Energy Efficient D2D Discovery for Proximity Services in 3GPP LTE- Advanced Networks

Transcription

1 (c) 2014 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works for resale or redistribution to servers or lists, or reuse of any copyrighted components of this work in other works. Energy Efficient D2D Discovery for Proximity Services in 3GPP LTE- Advanced Networks Athul Prasad, Nokia Research Center Andreas Kunz, Genadi Velev and Konstantinos Samdanis, NEC Europe Ltd. JaeSeung Song, Sejong University Device-to-device (D2D) communications facilitate promising solutions for service optimization and spectrum/capacity efficiency in 3rd Generation Partnership Project (3GPP) networks. A key enabler for D2D communications and proximity services is the device discovery process. Currently, many aggressive peer discovery methods discussed in 3GPP could induce high energy consumption. Therefore, a critical objective in efficient D2D proximity service deployments is how to prolong the battery life time of User Equipments (UEs), by managing discovery cycles intelligently. In this paper, D2D discovery mechanisms discussed in 3GPP are studied in terms of energy consumption aspects. In particular, we consider an efficient D2D discovery mechanism that takes advantage of the concept of proximity area, a dynamic geographical region wherein UEs activate their D2D capabilities. The mechanism enables UEs to perform D2D discovery procedures only when there is a high probability to find other UEs subscribed to the same service. The energy consumption profiles of various discovery mechanisms are evaluated using simulations and the results indicate that significant energy savings can be obtained using the considered discovery mechanism. Introduction The introduction of D2D communication for 3GPP Long Term Evolution (LTE) [1] was initially undertaken in Release- 11, and later followed by the technical work started in Release- 12 under the name of Proximity Services (ProSe) [2]. Currently, the work on ProSe in 3GPP concentrates on commercial scenarios e.g. social network among friends, shops advertising, etc., and public safety deployments e.g. police, ambulance, etc. The ProSe functionality is used by a specific application in the UE (called ProSe application), i.e. not all UE applications can access the ProSe functionality. According to the requirements specified in Release-12, the ProSe communications for commercial scenarios are allowed only among UEs within the LTE coverage. The proximity services in 3GPP envisage two basic functionalities: ProSe discovery and ProSe communication. ProSe discovery precedes ProSe communication in time, while ProSe communication may use information acquired during ProSe discovery. The term ProSe as specified in 3GPP implies (i) different discovery mechanisms (EPC-level, WLAN direct, LTE direct), (ii) communication through Wireless Local Area Network (WLAN) direct D2D communication and (iii) using LTE direct D2D communication [2], [17]. EPC-level discovery is based on location tracking in the EPC network and assisting the ProSe UEs with discovery information. The direct discovery mechanisms are based on UEs sending announcements over the proximity radio interface and on ProSe UEs that are monitoring those. Currently 3GPP System Architecture Working Group 2 (SA2) is investigating efficient discovery procedures for devices subscribed to the same proximity service [2]. Various ProSe discovery mechanisms have been introduced in the technical report and these mechanisms can be roughly classified into the following two categories: Core network / evolved packet core (EPC) assisted discovery: The main task involves location tracking of UEs in the network and alerting the corresponding users, once multiple UEs with the same ProSe application, subscription and authorization for discovery are close enough to each other. This method allows similar battery consumption savings as in the legacy LTE system, since UEs avoid scanning and announcing themselves periodically. Direct discovery: ProSe-enabled UEs announce and/or listen to dedicated/shared LTE radio resources in order to discover each other s presence. It is currently assumed that the announcements contain the identity of a specific application and the UEs identity. UEs using the same ProSe application, subscription, and necessary authorization can then discover each other. As mentioned above, in order to facilitate ProSe services, a UE could have two roles in the direct D2D discovery: (i) announcing (i.e., periodically transmitting ProSe discovery message) and (ii) monitoring (i.e., periodically listening to ProSe announcing UEs and possibly sending discovery response to an announcing UE). The ProSe discovery and response messages contain ProSe UE identity and ProSe

2 application layer identity. Regardless of its benefits such as permanent availability at any time as well as operation in out of macro network coverage situations, the main challenge for direct D2D discovery is the energy consumption during the announcing and monitoring procedures. In particular, announcing UEs would consume a lot of energy for permanent periodic transmission of the discovery message. This paper addresses such energy consumption challenges, enabling an efficient direct D2D discovery procedure. The purpose of this paper is to study opportunities for optimizing the power consumption in order to extend the battery life of the UE. Periodic Direct discovery increases the accuracy of device location however because of the high energy consumption there is a strong need to introduce a new Direct discovery mechanism. The new mechanism requires to provide the means to avoid permanent periodic Direct discovery without degrading the accuracy of the device location. In order to satisfy the requirements, the D2D discovery approach, introduced in this paper adopts the concept of proximity areas (which is elaborated extensively in the section for Device Discovery and Proximity Areas), taking advantage of a kind of pre-defined proximity area that enables UEs to perform direct D2D discovery procedures only when there is high probability to find each other. The rest of the article is structured as follows. First we provide the standardized ProSe architecture, and an overview of the direct D2D discovery mechanisms studied in 3GPP, elaborating also the state of the art in this field. Next, various D2D discovery mechanisms utilizing proximity area for energy savings are presented illustrating the associated procedures. Further, the simulation study and results are analyzed to provide a comparative insight into the energy consumed for the conventional and proposed D2D discovery procedures. Finally, we conclude this article and provide some further research directions. State Of The Art The evolving trend towards proximity-based applications and services has created a new market drive for short-range communications, which gave rise to direct device or D2D communications. Taking advantage of the device proximity, D2D communications provide efficient network resource utilization besides the service performance benefits, including reduced delays and increased data rates. The efficient D2D communications introduce certain challenges in terms of interference control, especially when licensed spectrum is used. Peer discovery, is yet another significant issue, where a trade-off exists between the accuracy of the device location and the device energy consumption. Such key enablers for realizing D2D communications within an LTE-based environment are elaborated in [3], which details device synchronization and discovery issues as well as interference avoidance by power control considering a network assisted solution. An operator controlled D2D communication use cases and business models are analyzed in [4] considering a range of different degrees of D2D control including fully controlled, wherein the network is in charge of setting-up D2D communications, and loosely control, wherein device have the option to autonomously set-up direct communication. Our work enhances such earlier device discovery proposals considering not only D2D communications between specific devices, but also exploiting group D2D and proximity services. The notion of proximity services is introduced in [5], by allowing a direct device discovery mechanism referred to as Aura-Sense. Aura-Sense enables D2D communication via a licensed Time Division Duplex (TDD)-based air-interface called FlashLinQ, in combination with a namespace system, which permits application-centered discovery among close devices. LTE-direct [6] is yet another proposal for proximity services wherein devices use expressions at the application/service layer for discovery, while the network assists in device authentication and in coordinating radio resources used for discovery. A bio-inspired distributed discovery scheme for D2D proximity services is introduced in [7] based-on the firefly algorithm, focusing mainly on the device synchronization issues. In [8], a study is conducted regarding the architectural and protocol enhancements for integrating D2D communication into the 3GPP LTE-Advanced (LTE-A). Figure 1 a) 3GPP ProSe Reference Architecture; b) D2D Discovery inside a P-Area. In 3GPP, proximity services are currently being studied in [2] and specified in [9] and [17]. The relevant EPC entities here are: the mobility management entity (MME), home subscriber server (HSS), the serving and packet data network gateway (Sand P-GW). As shown in Figure 1.a, the architecture for proximity services is integrated in the mobile network

3 architecture introducing six new reference points (i.e., from PC1 to PC6), as well as a new network functional entity (ProSe Function), and a new UE type. ProSe Application Server (ProSe AS): is an entity outside of the scope the 3GPP. The ProSe application in the UE communicates with the ProSe AS via the application layer reference point PC1. The ProSe AS is connected to the 3GPP network via PC2 reference point. ProSe Function: is part of the 3GPP s EPC and provides all relevant network services like authorization, authentication, data handling, etc. related to proximity services. For ProSe direct discovery and communication, the UE may obtain a specific ProSe UE identity, other configuration information, as well as authorization from the ProSe Function over the PC3 reference point. ProSe UE: is able to discover and communicate with other ProSe UEs via PC5 reference point. The direct D2D discovery procedures are part of the PC5 reference point procedures. 3GPP defines two modes of direct D2D discovery: Network independent (autonomous) direct D2D discovery: no network assistance is needed for generating the ProSe UE ID, i.e. the UE uses a pre-configured ProSe UE ID. Network authorized direct D2D discovery: network assistance is always needed for generating the ProSe UE ID. Beyond the current notion of 3GPP proximity services, the work done in [10] considers Machine-Type Communications (MTC) device grouping with the purpose of synchronizing devices, and allowing them to communicate directly via D2D. Device grouping in this case is performed by a centralized application server, based on a priori information obtained from devices and from the network with the objective to enhance QoS and network resource utilization. Such a scheme can be useful if proximity services involve devices with low mobility that need to communicate towards an application server via the LTE network. If devices need to share resources, then the mobile cloud proposal elaborated in [11] may provide a solution considering the formation of device-based cooperative local networks, with the generic purpose of resource pooling. D2D communications and proximity services provide a generic awareness of the surrounding environment and services, which is feasible only once the appropriate device discovery is in place; a process that can potentially be energy consuming for devices. A low power consumption direct device discovery based on recursive binary time partitioning is introduced in [12]. Such a proposal is more generic, and thus not fully aligned with the 3GPP perspective, since it is not relying on network discovery, but requires sophisticated scanning and additional security procedures. The work done in [13] evaluates the device discovery power consumption in a social cloud environment, in comparison to a network based device discovery approach. In our approach, we consider combining direct D2D discovery with network support, and utilize the knowledge about UEs subscribed to the same ProSe service in order to provide energy efficient device discovery. Device Discovery and Proximity Areas Typically, the location of one of the ProSe UEs is known in advance, e.g. the UE belongs to a shop, or the location where the UEs would meet is known a priori. Hence, it is not necessary to constantly query the location of both UEs in order to analyze the potential of proximity. In case of the D2D functionality being turned on constantly regardless of the probability of success in proximity detection, it results in unnecessary UE battery drain for announcing, and listening for other UEs in proximity. The current Direct discovery concept has some drawbacks in battery drain or signaling overhead. To tackle such a problem, we introduce the concept of Proximity- Area (P-Area), as the area where two or more devices subscribed to the same ProSe service can meet with a high probability. The P-Area may be described also similar to the fingerprint (FP) mechanism for small or home enbs (HeNBs) for more accurate location detection within a cell. A fingerprint maybe based on the unique combination of the available information of e.g. PLMN ID, location area/tracking area and cell ids of the surrounding cells. But the key difference here is that, while small cells are stationary, D2D devices could be mobile as well, thereby requiring a dynamic location tracking mechanism. We assume that this information is present at the ProSe Function, within the 3GPP network. To resolve such a discovery challenge, P-Area(s) should be defined on a per ProSe application basis and according to various criteria including time, type of UE, group membership, etc. Each P-Area can be mapped to a geographical region described by one or more cell IDs, or other location information. This geographic information can then be used by a UE autonomously or in a network assisted manner, in order to compare whether it is entering or leaving its P-Area(s). Figure 1.b shows an example of how a ProSe UE seeking for a specific service could discover devices providing the desired service when it enters a P-Area. Once a ProSe UE detects that it enters a P-Area, it turns on its ProSe discovery and announces itself, while it is listening to the announcements of other UEs (step ). While the UE moves around the P-Area, it receives advertisements from other UEs in the P-Area (step ) and hence detects another device, named UE-T, which provides the desired service when UE-A gets close enough (step ). When the UE-A leaves the P-Area it turns off its ProSe discovery to save battery power (step ). A. Determination of the Proximity Area Once a UE registers to the ProSe AS for a specific application, the ProSe AS in coordination with the ProSe Function, should determine the right P-Area(s) for the UE's ProSe service. For this purpose, the ProSe AS gathers location information about the users subscribed to the same ProSe application and takes several parameters into account. Firstly, the ProSe AS differentiates whether the ProSe application is for e.g. a chat service, where the presence of the buddy list needs to be tracked, or whether the application is for

4 e.g. a shoe shop that announces its best offers of the day. Additionally, the mobility of the users has to be known, in order to be able to predict potential proximity service opportunities. Specifically, the ProSe AS may try to determine specific location patterns over time based on empirical mobility information of the UE from the network. The ProSe AS could also estimate P-Areas for a specific time window, e.g. when the subscriber is at work; determine closest UEs of the subscribed service and estimate/lookup P-Areas for each of them. The ProSe Function then provides the P-Area(s) to the UE via the PC1 reference point. For the calculation of a P-Area, the ProSe Function considers the location of UEs subscribed to the same ProSe application. The ProSe AS/Function updates the closest UEs of the subscribed services (excluding UEs with low potential of involvement) with the own P-Area(s) of the UE. B. Proximity Discovery Scenarios In this section we illustrate how a P-Area can be used for achieving energy savings in Proximity discovery within LTE based-on the following two scenarios: Scenario 1: target location is known, e.g., emergency call scenario where police, ambulance or fire brigade move to the emergency location. Scenario 2: target location is unknown, i.e. medium or high mobility UEs. For the both scenarios, we assume that there are two UEs, UE- A and UE-B, which subscribe to the same proximity service. In the scenarios, UE-A is approaching to a place (known as a P- Area) where UE-B is located in so that UE-A and UE-B are defined as Guest UE and Host UE, respectively. Scenario 1: In this scenario two cases are considered 1) two UEs are moving towards a specified common location or 2) one UE exhibits low mobility/static e.g. a shop, with a known location. Such a location is known by the ProSe AS and can be provided to the UEs in advance so that the UEs can turn on their ProSe direct discovery at the right time to detect each other. Figure 2 illustrates the procedures of how D2D discovery is performed within LTE when the target location is known with UE-A as Guest UE and UE-B as Host UE. Scenario 2 extensions are shown with dotted lines. In this example, UE-B is a low mobility device and indicates this during the registration or subscription (step 1). The ProSe AS then determines the location and the subscribed services/applications of UE-A and UE B in order to find the closest UEs that match this profile and create the corresponding P-Areas (steps 2 5). The ProSe AS may update for instance the other UEs accordingly with the information about UE-A. The ProSe AS and the UE-A then exchange the Proximity Configuration including the P-Area to the UE-A (step 6a), or alternatively requests UE-A to report its current fingerprint to the ProSe function (step 6b), and related acknowledgement (step 7). The UE-A or ProSe function based on reports from UE-A now compares whether its current location matches to any of the received P-Areas. When the UE-A is located within a P-Area, ProSe function requests UE-A and B turns on their ProSe Direct Discovery functionality (step 8), and if UE-B is close enough to the UE-A, ProSe Direct Discovery and connection setup takes place (steps 9 and 10). Here the Gateway Mobile Location Centre (GMLC) is assumed to interact with the MME for supporting location based proximity detection. In case UEs move towards the same location, e.g. police and ambulance to an accident location, then this is similar to the flow above with the difference that all UEs known by the ProSe AS that move to the emergency location are instructed by the ProSe AS and share the same P-Area, equal to the emergency location. For UEs located in a fixed place or having low mobility characteristics (Scenario 1), P-Areas can easily be decided. However, for UEs moving without a known pattern or towards a not known location, the complexity to provide P-Areas updates increases significantly. Handling such scenario should involve UEs location tracking to be performed regularly or alternatively allowing such UEs to announce/listen permanently whether other UEs with the same subscriber services/applications are within proximity. Figure 2 Proximity Discovery within LTE coverage (Scenario 1 and 2).

5 Scenario 2: In this scenario we consider that two UEs are moving with medium or high mobility, assuming the network is still able to estimate the P-Areas where the probability is high to meet. Since the distance wherein Direct discovery can take place is within the magnitude of meters, it is necessary that both UEs are at the time of proximity detection close enough and not moving. However, if mobility is high, there is a high possibility to lose coverage of the device to device radio and proximity detection cannot take place. As compared to static UEs, high mobility UEs could have P- Areas that may be dependent on time, e.g. the time a subscriber is going to work. Then the P-Area would be around the working place during the day, while at night the P-Area would be around his home area. Since a UE should switch on the Direct discovery also when it resides in its own P-Area, this would result in a permanent switched on Direct discovery while waiting for other UEs with the same service to pass by. This makes sense in case of a shop or fixed installed UE, however it is not desired for devices with high mobility. For this reason, such UEs, called Host UEs, are recommended to turn off their Direct discovery while residing in their P-Area in order to save battery power and wait for other UEs, called Guest UEs, with the same subscribed service to join. When these Guest UEs enter the P-Area where a Host UE is waiting, they trigger the ProSe AS to notify the Host UE in its own P- Area to switch on the Direct discovery so that both Host and Guest UEs can detect each other. When the Host UE leaves its own P-Area, it should notify the ProSe AS about its new location so that the ProSe AS can update this P-Area for the new closest Guest UEs with the same subscribed service. Then these Guest UEs adjust their Direct discovery accordingly. In Figure 2 the extensions for the mobility status determination are shown in step 4 and the notifying procedures between the ProSe AS/Function and the Host UE in the steps 9a 9d respectively. The fingerprint information assumed here is the measurement report consisting of the reference symbol received power (RSRP) and the corresponding cell IDs sent by a UE to the network, similar to the ones used in [15], [16]. For EPC-assisted discovery, we assume that the fingerprint information of the Host UEs and Guest UEs are presented at the ProSe function, based on measurement reports sent over the PC3 interface. The function estimates the proximity of Host and Guest UEs by comparing whether the RSRP values in the fingerprints are within a finite measurement offset of δ db. If the fingerprint matches such a measurement offset, host UEs are requested to start sending discovery beacons, and guest UEs initiate related measurements for discovering each other. The main overhead for such discovery would be the periodic sending of measurement reports to the ProSe function, in order to track the approximate location of UEs. Carrier Frequency Macro-Cell ISD Basic Radio Configuration Parameters [14] Shadowing Standard Deviation [db] Spectrum Allocation Max Transmit Power [dbm] 2.0 GHz 500 m UE Speeds [km/h] UE Distribution Min. RSRP for D2D Comm. Min. Distance UE-eNB / UE-UE [m] Fingerprint measurement offset, δ UE Sleep Power [Units/Sub-frame] / [W] Macro Cell 8 D2D UE 7 10 MHz Carrier Macro Cell 46 UE 23 Uniform, 50 UEs/cell -112 dbm 35 / 3 5 db 0.01 / Mobility Specific Parameters [15] Cell Reselection Time Window, TCR_max Number of cell reselections for medium mobility state, NCR_M Number of cell reselections for high mobility state, NCR_H 120 s 2 5 Table 1 System-level simulation parameters. Figure 3 a) UE transmit and receive power consumption model [14], [16]; b) Simulation state transition model. In order to avoid frequent measurement/location reporting, an optimized discovery technique is introduced that sends the fingerprint of the Host UE along with the measurement offset, δ, to the Guest UE subscribed to the same service by the ProSe function, utilizing the PC3 interface. Such discovery is referred

6 to as Optimized Discovery, since it is an optimization of the EPC-assisted discovery. Thus, Guest UEs report proximity to Host UEs only when the fingerprint matches with its own neighbor cell measurement report. Furthermore, mobility state based device discovery can also be considered, where discovery is initiated only if the UEs are in normal or low mobility state. Since at medium and high mobility states, when the number of cell reselections (NCR) within a time window (TCR), is greater than that for medium (NCR_M) or high (NCR_H) mobility state [15], the time spent by a UE within D2D communication region is small, energy spent on discovery procedure could be avoided for such UEs. Simulation And Analysis In this section we introduce the simulation setup, and analyze mainly the energy consumption results for the various direct discovery mechanism introduced in the previous section. The potential energy saving gains using possible enhancements are also presented A. Simulation Assumptions System level simulations using LTE-A network settings for D2D discovery and communication were conducted for evaluating the energy efficiency of the device discovery mechanisms considered in this work. The values and assumptions used are mostly in line with the agreements within 3GPP [14]. We consider a system of seven macro enbs, each having three cells. The main system level simulation parameters used are as shown in Table 1. The simulation model used, as well as energy consumption models are as shown in Figure 3. Fifty users are distributed uniformly in the system, with different states, state durations, and state transition probabilities as shown in Figure 3.b. There is 50 % probability for each UE to be either in sleep mode, or in active mode, engaging in D2D discovery or communication. The P-Area is defined as 10 meters around UEs sending the discovery signal. In order to estimate the approximate UE battery power consumed for discovery, we consider the UE transmit power to actual power consumption values for LTE UEs used in [16]. We consider one unit of Tx/Rx power to be approximately 0.4 W of battery power. The power consumption model in units per subframe is based on the agreed model in 3GPP for D2D communication [14]. In order to simulate both scenarios 1 and 2 with both static and mobile UEs, we define three UE states as shown in Figure 3.b. When a UE is in active mode, it could be: Stationary and sending discovery beacons so that other UEs can discover it (D-UE state), Mobile, i.e. moving at a constant speed scanning for discovery beacons (M-UE state), Searching for discovery beacons while being stationary (S- UE state). Simulations are run as snapshots with duration of 1000 s, with all UEs being idle in the beginning of each snapshot. For the network to keep track of UEs approximate location, the fingerprint mechanism considered in [15], [16] is adopted. Here EPC-assisted discovery involves sending UEs fingerprint (measurement report) information to the ProSe function with a fixed pre-configured periodicity. The information is sent over one subframe, consuming energy depending on the UL transmit power, as considered in [16]. The Optimized scheme simulates the scenario where the fingerprint information of the UE sending discovery beacon is sent to the S/M-UEs by the ProSe function whenever there is a change in the fingerprint. This operation is assumed to consume one downlink subframe, similar to measurement report configurations, with receive power consumption of 0.4 W or 1 Unit per subframe. Direct discovery mechanism with D-UEs transmitting discovery beacons with a configured periodicity is considered as the baseline mechanism. It is assumed that M-UEs and S- UEs are aware of the periodicity of the discovery beacon based on the system information from the enbs. The radio resource configuration used for direct discovery is performed by the network using for e.g. system information blocks broadcasted in a cell. The basic discovery procedure is assumed to remain the same for direct discovery, and for the proximity-based discovery considered in this work, following the assumptions agreed in [14]. For EPC-assisted and optimized discovery, D- UEs are requested to send discovery beacons only when there is an active D2D UE in their proximity, leading to energy savings. Discovery beacons are transmitted by D-UEs with maximum transmit power. The energy overhead for the EPC assisted location tracking is also considered based on the transmit power of the UE sending the measurement report. The cost for receiving the P-Area sent by the ProSe AS for the optimized mechanism is assumed to be similar to receiving the optimized fingerprint, as considered in [16]. Since D2D communication range is limited, we also consider the use of Mobility State Estimation (MSE) to limit fast moving M-UEs, i.e. to suspend them from searching for beacon signals. This is accomplished by adopting the parameters used for MSE based cell search scheme considered in [15]. M-UEs having medium or high mobility state suspend their search in order to conserve battery power. The fingerprint measurement offset value, δ=5 db was used, since that value gave essentially the same discovery performance for direct and fingerprint based discovery mechanisms. B. Simulation Results and Analysis In this section, we compare the performance of direct discovery with the EPC assisted and the optimized P-Area based discovery introduced in this paper. Various performance metrics for these schemes are as shown in Figure 4. Figure 4.a shows the distribution of power consumption for performing device discovery based on the assumptions given in Figure 3.a. From the figure we can observe that distributed direct discovery, where UEs send and search for discovery beacons

7 with a configured periodicity, consumes a significant amount of power, depending on the periodicity with which the beacon is sent. While having a fingerprint mechanism at the EPC or UE reduces power consumption significantly, the performance of the mechanism depends on the periodicity with which UEs are configured to send their measurement report to the network. With the optimized mechanism, the power consumption is further reduced, essentially removing the dependency on measurement reporting periodicity. From figures 4.b and 4.c, we can observe that significant power savings of up to 78% can be obtained by optimizing the discovery procedure, with approximately 10 % additional savings for the considered scheme, compared to fingerprint based discovery. Since discovery resources are assumed to be configured by the enb on the UL spectrum of the serving cell, this would lead to proportional reductions in resource utilization too. In Figure 4.d, the mean battery power consumption values are presented to estimate the approximate cost of the UE battery life due to various discovery procedures. Figure 4 a) Distribution of UE power consumption; b) Mean power consumption; c) Mean power savings using the EPC- Assisted and optimized mechanisms; d) Mean UE battery power consumption. The D2D connection time values for M-UEs are as shown in Figure 5, where Direct-250 ms and optimized discovery mechanisms are used. The distribution of D2D connection time using direct discovery for different UE speeds is as shown in Figure 5.a. From the figure, we can observe that due to the short range of D2D communication, the connected times are reduced significantly as UEs speed increases. Hence, we suspend the beacon search for UEs in medium or high mobility state, as shown in Figure 5.b. As we can observe from the figure, the mean connected times indicate that viable communication is not possible for UE speeds above 3 km/h. The results also demonstrate that having a UE-based MSE mechanism as currently defined in LTE-A would also enable autonomous decision-making in terms of searching for beacons. From the figures, it could be further observed that the discovery performance of optimized scheme is almost similar to the direct discovery scheme at low mobility. Here we assume that D2D communication occurs only within the P- Area defined for the UEs. From the evaluation study performed, it could be concluded that having fully distributed direct discovery could lead to higher battery consumption, as well as sub-optimal network resource utilization. Sending discovery beacons infrequently, for e.g. every 1 s, can be a potential solution to overcome this problem but this may also lead to negative impacts on D2D connected time periods. Defining P-Areas based on the requirements and related use cases of ProSe can enable higher battery power savings with minimal impacts on standards. While having fingerprint based proximity determination, and discovery also leads to better system performance in terms of battery power consumption and uplink resource utilization, frequent location reporting by the UEs could significantly impact the performance. The reduced power consumption when using the proposed solution of P-Areas may reduce the success of ProSe discovery, depending on the accuracy of fingerprint information and the radio environment. This is especially valid in case of mobile Host UEs where fingerprints change and the proximity areas are not correctly configured. The frequent fingerprint information transmissions from ProSe function to Guest UEs and location information reporting from Host UEs to ProSe function for P-Areas determination would lead to higher signaling load in the network as compared to direct discovery. Another drawback of the proposed solution is that it does not work in case of out of cell coverage scenarios where the Guest and Host UEs cannot exchange information with the ProSe function, i.e. the P-Areas cannot be updated. Since such scenarios are considered relevant only for public safety scenarios, using UEs having higher transmit power, and battery capacity, direct discovery with discovery beacons sent infrequently could be used as an enabler for device discovery. Figure 5 a) D2D connected time distribution for M-UEs; b) Mean D2D connected time for various M-UE speeds.

8 Conclusions In this article, we presented an overview of D2D discovery mechanisms currently being studied and standardized in 3GPP. We have also considered an approach where UEs activate their D2D capabilities only when there is a high probability to be in the proximity of other UEs, so that UEs can achieve significant energy consumption savings, reduce overall signaling in the network and further optimize network resource utilization for the discovery procedure. In the simulation analysis, we evaluated the energy consumption for D2D discovery mechanisms, and the results show that by using the proposed proximity area based D2D discovery approach, the overall power consumption of a UE can be reduced up to 78% compared other cases where UEs use conventional D2D direct discovery mechanisms. Further work in this area could entail investigating the discovery performance in more challenging indoor scenarios, where UEs could be distributed in 3D space. Investigating the discovery performance when proximity region covers multiple cells and out-of-coverage public safety scenarios with D2D UEs acting as relays would also be interesting areas for further study. Network Deployments, IEEE Comm. Mag., vol. 51, no. 5, pp , May [16] A. Prasad, et al., Enhanced Small Cell Discovery in Heterogeneous Networks Using Optimized RF Fingerprints, in 24th Annual IEEE PIMRC, London, Sept [17] TS Proximity-based services (ProSe); Stage 2 References [1] TS Evolved Universal Terrestrial Radio Access (E- UTRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Overall description; Stage 2. [2] TR , Study on architecture enhancements to support Proximity Services (ProSe). [3] G. Fodor, et al., Design Aspects of Network Assisted Device-to- Device Communications, IEEE Comm. Mag., Vol.50, No.3, Mar [4] L. Lei, Z. Zhong, C. Lin, X. Shen, Operator Controlled Deviceto-Device Communications in LTE-Advanced Networks, IEEE Wir. Comm. Mag., Vol.19, No.3, Jun [5] M.S. Corson, et al., Towards Proximity Aware Interworking, IEEE Wir. Comm. Mag., Vol.17, No.6, Dec [6] S. Balraj, LTE Direct Overview, Qualcomm Research, Jul [7] S-L. Chao, H-Y. Lee, C-C. Chou, H-Y. Wei, Bio-Inspired Proximity Discovery and Synchronization for D2D Communications, IEEE Communications Letters, Vol.17. No.12, Dec [8] B. Raghothaman, et.al, Architecture and Protocols for LTEbased Device to Device Communication, IEEE ICNC, San Diego, Jan [9] TS General Packet Radio Service (GPRS) enhancements for Evolved Universal Terrestrial Radio Access Network (E- UTRAN) access. [10] K. Samdanis, A. Kunz, M.I. Hossain, T. Taleb, Virtual Bearer Management for Efficient MTC Radio and Backhaul Sharing in LTE Networks, IEEE 24th PIMRC, London, Sept [11] M.J. Salehi, et al., Mobile Cloud management: A New Framework, IEEE 23rd PIMRC, Sep [12] D. Li, P. Sinha, RBTP: Low Power Mobile Discovery Protocol through Recursive Binary Time Partitioning, IEEE Transactions on Mobile Computing, vol. 99, [13] A. Prasad, K. Samdanis, A. Kunz, and J. Song, Energy Efficient Device Discovery for Social Cloud Applications in 3GPP LTE- Advanced Networks, 19th IEEE ISCC, Madeira, June [14] TR , Feasibility Study on LTE Device to Device Proximity Services - Radio Aspects. [15] A. Prasad, et al., Energy-Efficient Inter-Frequency Small Cell Discovery Techniques for LTE-Advanced Heterogeneous