OTDOA Location Determining Technology for Universal Intelligent Positioning System (UIPS) Implementation in Malaysia

Keeratpal Singh, Mahamod Ismail Dept. of Electrical, Electronics and System Engineering, Faculty of Engineering, Universiti Kebangsaan Malaysia, 43600 UKM Bangi, Selangor D.E., MALAYSIA Tel: +603-89216322, Fax: +603-89216146, Email: singhjatt@hotmail.com, mahamod@eng.ukm.my

Abstract- Various location determining techniques have been proposed for Location Based Services (LBS) such as Cell Identification (Cell ID), Observed Time Difference of Arrival (OTDOA), Enhanced Observed Time Difference (EOTD), Network Multipath Analysis and Assisted Global Positioning System (A-GPS). This paper presents the development work by simulating OTDOA based location determining technology (LDT) with integration to Universal Intelligent Position System (UIPS) implementation in Malaysia's Telco network. The concept of UIPS for GSM and UMTS networks offers the key players of the entire mobile services value chain in Malaysia to benefit through its integration. It is expected with the usage of UIPS, Content Providers (CP) can connect easily to Telcos and offer variety of LBS services. It is hope that with UILPS, major key players to continuously promote LBS to make it a successful consolidated effort for navigation, information search and emergency tracking. Challenges faced on different geographical terrains calls for further study on random survey to construct a trained database with noise/loss models for accuracy improvement. This is to complement any fall back contingency.

LDT Cell ID UMTS)

TABLE 1. LDT COMPARISONS

Complexity Simple-Telco use Base station info to obtain location of mobile phone attached to the nearest/strongest BS Medium-Telco may need to implement LMU in network to synchronize clocking to obtain timning Medium-Telco may need to implement LMU in network to obtain accurate time of arrival Medium -Telco may need to implement smart antenna in network that provides angle and timing info Medium- Telco need AGPS server at base stations and users have to buy new AGPS phones Medium-Require to construct database on area points, cell info related to signals multipath /delay. Minor software changes to base stations.

(GSM/

Accuracy Low accuracy -20Gm (microcell in buildings) to 15km (rural or large cell size).

E-OTD (GSM)
OTDOA (UMTS) Angle Of Arrival (UMTS)
A-GPS (GSM/ UMTS) Method

Medium accuracy- 10Gm to 500m in urban and few km in rural. Need at least 3 BS. Medium accuracy- lOOm to 500m in urban and few km in rural. Need at least 3 BS. Medium accuracy-I 0Gm to 500m in urban and up to

few km in rural.

Database
UMTS)

(GSMI

High accuracy-30m to 300m. Need Line of Sight of GPS satellites for high accuracy. Medium to High accuracy50m in urban areas.

Key words: Location Based Services, UMTS, 3G, Wireless Technology, OTDOA

I. INTRODUCTION
.

----_

In Malaysia, the first LBS called Friend Finder was launched by Maxis in 2004. The accuracy and complexity to implement each LDT within a Telco' s network is very crucial for Telco's to launch the accurate Location Based Services. Table I illustrates the accuracy and complexity for each LDT. As part of improvement to the LBS, the design of Universal Intelligent Positioning System (UIPS) offers any Location Determining Technology (LDT) such as AGPS, Cell ID, and OTDOA to be integrated under its Positioning Determining Layer in a real Telco's GSM/GPRS/EDGE, 3G or beyond network [1. Figure 1 illustrates the concept. The OTDOA LDT method will be chosen here to be simulated. OTDOA or hyperbolic method is based on the mobile measurements of the relative arrival times of pilot signal from at least 3 base

L----------

Locations Content! Applications Layer

_ __ __ _

ICP

, Layer L_ $ OTDOA

Determining

Positioninigi
,

UIPS

Future Work

Network Layer

GSM

3G

L 4L

Malaysian Telcos: Maxis (2G,3G), Digi

(2G), Celcom (2G,3G)

stations.

Figure 1. Concept of UIPS in 2G, 3G and 4G. For this research, OTDA is selected as the LDT technique

1-4244-0000-7/05/$20.00 02005 IEEE.

1057

A pair of base station defines the hyperbola and with 3 base stations, making 2 pairs, the intersection or convergence can be obtained for the location coordinates [2]. This coordinates will be translated to the grid plans in actual location that Telco's already stored in their location database. Once this final info is obtained, the UIPS will send to the CP the respective info in an agreed format. And content provider will output the feedback message to the requester while requesting UIPS to charge the Telco's billing mediation/network. From this research, observations will be made to note the total time required for a simple round trip SMS query and service feedback message to the SMS requester. It is important to take into account the various LDT to be used in future as some requires more computations than others and this translates to a lot of query time between the UTRAN (BSS in GSM) and the UE (Mobile phone). 3GPP TS 25.215 version 6.0 .0 Release 6 [3], defines UMTS Physical layer Measurements for FDD networks, which includes UE or network side carrier received signal strength indicator (RSSI), UE transmitted power, observed time difference, GPS code phase, Signal to interference ratio, round trip time, propagation delay during PRACHJPCPCH access and others. Therefore for this research, OTDOA was utilized where in Release '99 of UMTS, all local Telcos have implemented this specifications with the availability of OTDOA-IPDL (Observed Time Difference Of Arrival-Idle Period Downlink). This specification is to avoid the near-far problems that are introduced along with the UMTS network, where during the idle period; the UE mobile could receive the neighbor station's pilot signal even though the best pilot signal on same frequency is strong from another base station. Location measurement unit (LMU) will be utilized in base stations so the network (Node B towards RNC) can calculate the UE (mobile) location in relative to transmission times of pilot signals from the neighbor or adjacent stations.

Figure 2 shows the Flowchart of the testing as initiated by the UE. The user sends an SMS message requesting for a location based service pertaining to his/her location to an advertised shortcode number, for example 22222. From the Telco' s SMSC, the message is then forwarded to the respective owner of the shortcode number, which is the Content Provider. The Content Provider's Server/Engine will then communicate via VPN network /dedicated lease line with UIPS. Since the format of the UIPS has been preset by the API used between Telcos and CP, the QoS associated with it and the billing charges are sent along to the Telco upon successful feedback. UIPS will then route the request to UTRAN or BSS based on the user's telephone network details as interrogated from the HLR. Since for this example, the UE is for 3G network, the RNC will be requested for timing measurements from all adjacent Node Bs (base stations).

Once the hyperbolic timing are calculated from LMU inputs at Node Bs (3 node B are required to generate 2 pair of hyperbolic equations), the coordinates of the convergence will be calculated.
Figure 3 illustrates an example of 7 Node Bs with 5 UEs present in the selected zone. Based on the example, Node B station 6, 7 and 1 provides the intersection or convergence point. This coordinate for the particular zone is sent to the database location server where relation of actual location in longitude and latitude is sent back to the content provider.
If the service was to request for bookstores, the data sent back to the CP will have latitude and longitude information so CPs will do the building/places mappings themselves. Once the request has been sent by UIPS to CP, and successful acknowledgement flag has been received by CP and has forwarded the message back to the user ("the nearest bookstore is Menara ABC"), the charging request will also be sent to Telco's billing mediation for further charging on the user's phone number.
From experience, and adding the simulation result time, this whole process approximately takes about 40-50 seconds for each location transactions and this is considered acceptable since other infotainment based requests like mono ringtone and picture message request via SMS takes about 20 seconds to 50 seconds (during peak hours) to receive requests at times. The chunk of time is due to heavy numerical calculations done by the LMU (simulation) and sending the coordinates to the database location server.

In this research the scope is assumed to have 3 best pilot signal at all times and that all UE are within the line of sight, which in ideal conditions may not typically be available. Study is being pursued to generate a trained database topology where propagation delays i.e. path loss will be added to the timing calculations to generate better accuracy in surveyed metropolitan areas.

II. METHODOLOGY
In this research the development and performance will be discussed as part of the design with the OTDOA simulation in a 7 hexagon cell settings, assuming LOS of UEs in relations to the 7 Node Bs. Work is in progress to develop the interface from CP (Content Provider) Application Programming interface (API) towards Core Network, MSC, HLR, SGSN, GGSN, BSS, and UTRAN and its integration to the respective location determining technology (under UIPS position determining layer).

1058

3000 r

2000

1000[
0

-1000
-2000
-9" -3000

-]u

-2000

-1000

1000

2000

3000

Figure 3. 7 cells with each node B centrally located in each cell. This example shows the circle convergence of UE at 3 cells (6, 7 and 1). Each x and y axis represents coordinates in meters.

III. RESULTS FROM SIMULATING OTDOA IN UIPS

For this study, Matlab was used as the simulation environment for generating a cluster of 7 neighbor cell, where in each cell one Node B is located. The UE will be generated randomly across any of the 7 hexagon cells. 3 base station calculations are required to obtain the 2 pair of hyperbolic equations. Based on IPDL, the measurements are generated to support the nearest 3 Node Bs. With the timing information, the distance is calculated for the interested UE (3 circle convergence). This info is then passed back to UIPS to send to CP.

The aim to generate at least 15% of location based traffic at peak hours within a zone of 7 hexagon cells was implemented. 100 UEs were generated that requested for location based services. Since the location of the UEs were known, and to check whether the OTDOA algorithm worked within a tolerable line of sight performance, the simulation outputs the reversible results to re-identify the locations of each of the 100 UEs. Table 2 shows the properties of the

simulation.

Figure 2. Process Flow of UIPS for SMS request example which is generated by UE and terminates as feedback SMS location based
service to the requester/originator

Figure 4 illustrates the error distance for 100 UEs as compared to the actual position of the simulated UE. The statistical result for median is 1.6460m for distance (in meters) of actual to predicted by OTDOA simulation. This shows that the hyperbolic equation [41 is very tolerable if line of sight is in place. The maximum inaccuracy is 1l.9m between predicted and actual UE location and the minimum is 0.0193m.

1059

No. I

TABLE 2. SIMULATION PARAMETERS

Description Cell Size
7 cluster area size Numbers of Request LDT method

2 3
4

(homogeneous cell size) 5000 meters 100 UEs (Mobile phones requesting for LBS services) OTDOA based on 3 circles
Cell ID based on strongest and nearest BTS used as fall back 2 dimension flat surface with Good line of Sight
convergence

Parameters 1000 meters

Secondary LDT
Terrain Environment

The simulated result distance error was less than 12m but in average cases, the inaccuracy could be more than lOOm due to non line of sight of Node Bs or additional time delay from multipath losses. Where there are no 3 adjacent stations, especially on straight highways, it is inappropriate to use OTDOA, and other location determining technology need to be considered for this case. Of course Cell ID or signal strength based technology is the easiest one to fall back to, but the accuracy will be lower i.e. 500m difference from the actual UE distance. It is crucial for OTDOA to be complemented with Cell ID in the event there are no proper alignments of 3 Node Bs. It is also very important that the site survey data be populated at certain random locations pertaining to which LDT to use at which sites. This will then be populated at the UIPS database server.

Empincal CDF
09

08 0.7
0.6

V. CONCLUSION
In this paper, the scope of UIPS integration work was discussed. An example of SMS request LBS service by an UE user was illustrated. To perform the OTDOA LDT, a Matlab program was simulated to work hand in hand with UIPS. This program simulates 100 UE in a 7 cell cluster and verifies its accuracy to detect the corresponding 100 UEs with the parameters of UTRAN as queried based on relative time from 3 base stations. The calculation of convergence or intersection point of the 2 hyperbolic pairs denotes the location of the mobile UE. This info will be sent to UIPS location mapping, which will be pushed back to CP and finally the mobile requester. From the experiment and experience, a 40-50 secs response time is considered acceptable if this testbed were to be put on a live environment. But at times when the environment is not ideal, complement of LDT is required and adjustment to the location calculation should be made by random survey to construct a trained database with noise/loss models. Work in this area is in progress. In general, CPs Location based services are dependent on how accurate Telco's LDT is and when Telco's role out plan for CP to interface with Telco's network. Therefore it is expected with the usage of UIPS, CPs can connect easily to Telcos via the LBS API and offer variety of LBS services.

xoD 5
0-4

0.3
02
I.......................:......

0.1

Figure 4. Cumulative Distribution Function of comparing the real 100 UE locations vs. the corresponding predicted 100 UE location as obtained from OTDOA algorithm

u. 0

7....

10

12

Another simulation parameter, Secondary LDT, which is based on Cell ID' s strongest and nearest signal of a few neighbor BTS (Node B) was added to ensure the algorithm compares final iteration distance in order to obtain the least distance errors. This acts as the control limit for the simulation. On going simulations on noisy environment are in progress to improve the initial location tracking algorithm.

IV. CHALLENGES FACED AND FURTHER STUDIES

In cities, ideal cases are not always observed due to timing delay from propagation and object mutipath losses [4]. It is therefore, recommended to develop a trained database with noise/loss table for interpolating the calculated distances. This will be done during site survey of few random locations and the table will be populated accordingly to mimic the random trends. Similar studies have been conducted for Database correlation Method for GSM location [5]. Work in this area is in progress.

ACKNOWLEDGEMENT
This work is supported by IRPA grant 04-02-02-0062EA289

1060

REFERENCES
[11 Keeratpal Singh, Mahamod Ismail, "Location Based
Services for GSM and 3G in Malaysia: Perspective, Challenges and Promises", Proceedings for M2USIC 2004 Conference, Putra Jaya, Malaysia, 7fh-8h October 2004, Document No. TS4C-5, pp. 17-20 Harri Holma, Antti Toskala, "WCDMA for UMTS Radio Access for Third Generation Mobile Communications", John Wiley & Sons Ltd, Third Edition, 2004, pp. 40-44 3GPP TS 25.215, "Technical Specification Universal Mobile Telecommunications System (UMTS); Physical Layer; Measurements (FDD) (3GPP TS25.215 version 6.0.0 Release 6)", V6.0.0 (2003-12), pp. 1-19 M.A. Spirito, "Accuracy of Hyperbolic Mobile Station Location in Cellular Networks", Electronics Letters, vol. 37, No. 11, 2001, pp. 708-710 H. Laitinen, J. Lahteenmaki, T. Nordstrom, "Database correlation method for GSM location", Proc. of IEEE 53rd VTC Spring 2001, pp.2504-2508

[21

[31

[41

[51

1061