DESCRIPTION
Title of Invention
METHOD FOR IMPROVING TRANSMISSION CAPACITY IN A DL MU-MIMO
COMMUNICATIONS SYSTEM
Technical Field
[0001]
The present invention relates methods for improving transmission capacity in a
communications system. In particular although not exclusively the present invention relates to
methods for enhancing transmission capacity in MU-MIMO based Communication Systems.
Background Art
[0002]
MIMO technology has attracted attention in wireless communicatiofis as it offers
significant increase in data throughput and link range without the need for additional bandwidth
or increased transmit power. This increase in throughput is achieved by spreading the same
total transmit power over the antennas to achieve an array gain that improves the spectral
efficiency (more bits per second per hertz of bandwidth) or to achieve a diversity gain that
improves the link reliability (reduced fading). Because of these properties, MGMO is an
important part of modern wireless communication standards such as IEEE 802. 1In (WiFi),
HSPA+, 4G, 3GPP Long Term Evolution (LTE), LTE-Advanced (LTE-A), WiMAX, etc.
[0003]
Multi-user MIMO or MU-MIMO is an enhanced form of MIMO technology that is
gaining acceptance. MU-MIMO enables multiple independent radio terminals to access a
system enhancing the communication capabilities of each individual terminal. MU-MIMO
exploits the maximum system capacity by scheduling multiple users to be able to simultaneously
access the same channel using the spatial degrees of freedom offered by MIMO.
Summary of Invention
Technical Problem
[0004]
A general MU-MIMO system 100 is shown in Fig. 1. As shown the transmitter 101
transmits data to different receivers 102, 103 on the same time-frequency from multiple transmit
antennas. To minimise interference between receivers 102, 103, the transmitter creates
transmission beams through pre-coding. At the receiving site, the receivers 102, 103 use
post-coding (decoding) to take its data. Pre-coding is very much depended on the channel
status. Mathematically, a MU-MIMO system is described as follows:
y ( = H( V( x( + å H( )V(* *) + n( ) (1)
=\ J ¹i ,. .
where: y(z) is the received signal at the i -th user, x(/) is the data signal for the i -th user,
H( ) is the channel matrix of the i -th user, V(/) is the precoder matrix of the -th user and
n(/) is the additive white Gaussian noise at the i -th user.
[0005]
Fig. 2 shows one possible transmission mechanism 200 utilised by the transmitter of Fig.
1 to transmit data 201 to different receivers. To minimise inter-user interference, the receivers
feedback their channel status information CSI 202 (which includes Precoder Matrix Indicator
PMI) to the transmitter. In a system having 2-stage codebook of PMI, the /-th receiver reports 2
PMIs: PMI#1 and PMI#2 denoted by W(/)i and W(/) 2. PMI#1 represents the long term or
wideband channel and PMI#2 the short term or instant channel.
[0006]
The transmitter then uses the reported PMIs to generate the precoder for the /-th receiver
as:
V(/) = W(/) W(/) 2 (2)
[0007]
While this form of pre-coding is effective it is not optimal. Additionally the use of
such direct pre-coding can fail in some cases, particular where the receivers are too close to each
other.
[0008]
Clearly it would be to provide a method for pre-coding which mitigates the likelihood of
failures. It would also be advantageous to provide a method for pre-coding which improves
transmission capacity between transmitters and receivers.
Solution to Problem
[0009]
Accordingly in one aspect of the present invention there is provided a method for
selecting User Equipment (UE) for scheduling/precoding within the coverage area of a
communications node having a predefined codebook said method including the steps of:
receiving at the communications node feedback information from each of the UEs
within the coverage area of the communications node wherein the feedback information includes
at least one precoder matrix indicator (PMI);
generating at the communications node a precoder matrix based on the reported
precoder matrix indicator from each UE;
determining at the communications node correlation values for each UE in the coverage
utilising the precoder matrix;
identifying a pair of UEs having minimum correlation values; and
selecting the identified UEs for scheduling/precoding if the minimum correlation value
is less than or equal to a threshold value.
[0010]
Suitably the predefined codebook is a 2-stage codebook and the feedback information
includes a first precoder matrix indicator PMIt and second precoder matrix indicator PMI2. In
such cases the procoder matrix W(i) for reported RMI Ϊ and PMI2 from each UE may be
generated by W( ) = W( ) i * W(/) 2 where z 1,..., F .
[0011]
Preferably the correlation values are given by
0-,;) = {[ ( ) ( ) [ *(i )W( )f where i = I,...,F -l, j = + 1,...,F . The minimum
correlation values may be determined by Ccorr (i , j ) = in{Ccorr ( , j )}, for
( ,7)= argmin{C ( , )}-
[0012]
In yet another aspect of the present invention there is provided a method of precoding
data transmissions in a communications node servicing User Equipment within the node's
coverage area the communications node having a predefined codebook servicing said method
including the steps of:
selecting a User Equipment (UE) from within the coverage area as candidates for
precoding wherein the selection of the UEs is based on correlation values calculated for each UE
within the coverage area utilising a precoder matrix generated from precoder matrix indicators
reported to the communications node by each UE;
determining correlation values for the selected UEs based on the reported precoder
matrix indicator and Channel Matrix obtained from a fixed codebook of representative channel
matrices;
selecting a Channel Matrix pair having the highest correlation values; and
generating precoders utilising the selected Channel Matrix pair.
[0013]
Suitably the predefined codebook is a 2-stage codebook and the feedback information
includes a first precoder matrix indicator PMIt and second precoder matrix indicator PMI . In
such cases the procoder matrix W(i) for reported PMIi and PMI from each UE may be
generated by W(/) = ( ), x W(/) where = 1,...,F .
[0014]
Preferably the correlation values are given by
where i =I,...,F l = i + 1,...,F . The minimum
correlation values may be determined by C o r (7, j ) - min{C rr (i, j)}, for
(7, J ) = arg v {C rr ( , j)}.
[0015]
The fixed codebook of representative channel matrices may be generated from the long
term PMI codebook and short term rank PMI codebook. Suitably the fixed codebook for
Rankl, W of representative channel matrices is generated from the long term PMI
codebook and short term rank#l PMI codebook. Preferably W , contains vectors of
size NTX x 1. Suitably Rank 2 the fixed codebook W of representative channel matrices
(CM) is generated from the long term PMI codebook and short term rank#2 PMI codebook.
RANK2 contains matrices of size N x .
[0016]
The fixed code books W and W may be utilised to identify the
representative channels matrices. For Rank 2 the representative channel matrices may be given
by
, H(«(7 j MC (6) with
n(7) = argma fr H ( ) (7 ) [H(«)W (7 )
n
n ) = argmax/r («)W ) [H( )W( )
n
[0017]
For Rankl a search may be conducted over for each of the -th UE, from the
PMI based matrix W(z') for vectors of size NTX x 1. These vectors may then utilised to
form the channel matrix H( ) of size x NT .
[0018]
Suitably the correlation values for the selected UEs are calculated by
C i,n) = r {[h( ) W( ) [ (n) ( ) n=1,...,Nvec
[0019]
Preferably the step of selecting a Channel Matrix having the highest correlation includes
sorting the correlation values C(z, n, ) > C(i, n ) > ...> C(i, nN ) to identify the largest
values corresponding to h(n,),h (n ),...,h(« ,w ) to form the channel matrix
H( = [h(/¾),h(« ),...,h ( OT ) .
[0020]
The step of generating the precoders in one embodiment of the invention may include
the steps of
a) initializing for all UEs the UE's post coder by setting G( (m=0) = J ( =
U where
J ( is Rl N
b) calculating the precoder V( )( + ) for all UEs using G(z) ;
c) calculating post coder G(z)(m+1) for all UEs using V( )(m+ ;
d) calculating E - G( )(m) || and comparing E to a convergence threshold e
e) setting m=m+l if E > e and repeat steps b) to d) until ||G(z)(m+I) - G(z)( ) | < e ; and
f outputting the precoder V(z')(m+1) .
[0021]
Suitably V(z)(m+I is calculated in accordance
withV(z) H ( z)G (z) where u is the Lagrange
multiplier and G(z)( + is calculated in accordance
withG(z) where N is the noise variance
[0022]
In some embodiments of the invention the Lagrange multiplier u may be calculated
according to the following steps:
a) calculating singular values l of the decomposition
VAVH=å H H(z)m+1 G (z) + G (0(m+ H(/) M+ ;
b) setting minimum u ) and maximum ( ) values of the Lagrange multiplier u based on
the calculated singular values l ;
c) setting the Lagrange multiplier as u = ( + )/2 ;
l-
d) calculating = ;
c) calculating \p - p \ where P is the total transmit power;
d) comparing P - P with the convergence threshold e ;
e) setting m = if \P - P \ is greater than e and if P is less than P;
f setting m = if P - P is greater than e and if P is greater than P;
g) repeating steps c) to f until \P~P\ T then the
(i ,j) pair are not selected as pair for scheduling/precoding and the process is terminated 303.
If the minimum correlation value is less than the correlation threshold T i.e.,Ccorr ( ,j) £ T
the correlation values of the reported PMI and CM are then calculated 304.
[0034]
The representative channel matrices (CM) in this instance are obtained utilising the
fixed codebook of representative channels 305. The fixed codebook of representative channels
differs for the rank. For Rank 1 the fixed codebook W , of representative channel matrices is
generated from the long term PMI codebook and short term rank#l PMI codebook.
contains vectors of size NT x 1. For Rank 2 the fixed codebook W NK of representative
channel matrices (CM) is generated from the long term PMI codebook and short term rank#2
PMI codebook. i R K 2 contains matrices of size NTX x N .
[0035]
The fixed code books W and W are utilised to identify the representative
channels matrices. For Rank 2 the representative channel matrices is given by
H(»( )) ft fl H(»(7)) J fl (6) with
n ) = argma r (n)W( [H(«)W(7)
n
n J ) = argmaxfr {H(H)W(7)]" [H( )W(7 )
n
i .
[0036]
For Rankl a search is conducted over W , for each of the i -th UE, from the PMI
- based matrix W(z) for vectors of size NTX x 1. These vectors are then utilised to form the
channel matrix H(z') of size x NTX .
[0037]
Once the representative channel matrices are identified the correlation values are then
calculated in accordance with the following
C(/ =frJh(») F W( [h(«) W( ) n=l,...,Nvec (7
[0038]
The correlation values are then sorted to find the largest correlation values
C(i,n ) >C(i,n2) >... >C(i,nNR ) corresponding to h (n, ),h(« ),..., ) to form a channel
matrix
H( = [h(/¾) h(¾),....h(/i ,itt )]tf (8)
[0039]
The Channel Matrix pairs having the maximum correlation values are then selected 306
and the precoders then calculated 307 utilising values for N (i ), N (j ) are then calculated
using CQI(i ,/), CQI{j ,1) (discussed in greater detail below). Then utilising the values for
N (i ), N (J) and H(«(/ )), H(n(j )) with Lagrange multiplier (discussed in greater detail
below) values for the precoders V(i ), V( ) computed.
[0040]
In this instance the resultant precoders are Minimum Mean Squared Error (MMSE)
pecoders which are computed based on the PMI feedback. Consequently information on the
channels is not required to produce the precoders. A more detailed discussion of the generation
of the MMSE precoders in accordance with an embodiment of the present invention is discussed
with respect to Fig. 4 below.
[0041]
With reference to Fig. 4 there is illustrated one process for generating the MMSE
precoders according to one embodiment of the present invention. The precoder is generated by
firstly initializing the post coder for all UEs 401 as follows:
[0042]
Here J( z) is the matrix RI N with the ,b) -th element being zero for a ¹ b
and being 1 for a=b and m) denotes the m-th iteration. The precoder V( )(m+1) is then
computed for all UEs 402 using G( ) for i = 1,..., . and the following equation:
[0043]
In equation 9 the variable u is the Lagrange multiplier obtained form step 505 in Fig.
5 which is discussed in greater detail below.
[0044]
The process then proceeds to compute for all UEs the post coder G(/) + 1 403 using
V( )( +1 for i=\,...,NUE in accordance with the following equation.
G(i) = V ( )H +N ( i (10)
[0045]
For less complexity, the post coder G( ) + for each UE can be calculated using the
following:
G( = V H (i ( [H ( )V( V (OH W(0 +N ( 1
[0046]
In the case of equations 10 and 11 N ( z)is the noise variance obtained form step 602 in
Fig. 6. The computation of the noise variance is discussed in greater detail below.
[0047]
The process then computes E = G( ) +i) - G( (m) | 404, here denotes
Frobenius norm. The process then determines if E is greater than or equal to the convergent
threshold e i.e. E ³ e . If E ³ e then the process increments m 405 and repeats the
calculations for V( )( + and G( + ) per steps 402, 403 to again calculate the value of E.
This process is then repeated until G(i) (m+1) - G( )( I at which stage the precoder
(=1
V(0 + is outputted 406.
[0048]
As can be seen form the above discussion the calculation of the precoder V(z)(m+I)
requires the use of a Lagrange multiplier u . The process of computing the Lagrange multiplier
u 500 is shown in Fig. 5. As shown singular values l of the decomposition
VAV = i) + G (i) + G i + (i m+l) are computed 501. The minimum and
maximum value of the Lagrange multiplier u are then set 502. The Lagrange
multiplier is then set 503 as u =(v x + )/2 .
[0049]
Once the Lagrange multiplier has been set the quantity is then
calculated 504. The process then proceeds to calculate the absolute value P less the total
transmit power P which is then compared with the convergence threshold e. If the absolute
value of is less than e i.e. LP - P ³ e the value of P is compared with the value of P, if P
is less than P then x = 506 before resetting the Lagrange multiplier per step 503. If is
greater than P then =u 507 before resetting the Lagrange multiplier per step 503.
[0050]
The steps of setting the Lagrange multiplier 503, calculation of P 504 and setting
ma
~ 506 or = u are repeated until P- P C{i,n ) >... >C(i,nN ) to
identify the ) to form the channel
matrix H(i) = [h(« ),h(« 2),...,h(« )]i .
[Claim 14]
The method of claim 3 wherein the step of generating the precoders includes the steps
of:
a) initializing for all UEs the UE's post coder by setting G( )(m=0) = J(i), i = 1,...,NUE where
J( is RI N
b) calculating the precoder V( )(m+1) for all UEs using G( ) ;
c) calculating post coder G( )(m+1) for all UEs using V( (m+ ;
d) calculating E=2 G(z)( + -G(i) (m I and comparing E to a convergence threshold e .
e) setting m=m+l if E > e and repeat steps b) to d) until å ( ) ,+ - G( )
f ) outputting the precoder V( V 1)
[Claim 15]
The method of claim 14 wherein V(z)( + is calculated in accordance
withV(z) = j j +v Hf (i)G f (i) where u is the Lagrange
multiplier and G(/)(m++ ) is calculated in accordance with
N,
G( =V ( )H H( )V( ) f ( + N ( where Nois the noise variance
[Claim 16]
The method of claim 15 wherein the calculating the Lagrange multiplier includes
the steps of:
a) calculating singular values l of the decomposition
H ( )m+ G ) G .m+l)
b) setting minimum ( ) and maximum ( i
a ) values of the Lagrange multiplier u based on
the calculated singular values ;
c) setting the Lagrange multiplier as u = ( + )/ 2 ;
d) calculating -
c) calculating P - P \ where P is the total transmit power;
d) comparing P - P with the convergence threshold e;
e) setting u, if is greater than e and if P is less than P;
f) setting = if P - P is greater than e and if P is greater than P;
g) repeating steps c) to f) until