Intercell Interference Coordination for D2D Discovery in ... - IEEE Xplore

IEEE WCNC'14 Track 3 (Mobile and Wireless Networks)

Intercell Interference Coordination for D2D Discovery in LTE-A HetNets Leonardo Babun, Meryem Simsek, and ˙Ismail Güvenç Department of Electrical & Computer Engineering, Florida International University 10555 West Flagler St. Miami, FL 33174 Email: lbabu002@fiu.edu, msimsek@fiu.edu, and iguvenc@fiu.edu

Abstract—Device-to-device (D2D) communication enables mobile devices to directly communicate with each other without the help of infrastructure and to reuse radio resources within cellular networks. When D2D user equipments (UEs) use the same radio resources for discovery and communication as the cellular network, intercell interference (ICI) may degrade the D2D discovery performance. Therefore, ICI between the cellular network and D2D UEs should be coordinated to avoid performance degradation. In this paper, we evaluate the benefits of timedomain intercell interference coordination (ICIC) approaches in D2D discovery by using almost blank subframes (ABS). Also, we present the benefits of using multi-hop discovery. Index Terms—D2D discovery , HetNets, Intercell Interference Coordination, LTE, System Level Simulations.

distance [m]

1000 800 600 400 200 0 −200 −400 −600 −800 −1000

I. I NTRODUCTION

D2D UE

−1000 −500

The demand for high speed wireless data communication in cellular networks is continuously increasing as new mobile communication services are introduced. In order to allow new types of services and at the same time improved network performance, recent research aims to increase the system capacity of cellular networks, e.g in Long Term Evolution (LTE)Advanced and WiMAX systems. Device-to-Device (D2D) communication technologies emerge as one of the solutions to offload traffic from cellular networks and to enhance the overall network performance [1]. D2D communication enables mobile devices to transmit and receive data directly without the help of infrastructure such as access points and base stations (BSs). Using D2D communications, wireless devices can communicate with each other via direct D2D links over licensed and/or unlicensed spectrum. D2D communications overlaid with cellular networks is expected to provide high capacity, guaranteed quality of service (QoS) over long ranges, and new proximity-based services. These technologies are particularly critical for scenarios with little or no infrastructure support such as emergency response and disaster relief operations. Based on its advantages including improved spectrum reuse and system throughput, offloading of the cellular network, improved energy efficiency, and extended coverage, D2D is considered to be the most promising technology in public safety systems and for handling proximity services (ProSe) [2]. In December 2012, a study item was created in the 3rd Generation Partnership Project (3GPP) radio access network (RAN) standardization group to study LTE D2D communications. Since then, important aspects of D2D such as design considerations for

978-1-4799-3083-8/14/$31.00 ©2014IEEE

regular PBS UE

MBS

0

distance [m]

500

1000

Figure 1: Layout option 5 according to [4] overlaid by small cell BSs.

D2D discovery, D2D group communications, physical layer enhancements and channel models for D2D deployments have been discussed and summarized in several 3GPP standardization meetings [1], [3]. D2D discovery is of critical importance as it enables proximity-based services. However, in high interference environments (e.g. heterogeneous network (HetNet) deployments), user equipment (UE) to UE (UE2UE) discovery can be affected by the influence of other stations (i.e. UEs, macrocells and picocells) which utilize the same frequency bands. In our previous work, we provided an overview of new agreements in 3GPP LTE RAN1 which are related to evaluation methodology and channel modeling for D2D discovery and communications. We also presented a D2D system level simulation environment based on the 3GPP simulation assumptions and a performance evaluation of three different D2D discovery algorithms [5]. In this paper, we extend the D2D system level simulator to a heterogeneous cellular network environment based on 3GPP simulation assumptions for HetNets [6]. The considered cellular network is depicted in Fig. 1, and consists of macrocells, picocells, regular UEs (communicating with macrocells/picocells), and D2D UEs. In this new model, we included macrocell and picocell-to-UE interference analysis in addition to the new intercell interference coordination (ICIC) approach

2202


II. D2D S YSTEM L EVEL S IMULATION E NVIRONMENT The D2D system level simulator including a heterogeneous cellular network deployment is based on the assumptions in 3GPP LTE-Advanced, including those on the radio frame structure, the physical resource structure and the channel modeling [3], [6], [8]. We developed a D2D system level simulator consisting of modular simulation blocks as shown in the flowchart in Fig. 2. For adapting the D2D simulator in our earlier work [5] to the assumptions in 3GPP LTE-Advanced, we modified the UE-to-UE initial channel propagation model and introduced the Macro-to-UE and Pico-to-UE multi-path links. These modifications directly impacted the ’Initialization’, ’Channel generation’ and ’SINR calculation’ blocks. For path loss, shadowing and fast fading calculation in the new scenarios we considered the channel models provided in [6]. The D2D system level simulator continuously simulates the temporal development of a LTE-Advanced based cellular network by calculating densely spaced snapshots, each corresponding to one subframe t with an iteration step size of T = 1 ms. Within each snapshot, the position of each UE changes according to its velocity, and the channels between each macro/pico base station (BS) to each UE and from each UE to each UE are newly calculated. D2D discovery is performed in each discovery time slot over NF × NT discovery resources (DR) based on different discovery methods, whereby DRs are multiplexed in both frequency and time domains. NF represents the number of available DRs in the frequency domain and NT is the number of discovery subframes in the time domain. The system level simulator consists of the following simulator blocks: 1) Initialization: In this simulation blocks the simulation parameters, discovery methods, and ICIC algorithms are set. The simulation parameters include the number of macrocells, the system bandwidth according to 3GPP LTE specifications,

Initialization t =1 Channel generation t = t+1

and the use of multi-hop. Even though D2D discovery is considered for the uplink transmission mode in 3GPP, we focus on the downlink transmission due to the possibility of reusing the existing enhanced intercell interference coordination (eICIC) techniques recently developed for HetNets [6], [7]. The main goal of the present paper is to evaluate the benefits of time-domain ICIC approaches in D2D discovery by using almost blank subframes (ABS) over several different scenarios and multiple ABS rates. We also present the benefits of using multi-hop discovery, in which a UE performs discovery by using another UE. The considered system model includes high interference scenarios and is evaluated for various ABS ratios for macrocells and picocells. The paper is organized as follows: In Section II, we present an overview of the HetNet based D2D system level simulator. In Section III, D2D discovery approaches together with ABS based ICIC algorithms and multi-hop discovery are discussed. In Section IV, D2D discovery performance results are evaluated for the proposed ICIC and multi-hop approaches. Finally, Section V concludes the paper.

Mobility

no

t = NT yes

RG selection SINR calculation D2D Discovery Mobility

Evaluation

Figure 2: D2D system level simulator flowchart.

the carrier frequency f0 , the transmit powers, the duplex mode (full duplex or half duplex)1, the number of discovery subframes NT . Also, the layout option for D2D discovery and communication based on [3] and, on the selected layout option, the number of UEs, and the UE distribution (uniform/hotspot). The following simulation layout options are agreed in 3GPP RAN1 for simulating D2D scenarios [4]: • • • • • •

Option 1: Urban macro (500 m inter-site distance (ISD)) + 1 remote radio head (RRH)/Indoor Hotzone per cell Option 2: Urban macro (500 m ISD) + 1 Dual stripe building per cell Option 3: Urban macro (500 m ISD); all UEs outdoor Option 4: Urban macro (500 m ISD) + 3 RRH/Indoor Hotzone per cell Option 5: Urban macro (1732 m ISD) Option 6: Urban micro (100 m ISD)

All layout options consider a hexagonal grid, 3 sectors per site, with 19 or 7 macrocell sites. The total number of active UEs per (active) cell are specified as 25 for options 1, 2, 4, and as 10 for options 3, 5, 6, whereby the total number of UEs is 150 for each layout option. 2) Channel Modeling: Time and frequency selective channel models considering multi-path fading effects are implemented. The simulator supports two different channel modeling methods for the BS to UE (BS2UE) channels and for the UE2UE channels according to [3]. All path loss models, shadowing factors/correlation, antenna gain models and multipath parameters are based on 3GPP assumptions. We consider that the total bandwidth (BW) is divided into subchannels with bandwidth Δf = 15 kHz. Orthogonal frequency division multiplexing (OFDM) symbols are grouped into resource blocks (RBs). Macro and picocells operate in the same frequency band and have the same amount NRB of available RBs, while UEs perform D2D discovery in the same frequency band, too. The main difference between the UE2UE channel and the 1 3GPP has recently decided to only consider the half duplex mode in D2D discovery and communication.

2203


BS2UE channel lies in the dual mobility of both devices in the former case, according to which the Doppler frequency calculation changes. For a multi-path channel with N paths and L subpaths for each path, the Doppler frequency for the l ∈ L-th subpath of the n ∈ N -th path of the BS2UE channel is given by [3], [9]: f0 BS2UE , (1) = vrx cos φAoA fDoppler,n,l n,l − θvrx c with vrx the velocity of the receiver/UE, φAoA n,l the angle of arrival from the l-th subpath of the n-th path, θvrx the random angle receiver’s/UE’s direction of movement, and c = 3 · 108 m/s. The Doppler frequency for the l ∈ L-th subpath of the n ∈ N -th path of the UE2UE channel is defined as [3]: UE2UE fDoppler,n,l = vrx cos φAoA n,l − θvrx

(2)

f0 , + vtx cos φAoD n,l − θvtx c with vtx the velocity of the transmitter/UE, φAoD n,l the angle of departure from the l-th subpath of the n-th path, and θvtx the random angle of direction of the transmitter/UE. The output of this simulation block is the channel transfer function from each transmitter to each receiver over all RBs and is calculated in each subframe t. 3) Resource Group Selection: Different RG selection algorithms are supported by the D2D system level simulator. The RG selection algorithms are described in Section III. 4) SINR Calculation: After each UE selects its RG based on a preselected algorithm, the signal-to-noise-plusinterference-ratio (SINR) is calculated for each UE on its selected RG. Hereby, the total interference is considered, i.e. the interference from macro/pico BSs and UEs performing discovery on the same RG. 5) D2D Discovery: After each UE selects its RG based on a preselected algorithm, the D2D discovery is performed based on a variable SINR threshold value, i. e. a UE is discovered when its SINR is larger than a SINR threshold. This simulation block can be extended to various D2D discovery approaches. 6) Mobility: A random walk process is considered, whereby all UEs in the network are assumed to be mobile. The velocity vu of a particular UE u can be set to correspond to any of {3, 30, 60, 120} km/h as proposed in [10] or is set according to 3GPP D2D layout option assumptions. The location lu (x, y, t) of UE u in position (x, y) at a velocity vu and a random direction of movement du = [cos(φ), sin(φ)] with φ ∈ [0; 2π] is updated in snapshot t according to: lu (x, y, tT + T ) = lu (x, y, tT ) + vu · T · du ,

(3)

(CDF) of total number of UEs discovered as a function of time. • Closed discovery (i.e. knowing the UEs to be discovered): Discovery probability as a function of time. • Average number of discovered UEs vs different number of DRs. The evaluation block is the last simulator block which is run after all snapshots. It provides averaged results over various random drops. III. D2D D ISCOVERY A LGORITHM AND ABS BASED ICIC It is assumed that the considered HetNet consists of a set of M = {1, ..., M } macrocells, a set of K = {1, ..., K} uniformly positioned picocells per macro sector, and a set of UEs U = {1, ..., U }. In the considered system, the macro BS m transmits with a maximum transmit power pm tx , the pico BS k transmits with a maximum transmit power pktx and the UE u transmits with a maximum transmit power putx . Assuming a uniform power distribution the corresponding transmit powers k u per RB r are pm tx (r), ptx (r), and ptx (r), respectively. Within the considered co-channel deployment scenario, in which the UEs perform D2D discovery in the same frequency bands as BSs are transmitting, a UE u’s total received power in RB r is given by: purx (r) =

M m=1

pm tx (r) +

K k=1

pktx (r) +

U

pw tx (r).

(4)

w=1,w=u

It is assumed that all UEs select a DR to periodically transmit their peer discovery information, whereby each DR is identified by a PDRID (Peer Discovery Resource ID). In each discovery slot with D = NF × NT DRs, each UE u ∈ U transmits its peer discovery information on one DR du ∈ D and listens to the remaining DRs to discover its peers. Hereby, the whole BW is used for discovery, so that NF ≈ NRB . The rationale behind this is to examine a worst case scenario with full interference over the whole BW. The following two PDRID selection approaches are proposed: • Random PDRID Selection: This is a baseline approach in which each UE u selects its PDRID du randomly out of D DRs in each discovery slot. • Smart PDRID Selection: In this approach, a UE u considers its received powers purx = [purx (1), . . . , purx (R), . . . , purx (D)] in one discovery slot and selects its DR d ∈ D based on the minimum received power according to: (5) arg min = (purx ) . d Considering the half duplex mode, in which UEs that selected DRs within the same subframe cannot discover each other, discovery is performed based on a threshold γthresh . A UE u is assumed to be successfully discovered if its SINR γ u (d) exceeds this threshold:

7) Evaluation:: The following metrics that are agreed upon to evaluate the performance of D2D discovery are considputx (d) · |hu(d),w(d)(t, d)|2 ered [11]: γ u (d) = > γthresh , v 2 2 • Open discovery: Total number of UEs discovered as a v=u ptx (d)|hu(d),v(d) (t, d)| + Imac + Ipic + σ (6) function of time and the cumulative distribution function

2204


1st-hop 2nd-hop 3rd -hop

RABS = 1

a) NF

ABS

NF

ABS

ABS

ABS

ABS

ABS

͘͘͘

RABS = 2 ABS

͘͘͘

ABS

UE3

RABS = 3 NF

ABS 1

2

ABS

ABS

UE2

͘͘͘

͘͘͘

UE1 NT

time

Discovery region of UE1

b) NFrame

͘͘͘

NFrame

͘͘͘

͘͘͘

NF NT

DR

Figure 4: Illustration of multi-hop discovery of far away UEs.

(7)

m

and the picocell interference p ptx (d)|hu(d),p(d) (t, d)|2 , Ipic =

UE4 Discovery region of UE3

time

Figure 3: (a) Frame structure for different ABS rates (RABS ), (b) DR distribution over NT × NF LTE-A subframe. with the macrocell interference 2 Imac = pm tx (d)|hu(d),m(d) (t, d)|

Discovery region of UE2

(8)

ICIC method, which is demonstrated based on system level simulations in the next section. Additionally, multi-hop discovery is considered. By using multi-hop, UEs can discover each other even when they are located in different discovery regions. As depicted in Fig. 4 we consider a multi-hop approach in which UE1 can discover in 1-hop UE2 in its own discovery region. In case of 2-hops UE1 can discover UE3 which is in UE2 ’s discovery region over UE2 . Using multi-hop D2D discovery a UE’s discovery region is virtually expanded.

p

IV. S IMULATION R ESULTS 2

where |hu(d),w(d) (t, d)| is the channel gain between UE u and UE w ∈ U in DR d at time t, |hu(d),m(d)(t, d)|2 is the channel gain between UE u and macro BS m in DR d, and |hu(d),p(d) (t, d)|2 is the channel gain between UE u pico BS p. It is obvious that the SINR threshold γthresh defines the discovery region of a UE in the proposed D2D discovery method. Different threshold values are examined in the next section to demonstrate the effect of it on the D2D discovery performance. To further improve the D2D discovery performance, we consider time-domain ICIC within the heterogeneous cellular network. Macro and pico BSs perform ICIC by ABSs as proposed in [6]. Within ABSs either the macro BSs or the pico BSs mute their transmission. Based on a selected ABS ratio RABS , BSs do not transmit on certain subframes out of [1, . . . , NT ] subframes, i.e., they cause no interference on these subframes. An illustration of the muting frame structure is depicted in Fig. 3(a) for different ABS ratios. Fig. 3(b) shows the DR distribution over NT subframes. Considering a decentralized approach in which the UEs are not informed about the ABSs ratio RABS , UEs select DRs based on the minimum received power according to the proposed metric in equation (5). Subframes with ABS will be preferred due to lower interference and can discover more UEs and more simultaneous DR transmissions by UEs. This enables UEs in close proximity to a macro/pico BS to perform D2D discovery in ABSs by selecting their DR in these subframes. However, muting subframes will lead to the case that many UEs select DRs in the ABSs due to the lower received power on these subframes. This will cause more interference by the UEs that transmit their peer discovery information on the same DRs so that consequently less UEs will be discovered. Hence, there is an optimum ABS ratio and SINR threshold in the proposed

This section presents system level simulation results for device discovery based on the 3GPP simulation and layout assumptions for HetNets and D2D discovery [1], [6], [7]. It is assumed that the considered HetNet consists of M = 1 macrocell, K = 2 uniformly positioned picocells per macro sector, and U = 450 (150 per macro sector) UEs which are uniformly dropped based on layout option 5 assumptions in [4]. We focus on this layout option since it is mandatory especially for public safety scenarios. Throughout the simulations we assume that NF = NRB (unless stated otherwise), i.e. all RGs in timeand frequency domain are available for discovery, whereby DRs are selected based on the PDRID selection algorithm. Further details about system level simulation parameters are provided in Table I. System level simulations have been performed for various random scenarios, whereby we consider the same random layouts and channels for each algorithm to enable a fair comparison. Fig. 5 depicts the cumulative distribution function (CDF) of the number of discovered UEs for the two D2D discovery approaches random and smart PDRID selection for different number of subframes in case of dedicated channel D2D and regular UEs for an SINR threshold γthresh . It can be observed that the smart PDRID approach yields in all cases a better performance than the random PDRID selection approach. To demonstrate the performance of the proposed methods in case of co-channel deployments and ABS based ICIC, we evaluate the performance of the average number of discovered UEs vs NT in Fig. 6 and Fig. 7 for macro cell based ICIC and pico cell based ICIC, respectively. The different ABS ratios RABS are the iteration step size of subframes over which the BSs mute, i.e. RABS = 1 means that the BS mutes every second subframe and RABS = 0 means that the BS does not

2205


Carrier frequency System bandwidth Subframe duration Number of RGs Number of macrocells/picocells Maximum transmit power k u pm tx ,ptx ,ptx Number of UEs per sector UE2UE path loss model UE2UE shadowing correlation BS2UE path loss model BS2UE shadowing correlation Shadowing standard deviation Minimum distance UE2UE Thermal noise density

Value Hexagonal grid, 3 sectors per cell, reuse 1 700 MHz 10 MHz 1 ms NF =50 M = 3/K = 2 46 dBm, 30 dBm, 23 dBm 150 (U = 450) Winner channel model (see [12]) Spatially correlated According to table B.1.2.1-1 in [6] Time correlated Independent and identically distributed (iid) 3m -174 dBm

1 0.9 0.8 0.7 0.6 CDF

Smart NT = Smart NT = Smart NT = Smart NT = Smart NT = Random NT Random NT Random NT Random NT Random NT

0.5 0.4

0.3 0.2 0.1 0 0

50

2 3 5 8 10 =2 =3 =5 =8 = 10

100 150 200 250 300 Number of discovered UEs

350

20 18 16 14 12 10 8 6 4 2 0 0

1

2

3

4

5 NT

Smart RABS = 0 Smart RABS = 1 Smart RABS = 2 Smart RABS = 3 Random RABS = Random RABS = Random RABS = Random RABS = 7 8 6 9

0 1 2 3 10

Figure 6: Average number of discovered UEs vs NT for various ABS ratios in case of macro ICIC and γthresh = 8 dB. Average number of discovered UEs

Parameter Cellular layout

Average number of discovered UEs

Table I: Simulation parameters.

50

Smart RABS = Smart RABS = Smart RABS = Smart RABS = Random RABS Random RABS Random RABS Random RABS

45 40 35 30 25

0 1 2 3 = = = =

0 1 2 3

20 15 10 5 0 0

1

2

3

4

5 NT

6

7

8

9

10

Figure 7: Average number of discovered UEs vs NT for various ABS ratios in case of pico ICIC and γthresh = 8 dB.

400

Figure 5: CDF of number of discovered UEs in the absence of ICIC for random and smart PDRID selection, and γthresh = 8 dB.

mute (see Fig. 3(a)). It can be observed that there is no gain in case of macro ICIC. This is due to the fact that the macro BS transmit power is high and UEs select their discovery resources mainly on subframes on which the macro BS mutes, which leads to a high UE2UE interference. In case of pico ICIC, we observe high performance improvements for the smart PDRID selection. This is because of the lower transmit power of pico BSs, which does not lead to overloaded ABSs. It can be seen that NT = 8 yields the best performance. Therefore, the following results will focus on performance evaluations for this number of discovery subframes. However, it has to be pointed out that NT = 8 may lead to delays in UE discovery. Fig. 8 and Fig. 9 depict the average number of discovered UEs vs SINR threshold γthresh for the smart PDRID selection algorithm for different ABS ratios. In case of pico ICIC the average number of discovered UEs is increasing for increasing number of ABSs while it is almost the same in case of macro ICIC. The wider a UE’s discovery range (the smaller γthresh ) is, the more UEs can be discovered.

To further improve the number of discovered UEs, we performed multi-hop based D2D discovery. The effect of interference from BSs on the discovery performance and the benefit of using multi-hop discovery in the absence of ICIC for 1-hop, 2-hop and 3-hop can be found in Fig. 10. Both a co-channel (without ICIC) and dedicated channel deployment of D2D and regular UEs are considered. It is observed that dedicated channel deployment of D2D UEs results in significantly larger number of UEs to be discovered. On the other hand, even with co-channel deployments, UEs are able to discover all the other UEs after three hops. Fig. 11 investigates the performance of multi-hop D2D discovery approach in co-channel deployments and with different ABS rates at the macrocell and picocells. Three different ABS ratios are considered: RABS = 1, RABS = 2, and RABS = 3. Comparing with Fig. 10, use of ICIC leads to larger number of discovered UEs, especially when implemented at picocells. The average number of discovered UEs increases with the number of hops, but also depends on the different ABS ratios in both macro and pico BSs. In all cases, all UEs are discovered after 3 hops. V. C ONCLUSION In this paper we describe a D2D system level simulator including a heterogeneous network deployment based on 3GPP

2206

40

Smart Smart Smart Smart

35

NT NT NT NT

= = = =

8; 8; 8; 8;

RABS RABS RABS RABS

= = = =

450

0 1 2 3




400 350

30

300

250

25

200 150

20

100

15 −8 −6 −4 −2

0 4 2 γthresh [dB]

6

8

10

12

Figure 8: Average number of discovered UEs vs γthresh for various ABS ratios in case of macro ICIC. Smart Smart Smart Smart

100 90

NT NT NT NT

= = = =

8; 8; 8; 8;

RABS RABS RABS RABS

= = = =

50 0

1

2 Number of hops

3

Figure 10: Average number of discovered UEs for multi-hop in the absence of ICIC.

0 1 2 3

450


110


Co-channel Dedicated Channel

Macro BS ICIC Pico BS ICIC

400

80

350

70

300

60

250

50

200

40

150

30

100

20 10 −8 −6 −4 −2

2 0 4 γthresh [dB]

6

8

10

12

Figure 9: Average number of discovered UEs vs γthresh for various ABS ratios in case of pico ICIC.

50 0

1 3 2 RABS = 1

3 1 2 RABS = 2

2 3 1 RABS = 3

No of hops

Figure 11: Average number of discovered UEs for multi-hop discovery and different ABS ratios. assumptions. We propose a received power based PDRID selection algorithm for D2D discovery and propose an ABS based ICIC coordination approach. In addition, we consider multi-hop based D2D discovery which leads to significant performance improvement. System level simulation results demonstrate that pico cell based ICIC leads to a better D2D discovery performance than macro cell based ICIC. In case of macro cell ICIC, the ABSs are overloaded and UEs suffer from other UE’s interference. While picocell-based ICIC is not common, it may be used to enhance the performance of high speed macrocell users as described in [7], [8]. Our future work includes evaluation of D2D discovery performance during uplink transmissions in a HetNet scenario. VI. ACKNOWLEDGMENT Part of the D2D system level simulator used in this paper has been developed while ˙Ismail Güvenç has been consulting to Motorola Solutions, Inc. The authors would like to thank Neiyer Correal for fruitful discussions. R EFERENCES [1] Qualcomm, Inc., “Study on LTE device to device proximity services,” 3GPP Work Item Description (RP-122009), Spain, Dec. 2012. [2] FCC, “Connecting America: The national broadband plan,” 2010. [Online]. Available: http://download.broadband.gov/plan/ national-broadband-plan.pdf

[3] 3GPP, “Feasibility study on LTE device to device proximity services radio aspects,” Technical Report (3GPP TR 36.843), 2013. [4] Alcatel-Lucent and Alcatel-Lucent Shanghai Bell, “LTE D2D dropping and association,” 3GPP RAN1 Standard Contribution (R1-131755), Chicago, IL, Apr. 2013. [5] M. Simsek, A. Merwaday, N. Correal, and I. Guvenc, “Device-todevice discovery based on 3GPP system level simulations,” in Proc. IEEE Global Commun. Conf., Workshop on Device-to-Device (D2D) Communication With and Without Infrastructure, Atlanta, GA, Dec. 2013. [6] 3GPP, “Evolved Universal Terrestrial Radio Access (E-UTRA); further advancements for E-UTRA physical layer aspects,” Technical Report (3GPP TR 36.814), 2010. [7] D. Lopez-Perez, I. Guvenc, and X. Chu, “Mobility management challenges in 3GPP heterogeneous networks,” IEEE Commun. Magazine, vol. 50, Issue: 12, pp. 70–78, 2012. [8] D. Lopez-Perez, I. Guvenc, and X. Chu, “Mobility enhancements for heterogeneous wireless networks through interference coordination,” in Proc. IEEE Int. Workshop on Broadband Femtocell Technologies, April 2012, pp. 69–74. [9] Winner, “IST-4-027756 WINNER II D1.1.2 V1.2 WINNER II channel model,” 2008. [Online]. Available: http://www.cept.org/files/ 1050/documents/winner2%20-%20final%20report.pdf [10] 3GPP, “Mobility enhancements in heterogeneous networks,” Technical Report (3GPP TR 36.843), 2012. [11] MCC Support, “Final report of 3GPP TSG RAN WG1 #72bis v1.1.0,” 3GPP RAN1 Standard Contribution (R1-132646), Chicago, USA, Apr. 2013. [12] ——, “Draft report of 3GPP TSG RAN WG1 #73 v0.2.0,” 3GPP RAN1 Standard Contribution, Japan, May 2013.

2207