Heterogeneous wireless networks mainly rely on RF network integration, such as cellular networks and WLANs or macro- and femto-cell cooperation as discussed in the previous chapters, to enhance the users perceived service quality and improve the achieved energy efficiency. However, such RF networks suffer from spectrum congestion. As a result, new network technologies that present larger spectrum availability should be introduced. In this context, VLC is considered to be a promising network technology that can offer high data rates with almost no transmission power consumption. Yet, VLC suffers from some technical limitations (such as absence of line of sight (LoS) and infeasibility of uplink transmission) that motivate its integration with RF network technologies to enhance the overall network performance. This chapter discusses RF and VLC integration in heterogeneous networks. Several integration objectives are presented such as load balancing, throughput maximization and uplink data transmission. Then, this chapter focuses on RF and VLC internetworking for green (energy-efficient) communications. A radio resource allocation mechanism that can improve downlink energy efficiency in an integrated RF–VLC network is presented and the challenging issues are discussed.
Using visible light for communications dates back to the early ages, when humans relied on beacon fires and lighthouses to convey messages [178]. In the 19th century, Arthur Aldis invented the signal lamp, which uses Morse code for data transmission [179]. In 1880, Alexander Graham Bell invented the photophone, a device that modulates light beams with human voice [178]. However, several technical limitations (e.g. the impact of natural obstacles such as fog and rain on the perceived signal quality) prevented Bell from pursuing his research. Recently, the invention of light-emitting diode (LED) has renewed the research in VLC. In 2004, data transmission was possible in the Nakagawa laboratory using LEDs and in turn this has raised a significant interest from both academia and industry in VLC [180].
In general, optical wireless communications refer to data transmission using infrared (IR), ultraviolet (UV) and visible light communications [181]. VLC is different from IR and UV in the sense that the same energy used for illumination is used for communications [181]. In particular, visible light is emitted from a LED light bulb, when excited by a direct current, as a stream of photons. The light intensity is proportional to the direct current value. By modulating the direct current with the data signal, the light intensity varies, which can be detected by a photo diode at the receiver. The varying light intensity, however, is undetectable to the human eyes. Hence, VLC is a data communication technology that uses the frequency spectrum in the range of 384–789 THz (i.e. corresponding to wavelengths 380–780 nm).
VLC offers many attractive features that motivates its wide adoption [178]. For instance, the VLC spectrum is 10,000 times more than the RF spectrum. In turn, this enables the support of data-hungry applications that require high throughput. Furthermore, there are no safety or health concerns for VLC. In addition, VLC offers a secure way of communication as the VLC signal is confined to the illumination area and cannot travel through walls. More importantly, VLC presents an energy-efficient means of communication, since the same energy used for illumination is used for communication, which implicitly means that zero transmission power is used in VLC. However, VLC technology is challenged by several technical limitations [178, 181]. For instance, in VLC, data communication deteriorates significantly in the absence of LoS signal, which is not the case for RF communications. Furthermore, interference from ambient (sun) light can significantly reduce the received SNR, and hence it degrades the communication quality. Moreover, realizing an efficient uplink communication can be problematic in VLC as MTs cannot support an optical uplink due to issues related to device orientation and energy constraints [182].
The aforementioned advantages and limitations suggest that VLC is better exploited if integrated with the existing RF communication technology rather than competing with it. In particular, there have been a few research efforts that aim to design a heterogeneous network that integrates RF and VLC networks in a unified framework to make use of the benefits of each technology and overcome the associated drawbacks. In this chapter, we discuss RF and VLC network integration for green (energy-efficient) communications. Towards this end, we first present some fundamentals related to VLC, followed by the integration scenarios of RF and VLC networks for load balancing and throughput maximization. Then, we present a radio resource allocation mechanism [183] that exploits the RF and VLC radio resources to support green communications. Finally, future research challenges in designing such a green VLC–RF network are discussed.
This section covers some background information about VLC networks. This includes the VLC transceiver, VLC channel, interference issues in VLC and VLC–RF internetworking.
A VLC transceiver is shown in Figure 6.1. In the following, we will describe the constitutive elements of the VLC transmitter and receiver.
From a lighting perspective, in practice, two types of LEDs can be used to generate the visible (white) light [181]. The first type is known as RGB LED, which mixes specific quantities of red, blue and green colours to generate the visible (white) light. However, this type of LED is challenged by inefficiency in green light generation ( efficiency in generating the green light as compared with and in generating blue and red lights, respectively). Another more efficient type of LED is the phosphor-based white LED. This type of LED is commonly used in lighting. It generates blue light, converts part of the blue light and mixes the converted and non-converted parts of the blue light to generate the desired shade of white light.
From a communications perspective, denote as the driving current of the LED, expressed in Amperes. The driving current is proportional to the modulating data. Hence, the driving power is the average of the squared driving current, that is, . In a multiuser system, MT is allocated a proportion of , which is denoted by . The allocated power represents one system parameter to be controlled when optimizing the system performance. The LED output is an optical signal (visible light), whose intensity is proportional to the driving current and is expressed as , where is the proportionality factor of the electrical-to-optical conversion. The average optical transmitted power is given by . The optical transmitted power specifies the LED illumination level.
The IEEE 802.15.7 standard, launched in September 2011, defines seven colour channels for the physical layer (PHY) in VLC [178]. Three PHY modes are introduced in IEEE 802.15.7, namely PHY I, PHY II and PHY III, which offer 11.67–266.6 kbps, 1.25–96 Mbps and 12–96 Mbps, respectively [184]. Both PHY I and PHY II are defined for a single light source and can support on–off keying (OOK) and variable pulse position modulation (VPPM). Only PHY III uses multiple optical sources with different frequencies (colours) and supports colour shift keying (CSK). Each PHY mode identifies mechanisms for light source modulation, run length-limited line (RLL) coding and channel coding. In addition, the achievable data rates can be significantly improved if more advanced modulation schemes are used such as optical orthogonal frequency division modulation (OOFDM) [180]. High data rates in the order of Gbps have also been achieved using multiple-input multiple-output (MIMO) techniques [180].
Illuminance is the most significant parameter that characterizes the white LEDs for lighting purposes [185]. Illuminance expresses the brightness of an illuminated surface [186]. At a specific point in the receiver plane, and assuming a Lambertian radiation pattern, the horizontal illuminance, as sensed by a photo-detector, is expressed as [185, 186]
where denotes the maximal (center) luminous intensity, denotes the irradiance angle, is the incidence angle and is the distance to the illuminated surface (i.e. the distance from LED to photo-detector). The Lambertian emission order is expressed as , where denotes the semi-angle at half-power (the viewing angle).
From a communication perspective, the received optical signal is first gathered by a concentrator. Different concentrators are used to gather LoS and non-LoS (NLoS) signals [181]. Optical filtering is then applied to alleviate the effect of the ambient light interference. The filtered signal is then fed to a photo-detector that converts the optical signal into an electrical signal (photo-current). Two types of photo-detectors can be used, namely photo-diode and image sensor [181]. The photo-current is then amplified, equalized and high pass filtered to remove the DC signal.
Denote the VLC channel power gain for MT by . The received optical signal by MT is given by . Let represent the photo-detector responsivity (in amp/W), which denotes the efficiency of the photo-detector in converting the received optical intensity into an electrical current. Hence, the received electrical signal, , for MT is given by . The average electrical power of the received signal is given by , and hence, it turns out that [183]
In the literature, two VLC channel power gain models are adopted. The first model captures only the LoS component, while the second model captures both the LoS component and NLoS first reflection. The two models are discussedbelow.
In [180], it has been shown that the average optical power received from the first reflection accounts only for of the optical power received from the LoS path. Hence, several works in the literature such as [185–189] and [190] represent the VLC channel via the LoS path and discard the reflection contributions. The optical channel gain for the LoS signal (DC gain) as received by MT is given by
where is the physical area of the photo-detector of MT , is the half-angle of MT field of view (FoV), is the optical filter gain and is the concentrator gain, which can be determined as
where denotes the refractive index.
However, most of the existing research do not account for the fact that the LoS component can be blocked due to obstacles. In [191], a random variable is introduced in (6.4) to denote the VLC LoS blocking event. The random variable is assumed to follow a Bernoulli distribution with blocking probability .
Although the NLoS component accounts for a small percentage of the optical power as compared with the LoS component, it is important in some scenarios such as when the LoS component is blocked and only the NLoS component is present or to account for the interference power from another AP. Therefore, it is important to model the NLoS channel power gain. In [191] and [192], the NLoS DC gain after the first diffuse reflection from a wall is expressed as
where and denote the distance between the AP and a reflecting surface and the distance between the reflecting surface and MT, respectively, denotes the wall reflection index, is a small reflective area, and denote the angle of irradiance to a reflective point and the angle of irradiance to the MT, respectively and . The NLoS channel power gain is then evaluated as follows
where the integration is over the four wall areas.
The channel power gain includes both an LoS component, , and NLoS component, , when the AP is in the FoV of the MT; otherwise, only the NLoS component is considered [192] (which can be the case for an interferer).
Two noise components should be accounted for, namely the shot noise due to the ambient light and thermal noise [191]. Hence, the VLC Gaussian noise has a total variance of
The shot noise variance is given by Jin et al. [191]
where is the electronic charge, is the received optical power by MT , denotes the equivalent noise bandwidth, denotes the background current due to the background (ambient) light and is a constant (in [191] ). The thermal noise variance is given by Jin et al. [191]
where denotes the Boltzmann's constant, is the absolute temperature, denotes the open-loop voltage gain, represents the fixed capacitance of the photo-detector per unit area, is the FET channel noise factor, is the FET trans-conductance and is a constant (in [191] ).
In VLC, a single AP is a collection of LEDs installed in the ceiling of a room. The AP purpose is twofold: illumination and communication. The AP coverage area is defined as an area free of illumination and communication dead zones. The illumination dead zone presents an illuminance level below a target value (e.g. for an indoor environment, the desired illuminance level is 100–500 lumens as standardized by GB 50034-2004 in China [186]). Similarly, the communication dead zone presents an SINR value below a target threshold. The AP illumination and communication coverage area depends on several parameters such as the LEDs semi-angle at half-power, height of the LEDs and FoV of the receiver (MT) [186].
A VLC cell can be defined within the VLC AP coverage area. Unity frequency reuse is adopted when the same frequency is reused across all VLC cells in a system with multiple APs. The drawback of such an approach is the resulting inter-cell interference (ICI) imposed by the neighbouring cells at the cell edge. For instance, the SINR for the cell-edge MT given the cell formation shown in the top left area of Figure 6.2 [188] is given by
where denotes the LoS component received by MT from AP A; , and denote the NLoS component (interference) received by MT from APs B, C and D, respectively (assuming that only AP A is within the field of view of MT ) and is given by (6.7). Hence, the MT receives interference from the NLoS component of all neighbouring cells. Different techniques are proposed in the literature to mitigate such ICI, as discussed below.
In this technique, adjacent cells are allocated different frequency bands for communications. The top right area of Figure 6.2 represents cell formation with a frequency reuse factor of 2 [180, 188]. Hence, the SINR for cell-edge MT is given by
In (6.11), the ICI from cells B and C is eliminated as compared with (6.10). However, using a frequency reuse factor larger than 1, a trade-off exists between the bandwidth efficiency (which will be degraded) and the SINR for cell-edge users (which will be improved). Rather than allocating separate frequency bands for the entire cell, and hence reducing the overall bandwidth efficiency, power control and subcarrier allocation can be exploited to mitigate interference for cell-edge users. Such an approach is similar in concept to the dynamic fractional frequency reuse that is deployed in RF networks [192].
In a combined transmission technique, a group of neighbouring APs forms a cluster and transmits the same information to MT [180, 188]. Hence, the received signals at the MT will be combined rather than being treated as interference. For instance, the bottom area of Figure 6.2 represents the combined transmission from two cells A and B to a cell-edge MT . In this case, the SINR is given by
In (6.12), the signal received from cell B is turned into a useful signal rather than being treated as interference as in (6.10), leading to an improved SINR. It should be noted that in (6.10) only the NLoS component of cell B contributes as an interference signal, while in (6.12) the LoS component is treated as a useful signal. One drawback with the combined transmission technique is that it also reduces the bandwidth efficiency, since only one user is served by several APs at a time.
Transmit pre-coding techniques (vector transmission) can be employed to enable simultaneous serving of multiple users [180, 188]. One example of such vector transmission is the zero-forcing approach, which is commonly used in multiuser MIMO RF communications.
As introduced earlier, VLC offers several advantages such as large available spectrum and energy-efficient and secure communications, while also it is challenged by several limitations such as signal deterioration in the absence of LoS and the infeasible uplink communications. Consequently, several works in the literature propose to integrate VLC and RF networks for improved system performance. In this context, VLC–RF internetworking has been employed for different objectives such as load balancing, throughput maximization, uplink support and energy efficiency, as discussed below.
The load balancing problem is basically a joint user association and resource allocation problem. In the literature, different integration scenarios are considered with load balancing objectives such as the integration between VLC and RF femto cells as in [185] and VLC and WLAN in [188]. One goal of load balancing is to ensure that no network is congested with users in a way that deteriorates the users' perceived QoS. A logarithmic utility function is commonly used as it can achieve load balancing and fairness among mobile users [185, 188]. The user association is a binary decision variable that is set to 1 if MT is assigned to network (VLC or RF). A constraint over ensures that MT is assigned to only one network (i.e. single-network assignment) [185, 188]. The radio resource allocation is a real variable that specifies the amount of allocated radio resources (e.g. power and/or bandwidth) to each MT with appropriate resource availability constraints. In most cases, the load balancing problem is an MINLP [188], which is an NP problem. In order to simplify the problem formulation, in [185], the authors consider that the effective load of an AP is represented by the number of MTs associated with that AP. In this case, the optimal resource allocation policy is to equally distribute the resources among the MTs associated with the AP. Hence, the load balancing problem is reduced to a pure association problem, that is, the problem is reduced from MINLP to binary programming (BP) and the problem is solved through variable relaxation. Furthermore, other heuristic association strategies can be employed such as the minimum distance user association and the maximum SINR user association. When both user association and resource allocation are considered, linear approximation can be used to reduce the problem from MINLP to linear programming (LP) as in [188].
In VLC, the achievable throughput spatially fluctuates within the indoor environment due to the absence of LoS caused by obstacles. Hence, the integration of VLC and RF networks is expected to improve the overall throughput, since VLC can support high data rates in certain areas while RF networks can support moderate data rates in larger areas. Such an integration with the objective of throughput maximization is studied in [190] between 5G RF AP (based on mmwave communications) and VLC, and in [191] between 4G RF femto cell and VLC. In [190], a central entity is assumed to be in place to monitor the network. By default, MTs connect to the VLC AP, and when the achievable throughput for a given MT is below a target threshold, the central entity connects this MT to the RF AP. In addition to investigating the achievable throughput in such an integrated system, the study in [190] investigates the throughput outage. It has been shown that the integrated system results in an improved achieved throughput and outage performance. Unlike [190], the work in [191] enables MTs to connect to both VLC and RF APs in a multi-homing approach. Furthermore, the work in [191] applies the effective capacity concept to convert the statistical delay constraints into equivalent average rate constraints. The objective of [191] is to design a radio resource allocation algorithm that maximizes the effective capacity of the integrated system and satisfies the users' required QoS. The problem is shown to be a convex optimization one that can be solved in a decentralized manner following the Lagrangian decomposition approach. Simulation results have demonstrated that in the presence of a VLC LoS component, the VLC AP is more reliable in satisfying the statistical delay guarantee than the RF AP, while in the absence of VLC LoS component, the RF AP becomes imperative.
VLC is ideal to support downlink communications in the coverage area of its AP. However, supporting uplink communications in VLC is challenging due to MT energy constraints, device orientation and interference. A heterogeneous RF–VLC network can resolve such a challenge where downlink communication is supported by VLC AP and uplink communication is supported by RF AP. Several implementation issues that should be addressed to achieve this vision are investigated in [182] and a test bed is developed as well.
LED is classified as a green lighting technology. Reports have indicated that the total lighting global power consumption will be reduced by if all light sources were replaced by LEDs [181]. In addition, VLC based on LED technology offers a green communication approach. This is mainly due to the fact that VLC exploits the illumination energy, which is already consumed for lighting, in high data rate transmission. However, due to reliability issues caused by the absence of LoS signal, VLC networks must be complemented by RF networks for energy-efficient and reliable network operation. In this context, the VLC AP employs its illumination power for data transmission while consuming additional power for data processing, while the RF AP consumes both data processing and transmission powers. MTs with multi-homing capability can receive data from both VLC and RF APs leading to an improved energy efficiency and reliability in the network performance. However, several challenging issues should be addressed first to fully exploit the energy efficiency and reliability benefits of such an integrated network. These challenging issues together with a radio resource allocation mechanism are discussed in the next section.
In this section, we first investigate the energy efficiency performance of an integrated VLC–RF network, then we discuss the challenging issues that need to be addressed in future research to fully exploit the energy saving and reliability benefits of this integrated network.
Consider an indoor downlink scenario with a single RF AP and VLC AP and a set of MTs in their coverage areas, as shown in Figure 6.3. The RF AP can be a femto-cell AP or WLAN AP. MTs with multi-homing capability can receive data from both VLC and RF APs to satisfy its required QoS. The total available bandwidths at the VLC and RF APs are denoted by and , respectively. The received bandwidths by MT from the VLC and RF APs are denoted by and , respectively. The allocated transmission powers to MT by the VLC and RF APs are given by and , respectively. Each AP has a maximum allowed total transmission power and . In addition to the transmission power, each AP has fixed power consumption components and for circuit operation and data processing before transmission. Furthermore, the VLC fixed power consumption accounts also for any required additional power to compensate for the losses in the LED efficiency due to data transmission. The achieved data rates by MT from each AP are denoted by and , respectively. The total data rate achieved by MT should satisfy a minimum required data rate of . The channel power gain between MT and the VLC AP, , is dominated by the LoS signal path loss and is given by (6.3). The noise power spectral density affecting the VLC receivers is denoted by , and is given by , where is given by (6.7). The channel power gain between MT and the RF AP is dominated by the signal path loss, and it is given by
where denotes the carrier frequency in GHz; , and are constants depending on the propagation model and is an environment-specific term. For the LoS scenario, , and . For the NLoS scenario, , , and in case of light walls or in case of heavy walls, where denotes the number of walls between the AP and MT. Hence, the RF channel power gain is given by
Denote the RF channel power gain in the presence of LoS by and for the NLoS scenario by . The LoS availability probabilities for RF and VLC systems denote the probability that there are no obstacles in the communication link between the MT and the corresponding AP, and are denoted by and , respectively. In the case of RF transmissions, the channel path-loss exponent increases with the LoS absence, as discussed earlier. For the case of VLC, the signal is degraded significantly in the absence of LoS and may result in unsuccessful data transmissions. It is assumed that the NLoS VLC transmissions are unsuccessful. Therefore, we focus only on the system performance with LoS VLC transmissions. This is mainly due to the fact that MTs operate in a multi-homing manner and the RF received signal will be significantly higher than the NLoS VLC signal.
The average received electrical SNR for MT from each AP is given by
where the value of is given by and to determine the corresponding and for the LoS and NLoS RF channels, respectively. The value of is calculated by dividing the received electrical power in (6.2) by the VLC noise power.
The achievable data rates from each AP are averaged over the probability mass function of LoS availability, and are given by
The total achieved data rate in the heterogeneous RF–VLC network is denoted by , and assumes the expression
The total communication power consumption is denoted by and its value is calculated as follows
where the first term in (6.18) represents the consumed power for the VLC AP and is calculated using the fact that the transmission power is the optical power used for illumination by design, and therefore only the fixed power consumption is accounted for as a communication power cost. The second and third terms represent the RF power consumption, which accounts for both the processing and transmission powers.
The objective of the radio resource allocation mechanism is to allocate transmission powers and bandwidths for the MTs from the available APs such that the total energy efficiency of the heterogeneous network is maximized and the MTs target QoS is supported while the bandwidth and power limitations of the APs are satisfied. This optimization problem is formulated as follows
where denotes the heterogeneous network total energy efficiency. The radio resource allocation problem (6.19) is a concave–convex fractional program [164] since the numerator of the objective function is concave with respect to the decision variables and the denominator is affine.
As in Chapter 5, the fractional program in (6.19) can be converted into a convex optimization problem in terms of parameter . Considering the same constraints as in (6.19), define
Using the Dinkelbach-type procedure in Algorithm 6.3.10, is updated and we can find the roots of , and hence the optimal solution for the optimization problem (6.20).
Problem (6.20) is a convex optimization problem. Applying the KKT conditions on the Lagrangian function of (6.20) as we did in Chapters 4 and 5, we can find (i) the optimal power allocations at the MT as a function of the Lagrangian multipliers , and and (ii) the optimal bandwidth allocation at each AP as a function of the Lagrangian multipliers , and [183]. The optimal values of the Lagrangian multipliers can be obtained by solving the dual problem using a gradient descent method [183].
Algorithm 6.3.11 finds the optimal VLC power allocation for a given allocated VLC bandwidth and .
Similarly, Algorithm 6.3.12 finds the optimal RF power allocation for a given allocated RF bandwidth and .
Given the allocated VLC power from Algorithm 6.3.11, the optimal VLC bandwidth can be allocated using Algorithm 6.3.13.
Similarly, given the allocated RF power from Algorithm 6.3.12, the optimal RF bandwidth can be allocated using Algorithm 6.3.14.
Using the optimal power and bandwidth allocated in Algorithms 6.3.11–6.3.14, the objective now is to jointly allocate the resources that satisfy the target data rate and maximize the resulting energy efficiency for a given . Algorithm 6.3.15 gives the optimal joint solution of (6.20) for a given value of . In Algorithm 6.3.15, we iterate over the power and bandwidth allocations until convergence to find the optimal joint bandwidth and power allocation solution that maximizes the energy efficiency in the heterogeneous VLC–RF network and satisfies the required QoS by all MTs. The optimal solution accounts for the RF and VLC characteristics (reliability and energy efficiency metrics).
Denote by the number of required iterations for the Algorithm 6.3.10 to converge. The complexity of the proposed resource allocation strategy is determined by calculating the number of dual variables in the dual problem. Therefore, the resource allocation framework computational complexity is given by , which is linear in the number of the MTs. The number of MTs that could be accessed using a single VLC AP for an indoor environment is small, and therefore the computational complexity of the framework is reasonably low.
Next, the energy efficiency of the integrated RF–VLC heterogeneous network (which is referred to in the simulation results by ‘RF–VLC’) is compared with two benchmark systems. The first benchmark represents a system consisting of a single RF wireless network (denoted by ‘RF-Only’) and hence no multi-homing is performed. The second benchmark represents a system comprising two RF APs over different frequency bands (denoted by ‘RF–RF’) and hence multi-homing is achieved only over RF links. In the system with two RF APs, one of the RF systems is assigned a bandwidth equal to that of the VLC system to ensure a fair comparison. In the following simulation results, we assume Mbps, W, W, W, W/Hz, W/Hz, MHz, MHz, W/amp, amp/W and . MTs are uniformly distributed away from the RF AP in the range between 1 and 1.5 m and are uniformly distributed away from the VLC AP in the range between 1.5 and 2 m. The maximum power of the VLC AP is given by the product of the number of LEDs used at the VLC source with the maximum power driving each LED. The number of LEDs is set to 38 with the maximum power to drive a LED to 300 mW. This value generates around 900 lumens from the VLC source, which is practically a suitable value for lighting.
Figure 6.4 shows the energy efficiency of the different systems against the number of MTs. The performance of the RF–VLC system is significantly better than the performance of the RF-only system thanks to the multi-homing capability of the MTs and the energy-efficient communication nature of the VLC AP. Through the multi-homing capability, MTs can enjoy communication links with better channel conditions with at least one AP, leading to high data rates and low power consumption. In addition, the performance of the RF–VLC system is better than that of the RF–RF system due to the lower power consumption cost of the VLC AP than the RF AP. The power consumption in RF APs is the sum of the fixed and transmission powers, while in the VLC AP, the power is due only to the fixed power component, since no power is dedicated for transmission as its transmission power is already used for illumination.
Figure 6.5 shows a plot of energy efficiency performance against the fixed power consumption of the VLC AP to investigate the effect of any increased fixed power consumption in the VLC AP. The energy efficiency of the RF–VLC system is equal to that of the RF–RF system when the fixed power is 6 W, which is nearly equal to the fixed power of an RF AP. One can conclude that the integration of a VLC AP in a heterogeneous networking with RF APs is beneficial only if the VLC AP fixed power consumption is less than an equivalent RF AP.
Figure 6.6 investigates the effect of the number of LEDs on the energy efficiency of the RF–VLC system. Increasing the number of LEDs allows a higher transmission power for the VLC system, which motivates the MTs to obtain most of their required data service from the VLC AP, reducing the transmission power consumption of RF AP, and hence improving the overall energy efficiency. In addition, the figure shows that introducing the VLC network even with a small number of LEDs (only 5 LEDs) enhances the energy efficiency significantly.
Figure 6.7 studies the case in which the LoS availability probabilities for RF and VLC systems are equal and shows the energy efficiency versus the LoS availability probability. The performance of the RF–VLC system is better than the benchmarks when the probability of LoS availability in the VLC AP is higher than 0.7 because of the good energy efficiency properties of the proposed RF–VLC system. Furthermore, the slope of the curve of the RF–VLC energy efficiency is higher than those of the benchmark systems because of the significance of the LoS availability in the VLC system compared with the RF system.
Figure 6.8 investigates the effect of LoS availability probability in the RF system on the energy efficiency of the RF–VLC heterogeneous system when . In the RF–VLC system, the MTs exploit the less costly VLC energy for data transmission and exploit the RF transmission power when needed. Consequently, the performance improvement with the increase of RF LoS availability probability presents a smaller slope in the integrated system than the RF benchmarks.
There are a number of challenging issues that should be addressed to enable more efficient integration of VLC and RF APs in a heterogeneous wireless network. In order to overcome these challenging issues, further investigations are required to deal with the users' spatial distribution, the APs' placement and the interference concerns in VLC networks. These issues are discussed next in more details.
High data rates in VLC APs can be achieved when there is an unobstructed LoS between transmitter and receiver. However, the data rate is reduced significantly in the absence of LoS, which can be due to the MT FoV misalignment with the VLC AP or due to obstacles. Consequently, data rate can be abruptly degraded once the LoS is obstructed, unlike the RF networks. Hence, in VLC APs, the network performance is very sensitive to the probability of the LoS availability, which in turn depends on the specific MT spatial distribution. The probability of the LoS availability can vary significantly with any slight change in the user locations. Therefore, the exact knowledge of the spatial distribution of the MTs is a crucial parameter in assessing the performance of the VLC network. Thus, the performance of the resource allocation mechanism in heterogeneous networks integrating RF and VLC APs is highly affected by the spatial distribution of the MTs. Hence, allocating network resources in the RF and VLC networks given the knowledge of the MTs spatial distribution can lead to better energy efficiency. Therefore, studying the MTs' spatial distribution and the techniques to employ this knowledge in resource allocation problems plays an important role in the system design stage.
Furthermore, it is necessary to design a resource allocation framework that is robust to uncertainties in MTs' locations. For instance, it is shown in [193] that a small error in the user location information can degrade significantly the data rate obtained from the VLC AP. Moreover, the achieved data rate at the VLC communication dead zones is relatively low. By examining the spatial distribution of MTs and studying the expected required rates of the users located in the VLC communication dead zones, more RF transmission power can be allocated to the MTs in these VLC dead zones while allocating more VLC transmission power to the VLC covered regions, which calls for a joint design of the VLC and RF APs coverage areas. The selection of the positions of the VLC and RF APs in both systems can considerably affect the system performance. While the RF AP placement is constrained mainly by the communication requirements, the VLC AP position is constrained by both the illumination and communication requirements [194]. Moreover, the transmission powers in both VLC and RF communication networks are highly affected by the path loss, and hence finding the optimal placement of the APs can highly benefit the achieved energy efficiency. Further research is needed for the joint placement optimization of RF and VLC APs to improve the energy efficiency while maintaining the communication and illumination requirements.
Communication over the light spectrum can be exposed to different types of interferences due usage of various light sources, for example, the ambient light. The visible light spectrum is wide enough to allow high data rates. In indoor scenarios, light does not penetrate walls. Hence, any portion of the visible light spectrum can be exploited inside some closed indoor spaces. Being unlicensed allows exploiting the spectrum for various services using different communication schemes. On the contrary, being unlicensed does not allow reserving certain portions of the spectrum for communication purposes only, and hence any light source or light reflection is considered as an interference source in VLC networks. This problem can be addressed by employing coding approaches in VLC systems. However, such an approach affects the energy consumption in VLC systems due to the extra processing requirements. By integrating both VLC and RF systems, interference mitigation solutions that exploit data transmissions over the two used spectra can improve energy efficiency. However, such a solution requires a further investigation.
Integrating RF and VLC APs into a heterogeneous wireless networking environment is a promising solution to support energy-efficient communications. The multi-homing capability of the MTs in a heterogeneous network with VLC and RF APs enables users to benefit from the huge unlicensed bandwidth of the visible light spectrum and the low cost of power transmission. Also, it provides a reliable means of communication in VLC systems, as RF communications are employed in the absence of VLC LoS. In this chapter, the superior performance of the heterogeneous integrated RF–VLC network is demonstrated and compared with the benchmarks that support an RF-only network or a heterogeneous network consisting of two RF systems, respectively. Further investigations are required to deal with the joint RF–VLC planning design (AP placement), exploit the MT spatial distribution information in the radio resource allocation and cope with the interferences present in a heterogeneous network with multiple VLC APs to support even greener communications.