The Vault

D2D Neighbor Discovery Interference Management for LTE Systems
Research Paper / Feb 2014

D2D Neighbor Discovery Interference Management for LTE Systems Yuxin Zhao, Benoit Pelletier, Paul Marinier, Diana Pani InterDigital Communications Montreal, Qc Canada {yuxin.zhao, benoit.pelletier, paul.marinier, diana.pani} Abstract— Direct Device-to-Device (D2D) communications in cellular systems has attracted a lot of interest for both commercial and public safety applications. To support D2D communications, a first step involving neighbor discovery is typically needed. This paper focuses on the study of neighbor discovery signal design aspects and interference management for D2D discovery under a typical LTE system deployment. System level results on detection rate and range comparing different interference management techniques are provided. The results indicate that practical interference management techniques have the potential to improve the performance of neighbor discovery for cell-edge devices significantly. Keywords—D2D; discovery signal design; interference management; resource hopping I. INTRODUCTION Motivated by the desire to determine the proximity between devices at a higher accuracy than current LTE positioning can provide, study item work for Device-to-Device Proximity Services (ProSe) has recently begun in 3GPP [1]. The work consists of studying Device-to-Device (D2D) neighbor discovery and communications. In D2D neighbor discovery, two (or more) devices determine their relative proximity based on direct radio communications. In D2D communications, two or more devices exchange data using a direct communication link, i.e. without having to go through the base station or Evolved Node B (eNB). One of the key issues to be studied is the additional interference introduced by these new D2D transmissions. In this paper, we focus on interference management and mitigation for D2D discovery. First, we address the design aspects of D2D discovery, including the choice of resources to be used for discovery and also the discovery signal structures design. We then analyze the interference scenarios for system with D2D transmissions and propose three interference management options to mitigate the interference caused by D2D transmissions. With simulation results we show that with practical management techniques in current LTE systems the impact of interference can be mitigated. This paper is constructed as follows: Section II introduces the design aspects of D2D discovery signals. Section III proposes three inter-cell interference management techniques, and simulation results for the proposed techniques are provided in Section IV. Section V concludes the paper. II. DISCOVERY SIGNAL DESIGN A. Resources for discovery One fundamental design consideration for D2D neighbor discovery and communications is the choice of resources to use for transmission. In principle, it should be possible to use either the uplink (UL) or downlink (DL) resources for D2D transmissions. It has recently been decided in 3GPP to use UL resources for the study item work. D2D transmissions will introduce additional intra-cell and inter-cell interference. Figure 1 a) and b) illustrate the interference scenarios for the cases where UL resources are used for D2D transmissions, and for the case where DL transmission are used for D2D transmissions, respectively. In the case where UL resources are used for D2D transmissions, as shown in Figure 1 a), the D2D transmissions cause inter-cell interference to neighbor eNBs; pretty much in the same way as in current systems. The D2D receiving devices are on the other hand interfered by inter-cell nearby cellular devices transmitting to the eNB. On the other hand in the case where DL resources are used for D2D transmission, as shown in Figure 1 b), the D2D transmitting devices interfere to other UEs receiving normal communications on the DL resources. Conversely, the eNBs in that case interfere with the D2D transmissions. D2D UL UL interferenceUL D2D D2D Interference (a) D2 D DL DL interference DL D2D interference (b) Figure 1 Impact of D2D when using UL (a) or DL (b) resources Globecom 2013 Workshop - International Workshop on Device-to-Device (D2D) Communication With and Without Infrastructure978-1-4799-2851-4/13/$31.00 ©2013IEEE 556 From the above discussion, and considering the practical implementation implications of a device using DL resources for transmission, we conclude that it is preferable to use UL resources for D2D transmission; more specifically, D2D transmission in this paper are assumed to use UL subframes (TDD) or UL spectrum (FDD). To simplify inter-cell coordination, it is further assumed that dedicated subframes are reserved for D2D transmissions (at least for discovery). To support discovery across cell borders, it is further assumed that the eNBs are coarsely synchronized and use the same reserved D2D subframes. In this context, D2D devices are assigned specific subframes to transmit discovery signal. During a D2D reserved subframe, in practice the entire Physical Uplink Shared Channel (PUSCH) region could be used in a D2D reserved subframe. Thus for a typical total bandwidth of 10 MHz, a maximum of 44 Resource Blocks (RBs) could be used for D2D communications when considering 6 RBs reserved for the PUCCH regions. As illustrated in Figure 2, we define a discovery bandwidth of Ndbw physical resource blocks (PRB) which consists of Ndru discovery resource unit (Ri, i=1,…,Ndru), where each resource unit consists of NPRB, where NPRB is the number of PRBs in one resource unit. PUCCH Resource for D2D Discovery System bandwidth of 10 MHz PUCCH R1 R2 R3 RNdru NPRB D2D subframe Ndbw Figure 2 D2D frequency allocation model in 10MHz system bandwidth B. Discovery signal design For neighbor discovery, the device to be discovered needs to transmit a signature sequences. This sequence can be derived from an existing UL LTE signal. It is also possible of course to use a new design; in this paper however we assume a discovery signal based on Zadoff-Chu (ZC) sequences (e.g., Sounding Reference Signal (SRS)/Random Access Channel (RACH) like signals or any other signal based on ZC sequences). ZC sequences have convenient correlation properties and are defined in LTE using the following equation [2]: π‘₯!(𝑛) = 𝑒(!!!!"! !!!! !!"!!" )   (1) where π‘ž ∈ {1,2,… ,𝑁!" − 1} is the ZC sequence root index, 𝑛 = 0, 1,2,… ,𝑁!" − 1, Nzc is the length of ZC sequence, and 𝑙 is any integer. In LTE, 𝑙 = 0 is used for simplicity. The correlation between two ZC sequences is given by: 𝜌!,! 𝜎 = π‘₯! 𝑛 π‘₯!∗ 𝑛 + 𝜎 =  π›Ώ 𝜎                      π‘˜ = π‘ž !!"!! !!! π‘₯! 𝑛 π‘₯!∗ 𝑛 + 𝜎 =   1𝑁!"!!"!!!!!                    π‘˜ ≠ π‘ž (2) where π‘˜, π‘ž ∈ {1,2,… ,𝑁!" − 1} , 𝜎 is the cyclic shift of the sequence, and 𝛿 βˆ™ is the delta function. Equation (2) indicates that the autocorrelation of two different ZC sequences generated from the same root is zero, regardless of the value of the cyclic shift. Thus with proper design [2], two discovery signals with different cyclic shifts of the same root can be easily detected without any interference/ambiguity. On the other hand, two ZC sequences generated from different sequence roots will interfere with each other. III. INTERFERENCE MANAGEMENT In this section, interference scenarios are analyzed and interference management techniques are proposed. To simplify description, and without loss of generality, only 2 discovery resource units (i.e., R1 and R2 in Figure 2) are used in the description of the proposed techniques. The proposed techniques are built from the assumptions for the resource allocation and discovery signal design of the previous Section. A. Inter-discovery signal interference (IDSI): As illustrated in Figure 3, both intra and inter-cell interference can be present for D2D. While intra-cell interference can be managed by each eNB scheduler, inter-cell interference requires coordination between eNBs. In general, because of this inter-cell interference, cell edge devices will suffer more than cell-center UEs. This paper focuses on the inter-cell interference management for devices at the cell-edge. Inter-Cell Interference Intra-Cell Interference Figure 3 Interference between D2D transmissions B. Interference management Three interference management options are proposed in this section. 1) Option 1: In this option, all the eNBs use the same dedicated frequency resources for discovery. The total number of available distinct discovery sequences is thus Ndru×Ncs for each eNB, where Ncs is the number of cyclic shifts for one root Globecom 2013 Workshop - International Workshop on Device-to-Device (D2D) Communication With and Without Infrastructure557 sequences. Different eNBs use different sets of sequence roots to avoid sequence ambiguity. When one eNB is running out of cyclic shifts for a root, another root is selected to generate a new set of Ndru×Ncs sequences. For a given device, the interference is thus caused by all other-cell transmitting (Tx) devices using different sequence roots, and potential intra-cell Tx devices using different sequence roots. This concept is illustrated in Figure 4. Root 1 Root 2 Root 3 Root 4 Root 5 Figure 4 Interference management Option 1 2) Option 2: In this option, a hopping pattern is assigned to the UE discovery sequence, such that they are transmitted in alternation between the Ndru discovery resources units (in this example R1 and R2) resulting in the interference being randomized on different resources. As in option 1, different eNBs assign sequences from different set of roots. The interference in this option is caused by Tx devices from other cells that transmit on the same resource but using different roots. When one eNB runs out of cyclic shifts on all Ndru discovery resources units, it selects another root to generate new sets of Ndru×Ncs sequences. Root 1 Root 2 Root 3 Root 4 Root 5 Figure 5 Interference management Option 2 3) Option 3: Observing that devices at the cell edge experience more inter-cell interference due to presence of non-orthogonal discovery sequences (due to different roots assigned by neighbor eNBs), we propose a resource allocation scheme to minimize the non-orthogonality among discovery signals of cell edge devices. This option is based on splitting of discovery resource unit among UEs at cell centre and UEs at cell edge, a concept illustrated in Figure 6. This option may be enabled by a centralized controller assigning discovery resources units (at least) for the cell edge devices. In this option, the cell center devices are defined as those within a radius of 𝑅!! of the eNB1; all other devices in the cell are defined as cell edge devices. A cell edge resource (central) controller coordinates the resources for cell edge devices of all eNBs in a coordinated area (e.g. several eNBs). In this option, the cell edge devices and cell center devices use 𝑅! and 𝑅! resources for D2D transmission, respectively. This eliminates interference from cell center UEs to cell edge UEs. The central controller assigns cyclic shift to all cell edge devices on resource 𝑅! (without loss of generality index “1” is used for description purposes here). When the central controller runs out of cyclic shift for a given root, it selects another root to generate new set of sequences. Each eNB assigns cyclic shifts to its cell center devices on resource  π‘…!,, 𝑛 ≠ 1, and different eNB use different sequence roots. When an eNB runs out of cyclic shifts, another root is selected to generate new sequences. Thus in this option, and provided the UEs are synchronized and the cyclic prefix is long enough, the interference for the cell edge devices is caused by other cell edge Tx devices using different sequence roots (i.e. the central controller has to exhaust all cyclic shifts of one root before inter-cell interference is experienced by the cell-edge devices). That the cell center devices also suffer from interference from the other cell center devices, however this interference is typically lower due to the larger distance separating the devices. Rcc Root 1 Root 2 Root 3 Root 4 Root 5 Root 6 Root 7 Root 7 Root 7 Figure 6 Interference management Option 3 IV. EXPERIMENTAL RESULTS A. Methodology and assumptions In this experiment, devices are either transmitting or receiving discovery signals, and the role of device does not change during simulation run time. D2D transmissions are taking place during reserved D2D subframes. To simplify the discovery signals are derived from ZC sequences using parameters from LTE PRACH signal. In the simulation, we use a 19 sites hexagon cell-site deployment where each site consists of 3 sectors. Within each sector, NTx D2D transmitting devices and NRx D2D receiving devices are dropped uniformly within the cell area. Several independent drops are performed to model UE mobility. For 1 The distance between UE and eNB can be estimated by the timing advance value. Globecom 2013 Workshop - International Workshop on Device-to-Device (D2D) Communication With and Without Infrastructure558 each drop, large scale fading path loss  π‘ƒπΏ!,!   between transmitting device l and receiving device m is calculated. The total inter-cell interference 𝑃!,! at receiving device 𝑙 can thus be expressed as: 𝑃!,! = πœŒβ„› ! ,𝒯 ! (π’ž π‘š − π’ž 𝑙 ) !!",!!"!,!!∈𝒰(!) (3) where β„› π‘š is the root sequence used by UE π‘š, 𝒯 𝑙 is the root of interest UE 𝑙 is trying to detect, π’ž π‘š is the cyclic shift of UE π‘š, 𝑃!",! is the transmit power of UE π‘š, and 𝒰 𝑙 is the set of all UEs for which the root used by the transmitting UE is different from the root of interest of the receiving UE 𝑙, i.e. 𝒰 𝑙 = ∀  π‘š  |  β„›(π‘š) ≠ 𝒯(𝑙) . Then, assuming the total interference at receiving device l is white2, the signal to interference and noise ratio (SINR) at device l for a discovery signal transmitted from UE π‘š is calculated based on the following equation: 𝑆𝐼𝑁𝑅! = !!",! !"!,!!!,!!!!! (4) where 𝜎!! is the additive white noise variance. Missed detection rate is then calculated as [4]: 𝑃! =   [𝐹!!(π‘Ž, 2𝑁! , 0)]!!!×𝐹!!(π‘Ž, 2𝑁! , 2𝑁!𝑁!"𝑆𝐼𝑁𝑅!) (5) where 𝐹𝝌𝟐(βˆ™) is the Cumulative Density Function (CDF) for a chi-squared distributed random variable, Na is the number of receiving antennas, D is the length of detection search window, π‘Ž is the detection threshold which is determined by false alarm rate:   𝑃!" = 1 − [𝐹!!(π‘Ž, 2𝑁! , 0)]! (6) At the receiver, the overall false alarm rate considering Ncs preamble sequences is:   𝑃!"_!!" = 1 − (1 − 𝑃!")!!" (7) In the simulations, 𝑃!"_!!"is fixed to 0.1% and the detection threshold π‘Ž is calculated from equation (6) and (7). The range here is defined as the distance within which 95% of devices will be successfully detected. The range is calculated based on (4) and (5) with (1-Pm) = 95%. TABLE I. lists the parameters used for the simulations. TABLE I. SIMULATION PARAMETERS Category Parameter Value Deloyment ISD 500 m Distribution of devices Uniformly Mobility 3 km/hr Centre carrier frequency 2 GHz System bandwidth (BW) 10 MHz Channel models Path loss model Indoor Hotspot NLOS [5] Shadowing standard deviation 4 dB Device parameters Antenna gain 0 dBi Noise figure 9 dB Antenna height 1.5 m Number of Tx and Rx antennas 1 Tx, 2 Rx Neighbor Length of discovery signal 800 µs 2 Sum of a large number of independent random variables is normally distributed according to central limit theory [4]. Category Parameter Value discovery Length of cyclic prefix 12.5 µs D 12.5 µs NRB 6 RBs Ndru 2 Discovery bandwidth Ndbw 12 RBs Number of cyclic shifts for one ZC sequence root (Ncs) 64 Target false alarm rate3 Pfa_Ncs = 0.1% Target detection rate4 Pd = 95% Target detetction range5 200 m NTx 30 per sector NRx 60 per sector Rcc Distance corresponding to the worst 10% cell edge UEs B. Simulation results Figure 7 illustrates the missed detection probability with respect to path gain (i.e., the inverse of path loss in dB) for all devices in the cell, only devices at cell edge (with option 1), and the missed detection probability for the case where no interference is present. It can be observed that cell edge devices in general suffer more from inter-cell interference than the average in the cell; for the same probability of missed detection, the cell edge UE require approximately 2dB more than the average and 3dB more than the case where no interference is present. In the following, the simulation results will focus on the cell edge devices only. Figure 7 Missed detection rate for all devices in the cell and cell edge devices Figure 8 shows the missed detection probability versus path gain for the three interference management options described in Section III. The results show that when compared to the baseline where there is no interference, an approximate detection performance loss of 3 dB is observed at Pm = 0.1% 3 This probability is used to determine the target detection range within which a device can be detected with a probability of Pd in AWGN. 4 This probability is used to determine the detection range. 5 Transmit power will be adjusted so that the target detection range can be achieved in AWGN. Globecom 2013 Workshop - International Workshop on Device-to-Device (D2D) Communication With and Without Infrastructure559 for option 1. From Figure 9, it can be observed that this corresponds to an approximate loss of 20 meters in range (here we ignore the shadowing effects). With the resource hopping in option 2, the performance is improved by 1 dB as compared to option 1, and with centralized cell edge device controller and resource splitting between cell centre and cell edge devices, the interference is almost eliminated. This is because the number of Tx devices in close proximity to the receiving devices trying to detect a UE from a different root sequence will be much smaller on average. We note that in option 3, as the number of devices increases, more root sequences will be assigned. An advanced sequence management scheme may be used (e.g., assign cell edge devices in the same geographic area with the same root and devices in other areas will be assigned with different root sequences). Figure 8 Missed detection probability versus path gain Figure 9 Interference management schemes comparison From above figures and discussions, we can conclude that interference among D2D transmissions will greatly degrade detection performance and needs to be coped with. Option 1 is the simplest interference management scheme but it results in performance degradation. Option 2 offers less gain than option 3 while it is simpler to implement as it only relies on some semi-static coordination between the cells and a known hoping pattern between the devices and Node Bs. In addition, for option 2 if more than 12RBs are available, additional resource hoping can be achieved and therefore, a larger gain can be anticipated. For option 3, even though it approaches and realizes the ideal scenario of no interfence at all, the level of complexity should be taken into consideration. In LTE networks we cannot always assume the presence of a fast central controller managing the resources of a group of eNBs. Additionally, separating cell edge and cell center devices requires the network to be aware of device locations within the cell at all times. This comes at the cost of increased signaling overhead due to frequent device measurement reporting, to assist the network with location determination. V. CONCLUSION In this paper, we first introduce the D2D neighbor discovery and the design aspects of discovery signals. Then, the interference caused between discovery signals is analyzed and three interference management techniques are proposed and compared. Final results show that centralized resource allocation method yields significant gains compared with other two techniques. We also demonstrate that using resource hopping is a feasible and effective way to manage inter-cell interference in practical systems. REFERENCES [1] RP-122009, “Study on LTE Device to Device Proximity Services,” 3GPP RANP#58, Dec. 2012. [2] S. Sesia, I. Toufik, M. Barker, “LTE-The UMTS long term evolution,” 2nd ed., John Wiley & Sons, Jan. 2011. [3] J. Rice, “Mathematical Statistics and Data Analysis,” 2nd ed., Duxbury Press, 1995. [4] R1-060998, “E-UTRA random access preamble design,” Ericsson, 3GPP TSG-RAN WG1 #44bis, Athens, Greece, Mar. 27-31, 2006. [5] 3GPP TR 36.814, “Further advancements for E-UTRA physical layer aspects”, V9.0.0, Mar. 2010. Globecom 2013 Workshop - International Workshop on Device-to-Device (D2D) Communication With and Without Infrastructure560