ul/dl mode selection and transceiver design for dynamic...

2016 Tyrrhenian International Workshop on Digital Communications, Livorno, Italy , 14 Sept, 2016UL/DL Mode Selection and Transceiver Design for Dynamic TDD Systems 1

UL/DL Mode Selection and Transceiver Design for

Dynamic TDD Systems

Antti Tolliwith Ganesh Venkatraman, Jarkko Kaleva and David Gesbert

e-mail: [email protected]

Centre for Wireless Communications, University of Oulu, Finland

2016 Tyrrhenian International Workshop on Digital Communications, Livorno, Italy

14 Sept, 2016

A. Tolli, J. Kaleva, G. Venkatraman & D. Gesbert, ”Joint UL/DL Mode Selection and Transceiver Design for

Dynamic TDD Systems”, in Proc. 2016 IEEE Global Conference on Signal and Information Processing

(GLOBALSIP), Washington, D.C., USA, Dec. 7–9, 2016

c©Antti Tolli, CWC


Dynamic TDDUL and DLcontrolchannels UL or DL data channels

time

freq

uenc

y

Figure: Flexible TDD frame structure1

Significant load

variation between

adjacent cells

Flexible UL/DL

allocation provides

large potential gains

in spectral efficiency2

More challenging

interference

management

1Nokia Networks, ”5G radio access system design aspects”, Nokia white paper, Aug. 2015. Available:

http://networks.nokia.com/file/37611/5g-radio-access

23GPP TSG RAN WG1, ”Study on scenarios and requirements for next generation access technologies TR 38.913,” 3rd

Generation Partnership Project 3GPP, www.3gpp.org, 2016

c©Antti Tolli, CWC

http://networks.nokia.com/file/37611/5g-radio-access


Dynamic TDD

Figure: UL-DL/DL-UL interference in Dynamic TDD

Additional UL-to-DL and DL-to-UL interference associated with the

dynamic TDD

Interference mitigated by coordinated beamforming.

More measurements and info exchange also at the terminal side

Similar interference scenarios in underlay D2D transmission3

3A. Tolli, J. Kaleva & P. Komulainen, ”Mode Selection and Transceiver Design for Rate Maximization in Underlay D2D

MIMO Systems”, in Proc. IEEE ICC 2015, London, UK, June, 2015

c©Antti Tolli, CWC


System Model & Problem Formulation

OFDM system with N

sub-channels and NB

BSs, NT TX antennas

per BS

K users each with NR

antennas

Goal: minimize the number of packets in BS/user queues via joint

uplink (UL) / downlink (DL) cell mode selection, TX/RX design and

resource allocation over spatial and frequency resources

c©Antti Tolli, CWC


Queueing Model

Each user is associated with backlogged packets of size Qk.

Queued DL (UL) packets Qk (Qk) of each user follows dynamic

equation at the ith instant as

Qk(i+ 1) =[Qk(i)− tk(i)

]++ λk(i) (1)

where tk =∑N

n=1

∑Ll=1 tl,k,n denotes the total number of

transmitted packets corresponding to user k

λk represents the fresh arrivals of user k at BS bk

Separate user specific queues for UL and DL traffic

c©Antti Tolli, CWC


Objective

Minimize the total number of backlogged packets in DL and UL4

minimizetk,tk

∑k∈U

αk|vk|q + βk|uk|q (2)

where αk, βk are arbitrary priority weights and

vk = Qk − tk = Qk −∑N

n=1

∑L

l=1log2(1 + γl,k,n) (3)

uk = Qk − tk = Qk −∑N

n=1

∑L

l=1log2(1 + γl,k,n) (4)

q = 1, 2, . . . ,∞ plays different role based on the value it assumesI Inherent maximum rate constraint:

∑Nn=1

∑Ll=1 tl,k,n ≤ Qk

Special cases (when Qk >∑N

n=1

∑Ll=1 tl,k,n ∀ k):

I q = 1: Sum rate maximizationI q = 2: Queue-Weighted Sum Rate Maximization (Q-WSRM)

4G. Venkatraman, A. Tolli, L-N. Tran & M. Juntti, ”Traffic Aware Resource Allocation Schemes for Multi-Cell

MIMO-OFDM Systems”, IEEE Transactions on Signal Processing, vol. 64, no. 11, pp. 2730–2745, June 2016.

c©Antti Tolli, CWC


Spatial Overloading in SINR

DL SINR5

Γl,k,n =

∣∣∣wHl,k,n Hbk,k,n ml,k,n

∣∣∣2N0 +

∑i∈U\{k}

L∑j=1

|wHl,k,nHbi,k,nmj,i,n|2

︸︷︷︸DL-DL interference

+∑

i∈U\Ubk

L∑j=1

|wHl,k,nHi,k,nmj,i,n|2

︸︷︷︸UL-DL interference

(5)

UL SINR

Γl,k,n =

∣∣∣wHl,k,n HT

bk,k,nml,k,n

∣∣∣2N0 +

∑i∈U\{k}

L∑j=1

|wHl,k,nHT

bk,i,nmj,i,n|2

︸︷︷︸UL-UL interference

+∑

i∈U\Ubk

L∑j=1

|wHl,k,nHbi,bk,nmj,i,n|2

︸︷︷︸DL-UL interference

(6)

5Note that UL-DL and DL-UL interference terms in (5), and (6), respectively, include potential interference from all

other-cell users. UL/DL mode selection per BS/user is handled separately via (relaxed) binary selection.

c©Antti Tolli, CWC


Queue Minimization with UL/DL Mode Selection

min. ‖v‖q + ‖u‖q (7a)

s. t. γl,k,n ≤ Γl,k,n ∀ l, k, n (7b)

γl,k,n ≤ Γl,k,n ∀ l, k, n (7c)∑N

n=1

∑k∈Ub

∑L

l=1‖ml,k,n‖2 ≤ xbPmax ∀ b (7d)∑N

n=1

∑L

l=1‖ml,k,n‖2 ≤ xbkP

UEmax ∀ k (7e)

xb + xb = 1 ∀ b, xb ∈ {0, 1}, xb ∈ {0, 1} (7f)

where vk , a1q

k (Qk −∑N

n=1

∑Ll=1 tl,k,n) and tl,k,n = log(1 + γl,k,n)

Nonconvex (difference of convex) SINR constraints, and integer

UL/DL selection constraints

c©Antti Tolli, CWC


Approximation of the SINR Constraints

The DL SINR constraints in (7b) are relaxed as6 (UL similarly)

γl,k,n ≤

∣∣∣wHl,k,n Hbk,k,n ml,k,n

∣∣∣2βl,k,n

=p2l,k,n + q2

l,k,n

βl,k,n

βl,k,n ≥ N0 +∑

i∈U\{k}

∑L

j=1|wH

l,k,nHbi,k,nmj,i,n|2

+∑

i∈U\Ubk

∑L

j=1|wH

l,k,nHi,k,nmj,i,n|2

(8)

(9)

where

pl,k,n , <(wHl,k,n Hbk,k,n ml,k,n), ql,k,n , =(wH

l,k,n Hbk,k,n ml,k,n)

Difference of convex constraint solved via successive convex (linear)

approximation (SCA)6G. Venkatraman, A. Tolli, L-N. Tran & M. Juntti, ”Traffic Aware Resource Allocation Schemes for Multi-Cell

MIMO-OFDM Systems”, IEEE Transactions on Signal Processing, vol. 64, no. 11, pp. 2730–2745, June 2016.

c©Antti Tolli, CWC


Binary Relaxation

Binary variables xb, xb ∈ {0, 1} are replaced by continuous variables

xb, xb ∈ [0, 1], ∀bProblem (7) becomes convex (for fixed receivers, at any given

linearization point of the SINR constraints)

Sparsity must be enforced! → Use a regularization function7

minimize ‖v‖q + ‖u‖q + ψ

NB∑t=1

(log(xb + ε) + log(xb + ε))︸︷︷︸Concave

. (10)

successively linearized as

minimize ‖v‖q + ‖u‖q + ψ

NB∑b=1

(xb − x

(i)b

x(i)b + ε

+xb − x

(i)b

x(i)b + ε

)(11)

7E. J. Candes, M. B Wakin, and S. Boyd, ”Enhancing Sparsity by Reweighted l1 Minimization,” Journal of Fourier

analysis and applications, vol. 14, no. 5-6, pp. 877–905, 2008.

c©Antti Tolli, CWC


Numerical Example

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Arrival rate (A).

0

0.5

1

1.5

2

2.5

Avera

ge q

ueue b

acklo

g (

bits).

50/50 Fixed ( α = 0.4, β = 10)

Relaxed Binary ( α = 0.4, β = 10)

50/50 Fixed ( α = 0.2, β = 10)


50/50 Fixed ( α = 0.2, β = 5)


Figure: Average number of queued bits per user with varying packet arrival rates.

The mean arrival rate across all low and high rate demand users is

(1− α)A+ αβA.

c©Antti Tolli, CWC


Next Steps

Decentralization, decoupling the problem

Inter-carrier, inter-sector UL-DL interference

Signalling, CSI acquisition

Time-scale of changing UL/DL allocation?

Impact of more practical traffic models

c©Antti Tolli, CWC

ul/dl mode selection and transceiver design for dynamic...

Documents