Slides Downlink BF

Multiuser Downlink Beamforming:
Rank-Constrained SDP
Daniel P. Palomar
Hong Kong University of Science and Technology (HKUST)
ELEC5470 - Convex Optimization

Fall 2016-17, HKUST, Hong Kong
Outline of Lecture
• Problem formulation: downlink beamforming
• SOCP formulation, SDP relaxation, and equivalent reformulation
• General framework: rank-constrained SDP
• Numerical results
• Summary
Daniel P. Palomar 1
Problem Formulation:
Downlink Beamforming
Daniel P. Palomar 2
• Consider several multi-antenna base stations serving single-antenna
users:
Daniel P. Palomar 3
Signal Model
• Signal transmitted by the kth base station:
X
xk (t) = wmsm (t)
m∈Ik
where Ik are the users assigned to the kth base station.
• Signal received by mth user:

X
rm (t) = hH
m,κm xκm + hH
m,k xk + nm (t)
k6=κm
where κm is the base station assigned to user m, hm,k is the channel

vector from base station k to user m with correlation matrix Rm,k .
Daniel P. Palomar 4
SINR Constraints
• SINR of the mth user is
H
wm Rmmwm
SINRm = P H 2
.
l6=m wl Rml wl + σm
• We can similarly consider an instantaneous version of the SINR, as

well as the SINR for a dirty-paper coding type of transmission.
• Optimization variables: beamvectors wl for l = 1, . . . , L.
• The SINR constraints are
SINRm ≥ ρm for m = 1, . . . , L.
Daniel P. Palomar 5
Formulation of Optimal Beamforming Problem
• Minimization of the overall transmitted power subject to SINR

constraints:
PL 2
minimize l=1 kw l k
{wl }
H
wm Rmmwm
subject to P H 2 ≥ ρm , m = 1, . . . , L.
l6=m wl Rml wl +σm
• This beamforming optimization is nonconvex!!
Daniel P. Palomar 6
SOCP Formulation
Daniel P. Palomar 7
• Assume rank-one direct channels: Rmm = hmhH
m.
H
• By adding, w.l.o.g., the constraint wm hm ≥ 0, we can write the
beamforming problem as
PL 2
minimize
{wl } l=1 kwl k
P
H
2 H 2
subject to wm hm ≥ ρm w
l6=m l R ml w l + σ m , ∀m
H
wm hm ≥ 0.
• Observe we can write

X 2
H 2 e 1/2
wl Rmlwl + σm = Rm wm

e
l6=m
Daniel P. Palomar 8
where
 
  Rm1
w1 
 .. 

 ..  
0

w
em =  and R
e m = diag  .
  
 wL



 .. 
 
1  RmL 
2
σm
• The beamforming problem can be finally written as the SOCP:

PL 2
minimize
{wl } l=1 kwl k

H e 1/2
subject to wm hm ≥ ρm R w m , ∀m.

m e
Daniel P. Palomar 9
SDP Relaxation
Daniel P. Palomar 10
• Let’s start by rewriting the SINR constraint
H
wm Rmmwm
P H 2
≥ ρm
as
X
H
wm Rmmwm − ρm wlH Rmlwl ≥ ρmσm
2
.
l6=m
• Then, define the rank-one matrix: Xl = wlwlH .
• Notation: observe that
wlH Awl = Tr (AXl) , A • Xl.

• The beamforming optimization can be rewritten as
PL
minimize l=1 I • Xl
X1,...,XL
2
P
subject to Rmm • Xm − ρm l6=m R ml • X l ≥ ρ m σ m, ∀m
Xl 0
rank (Xl) = 1
• This is still a nonconvex problem, but if we remove the rank-one

constraint, it becomes an SDP!
• Thus, a relaxation of the beamforming problem is an SDP which

can be solved in polynomial time.
Equivalent Reformulation
• An equivalent reformulation was obtained in the seminal work:
– M. Bengtsson and B. Ottersten, “Optimal and suboptimal trasmit beamforming,” in Handbook
of Antennas in Wireless Communications, ch. 18, L.C. Godara, Ed., Boca Raton, FL: CRC
Press, Aug. 2001.
– M. Bengtsson and B. Ottersten, ”Optimal transmit beamforming using semidefinite
optimization,” in Proc. of 37th Annual Allerton Conference on Communication, Control,
and Computing, Sept. 1999.
• Theorem: If the SDP relaxation is feasible, then it has at least one

rank-one solution.
• The proof is quite involved as it requires tools from Lagrange duality

and Perron-Frobenius theory on nonnegative matrices.
• Lemma: The following two problems have the same unique optimal
solution λ:
PL 2
minimize l=1 λ l ρ l σ l
λ,{um}
subject to uH
P
m I − λmRmm + l6=m λl ρl Rml ui = 0, ∀m
kumk = 1
λm ≥ 0
and
PL 2
maximize l=1 λ l ρ l σ l
λ P
subject to I − λmRmm + l6=m λl ρl Rml 0, ∀m
λm ≥ 0.
• The next two steps of the proof as the following:
1. show that the dual problem of the beamforming problem is exactly
the maximization in the previous lemma (this is quite simple)
2. show that the beamforming problem can be rewritten as the
minimization in the previous lemma (this is quite challenging and
requires Perron-Frobenius theory).
• At this point, one just needs to realize that what has been proved
is that strong duality holds. This implies that an optimal solution
of the original problem is also an optimal solution of the relaxed
problem; hence, the relaxed problem must have a rank-one solution.
More General Problem Formulation:
Soft-Shaping Interference Constraints
• Constraint to control the amount of interference generated along
some particular direction:
N h i
X 2
E hH

m,k xk (t)
≤ τm for m = L + 1, . . . , M.
k=1
• Defining Sml = hm,κl hH

m,κl , we can rewrite it as
L
X
wlH Smlwl ≤ τm for m = L + 1, . . . , M.
l=1
• We can similarly consider long-term constraints, high rank matrices

Sml, total power transmitted from one particular base station to one
external user.
Individual Shaping Interference Constraints
• Some interference constraints only affect one beamvector and we
call them individual constraints.
• We consider the following two types of constraints:
wlH Blwl = 0 for l∈E

wlH Blwl ≥ 0 for l∈
/ E.
• The following global null shaping constraint is equivalent to individ-

ual constraints (assuming Sml 0):
L
X
wlH Smlwl ≤ 0.
l=1
Formulation of Optimal Beamforming Problem
• Minimization of the overall transmitted power subject to all the

previous SINR and interference constraints:
PL 2
minimize
{wl } l=1 kwl k
H
wm Rmmwm
subject to P
w H R w +σ 2 ≥ ρm , m = 1, . . . , L
ml l m
Pl6L
=m l
H
l=1 l Sml wl ≤ τm,
w m = L + 1, . . . , M
wlH Blwl = (≥) 0, l ∈ (∈)
/ E1
wlH Dlwl = (≥) 0, l ∈ (∈)
/ E2.
• This beamforming optimization is nonconvex!!
SDP Relaxation
• Let’s start by rewriting the SINR constraint

H
wm Rmmwm
P H 2
≥ ρm
as X
H
wm Rmmwm − ρm wlH Rmlwl ≥ ρmσm
2
.
l6=m
• Then, define the rank-one matrix: Xl = wlwlH .
• Notation: observe that
wlH Awl = Tr (AXl) , A • Xl.

• The beamforming optimization can be rewritten as
PL
X1,...,XL
2
P
subject to Rmm • Xm − ρm l6=m Rml • Xl ≥ ρmσm , m = 1, . . . , L
PL
l=1 Sml • Xl ≤ τm, m = L + 1, . . . , M
Bl • Xl = (≥) 0, l ∈ (∈)
/ E1
Dl • Xl = (≥) 0, l ∈ (∈)
/ E2
Xl 0, rank (Xl) = 1 l = 1, . . . , L.
• This is still a nonconvex problem, but if we remove the rank-one

constraint, it becomes an SDP!
• Thus, a relaxation of the beamforming problem is an SDP which

can be solved in polynomial time.
Existing Literature
• M. Bengtsson and B. Ottersten, “Optimal and suboptimal trasmit beamforming,” in Handbook
of Antennas in Wireless Communications, ch. 18, L.C. Godara, Ed., Boca Raton, FL: CRC
Press, Aug. 2001.
• M. Bengtsson and B. Ottersten, ”Optimal transmit beamforming using semidefinite optimization,”

in Proc. of 37th Annual Allerton Conference on Communication, Control, and Computing,
Sept. 1999.
• D. Hammarwall, M. Bengtsson, and B. Ottersten, ”On downlink beamforming with indefinite

shaping constraints,” IEEE Trans. on Signal Processing, vol. 54, no. 9, Sept. 2006.
• G. Scutari, D. P. Palomar, and S. Barbarossa, “Cognitive MIMO Radio,” IEEE Signal Processing
Magazine, vol. 25, no. 6, Nov. 2008.
• Z.-Q. Luo and T.-H. Chang, “SDP relaxation of homogeneous quadratic optimization: approxima-
tion bounds and applications,” Convex Optimization in Signal Processing and Communications,
D.P. Palomar and Y. Eldar, Eds., Cambridge University Press, to appear in 2009.
Curiosity: Multicast Downlink Beamforming
• This small variation was proven to be NP-hard:

– N. D. Sidiropoulos, T. N. Davidson, Z.-Q. (Tom) Luo, “Transmit Beamforming
for Physical Layer Multicasting”, IEEE Trans. on Signal Processing, vol. 54,
no. 6, June 2006.
Rank-Constrained SDP
Separable SDP
• Consider the following separable SDP:

PL
minimize
X1,...,XL l=1 Cl • Xl
PL
subject to l=1 Aml • Xl = bm, m = 1, . . . , M
Xl 0, l = 1, . . . , L.
• If we solve this SDP, we will obtain an optimal solution X?1 , . . . , X?L

with an arbitrary rank profile.
• Can achieve a low-rank optimal solution? (recall that for the

beamforming problem we want a rank-one solution)
Rank-Constrained Solutions
• A general framework for rank-constrained SDP was developed in
– Yongwei Huang and Daniel P. Palomar, “Rank-Constrained Separable Semidef-
inite Programming for Optimal Beamforming Design,” in Proc. IEEE Inter-
national Symposium on Information Theory (ISIT’09), Seoul, Korea, June
28 - July 3, 2009.
– Yongwei Huang and Daniel P. Palomar, “Rank-Constrained Separable Semidef-
inite Programming with Applications to Optimal Beamforming,” IEEE Trans.
on Signal Processing, vol. 58, no. 2, pp. 664-678, Feb. 2010.
• Theorem: Suppose the SDP is solvable and strictly feasible. Then,

it has always an optimal solution such that
L
X
rank2 (Xl) ≤ M.
l=1
Purification-Type Algorithm
Algorithm: Rank-constrained solution procedure for separable SDP
1. Solve the SDP finding X1, . . . , XL with arbitrary rank.

PL
2. while l=1 rank2 (Xl) > M , then
3. decompose Xl = VlVlH and find a nonzero solution {∆l} to
XL
VlH AmlVl • ∆l = 0 m = 1, . . . , M.
l=1
4. choose maximum eigenvalue λmax of ∆1, . . . , ∆L

1
5. compute Xl = Vl I − λmax ∆l VlH for l = 1, . . . , L.
6. end while
A Simple Consequence
√
• Observe that if L = 1, then rank (X) ≤ M (real-valued case by
[Pataki’98]).
• Furthermore, if M ≤ 3, then rank (X) ≤ 1:
minimize C•X
X0
subject to Am • X = bm, m = 1, . . . , M.
• Corollary: The following QCQP has no duality gap:
minimize xH Cx
x
subject to xH Amx = bm, m = 1, 2, 3.
Rank-One Solutions for Beamforming
• Our result allows us to find rank-constrained solutions satisfying

PL 2
l=1 rank (Xl) ≤ M .
• But, what about rank-one solutions for beamforming?
• Recall that to write the beamforming problem as an SDP we used

Xl = wlwlH .
• Corollary: Suppose the SDP is solvable and strictly feasible, and

that any optimal primal solution has no zero matrix component.
Then, if M ≤ L + 2, the SDP has always a rank-one optimal
solution.
• The following conditions guarantee that any optimal solution has no
zero component:
L ≤ M
−Aml 0, ∀l 6= m, m, l = 1, . . . , L
bm > 0, m = 1, . . . , L.
• These conditions are satisfied by the beamforming problem:

PL
X1,...,XL
2
P
subject to Rmm • Xm − ρm l6=m Rml • Xl ≥ ρmσm , m = 1, . . . , L
PL
l=1 Sml • Xl ≤ τm, m = L + 1, . . . , M
Xl 0, l = 1, . . . , L.
Contribution Thus Far
• Original result:
– original seminal 2001 paper by Bengtsson and Ottersten showed
strong duality for the rank-one beamforming problem with M = L
– the proof relied on Perron-Frobenius nonnegative matrix theory
and on the uplink-downlink beamforming duality
– it seemed that strong duality was specific to that scenario.
• Current result:
– applies to a much wide problem formulation of separable SDP
– in the context of beamforming, it allows for external interference
constraints
– proof based purely on optimization concepts (no Perron-Frobenius,
no uplink-downlink duality).
Rank-Constrained SDP with Additional
Individual Constraints
Separable SDP with Individual Constraints
• Consider the following separable SDP with individual constrains:
PL
minimize l=1 Cl • Xl
X1,...,XL
PL
subject to l=1 Aml • Xl = bm, m = 1, . . . , M
Bl • Xl = (≥) 0, l ∈ (∈)
/ E1
Dl • Xl = (≥) 0, l ∈ (∈)
/ E2
Xl 0, l = 1, . . . , L.
• By directly invoking our previous result we get:

L
X
rank2 (Xl) ≤ M + 2L.
l=1
• Can we do better?
Low-Rank Solutions
• Theorem: Suppose the SDP is solvable and strictly feasible. Then,

it has always an optimal solution such that
L
X
rank (Xl) ≤ M.
l=1
• Observe that the main difference is the loss of the square in the
ranks. This implies that solutions will have higher ranks.
• Can we do better?
• We can sharpen the previous result is we assume that some of the
matrices (Bl, Dl) are semidefinite.
• Theorem: Suppose the SDP is solvable and strictly feasible, and

that (Bl, Dl) for l = 1, . . . , L0 are semidefinite. Then, the SDP has
always an optimal solution such that
0
L
X L
X
rank2 (Xl) + rank (Xl) ≤ M.
l=1 l=L0+1
• Observation: individual constraints defined by semidefinite matrices

do not contribute to the loss of the square in the rank.
A Simple Consequence
• Corollary: The following QCQP has no duality gap (assuming B

and D semidefinite):
minimize xH Cx
x
subject to xH Amx = (≥) bm, m = 1, 2, 3
xH Bx = (≥) 0
xH Dx = (≥) 0.
Rank-One Solutions for Beamforming
• What about rank-one solutions for beamforming?
• Recall that to write the beamforming problem as an SDP we used

Xl = wlwlH .
• Corollary: Suppose the SDP is solvable and strictly feasible, and

that any optimal primal solution has no zero matrix component.
Then, the SDP has a rank-one solution if one of the following hold:
– M =L
– M ≤ L + 2 and (Bl, Dl) are semidefinite for l = 1, . . . , L.
Numerical Simulations
Simulation Setup
• Scenario: single base station with K = 8 antennas serving 3 users

at −5◦, 10◦, and 25◦ with an angular spread of σθ = 2◦.
• SINR constraints set to: SINRl ≥ 1 for l = 1, 2, 3.
• External users from another coexisting wireless system at 30◦ and

50◦ with no angular spread.
• Soft-shaping interference constraints for external users set to: τ1 =

0.001 and τ2 = 0.0001.
Simulation #1
• With no soft-shaping interference constraints (15.36 dBm):
Radiation pattern at the transmitter
−15
−20
−25
dBW
−30
−35
−40
−45
−90 −60 −30 0 30 60 90

azimut in degrees
Simulation #2
• With soft-shaping interference constraints (18.54 dBm):
−10
−15
−20
−25
dBW
−30
−35
−40
−45
−50
−90 −60 −30 0 30 60 90

azimut in degrees
Simulation #3
• With soft- and null-shaping interference constraints (at −20◦, 30◦
and 50◦, 70◦) (20.88 dBm):
−10
−20
−30
dBW
−40
−50
−60
−70
−90 −60 −30 0 30 60 90
azimut in degrees
Simulation #4
• With soft- and null-shaping interference constraints (at −20◦, 50◦, and 70◦ with
τ1 = 0.00001, τ2 = 0, τ3 = 0.000001) as well as a derivative null constraint at
70◦ (15.66 dBm):
−20
−30
−40
dBW
−50
−60
−70
−80
−90 −60 −30 0 30 60 90
azimut in degrees
Summary
• We have considered the optimal downlink beamforming problem

that minimizes the transmission power subject to:
– SINR constraints for users within the system
– soft-shaping interference constraints to protect users from coex-
isting systems
– individual null-shaping interference constraints
– derivative interference constraints.
• The problem belongs to the class of rank-constrained separable SDP.
• We have proposed conditions under which strong duality holds as

well as rank reduction procedures.

Slides Downlink BF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Slides Downlink BF

Uploaded by

Copyright:

Available Formats

Multiuser Downlink Beamforming:

ELEC5470 - Convex Optimization

• Problem formulation: downlink beamforming

• SOCP formulation, SDP relaxation, and equivalent reformulation

• General framework: rank-constrained SDP

where Ik are the users assigned to the kth base station.

• Signal received by mth user:

where κm is the base station assigned to user m, hm,k is the channel

• We can similarly consider an instantaneous version of the SINR, as

• Optimization variables: beamvectors wl for l = 1, . . . , L.

• The SINR constraints are

• Minimization of the overall transmitted power subject to SINR

• This beamforming optimization is nonconvex!!

• Observe we can write

• The beamforming problem can be finally written as the SOCP:

• Then, define the rank-one matrix: Xl = wlwlH .

• Notation: observe that

wlH Awl = Tr (AXl) , A • Xl.

• This is still a nonconvex problem, but if we remove the rank-one

• Thus, a relaxation of the beamforming problem is an SDP which

• Theorem: If the SDP relaxation is feasible, then it has at least one

• The proof is quite involved as it requires tools from Lagrange duality

• Defining Sml = hm,κl hH

• We can similarly consider long-term constraints, high rank matrices

• We consider the following two types of constraints:

wlH Blwl = 0 for l∈E

• The following global null shaping constraint is equivalent to individ-

• Minimization of the overall transmitted power subject to all the

• This beamforming optimization is nonconvex!!

• Let’s start by rewriting the SINR constraint

• Then, define the rank-one matrix: Xl = wlwlH .

• Notation: observe that

wlH Awl = Tr (AXl) , A • Xl.

• This is still a nonconvex problem, but if we remove the rank-one

• Thus, a relaxation of the beamforming problem is an SDP which

• M. Bengtsson and B. Ottersten, ”Optimal transmit beamforming using semidefinite optimization,”

• D. Hammarwall, M. Bengtsson, and B. Ottersten, ”On downlink beamforming with indefinite

• This small variation was proven to be NP-hard:

• Consider the following separable SDP:

• If we solve this SDP, we will obtain an optimal solution X?1 , . . . , X?L

• Can achieve a low-rank optimal solution? (recall that for the

• Theorem: Suppose the SDP is solvable and strictly feasible. Then,

1. Solve the SDP finding X1, . . . , XL with arbitrary rank.

4. choose maximum eigenvalue λmax of ∆1, . . . , ∆L

• Furthermore, if M ≤ 3, then rank (X) ≤ 1:

• Corollary: The following QCQP has no duality gap:

• Our result allows us to find rank-constrained solutions satisfying

• But, what about rank-one solutions for beamforming?

• Recall that to write the beamforming problem as an SDP we used

• Corollary: Suppose the SDP is solvable and strictly feasible, and

• These conditions are satisfied by the beamforming problem:

• By directly invoking our previous result we get:

• Theorem: Suppose the SDP is solvable and strictly feasible. Then,

• Theorem: Suppose the SDP is solvable and strictly feasible, and

• Observation: individual constraints defined by semidefinite matrices

• Corollary: The following QCQP has no duality gap (assuming B

• What about rank-one solutions for beamforming?

• Recall that to write the beamforming problem as an SDP we used

• Corollary: Suppose the SDP is solvable and strictly feasible, and

• Scenario: single base station with K = 8 antennas serving 3 users

• SINR constraints set to: SINRl ≥ 1 for l = 1, 2, 3.

• External users from another coexisting wireless system at 30◦ and