The present disclosure is related to alignment of a transmit antenna beam in a millimeter wave (mmWave) band using ultra wide band (UWB) signals.
A frequency band of mmWave communication occurs in a range of frequencies extending from 28 GHz to 100 GHz (“mmWave band”). In this frequency band, there are low delay spreads from multipath components created by reflectors within a single room.
A problem exists with communication in the mmWave band because multipath components may be strongly attenuated. Communication using a direct path overcomes this problem (if the path is not blocked). However, beam alignment is difficult with a small number of antennas at a User Equipment device (UE) because a beam of an antenna array with a small number of antennas is wide.
In embodiments, mmWave communication is performed using antenna beams that are aligned using signals at two or more ultra wide-band (UWB) frequencies. Embodiments provide a technique for mmWave beam alignment of the UE using a UWB radio in the UE. An example UWB radio transmits pulses with a bandwidth of 500 MHz at carrier frequencies in a range from 3.5 GHz to 10 GHz. Embodiments perform the alignment using two or more UWB carrier frequencies in order to overcome a limitation of a small number of antennas at the UE.
Provided herein is a method of aligning mmWave beam transmissions, the method including: receiving, at a second antenna array of a second device, a first plurality of ultra wideband (UWB) pulses from a first antenna element of a first device, wherein the first plurality of UWB pulses has a first carrier frequency, and a first antenna array of the first device includes the first antenna element; receiving, at the second antenna array of the second device, a second plurality of UWB pulses from the first antenna element, wherein the second plurality of UWB pulses has a second carrier frequency; computing, based on the first plurality of UWB pulses and the second plurality of UWB pulses, an angle of arrival (AoA) spectrum at the second device; and aligning a mmWave beam at the second device based on the AoA spectrum.
Also provided herein is a device including: an antenna array; one or more memories; and one or more processors, wherein the one or more processors are configured to access a computer program stored as plurality of instructions stored in the one or more memories, wherein the computer program is configured to cause the one or more processors to: receive, at the antenna array, a first plurality of ultra wideband (UWB) pulses from a first antenna element of another device, wherein the first plurality of UWB pulses has a first carrier frequency, and another antenna array of the another device includes the first antenna element of the another device; receive, at the antenna array, a second plurality of UWB pulses from the first antenna element of the another device, wherein the second plurality of UWB pulses has a second carrier frequency; compute, based on the first plurality of UWB pulses and the second plurality of UWB pulses, an angle of arrival (AoA) spectrum at the device; and align a mmWave beam at the device based on the AoA spectrum.
Also provided herein is a non-transitory computer readable medium storing a computer program, wherein the computer program is configured to cause a device to: receive, at an antenna array, a first plurality of ultra wideband (UWB) pulses from a first antenna element of another device, wherein the first plurality of UWB pulses has a first carrier frequency, and another antenna array of the another device includes the first antenna element of the another device; receive, at the antenna array, a second plurality of UWB pulses from the first antenna element of the another device, wherein the second plurality of UWB pulses has a second carrier frequency; compute, based on the first plurality of UWB pulses and the second plurality of UWB pulses, an angle of arrival (AoA) spectrum at the device; and align a mmWave beam at the device based on the AoA spectrum.
Provided herein is another method of aligning mmWave beam transmissions, the another method including: receiving, at a second antenna array of a second device, a first plurality of ultra wideband (UWB) pulses from a first antenna element of a first device, wherein the first plurality of UWB pulses has a first carrier frequency, and a first antenna array of the first device includes the first antenna element; receiving, at the second antenna array of the second device, a second plurality of UWB pulses from the first antenna element, wherein the second plurality of UWB pulses has a second carrier frequency; computing, based on the first plurality of UWB pulses and the second plurality of UWB pulses, an angle of arrival (AoA) spectrum at the second device, wherein the AoA spectrum has a plurality of intensity values at a corresponding first plurality of angle values; calculating a confidence value based on the AoA spectrum; when the confidence value is higher than a threshold: aligning a mmWave beam at the second device based on the AoA spectrum; when the confidence value is not higher than the threshold: converting the AoA spectrum to a two dimensional image; estimating, using a convolutional neural network applied to the two dimensional image, a second plurality of fine grained angle values; and aligning the mmWave beam at the second device based on the second plurality of fine grained angle values.
The text and figures are provided solely as examples to aid the reader in understanding the invention. They are not intended and are not to be construed as limiting the scope of this invention in any manner. Although certain embodiments and examples have been provided, it will be apparent to those skilled in the art based on the disclosures herein that changes in the embodiments and examples shown may be made without departing from the scope of embodiments provided herein.
UE 1-1 includes antenna array 1-2 and AP 1-11 includes antenna array 1-12.
An x, y, z, coordinate system 1-15 is indicated in
Generally, the signal 1-3 is the form of a narrow beam, see example beam 1-18 in
An azimuth angle, θ 1-20, is indicated in
In comparative approaches to the problem of beam alignment, AP 1-11 would sweep the beam 1-18 from antenna array 1-12 and UE 1-1 would record observations. The sweeping by the AP 1-11 would be done by controlling the phase of signals provided to each of antenna elements 4, 5, 6, 7 (see
Embodiments use UWB pulses 1-13 at two or more carrier frequencies in order to find good values for the pair (θalign 1-15, θalign 1-17), that is, achieve beam alignment.
Referring again to
AP 1-11 then repeats the transmission of UWB pulses from the same antenna element but with a different carrier frequency. The difference between one carrier frequency and the next may be, for example, about 500 MHz. UE 1-1 receives the second group of transmitted pulses and stores observed intensity and phase values. The vector may be extended to include the second group.
A correlation matrix representing a second moment of the vector may then be formed. Next, a noise subspace of the correlation matrix may be determined. Finally, an angle of arrival spectrum may be obtained based on an inverse of a quadratic form with the kernel of the quadratic form based on the noise subspace. Before describing the full details of embodiments, some preliminary expressions are developed.
Math Preliminaries
A phase difference between an mth antenna element and a reference antenna element is given in Eq. 1.
where the antenna elements are spaced by distance d 1-21, λ is the wavelength of the carrier wave impinging on the array, m is the index of an antenna in an array of M antennas and p is the index of the arriving carrier wave (there may be L such waves, with L>1 if scattering exists).
The received signal at the mth antenna is given in Eq. 2.
xm=Σspe−jω(m,θ
where the summation is over the index p from 1 to L, and s is a complex pulse shape, attenuated for the pth path, possibly with data modulation. The function ω(⋅,⋅) is defined in Eq. 1.
There are M elements in the antenna array being considered as receiving a signal. The received signal vector X at one moment in time sampled over the M antennas is given in Eq. 3.
X=AS+n Eq. 3
Where n is a noise vector, A is a steering matrix with columns a(θp) given in Eq. 4.
a(θp)=[1,e−jω(2,θ
where (⋅)T denotes transpose.
An algorithm known as MUSIC estimates angles of arrival (AOAs) by analyzing the eigenstructure of the correlation matrix R=[XXH] of the received vector X, where (⋅)H denotes Hermetian transpose. The correlation matrix R has M eigenvalues. The largest L (L<M) eigenvalues correspond to L incoming signals while the rest, M−L eigenvalues, correspond to noise.
The eigenvectors corresponding to the M-L eigenvalues form the basis of a noise vector subspace UN. Due to the orthogonality between the steering vector and the noise subspace, a(θp)HUN is close to zero. The AoA spectrum of MUSIC is given in Eq. 5.
Solution, Multiple Frequencies
As shown above, AoA can be estimated because different phase differences at elements of an antenna array depend on angle of arrival (AoA). Embodiments use the fact that phase difference values at different frequencies also depend on angle of arrival (AoA). Embodiments use phase differences of multiple frequencies to improve AoA estimation accuracy with a limited number of antennas.
Embodiments extend the MUSIC algorithm to multiple carrier frequencies. Suppose that there are K frequency bands in UWB and their center frequencies are {f1, . . . fk, . . . , fK}. The AP 1-11 transmits two signals with frequencies of f1 and fk. Then, the two signals will arrive at antenna array 1-2 with the same AoA θp. The signal phase difference ϕ(m,fk) of the two frequencies measured at the mth antenna is given in Eq. 6.
where c is the speed of light, d 1-21 is the inter-element spacing and D 1-22 is the distance between the transmitter and the first antenna. D 1-22 can be measured by UWB radios accurately.
For a given geometry of UE 1-1, AP 1-11, the antenna array 1-2, fk, f1, D 1-22, and d 1-21 are constants. Then, the phase difference in Eq. 6 can be simplified as shown in Eq. 7.
ϕ(m)=C1˜cos(θp)+C2 Eq. 7
where C1 and C2 are two constants related to fk, f1, D 1-22, and d 1-21. Eq. 7 implies that, if the phase difference of the mth antenna in the given two frequencies is known, then the AoA θp (θ 1-20 in
Using multiple frequencies in the frequency domain can increase the accuracy and resolution of AoA estimation. Comparative AoA estimation methods (e.g., MUSIC) use only a single frequency and multiple antennas in the spatial domain.
An example of measureable phase differences is provided here based on Eq. 1 and Eq. 6. In the example, for AoAs of 0-180 degrees, a measurable range of phase differences using two frequencies in the frequency domain is increased by 5:83/3:14=1.86, compared with the phase difference measured at two antennas in the spatial domain. In this example, Eq. 1 of spatial domain has been compared with Eq. 6 of frequency domain for example parameter values of f1=3494.4 MHz, fk=9984.0 MHz, and M=2 (i.e., two antennas). Overall, the phase differences vary by different AoAs, which is important for AoA estimation. Also, the example demonstrates that using multiple frequencies can improve the accuracy of AoA estimation.
Generalizing these ideas, consider an antenna array with M antennas. Each antenna can work with K frequency bands. Then, embodiments form a virtual array composed of M x K virtual antennas. Consider the first frequency (i.e., f1) as the reference frequency and the first antenna as the reference antenna. Let sp,f
The summation in Eq. 8 is over the index p ranging from 1 to L.
The received signal vector X′ at the antenna area is given in Eq. 9.
The steering vector, similar to Eq. 4, is given as
Applying the same steps as used to progress to Eq. 5, Eq. 9 can be used to find a noise subspace UN′ based on a correlation matrix R′, where R′=[X′(X′)H]. The AoA spectrum P′(θp) is given in Eq. 10.
P′(θ) shows peaks at AoAs of incoming signals. Well-separated angles show sharper peaks.
Varying the free variable of the quadratic form over candidate angles of arrival (AoA or θ) provides values of the AoA power spectrum, P′(θ). In some embodiments, the UE 1-1 may then configure phase values of a steering vector applied to the elements of the antenna array 1-2 so that the antenna array 1-2 forms a beam 1-16 in a direction θalign 1-15. θalign 1-15, in some embodiments, is given as
Operation 2-2 includes transmitting a first group of UWB pulses from a first antenna of a first device (omnidirectional). The first group of UWB pulses has a first carrier frequency. For example, this first antenna could be any of the antenna elements 1-4 of antenna array 1-12 of AP 1-11 shown in
When there are multiple arriving waves, such as signals 1-3, 1-5 and 1-7 of
However, due to possible penetration loss due to the direct path passing through an attenuating object, an indirect path may be better. Embodiments provide an algorithm for determining which path to align the beam with.
Specifically, when the direct path is clear, the direct path angles are the best alignment angles, since it has the minimum energy attenuation than other paths. When the direct path is blocked, mmWave radios have to find a reflection path for communications. To identify if the direct path is blocked or not, a method compares the received UWB signal power against a pre-obtained baseline signal power when the direct path is clear, since a blockage, if present, will decrease the signal power significantly. Embodiments use a UWB estimated distance and propagation models to estimate the baseline signal power. As an example, when the direct path is clear, the received signal power Prx equals to the transmitted power Ptx minus the attenuation loss ΔR over a propagation distance D as shown in Eq. 11.
Prx=Ptx−ΔR=Ptx10γ log D Eq. 11
where γ is the rate of signal power attenuation and is related to the environment.
Existing UWB technology can estimate D accurately with mm-level accuracy and Ptx is available at the transmitter side. Thus, Eq. 11 implies that the estimation accuracy of baseline signal power depends on γ which varies over different environments. For a given environment (γ), embodiments calculate the baseline signal power PBrx of the direct path using Eq. 11, and measure the direct path signal power PMrx, If |PBrx−PMrx| is smaller than a threshold Th (Th=3 dB, for example), embodiments identify the direct path as clear; otherwise, the direct path is blocked. The beam alignment angle selection method is summarized in Table 1 below.
The beam alignment selection algorithm is also provided by the logic 3-9 of
At operation 3-2, groups of UWB pulses are received at corresponding carrier frequencies, one carrier frequency per group of UWB pulses. At operation 3-4, an AOA spectrum is computed, based on the groups of UWB pulses. Operation 3-6 includes detecting a first peak at 01 of the AoA spectrum, and detecting a second peak at 02 of the AoA spectrum. Operation 3-8 includes determining an expected received power PB based on range R and a propagation coefficient γ and determining an actual received power PM. Operation 3-10 includes choosing θ1 or θ2 based on PB, PM, D and γ.
At operation 4-2, the logic 3-9 is performed with UE 1-1 determining θalign 1-15. At operation 4-4, UE 1-1 uses θalign 1-15 determined using UWB to adjust beam 1-6. At operation 4-6, the logic 3-9 is performed with the AP 1-11 determining θalign 1-17 using UWB. AP 1-11, at operation 4-8, adjusts beam 1-18 using θalign 1-17. At operation 4-10, mmWave data is transmitted by UE 1-1 and/or AP 1-11 using the aligned beams (beam 1-16 and beam 1-18).
In a radio antenna pattern, the half power beam width is the angle between the half-power (−3 dB) points of the main lobe, when referenced to the peak effective radiated power of the main lobe. Beamwidth is usually expressed in degrees and for the horizontal plane. As an example, a misadjustment in beam alignment may cause an impinging wave (signal such as 1-3) to be received at the half power beam width or at an angle further off the maximum of the beam with even more attenuation. Receiving an attenuated wave reduces the received signal power.
The maximum data rate that can be supported depends on the received signal power, the interference and the noise. The interference and noise do not decrease when the beam is mis-aligned. An upper limit on the achievable data rate is given by the Shannon-Hartley expression in Eq. 12.
B is the channel bandwidth, PS is the received signal power and PN represents the interference and noise power. For misalignment of a small amount so that the received signal arrives at the shoulder of the beam (half power beam width), PS/PN is reduced by ½. At low SNR of PS/PN=1, the maximum Data Rate is then reduced to log2(1+0.50*1)=0.58 B from 1.0 B, a 42% reduction in the maximum (Eq. 12). The benefit of accurate beam alignment using multiple frequencies is further illustrated in
As shown in Table 2, curve 5-3 is the result of the MUSIC algorithm with M=16 antennas and K=1 frequency and curve 5-6 is for M=3 antennas and K=1 frequency.
The computed AoA spectra of some example embodiments is shown by curves 5-4 and 5-3. Curve 5-4 is for M=3 antennas and K=9 frequencies. Curve 5-5 is for M=3 antennas and K=5 frequencies. The reduction in maximum data rate depends on the radio antenna pattern of the antenna array.
The non-linearity of the MUSIC algorithm is reflected in curves 5-5 and 5-6 not being symmetrical with respect to the average of these angles, 75 degrees.
When multiple signals arrive at close angles, the AoA spectrum of a weak signal will be submerged in strong signals. Thus, existing methods are hard to estimate angle-of-arrival (AoA) of all signals, and will miss some signal's AoA.
To better identify all paths, embodiments provide an angle refinement method using a convolutional neural network (CNN). A CNN is a class of deep neural network. A CNN includes a shared-weight architecture of convolution kernels or filters that slide along input features and provide responses known as feature maps. A convolutional neural network consists of an input layer, hidden layers and an output layer. In any feed-forward neural network, any middle layers are called hidden because their inputs and outputs are masked by the activation function and final convolution. In a convolutional neural network, the hidden layers include layers that perform convolutions. Typically this includes a layer that performs a dot product of the convolution kernel with the layer's input matrix. This product is usually the Frobenius inner product, and its activation function is commonly ReLU. As the convolution kernel slides along the input matrix for the layer, the convolution operation generates a feature map, which in turn contributes to the input of the next layer. This is followed by other layers such as pooling layers, fully connected layers, and normalization layers. A CNN is familiar to one of ordinary skill in the art.
To determine whether the CNN should be applied, embodiments define a confidence rate (CR). CR=h/w, where h is the difference between the (peak spectrum value) and the 1.05*(Min spectrum value); w is the angle width when the line y=1.05*(minimum spectrum value) cut through the spectrum.
When two signals arrive at close angles, the peak spectrum does not indicate the two AoAs correctly.
The CR can detect the AoA missing issue. For example, in
In
Each spectrum is formatted as an image 7-7.
The image 7-7 is input to the CNN 7-1. At the output of the CNN 7-1, classifications 7-3 include estimates of a number of waves and AoA values for each wave.
Operation 7-22 includes receiving, at a second antenna array of a second device, a first group of UWB pulses from a first antenna element of a first device. The first group of UWB pulses has a first carrier frequency, and a first antenna array of the first device includes the first antenna element.
Operation 7-24 includes receiving, at the second antenna array of the second device, a second group of UWB pulses from the first antenna element, wherein the second group of UWB pulses has a second carrier frequency.
Operation 7-26 includes computing, based on the first plurality of UWB pulses and the second plurality of UWB pulses, an AoA spectrum at the second device. The AoA spectrum has intensity values at a first group of angle values.
Operation 7-28 first determines CR (see the discussion of
If the CR is not above the threshold, the logic flow 7-29 proceeds to operations 7-32, 7-34 and 7-36.
Operation 7-32 includes converting the AoA spectrum to a two dimensional image, for example, image 7-7 of
Operation 7-34 applies the image to the CNN 7-1 and obtains a number of peaks and the angle value corresponding to each peak (see 7-3 of
At operation 7-36, the mmWave beam is aligned at the second device based on the group of fine grained angle values. The selection of θalign may including using Algorithm 1 (Table 1) (not shown in
This application claims benefit of priority of U.S. Provisional Application No. 63/090,527 filed Oct. 12, 2020, the contents of which are hereby incorporated by reference.
Number | Name | Date | Kind |
---|---|---|---|
8965282 | Barbotin | Feb 2015 | B2 |
10408930 | Karls et al. | Sep 2019 | B2 |
10461421 | Tran | Oct 2019 | B1 |
20090213901 | Berens | Aug 2009 | A1 |
20190069294 | Va et al. | Feb 2019 | A1 |
20200049790 | Hollar | Feb 2020 | A1 |
20200088869 | Pefkianakis et al. | Mar 2020 | A1 |
20200382265 | Fukui | Dec 2020 | A1 |
Number | Date | Country |
---|---|---|
111650575 | Sep 2020 | CN |
Entry |
---|
Communication dated Jan. 7, 2022 issued by the International Searching Authority in counterpart Application No. PCT/KR2021/013890 (PCT/ISA/220, PCT/ISA/210, and PCT/ISA/237). |
Ju Wang et al., “Eliminating Millimeter-Wave Beam Alignment Cost With UWB”, Oct. 10, 2020, 7 pages total. |
Number | Date | Country | |
---|---|---|---|
20220116088 A1 | Apr 2022 | US |
Number | Date | Country | |
---|---|---|---|
63090527 | Oct 2020 | US |