9.2.1 Network Architecture

The hierarchical HetNet comprises the following tiers:

• Tier 1: Macro-cell users connected to the MBS.
• Tier 2: Small-cell users connected to the SBS.
• Tier 3: D2D users connected to MBS and SBSs.

The following subsections will present the assumptions underlying the user distribution in the macro-cell and small-cell networks and the integration of D2D communications in heterogeneous networks.

9.2.1.2 Small-Cell Network

Let denote the number of users in each SBS and define the user density of the SBS. The number of SBSs required to cover the MBS coverage area is

9.1

where is the smallest integer not less than and denotes the number of users in the SBS.

Let denote the content exchange information of the SBS, where the corresponding users are involved in D2D communication for device-centric and low-mobility indoor activities (gaming, ultrahigh-definition video sharing, etc). In this case, yields the number of SBS cellular users (not involved in D2D communications), whereas the total number of users involved in D2D communication in the entire small-cell network can be expressed as

9.2

where denotes the total number of D2D pairs in the small-cell network. The remaining users of all SBSs, not involved in D2D communication, are given by

9.3

9.2.2 D2D User Density in Hierarchical HetNets

The probability of users for D2D communication depends on many factors, including channel conditions and common contents. In order to choose D2D pairs in MBS () and SBS (), the cumulative distribution function (CDF) for MBS and SBS users is approximated as shown in Figure 9.2.

Since MBS users are non-nomadic and fast moving as compared with SBS users; therefore, the value of is chosen and it shows approximately 50% probability for D2D users. For SBS users, the value of shows 60% probability for D2D users in a small-cell. For illustrative purposes, the value for and is chosen, and it generates approximately 5,000 users among which 1,000 (20%) are MBS and the rest (80%) are SBS users. In Figure 9.2, the value of corresponds to approximately 660 MBS D2D users, whereas corresponds to approximately nine D2D users per SBS.

An ultra-dense environment is simulated by increasing the user density from 1 to 20 milli users/m. In order to deploy SBSs uniformly into the coverage area of MBS, the whole disc of radius is divided into circular rings. For illustrative purposes, the rings of the hierarchical HetNet showing a three-tier network are shown in Figure 9.3.

In such a hierarchical network, the whole area is covered by the MBS with black circles showing SBS deployment. The small circles and plus signs show MBS and SBS users directly connected to the respective BSs. The crosses show D2D users either in MBS or SBSs as tier 3 network.

9.2.3 Spectrum Partitioning in Hierarchical HetNets

We assume dedicated carrier deployment in the communication network, where the MBS, SBS and D2D communication users operate on separate bandwidths based on the active number of users associated with each technology. Let the total available spectrum be (Hz). It follows that for the traditional HetNet,

whereas for the hierarchical HetNet,

where and are the dedicated channels of each MSB and SBS user in the traditional HetNet, respectively. Similarly, , , and are the dedicated channels of eachMBS, SBS, D2D with MBS and D2D with SBS user in the hierarchical HetNet, respectively. The number of channels in MBS and SBSs are assumed to be equal to the number of users they contain, and each channel is allocated to a single user [256]. Hence, the interference received at the MBS or SBS is from the mobile users in each of the neighbouring co-channel macro or small-cells that transmit on the same channel, while the interference in each D2D communication link is assumed to be from the closest D2D communication user that is not part of that communication link. This assumption was made because mobile devices engaged in D2D communication usually transmit with very low power, which causes reduced interference.

9.2.4 Power Control over D2D Links

The received signal power at a distance between MBS or SBSs and one of the devices engaged in D2D communication is given by

9.4

where and denote the basic and additional path-loss exponents, respectively, and denotes a path-loss-dependent constant. The parameter (m) is the break point of the path-loss curve, (m) represents the BS antenna height, (m) denotes the mobile user antenna height and (m) denotes the wavelength of the carrier frequency . Both the small-cell and D2D communication users are assumed to transmit with adaptive power while maintaining a certain received signal threshold. The adaptive transmission power of a user is given by

9.5

where and represent the maximum transmission power of a user, received signal power threshold, link distance and path-loss exponent, respectively. The assumption that is also considered due to the short link distances. This implies that

9.6

for all users in the network.

9.3.1 Traditional HetNet

The sum rate of traditional HetNet (without D2D communication) consists of the individual sum rates of MBS and SBSs:

9.7

where (bits/s) denotes the sum rate of MBS and (bits/s) denotes the sum rate of SBSs. The achievable capacity of the user located in the MBS of a traditional HetNet is given by

9.8

where denotes the PDF of and is the signal-to-interference ratio (SIR) of the desired link. Assuming the thermal noise power is negligible compared with the co-channel interference power, the SIR of the user located in the macro cell is given by

9.9

where (W) denotes the received power at the macro cell from the user and denotes the sum of the individual interfering power levels received at the reference MBS from the interfering mobile users , which are located in each of the interfering MBSs. Substituting (9.4) into¹ (9.9), it turns out that the SIR of a macro-cell user is given by

9.10

Similarly is the achievable capacity of the user in the small-cell, and it is given by

9.11

where denotes the PDF of and denotes the SIR of the user in the small-cell, and is expressed as

9.12

The parameter (W) in (9.12) represents the received powers at the small BS from the user and is the sum of the power received at the small BS from the interfering small-cell users located in the neighbouring interfering small BSs in HetNet. Substituting (9.4) into (9.12), the SIR of the small-cell user is expressed as

9.13

9.3.2 Hierarchical HetNet

The capacity of the hierarchical HetNet depends on the cellular and D2D users in both MBS and SBSs. In case of MBS, we have cellular and D2D users, whereas for each SBS, we have cellular and D2D users. The total capacity (bits/s) of the hierarchical HetNet is given by

9.14

where consists of the capacity of cellular and D2D users of MBS. Similarly, the capacity is the capacity of cellular and D2D users of each SBS. Variables and represent the achievable capacity of MBS and SBS cellular users calculated similarly to (9.8) and (9.11), respectively.

The achievable capacity of the D2D communication user in MBS or SBS is expressed as

9.15

for . Let denote the PDF of the desired SIR of the D2D communication user in MBS or SBS. Then,

9.16

where denotes the D2D user's received power at its D2D partner in MBS or SBS and denotes the received interference power at the D2D user from the interfering D2D user .

Substituting (9.4) into (9.16), the SIR of the mobile user () is expressed as

9.17

The capacity enhancement of the hierarchical HetNet is compared with the traditional HetNet in Figures 9.4 and 9.5. Figure 9.4 illustrates the sum rate (bps/Hz) versus the variable user density for MBS and SBSs for the two cases of non-D2D and D2D users. The sum rate capacity increases with an increase in the number of D2D users in the hierarchical HetNet. This is due to the frequency reuse, whereas the traditional HetNet shows a constant sum rate. By increasing the number of users in a traditional HetNet, the channel bandwidth per user reduces to accommodate the new users in a fair and uniform manner. However, the sum rate calculated for the increased number of users under fixed-system bandwidth will remain constant as validated by the simulation results. For the hierarchical HetNet, the channel bandwidth for a cellular user decreases, but the D2D communication reuses the channel bandwidth and results in a sum rate enhancement. An interesting cross-over point is observed at 11 milli users/m, where single MBS with D2D links shows higher capacity than huge deployments of SBSs with non-D2D links. This cross-over point can be reached at low user density if the number of D2D links is increased further. However, D2D pairs can be exploited opportunistically depending on different factors, for example, shortest distance, channel conditions and common content information.

The overall system gain of the hierarchical HetNet depicted in Figure 9.5 shows significant capacity enhancements compared with the constant sum rate of the traditional HetNet. These capacity gains can further be enhanced by using non-orthogonal spectrum sharing and smart interference management techniques. In such a scenario, the optimum number of D2D pairs can be found, for example, by achieving the target SIR at the desired node (cellular or D2D).

In Figure 9.6, the interference geometry is drawn for a traditional and hierarchical HetNet. Two user densities are simulated: 1 milli user/m² and 10 milli users/m² (closer to the cross-over point). In both cases, the CDF plot shows significant improvements in terms of required SIR and outage probability. For example, to ensure an outage probability of 10% in case of 1 milli user/m², the HetNet with D2D links requires = 26.66 dB less SIR than a traditional HetNet. Similarly, in case of 10 milli users/m², the SIR gain = 32.74 dB was observed.

In the next section, the mathematical analysis to compute the average transmission power of a user in the network is presented.

9.4.1 Discussion on Transmission Power Analysis of D2D Users

This section presents the transmission power performance of our proposed network deployment. We only consider the effect of path loss in our simulation. The simulation parameters are summarized in Table 9.1.

A comparison of the simulation and analytical results of the proposed deployment in terms of average user transmission power is illustrated in Figure 9.7. The simulation results are further compared with the results corresponding to a full small-cell deployment network and maximum transmission power. Figure 9.7 shows that the simulation and analytical results match and that the proposed deployment has a considerably lower average user transmission power than the full small-cell deployment. This is attributed to the lower transmission powers of the D2D communications required by the shorter communication link. However, the average user transmission power of both deployments increases as the received signal threshold increases. This is due to the fact that the users must transmit with higher power to overcome the effect of path loss and achieve the minimum received signal power at the receiver (of the D2D communication or SBS). It turns out from the Figure 9.7 that our proposed deployment achieves up to 25% reduction in average user transmission power compared with the full small-cell deployment.

Figure 9.8 shows the average user transmission power saving of the proposed scheme and small-cell deployment. The transmission power saving depicts how much power a typical user is able to conserve by incorporating D2D communication in the network. It can be observed that the average user transmission power saving decreases as the received signal power threshold increases. This is aresult of the increased transmission power of the users as the minimum received signal power increases. It turns out that the proposed deployment achieves a higher average user transmission power saving than the full small-cell deployment.

Figure 9.9 shows the average user transmission power of the proposed scheme against the full small-cell deployment as the user density increases at μW. It can be inferred from the figure that the average transmission power of a small-cell user is constant as the user density increases. This is because the user density does not affect the distance between the user, the SBS and the average transmission power (9.20). On the contrary, as the user density increases, the number of potential D2D communications increases, which results in the mobile users being closer. The shorter link distances lead to the reduction of average user transmission power in the considered deployment as the user density increases as shown in (9.24). Even though the average transmission power of the small-cell user is the same as that of the full small-cell deployment, the incorporation of D2D communications lowers the average transmission power of users.

9.5.1 Backhaul Power Consumption

The backhaul power consumption, which is the power needed to carry user traffic to the core network, depends on the type of deployment and the small-cell technology used. D2D communication has no backhaul power requirement, because D2D communication user traffic is not routed to the core network, as the mobile users engage in direct communication without the need for any intermediary node. Therefore, the total backhaul power requirement of the network with D2D communication is simply the backhaul power requirement of the macro-cell BS, and is expressed as [257]

9.27

where represents the maximum number of downlink interfaces at the macro-cell BS aggregation switch and it is used to compute the number of aggregation switches needed. Variable denotes the power consumed by a downlink interface at the macro-cell aggregation switch, and it is used to receive the backhaul traffic. Variables and represent the total number of uplink interfaces and the power consumption of one uplink interface, respectively. The number of uplink interfaces is given by Skubic and Ericsson [257]

9.28

where is the aggregate traffic at the macro-cell BS switch(es) and is the maximum transmission rate of an uplink switch. The term denotes the power consumption of the aggregation switch, and is expressed as

9.29

where is the maximum power consumption of the switch, represents the maximum traffic that the switch can carry, and denotes the weighting factor [257].

We assume that the traffic from the small-cells (femto cells) is routed straight to the core network via the Internet, without going through the aggregation node at the macro-cell BS. The access network of the small-cells is assumed to be a passive optical network (PON). A single fibre cable from the core network, which serves a group of small-cells, is fed into an optical line terminal (OLT), which may be located at the local exchange. A passive curb at the local exchange splits the fibre cable from the OLT into several fibres, each connected to an optical network unit (ONU). Each ONU then serves a single small-cell. The OLTs are connected to the edge routers, which serve as the small-cell gateways for transmission to the core network. The power consumption of the small-cell backhaul is expressed as

9.30

where denotes the number of ONUs that connect to one OLT, represents the total traffic of the small-cells, denotes the power consumption of the OLT, denotes the power consumption of the ONU [258] and represents the power consumption of the edge router, which can support up to 40 OLTs [259].

9.5.2 Backhaul Energy Efficiency

The backhaul energy efficiency (BEE), which shows the energy utilization of the backhaul technology, is a key performance indicator for future mobile communication systems. BEE expressed as the maximum amount of bits that can be transmitted per joule of energy consumed by the backhaul network, and it is measured in bit/Joule [260]. BEE is important particularly when choosing the type of backhaul technology to use during network planning to bring down the operational expenditure (OPEX) of the network. BEE can be also expressed as

9.31

where is the achievable throughput of the network and represents the resultant backhaul power consumption of the network expressed as the sum of the power consumption of the backhaul network and the downlink power consumption:

9.32

Variable represents the total transmission power of the D2D communication users. The total power consumption of the full small-cell network is expressed as

9.33

where

9.34

and

9.35

The parameters and denote the power consumptions of the macro-cell BS and each small-cell BS, respectively. The parameters and represent the slope of the load-dependent power consumption of the macro-cell BS and small-cell BS, respectively. Variables and denote the transmission power of the macro-cell BS and small-cell BSs, respectively. Furthermore, and denote the overhead power consumption of the macro-cell and small-cell BS, respectively [261].

9.5.3 Considerations on Backhaul Energy Efficiency of Hierarchical HetNet

This section compares the performances of the proposed network with D2D communication against the network with full small-cell deployment in terms of the backhaul power consumption and BEE. The simulation parameters are summarized in Table 9.1.

Figure 9.10 depicts the backhaul power consumption of the proposed network with D2D communication against a network with full small-cell deployment by assuming the throughput constant and a varying macro-cell radius from m. Figure 9.10 indicates that the network with D2D communication presents a significantly lower backhaul power consumption than the network with full small-cell deployment. This is because D2D communication users have no need for any backhaul network to convey their traffic to the core network and only the macro-cell users have their traffic carried by the backhaul network from the macro-cell BS to the core network. On the contrary, the backhaul power requirement of the network with full small-cell deployment increases as the radius of the macro-cell increases. This is due to the increase in the population of small-cells in the network as the macro-cell radius increases and each small-cell has its own backhaul power requirement. It turns out that the network with full small-cell deployment presents about 4–20 times higher backhaul power consumption than that of the network with D2D communication, depending on the radius of the macro cell.

Figure 9.11 illustrates the BEE comparison of the network with D2D communication and the full small-cell deployment. The network radius and the throughput of the macro cell were fixed at m and Mbps, while the total throughput of the network was varied from 10 to 100 Mbps. It can be seen that the BEE of both networks increases as the throughput of the network increases. This is because BEE is a function of the throughput and the total power consumption of the backhaul network. The BEE of the network with D2D communication is at least 260% higher than that of the network with full small-cell deployment. The higher BEE of the D2D communication is due to the lack of backhaul power consumption for the D2D communication users and only the macro-cell users' traffic is backhauled to the core network. However, the backhaul power consumption of each small-cell in the network with full small-cell deployment has to be considered in calculating the BEE, which results in a lower BEE of the network.

Figure 9.12 depicts the total transmission powers of the tier 2 D2D communication users and small-cell users against the macro-cell radius. It turns out that the D2D communication users exhibit a lower transmission power compared to the small-cell users. The lower transmission power is due to the shorter transmitter–receiver link in the D2D communication relative to the small-cell access distance, and the mobile users transmit with just enough power to overcome the effect of path loss via power control. The D2D communication users achieve up to 250% transmission power reduction at a macro-cell radius of 600 m compared to the small-cell users. However, the sum transmission powers of both schemes assume larger values because of the increase in the number of users as a result of the increase in the macro-cell radius.

Figure 9.13 shows the downlink power consumption comparison of the network with D2D communication with the network with the full small-cell deployment at different macro-cell radii. It turns out that the full small-cell network presents a much higher downlink power consumption that increases as the macro-cell radius increases. This is due to the increase in the population of small-cells in the network as the radius of the macro cell increases. Although the downlink power consumption of the network with D2D communication appears to be constant, there is a marginal increase in the downlink power consumption due to the increased user population as the macro-cell radius increases. The network with D2D communication achieves a downlink power consumption reduction of up to 400% at a macro-cell radius of 600 m. Even though the transmission power of the D2D communication users is very low, transmissions over long periods of time (as is the case with mobile multiplayer gaming) may have significant impact on the battery life of the D2D communication terminals.