process in order to obtain the results. We have modified parts of the program and we have added others in order to satisfy the requirement related to the length of the signal to be processed, reducing therefore the time needed to get a first fix. The biggest issue in GPS computations is related to time and to the receiver clock errors that have to be addressed; thus, we have developed two methods that attempt to solve this problem. In one of the methods we process 6 s of the signal in order to decode the TOW and obtain precise timing that will be used further on in computations. In the second approach, we followed the method presented by Frank Van Diggelen in his book [van Diggelen, 2009]. The time error is solved by the use of the pseudorange rate speed of each satellite. Both methods will be described in greater detail in the following sections. 21

36 CHAPTER 4. ALGORITHM DESCRIPTION 4.1 Design and Algorithms Decoding the TOW The first approach to solving the time issue is accomplished by decoding the TOW from the message. This involves acquiring at least 6 s of the signal before computing the actual position. While this method implies more time than the method developed by Van Diggelen, it is more precise because it uses the time of the message instead of computing the time based on the pseudorange rate speed of the satellites. In order to compute the TOW we need to detect the preamble first. The preamble is a sequence of bits ( ) that marks the start of every subframe. Normally we would need at least 12 s to make sure we have two preambles in the interval that are 6 s apart, in order to make sure that the 8 bits segment is truly the correct preamble and not another identical sequence of bits that represents something else in the message. There are several other ways to check if we have obtained the correct position of the preamble, from which our version implies the following steps: Detect the possible preambles by correlation with the preamble sequence ( ) For the possible preambles found we check the supposed parity bits for the bit sequence that follows the possible preamble and we recompute the parity bits for that sequence, and in case they match with the supposed parity bits we proceed to the next step We check if the last two bits of the possible Hand Over Word (HOW) are equal to 0, and in case they are we have the confirmation that we have found the real preamble. A problem that has to be taken into account is the fact that the preamble can be inverted ( ) due to the I/Q tracking. We also check the possible inverted preamble with the same method as before and since the initial preamble has 0 values instead of -1 values, the preamble becoming ( ), we multiply the whole possible TLM+HOW word with the value of the first bit of the preamble, thus obtaining the normal case that was solved previously. 22

37 CHAPTER 4. ALGORITHM DESCRIPTION Figure 4.1: Preamble decoding [NAVSTAR, 1995] Another idea for increasing the precision of the result would be by checking the sequence of bits which are supposed to represent the sub-frame id. The possible values should be 001, 010, 011, 100, and 101, since there are only five frames. Thus the other values, 000, 110, and 111, would confirm that we have detected something else than the preamble. 23

38 CHAPTER 4. ALGORITHM DESCRIPTION The obtained TOW is a sequence of 17 bits, which we will transform into a decimal value, and afterwards multiply it with 6 (TOW works in increments of 1, so for a whole subframe we get the result by multiplying it with the length of the subframe which is 6 s) in order to get the time of the next subframe. So in order to obtain the time of the current subframe we need to subtract another 6 s. Figure 4.2: Navigation Data [kom.aau.dk] From the satellites we acquire a sequence of about 6 s, but that doesn t imply that we will have the preamble at the beginning of the sequence. Since the code phase detection starts from the beginning of the sequence, we need to get the time of the detection, so we subtracted the time with the amount the index containing the preamble provides us with Solving the Coarse Time Error The other approach eliminates the need of decoding the TOW. Since it does so, it needs to receive extra assistance data from the BS. With all the assistance data, we will try to speed up the time needed to get a first fix. We eliminated the need of decoding the ephemeris for pseudorange computation, by getting from the BS the integer part of the pseudorange and adding to it the fractional part which is obtained from the C/A code phase 24

39 CHAPTER 4. ALGORITHM DESCRIPTION computed in the acquisition function. But we would still need to process at least 6 s of signal in order to obtain the TOW, for timing purposes. In order to obtain a faster computation we have to shorten the length of the signal, thus we thought of trying the approach presented by van Digglen in [van Diggelen, 2009], so we could get to the point where only approximately 20 ms of signal would be required. Figure 4.3: The azimuth and elevation angle [searchcio-midmarket.com] He presents in one chapter his approach to the coarse time problem. By computing the time through means, other then decoding the signal, we reduce the ms needed to be processed from 6500 ms to just a few. He presents his approach in 4 steps: 1. Start with an a priori estimate of the state (the coordinates of the position and the common clock bias) 2. Predict the pseudoranges obtained from that state 3. Take the actual pseudoranges 4. Update the a priori state based on the difference between the predicted and the actual pseudorage measurements The state vector is defined as: δ x δx = δ y δ z (4.1) δ b 25

40 CHAPTER 4. ALGORITHM DESCRIPTION where δ x, δ y, and δ z are the coordinates of position while δ b is the common bias found in the pseudoranges. The vector of a priori measurement residuals is defined as: δz = z ẑ, with z being the vector of measured pseudoranges and ẑ the vector of predicted pseudoranges. In order to achive the forth step of the method we need the relationship that links the a priori pseudorange residuals δz to the state update δx: δz (k) = e (k) δx xyz + δ b + ɛ k (4.2) where e (k) is the unit vector from the a priori position to the satellite and ɛ (k) contains the measurement errors, including the unmodeled atmospheric delays. For all K satellite we get the matrix equation: where H, the observation matrix is: e (1) 1.. H =.... e (K) 1 δz = Hδx + ɛ (4.3) The equation can be solved for x, by using the least squares method: The predicted pseudorange ẑ (k) is defined as: (4.4) δˆx = (H T H) 1 Hδz (4.5) where: ẑ (k) = x (k) ( ˆ t tx ) x 0 δ (k) t (ˆt tx ) + b 0 (4.6) ˆt tx is the estimated time of transmission of the signal measured x (k) ( ˆ t tx ) is the calculated satellite position at time ˆt tx x 0 is the a priori receiver position δ (k) t (ˆt tx ) is the satellite clock error, at time ˆt tx b 0 is the a priori estimate of the common bias. 26

41 CHAPTER 4. ALGORITHM DESCRIPTION In the situation in which the satellite measurements are affected by the common bias all measurements will be different by the same amount; in the situation of a coarse-time error the satellite measurements will be affected by different values. Therefore, if we take into consideration also the coarse-time error things will get more complicated and the state-update vector will have the coarse-time error included. Having the relative velocity and the clock rate of the satellites, we can compute the pseudorange-rate obtaining the coarse-time error: where: ẑ (k) (ˆt tx ) ẑ (k) (t tx ) = ẑ (k) (ˆt tx ) ẑ (k) (ˆt tx + δ tc ) = ν (k) δ tc (4.7) δ tc is the update to the a priori coarse time state t tx is the actual time of transmission ˆt tx is the coarse time estimate of t tx ν (k) = (e (k) v (k) δ (k) t ) is the pseudorange rate (based on the azimuth and elevation angle -Figure 4.3) e (k) is the unit vector from the receiver to the satellite k δ (k) t is the satellite clock error rate So the state-update vector, now becomes: δx δy δx = δz δb δ tc This means that equation 4.2 will become: (4.8) δz (k) = e (k) δx xyz + δ b + ν (k) δ tc + ɛ k (4.9) For all available satellites, we get the matrix equation: where: δz = Hδx + ɛ (4.10) 27

42 CHAPTER 4. ALGORITHM DESCRIPTION e (1) 1 ν (1)... H = e (K) 1 ν (K) (4.11) The equation can be solved for δx provided that there are at least five independent rows in H, meaning that we need measurements from at least five diferent satellites in order to produce the five lines in matrix H. Figure 4.4: Flowchart of coarse time solution 28

43 CHAPTER 4. ALGORITHM DESCRIPTION 4.2 Program Description For the development of the project we have used Matlab version 8.0 provided by Aalborg University. Testing on the Satel radio modems was done on Satelline Saterm 4.0.7, but the communication currently works only based on the Matlab programs. The software used with the Topcon receivers was PC-CDU, a freeware downloaded from the Topcon website. Our program uses code from both the EASY suite developed by Prof. Dr. Kai Borre, which can be downloaded for free from the Internet address: website and the code developed for the GNSS SDR receiver [Borre et al., 2007]. The EASY suite helps in providing the receiver s position in the TOW case and provides the satellite s position in the Van Diggelen case. The GNSS software is used for decoding the data recorded by the SDR frontend. Both EASY and GNSS software suffered slight changes in order to adapt them to our case. The description of both packages is done in the previously mentioned website and in the [Borre et al., 2007]. The program package is distributed in two segments, one at BS and one at the MS. The BS holds the function necessary to select the latest data from the ephemeris, and deletes the previous written files when it receives a call from the BS. After selecting the proper ephemeris file and formating both the assistance data file and the ephemeris file it sends the information to the base station. The selection of the file is done by the program delete_file.m. It works by selecting the.tps file created in the last minute (using the computer s time) by the receiver while checking the time of the file, and afterwards it deletes all other.tps files in the folder. The.tps files are generated by the Topcon receiver program every one minute, with respect to the settings made in the PC-CDU program. After the assistance data is generated, we transmit first the data file which comprises the assistance data mentioned before, and since the total amount of information is well under 1 KB we can send it with one step only. The more difficult part comes with the outputfile which contains ephemeris data which is organized in 22 lines, as we have 22 ephemerides overall for each satellite. The number of columns is the same as the one of the detected satellites by the receiver. Since one line contains less information than 1 KB, we will transmit the information line by line and receive it line by line at the other end and compose the whole ephemeris matrix afterwards. 29

44 CHAPTER 4. ALGORITHM DESCRIPTION Figure 4.5: The SV signal strength After the whole information is received, the BS s work is done and the program focuses only on the MS job. The MS acquires the satellites after receiving the assistance data, and because of the urban condition and of the quality of the antenna it will detect just a part from the total number of satellites observed by the Topcon receiver. Our experiments have shown that while the Topcon acquires 11 satellites the SDR receiver will detect just 6 or 7 of them, which would be just enough considering that the van Digellen formula needs only 5 satellites to get a fix. After detecting the available satellites the program computes the code phase for all of them and selects the coresponding ephemeris data from the outputfile file while deleting the unused satellites data. Since the sampling rate of the SDR frontend is about 16 MHz we expect a precision of data of 20 m for each pseudorange. We compute the code phase by dividing the obtained code delay to the sampling frequency and multiplying the result with the chip rate of the CA code. The obtained code phase will eventually be multiplied by the speed of light and divided to the chip rate in order to get the sub-integer part of ms. In order to use the van Diggelen formula we need to obtain the measured pseudoranges from the receiver to the satellite at the coarse time moment and to obtain also the elevation angle and the azimuth for each satellite. Both these problems can be solved by using Kai Borre s 30

45 CHAPTER 4. ALGORITHM DESCRIPTION satpos.m function. We will further on compose the H matrix necessary for the computation of the correct position and we will construct the pseudoranges from the integer part and the computed sub-integer ms part. The final part is to iterate the 4 steps provided by van Diggelen as long as the difference between the last two coarse time errors is less then 1 ms. The final position obtained when the condition becomes true is the computed position of the MS. The program also computes the position using the whole 6500 ms using the TOW version. This way we eliminate the need of computing the elements of the H matrix but the TTFF is increased by almost 6 s. For this case we use Kai Borre s recpo_ls.m file, which computes the receiver s position based on the pseudoranges, fine time, and ephemeris data. This method provides us more accurate results also, while it uses precise time instead of iterating till it gets a good estimate. The computational power is also substantially reduced. An assumption that we made is that the receiver s clock precision is within 1-2 s from the GPS time, as the GSM time accuracy is mentioned in [3GPP, 2006]. 4.3 Description of Ephemeris Data The ephemerides are transmitted as part of the GPS navigation message in form of predicted orbital parameters. A network of 12 monitor stations collect GPS data on a continuos basis. The data is then transmitted to the GPS MCS where it is fed into a Kalman filter to produce orbital parameters. After applying all these corrections, a new set of parameters is obtained, which is used to compute the satellite coordinates at the time of transmission. Ephemeris records are updated every hour, therefore to ensure high positioning accuracy only fresh ephemeris records must be used. A user must examine the IODC and IODE parameters to detect any changes in the clock correction and ephemeris parameters, respectively. The ephemeris data can be obtained from the navigation message in the GPS receiver. Table 4.1 shows an example of ephemeris information from the navigation file SITE247J.01N. Once the beginning of subframe 1 is found the following information can be obtained: the week number (starting from midnight January 5, 1980), 31

46 CHAPTER 4. ALGORITHM DESCRIPTION Line Ephemeris information D D D D D D D D D D D D D D D D D D D D D D D D D D D D + 06 Table 4.1: Ephemeris data T GD, the satellite clock corrections: T oc, a f2, a f1, a f0, IODC (the LSB of the IODC will be compared to the IODE of subframe 2 and 3 and whenever these three data sets don t match new data must be collected). The navigation data obtained from subframe 2 are: IODE, C rs, n, M 0, C uc, e, C us, A, and the reference time t oe. From the 3 rd subframe are extracted the following parameters: C ic, Ω 0, C is, i 0, C rc, ω, Ω, IODE, and IDOT. Table 4.2 describes the information that can be extracted from a navigation message. These parameters will be used for computing satellite position and also for correcting time errors [Strang & Borre, 1997]. Since we were limited by the fact that one radio-modem didn t receive packages properly we decided to simulate the MS-Based case, but otherwise we could have sent back the composed pseudoranges or just the code phase and let the Base Station compute the final position and provide it back to the end-user as we initially planned, the MS-Assisted case. 32

47 CHAPTER 4. ALGORITHM DESCRIPTION Line Description V ariable V alue Unit 1 Space Vehicle Number SVN 7 Epoch: Time of Clock T oc 2001/9/4 at 9:59:44 Clock bias a sec Clock drift a sec/sec Clock drift rate a 2 0 sec/sec 2 2 Issue of Data, Ephemeris IODE 228 Amplitude of the sine harmonic correction term to the C rs meters orbit radius Mean motion difference from computed value n radians/sec Mean anomaly at reference time M radians 3 Amplitude of the cosine harmonic correction term to the C uc radians argument of latitude Eccentricity e Amplitude of the sine harmonic correction term to the C us radians argument of latitude Square root of the semi-major axis A meters 4 Time of ephemeris T oe sec Amplitude of the cosine harmonic correction term to the C ic radians angle of inclination Longitude of ascending node of orbit plane at weekly Ω radians epoch Amplitude of the sine harmonic correction term to the C is radians angle of inclination 5 Inclination angle at reference time i radians Amplitude of the cosine harmonic correction term to the C rc meters orbit radius Argument of perigee ω radians Rate of right ascension Ω radians/sec 6 Rate of inclination angle IDOT radians/sec Codes on L2 channel 0 discrete GPS Week Number WN 1130 week Data Flag for L2 P-Code 0 discrete 7 Space Vehicle Accuracy (index for User Range Accuracy) SVA index Space Vehicle Health (index:0=all data are OK, 1= some SVH 0 index or all data are bad) Estimated Group Delay Differential (L1-L2 correction T GD sec term) Issue of Data, Clock IODC Transmission time of message sec Table 4.2: Ephemeris parameters 33

48 Chapter 5 The Equipment 5.1 The Base Station For simulating the functionality of the Base Station we have used a computer that is connected to a receiver in order to collect real time data from the antenna and to transmit through a radio link the data collected The Receiver The receiver we used is a dual frequency Legacy-E Topcon receiver. The Legacy-E is developed around a full function Euro Card design, which allows use of GPS and/or GLONASS L1 and L1/L2. Other feature include: Low signal tracking Multipath Mitigation for both code and carrier Allow GPS L2 tracking in presence of Anti-Spoofing L2 C/A code ready Two RS232 serial ports, and can be equipped with up to four. Courtesy of topconeurope.com 34

49 CHAPTER 5. THE EQUIPMENT The receiver acquires data from the antenna positioned on the roof of the GPS Department in the AAU campus. Using the PC-CDU program provided by Topcon we have acquired the satellite information in the form of.tps extension files. Using a basic application tps2rin.exe these files are converted into navigation and observation files. Every minute a new file is generated and the previous ones are deleted. These files are then read by a Matlab program that selects only the necessary information. Due to the fact that the transmission speed is not very high and that the buffer of the transceivers is limited to 1 KB we decided to eliminate any information that is not necessary, or that is considered redundant. Therefore, after processing the files generated by the PC-CDU program, by using a MatLab program there will be generated another two files much smaller that contain only the data needed to be transferred. Using the Matlab function extract_site.m we will extract from the observation file the PRNs for each satellite acquired, the pseudoranges, the Doppler, and the a priori position. The program starts by loading the observation file. It skip all the header because there is no information needed for later computations and goes directly to the last epochs. It reads the pseudoranges from the last two epochs in order to check the difference between the pseudoranges. If the difference between the same pseudoranges from the two epochs is bigger then 280 km then the pseudoranges from the last epoch, the ones that will be used for later computations, will be corrected by that difference. The IDs of the tracked satellites, the checked pseudoranges, the Doppler values, and the a priori position are all saved in the data file, to be used at the MS site. From the observation file name is determined the navigation file name, because only the last letter of the extension differs. The same thing happens also for the navigation file, from which we extract the ephemeris data from the last epoch, neglecting the others. Twenty one ephemerides values are extracted from the navigation file, which are afterwards saved in the outputfile for further computation. Some of the ephemerides that appear in the navigation file are neglected due to the fact that they will not be needed later, or because for the current generation of satellites that are on the sky their values will always be zero. 35

50 CHAPTER 5. THE EQUIPMENT Figure 5.1: The equipment The Radio Link In order to simulate the transmission between the BS and the MS we have used two Satel Modems connected to the computers which communicate through Matlab using the computer s serial ports. Basically only one of the receiver will transmit data, and the other one s job is just to send a start request and record the received information. The drawback of the Satel modems is the fact that the top Baud rate where the transmission is reliable is baud per second, which limits the speed of transmitting the whole necessary data. It will need a few seconds to transmit the ephemeris and assistance data. The size of the total information to be sent is about 5 KB, most of it being the ephemeris data (around 4.6 KB). Also the buffer size is 1 KB, which forces us to sent several packages in order to transmit the data and afterwards recompose the contents. The assistance data to be sent will comprise the available satellites together with the corresponding pseudoranges, Doppler frequency, and a priori position. 36

51 CHAPTER 5. THE EQUIPMENT The ephemeris information consists of 22 elements for each satellite, this way eliminating the unnecessary information from the ephemeris which would just increase the amount of data to be sent. While at the BS we have used a Topcon receiver from which we decode the necessary data, at the MS the job was done by the software receiver. Since the system simulates the real-time case, but as long as Matlab (and not any dedicated hardware) will do all the computations, we can just estimate the total time. The software receiver will record a sequence of 6.5 s which will be used later on for both methods presented previously. In the TOW method case the whole 6500 ms will be used, while in the van Diggelen method we will use just the first 20 ms bits, since we need to decode just the code phase. The code phase will be used to recompose the pseudoranges from the integer part of pseudoranges in ms computed from the assistance data. The code phase represents the amount the pseudorange has travelled in a one ms interval besides the integer part of ms. 5.2 The Mobile Station - The Rover At the MS site we have the computer that gets the data transmitted by the Base Station, and passes it on to the SDR receiver that is connected to it. With all the assistance data, we will try to speed up the time needed to get a first fix. We eliminated the need of decoding the ephemeris for pseudorange computation, by getting from the Base Station the integer part of the pseudorange and adding to it the fractional part which is obtained from the C/A code phase computed in the acquisition function. We still need to process at least 6 s of signal in order to obtain the TOW, for timing purposes. In order to obtain a faster computation we have to eliminate the need to decode so many seconds, so we thought of trying the approach presented by van Digglen in his book [van Diggelen, 2009], so we could get to the point where only approximately 20 ms of signal would be required. He presents in one chapter his approach to the coarse time problem. By computing the time through means, other then decoding the signal, we reduce the ms needed to be processed from 6000 ms to just a few. 37

52 Chapter 6 Measurements and Results In order to evaluate the accuracy of the system, two tests were developed. In the first situation the MS and the BS are in the same place, while in the second test the BS and the MS were approximately 1 km apart. The purpose of the first test is to examine the behaviour of the system in an ideal situation, in which there are no buildings or trees obstructing the signals. The following test takes place in an urban location, within Aalborg University Campus (green spot),with the GPS Center antenna acting as the reference station. We have moved out with the mobile station at about one km away, which is the approximate interval of GSM antennas in a normal city (it varies from a few hundred meters to up to 1 or 2 km). Using Google Maps and adjusting the scale we have obtained a plot with the preliminary positions which converge to the final result. It can be clearly observed that the preliminary locations follow the path of a line towards the actual location, with the density of points increasing towards the real position. This is also shown in the 3D plot in Figure 6.1, which takes the Z axis into account, and follows the same line pattern, but this time in three dimensions. It can be noticed on both Figures 6.2 and 6.3 that the overall position error is less than 50 m, which matches the results of the A-GPS 5-50 m precision studies. The algorithm iterates the computations of the position untill it fits the fine time condition. This means that the difference of time obtained between two iterations will be less than 1 ms. Another fact that can be observed from the graph is that some of the intermediate positions are closer to the actual position measured with the professional receiver (blue spot). If we change the iteration condition for this particular case we could obtain a 38

53 CHAPTER 6. MEASUREMENTS AND RESULTS Figure 6.1: 3 Dimension Plot result which will vary by only 1-2 m from the actual position, but overall it would not benefit the program, because our experiments have proved that the s limit is the best condition. We plotted the results over two Figures (6.2 and 6.3), the first one includes the AAU reference station (green spot) and the second one is the zoomed in case where we can clearly observe that the final results converge to a single point. By increasing by a factor of 10 (0.1 ms) the stop condition of the iteration, the position will differ by just a few decimeters but the computation power and time needed increase drastically, making the coarse time case become less feasible than the TOW method. Method Deviation X Axis Y Axis Z Axis TTFF Coarse Time m m m 1-2 s TOW m m 5-20 m 6-7 s Standard GPS 5-10 m 5-10 m 5-20 m s Table 6.1: A-GPS performance evaluation 39

54 CHAPTER 6. MEASUREMENTS AND RESULTS Figure 6.2: Location Plot Figure 6.3: Location Zoomed Plot 40

55 CHAPTER 6. MEASUREMENTS AND RESULTS While with the TOW method we obtain a position that is within 25 m from the actual position, the second method provides a result that is 45 m away but the needed time is 1-2 s instead of 6 s, and it s good enough for a preliminary guess. Afterwards we will switch to the TOW position to obtain a more precise result if needed in specific applications. One drawback of the algorithm implemented is that it doesn t provide positioning with an accuracy of up to a few meters. Our opinion is that this is due to the fact that the code phase is not obtained accurately by the receiver and that we don t have professional ionospheric and tropospheric delay adjustments implemented in the program. Future work could focus on optimizing the code and the position result by correcting the errors caused by the previously mentioned factors. 41

56 Chapter 7 Conclusion and Outlook In our previous work in A-GPS [Marinescu & Cătălin, 2009],we focused on simulating the functionality of an A-GPS system by using Matlab. Furthermore, we studied and obtained several measurements and results to indicate the accuracy of an assisted system in comparison to a standard one and also in comparison to the other popular navigation systems like AOA and TDOA or Cell ID. We also worked on reducing the time neccesary to obtain a position fix and we analyzed the feasibility of a practical implementation and its requirements. In the work presented in this report, we took our studies on A-GPS one step further by designing and implementing such a system. Our studies are relevant in terms of precision and time to first fix, because further analysis on mobile devices hasn t been done yet. This could also be part of a future project where we obtain a device which is programmable at the core level of the GPS module in order to be able to allow assistance data for the phone. The experiments have proved that the position precision is within 5-40 meters of the true value. Also in the end the accurate time is provided to the user. The total time needed to compute the first position in the van Diggelen case is about 2-3 s, while in the TOW case we would need about 6-7 s. A diagram with the position accuracy is shown in the Figure 6.3. All the tests were done in the Aalborg University campus by using as reference station a receiver connected to the GPS antenna from the Danish GPS Center. In a future work, the implementation code could be ported from Matlab to C in order to make a real-time system. Another improvement could be in the field of precison, by trying to correct the pseudoranges using EGNOS messages, which should provide a better accuracy to the measurements. 42

57 Appendix A Work Description A.1 Project Journal Flow of project progress during months: February March - Start of the second semester - Thinking about the project options - Decided to continue the project from the last semester and gathering information from [van Diggelen, 2009] and [3GPP, 2006] - Trying a mobile phone oriented approach. Started studying mobile phone programming. Obtained a Nokia N95 (A-GPS compatible) device from the Mobile Communications department to make tests on. - Started studying and working in the Java for Mobile environment. Developing basic programs that use the Location Based Services library - Trying to program the core level of the GPS module in java, consulting with Postdoc Gian Paolo Perucci from the Mobile Communication department. Eventually we didn t manage to implement functions to modify the core level of the GPS module because the access is restricted to normal users. 43

58 APPENDIX A. WORK DESCRIPTION April - Decided to move on to an alternative hardware simulation. Using the devices provided by our department - Studying first the transmission system used together with the devices manuals and developing programs to make the transmission between two computers using Matlab connected through a radio-link - Focusing on developing the Matlab program to work with the current hardware system. As the transmision program was finished we moved on to the actual A-GPS algorithm implementation - Modifying code from the EASY suit and the GNSS SDR in order to adapt it to our pograms - Managed to finish the main program but the results are not satisfactory. Tyring to find the cause. - After a week of failures in finding the cause the solutions are found for both methods and the results obtained are accurate enough. - Minor details in the program are added May - Starting doing field tests in the campus in different scenarios. Analyzing results - Focusing on report writing and preparing it for delivery for the 20 th of May 44

59 APPENDIX A. WORK DESCRIPTION A.2 Time Planning and Scheduling Figure A.1: Project timeflow 45