Online Benchmarking tool
1. Purpose of Project:
The purpose of the project is to study existing Network Based techniques used to estimate location of a mobile user and then to develop an online benchmarking tool for major network based techniques. The tool must evaluate results, in a systematic manner, using different techniques by taking input from user. It must be capable of comparing results uploaded by the user to results computed by its own algorithms. The parameters of performance may include availability, accuracy, robustness, reliability and cost. The techniques may include Cell ID, time/angle of arrival, signal strength etc.
2. Introduction:
The positioning techniques have gained much importance over last few years. The rapid growth in the market for consumer positioning devices is stimulating research into ways to extend positioning coverage into indoor environments, where satellite signals are not available. The location based services is expected to be the next market for mobile operators and users. This rapid growth is mainly because of two reasons. Firstly, the mobile operators are ordered by FCC and EU to locate an emergency caller, with certain accuracy. Due to large number of mobile users, a lot of emergency calls originate from such users. Thus any mobile operator should be able to locate an emergency user with accuracy of less than 125 m for 67% of the cases. Second interest in exploring the position is commercial prospective. Thus it can be used to provide Location Based Services (LBS) to a mobile user. Hence there is a lot of interest in this field as seen from operators as well as users point of view. A mobile operator is interested in amount of capital which must be interested to deploy certain project. From this view point LBS has a vast market.
Some of the applications of Mobile position location are as follows:
- Network performance
- Subscriber safety
- Location sensitive billing
- Intelligent transport systems
It is believed that location based services will play a key role in future wireless market. It is estimated that location based services (LBSs) will generate annual revenues of the order of US$15 billion worldwide [1]. In the United States alone, about 170 million mobile subscribers are expected to become covered by the FCC mandated location accuracy for emergency services [1].
The location techniques can be grouped in two main categories, Handset Based positioning and Network based positioning. Handset Based techniques needs special type of handset hardware or the installation of specialized network software in existing handsets e.g. GPS and A GPS. In it GPS receiver determines its own position by sending and receiving signals from at least four satellites. The time to reach satellite signal to GPS receiver is used as parameter. The accuracy of GPS based systems is very high. And the coverage of satellite is very good in outdoor environment. However, drawback is inability of GPS systems to operate in indoor environments where signals of the satellite cannot be fully covered. Also Embedding a GPS receiver into mobile devices leads to increased cost, size, and battery consumption [1]. Thus hundreds of millions of handsets in market need to be replaced or modified. In contrast, Network Based Techniques determine the position of a mobile user by measuring its signal parameters when received at the network Base Station (BS). Here BS receives signal from Mobile Station (MS) and sends them to a central site where location estimation algorithms are used to estimate location. In this method there is no need to change or replace existing handsets. It would require change in a few thousands of network nodes (Base Stations). Hence, it can be implemented easily, less costly & using existing technology. And still can provide a very good estimate of position of a mobile user. The Network based positioning is also feasible for network operators as it would help them to implement location sensitive billing and location related services. However the main isuue with this technique is its accuracy. Due to multipath propagation, fading and low SNR there may be large errors in position estimate. The project will emphasize on comparing various existing network based location techniques. And then to develop and implement an online benchmark for comparing location evaluation estimates using these techniques.
2.1. Self Positioning:
In self positioning systems geographically separated receivers synchronizes themselves with transmitter and locate their position. An example is GPS, where a specific handset with GPS hardware locates itself. A GPS is also an example.
2.2. Remote Positioning:
In remote positioning systems a number of transmitters locates geographically distributed receivers. The Signal between receiver and transmitter are used for this purpose. It may further be classified to Network Based Technique & Handset Based Technique. In the former Network Nodes (BS) estimates mobile location using some signal parameters while in the latter case MS measures its position by using satellite signals.
2.3. Indirect Positioning:
In indirect positioning the estimate of location can be made at self positioning receiver and then transmitted to a remote receiver & vice versa. It can be classified in to Indirect Remote positioning or Indirect Self positioning. A self positioning receiver that sends position data to remote location will be indirect remote positioning [2]. A remote positioning system transmitting an object's position to the object is called Indirect self positioning [2].
3. Review of Existing Methods:
Various techniques have been devised in recent years to locate MS using received signal parameters at BS. Usually, for location estimation, two operations must be performed at the BS. The BS has to measure some signal parameters (such as the time of arrival or the angle of arrival) of the received MS signals. Then, the measured signal parameters are combined in a data fusion stage to provide the final estimate for location [1]. The major issue is to deal with multipath and fading channels. The CGI+TA and UL TOA positioning techniques were standardized in May 1999 [8].
The overview of all of these methods is given below.
3.1. Cell ID : [1,3]
3.1.1. Basic Cell ID:
It is easiest way to locate a mobile user. In this method the longitude and latitude of serving antenna are used to estimate the position. It is simple and easy to implement. It also does not need any network or handset upgrade. But because the accuracy depends on cell size thus the mobile can be anywhere within a cell. The accuracy may vary from tens of meters in urban areas to tens of kilometres in suburban areas. The performance of cell id is also affected by the fact that MS is not always connected to closest cell. Hence cell id cannot be used, in its pure form, for a good estimate of position of MS. The cell ID can be made little better by using sectored cells. Thus by dividing area of each cell into sectors the position of mobile will be better estimated.
3.1.2. Enhanced Cell ID:
Certain enhancement in basic cell id approach has been made to improve its performance. These include Cell ID + RSCP (Received Signal Code Power), Cell ID + RTT (Round Trip Time) for UMTS, Cell ID + TA (Timing Advance) for GSM. RTT is the time that a signal takes to travel from mobile station (MS) to base station (BS) or vice versa. It can be implemented alternatively by using the time signal takes from transmitter to receiver and then back to transmitter. The resulting time will be twice the time of single side. The former method requires an accurate synchronization between transmitter and receiver. Thus a BS should know accurate time at which mobile started to transmit. The latter does not require a strict timing requirement. It is thus most commonly used. The accuracy of RTT is better than the TA due to shorter chip duration and larger bandwidth in UMTS. The accuracy is also enhanced when LOS propagation becomes dominant.
3.2. Angle of Arrival or Direction of Arrival:
This technique is very similar to TOA except that the angle of received signal from MS is used to estimate the location. It is also called direction of arrival because we can estimate the direction of arrival of signal. Ideally only two BS will be enough to estimate location of MS using AOA. However, in practical system more measurements should be taken to improve accuracy. The AOA measurement from MS to BS will give a straight line. Another measurement will give another straight line. The two lines will intersect at one point which will be the location of MS.
3.3. Time Delay:
EM waves travel at a constant speed (equal to speed of light) in free space. One can estimate distance between two points between which waves travel. The method is used almost universally by the satellites. The time delay method makes use of the fact that time taken by the waves to travel between MS and BS can be used as a measure of distance between them.
The time delay may be of two types:
- Time of Arrival or Absolute time
- Time difference of arrival or Differential time
3.3.1. Time of Arrival:
In this method we take the absolute time a wave takes to travel between a transmitter and receiver or vice versa. Alternatively, it may involve half the time a wave takes to travel from transmitter to receiver and then echoed back to transmitter or vice versa. However, in former case a strict timing requirement is needed, thus the transmitter and receiver must be synchronized. In the latter case a strict timing is not required. It is thus most commonly used measurement. Time of arrival has the advantage of accuracy.
3.3.2. Time difference of arrival or Differential time:
The time of arrival technique uses a strict timing requirement. In practical systems there is timing error due to noise and multipath. To overcome this problem, Time Difference of Arrival is used. In this technique several transmitters (Base Stations) are synchronized to a common time base. The time difference of arrival from MS (Mobile Station) is then measured at all of BS. This method further improves the accuracy of TOA technique. In 3G systems time difference of arrival is known as Observed Time Difference of Arrival.
3.4. Database Techniques:[5,6]
All of the AOA, TOA & TDOA techniques are based on assumption of LOS path. However, in most of practical case direct LOS path may not be available due to multipath and fading. The database technique is efficient in such cases. In this technique the signal information seen by MS due to different BS is stored in database. The database is often called fingerprint database. To estimate location the signal samples are compared with fingerprint database to get accurate estimate. Here use of the fact that signal strength is dependent on the position. The database correlation method is best for heavily populated urban areas. It also can be easily mitigated for NLOS and fading problems. However, the major issue is to create such a database for big wireless networks.
3.5. Pilot Correlation Method:
This method is based on pre measured stored values of all visible pilots. PCM (pilot correlation method) require a database with stored pre measured samples over the network coverage area [7]. The database is prepared by dividing the area into smaller regions also called positioning regions. The size of region depends on required accuracy. The received signal code power (RSCP) for most common pilot channels for each region is stored in database. For estimation of location measured values are compared with stored values. Finally Least Square method is applied to get a location estimate. The PCM method can be easily implemented.
4. Online benchmarking tool:
The Main purpose of the project is to develop an online benchmark tool. This tool should facilitate user by evaluating comparative results of different positioning techniques. The specifications of such a tool are provided by 'Vodafone Greece' and are shown in [9].
The two main requirements are:
- Software requirement
- Algorithm requirement
The Software can be Java, C++ or Matlab in which the algorithm will be implemented. The reason to use these languages is that they are more flexible and easy to implement and understand.
The Choice of Algorithms depends on number of different factors. The Most important of them are availability, easiness, robustness and accuracy and implementation cost. A number of algorithms are discussed in section 5.
The functions of the benchmark tool should be as under:
4.1. Datasets Of Network Measurements:
To locate the position of a mobile use is made of the measurements made by mobile during its normal operation mode. These measurements may include received signal strength by different nearby base stations, Cell ID of each base station and more measurements regarding network parameters. [9]
The second part is that of network description. As the operator only knows the network description so the positioning algorithm must run at network side. [9]
4.2. Download of Datasets:
The Network Measurement Reports and Network Description will be available to download as a simple file. [9] The NMR's and network description should be altered properly so as not to disclose network related sensitive information. Thus the security of network operator information must not be lost.
The datasets can further be categorized in two parts:
4.2.1. Testing Dataset:
The function of testing dataset is to develop new positioning algorithm. These dataset will consist of NMR and network description files. [9]
4.2.2. Validation Dataset:
The function of validation dataset is to benchmark evaluation of an algorithm. These dataset will also consist of NMR and network description but NMR will be slightly different. The NMR in this case will not include longitude & latitude information. This is due to the reason that this information must be provided by algorithm. The measurements at different times from algorithm & network surveying were then compared so as to validate performance and accuracy of algorithm.
4.3. Uploading of positioning results
The benchmark user must be able to upload results of his testing algorithm, in order to be validated. The Uploaded results must be in format as shown below:
Sample Time |
Positioning Longitude |
Positioning Latitude |
59:02.5 |
23.73170505 |
38.05285935 |
It shows the position (long., lat.) at a certain time as estimated by algorithm. The tool should be able to give user the flexibility to comment on his results, upload of files, or even the code used by him for positioning. [9]
4.4. Automatic Evaluation of submitted solution
After the upload of positioning results from user the most important part of the benchmark tool starts. This is the basic and most important function. Now it should perform an automatic assessment & evaluation of the performance of the user submitted results. [9] To evaluate the results we must set some criterion. Such type of criterion, known as, Key Performance Indicators (KPI), make sure that the results are up to standard. Such KPIs may include Accuracy, Reliability, Availability, Robustness, Flexibility, Time to execute etc. [9] The implementation of many criterion at same time is however a challenge in automatic assessment systems. [9] For this reason the evaluation of accuracy based KPIs is first implemented and assessed on a server side running engine. [9]
4.5. Accuracy based KPIs:
A list of KPIs should be extracted that indicate different measures of positioning accuracy. [9] The matric for the algorithm is the distance between distance estimated by algorithm and the actual position of terminal. [9] The following KPIs should be derived and calculated: [9]
4.5.1. Sample Statistics:
The following statistics of positioning errors should be calculated: mean,Standard deviation, maximum values. [9]
4.5.2. Frequency Distribution:
The distribution of frequencies at which the different positioning errors occur should be calculated. [9] Hence the error distribution for the median,the 68th and the 95th percentiles should be calculated. [9]
5. Algorithms and implementation:
Several algorithms have been developed to implement the above mentioned techniques. Due to large number of such algorithms in literature only most important and popular have been used. The project specification provided by Vodafone Greece provided a guideline for them.
For each of the technique two important algorithms have been studied and one of them has been selected. The selection criteria are availability, accuracy, ease of implementation, modification requirement and environment.
Below is a description of algorithms for each technique.
5.1. Cell ID Based:
Two important algorithms for CID +TA/RTT method are in [1a, 2a]. The [1a] is to implement the CID with timing advance in GSM to improve accuracy of basic cell id technique. The [2a] is used to implement CID with round trip time in 3G networks. It is based on estimating the time taken by the signal from BS to MS and back to BS. This is done by using forced soft handover (SHO) technique. The accuracy is greatly improved using soft handover (SHO).
5.2. Time of arrival:
The time of arrival algorithms are described in [4a, 5a]. First algorithm is primarily for NLOS condition. It is a deterministic model to estimate location of mobile. A modified deterministic model is also proposed to increase accuracy of the algorithm. Second is an algorithm used to implement narrow band TDMA signal. The performance is measured over several environments. The algorithm works well for GSM signals. Multipath propagation error decreases using this algorithm.
5.3. Time Difference of Arrival:
It makes use of the difference of time of arrivals from different BS at MS. An algorithm is proposed in [3a] using Taylor series. The algorithm produces much better and accurate results and is robust with good convergence time. However the drawback is that some initial approximation of user position must be provided in Taylor series.
5.4. Database Correlation method:
The algorithm for database correlation technique is shown in [6a]. The algorithm explains how database can be used to estimate mobile location. It works best in heavily populated areas e.g. a big city. The method is useful for estimating location especially in NLOS conditions. The cost for this method is also low as there is only cost of creating and maintaining database.
5.5. Pilot Correlation Method:
The algorithm to implement PCM is shown in [7a]. The algorithm uses the received signal code power (RSCP) at MS. Least square method is then used to get a close match to some pre computed pilot values stored at location server. The method can be applied easily in any network. The accuracy for the method is also carried out in certain environment and is very high.
All of the above discussed algorithms have their advantages and limitations. All of them are tested under certain conditions. The choice of algorithm depends on various factors. The TDOA or PCM can be easily implemented. However we can use the hybrid techniques algorithms as in [7a, 8a]. Usually the results of the hybrid techniques are very good.
6. Conclusion:
The project aim was to study Network based mobile positioning methods and to develop online benchmark tool. Presently popular methods are studied in detail. All of the methods have their own advantages and limitations. The specifications of online benchmarking and algorithms to develop such a tool are also explored. After detailed study of literature and algorithms it is found that the network based positioning is easy to implement in present technology, sufficiently accurate, low cost to implement and does not require upgrades in handsets. Also it can be made further better to enhance its performance.
