1 © 2012 InterDigital, Inc. All rights reserved. © 2012 InterDigital, Inc. All rights reserved. 

Y. Zhao, B. Pelletier,  

P. Marinier and D. Pani 

Globecom 2013 

D2D NEIGHBOR DISCOVERY 

INTERFERENCE 

MANAGEMENT FOR LTE 

SYSTEMS 

2 © 2012 InterDigital, Inc. All rights reserved. 

Outline 

•  Background 

•  System Model 

•  Interference Management Techniques 

•  Methodology 

•  Experimental results 

•  Summary and Conclusion 

3 © 2012 InterDigital, Inc. All rights reserved. 

Background 

•  Proximity detection is motivated by the desire to provide 

new services to mobile devices targetting: 

•  Social applications 

•  Advertisement 

•  Local-based services 

•  Current cellular-based systems such LTE do not have 

mechanisms for precise location information 

•  Precise location systems rely on combination of cellular, 

GPS and WiFi 

•  This approach is inefficient from a battery and resource 

perspective 

•  In D2D discovery, two or more devices determine their 

proximity based on direct radio communications 

•  Implemented in LTE, D2D discovery has the potential to 

provide energy efficient, low-cost ubiquitous proximity 

information and further enable D2D communications 

4 © 2012 InterDigital, Inc. All rights reserved. 

System Model  LTE Resource Allocation 

•  Uplink frequency is used for 

transmission of D2D discovery 

signal 

•  Set of UL resources are reserved 

for D2D discovery 

•  Neighbor eNBs use the same 

resources and are assumed 

coarsely synchronized 

•  Individual discovery signals 

occupy NPRB resource blocks 

•  Discovery signals are transmitted 

over specific discovery resource 

units Ri 

5 © 2012 InterDigital, Inc. All rights reserved. 

System Model  Signal Design 

•  Assume a signal-based discovery carrying no payload 

•  Signal based on Zadoff-Chu (ZC) sequences of a specific length ​𝑁↓𝑧𝑐    

•  Discovery signals ​𝑥↓𝑘 (𝑛+𝜎) is characterized by root (k) and cyclic 

shift (𝜎)  combination, with correlation property ​𝜌↓𝑘,𝑞 (𝜎) such that: ​𝜌↓𝑘,𝑞 (𝜎)={█■∑𝑛=0↑​𝑁↓𝑧𝑐 −1▒​𝑥↓𝑘 (𝑛)​𝑥↓𝑞↑∗ (𝑛+𝜎)=  𝛿(𝜎)                      𝑘=𝑞  ⁠∑𝑛=0↑​𝑁↓𝑧𝑐 −1▒​𝑥↓𝑘 (𝑛)​𝑥↓𝑞↑∗ (𝑛+𝜎)=   ​1/√⁠​𝑁↓𝑧𝑐                        𝑘≠𝑞   

•  Interference between discovery signals caused by: 

1.  Signals with same ZC roots and same cyclic shift (collision) 

2.  Signals with different ZC roots, regardless of cyclic shifts 

•  Up to ​𝑁↓𝑧𝑐  cyclic shifts can be allocated for each ZC root 

6 © 2012 InterDigital, Inc. All rights reserved. 

System Model  Interference Model 

•  In practice a limited number of discovery resource units are 

allocated 

•  Baseline assumption: each eNB assigns a different set of root 

sequences 

•  Inter-cell interference arises from all D2D UEs transmitting in the 

other cell over the same discovery resource unit 

•  Intra-cell interference arises from UEs allocated with a different 

root sequence (due to e.g. running out of cyclic shifts for a given 

root) also transmitting in the same discovery resource unit 

7 © 2012 InterDigital, Inc. All rights reserved. 

Interference Management Techniques – Option 1 

•  Each eNB assigns the 

discovery resources and uses 

a different pool of root 

sequences 

•  All cyclic shifts of a given root 

sequence are allocated before 

a new discovery resource unit 

is allocated 

•  All discovery resource units 

are allocated before a second 

root sequence is allocated on 

an already allocated discovery 

resource unit 

8 © 2012 InterDigital, Inc. All rights reserved. 

Interference Management Techniques – Option 2 

•  Similar to Option 1:  

•  Each eNB assigns the discovery 

resources and uses a different 

pool of root sequences 

•  All discovery resource units are 

allocated before a second root 

sequence is allocated on an 

already allocated discovery 

resource unit 

•  Discovery resource unit are 

assigned in alternance to 

randomize interference in 

frequency 

9 © 2012 InterDigital, Inc. All rights reserved. 

Interference Management Techniques – Option 3 

•  Option 3 is motivated by 

the observation that the 

cell-edge UEs suffer more 

from inter-cell interference 

than cell-center UEs 

•  Discovery resource units 

are seggregated between 

cell-edge UEs and cell-

center UEs 

•  Resources are assumed to 

be assigned by a 

centralized controller 

10 © 2012 InterDigital, Inc. All rights reserved. 

Methodology 

The experiment consists of a set of system-level simulations where: 

•  ​𝑁↓𝑇𝑋  transmitting and ​𝑁↓𝑅𝑋   monitoring UEs are dropped uniformly 

in each sector 

•  The path loss between each transmit-receive UE pair (𝑙,𝑚)  is 

calculated ( ​𝑃𝐿↓𝑙,𝑚 ) 

•  The resource allocation according to one of the option is carried out 

•  The SINR at receiver 𝑙 for a given target transmit UE 𝑚 with TX power ​𝑃↓𝑇𝑋,𝑚  is calculated as follows: ​𝑆𝐼𝑁𝑅↓𝑙,𝑚 = ​​​𝑃↓𝑇𝑋,𝑚 ∕​𝑃𝐿↓𝑙,𝑚  /​𝑃↓𝐼,𝑙 + ​𝜎↓𝑤↑2   

where ​𝜎↓𝑤↑2  is the thermal noise power and ​𝑃↓𝐼,𝑙  is the interference 

term derived from the discovery signal properties as follows: ​𝑃↓𝐼,𝑙 =∑𝑚∈𝒰(𝑙)↑▒​𝜌↓ℛ(𝑚),ℛ(𝑙) (𝒞(𝑚)−𝒞(𝑙)) ​​𝑃↓𝑇𝑋,𝑚 /​𝑃𝐿↓𝑙,𝑚    𝒞(𝑚) and ℛ(𝑚) are the cyclic shifts and root sequences of UE 𝑚, 

respectively. 

•  Statistics are accumulated over multiple drops to emulate UE mobility 

11 © 2012 InterDigital, Inc. All rights reserved. 

Methodology 

•  Probability of missed detection ( ​𝑃↓𝑚 ): 

•  In this experiment ​𝑃↓𝑚  is determined analytically based on the 

properties of the discovery signal and the characteristics of each drop 

•  This model assumes that the total interference is spectrally white 

•  False alarm probability ( ​𝑃↓𝑓𝑎 ): 

•  The false alarm probability is the probability that a UE detects the 

presence of another UE in proximity, where in fact there is no such 

UE present 

•  ​𝑃↓𝑓𝑎  is also determined analytically 

•  Range: 

•  The distance for which 95% or more of the discovery signals are 

discovered at a target false alarm probability of 0.1% 

12 © 2012 InterDigital, Inc. All rights reserved. 

Simulation assumptions 

Cat.

   Parameter

   Value

De

pl

oy

m

en

t

ISD 500 m 

Distribution of devices Uniformly 

Mobility 3 km/hr 

Centre carrier frequency 2 GHz 

System bandwidth 

(BW) 10 MHz 

Ch

an

ne

l

m

od

el

Path loss model Indoor Hotspot NLOS 

Shadowing standard 

deviation 4 dB 

De

vi

ce

  p

ar

am

s.

Antenna gain 0 dBi 

Noise figure 9 dB 

Antenna height 1.5 m 

Number of Tx and Rx 

antennas 1 Tx / 2 Rx 

Cat.

   Parameter

   Value

Di

sc

ov

er

y

sig

na

l

  p

ar

am

et

er

s

Length of discovery signal 800 µs 

Length of cyclic prefix 12.5 µs 

D 12.5 µs 

NRB 6 RBs 

Ndru 2 

Discovery bandwidth Ndbw 12 RBs 

Ncs 64 

Target false alarm rate 0.1% 

Target detection rate 95% 

Target detection range 200 m 

NTx 30 / sector 

NRx 60 / sector 

13 © 2012 InterDigital, Inc. All rights reserved. 

Experimental Results  Pm for cell-edge devices 

•  Cell-edge devices 

are those worst 

10% UEs w.r.t. the 

distance to the eNB.  

•  Cell-edge device 

suffer a 2dB loss at ​𝑃↓𝑚 =1% compare 

to the cell average 

14 © 2012 InterDigital, Inc. All rights reserved. 

Experimental Results  Pm for Options 1-3 

Opt.

   Loss

1

   3dB

2

   2dB

3

   0dB

Impact

  of

  Interference

at

  Pm=1%

15 © 2012 InterDigital, Inc. All rights reserved. 

Experimental Results                  Range for Option 1-3 

20m

10m

16 © 2012 InterDigital, Inc. All rights reserved. 

Summary and Conclusions 

•  This studies interference management for D2D 

discovery in the context of LTE systems 

•  Various practical options for resource allocations are 

studied 

•  The experimental results show that: 

•  D2D discovery signal inter-cell interference has the 

potential to significantly impact the D2D discovery 

detection range 

•  allocating the D2D discovery resources for cell-edge 

UEs using a centralized approach has to potential to 

(almost) eliminate the loss in D2D discovery detection 

range due to inter-cell interference