CC BY
66
MIMO CHANNEL PREDICTION AND RECEIVE ANTENNA SELECTION FOR THE MOVING LINEAR ARRAY
Mikhail Yu. Starovoytov,
Manager Nokia, Moscow, Russia, mikhail.starovoytov@nokia.com
Keywords: Single User MIMO, spatial multiplexing, linear array, receive antenna selection, channel matrix prediction.
The Single User MIMO Spatial Multiplexing (SU MIMO SM) mode allows for reaching the highest spectral efficiency per transmit antenna for the fixed number of transmit and receive (RX) antennas per user.
The target considered architecture at the reception end is the typical scheme with regular linear antenna array installed on the roof of the mobile vehicle, and user equipment with a limited number of receiving circuits of RX signal detection/processing. The vehicle is equipped with on-board system for defining the velocity and direction of motion at every given moment.
For stationary radioconditions the solution for the problem of choosing the optimal subset of receive antennas is known since long ago. However for the case of fast moving RX antennas, when the change of radioconditions during one subframe interval (1 ms) is not negligible and the channel is essentially non-stationary, the RX antenna selection for MIMO is not addressed in literature. As the reason for that we see the principal difficulty of prediction of channel matrix elements for 1 ms ahead - in the situation that the number of RX antennas is more than the number of receiving circuits. The new method proposed in this article solves the principal problem of prediction of MIMO channel matrix elements with due accuracy in a particular case when the radio system satisfies certain additional conditions imposed on it.
At that the accurate velocity and direction of motion data received from on-board measuring system in the real time - are of key importance for the proposed method to work. The topic of defining the kinematic parameters of motion of vehicles with high precision has for today a rich coverage in scientific and technical literature; its development owes to the perspective of introduction of automatic steering of all moving transport objects including vehicles on the highway. We are proposing the non-traditional way to utilize the data from on-board measuring system for improving the MIMO communication system performance. In the model experiments for MIMO 4x4 Spatial Multiplexing mode, in the case of the new method applied to 4 out of 8 antennas selection - the 2 dB energy efficiency performance gain is availed compared to the standard 4x4 without antenna selection.
Для цитирования:
Старовойтов М.Ю. Предсказание характеристик канала MIMO и алгоритм выбора антенн на приеме при движении линейной антенной решетки // T-Comm: Телекоммуникации и транспорт. 2017. Том 11. №2. С. 56-62.
For citation:
Starovoytov M.Yu. (2017). MIMO channel prediction and receive antenna selection for the moving linear array. T-Comm, vol. 11, no.2, рр. 56-62.
SÓ
Introduction
The target architecture on the receiving side is assumed the standard scheme with regular linear antenna array placed on the top of mobile vehicle, and the mobile terminal with a fixed number of receiving circuits (circuits of signal processing). The vehicle is equipped with cm-board measuring system with the capabilities to determine the current velocity and direction of motion of the vehicle in real time.
For the Ml MO model the known problem f 1] of RX antennas selection is formulated as follows: number of RX antennas is more than number of receiving chains; it's necessary to switch the receiving chains to the same number of RX antennas so thai to get the energy efficiency gain. A vast choice of literature is available on the topic of RX antennas selection for stationary radio channel conditions [I, 2J. The dominant criterion used in the known up to date antenna selection algorythms in Ml MO systems, is the criterion of maximization of mutual information in the MIMO channel [3].
However, the RX antenna selection methods in cellular networks for the case of vehicle moving at medium and high speed (30 km/li and higher) - is not covered in literature. It seems that up to present time the application of RX antenna selection in the case of non-stationary channel with the nessessity to refresh the channel measurements data for every RX antenna at a pace of once per 1 millisecond — was not considered as feasible. Simultaneously we shall note that the use of MIMO Spatial Multiplexing mode for the fixed RX antenna set (without antenna selection) - is well covered in literature, up to the receiver moving speeds at 250 km/h [4,6].
In this article the problem is stated in the way: without increasing the share of resources assigned for transmitting the piloting information and without the change in the basic spatiotemporal characteristics of LTE standard, we shall find the evaluation algorithm that will predict the MIMO channel matrix for the period of I ms ahead for the case of fast moving RX linear array, to further apply the criterion of maximization of mutual information to the predicted channel data for finding the optimal subset of selected RX antennas.
The article is further organized as follows: ¡n Chapter I the model of Single User MIMO Spatial Multiplexing is described, and the mathematical statement for the problem of selecting the optima) RX antennas subset on the basis of the known channel information is given. Chapter 11 is split in paragraphs. In paragraph 2.1 the conditions, imposed on the system to be satisfied for applicability of the newly developed methods are described. In paragraph 2.2 the MIMO channel matrix evaluation and prediction is presented. Paragraphs 2.3 and 2.4 are dedicated to the description of measures taken to make numerical modelling possible, and the modelling Program is given in express form. Chapter III presents the numerical modelling results and their discussion,
Notational conventions in this article:
Upper- "//"designates Hermitian transform;
ckl(X) - matrix X determinant;
X (3,:)— line vector builtof all elements of matrix X in row 3;
/» = [230] - line vector with three elements 2, 3, 0;
/» = [1:/?] - line vector with whole number elements from I to R in increasing order;
X = 0 - a matrix with zeros in all positions; - sign of unity of sets;
A' ( />,:) - matrix composed of rows of matrix X with row numbers designated by (positive whole number) elements of vector p;
IK - all-ones on diagonal and zero elements the rest matrix of size K x K ;
" roundup (X)" - the rounding operator of real number X up
to the nearest whole number;
)« - mathematical expectation,
I. System model and mathematical problem statement
In LTE standard the lime frame is composed of 10 subframes of length 1 ms each, each frame in its turn could be simplistically seen as 14 symbols of equal length T [71. In the frequency domain the whole resource is spanned by 180-k!Iz blocks (PRBs), each block consisting of !2 subcarriers with c 15 kHz step. The Time-Frequency domain composed of 168 slots, in each slot (12 frequency subcarriers x 14 lime symbol intervals) we will call the "resource unit". In each "resource unit" the agreed and fixed amount of slots in certain places on lime-frequency plane is taken by pilot channels, that are used by the system for radio channel evaluation. The rest slots in "resource unit" are used for pay load data. The 1 ms long subframe interval is the - smallest unit of time, during which the pilot information is read and processed, and during which the channel characteristics are calculated and extrapolated over all slots of "resource units", and extended to the whole time-frequency domain containing all "resource units" in consideration.
We assume that the receiving terminal contains N "receiving circuits", each of which is switched (by means of high-speed antenna switch) to one of PN (P = 2,3,...) RX antennas of linear array, and slays connected to the chosen antenna throughout any given subframe.
From here on we introduce the designations and concessions, necessary to describe the SU MIMO SM channel:
T and 1 AT - lengths in time of one symbol and one subframe of symbols accordingly; in the case of LTE [7J we have: UT = \mc.
XI2 — half-wave length; in the 2 GFIz band X / 2-1,5 sm ■ Uniform length between elements of regular antenna array we'll take equal X/2 [8].
v - vector of speed of motion of vehicle (and antenna array on its top).
t>., - estimation of vector of speed, made by on-board measuring system.
<Jv = v—vej( - {unknown to the on-board measuring system)
error of speed estimation.
'H - set of 2-dimensional square complex matrices of any size, with all elements Gauss-distributed independent zero-mean unit-variance.
H - MIMO channel matrix of size PiWxN. He"H-
H - evaluation/prediction of matrix H for the next subframe period (1 ms).
.v — normed column vector of independently distributed symbols in space Cv, elements of which can with equal prob-
7TT
ability take the value from a complex discrete set x in accordance with the used standardt modulation scheme
(QPSKAbOAM MOAM ), S(xx") = /„ ■
n - nomied column vector of noise in space C;V, with all elements Gauss-distributed independent zero-mean, E(mr" ) = /, .
y -the observed RX column vector of size N bC*.
d - antenna switching vector of size N, containing in succession the numbers from the set I ...PN of selected RX antennas in linear array connected to 1,2,..., N -th receiving circuits.
The known narrowband Ml MO NxN model with N rows, selected from PN rows of matrix H [2, 3]:
y = yfSNR-H(dt)x+n (1)
Scalar factor yfSNR in (1) carries the traditional interpretation of Sienal-to-Noise ratio in relation to one transmit antenna [1].
Expression for Mutual Information - MI for model (1) of M1MO channel and for the given antenna switching vector d is
[3}i
MI (JSNR. H ( lI,:)] = logfdet (if (d,:)" H {</,:)+ f v ))
The classic optimization problem for choosing the RX antennas in MIMO stationary channel model (1) by the criteria of maximization of mutual information in MIMO channel for te given SNR and channel matrix H is stated as [1,2]:
d tb argmax( MI (y[SNR, H ($,:))) (2)
,yj;e(s}=jV!
Here the maximum is searched over all veetors s of size N , containing the subset of N indexes from the set of indexes [ I: /VP]. And the channel evaluation through processing of pilots could be assumed ideal - due to channel stationarity.
In non-stationary channel, the algorithm of N out of PN antennas selection with formula (2) meets a specific obstacle that exposes itself in fast motion of the RX array. As we have mentioned already, for reliable channel estimation through processing of pilots the N "receiving circuits" mapping to N out of PN RX antennas of linear array shall stay constant for the duration of any given subframe (147"). Therefore for Ihe case of the number of antennas PN more than N , for the pilot channels information to be read by N receiving circuits, a minimum of P subframe periods would be required during each of which the antenna switching vector would stay constant. So reading the pilot information from all PN antennas in one subframe duration is impossible in principle. Hence, when the receiver is moving fast enough, using the accurate H value in formula (2) is impossible, and we shall use the channel matrix evaluation H in the argument, which will diverge from H the more the higher will be the speed of motion.
In this work the solution for getting the evaluation/prediction H of channel matrix for the duration of next subframe period (I ms) is proposed. Then the found matrix H to be used in the argument of mutual information in the modified criterion:
d = argmax f Ml [-JSNR, H (s,:))j (2' ^
iCf(j)=.V|
II. The proposal
2.1. Necessary conditions imposed on the system to apply the channel prediction method
From here on we shall only consider the case of N = 4, P -2, MIMO 4x4 with selection of 4 RX antennas out of 8 (N = 4, PN = 8).
Let At - be the 1-st antenna in the direction of travel of the linear antenna array, решетки. To be able to predict channel charcteristics on the rest antennas we will demand
the CI and C2 conditions to be fulfilled:
CI: Considered are only the route parts where the vehicle moves along the straight line, where the axis along linear antenna array is oriented parallel to the ground plane (which is considered ideal), antenna array moves in the plane parallel to the ground plane with the constant speed V . On these direct parts of the route it's technically possible to orient the linear antenna array axis ideally along the vector of estimated direction of motion e. =v /1 v in zero time.
L est | f v I |
The plane parallel to the (ideal) ground plane, containing all possible positions and routes of antennas in space, we Will further call the "work plane".
Let's introduce two parameters: f tl = r0WK]Up{ f-y
From the CI condition it follows that, given <5v| |v|: in the moment of time t+ fT antennas with numbers 2...S will take approximately those positions in space that antennas with numbers 1 ...7 are holding in the moment /. It's illustrated on Figure I (top view at the work plane): the lower horizontal line designates the linear antenna array axis and simultaneously the vector of estimated direction of motion e, = v /|v I- Positions of an-
L est | est |
tennas A,,<i,,A},... at / = 0 are shown by bold circles uniformly Л/2 - distanced between neighbors.
Thin solid lines form the skewed lattice that is stretched on the veetors of the forecasted motion fT-vea (all horizontal solid
lines) and vectors of forecast error fT-Sv (all inclined solid lines). De-facto the movement of antennas in the plane of the Figure 1 goes along the dotted lines, col linear with the vector of actual speed v.
C2: The radio field characteristics values depend only on the point of measurement. If two antennas passed in different points in time through the same point on the work plane, then the radio field measurements from these two antennas will match upto multiplication factor - complex exponent.
The fulfilled condition C2 opens the possibility to establish a one-to-one correspondence between every point on the work space defined by radius-vector r related to a single non-moving coordinates origin point (on the ground), and complex string vector of channel characteristics of size N.
So, if the positions of antennas A,A:,...,jL in a given moment in time are defined by vectors г.,л,...,г„ on the work plane, then the channel matrix H in this moment in time is composed of eight rows А{г.): 1/(е,-)=Н{гг),е=%,2.....8-
\t = 3fT
fr- 61)
fT-bv
ML
"Axis"
West
the linear antenna array axis and 0 simultaneously
Center of coordinates №cCor estimated direction of motion on the "Axis"
Figure
Now we'll explain the complex exponent factor mentioned in the formulation of C2. Let's assume matrix H is formed of rows, measurements for each of which were made ¡11 points 011 work plane r,, r,,..., n. in different moments in time: r,,t2r,..•
Then:
_ i-2itvtfi X
h(r,),e= 1,2,..„8
offsets (e) = roundup
\vJ-T
+ \,e- 1,2.....f
(3)
However, the resulting matrix H will influence the search of vector d by means of formula (2') indirectly - only through the function mUsNR,h) • At that function Ml is immune to
multiplication of any selection of rows of matrix H by different complex factors of type e'"''. The similar discourse is also true for matrix H , and for line vector h. Therefore everywhere further: equations involving rows from matrices H , //, or vector-lines h. will be understood as equations with line-by-line modulo equality sign (equality sign subject to multiplication by a relevant complex exponent factor line-by-line).
Now let's consider the moving reference system associated with the vehicle. From the knowledge of on-board measuring system Sv = 0. It means, that on-board measuring system sees the trajectories of all antennas in the linear array lying on one line, that coincides with the linear array axis. Array axis on the vvork plane with the center of coordinates 0, associated with the current position of antenna Ar and with the positive direction
defined by ort (-£>,), we will further 011 designate as "Axis" (Axis is shown with bold horizontal line on Figure 1). Let's sequentially lay the segments of equal length |vtJ,|-7' on the Axis in
the direction (-<?) stalling from zero, and assign numbers to them. For our purposes it will do to take segments with numbers
from 1 to sizematrix= roundup1 ' 35 lhis succession
IWTJ
covers (from the point of view of on-board measuring system) the positions of all antennas At to As-
Coordinates on the Axis of all antennas A2,4>>.,. relative to At are fixed. Vector offsets containing the succession of all coordinates of positions of antennasAiA2,...,A on the Axis is written as:
It's assumed that the center of coordinates 0 of the Axis belongs to the segment # I -eL -jv„,\-T\Q].
Let's compose matrix H of size sizematrix x 4, where in any £-th row out of k e \Y,siz&natrix\ rows is contained the last saved
measurement result of vector h of size 4, related 10 the position
-«vM'^-O7,011 Axis-
2.2. Ml MO channel matrix estimation/prediction, modification of antenna selection criterion.
The general principle of work of iterations in the Program is shown on Figure 2. Arrows with dotted lines show the logical dependences in calculations.
for packet = 1: F
H
1 d
for svmb= 1:14
Pi,- ,P@\
I I \
\ 11(1.s).....H(8,:> \
end
------t H
end
Figure 2
At the start of every external cycles in the Program (corresponding to the beginning of every new 14-symbols subframe), the task is to find the forecast for matrix H based on the saved
previous data //, and based 011 that define vector d for the new subframe. We will consider the predicted value for the symbol #7 in the middle of the coming new subframe the best evaluation
for the predicted channel matrix in all the symbol intervals from I to 14 in the coming new subframe. Then for each antenna from the set of antennas except l-st the estimation/prediction of the channel characteristics vector could be made by reading from
memory the row of matrix // corresponding to the coming position of each antenna on the Axis:
tf{2:8,:) = H(offsets(2:8}-7,:) (4)
Relating to the l-st antenna. Channel vector estimation/prediction for it for the #7 symbol in the coming subframe could be made by a number of known approaches and regression methods. We will take the simplest possible approach - the predicted value for a, shall be equal to the latest in lime measured
value on a:
H( 1,:) = //(!,:) (5)
The accuracy of such prediction will naturally be the worse the higher is the speed of motion of the vehicle.
Let's introduce the restriction into the antenna selection rule (2'): antenna A shall be included in the selected set of chosen
RX antennas always, so we demand lei/- always .The remaining 3 antennas for the resulting antenna sw itching vector shall be chosen out of the set A,...A,- Thus we are substituting formula
(2*) in the search of global optimum in the problem stated in the form (1 )-(2') by the formula (6) of the search of local optimum:
d = argmax [mlUSNR,H(s,\))\ (6'
The rule of I ed , except improving the forecast quality of f ormula (5) for antenna a, (at the price of switching from global
to local optimum search), carries also the mission of providing conditions for the work of the proposed estimation/prediction method for antennas A2.....A expressed by formula (4).
The Program is developed to establish the characteristics of the proposed method. In its internal cycle the increment of time variable with step T is done 14 times, with each increment the coordinates of all 8 antennas p ,e = |,2.....8, in work plane are
refreshed, all 8 lines of matrix // are recalculated. With each increment all rows of matrix H are moved down by one row . H (2: sizematrix,:) = H (1: (sizematrix—l),:l (?)
The antenna switching vector d found in the external cycle defines the array numbers of antennas which in every given symbol period of the current subframe period shall be connected to receiving circuits and shall be receiving data and pilot information, With every T increment the renovation of information
in 4 lines of matrix //is done:
H(offsets{d),)=H{d,-) (8)
It's easy to see that the rule led, in combination with operations (7), (8) in the internal cycle of the Program, do provide the existence of non-zero estimation of elements in each of
sizematrix rows of matrix H (on each of sizematrix segments on the Axis) throughout the whole working period of the algorythm after having passed the short initialization period of time of length not more than InT.
Note that formula (9) means the inclusion of elements from the chosen rows of matrix H into the matching rows of matrix
H accomplished in real time during the run of the algorithm. This formula realization implies the presence of the given field of channel characteristics vector values h( r ) of size N in each
point r on the work plane. In the field experiments the assumption about given vector field for h is fulfilled automatically. However, for executing the computer modelling we have to find a way to set the (random) vector field for h on a big continuous area of the work plane, and for convenience of analysis and comparison it is desirable to provide that each element A(r) in
each work plane point r would belong to 'H . The way to do this is proposed in Paragraph 2.3.
2.3. MIMO ctianncl characteristics vector field generation in (lie work plane
Let's produce a geometrical construction: make a regular lattice with generating shape in the form of equilateral triangle, covering without breaches or over lappings the whole work plane; side of the equilateral triangle put equal to /1/2; Index the bundles of the resulting lattice in the form xy (x,y - even numbers): in the bundle with index 00 place the center of coordinates for radius-vectors. Vector pointing at the bundle xy let's designate ¡- . Then we can make a correspondence between any
vector in work plane r and the smallest equilateral triangle inside which the end of this vector is contained.
If for example r is contained inside the triangle with vertices rnf>ru'rQ2' t'ien thee do exist three real positive or zero factors
»(,,,<»„,<»„, 0m +<»K = 1, such that the representation
r^&^+m^+m^ - is unique [9]. Let's generate random
veetors of size 4 w ith elements from class 'H ■ p e H - one for
each bundle r • Now we formulate the
xy
Proposition:
Complex channel characteristics vector h (in our case row vector of size 4) for the antenna placed in the end of the pointing vector r on the work plane, could be for the modelling purposes represented as: /,(>) = + + , where coef-
ficients 6>ab,vcd,t».f> 0}ai + + (Oj = 1 - are baricentric coordinates of vector r = a ihrtl, + a> ,r , + a , r , - inside the minimal equilateral triangle of the regular lattice w ith vertices rah,r^,rtj ■ containing r.
It's immediately established that h e H, /,(r )-p and h(r)~> p -
2.4. Program for the numerical modelling of Sli MfMO 4x4 SM mode with the selection of 4 out of 8 RX antennas
For the convienee of further Program description let's introduce new functions:
"GEN" — function accomplishing the generation of random vectors of size 4 with elements from H : p e"H - one vector
per each lattice bundle r ;
"hV/" — the block of actions, when oil the work plane the random unit length vectors g and en. are generated, random
radius-vector p is generated, vectors v and v are defined, and
all 8 vectors p pointing to the initial positions of antennas
/(,,£ = 1,2.....8 arc ealculaled:
"fcr^t» Sv " H' «*,; * = + ;
"triang(r)" - function using vector r as input and giving the function from the Proposition in Paragraph 2.3 as an output. Then the triang(p ) is the row H(e,:)-
Program description:
Initial part: defining TtM2,\vtjl\t\8v\, SNR;
defining the set of bundles r ;
n = roundup
f ¿12 N \itm)rT.
; sizematrix - roundup
1l-ktl^
+ 1;
offsets (e) = roundup
{<¡-1)4/2
+ 1,6 = 1,2.....8;
I'-l-r
% formula (3) FOR j = 1:20000
% 20000 times the generation of a new channel matrix
GEN;:
INI;
F = 20; % any number > 8, does not influence the result
H[ 1: size matrix, 1:4) = 0; ¿ = [1357]; for packet = 1: F if packet 14 >2 n
//(],:)=//(!,:): % formula (5) H (2:8,:) = H (offsets (2:8) - 7,:); % formula (4)
d4 argmax {mi{4sNR,H {s,:)\\, <6>
{«(iq»']!:« v ''
s!:ii j)=,V)
else end
for symb = 1:14 for e = 1:8 P< =Pe+v-T\
H(e,-) = triang{p)\ end
H (2 : sizematrix,-) = H (I: [xizematrix-1),:}; % .formula (7)
H {offsets {d ),:) = //(</.:); % formula (8)
if packet = F
Rsymh = H{d,-)\
else end
end end
Standard error rate analysis for the given SNR - for all 14 channel matrices
END
Note that the initial vector value d=[1357]
Is fixed during initial iterations of the external cyclc until the condition packet4 > 2/7
Is reached: this has a purpose of having all the sizermatrix rows of matrix H tilled in by nonzero values for the further working period of the Program after having passed the initialization period of time 2n T.
III. Modelling experiments
In modelling experiments the actual for practical use in ccliu-lar mode SU MIMO 4x4 SM, modulation \6QAM, band
2 GIIz was tested. For each SNR value 20000 numerical modelling experiments were run in accordance with the Program described in Paragraph 2, with parameters: |<5v|/|vfJ =0,01;
|,.oi| = 70[Arm//i]. System model (1) used, search of the antenna
switching vector by using formula (6).
Demodulation after Program cycles was done according to the Maximum Likelihood criterion [1).
The designation SER on the vertical axis of the plot on Figure
3 means Symbol Error Rate (the ratio of wrongly demodulated symbols to all symbols sent), Error correction was not applied.
The energy efficiency curve corresponding to the channel matrix R" with the worst energy efficiency out of all 14 matrices Rl, R2,..., R<4 - is the dotted line on Figure 3. The choice of
number 14 is explained by the worst channel forecasting accuracy for the 14-th symbol for 1-st antenna according to formula (5)). In solid line the curve of comparison base is shown - for SU MIMO SM 4x4 without antenna selection.
Energy efficiency performance gain from using the proposed method related to the standard MIMO (without antenna selection) showed to be 2 dli. Getting this gain is restricted by satisfying conditions CI and C2 imposed on the system on direct travel paths of the vehicle.
Rayleigh fading channel 4t4r, arraysize -82 GHz, [Vesl| - 70 kmfli, |dV|/|Vestj = 0 01 1(T
10
10
<£ LU (C
10"
10
10"
■ The proposed method
■ MIM04x4 w/o antenna selection
J_I_I_L.
6 8 10 12 14 16 18 SNR [dB]
Figure 3
Conclusion
The application of the proposed radio channel characteristics prediction method for 4 out of 8RX antennas selection, provided that the described requirement on the channel are satisfied - will allow for getting the 2 dB energy efficiency performance gain in the M1MO 4x4 Spatial Multiplexing mode, compared to the standard MIMO 4x4 without antenna selection, at the speed of motion of the vehicle with linear antenna array of 70 km/h.
1. Hampton. J R. 2014, 'Introduction lo MIMO com ¡nun ¡canons', UK, Cambridge University Press, 288 p.
2. Molisch. A F& Moe. Z & Win. M 7. & Choi. Y-S <& Winters. J II. July 2005, 'Capacity of MIMO systems with antenna Selection' IEEE Transactions on Wireless Communications, vol. 4, no. 4, pp. 1759-1772.
3. Gorokhov, A t£ Dhananjay, A t£ Gore. D A & Pauiraj. A J, November 2003, 'Receive antenna selection tor MIMO spatial multi-
plexing: theory and algorithms' IEEE Transactions on Signal Processing, vo!. 51, no. 11, pp. 2796-2807.
4. Telatar, IE, November-December 1999, 'Capacity of multiple-antenna Gaussian channels' European Transactions on Télécommunications, vol. 10, no. 6. pp. 585-595.
5. Rose Qingyang Hit. R Q & Qian. >', May 2013, 'Heterogeneous cellular networks' Chichester, West Sussex, United Kingdom: John Wiley & Sons, 378 p.
6. Merz R & Wenger. D & Scanferla. D & Mamon. S, 2014, 'Performance of LTE in a high-velocity environment: A measurement study' in Proceedings of the 4th Workshop on All Things Cellular: Operations, Applications, & Challenges, scr. AllThingsCellular '14. New York, NY, USA: ACM, pp. 47-52.
7. 3GPP TS 36.211 : 'Evolved Universal Terrestrial Radio Access (E-UTRA); Physical channels and modulation'. htip;//www,3gpp.org/, Retrieved: 20.11.2016.
8. Lee W.C.Y., 1997, 'Mobile Communications Engineering.' Ney York: McGraw-Hill, 689 p.
9. Coxeter. H.S.M.. 1969, 'Introduction lo geometry (2nd ed.).' John Wiley and Sons. pp. 216-22).
References
ПРЕДСКАЗАНИЕ ХАРАКТЕРИСТИК КАНАЛА MIMO И АЛГОРИТМ ВЫБОРА АНТЕНН НА ПРИЕМЕ ПРИ ДВИЖЕНИИ ЛИНЕЙНОЙ АНТЕННОЙ РЕШЕТКИ
Старовойтов Михаил Юрьевич, Менеджер, ООО Нокиа Солюшнз энд Нетворкс, Москва, Россия,
mikhail.starovoytov@nokia.com
Аннотация
Режим однопользовательского MIMO с пространственным мультиплексированием позволяет получить для случая фиксированного количества линий передачи и приема информации наивысшую спектральную эффективность в расчете на одну передающую антенну. Исходной архитектурой на приемной (абонентской) стороне является типовая схема с регулярной линейной антенной решеткой, установленной на крыше транспортного средства, и мобильный терминал с фиксированным количеством "приемных цепей" (трактов обработки сигнала). Транспортное средство оснащено бортовой системой с возможностями определения текущей скорости и направления движения. Для стационарных радиоусловий задача оптимального выбора приемных антенн давно решена. Однако для случая быстрого движения антенн приемника, когда изменениями радиосреды за один пакетный интервал (за 1 миллисекунду) нельзя пренебречь и радиосреда существенно нестационарна, выбор подмножества антенн для MIMO в литературе не рассматривался. Причина нам видится в неизученности вопроса предсказания элементов матрицы канала на 1 миллисекунду вперед - при количестве "приемных цепей" меньшем, чем количество антенн в решетке. Предложенный в статье новый метод решает проблему предсказания элементов матрицы канала MIMO с достаточной точностью - в случае выполнения определенных добавочных условий, наложенных на радиосреду в пространстве по ходу движения антенной решетки. При этом данные от бортовой системы о точных значениях текущей скорости и направления движения транспортного средства имеют для применения нового метода ключевое значение. Тема точного определения кинематических параметров движения транспортного средства имеет на сегодня широкое покрытие в научно-технической литературе; ее разработка связана с перспективой внедрения автопилотов в управление всеми тринспортными средствами включая автомобили. Мы в предлагаемом методе находим способ по-новому использовать данные бортовой системы для улучшения параметров работы режима MIMO. Выигрыш в помехоустойчивости от применения предложенного метода с выбором 4-х антенн из 8-ми по отношению к базовому MIMO (без выбора антенн) составляет 2 дБ.
Ключевые слова: MIMO, пространственное мультиплексирование, линейный регулярный антенный массив, выбор антенн, предсказание матрицы канала.
Литература
1. Jerry R.Hampton. Introduction to MIMO Communications, UK, Cambridge University Press, 2014, 288 p.
2. Andreas F. Molisch, Moe Z. Win, Yang-Seok Choi, Jack H. Winters. Fellow "Capacity of MIMO Systems With Antenna Selection" IEEE Transactions on Wireless Communications, vol. 4, no. 4, pp. 1759-1772, July 2005.
3. Alexei Gorokhov, Dhananjay A. Gore, and Arogyaswami J. Paulraj. Receive Antenna Selection for MIMO Spatial Multiplexing: Theory and Algorithms // IEEE Transactions on Signal Processing, vol. 51, no. 11, pp. 2796-2807, November 2003.
4. Emre Telatar. Capacity of multiple-antenna Gaussian channels // European Transactions on Telecommunications, vol. 10, no. 6, pp. 585-595, November-December 1999.
5. Rose Qingyang Hu; Yi Qian. Heterogeneous cellular networks. Chichester, West Sussex, United Kingdom : John Wiley & Sons, 2013.
6. R. Merz, D. Wenger, D. Scanferla, and S. Mauron. Performance of LTE in a high-velocity environment: A measurement study // Proceedings of the 4th Workshop on All Things Cellular: Operations, Applications, & Challenges, ser. AllThingsCellular '14. New York, NY, USA: ACM, 2014, pp. 47-52.
7. 3GPP TS 36.211: "Evolved Universal Terrestrial Radio Access (E-UTRA); Physical channels and modulation". http://www.3gpp.org/. Retrieved: 12.09.2016.
8. Lee W.C.Y. Mobile Communications Engineering. Ney York: McGraw-Hill, 1997.
9. Coxeter, H.S.M. Introduction to geometry (2nd ed.). John Wiley and Sons. pp. 216-221. ISBN 978-0-471-50458-0. Zbl 0181.48101, 1969.
