on the fingerprint-based position algorithm enhanced by

4
On the Fingerprint-based Position Algorithm Enhanced by Round Trip Time Measurement in Radio Access System Junwei Lv* 1 , Jianwu Dou 2 , and Xi Yuan 3 1 ZTE Corp. Ltd., 889#, Bibo Road, Pudong District, Shanghai, P.R. China, [email protected] 2 ZTE Corp. Ltd., 889#, Bibo Road, Pudong District, Shanghai, P.R. China, [email protected] 3 ZTE Corp. Ltd., 889#, Bibo Road, Pudong District, Shanghai, P.R. China, [email protected] Abstract This paper presents a new algorithm of Fingerprint-based position enhanced by Round Trip Time (RTT) measurement in Radio Access System. By confining the matching points to the area defined by RTT position, the location accuracy is improved. Some key influence factors of the algorithm are also discussed, and the simulation results are presented in a scenario of METIS TC2. 1. Introduction Position system becomes more and more important in wireless communication system. It can not only provide the location services for the terminal application, but also be used in radio network management and optimization, e.g. handover failure tracing and blind area detection. In 3GPP, it has proposed several positioning methods: Cell-ID/Cell- ID enhancedObserved Time Difference Of Arrival(OTDOA) and Assisted Global Positioning Systems(AGPS)[1]. AGPS is usually used to provide location-based services in outdoor scenarios. However, very low percentage ratio of GPS is on available state when Radio Access Network initializing AGPS measurement due to privacy issues, so the location information is not enough for radio network optimization. RTT-based position method, which finds the UE location via trilateration, meets the same situation because of lacking enough Radio links as well as poor performance in Non-Light-of-sight (NLOS) scenarios, e.g. in urban blocks. Finger-Print-based (FP) Position algorithm is a popular location technique and robust in NLOS scenarios [2]. It establishes a database which stores measurements of network characteristics from different locations, and estimates User Equipment (UE)’s location by inspecting currently measured network characteristics. Typically, network characteristics are Base Station (BS) identifiers and the received signal strength. With a large fluctuation in the received signal strength at the same location influenced by moving obstacles or the weather, fingerprint-based position algorithm would degrade performance in outdoor scenarios. Therefore, we propose a new fingerprint-based position method enhanced by RTT measurement. A regional range of location is calculated by RTT measurement, and then the specific location is determined by FP position. The FP position margin of distance error could be limited by RTT measurement. The paper is organized as follows. In Section 2, the fingerprint-based position method enhanced by RTT measurement is introduced. A simulation process of this method is presented in Section 3. It is found that accuracy of FP can be improved by RTT position and several influence factors are discussed. Conclusions are drawn in Section 4. 2. The fingerprint-based position algorithm enhanced by RTT There are two phases in FP position. One is offline phase, when a radio map storing locations and corresponding RSCP vectors are established. The other is online phase, when location of UE is given by comparing the reported current RSCP vector with samples in the radio map. RTT position is implemented in the online phase. After determination of a Region Of Interest (ROI) that includes relevant samples, such as RSS-based ROI [3], ROI would be refined by RTT position. Samples in ROI which is out of the region defined by RTT position should be filtered out. The way to determine the matching points region with RTT position is described now. In a radio communication network, RTT information can be acquired when UE is in connected mode. RTT can be used to calculate the distance d between UE and BS antenna when a Line Of Sight (LOS) path exists. The estimate position of UE is a circle ring belt with antenna’s location as center and away from the center with the interval of [d-, d +], while is a distance deviation due to the measurement accuracy of RTT, which is less than a chip in UMTS. So is approximately 78m. In fact, when an obstacle is located between UE and BS antenna, the distance d may be 978-1-4673-5225-3/14/$31.00 ©2014 IEEE

Upload: others

Post on 13-Nov-2021

6 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: On the Fingerprint-based Position Algorithm Enhanced by

On the Fingerprint-based Position Algorithm Enhanced by Round Trip Time Measurement in Radio Access System

Junwei Lv*1, Jianwu Dou2, and Xi Yuan3

1ZTE Corp. Ltd., 889#, Bibo Road, Pudong District, Shanghai, P.R. China, [email protected]

2ZTE Corp. Ltd., 889#, Bibo Road, Pudong District, Shanghai, P.R. China, [email protected]

3ZTE Corp. Ltd., 889#, Bibo Road, Pudong District, Shanghai, P.R. China, [email protected]

Abstract

This paper presents a new algorithm of Fingerprint-based position enhanced by Round Trip Time (RTT) measurement in Radio Access System. By confining the matching points to the area defined by RTT position, the location accuracy is improved. Some key influence factors of the algorithm are also discussed, and the simulation results are presented in a scenario of METIS TC2.

1. Introduction Position system becomes more and more important in wireless communication system. It can not only provide the location services for the terminal application, but also be used in radio network management and optimization, e.g. handover failure tracing and blind area detection. In 3GPP, it has proposed several positioning methods: Cell-ID/Cell-ID enhanced、Observed Time Difference Of Arrival(OTDOA) and Assisted Global Positioning Systems(AGPS)[1]. AGPS is usually used to provide location-based services in outdoor scenarios. However, very low percentage ratio of GPS is on available state when Radio Access Network initializing AGPS measurement due to privacy issues, so the location information is not enough for radio network optimization. RTT-based position method, which finds the UE location via trilateration, meets the same situation because of lacking enough Radio links as well as poor performance in Non-Light-of-sight (NLOS) scenarios, e.g. in urban blocks. Finger-Print-based (FP) Position algorithm is a popular location technique and robust in NLOS scenarios [2]. It establishes a database which stores measurements of network characteristics from different locations, and estimates User Equipment (UE)’s location by inspecting currently measured network characteristics. Typically, network characteristics are Base Station (BS) identifiers and the received signal strength. With a large fluctuation in the received signal strength at the same location influenced by moving obstacles or the weather, fingerprint-based position algorithm would degrade performance in outdoor scenarios. Therefore, we propose a new fingerprint-based position method enhanced by RTT measurement. A regional range of location is calculated by RTT measurement, and then the specific location is determined by FP position. The FP position margin of distance error could be limited by RTT measurement.

The paper is organized as follows. In Section 2, the fingerprint-based position method enhanced by RTT measurement is introduced. A simulation process of this method is presented in Section 3. It is found that accuracy of FP can be improved by RTT position and several influence factors are discussed. Conclusions are drawn in Section 4.

2. The fingerprint-based position algorithm enhanced by RTT

There are two phases in FP position. One is offline phase, when a radio map storing locations and corresponding RSCP vectors are established. The other is online phase, when location of UE is given by comparing the reported current RSCP vector with samples in the radio map. RTT position is implemented in the online phase. After determination of a Region Of Interest (ROI) that includes relevant samples, such as RSS-based ROI [3], ROI would be refined by RTT position. Samples in ROI which is out of the region defined by RTT position should be filtered out. The way to determine the matching points region with RTT position is described now. In a radio communication network, RTT information can be acquired when UE is in connected mode. RTT can be used to calculate the distance d between UE and BS antenna when a Line Of Sight (LOS) path exists. The estimate position of UE is a circle ring belt with antenna’s location as center and away from the center with the interval of [d-∆, d +∆], while ∆ is a distance deviation due to the measurement accuracy of RTT, which is less than a chip in UMTS. So ∆ is approximately 78m. In fact, when an obstacle is located between UE and BS antenna, the distance d may be

978-1-4673-5225-3/14/$31.00 ©2014 IEEE

Page 2: On the Fingerprint-based Position Algorithm Enhanced by

estimated based on a NLOS path which is much longer than the real one. Therefore, (d-∆) may be meaningless, which may be longer than the real distance between UE and BS antenna. However, (d+∆) is always not less than the LOS distance between UE and BS antenna, so it can be used as the outer boundary of the matching points region. Moreover, UEs located in the handover area, can acquire two or more RTT measurements from different BS antennas. In this case, multiple RTT measurements can be used to determine the matching points region more accurately. The matching points region defined within the outer boundary (d+∆) is showed in Figure 1.

Figure 1 The defined matching points region by RTT information

3. Location Estimation Via FP and RTT in UMTS

3.1 Simulation Scenario Descriptions

The simulation scenario is provided on METIS website [4] as “Ray Tracing files for TC2 (macro cells)”, which is an urban street scenario with 27 Macro cells as Figure 2.a. To simplify the simulation, we use deterministic technique which applies deterministic inference to estimate a UE’s location. In the offline phase, several radio maps, which store the large scale fading value of samples, are generated by using data of TC2 (macro cells) with different sampling steps. In the online phase, a new large scale fading value samples extracted from TC2 (macro cells) and disturbed by small scale channel model TU3 [5], are compared against the radio map with calculated Euclidean distances[6], and given estimated locations by Weighted K-Nearest Neighbor (WKNN) algorithm. In order to make the test points different from the sample points in radio map, their positions are interpolated as figure 2.b.

Figure2.a The simulation scenario

Figure 2.b The distribution of samples and test points

Suppose that N is the number of BS antennas for positioning, RSCPi is Received Signal Chip Power (RSCP) value of test point from BS antenna i, and is the value of samples in radio map from antenna i. The Euclidean distance (ED) is (1):

2

1( )

N

i ii

ED RSCP RSCP=

= −∑ (1)

Suppose EDj is the Euclidean distance between test point and sample j, keep K samples of shortest Euclidean distance to estimate the location of test point by WKNN. Let (xj, yj) denote the position coordinates of samples and wj is a weighting factor which is inversely proportional to EDj. Then the position coordinates (X, Y) of test points are calculated as (2):

iRSCP

Page 3: On the Fingerprint-based Position Algorithm Enhanced by

1

1

1 1( , ) ( , )0.00001

K

j j j jKj j

jj

X Y w x y wEDw =

=

= • =+∑

∑ (2)

Based on ZTE simulation platform, we implement the FP Algorithm enhanced by RTT Measurement (ERTT). RTT distance d is calculated with the ray tracing method based on the following constraints: 1) The LOS path which penetrates 5 walls is considered unreachable. 2) d is the shortest one of the paths whose receiving signal power level is above receiver sensitivity. In this way, d does not need to be corrected by ∆. The rules of RTT distance selection are as follows: 1) Take the BS antenna with the strongest RSCP value as the serving cell; 2) When the RSCP difference between the serving cell and the neighboring cell is less than 3dB, UE should be considered in the handover region,and multiple RTT values can be acquired . The samples in the radio map which are out of the region defined by RTT positioning will be filtered. The simulation parameters are set as Table 1:

Table 1 The simulation parameters Parameter type value Parameter type value

receiver sensitivity -120dBm Sampling steps for radio maps 5m, 10m, 15m, 20m K value of WKNN 3 Sampling step for test points 10m

Antennas’ number for positioning (Ante_num) 2,3,4,5 Number limit of RTT in

handover region(RTT_num) 2

RTT measurement accuracy(∆) 0m

3.2 Performance of Deterministic FP Algorithm Enhanced by RTT Measurement Comparing ERTT with the method without RTT positioning (Non-RTT), we can find that ERTT enhance the accuracy at 90th percentile obviously in Table 1. In Figure 3, the performance comparison of ERTT vs Non-RTT is showed, with the sampling step of 5m. Figure 3.a shows that the performance of ERTT is related to the difference between d and LOS (denoted by (d - LOS)). (Non-RTT - ERTT) denotes that error distance of Non-RTT subtracts one of ERTT. When (d - LOS) is less than 150m, ERTT is active. As (d - LOS) increases, ERTT return to non-RTT. It appears that ERTT is worse than non-RTT sometimes. That is because some optimum matching nodes are filtered by ERRT. When (d - LOS) is small, the accuracy can be improved, meanwhile, errors are introduced, since some related nodes near the test points are more likely to be filtered out in the certain scenes (e.g., Figure 4.a). When the (d - LOS) is large, the reason for the error introduced is that RSCP feature of test points is very different from one of their neighbors. After the optimum matching points which calculated by non-RTT filtered by ERTT, the worse points are selected. Figure 3.b describes four cases with different values of Ante_num and RTT_num set as Table 3. To reduce errors, a feasible method is that keep the related sample points as much as possible when (d - LOS) is small, for example, adding a protection distance △d. Furthermore, considering that streets in urban area are almost straight, it may be more efficient to use a square filtering area instead of a circular one, such as Figure 4.a, since computing a square filtering area is much easier. Figure 4.b shows the results of ERTT using a square filtering area with △d=0.

Table 2 The 90th, 75th, and 50th percentile values of the error distance Sampling step Of radio maps

90th(meter) 75th(meter) 50th(meter) Non-RTT ERTT Non-RTT ERTT Non-RTT ERTT

5m 110.15 83.61 47.28 41.04 19.15 18.60 10m 108.55 81.08 46.24 40.66 19.19 17.99 15m 110.98 83.78 46.97 42.47 20.74 19.49 20m 138.56 103.33 53.62 47.59 24.54 23.28

4. Conclusions

In this paper, we have studied the performance of fingerprint-based position algorithm enhanced by RTT position, and accuracy would be better if one or more of the following conditions are satisfied: 1) multi-LOS paths in

Page 4: On the Fingerprint-based Position Algorithm Enhanced by

the scene. 2) RTT measurements to several geographically dispersed NodeBs are available. 3) The region define by RTT position could keep the closely relevant sample points as much as possible.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

200

400

600

800

probability(%)

erro

r dis

tanc

e(m

)

0 100 200 300 400 500 600 700 800-500

0

500

1000error distance difference distribution

d - LOS (m)

nonR

TT -

ERTT

ERTT

nonRTT

worse

equalbetter

Figure 3.a Accuracy of ERTT Vs Non-RTT

0 0.5 10

200

400

600case 1

erro

r dis

tanc

e(m

)

0 0.5 10

200

400

600

800case 2

erro

r dis

tanc

e(m

)

0 0.5 10

100

200

300case 3

probability

erro

r dis

tanc

e(m

)

0 0.5 10

200

400

600

800case 4

probability

erro

r dis

tanc

e(m

)

ERTT

nonRTT

Figure 3.b Cases with different Ante_num and

RTT_num

Table 3 The parameters of Figure 3.b Case 1 Case 2 Case 3 Case 4

Ante_num 2 2 >2 >2 RTT_num 1 2 1 2

Figure 4.a Filtering region defined by RTT

0 0.5 10

200

400

600case 1

erro

r dis

tanc

e(m

)

0 0.5 10

200

400

600

800case 2

erro

r dis

tanc

e(m

)

0 0.5 10

100

200

300case 3

probability

erro

r dis

tanc

e(m

)

0 0.5 10

200

400

600

800case 4

probability

erro

r dis

tanc

e(m

)

ERTT

nonRTT

Figure 4.b Results with square filtering areas

5. References

1. 3GPP, TS 25.305 V12.0.0: Stage 2 functional specification of User Equipment (UE) positioning in UTRAN(Release 12), 2013 2. M.B.Kjærgaard, A taxonomy for radio location fingerprinting, Lecture Notes in Computer Science, 2007, pp.139-156. 3. Azadeh Kushki, Konstantinos N.Plataniotis, Anastasios N.Venetsanopoulos, WLAN positioning systems: principles and applications in location-based services, Cambridge University Press, 2012, pp. 118-119. 4. Https://www.metis2020.com/documents/simulations/. 5. 3GPP, TR 25.943 V11.0.0: Deployment aspects (Release 11), 2012. 6. P. Bahl, V.N. Padmanabhan, RADAR: An In-Building RFbased User Location and Tracking System, Proceedings of IEEE Infocom 2000, Tel Aviv, Israel, March 2000.