Huawei Technologies Co., Ltd, Beijing, P. R. China. Email: {liujinnan, shulan.feng }@huawei.com. AbstractâWe studies the influence of multiple reference ...
RSTD Performance for Small Bandwidth of OTDOA Positioning in 3GPP LTE Jinnan Liu , Shulan Feng Research Department of Hisilicon, Huawei Technologies Co., Ltd, Beijing, P. R. China Email: {liujinnan, shulan.feng }@huawei.com Abstract—We studies the influence of multiple reference signals on different transmit ports for RSTD measurements in small bandwidth scenarios. Because of scalable measurement bandwidth in long term evolution (LTE), the performance of reference signal time difference (RSTD) suffers from small bandwidth. Conventional RSTD scheme is based on positioning reference signals (PRS) which use one transmit antenna and two receive antennas (1x2). In order to reduce the loss of accuracy caused by small bandwidth, three new schemes are proposed. Compared with the conventional scheme, the accuracy of three proposed schemes has been improved in ETU channel. We verify it with our simulation results. The detection probability of one proposal 2x2 scheme which splits PRS symbols on two transmit antennas is decreased by half the number of accumulating symbols on each transmit antenna. Other two schemes which accumulate cell-specific reference signals (CRS) and PRS noncoherently or coherently not only provide better RSTD accuracies but also improve detection probabilities. Index Terms—LTE; RSTD; PRS; CRS; Small bandwidth.
I. INTRODUCTION Since the GPS receivers and Wi-Fi modules are integrated into smart phones, satellite and Wi-Fi positioning methods can provide mobile location as cellular systems do[1]. The cellular location systems are the cost-effective utilization of infrastructures and complementary positioning methods for the required level of accuracy in all environments, especially in indoor and urban canyon environments. Since cellular systems are not originally designed for positioning, the observed time difference of arrival technology (OTDOA)[2] is designed to reuse current reference signals and minimize the influence of current standards. In long term evolution (LTE), OTDOA method is supported [3][4] and based on measurements of downlink reference signals. Some candidates of existing downlink reference signals are not suitable for the hear-ability and accuracy. Primary synchronization signals (PSS) and secondary synchronization signals (SSS) are discussed which are used for synchronization and physical cell ID detection in release 8 (R8). However, the positioning performance of PSS and SSS is not good enough for E911 requirements for small accumulating time and frequency resources [5]. And cellspecific reference signals (CRS) are also considered. While CRS do not satisfy the hear-ability that at least three base
states (BS) signal should be detected by user equipments (UE) [7]. To enable positioning in LTE and facilitate positioning measurement with an acceptable quality, new physical reference signals named positioning reference signals (PRS), have been introduced in release 9 (R9). Furthermore, to enhance the hear-ability from multiple sites, the positioning subframes are designed as a low interference pattern which means no data transmission in physical downlink share channel (PDSCH). In this way, PRS shall be only interfered by the reference signals from other cells in the same time and frequency resources. LTE provides a scalable system bandwidth and enables flexible network deployments from 1.4MHz to 20MHz. PRS bandwidth can be less than or equal to system bandwidth. By non-coherent accumulation of different positioning subframes and coherent accumulation of different PRS symbols within one positioning subframe, the performance of reference signal time difference (RSTD) measurements can be achieved in large PRS bandwidth, such as 20MHz [6]. Unfortunately, even for accumulation of maximum number of PRS subframes, the performance of the small PRS bandwidth is poor with the simulation assumption of standard [8]. In this paper, we discuss the gain from multiple transmit antennas and multiple reference signals. The UE can estimate RSTD from non-coherent summation of different receive antennas and other reference signals from different transmit antennas. In LTE R9, 1x2 (1 transmit antenna and 2 receive antennas) are used for RSTD measurement to get the receive diversity gain. With a limited UE capacity, we do not discuss to increase the number of receive antennas and number of PRS subframes in this paper. Three new schemes are proposed to take advantage of coherent or non-coherent accumulation gain. The remainder of this paper is organized as follows. Section II describes the OTDOA design in LTE. Section III illustrates the flow chart of the RSTD measurement which is implemented by UE for OTDOA positioning. In the following two sections, we explain different transmit methods and show their performance respectively. In Section VI, we give a brief conclusion. Throughout this paper, Ts is expressed as a number of time seconds, · the units, Ts 1/ 15000 2048
conjugate operation, | · | the modular squaring.
978-1-4673-6187-3/13/$31.00 ©2013 IEEE
III. RSTD MEASUREMENT
II. OTDOA IN LTE For LTE, OTDOA positioning impacts on both positioning protocol and physical layer. PRS configuration of neighbour cells will be informed to UE by assistant data messages. And BSs of reference cell and neighbour cells will transmit PRS according to PRS subframe structure in each positioning occasion (PO) periodically. A. LPP LTE positioning protocol (LPP) [3] is used between a location server and a target device in order to positioning the target device by position-related measurements, such as RSTD. For OTDOA positioning, OTDOA assistant data messages are sent by the location server to inform UE the PRS parameters of reference cell/multiple neighbour cells. According to OTDOA assistant data, UE can generate the local PRS and estimate the RSTD by cross-correlation of received signals and local PRS. The location server can estimate UE location by hyperbolic positioning [2]. B. Physical layer Concepts B1. Frame structure In time-domain structure for LTE transmission, each radio frame is 10ms and consists of 10 equally sized subframes. A subframe is defined as two consecutive slots where subframe i consists of slots 2i and 2i+1. And each slot has N DL orthogonal frequency division modulation (OFDM) symbols. The resource unit for scheduling is two resource blocks (RB), defined as N DL OFDM symbols in time domain and 12 subcarriers in frequency domain. Each element in the RB is called a resource element (RE).
,
is defined by the sequence of reference signals from u BS on pt transmit antenna in frequency domain, and R , is , transmission which the OFDM symbol indice set of , consists of , OFDM symbols in Fig. 3. is generated by UE as local reference signals (RS) which are transmitted on pt transmit antenna in i PRS subframe from u BS as following: s
x
,
x
,
,
,
,
, xNDL
cp ifft R
n
,x
,
,
,
,
, xNDL
, ifft R , , 0, ,0 ,
(1)
,
other
0 n N NCP 1 where ifft · and cp · for modulating OFDM and adding , cyclic prefix (CP) respectively. x n is the time samples of NCP samples, where an OFDM symbol which has N is the size of ifft · for a OFDM symbol and NCP is the N length of CP.
Fig. 1 RSTD measurement
B2.CRS CRSs are transmitted on port0 ~port3. According to the number of physical broadcast channel (PBCH) transmit antenna, CRS can be transmitted on 1,2or 4 antenna ports. UE can also generate local CRS of reference cell/neighbour cells according to OTDOA assistant data message ,which is similar as the local PRS generation.
UE obtains overlapping signals y n from different cells on pr receive antenna within a search window W and measures the relative timing difference of reference signals between a neighbour cell and the reference cells in Fig. 1. The module of power delay profile (PDP) estimation can be expressed as (2) and illustrated in Fig. 2. The module of RS generator will generate RS of cell u by multiple steps In Fig.2. It will firstly
B3.PRS and PRS subframe PRSs are transmitted on one antenna port, i.e. port6. Since CRS enable UE to estimate channel and track synchronization in R8, the transmission of PRS should avoid the OFDM symbols occupied by CRS. Also, the number of OFDM symbols used for PRS transmission will be constrained by the OFDM symbols occupied by CRS in one subframe. Furthermore, to reduce inter-cell interference and improve the hear-ability, low interference subframes, defined as PRS subframes[4], have been used. In a PRS subframe, no data transmission is allowed except for common information, such as PBCH, PSS, SSS regardless of their antenna port p. A PO can be configured by N PRS consecutive downlink
generate R in frequency domain and then map into , zero symbols. Finally, , RS symbols and symbols will consist of a subframe according to R , . To estimate time delay, cross-correlation results of different transmission antennas, receiver antennas and subframes can be non-coherently accumulated. While the correlation results of multiple OFDM symbols in one subframe on the same transmit antenna port are accumulated coherently in (2)
PRS subframes and 12NRB central consecutive According to the physical layer design [3], NPRS PRS NRB 6,15,25,50,75,100.
subcarriers. 1,2,4,6 and
u , pt
Pu (τ ) =
1 N PRS mu
nu
mu N PRS
∑∑ ∑ N pr =1 pt =1 i =1
1 u , pt
y
pr i
(n + τ ) ⋅ s (n) u , pt i
*
0 τ 1 Then, the estimated time delay can be express as
τˆu , fstP = arg min{Pu (τ ) ≥ au max(Pu (τ ))} τ
(3)
2
(2)
if SNR( Pu (τ )) ≥ Th u , n , a successive detection
much as possible false alarm with a sufficient level of detection.
influence of different reference signals may get the noncoherent gain with different antenna ports or coherent gain with the same antenna ports for LTE. It is well known that the SNR of coherent summation is increased by the coherent numbers of complex symbols, i.e.
au is a first path scaling coefficient for multiple path fading
N u , pt . And the non-coherent accumulation will compensate
else , a failure detection Where Thu,n must be above the noise level in order to limit as
the rotation from the multipath fading channel and the Doppler frequency offset.
channel environments.
B. SNR of The SNR of P τ is estimated by P τ (2) in the range τ , N ,τ , N and τˆ u, mP = arg max {Pu (τ )} .In this τ
range, we sum a given number of the largest paths as signals and sum the rest as noise to get the SNR. IV. PROPOSED SCHEMES The PRSs are introduced for positioning in LTE release 9. In general assumption, the PRSs are transmitted on one antenna and UE has two receive antennas. So the performance of RSTD measurement is evaluated by 1x2 PRS scheme as the baseline in Case A. The mapping of PRS sequence in one PRB is shown in Fig.3.
Case A: Baseline In this case, PRSs are transmitted on one antenna, noted as port 6. And the PRSs are designed for diagonal shift in one subframe to get good auto correlation function (ACF)
Fig. 2 PDP estimation
A. Crámer–Rao low bound (CRLB) The CRLB of delay estimation is given in [9]
u ,6
(4)
performance. In frequency domain, the PRS sequence R is a kind of golden sequence which is initiated by the Cell id, slot number and symbol number [4].
Where F is the mean square bandwidth (MSB) of the reference signals,
OFDM And at the receiver, the correlation of N symbols in one PRS subframe can be accumulated coherently and N PRS PRS subframes can be accumulated non-coherently.
var (τ ) ≥ CRLB (τ ) =
1
γF
2
u ,6
2
∞
F
2
=
∫ (2 π f ) s ( f ) ∫ s ( f ) df 2
∞
∞
2
df
(5)
2
For normal CP, N
u ,6
is 8 and
⎫ ⎧ ⎪ ⎧ ⎫ ⎪ ⎨ l u , 6 ⎬ = ⎨ 3, 5, 6 ,1, 2 ,3, 5, 6 ⎬
⎪ ⎩ R ⎭ ⎪ ⎩ slot 2 i slot 2 i +1 ⎭
when
PBCH antenna ports are one or two.
∞
≈ βB 2
β
is 4π / 3 when the power spectral density (PSD) of the 2
OFDM signals is rectangular and
β is
4.4 when the PSD of
transmitted signal is sinc squared function within B [10]. Also
2
γ is the SNR of received signal, h is the mean
square power of estimated first path for multiple path channel model. If in AWGN channel,
2
h equals 1. 2
γ =
h Es
(6)
N0 / 2
From (4), the larger γ or B , the lower the CRLB. In MIMO system, different transmit and receive antennas will be summed non-coherently. So the influence of receive antennas is similar as that of non-coherent accumulated subframes. The
l=0
l=6 l=0
l=6
Fig. 3 Mapping of PRS (normal cyclic prefix)
Case B: Split PRS into 2 ports In case B, PRSs are transmitted on two antennas. We u ,6
separate N OFDM symbols into 2 transmit antennas, i.e. ＇ port 6 and 6".
For normal CP, each transmit antenna sends 4 OFDM symbols when PBCH antenna ports are one or two. And the sets of OFDM symbols indices in one subframe on two ⎫ ⎧ ⎫ ⎪ ⎧ ⎪ , 5, 6, 1N ⎬ ⎨l ' ⎬ = ⎨3
u , 6 ⎩R ⎭ ⎪ ⎪ slot 12 i + 1 ⎭ ⎩ slot 2 i
transmit antenna are
and
⎫ ⎧ ⎫ ⎪ ⎧ ⎪ , 3, 5, 6 ⎬ ⎨l '' ⎬ = ⎨ 2
u , 6 ⎭ ⎪ ⎩R ⎪ ⎩ slot 2 i+1 ⎭
individually.
Case C: Combine PRS with CRS Since the CRSs are kept in PRS subframes shown in Fig.3. And also the symbols of CRSs are low interferences in PDSCH. So the UE can accumulate correlation results of CRS and PRS, which are transmitted on port 0~3 and port 6 respectively. Both CRS and PRS sequences are a kind of gold sequence initialized with slot index, physical cell id, cyclic prefix type, and symbol index which can get from OTDOA assistant data message for local PRS or CRS generation as in Fig. 2. To simplify, receiver combines only CRS
R u ,0 on port 0
u ,6
with PRS R on port 6 in case C. Because the port 0 is always used regardless of number of antenna is 1,2or 4. Furthermore, CRSs in PDSCH are adopted for low interference consideration instead of CRSs in control channels at #0 symbol and
even
slots.
For
⎫ ⎧ ⎪ ⎧ ⎫ ⎪ ⎨l u , 0 ⎬ = ⎨ 4N , 0N , 4 ⎬ when ⎩R ⎭ ⎪ ⎩ slot 2 i slot 2 i +1⎪⎭
And
PRS
normal
N u , 0 is 3 and
PBCH antenna ports are one or two.
configuration
⎫ ⎧ ⎪ ⎧ ⎫ ⎪ ⎨ l u , 6 ⎬ = ⎨ 3, 5, 6 ,1, 2 ,3, 5, 6 ⎬
⎪ ⎩ R ⎭ ⎪ ⎩ slot 2 i slot 2 i +1 ⎭
CP,
is
same
as
Case
A
and
.
Case D:Share CRS port with PRS In case D, we assume the antenna physical configuration of port 6 is as same as that of port 0 and the physical configuration is informed to UE explicitly or implicitly. So the receiver can accumulate OFDM symbols on the common port of CRS and PRS in one PRS subframe coherently. Similarly as case C, only CRSs on port 0 in PDSCH are used. For normal CP, N port 0 port 6 ⎧ ⎫⎪ ⎧⎪ ⎫⎪ ⎧ ⎫ ⎪ , 4 ⎬ ∪ ⎨ 3, 5, 6 ,1, 2, 3, 5, 6 ⎬ ⎨ l u , 0+6 ⎬ = ⎨ 4N , 0N
⎩R ⎭ ⎪ ⎩ slot 2 i slot 2 i +1 ⎪⎭ ⎪⎩ slot 2 i slot 2 i +1 ⎪⎭
u ,0+ 6
TABLE I. SIMULATION PARAMETERS Parameter Cell ID combinations Network synchronization Duplex modes Cyclic prefix Carrier frequency UE speed Carrier bandwidth Channel model Ês/Noc for three cells, [dB] PRS Number of transmit CRS antennas Power ratio Number of receive antennas Positioning subframes Number of consecutive positioning subframes PRS pattern PRS transmission bandwidth Measurement bandwidth
Value Synchronous with time shifts FDD Normal 2 GHz 3 km/h 1.4 /5MHz AWGN/ETU , 1 1 1 (PRS and CRS are transmitted equal power) 2 equal-gain uncorrelated antennas Low Interference Scheme (no presence of PDSCH in PRBs containing PRS), full or partial alignment 6/4 non-coherent accumulation 6-reuse in frequency, vshift = mod(PCI,6) Full carrier bandwidth Full carrier bandwidth
A. AWGN channel To simplify problem, the performance in AWGN channel is studied and shown Fig.4 and Fig.5. The parameters of those simulations, such as Thu ,n and au , are shown in Table II. Absent of the phase rotation from the multiple path channel, the performance is determined by the number of coherent accumulating symbols in AWGN channel. There are only 4 OFDM symbols in case B (red curve) and 11 OFDM symbols in case D (blue curve) in one subframe on one antenna in Fig. 4. TABLE II. THRESHOLDS FOR VARIOUS SCENARIOS PRS bandwidth
N PRS
1.4MHz
6
5MHz
4
Transmit Schemes
Thu ,n (dB)
au (dB)
Case A/D Case B/C Case A/D Case B/C
4.5 3.3887 4.0 3.0
-1.5 -4
is 11 and
when PBCH antenna ports
are one or two. V. PERFORMANCE In this part, we study the AWGN channel and ETU channel. The performance of proposed schemes is evaluated by cumulative distribution function (CDF) of RSTD error and detection probability over 1000 Monte Carlo simulations in
each scenario, where the occasions when a cell is not detected are excluded from the CDF. And the simulation is set up by 3 cells with the normal CP and the search window is two times of length of CP. The details are given in Table I. In this part, Ts is defined as a number of time units and about 9.8m according to the speed of light.
Fig. 4 RSTD Performance for AWGN channel at 1.4MHz bandwidth
transmit diversity gain. And case C and case D have similar CDF but the detection probability of case D is higher for coherent accumulation gain. VI. CONCLUSION AND FUTURE WORK
Fig. 5 RSTD Performance for AWGN channel at 5MHz bandwidth TABLE III. Pd 1.4MHz 5MHz
Case A 100% 100%
Detection Probability of AWGN Case B 99.8% 100%
Case C 100% 100%
Case D 100% 100%
Case D achieves gain to increase the number of coherent OFDM symbols in one subframe. From Fig.5, the proposal case D scheme has margin impacts on positioning performance for 5MHz PRS bandwidth. And case C have no gain and case B have a loss for no increasing and decreasing number of coherent OFDM symbols in one subframe respectively in AWGN channel.
B. ETU channel Used by a linear interpolation, we estimation first path delay from PDP for multiple path channel environments. From the conclusion of AWGN channel, we only show the simulations of 1.4MHz bandwidth for multipath fading channel. TABLE IV. DETECTION PROBABILITY OF ETU 1.4MHz Pd RSTD error[Ts] at 90%
Case A 76% 12
Case B 55% 5.8
Case C 80% 7.3
Case D 86% 8.9
Fig. 6 RSTD Performance for ETU channel at 1.4MHz bandwidth
Unlike AWGN channel, non-coherent summation compensates the rotation from the multipath fading channel. In multipath fading channel, case B has the best accuracy for
Multiple transmit antenna scheme of PRS in case B will decrease the detection of probability. Because the total number of symbol is kept, the split scheme (case B) will decrease the number of symbol on each antenna for coherent gain. In case C, PRS and CRS are combined non-coherently as independent transmit antennas. And in case D, PRS and CRS are configured to transmit on the same physical port. Both schemes with coherent and non-coherent accumulation between PRS and CRS achieve better performances than the traditional method of case A does in ETU channel. With the increasing of PRS bandwidth, the coherent or non-coherent gain from CRS is decreasing, since the length of coherent or non-coherent accumulation is large enough. We only consider small PRS bandwidth case in this paper. We do further work about how to decrease the system overhead and computation complexity of large PRS bandwidth for low cost and power efficiency [11]. REFERENCES [1] Denis. Huber,"Background Positioning for Mobile devicesAndroid vs. iphone", Joint Conference of IEEE computer & communication societies, 2011 [2] Chan, Y. T., and K. C. Ho. "A simple and efficient estimator for hyperbolic location." Signal Processing, IEEE Transactions on 42.8 (1994): 1905-1915. [3] 3GPP TS 36.355 "Evolved Universal Terrestrial Radio Access (EUTRA); LTE Positioning Protocol (LPP) ", version 9.4.0, 122010 [4] 3GPP TS 36.211 "Evolved Universal Terrestrial Radio Access (E-UTRA); Physical channels and modulation", version 9.1.0, 03-2010. [5] C. Mensing, S. Sand, A. Dammann, and W. Utschick "Interference-Aware Location Estimation in Cellular OFDM Communications Systems", IEEE International Conference on Communications (ICC), pp. 2213-2217,June 2009. [6] J. Medbo, I. Siomina, A. Kangas and J. Furuskog "Propagation channel impact on LTE positioning accuracy: A study based on real measurements of observed time difference of arrival", in Proc. IEEE 20th Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), pp. 2213-2217, 2009 [7] A. Oborina1, T. Henttonen and V. Koivunen "Cell Hear-ability Analysis in UTRAN Long Term Evolution Downlink", Proceedings of the 43rd Asilomar Conference on Signals, Systems and Computers, California, USA, 2009, pp. 991-995. [8] Huawei, HiSilicon "R4-121268: Remaining Issues of RSTD Measurement Accuracy Requirements", 2012 [9] S. Kay, "Fundamentals of Statistical Signal Processing: Estimation Theory". Prentice-Hall PTR, pp 54–55,1993 [10] R. Martin, "Comments on “OFDM Transmission forTime-Based Range Estimation” "IEEE Signal Processing Letters, Feb. 2011 [11] J. Liu, S. Feng, "RSTD Performance for Scalable Bandwidth of OTDOA Positioning in 3GPP LTE" ICL-GNSS, June 2013
