Multi-user downlink linear MIMO precoding systems
View Patent ↗Multi-user (MU-) MIMO systems with quantized feedback are designed to maximize the sum-rate via scheduling and linear precoding. To maximize throughput over the network, quantized CSIT is sent through a low-rate feedback link feedback from a plurality of users back to a base station. The base station then determines a subset of the plurality of users to transmit one or more signals to based on the received feedback and determines a preceding matrix based on the received feedback from the plurality of users wherein the precoding matrix maximizes a sum-rate throughput for the subset of the plurality of users. Additionally, based on the received feedback, the base station designs a quantization codebook. This codebook may be designed off-line and/or online. The codebook and/or precoding matrix are used to transmit signals to the users.
1. A method of operation of a downlink multi-user MIMO system comprising:
receiving quantized feedback from a plurality of users;
determining a subset of the plurality of users to transmit one or more signals to based on the received feedback; and
determining a precoding matrix based on the received feedback from the plurality of users wherein the precoding matrix maximizes a sum-rate throughput for the subset of the plurality of users, said
determining the precoding matrix with an iterative algorithm comprising:
(a) initializing a precoding matrix G (0) ;
(b) determining a precoding matrix G (k) at a k th iteration;
(c) determining G (k+1) E HεD [(H*ψ*AψH) −1 H*ψ*(1+i)A]| G (k) wherein H is a channel matrix of transmission channels to the plurality of users, E is an average power of the channel matrix, D is a generic Voronoi region,
A
=
diag
(
1
A
1
,
1
A
2
,
…
,
1
A
L
)
,
H=[H 1 T H 2 T . . . H L T ] T , ψ=diag(T 1 , T 2 , . . . , T L ), i =diag(T 1 H 1 g 1 , T 2 H 2 g 2 , . . . , T 2 H L g L ), and
T
l
=
(
H
l
g
l
)
*
∑
l
-
1
,
1
≤
l
≤
L
iteratively repeating steps (b) and (c) until G converges.
2. The method of claim 1 further comprising:
scheduling transmission to the plurality of users using the precoding matrix.
3. The method of claim 2 further comprising:
transmitting one or more signals over a plurality of channels to the plurality of users according to the transmission schedule.
4. The method of claim 1 further comprising:
determining a transmission signal based at least in part on the determined precoding matrix.
5. The method of claim 4 wherein determining the transmission signal comprises determining the result of x=Gu wherein:
x is the transmission signal;
G is the determined precoding matrix; and
u is a data stream to be transmitted.
6. The method of claim 1 wherein the precoding matrix comprises:
determining the precoding matrix with a gradient-descent search algorithm.
7. The method of claim 1 wherein the precoding matrix is determined off-line and further comprising:
determining from a look-up table a precoder rank at a base station based at least in part on an average channel condition and a transmit signal-to-noise ratio that results in the highest average sum-rate performance.
8. The method of claim 1 wherein scheduling transmission comprises:
determining a subset K of the plurality of users which minimizes
ϕ
(
V
_
^
)
=
∑
i
=
1
L
∑
j
=
1
,
j
≠
i
L
v
^
i
*
v
^
j
2
for all matrices {circumflex over (V)} =[{circumflex over (v)} 1 , {circumflex over (v)} 2 , . . . , {circumflex over (v)} L ] T composed of L user streams out of KN quantized eigenvectors.
9. A method of operation of a downlink multi-user MIMO system comprising:
receiving quantized feedback from a plurality of users;
determining a subset of the plurality of users to transmit one or more signals to based on the received feedback; and
determining a precoding matrix based on the received feedback from the plurality of users wherein the precoding matrix maximizes a sum-rate throughput for the subset of the plurality of users, said precoding matrix being determined with a gradient descent search algorithm comprising:
determining an (l,j) th element of the gradient of an objective function given by
∂
J
∂
g
lj
=
E
H
∈
D
[
2
A
l
(
H
l
g
l
)
*
∑
l
-
1
h
j
(
l
)
+
∑
k
=
1
,
k
≠
l
L
2
A
k
(
H
k
g
k
)
*
[
-
∑
k
-
1
h
j
(
k
)
(
H
k
g
l
)
*
∑
k
-
1
]
(
H
k
g
k
)
]
where
∑
k
=
I
+
∑
i
=
1
,
i
≠
k
L
(
H
k
g
i
)
(
H
k
g
i
)
*
,
A
k
=
1
+
(
H
k
g
k
)
*
∑
k
-
1
(
H
k
g
k
)
,
h
j
(
k
)
is
a
j
th
row if H k and g lj is the j th element of the column vector g l. ;
determining ∇ G J=E HεD [H*ψ*A+H*ψ*i A−H*ψ*AψHG] where
A
=
diag
(
1
A
1
,
1
A
2
,
…
,
1
A
L
)
,
H=[H 1 T H 2 T . . . H L T ] T , ω=diag(T 1 , T 2 , . . . , T L ), i =diag(T 1 H 1 g 1 , T 2 H 2 g 2 , . . . , T 2 H L g L ), and
T
l
=
(
H
l
g
l
)
*
∑
l
-
1
,
1
≤
l
≤
L
determining the precoding matrix with a gradient search algorithm using a recursion G (k+1) =G (k) −μ k ∇ G J(G (k) ).
10. The method of claim 9 further comprising:
normalizing G(k+1) with
G
(
k
+
1
)
=
P
G
~
(
k
+
1
)
tr
(
G
~
(
k
+
1
)
G
~
(
k
+
1
)
*
)
to satisfy a power constraint.
11. The method of claim 10 , wherein the power constraint is
Σ {circumflex over (V)} Pr ( HεD ( {circumflex over (V)} )) tr ( G ( {circumflex over (V)} ) G ( {circumflex over (V)} )*)≦ P.
12. A method of data transmission in a downlink MIMO system comprising:
receiving quantized feedback from one or more users;
selecting one or more precodes from a precoding matrix generated based on the received feedback wherein the precoding matrix maximizes a sum-rate throughput transmission rate for a subset of the one or more users; and
determining a subset K of the plurality of users which minimizes
ϕ
(
V
_
^
)
=
∑
i
=
1
L
∑
j
=
1
,
j
≠
i
L
v
^
i
*
v
^
j
2
for all matrices {circumflex over (V)} =[{circumflex over (v)} 1 , {circumflex over (v)} 2 , . . . , {circumflex over (v)} L ] T composed of L user streams out of KN quantized eigenvectors.
13. The method of claim 12 further comprising:
determining a transmission schedule for the subset K of the one or more users; and
transmitting one or more signals to the subset K of the one or more users according to the transmission schedule.