Smartphones are equipped with multiple radio interfaces that enable them to access different types of wireless networks, including WLANs, Bluetooth and Zigbee, besides cellular networks. Emerging device-centric systems (DCS) such as devices-to-device communications are considered standard components of future mobile networks, where operators/consumers involve their devices in direct communications to improve the cellular system throughput, latency, fairness and energy efficiency. However, the battery life of the mobile devices involved in such communications is crucial for 5G smartphone users to explore the potential of emerging applications in DCS. It is anticipated that the owners of 5G-enabled smartphones will use their devices to talk, text, e-mail and surf the Internet more often than the customers with 4G smartphones and traditional handsets, which puts a significantly higher demand on the battery life. This chapter introduces a new scheme to support emerging features in DCS, where a device-to-device (D2D)-enabled mobile device (sink device or a content requester) aggregates the radio resources of multiple mobile devices (source devices or content providers) to improve the file transfer latency (FTL), energy efficiency and battery life. This scheme is referred to as devices-to-device (Ds2D) communications. In such a networking setting, this chapter discusses a network-controlled algorithm for optimal selection of source devices and their respective radio interfaces to support green Ds2D communications. Ds2D communications ensure an optimal packet split among the source mobile devices to reduce the FTL and hence to prolong the mobile battery life. Simulation results demonstrate that the proposed optimal packet split scheme guarantees an improvement in the mobile battery life over a wide range of data rate levels in comparison with the random packet split strategy and the traditional D2D communication paradigm between the sink and source mobile devices.
The recent widespread use of mobile Internet complemented by the advent of many smart applications has led to an explosive growth in mobile data traffic over the last few years. This remarkable growing momentum of the mobile traffic will most likely continue on a similar trajectory, mainly due to the emerging need for connecting people, machines and applications in an ubiquitous manner through the mobile devices. Every new release of an iPhone and Android smartphone spurs new applications and services, with advanced display screens to deliver an exceptional quality of experience to theend user. As a result, the current and projected dramatic growth of mobile data traffic necessitates the development of fifth-generation (5G) mobile communications technology. The 5G communications will provide us with the promise of a mobile broadband experience far beyond the current 4G systems. The 5G technology has a broad vision and envisages design targets that include 10–100 peak date rate, 1,000 network capacity, 10 energy efficiency and 10–30 lower latency [264]. In order to achieve these expectations, operators and carriers are planning to leverage emerging device-centric systems (DCS) such as device-to-device (D2D) communications, small-cells and nano and elastic cells to improve the user experience and consequently improve the overall network performance. However, the evolution of mobile devices to support the emerging features in DCS comes at a cost that places stringent demands on the mobile device battery life and energy consumption [265]. Hence, there are considerable market interests on the development and deployment of innovative green and smart solutions to support emerging features in DCS in ultra-dense heterogeneous networks.
From 2G to 4G, systems are based on network-centric approaches, but 5G systems will drop this assumption and move toward DCS. It is envisioned that the 5G networks will be mostly deployed for data-centric applications rather than voice-centric applications. The main drivers of DCS are the Internet of things (IoT), machine-to-machine communications and BigData applications, which will exploit the intelligence at the mobile device side to support the emerging device-centric communication paradigms and ensure ubiquitous connectivity.
D2D communication is considered a promising technology to complement the 5G DCS. As shown in Figure 10.1, traditional D2D communications take place among two devices, that is, a pair of devices and such that a direct communication link is established between the two mobile devices without any interaction from the BSs or the core of the cellular network. In [266], the authors have provided a literature review on D2D communications, including new insights concerning existing works and emerging protocols. This study includes a review on the inband (underlay or overlay in cellular spectrum) and outband (unlicensed spectrum) integration of D2D communications. In the literature, outband D2D communication uses a cellular interface to set up the connection and the WiFi interface for data transmission between the two devices involved in the D2D communication. Another form of D2D communication involves a pair of devices communicating over multiple interfaces, that is, a pair of devices and performing data transmission over both cellular and WiFi interfaces in a D2D set-up (multi-homing D2D pair). Researchers are still formulating the design objectives as optimization problems, but leaving them unsolved due to their NP-hardness. Consequently, most of the proposed algorithms such as the heuristic algorithm [267] and linear/nonlinear/dynamic algorithms [267–269] are subjects open for investigation for new optimal solutions for pairing the devices involved in such communications. D2D communication is also considered as a traffic offloading technology and has received much attention from the operators. However, the feasibility of its large-scale implementation and integration into an ultra-dense heterogeneous communication infrastructure is still an open research problem.
Direct D2D communication between cellular equipments is proposed to increase data rate and extend conventional cellular coverage. In an underlay scheme, the D2D communication may generate interference to the neighbouring cells due to the reuse of the same resources. Therefore, in the underlay approach, D2D links may only existif they do not harm the SINR at the BSs (uplink) or at the other devices (downlink) in the conventional communication approach. Researchers have proposed different interference management algorithms to increase network capacity [270 271]. For instance, the authors of [270] proposed that the D2D users monitor the received power of the downlink control signals to control their uplink transmit power below a threshold to avoid high interference to cellular users. If the required transmission power for a D2D link is higher than an interference threshold, then the D2D link is forbidden. One of the proposed solutions for future applications and services in DCS is to reduce the time-average interference power over different networks for both conventional users (communicating through BSs and access points (APs)) and D2D users.
In [272], the authors proposed a new LTE-A-based D2D communication network architecture. They have introduced a new reference point between D2D-enabled devices named ‘Di interface’ using enhanced radio protocols. The following D2D-specific functionalities are supported by many functions of this interface: (i) the D2D scheme should have the ability to measure the distance between two mobile devices to assess the feasibility of direct connection; (ii) the devices in the D2D architecture should be covered by the eNodeBs to maintain control and signalling and (iii) D2D data transmission between the devices should utilize a physical channel similar to the LTE-A uplink/downlink shared channel.
Some challenges to implement and integrate D2D communications into 5G networks are listed below [266]:
The opportunity of enabling multiple radio interfaces including WLANs, Bluetooth and Zigbee, besides cellular networks, is not fully exploited in D2D communications, since the D2D communications take place over a single link between two mobile devices involved in a direct communication. Enabling D2D data transmission between multiple source mobile devices and a sink mobile device over multiple radio interfaces is referred to as devices-to-device (Ds2D) communication. As an example, Figure 10.1 shows that the source mobile devices , and are involved in Ds2D communication with a sink mobile . Ds2D communication can take advantage of the diverse resources available at different radio interfaces (e.g. the supporting bandwidth). Aggregating such radio resources at the sink device allows for an improved system performance in terms of the achieved throughput, latency and energy efficiency.
Consider a system model with a single-sink mobile device and a set of candidate source mobile devices. The sink mobile device is required to download a file (content), which is cached in the source mobile devices. Let denote a set of mobile devices that are in the coverage area of a single cellular network base station (BS). Four communication modes can be distinguished in such a network setting, as shown in Fig. 10.1:
A network-controlled Ds2D communications approach is considered. Hence, in Ds2D communications, the sink mobile device requests a given (popular) file from the BS and indicates that it can operate in a Ds2D communication mode. The BS broadcasts the file request message to the mobile devices within the sink device proximity. On the basis of the mobile devices feedback, the BS defines a set of candidate source devices that (i) are within the proximity of the sink device, (ii) have a copy of the (popular) file required by the sink device and (iii) are willing to contribute in such a Ds2D communication. Then, the BS selects (from the available candidate source devices) the optimal source devices and their respective radio interfaces that deliver the required file to thesink device in the most energy-efficient manner. After optimal selection of source devices and their respective radio interfaces, the BS coordinates which source device transmits which chunk of the required file. The sink device aggregates the data chunks transmitted by different source devices. This approach can support data hungry applications such as file download or video streaming.
As a first step of research, we consider a system model with a single sink device and a set of candidate source devices. Let with representing the sink device and representing the candidate source devices. Each mobile device has a set of distinct radio interfaces . Radio interface in all mobile devices employs the same access technology. For instance, represents cellular radio interface in all mobile devices, represents an LTE direct radio interface, represents a WiFi direct radio interface and so on. Let be a binary variable that indicates if the sink device communicates with source device over radio interface for data transfer.
The transmission bandwidth that can be supported at radio interface for is denoted by . Each source device communicates with the sink device over radio interface using transmission power . Let represent the power amplifier efficiency for each source device. The circuit power consumption for source device and radio interface scales with the transmission data rate via [273]:
where and are two constants, measured in watts and watts per bit per second (bps). The total power consumption for source device to communicate over its radio interface is given by
Let and represent the distance and path-loss exponent between the sink device and source device , respectively. Denote by the Rayleigh random variable associated with the channel between the sink device and radio interface of source device . The channel power gain is given by
The average channel power gain between the sink device and radio interface of source device is denoted by .
Each radio interface of the sink device suffers from interference imposed by other mobile devices communicating over that specific band. Let denote the set of mobile devices interfering with the sink device file reception over radio interface . The distance between the sink device and the source of interference is denoted by and denotes the path-loss exponent. Let denote the transmission power of interferer over radio interface . The interference power over radio interface of the sink device is approximated by a Gaussian random variable with zero mean and variance . The one-sided noise power spectral density is represented by .
In this section, the problem of optimal selection of source devices and radio interfaces is formulated and an algorithm is presented to solve it.
The selection criterion of a given radio interface of source device is the average achieved energy efficiency , which is a ratio between the average achieved data rate and the average power consumption. Using Shannon's formula, the achieved data rate over radio interface of source device is given by
The average achieved data rate on the link between the sink device and source device over radio interface is given by [224]
where denotes the expectation and denotes the exponential integral. From Lemma 2.1 in [224], a lower bound of the average achieved data rate is given by
Hence, the average achieved energy efficiency on the link between sink device and source device over radio interface is given by
The objective is to select the source devices and their respective radio interfaces that maximize the total energy efficiency, that is,
The total number of links used for data transmission is upper bounded by the maximum number of available radio interfaces , excluding the cellular radio interface that is used for coordination, that is,
Furthermore, only one source device is allowed to communicate with a given radio interface of the sink device, that is,
For Ds2D communications, each source device employs only a single radio interface for data transmission; thus, we have
The summation over in (10.11) excludes the cellular radio interface, which is used for coordination.
Hence, the optimal selection of source devices and radio interfaces for green Ds2D communications is obtained by solving the optimization problem
One way to solve (10.12) for Ds2D communications is based on the ascending proxy auctions [274]. In this context, each source device defines a set that includes pairs of candidate radio interface and the achieved average energy efficiency over that interface, that is, , which excludes the cellular radio interface that is used for coordination. Define one element of by , for example, and a selection is given by , that is, . Each source device ranks based on . Let denote a strict preferenceordering over based on . All candidate source devices report such a preference order over the cellular radio interface to the cellular BS, which will be in charge of selecting the optimal combination of source devices and radio interfaces.
Let set denote a feasible selection set of source devices and their respective radio interfaces that satisfies the constraints in (10.9)–(10.11). The BS can form the feasible selection set by considering possible combinations of elements for all () and eliminating those combinations that do not follow the constraints in (10.9)–(10.11). For a given source device, if , then device is not selected to contribute to the Ds2D communication session (i.e. for that device ). Furthermore, means that no source device contributes to the Ds2D communication session and the sink device receives the requested file from the cellular BS via cellular communication. The cellular BS specifies a preference ordering over the set of feasible selection profile based on the total average energy efficiency (i.e. ).
The ascending proxy auction works over iterations () until the optimal selection of source devices and their respected radio interfaces is obtained. Define a bid as the proposed element from devices at iteration , i.e., and . Define as the set of bids (radio interfaces and average energy efficiencies) offered by source device till iteration , that is, . Let . The set of available new bids by device is denoted by , that is, feasible radio interface and corresponding energy efficiency that have not been offered till iteration . The optimal selection of source devices and their respective radio interfaces for Ds2D green communication is described by Algorithm 10.4.16, which is executed by the cellular BS. From Theorem 1 in [274], the selection made by Algorithm 10.4.16 is a stable (NTU-core) selection with respect to the reported preferences.
In Algorithm 10.4.16, each source device first updates its new available bids that can be offered in iteration . If there exists a source device with and still has new bids to offer (i.e. ), the source device will offer the most preferred radio interface to participate in the Ds2D communication (the preference order here is based on the source device most energy-efficient radio interface). The source device also updates the set of bids offered until iteration (). All other devices make no new bid at this iteration. The BS updates the set of feasible bids at the current iteration () and then selects the most energy-efficient set of source devices and radio interfaces (the selection here is made based on the total average energy efficiency ).
This section presents comparative simulation results for green Ds2D, multi-homing D2D and conventional D2D communications. The optimal selection of source devices and their respected radio interfaces for the Ds2D is implemented using Algorithm 10.4.16. For conventional D2D communications, only the source device and radio interface offering the maximum energy efficiency are selected for data transfer. For multi-homing D2D, the source device achieving maximum total (sum) energy efficiency across all its radio interfaces is selected for data transfer. All mobile devices have two radio interfaces besides the cellular radio interface (i.e. ). In all three modes, coordination is established over the cellular radio interface () and data transfer can take place over the other radio interfaces (). The candidate source devices are uniformly distributed within the proximity of [50,100] m away from the sink device. The supporting bandwidth for the radio interfaces used for data transmission are MHz and MHz. Each radio interface of the sink device is subject to a random numberof interferers uniformly distributed in the range [5,10]. The interferers are assumed to be close to the sink device (for a worst-case scenario), that is, uniformly distributed within the proximity of [50, 60] m away from the sink device. The transmission power is 100 mW for and . The power amplifier drain efficiency is . The circuit power constants are mW, mW, W/bps and W/bps. The path-loss exponent equals 4 for and , and dBm/Hz.
Fig. 10.1 shows the achieved average energy efficiency versus the number of candidate source devices. With more candidate source devices, a better energy efficiency can be achieved due to the diverse channel conditions among the candidate source devices and the sink device. Both Ds2D and multi-homing D2D communications exhibit an improved energy efficiency performance compared with the conventional D2D communication (up to improvement in energy efficiency). This is mainly due to the aggregated resources at the sink device from multiple radio interfaces, which allows for higher achieved throughput and hence improved energy efficiency. Such an improvement is also due to spatial diversity as some differences are expected in the channel conditions among the sink device and different source devices for Ds2D communications. As shown in Fig. 10.1, Ds2D communications exhibit a closer performance to multi-homing D2D communications as the number of candidate source devices increases. This is due to the higher probability of having more than one source device with good channel conditions with the sink device. While Fig. 10.1 shows a slightly improved performance for multi-homing D2D over Ds2D communications in terms of the total energy efficiency, the next result shows that Ds2D communications is an attractive alternative as it exhibits a much lower energy consumption per source device. Such an option motivates source devices to contribute to D2D communications.
Fig. 10.2 shows the average energy consumption performance per source device to transfer a 1 Mbit-file to the sink device versus the number of candidate source devices. The worst energy consumption performance per source device is for the conventional D2D communications approach, since only one radio interface is used for data transfer, which results in a longer latency to transfer the file to the sink device, and that results in a higher energy consumption. On the contrary, Ds2D communications exhibit the least energy consumption per source device (up to compared with the conventional D2D communications and up to compared with the multi-homing D2D communications). This is mainly because Ds2D communications split the total energy consumption burden over different source devices contributing to the file transfer, while multi-homing D2D communications relies on a single source device for file transfer, which incurs a higher energy consumption to activate all radio interfaces and transmit across them. With more available radio interfaces at the sink device, additional energy saving is expected per source device when compared with multi-homing D2D, as more source devices will be involved in the file transfer.
After optimal selection of source mobile devices and their respective radio interfaces, the BS coordinates with the source mobile devices to transfer the desired data packets to the sink mobile device in a distributed manner. The sink mobile device aggregates the data packets transmitted by different source mobiledevices to reconstruct the required file. This approach can support data hungry applications such as file download or video streaming of a popular content.
The optimal packet split algorithm should specify the packet distribution ratio among the source devices based on the achieved data rates over their respective radio interfaces. Consider that the desired file has long data packets, which should be transmitted from the source mobile devices (e.g. and as shown in Figure 10.1) to the sink mobile device () over a set of two different radio interfaces , as shown in Figure 10.1. Let denote the optimal packet split ratio (OPSR) that splits the requested file into two sets of data packets based on the achieved data rate for each selected source device. Set 1 of data packets contains data packets that are transmitted by source mobile device through radio interface . Similarly, set 2 of data packets contains data packets that are transmitted by the source mobile device through radio interface . The sink mobile device receives the packets from both source mobile devices simultaneously over two different radio interfaces () and combines them to restore the requested file.
The two source mobile devices and transmit with different data rates and , respectively, depending on the SINR of each source mobile device at the corresponding radio interface.1 The file transfer latency at the sink mobile device is defined as the duration required to transfer the desired data packets from all source mobile devices to the sink mobile device by aggregating the multiple radio resources, and is given by
where denotes the number of data packets transmitted over the radio interface (using , we have and ), denotes the data rate over the radio interface and denotes the number of bits per data packet. It is assumed that each data packet contains 1,500 bytes. From (10.13), the file transfer latency is minimum if all source devices complete their data transmissions at the same time. Hence, the main rationale behind the search of is to ensure that the source devices involved in Ds2D communications complete the file transfer at the same time such that the sink mobile device does not have to wait for one source mobile device to complete the transmission of its assigned data packets, which elongates the communication session and leads to a higher energy consumption. Thus, can be found by solving .
In order to evaluate the effectiveness of the proposed optimal packet split strategy over the two radio interfaces of two source mobile devices, we consider an average monthly data usage capability for each mobile subscriber of about 2.5 GB with the daily download capability of 80 MB. Given the fact that each data packet has 1,500 bytes, the file (requested content) has K data packets. The average data rate achieved for the second source mobile device () over radio interface is assumed to be 1.646 Mbps. Moreover, the average achieved data rate for the first source device () over radio interface () is varied for performance evaluation. Figure 10.3 shows the optimal packet split between a pair of source devices () over two radio interfaces for different data rates achieved by . The optimal packet split algorithm divides the data packets among the two source mobile devices based on the achieved data rate for each source mobile device. This is mainly because the optimal packet split algorithm ensures the same FTL at each source mobile device, as shown in Figure 10.4. It turns out that the FTL is dominated by the device suffering from the maximum FTL.
Another performance evaluation criterion is the relative percentage reduction in FTL (i.e. relative gain) as compared with the transmission over the conventional D2D communication paradigm, where only one source mobile device transmits the complete file to the sink mobile device, that is, direct D2D communication between a pair of devices ( and ). The performance of the optimal packet split algorithm is evaluated against a random packet split benchmark. The random packet split benchmark algorithm randomly divides the file among the two source mobile devices and each source mobile device transfers the packets to the sink mobile device that combines both sets to restore the requested file. It can been seen clearly from Figure 10.5 that Ds2D transmission with optimal packet split has a lower transmission FTL than the conventional D2D paradigm (there is always a gain, which ranges from to FTL reduction). Moreover, the optimal packet split is necessary for performance improvement in the Ds2D paradigm, as the random packet split can have an FTL performance worse than the conventional D2D paradigm (gain is below for the data rate levels contained in the range 8–15). Furthermore, as shown in Figure 10.5, with the increase in the data rate level for , the achieved gain is reduced. This is mainly because with high data rates achieved for , a single transmission link (between and ) is already sufficient to achieve a lower FTL than the Ds2D communications (among , and ).
In this subsection, we present simulation results to show the performance of the proposed Ds2D communication under optimal packet split and random packet split schemes. The energy consumption of Ds2D communication is compared with the D2D communication scenario when the sink mobile device receives the complete file from only one of the source mobile devices over the direct communication link. As an example, consider a source mobile device such that its battery holds a charge of mAh with Wh [265].
Energy consumption of the source mobile device for transferring a file to a sink mobile device can be determined as follows:
where is the FTL per source mobile device measured in seconds to transfer the desired content to the sink mobile device. Figure 10.6 shows the energy consumption per source device involved in transferring optimally the assigned data packets out of the file of size 80 MB (or equivalently K data packets) to a sink device over the range of date rate levels. Compared with the direct D2D communication between a pair of devices, the proposed Ds2D communication offers reduced energy consumption per source mobile device, since each source device only transmits a fraction of packets of the requested file. However, the energy consumption of the source mobile devices involved in Ds2D communication with an optimal packet split scheme outperforms the energy consumption of the source mobile devices that adapt the random packet split scheme. Moreover, at lower date rate levels, the energy consumption of source mobile devices involved in Ds2D with an optimal packet split scheme is significantly reduced in comparison with the energy consumption of the source mobile devices involved in Ds2D communication with a random packet split scheme and traditional D2D communications. The improvement is due to the fact that the source mobile device is engaged with the sink mobile device for a relatively longer duration at a lower data rate level to complete the transfer of the required file under direct D2D communications in comparison with the source mobile devices involved in Ds2D communication. As an example, at a rate level 2, that is, kbs, the source devices can achieve reductionin energy consumption under the optimal packet split scheme and reduction under the random packet split scheme in comparison with the source device involved in D2D communications. As shown in Figure 10.6, Ds2D communications with optimally assigned data packets exhibit a closer performance to D2D communications at higher data rate levels. This is due to the fact that with the high data rates exhibited by the source mobile device, a single transmission link (e.g. between and ) is already sufficient to achieve low FTL as compared with the Ds2D communications.
Reducing energy consumption of mobile devices lowers the electricity cost for charging devices, and thereby, results in financial cost savings to the consumers if the energy savings offset any additional costs for implementing an energy-efficient framework. The monthly cost of electricity that is associated with the implementation of Ds2D communications is calculated by assuming 1 kWh = 12 cents and it assumes the expression
Figure 10.7 shows the monthly electricity cost associated with the energy savings achieved per source mobile device for the transfer of a 80-MB file over the considered device-centric framework. It can be seen clearly that at an average price of 12 cent/kWh, the mobile device costs approximately 150 USD in addition to the monthly electricity bill of the consumers who assume a pair of devices that are involved in D2D communications with an average daily data usage or file/content transfer of size 80 MB. On the contrary, the electricity cost decreased to 70 and 19 USD, when the devices are involved in Ds2D communication with random packet split and optimal packet split schemes, respectively, assuming the same amount of daily data transfer. In the presence of high data rates achieved by the source mobile devices, a single transmission link (e.g. between and ) is sufficient to achieve low FTL as compared with the Ds2D communications. This fact explains the close performance of Ds2D and D2D communications at high data rate levels.
Energy efficiency has been recently marked as one of the alarming bottlenecks in the telecommunications growth paradigm mainly due to two major reasons, namely (i) slowly progressing battery technology [276] and (ii) dramatically varying global climate [9]. A recent survey reports that up to 60% of the mobile users in China complained that the battery consumption is the greatest hurdle while using 4G services [277]. Emerging device-centric frameworks can offer a longer battery life, while consumers can enjoy high data rate 5G services and applications. The battery life or battery capacity can be calculated from the input current rating of the battery and the load current of the battery charging circuit [265]. Battery life will be high when the load current is low and vice versa. The capacity of battery is given by DigiKey Electronics [278]
where is the battery capacity in mAh, and is the load current drawn by the source mobile device for transferring the file to the sink mobile device. Here, the factor 0.70 represents external factors that can affect the mobile device battery life [278]. Figure 10.8 shows the mobile battery life (h) over the range of data rate levels for a mobile device involved in D2D and Ds2D communications. Overall, as the data rate level increases, the FTL is decreased, and hence the battery life is prolonged. However, it can be seen clearly that the battery life of a mobile device involved in Ds2D communications with an optimal packet split scheme is significantly higher than the battery life of a mobile device involved in Ds2D communication via a random packet split scheme and traditional D2D communication. Moreover, the battery life of the mobile device involved in Ds2D communication and that assumes the random packet split scheme degrades at high data rate levels, since the FTL performance of the random packet split scheme is worse than that of the traditional D2D communications (as can be seen from Figure 10.4).
In general, Ds2D communications can be established among any sink mobile device and multiple source mobile devices over multiple radio interfaces. Selection of source mobile devices is highly dependent on the availability of the file orcontent and its close proximity with the sink mobile device. As discussed earlier, coordination among the involved mobile devices is required for the successful implementation of Ds2D communication and optimal distribution of the desired content (data packets) among the source mobile devices. There are two possible implementation approaches to achieve the coordination among the mobile devices and set up Ds2D communications, as described below.
Under the centralized Ds2D set-up, cloud radio access networks (CRAN) perform source mobile device selection, Ds2D link establishment and data packet distribution among the source devices with a limited or full supervision of cellular network. Devices involved in Ds2D communication perform full or limited information exchange and signalling with the cellular network using the LTE–Uu interface (i.e. cellular link). Since the cellular interface for all devices is reserved for information exchange and signalling, data transmission can be established between a sink mobile device and source devices over radio interfaces. Therefore, the devices have at least two active interfaces (cellular interface for control and an additional radio interface for data transmission). Mobility of the devices involved in Ds2D communication, interference management and content availability is considered as an advantage to integrate Ds2D communications under the centralized CRAN-enabled cellular system. Inter-network caching plays an important role to efficiently exploit the benefits of Ds2D communications. However, the centralized approach imposes additional challenges to the fronthaul requirements such as high data rate and latency due to the information exchange and signalling overheads between the devices involved in Ds2D communications. Moreover, devices cannot establish Ds2D communication links without full or limited intervention and approval of the request by the cellular BS.
Under the decentralized Ds2D set-up, devices involved in Ds2D communications can exchange control signalling for selection of source mobile devices, Ds2D communication establishment and content distribution among the devices without any intervention from the cellular BS. Therefore, devices can establish Ds2D communications over a relatively short time period under a decentralized system in comparison with the time required to set up Ds2D links in a centralized manner. The cellular network does not have any supervision over the functionalities used by the devices involved in Ds2D communications, such as resource allocation and interference management. Devices can use the PC5 interface, which is allocated by the LTE standard for device discovery and Ds2D communication between users.Moreover, the fronthaul requirements can be relaxed due to the reduced signalling information exchange between the devices and access network. Long-term availability of the desired content due to sustainability of connection and mobility is one of the challenges to integrate Ds2D communications in a decentralized manner.
Smartphones will play an important role to enable device-centric communication paradigms in 5G networks, such as D2D communications. This chapter focused on the implementation perspectives of such a device centric architecture, including energy consumption and battery life aspects of the devices involved in communication. A new device-centric scheme, Ds2D communication was discussed, and it incorporates several source devices and multiple radio interfaces for data transfer to the sink device. An optimal algorithm for source device and radio interface selection was presented based on the ascending proxy auctions mechanism. The proposed mechanism achieves a higher energy efficiency compared with the conventional D2D communications approach and a lower energy consumption per source device compared with the multi-homing D2D communications approach. The proposed Ds2D communication scheme guarantees an optimal data packet distribution among the source mobile devices and it ensures improvements in the file transfer latency, energy consumption and battery life of the source mobile devices involved in communication. Simulation results evaluated the quantitative gains as exhibited by the traditional D2D and Ds2D communications via random data packet distributions. It illustrated that performance metrics associated with the source mobile devices such as file transfer latency, energy consumption and battery life can be effectively optimized through an optimal packet split strategy among the source mobile devices and their respective radio interfaces involved in Ds2D communications in DCS.