Patent Application
20160381592
This application claims priority under 35 U.S.C. §119 or 365 to European Application No. 15306019.9, filed Jun. 26, 2015. The entire teachings of the above application are incorporated herein by reference.
This invention relates to ambiguity resolution in positioning measurements. It is particularly suitable for frequency domain implementation of OTDOA measurements in an LTE network.
The Observed Time Difference of Arrival (OTDOA) feature was introduced in the LTE R9 standard as a user equipment (UE)-assisted positioning method. It comprises estimating the location of a UE by using the positioning reference signal arrival times that the UE measures from a number of surrounding eNodeBs.
A UE is provided with a list of cells by the location server of a cellular network to measure the relative arrival time with respect to a pre-configured cell (most often the serving cell). The relative arrival time is referred to as reference signal time difference (RSTD), and the pre-configured cell is referred to as the reference cell. Each cell in the list is associated with a priori information on its RSTD. This information is composed of an expected RSTD, which is estimated by the network, and an uncertainty window centered around the expected RSTD value which helps the UE to localize its search in the uncertainty window only.
The network, or precisely the location server, will use the set of RSTDs measured by the UE to obtain the final estimate of the UE position. Therefore, the quality of this positioning method depends closely on the accuracy of the RSTD measurement performed by the UE.
The UE measures RSTD using the Positioning Reference Signal (PRS) that is transmitted on antenna port (AP) 6 of a measured cell. As per standard specification, the size of the uncertainty window can be 2×3069 Ts and the expected RSTD can be 3×8192 Ts (approximately 11 OFDM symbols). Therefore, it is common for the PRS subframe of a measured cell not to be aligned in time with that of the reference cell. Common reasons for such a misalignment include high distance between the serving and measured eNodeB, or asynchronous operation.
Since frequency-domain subframe-based processing is the reference baseband processing used in LTE, a frequency-domain implementation should be less complex than a time-domain approach for RSTD estimation by the UE.
Typically, the underlying baseband processing is synchronized with the serving cell. Keeping the same data path, the frequency-domain data is only available per useful part of an OFDM symbol (i.e. without the cyclic prefix) aligned in time with the serving cell. Thus, a frequency-domain implementation of OTDOA measurement typically includes the following parts:
1. Determine the reference OFDM symbol that corresponds best to the positioning symbol of a measured cell by using an estimation technique,
2. Descramble the selected reference OFDM symbol with the scrambling sequence that is used for PRS of the measured cell,
3. Estimate the PRS arrival time of the measured cell, for example by using an IFFT transform of the resulting descrambled frequency-domain data.
A problem of frequency-domain implementation is that each OFDM symbol can only cover a maximum timing range of 2048 Ts. This is due to the fact that in the frequency domain the timing measure is circular so that we can only get a time measure modulo 2048 Ts which is the FFT size (to enable frequency-domain processing as would be understood).
Taking the following example to illustrate the limiting range, in
For ease of explanation, it is assumed that the channel is composed of only one tap. Then, the power-delay profile (PDP) obtained from the reference OFDM symbol n is as in
Since the measured delay is cyclic and only measured modulo 2048 Ts, and with the range covered by a window, for example [0, 2048 Ts], or [−1024 Ts, 1023 Ts], a delay not falling in this range will be estimated wrongly, i.e. with an error that is a multiple of 2048 Ts. In the example of
Moreover, by convention, if only timings in [−1024 Ts, 1023 Ts] are measured, the FFT window is represented by inverting its left and right halves so that the 0Ts delay appears to be in the center of the FFT window. This representation is used in
For instance, the positioning symbol at (b) results in a channel tap at delay 0Ts. At position (d), the channel tap appears in the negative part of the PDP at position (τ−1)Ts from −1024 Ts due to phase wrapping. This situation shows the fact that it is ambiguous whether channel tap in the resulting PDP corresponds to a real negative delay or a wrapping positive delay. Similarly, there is the same issue the other way round, where a signal having a real negative delay of (−1024−τ)Ts will appear as having a wrapped delay of (1023−τ)Ts
In relation to position (d) in
The above examples can be generalized to show that a frequency-domain implementation faces ambiguity regions as illustrated in
The aforementioned ambiguity regions are inherent to the limiting timing range of the PDP obtained from one reference OFDM symbol. Could this ambiguity problem be avoided by using a PDP that is computed from a combination of several reference OFDM symbols instead of from only one?
Let us briefly describe some (but not all) types of combinations to see what combination means. Take example (c) of
For such a combined PDP, it is beneficial but not essential to define its timing range as [0,2047], as shown in
In light of the above, ambiguity still occurs even with a combined PDP. Let us consider combined PDP obtained from reference OFDM symbols n and n+1 as in
It can be generalized to show that the PDP obtained from the symbol combined from reference symbols n and n+1 has ambiguity regions as shown in
As a consequence, ambiguity is an issue of the frequency-domain approach.
In a typical frequency-domain implementation of OTDOA receiver, PRS pilot tones corresponding to each eNodeB are extracted in the frequency domain, descrambled and transformed back into time domain on an OFDM symbol basis. Then, accumulation or interpolation of the time domain symbols is performed for all PRS-carrying symbols of a particular subframe to obtain one power delay profile on which timing estimation will be done. The operations of interpolation/accumulation and iFFT may be inverted i.e. carried out in a different order. Then a UE search window is placed around the highest correlation value for the search of the first channel path.
However, all frequency-domain implementations proposed so far have only considered small uncertainty range for which the ambiguity issue is not addressed, whereas time-domain implementations allow performance gain but are known to be computationally more costly.
Accordingly, there is a need to provide a method for ambiguity resolution of RSTD estimation in a frequency-domain implementation of OTDOA measurements in an LTE network.
According to a first aspect there is provided a method as defined in claim 1 of the appended claims. Thus there is provided a computer implemented method of providing RSTD data comprising:
receiving an uncertainty window centered around an expected RSTD value, determining the PDP of each reference OFDM symbol within the uncertainty window, obtaining a main PDP by calculating a parameter indicative of signal quality for each determined PDP, the main PDP having the highest signal quality, obtaining a preceding PDP of the Main PDP, obtaining a succeeding PDP of the Main PDP, determining PDP metrics comprising:
Optionally, the method wherein the signal quality is indicative of whether a respective reference OFDM symbol contains positioning data of a measured cell.
Optionally, the method wherein the PDPs comprise combined PDPs.
Optionally, the method wherein combined PDPs are obtained by at least one of coherently combining frequency-domain data; or combining two data sets output from an IFFT of each reference OFDM symbol by complex value combination or power combination.
Optionally, the method wherein if the main PDP is a combined PDP, coherence testing is performed on the PDP metrics and delays to detect any ambiguity in delay estimate of the main PDP, and any ambiguity is corrected.
Optionally, the method wherein the PDPS are calculated from at least one of a positioning symbol, a positioning subframe, a predetermined positioning frame, a positioning occasion.
Optionally, the method wherein ambiguity is corrected by adding an FFT size delay to the delay estimate of the main PDP if:
Optionally, the method wherein ambiguity is corrected by subtracting an FFT size delay from the delay estimate of the main PDP if:
Optionally, the FFT size delay is equivalent to the FFT window size.
Optionally, the method wherein the RSTD is measured using PRS.
Optionally, the method wherein the uncertainty window is longer than the FFT window used for frequency domain processing.
Optionally, the method wherein correlation with an incoming pilot sequence is executed in the frequency domain.
Optionally, the method further comprising determining a set of consecutive FFT windows for processing in the frequency domain.
Optionally, the method further comprising determining an RSTD estimate based on the delay estimate of the main PDP.
Optionally, the method wherein if the main PDP is a combined PDP the performing step is executed.
According to a second aspect there is provided an apparatus as defined in claim 14.
Optionally, the apparatus may comprise a mobile device. Optionally the apparatus may comprise a UE. Optionally the apparatus may comprise a processor.
According to a third aspect there is provided a computer readable medium as defined in claim 15.
With all the aspects, preferable and optional features are defined in the dependent claims.
Embodiments will now be described, by way of example only, and with reference to the drawings in which:
In the figures, like elements are indicated by like reference numerals throughout.
The disclosed embodiments comprise detecting and correcting ambiguity of a timing estimate obtained from a main power-delay profile (PDP).
By ‘main PDP’, we define a PDP that is selected as the best between all possible candidate PDPs for RSTD estimation, and for which ambiguity may be present. To select the best PDP, a metric is chosen that is representative of the signal quality (for example peak power of the PDP, or peak SNR on the PDP). The PDP having the maximum metric is chosen.
Note, however, that:
The method is not limited to any possible extension on how a PDP is computed from frequency-domain data, for instance a PDP can be obtained from a positioning symbol, or from a positioning subframe, or from a predetermined (given) positioning frame, or from a positioning occasion. For the sake of simplicity, in the following, we consider that a PDP is computed from a positioning subframe, however no change is required if the PDP is calculated by a different method.
The proposed method is not limited to any specific method for determining the possible candidate PDPs from the provided uncertainty window.
Also, the proposed method is not limited to any specific way of electing the main PDP from all possible candidate PDPs. The main PDP is chosen that has the maximum metric, where the metric is a quantity proportional to how much the considered OFDM symbol contains the positioning data of the measured cell.
In summary, the method comprises
1. Retrieving the PDP that immediately precedes the main PDP, this is defined as the ‘preceding PDP’.
2. Retrieving the PDP that immediately succeeds the main PDP, this is defined as the ‘succeeding PDP’.
3. Estimating metrics from each PDP,
4. Performing coherent testing on the estimated metrics to detect ambiguity,
5. Correcting ambiguity.
The resulting method provides Robustness against channel conditions (channel profile, SNR) over the maximum range of standardized uncertainty. (It is noted that, as per LTE standard specification, the size of the uncertainty window can be 2×3069 Ts).
Implementation independent of the underlying frequency-implementation, e.g. with or without different types of combination, and flexible channel metrics.
Low complexity and processing overhead. The proposed ambiguity resolution only adds a few calculations for coherence testing based on existing computations of a typical frequency-domain implementation (e.g. IFFT). These calculations are negligible in comparison to existing computations. The computational overhead is therefore negligible.
Throughout this specification, the following definitions apply:
A PDP can be directly computed from a single reference OFDM symbol, or from a combination of several reference OFDM symbols through a combination technique as previously described in relation to
The proposed method is not limited to whether the underlying frequency-domain implementation uses any specific combination or not. For completeness, we describe the proposed method both with and without combination.
Throughout this specification, a PDP is defined as ‘original PDP’ when it is computed from a single OFDM symbol. PDPn denotes the original PDP obtained from reference OFDM symbol n.
A PDP is defined as ‘combined PDP’ when a combination is applied and the combined PDP is obtained from a combination of several OFDM symbols. PDP(n,n+1) denotes the combined PDP obtained from reference OFDM symbols n and n+1.
By convention, the timing range of an original PDP is [−1024, 1023], and of a combined PDP is [0, 2047] with the origin as shown in
When combination is NOT applied:
The preceding PDP of original PDP is defined as original PDPn−1.
The succeeding PDP of original PDP is defined as original PDPn+1.
When combination is applied:
The preceding PDP of original PDP is defined as the combined PDP(n−1,n).
The succeeding PDP of original PDP is defined as the combined PDP(n,n+1).
The preceding PDP of combined PDP(n,n+1) is defined as original PDPn.
The succeeding PDP of combined PDP(n,n+1) is defined as original PDPn+1.
For each PDP (the main, the preceding, and the succeeding PDPs), the following is calculated:
Channel metric: this may be channel power, channel-power-to-noise ratio, or any other quantity that can be used to estimate the quality of the positioning symbol of interest on this OFDM symbol
Channel main tap: this is the tap with the maximum power between the different taps of the channel region (see below for the definition of channel region)
Delay estimate: this is the RSTD estimate obtained from the PDP in question.
Turning to the example as shown in
{circumflex over (d)}
n+1=(2048+CPA)+τ.
So, let us denote by:
μ0, τ0, {circumflex over (d)}0: channel metric, main tap delay, and delay estimate obtained from the main PDP respectively.
μ−1, τ−1, {circumflex over (d)}−1: channel metric, main tap delay, and delay estimate obtained from the preceding PDP respectively.
μ+1, τ+1, {circumflex over (d)}+1: channel metric, main tap delay, and delay estimate obtained from the succeeding PDP respectively.
For a given PDP, let N be the number of its taps. N is essentially the size of the IFFT used for frequency-domain to time-domain transform.
Denote:
{pk, k=0, . . . , N−1}: is defined as the set of tap powers of the PDP.
{dk, k=0, . . . , N−1}: is defined as the set of tap delays of the PDP.
When 0<1≦N and 0≦i<N, we define:
Ri(l)((i+k) mod N, k=0, . . . , 1−1).
Here, the window Ri(l) is a candidate channel region, the index 1 stands for the length of this candidate channel region and the index i for its position inside the PDP (see section “control parameters” below for additional information).
Then,
Pi(l)ΣjεR
warg maxjεR
From the above notation, an estimate of the channel region with length l of a given PDP, say A, is determined as:
Λ=Rj(l)|j=arg maxi=0, . . . ,N−1{Pi(l)}, given l
Then,
The channel power is the power of Λ.
The channel main tap is the main tap of Λ. Its delay is τ=dw.
Coherence testing checks the coherence between the estimates obtained from the main, the preceding, and the succeeding PDPs to detect ambiguity.
The below describes ambiguity resolution when symbol combination is used (when symbol combination is not used, the metric estimation on each PDP is the same but coherent testing on the estimated metrics to detect ambiguity and the actual ambiguity correcting should be adapted as would be understood):
The case where a non-combined PDP is selected as the main PDP statistically occurs in situations where the signal to be measured is nearly aligned with the FFT window (in the examples herein the size of the FFT window is 2048Ts) used to generate this main PDP. Actually, for high delays (typically higher than +/−400 Ts), the measured signal will have significant overlap with two consecutive OFDM symbols and there is a high probability that the corresponding combined PDP will be selected as the main PDP.
Thus, in the case of main PDP being a non-combined PDP, we can make the assumption that there is no significant overlap, and consider that the estimated timing does not require any ambiguity correction.
First, following the previous section, we can add a pre-processing stage where, if the signal to be measured is nearly aligned with the FFT window used for PDP computation, the timing measure is carried out on the corresponding non-combined PDP and the ambiguity resolution described below is optional.
This occurs typically when the measured delay is in a neighborhood of symbol time, i.e. in the case where the measured delay is close to either 0 or 2047. This can be expressed as min(τ0, 2047−τ0)<ξ. The value of is configurable, and can be typically equal to a few hundreds of Ts, e.g. 400Ts.
In the case where the delay is, for example, more than 400Ts, the procedure below may be applied to remove the ambiguity. It follows that, the main tap delay obtained from the main PDP is τ0ε[0Ts, 2047Ts] as the main PDP is a combining PDP. An ambiguity in τ0 will result in ambiguity of the delay estimate {circumflex over (d)}0. We examine τ0 for the two following cases:
When τ0Σ[0,1023] as shown in
This main tap can correspond to two possibilities as shown in
2047+τ0+1−τ0=2048 Ts.
It means that this ambiguity should be corrected as
{circumflex over (d)}
0
:={circumflex over (d)}
0+2048. (1)
This is a correction of adding a delay equivalent to the FFT window size and can be generically thought of as such.
The point here is that this correction must not be applied if the real channel tap is positioned as in
Firstly, as previously described, the succeeding PDP is the original PDPn+1 obtained from the reference OFDM symbol n+1. Its delay estimate {circumflex over (d)}+1 is taken for case A. Thus, it is easy to verify that {circumflex over (d)}+1={circumflex over (d)}0+2048.
For the fact that channel may change over OFDM symbols, this condition can be re-expressed as
|{circumflex over (d)}0+2048−{circumflex over (d)}−1|≦ε (2)
where ε≧0 is a configurable matching threshold. A possible value for ε is 72 Ts (half the size of the CP).
Secondly, the channel tap for case A implies that reference OFDM symbol n+1 should contain the positioning symbol of the measured cell whereas reference OFDM symbol n should not. Thus, the channel metric measured on the succeeding PDP, μ+1, should be better than that measured on the preceding PDP, μ−1. This is expressible as
α(μ0−μ+1)<(μ0−μ−1), (3)
where α≧1 is a configurable threshold to compensate the effect of noise. A possible value for α is 1.5, which provides good results in the tested configurations.
That is the difference between the main PDP channel metric and the preceding PDP channel metric divided by the difference between main PDP channel metric and succeeding PDP channel metric being higher than α.
Generically, (3) could be the result of a two variable function having channel metric of preceding PDP as a first argument and channel metric of succeeding PDP as a second argument being higher than α.
Thus, correction (1) shall be applied when detecting conditions (2) and (3) hold.
We refer to the subsequent section for how to choose the involving parameters.
When τ0ε[1024, 2047] as shown in
The channel tap can correspond to two possibilities as shown in
τ0−(τ0−2048)=2048Ts.
Thus, the delay estimate {circumflex over (d)}0 should be corrected as
{circumflex over (d)}
0
:={circumflex over (d)}
0−2048. (4)
This is a correction of subtracting a delay equivalent to the FFT window size and can be generically thought of as such.
Similar to the cases of
Firstly, the preceding PDP is the original PDP obtained from DLP symbol n. Its delay estimate {circumflex over (d)}−1 is taken at position (a), resulting in
{circumflex over (d)}
−1
=d
0−2048.
To take into account channel variation over OFDM symbols, this condition can be re-expressed as
|{circumflex over (d)}0−2048−{circumflex over (d)}−1|≦ε. (5)
Secondly, channel tap at position of case A of
α(μ0−μ−1)<(μ0−μ+1). (6)
That is the difference between the main PDP channel metric and succeeding PDP channel metric divided by the difference between main PDP channel metric and preceding PDP channel metric being higher than a.
Generically, (6) could be the result of a two variable function having channel metric of the succeeding PDP as a first argument and channel metric of the preceding PDP as a second argument being higher than a.
Thus, ambiguity correction (4) shall be applied when detecting conditions (5) and (6) hold.
The described methods use three parameters:
Channel region length, l≧1
Matching threshold, ε≧0
Metric coefficient, α≧1.
While the proposed method remains flexible for parameter setting, guidelines are provided for the configuration of these parameters.
Although the channel length l should be proportional to the channel delay spread, evaluation over different channel types has shown that l=3 in 512 IFFT basis is an efficient tradeoff for all channels. Preferably, l is tuned to the exact delay spread of the channel since this measures the width of the window containing the useful signal. Therefore, preferably, l is equal to one tap for single tap channel and the value of l is increased to values greater than 3 for wider channels. In practice, using the same value l=3 for all channels provides a good tradeoff.
Matching threshold is aimed at taking into account channel variation over symbols, as well as possible imperfection of the channel tap estimation. Evaluation has shown that the proposed method is not sensitive to ε (where we recall that ε≧0 is the configurable matching threshold, see above). One can set ε to some tens of Ts, even to one hundred Ts.
Metric coefficient α is used to take into account the effect of noise. However, the comparison of relative channel metrics as proposed in Eq (3) and Eq (6) has been shown to be significantly robust against simulated signal-power-to-noise ratio (SNR) over different channels. Setting α=1.5 gives the best performance tradeoff over different scenarios.
The proposed ambiguity resolution method with the parameter setting described above has been challenged over different scenarios and its performance is shown for Additive White Gaussian Noise (AWGN) in
These evaluation scenarios have been set for the most challenging configuration with the maximum uncertainty window of [−3×1023, +3×1023] Ts, and input RSTD varying from −3000 Ts to 3000 Ts. This case is the most challenging one because it corresponds to the maximum parameter values authorized by the LTE standard.
In addition, the SNR of the measured cell was set as low as −13 dB. The performance was evaluated as the percentage of remaining RSTD measures that contain ambiguity.
The evaluation results show that the proposed ambiguity resolution method enables frequency-domain implementation to work over a wide range of delay and of uncertainty. In each graph of
As discussed previously, the methodology disclosed herein provides benefits of:
Robustness against channel conditions
An implementation that is independent of the underlying frequency-implementation
Low complexity and processing overhead.
The various embodiments described above may be implemented by a computer program product. The computer program product may include computer code arranged to instruct a computer (processor) to perform the functions of one or more of the various methods described above. The computer program and/or the code for performing such embodiments may be provided to an apparatus, such as a computer (processor), on a computer readable medium or computer program product. The computer readable medium may be transitory or non-transitory. The computer readable medium could be, for example, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, or a propagation medium for data transmission, for example for downloading the code over the Internet. Alternatively, the computer readable medium could take the form of a physical computer readable medium such as semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory (RAM), a read-only memory (ROM), a rigid magnetic disc, and an optical disk, such as a CD-ROM, CD-R/W or DVD.
An apparatus such as a computer (processor) may be configured in accordance with such code to perform one or more processes in accordance with the various embodiments discussed herein.
| Number | Date | Country | Kind |
|---|---|---|---|
| 15306019.9 | Jun 2015 | EP | regional |
