Network selection method based on MADM and VH-based multi-user access scheme for indoor VLC hybrid networks

Tao Shang; Zhi Peng Sun; Zan Yang Dong; Qian Li

doi:10.1364/OE.26.030795

1. Introduction

Recently, with the development and popularity of various mobile terminals, cloud-based data and multimedia services have grown rapidly. Meanwhile, users put forward higher requirements for data transmission rate, system capacity, seamless connection, and network security. According to the statistics of the operators, 80 percent of mobile traffic occurs indoors [1]. However, in indoor environment, the traditional wireless communication based on radio frequency (RF) is facing with the problem of limited spectrum resource and bandwidth. Moreover, it cannot be better applied to RF sensitive areas such as hospitals, airports, and mines etc.

As an emerging optical wireless communication technology, visible light communication (VLC) has the advantages of large capacity, high bandwidth, anti electromagnetic interference, high security, free spectrum application, and green energy saving etc [2–6]. The access point (AP) of VLC cannot only provide illumination, but also support high speed data transmission in indoor environments. However, if VLC is used as a single wireless communication system, there are many shortcomings such as the communication link is easily blocked, the uplink is difficult to implement, and the modulation bandwidth of the LED is limited etc. Therefore, more and more researchers combine VLC with other wireless communication technologies, such as WiFi and infrared (IR), to build indoor hybrid networks [7–9].

In VLC-based indoor hybrid networks, when the user initially enters the hybrid networks or when the threshold condition of handover is triggered, it is important to select an optimal network that best suits the requirements of the user. In recent years, the network selection algorithm of hybrid networks has been studied by more and more researchers [10–14]. An effective network selection algorithm is proposed, which combines the analytic hierarchy process (AHP) method to determine the importance of the network parameters and the total order preference by similarity to the ideal solution (TOPSIS) method to rank the candidate networks [10]. A novel network selection scheme is presented, in which the AHP method is used to decide the relative weights of criteria and the grey relational analysis (GRA) method is used to rank the network alternatives [11]. A fuzzy-AHP based network selection method is proposed, which considers the user preferences, network conditions, and energy consumption [12]. An intelligence access selection scheme is proposed, which makes decision based on context-awareness and predictable dynamic attributes [13]. A network selection scheme considering user centric requirements and the load balance of network is presented in [14]. Meanwhile, some researchers have proposed to combine the subjective preference of the user with the objective performance of the networks for network selection. In [15], a realistic hybrid RF-VLC system is proposed that considers both the user's application requirement and the objective network performance. The system presented performs the handovers in an iterative manner, where the users first select the access points based on their trajectory and the access points select the users according to the dynamic load graph. But how to effectively combine the subjective preference of the user with the objective performance of the network and dynamically assign weights to these two factors is also an important issue in the optimal network selection in VLC-based hybrid networks.

In the process of multi-user competing to access the network, the mechanism of carrier sense multiple access with collision avoidance (CSMA/CA) is widely adopted to make the best use of the wireless channel resources, but once the number of retransmissions reaches a certain number of times, the contention window will be larger, and the delay for users to access the network will be longer. In addition, the number of retransmissions is limited, if the packet transmission is still unsuccessful after maximum attempts, the packet will be discarded. To overcome the drawbacks of the CSMA/CA mechanism, some researchers have improved the backoff mechanism of CSMA [16–18], but they are only improved for a single network. At present, the multi-user access protocol of VLC-based indoor hybrid networks has been studied. A parallel transmission MAC protocol which combines the CSMA/CA algorithm and the concept of parallel transmission is proposed in [19]. In this work, the RF channel is used as the uplink to send control packets via CSMA/CA algorithm, and the VLC channel and extra RF channel are used as downlink to transmit data packets. A CSMA/CA-based uplink MAC protocol for hybrid VLC/WiFi networks is proposed, which allows the user dynamically select channels based on channel aware and traffic aware [20]. For the problem of multi-user access in indoor VLC-based hybrid networks, how to reasonably use different channel resources and MAC information to improve the performance of hybrid networks is still a research hotspot.

In this paper, we mainly focus on how to select an optimal network and how to effectively realize multi-user access in VLC-based indoor hybrid networks. First, a network selection method based on multiple attribute decision making (MADM) is proposed. Considering the service requirement of user and the objective performance of network, we use the analytic hierarchy process (AHP) method, the standard deviation method and the simple addictive weighting (SAW) method to obtain the subject weight of the decision parameter, the objective weight of the decision parameter, and the evaluation value of network, respectively. In order to make the evaluation value consistent with the user’s requirement, and also with the network performance, a minimization mathematical model is established to get the partition coefficients of the subjective and objective weights. Second, a VH-based (virtual handover) multi-user access scheme is proposed, which comprehensively considers the channel and MAC information of various wireless communication modes in the indoor hybrid networks. A concept of backoff lock is presented to control the access request of the user to different channels. The concepts of VH and minimum waiting access delay (MWAD) are also presented to reduce the access delay and improve the access success ratio of the user. Unlike the previous studies, when the conditions of VH is satisfied, the user utilizes backoff lock to lock the current access request, and selects an optimal or the least MWAD network to send access request packet to other networks. But at this time, no network is switched. Moreover, the formulas of collision probability, access delay and access success ratio in the VLC-based hybrid networks are given. Final, under the same conversational service and different application requirements, the network selection method is analyzed by simulation. The results show that the proposed network selection method is effective to meet the user’s requirement and network performance. Meantime, the performances of the collision probability, access delay and access success ratio are also simulated for the proposed VH-based multi-user access scheme. The results show that, comparing to IR network and WiFi 802.11b network, the multi-user access scheme can reduce the collision probability and the access delay of the user, and improve the access success ratio.

The rest of this paper is organized as follows. The VLC-based indoor hybrid network model is presented in Section II. The network selection method is proposed in Section III. The VH-based multi-user access scheme is proposed in Section IV. The simulation results and analysis are given in Section V. Section VI concludes this paper.

2. VLC-based hybrid network model

An indoor VLC-based hybrid network model is shown in Fig. 1, which consists of one router, one server, three switches, N VLC APs, N^′ IR APs, and one hotspot of WiFi. In this model, All the VLC APs, IR APs and the hotspot of WiFi are linked to their respective switches via wired connections, and the three switches are connected to the server via wired connections. The server has the routing information of all the stations, and can respond to the request of users accordingly.

Fig. 1 Indoor VLC-based hybrid network model.

Download Full Size | PDF

In general, the half-power angle of white light LED used by VLC is up to 60 degrees, while the half-power angle of the infrared LED used by IR is 15 to 30 degrees. In order to achieve full coverage of the indoor hybrid networks and satisfy the requirement of illumination, multiple VLC and IR APs need to be mounted. For simplicity, the VLC and IR transmitter in our model adopt the pattern of Lambertian optical source [21]. In adjacent areas, in order to reduce the interference of adjacent VLC Aps, the adjacent VLC Aps are set to work in different frequencies bands, and the system settings of infrared AP are similar to those of VLC AP.

As shown in Fig. 1, IR and WiFi are used for uplink transmission, and VLC and WiFi are used for downlink transmission. In order to ensure the reliability of communication and the continuity of data transmission, the optical link (VLC or IR) can be switched to the WiFi link when it is blocked for a long time or its communication performance deteriorates seriously. Especially, we can close the hotspot of WiFi, and establish an all-optical network to apply to electromagnetic sensitive areas, which only consists of IR uplink and VLC downlink. For the convenience of analysis and comparison, the standard of WiFi, IR, and VLC adopted are IEEE 802.11b, IrDA: Advanced Infrared (AIr) - Version 1.0, and IEEE 802.15.7 respectively, and the channel access schemes of the three communication modes are all based on the RTS/CTS CSMA/CA mechanism.

In Fig. 1, taking User A as an example, when User A is on the coverage of the hybrid network model, it should decide which the most suitable network is. A network selection method based on MADM is presented in Section III. After User A completes the access selection, it maybe confronts with the problem of excessive access delay and even access failure in the procedure of access, so a VH-based multi-user access scheme is given in Section IV.

3. Network selection method based on MADM

In this section, a network selection method based on MADM for hybrid networks is presented. Based on the user’s service requirement and preference, we use AHP to obtain the subjective weight of each decision parameter. According to the concrete performance parameters of network, we use the standard deviation method to obtain the objective weight of each decision parameter. Then, a minimization mathematical model is established to obtain the partition coefficients of the subjective and objective weights. After obtaining the comprehensive weight and value of each parameter, we use SAW method to get the evaluation value of each network. The network with the highest evaluation value is the one that the user should choose to access.

During the procedure of network selection, in order to guarantee the QoS of the user, we consider seven representative decision parameters: bandwidth, average delay, throughput, average transmission rate, bit error rate (BER), received signal strength (RSS), and cost.

3.1 AHP for subjective weight

AHP method was proposed by T. L. Saaty, which combines quantitative and qualitative analysis [22]. It is commonly used in the field of decision analysis to assign the weight of each parameter.

In order to obtain the subjective weight of each decision parameter, based on AHP method, the main steps are presented as follows:

1) Construct judgment matrix
To satisfy the diverse service requirements, the user can set preference values for the decision parameters. A judgment matrix $A$ is constructed as:
$A = [\begin{matrix} x 11 & x 12 & ... & ... & x 1 n \\ x 21 & x 22 & ... & ... & x 2 n \\ ... & ... & ... & x i j & ... \\ ... & ... & x j i & ... & ... \\ x n 1 & x n 2 & ... & ... & x n n \end{matrix}] x_{j i} = \frac{1}{x_{i j}}, x_{i j} \neq 0$
where, $n$ is the total number of the decision parameters considered in the network selection; elements $x_{i j}$ is the importance of parameter $i$ relative to parameter $j$ , and its value can be determined according to Saaty Rule [22]. The Saaty Rule is given in Table 1.
Table 1. Scale of importance in Satty Rule.
View Table | View all tables in this article
2) Calculate the subjective weight
The subjective weight $w_{j}^{'}$ of the decision parameter can be expressed as
$A w^{0} = λ_{\max} w^{0}$ $w_{j}^{'} = {\frac{w_{j}^{0}}{\sum_{i = 1}^{n} w_{j}^{0}}}_{}$
where, $λ_{\max}$ is the largest eigenvalue of the matrix $A$ , $w^{0}$ is the eigenvector corresponding to $λ_{\max}$ , $w_{j}^{0}$ is the weight of decision parameter $j$ in $w^{0}$ , and $w_{j}^{'}$ is the subjective decision weight of parameter $j$ .
3) Check consistency
To avoid potential comparative inconsistency in the construction of judgment matrix $A$ , a consistency ratio ( $C R$ ) is given as:
$C R = \frac{C I}{R I} = \frac{λ_{\max} - n}{(n - 1) R I}$
where, CI is the consistency index of the judgment matrix A, it can be calculated as:
$C I = \frac{λ_{\max} - n}{n - 1}$
RI is a random index dependent on the size of the matrix A, its value can refer to Table 2 [22].
Table 2. The value of Random Consistency Index.
View Table | View all tables in this article

If the value of $C R$ is less than or equal to 0.10, the inconsistency of judgment matrix is acceptable; otherwise, the judgment matrix should be revised.

3.2 The standard deviation method for objective weight

The standard deviation method [23] was proposed by Y. M. Wang in 2003, which uses mathematical variance information to solve the problem of MADM. Its main idea is to use standard deviation to reflect the discreteness of the data sets.

In order to obtain the subjective weight of each decision parameter, apply the standard deviation method, the main steps are presented as follows:

1) Construct normalized decision matrix
Taking network performance into account, the decision matrix is constructed as:
$S = [\begin{matrix} s_{11} & s_{11} & ... & ... & s_{11} \\ s_{21} & s_{22} & ... & ... & s_{2 n} \\ ... & ... & s_{k i} & ... & ... \\ ... & ... & ... & ... & ... \\ s_{m 1} & s_{m 2} & ... & ... & s_{m n} \end{matrix}]$
where, $m$ is the total number of the candidate networks, $n$ is the total number of the decision parameters considered in the network selection, elements $s_{k i}$ is the normalized value of decision parameter $i$ in network $k$ . The normalization formula for the benefit parameters is expressed as [24]:
$s_{k i} = \frac{b_{k i} - \min {b_{k i} | (1 \leq k \leq m)}}{\max {b_{k i} | (1 \leq k \leq n)} - \min {b_{k i} | (1 \leq k \leq m)}}$
where, $b_{k i}$ is the value of decision parameter $i$ in network $k$ .
The normalization formula for the cost parameters is expressed as [24]:
$s_{k i} = \frac{\max {b_{k i} | (1 \leq k \leq m)} - b_{k i}}{\max {b_{k i} | (1 \leq k \leq m)} - \min {b_{k i} | (1 \leq k \leq m)}}$
2) Calculate the objective weight
The objective weight $w_{j}^{' '}$ of the decision parameter can be calculated as:
$w_{j}^{' '} = \frac{\sqrt{\sum_{j = 1}^{n} {(s_{i j} - \frac{1}{m} \sum_{i = 1}^{m} s_{i j})}^{2}}}{\sum_{j = 1}^{n} \sqrt{\sum_{j = 1}^{n} {(s_{i j} - \frac{1}{m} \sum_{i = 1}^{m} s_{i j})}^{2}}}, i=1,2 ... m, j = 1, 2... n$

3.3 Minimization mathematical model for partition coefficients

In order to effectively combine the user’s subjective preference with the objective performance of the network, we establish a minimization mathematical model to obtain the partition coefficients of the subjective and objective weights. In this mathematical model, we consider the consistency of the subjective and objective evaluation value and the optimality of evaluation results. Wherein, the consistency is determined by the deviation of the subjective and objective evaluation results of each candidate network, and the optimality is determined by the evaluation results of all candidate networks. The mathematical model is constructed as follows:

{\begin{cases} \min F = \sum_{i = 1}^{m} {{(\sum_{j = 1}^{n} α s_{i j} w_{j}^{'} - \sum_{j = 1}^{n} β s_{i j} w_{j}^{' ​'})}^{2}} - \sum_{i = 1}^{m} \sum_{j = 1}^{n} s_{i j} (α w_{j}^{'} + β w_{j}^{' ​'}) \\ s . t . α + β =1,0 \leq α \leq 1, 0 \leq β \leq 1 \end{cases}

where,

α

and

β

are the distribution coefficient of the subjective and objective weight respectively.

To solve the above mathematical model, we can construct a Lagrange function as:

\begin{array}{l} L (α, β, λ) = \sum_{i = 1}^{m} {{(\sum_{j = 1}^{n} α s_{i j} w_{j}^{'} - \sum_{j = 1}^{n} β s_{i j} w_{j}^{' ​'})}^{2}} \\ - \sum_{i=1}^{m} \sum_{j=1}^{n} s_{ij} (α w_{j}^{'} + β w_{j}^{' ​'}) + λ (α + β - 1) \end{array}

where,

λ

is the Lagrange coefficient.

The subjective weight $w_{j}^{'}$ and objective weight $w_{j}^{' '}$ of each decision parameter can be calculated by Eqs. (2), (3), and (9). Taking a derivative with respect to $α$ , $β$ , and $λ$ respectively, we can get the following equations:

{\begin{cases} \frac{\partial L}{\partial α} = \sum_{i = 1}^{m} [2 α {(\sum_{j = 1}^{n} s_{i j} w_{j}^{'})}^{2} - 2 β (\sum_{j = 1}^{n} s_{i j} w_{j}^{'}) (\sum_{j = 1}^{n} s_{i j} w_{j}^{' ​'})] - \sum_{i = 1}^{m} \sum_{j = 1}^{n} s_{i j} w_{j}^{'} + λ = 0 \\ \frac{\partial L}{\partial β} = \sum_{i = 1}^{m} [2 β {(\sum_{j = 1}^{n} s_{i j} w_{j}^{' ​'})}^{2} - 2 α (\sum_{j = 1}^{n} s_{i j} w_{j}^{'}) (\sum_{j = 1}^{n} s_{i j} w_{j}^{' ​'})] - \sum_{i = 1}^{m} \sum_{j = 1}^{n} s_{i j} w_{j}^{' ​'} + λ = 0 \\ \frac{\partial L}{\partial λ} = α + β - 1 = 0 \end{cases}

Solving this equation, we can get the result of $α$ and $β$ as follows:

α = \frac{\sum_{i = 1}^{m} [(\sum_{j = 1}^{n} s_{i j} w_{j}^{' ​'}) (\sum_{j = 1}^{n} s_{i j} w_{j}^{'} + \sum_{j = 1}^{n} s_{i j} w_{j}^{' ​'})] - \frac{1}{2} \sum_{i = 1}^{m} \sum_{j = 1}^{n} s_{i j} (w_{j}^{' ​'} - w_{j}^{'})}{\sum_{i = 1}^{m} {[\sum_{j = 1}^{n} s_{i j} (w_{j}^{'} + w_{j}^{' ​'})]}^{2}}

β = \frac{\sum_{i = 1}^{m} [(\sum_{j = 1}^{n} s_{i j} w_{j}^{'}) (\sum_{j = 1}^{n} s_{i j} w_{j}^{'} + \sum_{j = 1}^{n} s_{i j} w_{j}^{' ​'})] + \frac{1}{2} \sum_{i = 1}^{m} \sum_{j = 1}^{n} s_{i j} (w_{j}^{' ​'} - w_{j}^{'})}{\sum_{i = 1}^{m} {[\sum_{j = 1}^{n} s_{i j} (w_{j}^{'} + w_{j}^{' ​'})]}^{2}}

Then, the comprehensive weight

w_{j}

of each decision parameter can be written as:

w_{j} = α w_{j}^{'} + β w_{j}^{' ​'}

3.4 SAW for evaluation value

The SAW method [25] is widely used for network selection. Its main idea is to make a simple weighting of the decision parameters.

In our network selection method, in order to get the evaluation value of each network, referring to the SAW method, we make a simple weighting of the decision parameters and comprehensive weight. According to the weighting score, the rank of the optional networks is provided to the users. Therefore, the network got highest score is the access network. The score $V_{i}$ of each network can be expressed as:

V_{i} = \sum_{j = 1}^{n} w_{j} s_{i j}

In order to reflect the performance of each network more intuitively, we normalize the score of each network. The evaluation value $G_{i}$ of each network can be expressed as:

G_{i} = \frac{V_{i}}{\sum_{k = 1}^{m} V_{k}}

Then, the network with the highest evaluation value is the one that the user should choose to access.

4. VH-based multi-user access scheme

After completing the network selection, the user confronts the problem of access collision during the procedure of accessing the network, which may cause excessive access delay, even failure to access the network. In order to ensure multi-user effectively access the network, referred to CSMA/CA protocol, a VH-based multi-user access scheme is proposed, which comprehensively considers the channel resource and MAC information of various wireless communication modes in the indoor VLC-based hybrid networks. In this section, VH and its related concepts are presented first, and then the VH-based multi-user access method is proposed. Then the process of VH judgment is depicted in detail. Finally, the corresponding performance indicators are deduced.

4.1 VH and its related concepts

In the mechanism of CSMA/CA, when users need to transmit data, they need to monitor the channel information and maintain a backoff counter. The counter will be self-subtraction only when the channel is idle. If the channel is sensed idle again for a guard period DIFS after the backoff counter is reduced to zero, the users can transmit data. Only when the counters of two or more stations reach zero in the same slot, a collision will occur. Besides, once the number of retransmissions reaches a certain number of times, the contention window will be larger, and the delay for users to access the network will be longer. In addition, the number of retransmissions is limited, if the packet transmission is still unsuccessful after maximum attempts, the packet will be discarded.

We first set the retransmission threshold number k for each network according to the backoff algorithm, average delay and collision probability of each network. Once k is reached, the contention window will be relatively large, and the delay for the user to wait for accessing the network will be relatively long. Then, we set the threshold value THRESH of the network evaluation. If the value of network evaluation is lower than THRESH, it means that the current network has already deteriorated. Here, without loss of generality, we assume that the mobile terminal is multimode and the terminal can sense the channel information of various wireless access patterns, and that the MAC layers in the multimode terminal can exchange information with each other.

Based on the above analysis, we present the concept of backoff lock to control user’s access request for different channels. Its main idea is to use the lock to control the self reducing operation of the backoff counter. When the backoff counter is locked, it can’t be self reduced even if the channel is idle. However, if the backoff counter is unlocked, it can perform normal self reducing operation based on the channel condition.

On this basis, we put forward the concept of virtual handover (VH). Unlike the traditional handover, when the network satisfies the VH conditions, the user can use backoff lock to lock the backoff counter of the current network and send access request packets to VH network, but at this time, no network is switched. The VH conditions include two cases.

One case is that the retransmission number is between k + 1 and the maximum retransmission number M. In this case, the user needs to compare the evaluation value of current network with other networks before retransmitting the access request packet. If the evaluation value of current network is not the highest, but lower than THRESH, the network with the highest evaluation value is selected as the VH network. If the evaluation value of current network is not the highest, but higher than THRESH, the network with the minimum MWAD that represents the minimum waiting delay for the user to access the channel, is selected as the VH network.

Another case is that the retransmission number reaches M. In this case, the user fails to access the current network. But different from the traditional CSMA/CA mechanism, the access request packet will not be discarded immediately. At this time, the network with the highest evaluation value in the remaining candidate networks is selected as the VH network.

The definition of MWAD is as follows:

T_{M W A D} = N A V + B O T + D I F S

where,

N A V

is the estimated time to occupy the channel, and its value can be obtained from the Duration fields of the MAC frame header. The user monitors all the received frame headers and updates the

N A V

of Duration field continuously. The user will not transmit data during the period of

N A V

, thus it greatly reduces the probability of collision.

D I F S

is the guard period defined by the CSMA/CA protocol.

B O T

is the random backoff time based on the backoff mechanism of CSMA/CA, and its value can be calculated as:

B O T = R a n d o w (C W) \cdot slotTime

where,

C W

is the contention window based on the number of retransmission, the value of Random function is a random number that obeys the uniform distribution of [0,

C W

-1], and the

s l o t T i m e

is the slot interval determined by the physical layer.

4.2 VH-based multi-user access method

Based on the concept of VH and backoff lock, we propose a VH-based multi-user access method. The method mainly consists of the following steps:

Step 1: The user executes the network selection method and selects the optimal network A.

Step 2: The user executes the backoff algorithms of the network A and sends out the access request packet to network A. If the packet is transmitted successfully within k retransmissions, the user accesses network A successfully; otherwise, the user enters Step3.

Step 3: From k + 1 retransmission to the maximum retransmission number of M, the user needs to make a judgment whether to execute VH before retransmitting the packet. If do not need VH operation, the user enters Step4; otherwise, enters Step5.

Step 4: The user continues to execute the backoff algorithms of network A and retransmits access request packet to network A. If the packet is transmitted successfully, the user accesses network A successfully. At this moment, the user needs to clear the access request records of other networks; otherwise, the number of retransmissions should be counted. If the retransmission number is less than M, then the user enters Step3. If the retransmission number reaches M, the access request to network A is failed, then the user enters Step11.

Step 5: After selecting the VH network B according to the result of VH judgment, the user needs to use a backoff lock to lock the backoff counter of the current network and saves the access request records of current network, and then executes the VH operation to network B. If the user doesn't have the access request records of network B, it enters Step6; otherwise, enters Step9.

Step 6: The user executes the backoff algorithms of network B and sends out the access request packet to network B. If the packet is transmitted successfully within k' retransmissions, the user accesses network B successfully; otherwise, the user enters Step7.

Step 7: From the k' + 1 retransmission to the maximum retransmission number of M', the user needs to make a judgment whether to execute VH before retransmitting the packet. If do not need VH operation, the user enters Step8; otherwise, enters Step5.

Step 8: The user continues to execute the backoff algorithms of network B on the basis of the access request records and retransmits access request packet to network B. If the packet is transmitted successfully, the user accesses network B successfully. At this moment, the user needs to clear the access request records of other networks. Otherwise, the number of retransmissions should be counted. If the retransmission number is less than M', then the user enters Step 7. If the retransmission number reaches M', the access request to network B is failed, then the user enters Step11.

Step 9: After reading the access request records of network B, the user need to open the backoff lock of network B. If the retransmission number is less than k', then the user enters Step 10; otherwise, enters Step8.

Step 10: The user continues to execute the backoff algorithms of network B and sends out the access request packet to network B. If the packet is transmitted successfully within k' retransmissions, the user accesses network B successfully; otherwise, the user enters Step7.

Step11: The user needs to make a judgment on VH. If there is a VH network satisfying certain conditions, the user enters Step5; otherwise, the user fails to access the hybrid network.

The flow chart of our multi-user access method is shown in Fig. 2.

Fig. 2 Flowchart of our multi-user access method.

Download Full Size | PDF

4.3 Process of VH judgment

During the procedure of multi-user access the hybrid network, the user needs to make a judgment whether to execute VH when the determinant conditions of VH are triggered. In the process of judgment, we consider the performance of the network, the MWAD, and number of retransmission. The process of judgment includes two cases, one case occurs when the retransmission number is between k + 1 to M, and the other one occurs when the retransmission number reaches M.

1) Retransmission number between k + 1 to M
If the retransmission number is between k + 1 to M, the user first needs to update the evaluation value of each network according to the network status in real time, and then compare the evaluation value of the current network (Net_Current) with other networks, either of the following three conditions may occur:
- a) The evaluation value of Net_Current is still the highest, so no VH operation is required.
- b) The evaluation value of Net_Current is not the highest, and it is lower than the threshold THRESH. On this condition, the user needs to read the access request records and determine whether there are other networks that do not reach their maximum retransmission numbers. If none of the networks satisfy the conditions, then no VH operation is required; otherwise, the user needs to select the network with the highest evaluation value as the VH network in the networks that satisfies the conditions.
- c) The evaluation value of Net_Current is not the highest, but it higher than THRESH. On this condition, the user needs to execute the following five steps:
  Step1: Read the access request records and screen out networks that do not reach their maximum retransmission numbers. If such a network doesn’t exist, no VH operation is required. If exists, enter Step2.
  Step2: In the selected networks in Step1, screen out the networks with an evaluation value higher than THRESH and add them to list FILTER_LIST.
  Step3: Read the access request record of each network (Net_i) in FILTER_LIST, if Net_i doesn’t have a backoff window, then the initial backoff value should be generated based on the initial backoff window of Net_i.
  Step4: Calculate the MWAD of each network in FILTER_LIST.
  Step5: Compare the MWAD of Net_Current with the MWAD of each network (Net_i) in FILTER_LIST. The determine formula is expressed as:
  $T_{M A A D} (N e t_C u r r e n t) - T_{M A A D} (N e t_i) \geq Δ T$
  where, $Δ T$ represents the threshold of MWAD, and its value can be set according to the average delay and collision probability.
  If none of the networks in FILTER_LIST satisfies Eq. (20), then no VH operation is required. At this time, we need to use the backoff lock to lock the backoff counter of each network in FILTER_LIST, and the access request records should also be saved.
  If there are networks that satisfy Eq. (20), we need to select a network with the least MWAD in FILTER_LIST, which is the VH network. Meanwhile, we need to use the backoff lock to lock the backoff counter of other networks in FILTER_LIST, and the access request records should also be saved.
2) Retransmission number reaches M
If the retransmission number reaches M, the user needs to read the access request records and determine whether there are networks that do not reach their maximum retransmission numbers. If none of the networks satisfy the conditions, then no VH operation is required, and the user fails to access the hybrid network; otherwise, the user needs to select the network with the highest evaluation value as the VH network in the networks that satisfies the conditions.

4.4 Performance indicators analysis

In order to evaluate the proposed multi-user access scheme, we analyze the performance of collision probability, access delay, and access success ratio. For facilitating the analysis, IR and WiFi channels are used for uplink transmission, and VLC channel is used for downlink transmission. We assume that n users compete to access the hybrid network, in which n₁ users compete to access the infrared channel, and n₂ users compete to access the WiFi channel.

Based on the IrDA: Advanced Infrared (AIr)–Version 1.0 protocol, the collision probability $p_{1}$ is expressed as [26]:

p_{1} = 1 - {(1 - τ_{1})}^{n_{1} - 1}

where,

τ_{1}

is the transmission probability of the packet, it is given by [26]:

τ_{1} = \frac{2}{(C W_{\min} + 1) + 4 p_{1} (\frac{{(1 - p_{1})}^{m + 1} + (2 m + 1) p_{1}^{m + 1} - (m + 1) p_{1}^{m}}{({(1 - p_{1})}^{m + 1} - p_{1}^{m + 1}) (1 - 2 p_{1})})}

where,

C W_{\min}

is the minimum collision window, and m is the number of backoff stages corresponding to the maximum collision window.

Based on the WiFi IEEE 802.11b protocol, the collision probability $p_{2}$ is written as [26]:

p_{2} = 1 - {(1 - τ_{2})}^{n_{2} - 1}

where,

τ_{2}

is the transmission probability of the packet, it is given by [26]:

τ_{2} = \frac{2 (1 - 2 p_{2})}{(1 - 2 p_{2}) \cdot (C W_{m i n}^{'} + 1) + p_{2} \cdot C W_{m i n}^{'} \cdot (1 - {(2 p_{2})}^{m^{'}})}

where,

C W_{\min}^{'}

is the minimum collision window and

m^{'}

is the number of backoff stages corresponding to the maximum collision window.

Due to the collision avoidance mechanism of each network is different and the number of users competing to access each network is also different, the collision probability of the hybrid networks can be expressed as:

p_{h y b r i d} = \frac{\sum_{i = 1}^{n_{1}} p_{1} (i) + \sum_{j = 1}^{n_{2}} p_{2} (j)}{n}

where,

p_{1} (i)

is the collision probability of user

i

in IR AP, and

p_{2} (j)

is the collision probability of user

j

in WiFi hotspot.

In addition, a new concept of waiting access delay (WAD) is defined in this paper, which represents the delay of the user to wait for access the channel. Its definition is as follows:

T_{W A D} = t_{e n d} - t_{s t a r t}

where,

t_{s t a r t}

is the moment when the user initiates an access request to the network,

t_{e n d}

is the moment when the user successfully accesses or fails to access the channels of the hybrid networks.

Referred to literature [27], the access delay is written as:

T_{A c c e s s} = T_{W A D} + T_{t r a n s}

where,

T_{t r a n s}

is the transmission delay of data packets. When IR is selected as the uplink, the transmission delay of data packets can be expressed as:

\begin{array}{l} T_{t r a n s_u p l i n k_I R} = C A S_{I R} + R T S_{I R} + S I F S_{V L C} + C T S_{V L C} + T A T_{I R} \\ + \frac{Payload}{{Rate}_{I R}} {+EOB}_{I R} {+SIFS}_{VLC} {+ACK}_{VLC} \end{array}

where,

C A S_{I R}

is the collision avoidance slot (CAS) defined to cope with collisions caused from hidden station;

R T S_{I R}

is the time required for transmitting the

R T S

frame of IR;

S I F S_{V L C}

is the short inter-frame space of VLC;

C T S_{V L C}

is the time required for transmitting the CTS frame of VLC;

T A T_{I R}

is the turn-around time delay of IR;

P a y l o a d

is the size of data packets;

R a t e_{I R}

is the transmission rates of IR;

E O B_{I R}

is the time required for transmitting end-of-burst frame of IR;

A C K_{V L C}

is the time required for transmitting the ACK frame of VLC.

When WiFi is selected as the uplink, the transmission delay of data packets can be expressed as:

\begin{array}{l} T_{t r a n s_u p l i n k_W i F i} = D I F S_{W i F i} + R T S_{W i F i} + S I F S_{V L C} + C T S_{V L C} + S I F S_{W i F i} \\ + \frac{Payload}{{Rate}_{WiFi}} {+SIFS}_{VLC} {+ACK}_{VLC} \end{array}

where,

D I F S_{W i F i}

is the distributed inter frame space (DIFS) of WiFi;

R T S_{W i F i}

is the time required for transmitting the

R T S_{}

frame of WiFi;

S I F S_{W i F i}

is the short inter-frame space of WiFi;

R a t e_{W i F i}

is the transmission rates of WiFi.

Due to the delay for the user competing to access different channels is different and the number of users competing to access each network is also different, the access delay of the hybrid networks can be expressed as:

T_{A c c e s s_h y b r i d} = \frac{\sum_{i = 1}^{n_{1}} {T_{W A D} (i) + T_{t r a n s_u p l i n k_I R} (i)} + \sum_{j = 1}^{n_{2}} {T_{W A D} (j) + T_{t r a n s_u p l i n k_W i F i} (j)}}{n}

where,

T_{W A D} (i)

is the delay of user

i

to wait for access the IR channel,

T_{t r a n s_u p l i n k_I R} (i)

is the packet transmission delay of user

i

using IR uplink;

T_{W A D} (j)

is the delay of user

j

to wait for access the WiFi channel, and

T_{t r a n s_u p l i n k_W i F i} (j)

is the packet transmission delay of user

j

using WiFi uplink.

Besides, it is assumed that there are $n^{'}$ users fail to access the hybrid network, so the access success ratio can be written as:

p_{s u c e s s} = \frac{n - n^{'}}{n}

5. Simulation and analysis

The simulation scenario is shown in Fig. 1. We use the microcell where user A is located as an example to analyze the performance of network selection and multi-user access in the VLC-based indoor hybrid networks. In our scenario, three candidate networks are available for access selection, that is, WiFi, VLC, and IR.

Referred to literature [28–30], the parameters of WiFi are set. Refereed to literature [31–33], the parameters of IR are set. Refereed to literature [34–36], the parameters of VLC are set. As for the average delay of VLC, we assume that the propagating and processing delay are equivalent to those of IR. Without loss of generality, the transmission delay is inversely proportional to the average transmission rate, so the average delay of VLC is set as 40ms. We normalized the throughput, and its value can be selected between 0 and 1 according to the actual conditions of network. In this paper, we set the normalized throughput of WiFi, IR, and VLC is 0.8, 0.5, and 0.1, respectively. In addition, according to the operator’s statistics and the recent developments of optical wireless communication [33,37], the cost of WiFi, IR, and VLC is set as 0.8, 0.2, and 0.1 respectively. The network parameters of each network are listed in Table 3.

Table 3. Network parameters of candidate networks.

View Table | View all tables in this article

Moreover, we consider three applications, namely conversational service, video stream service, and background service [38]. As for conversational service, users need to communicate with each other, so the limit for acceptable delay is very strict, but there are less hard requirements of BER and bandwidth; As for real-time video stream service, the requirements of bandwidth, end-to-end delay and BER are comprehensively tighter, while there are less stringent on cost; As for background service, the only requirement is that the information should be delivered error free, and users prefer to a lower cost network. According to the characteristics of various services and the scale of importance in Table 1, the preference values of the judgment matrix for conversational service, video stream service, and background service is listed in Table 4, Table 5, and Table 6, respectively.

Table 4. Preference values of the judgment matrix for conversational service.

View Table | View all tables in this article

Table 5. Preference values of the judgment matrix for video stream service.

View Table | View all tables in this article

Table 6. Preference values of the judgment matrix for background service.

View Table | View all tables in this article

After the network parameters of each network and the preference values of the judgment matrix are determined, the subjective weight, objective weight, and comprehensive weight of each decision parameter can be calculated firstly by our proposed optimal network selection method. Taken conversational service as an example, the weights are shown in Fig. 3. The comprehensive weights for different service are shown in Fig. 4. Then under the same requirements of conversational service, the simulation result of selecting an optimal network in the hybrid networks is shown in Fig. 5. Moreover, under different application requirements, the proposed network selection method is adopted to select an optimal network, and the comparison is simulated as shown in Fig. 6.

Fig. 3 Weights for conversational service.

Download Full Size | PDF

Fig. 4 The comprehensive weights for different service.

Download Full Size | PDF

Fig. 5 Network selection under the same conversational requirement through three methods.

Download Full Size | PDF

Fig. 6 Comparison of network selection under different application requirements through the proposed network selection method

Download Full Size | PDF

In Fig. 3, it is observed that the subjective weight of each decision parameter obtained by AHP method is different, which shows that the user has different preference value for different decision parameters. The objective weight obtained by the standard deviation method is also different, which shows the different performance for each network. However, the comprehensive weight obtained by our proposed method is always between the subjective weight and the objective weight, which shows that our comprehensive weight considers both the preference value and performance of each network. Taking the decision parameter of bandwidth and RSS as examples, as for bandwidth, the subjective weight obtained by AHP method is 0.0605 that is less than the objective weight of 0.1436 obtained by the standard deviation method, while the comprehensive weight obtained by our method is 0.1059; as for RSS, the subjective weight obtained by AHP method is 0.2118 that is higher than the objective weight of 0.1369 obtained by the standard deviation method, while the comprehensive weight obtained by our method is 0.1708. In addition, the comprehensive weights of average delay and RSS are higher than those of other decision parameters, which is consistent with the requirement of the user’s conversational service.

In Fig. 4, it is observed that the comprehensive weight of each decision parameter for different service is different, which illustrates that the comprehensive weight obtained by our proposed method can change dynamically with the user’s different service requirement. For example, as for video stream service, the comprehensive weight of BER is 0.1134; while as for background service, the comprehensive weight of BER is 0.2027. In addition, it can be seen that even under the same service requirement, the comprehensive weights of different decision parameters are still quite different. The reason is that the parameter value and the preference value of each decision parameter are different.

In Fig. 5, it is observed that in case of conversational service, comparing with IR and WiFi, the evaluation value of VLC is the highest when using our proposed network selection method, so VLC is selected as the optimal network in the hybrid networks. According to the network parameters listed in Table 3, the average delay of VLC is the lowest, and the parameters such as transmission rate, bandwidth, and cost are also better, so the result of network selection is reasonable. In addition, it can be seen that, as for the IR network, the evaluation value obtained by AHP method is 0.1853, which is less than the value of 0.2828 obtained by the standard deviation method, and the evaluation value obtained by our method is 0.2367. While, as for the WiFi network, the evaluation value obtained by AHP method is 0.2808, which is higher than the value of 0.2300 obtained by the standard deviation method, and the evaluation value obtained by our method is 0.2540. Therefore, the evaluation value of our proposed network selection method is always between that of the AHP and the standard deviation method, which illustrates that our optimal network selection method takes account of the subjective application requirement of users and the objective performance of each network. Moreover, the evaluation value of each network is quite different when using the same method to select the candidate network in the hybrid networks. For example, when our proposed method is used for network selection, the evaluation values of IR, WiFi, and VLC is 0.2367, 0.2540, and 0.5093, respectively. The reason is that the network parameters set for each candidate network are different.

Then, under different application requirements, the comparison of selecting an optimal network by the proposed network selection method is shown in Fig. 6. When the user prepares to access the hybrid networks, VLC network is selected as the optimal network in different applications. As for the conversational application, the evaluation value of WiFi is 0.2540, which is higher than that of IR (0.2367), so WiFi is selected as the suboptimal network. While as for the background application, the evaluation value of WiFi is 0.2059, which is lower than that of IR (0.3343), so IR is selected as the suboptimal network. Therefore, for different application requirements, the optimal network selection is different.

We also simulate the performance of the collision probability, access delay, and access success ratio through the proposed VH-based multi-user access scheme. To simplify the analysis, We assume that: 1) IR and WiFi channels are used for uplink transmission, and VLC channel is used for downlink transmission, so the user can access the network through IR or WiFi channel; 2) when the determinant conditions of VH are triggered, the evaluation value of current network is not the highest, but higher than threshold value, so the THRESH is not discussed here (In practical application, the threshold value THRESH can be set according to the network conditions); 3) the user has a probability p to access the WiFi channel during the initialization phase of access selection.

Referred to literature [26], [39], [40], the transmission rate, the time required for transmitting RTS frame, the turn around time (TAT) delay, the time required for transmitting CTS frame, the time required for transmitting end-of-burst (EOB) and end-of-burst-confirm (EOBC) frame, the collision avoidance slot(CAS) defined to cope with collisions caused from hidden station, the payload size, the minimum collision window, and the maximum number of attempts of IrDA AIr are listed in Table 7. Besides, the retransmission threshold number k and the threshold of MWAD ∆T cannot be too large or too small. If k is greater than 10 or ∆T is higher than 1ms, the user will not execute the VH operation with a larger probability, which results in no noticeable improvements of the network performance. If k is less than 3 or ∆T is less than 100us, the user will frequently trigger VH judgment. Without loss of generality, we set k as 5 and ∆T as 400us. The MAC and PHY parameters of IrDA AIr are listed in Table 7.

Table 7. AIr MAC and PHY parameters.

View Table | View all tables in this article

Referred to literature [26–28], the RTS, CTS, ACK, slot time, SIFS, DIFS, minimum collision window, and maximum number of attempts of 802.11b are set. Meantime, 8-chip complementary code keying (CCK) is employed as the modulation scheme to provide the high transmission rates of 5.5Mbit/s and 11Mbit/s. In order to have a valid comparison between the two protocols, the transmission rate of 802.11b is set as 5.5Mbit/s; the payload size is also set as 16384bits. Similar to Table 7, when the retransmission threshold number k^’ is greater than 6, the user will not execute the VH operation with a larger probability, so k′ is set as 3; the threshold of MWAD is set as 400us. The MAC and PHY parameters of 802.11b are listed in Table 8.

Table 8. 802.11b MAC and PHY parameters.

View Table | View all tables in this article

Referred to literature [35,41], the MAC and PHY parameters of VLC are listed in Table 9.

Table 9. VLC MAC and PHY parameters.

View Table | View all tables in this article

Through the proposed VH-based multi-user access scheme, when p is set as 0.25, 0.5 and 0.75, the performances of the collision probability, access delay and access success ratio of the VLC-based hybrid networks are analyzed by simulation, and comparison with IR and WiFi 802.11b network are carried out. The results are shown in Fig. 7, Fig. 8, and Fig. 9, respectively.

Fig. 7 Comparison of collision probability for different networks.

Download Full Size | PDF

Fig. 8 Comparison of access delay for different networks.

Download Full Size | PDF

Fig. 9 Comparison of access success ratio for different networks.

Download Full Size | PDF

In Fig. 7, it can be seen that the collision probabilities of IR network, WiFi network, and VLC-based hybrid networks increase when the number of the users increases, but the collision probability of the VLC-based hybrid networks is always lower than those of IR and WiFi network. Even though the number of the users reaches 28, the collision probability of VLC-based hybrid networks is up to 0.420, while the collision probability of IR network is 0.536 and the one of WiFi network is 0.448. When the probability p is different, the collision probability of VLC-based hybrid networks is different too. The reason is that the number of the users initially assigned to each network in hybrid networks is different and the collision avoidance mechanism of each network is also different. It can also be seen that, when the number of the users is 4, the collision possibility of IR network is still up to 0.36. The reason is that the initial collision window defined by the protocol of IrDA is 8 and the backoff algorithm employees a linear adjustment of the contention window.

In Fig. 8, it can be seen that the access delay becomes long when the number of the users increases, but the access delay of VLC-based hybrid networks is always lower than those of IR and WiFi 802.11b network. For the VLC-based hybrid networks, when the number of the users reaches 32, the access delay is low as 0.104s, which meets the requirement of the user access latency; even though the number of the users reaches 60, the access delay is less than 0.469s. In addition, it can be seen that, the results of the access delay are fluctuant. The reason is that the backoff window selected by each user is random, so the access delay for the user to wait for access the channel is different. Because the CAS is 800us and the collision possibility of IR network is relatively large, the access delay of IR network is relatively higher than those of other networks.

In Fig. 9, it can be seen that, for the VLC-based hybrid networks, the access success ratio can reach 100% when the number of the users is below 60. The reason is that the users can send access request to other networks when the user fails to access one network, which is the main advantage of hybrid networks. It can also be seen that, when the number of the users is more than 32, the access success ratio of IR network is higher than that of WiFi network in most cases. The reason is that the maximum retransmission number of IR network is 20, while the one of WiFi network is 7. Similar to the access delay, the results of the access success ratio are also fluctuant. The reason is that the backoff window selected by each user is random, and the collision is different for each specific access process.

6. Conclusion

Recently, with the development of VLC technology, more and more researchers combine VLC with other wireless communications to build indoor hybrid networks. In order to realize indoor VLC-based hybrid networks, it is important to select an optimal network that best suits the requirements of the users and to realize multi-user access.

In this paper, firstly, a network selection method based on MADM is proposed. The subjective weight of the parameters, the objective weight of the parameters, and the evaluation value of each network is obtained, respectively. Then, a minimization mathematical model is established to obtain the partition coefficients of the subjective and objective weights of each parameter. Secondly, a VH-based multi-user access scheme is proposed, which comprehensively considers the channel and MAC information of various wireless communication modes in indoor hybrid networks. By controlling the self subtraction of backoff counter, a concept of backoff lock is presented to control the access request of the user to different channels. Meanwhile, the concept of MWAD and VH are also presented to reduce the access delay and improve the access success ratio. When the determinant conditions of VH are triggered, the user can use the backoff lock to lock the current access request, and select an optimal or the least MWAD network to send access request packet. Even if the user fails to access a network, the access request packet will not be discarded immediately, and the user can still send access request packet to other networks. The formulas of the performance indicators are deduced, such as collision probability, access delay, and access success ratio.

Finally, under the same conversational service requirements and under different application requirements, the optimal network selection is analyzed by simulation, respectively. The results show that the proposed network selection method can satisfy the users' application requirement, and the evaluation value obtained by our network selection method is also in accordance with the objective network performance. The performances of the collision probability, access delay and access success ratio for the VH-based multi-user access scheme are also simulated. The access delay is low as 0.104s when the number of the users reaches 32, and the access success ratio can reach 100% when the number of the users is below 60. The results show that the VH-based multi-user access scheme can reduce the collision probability and access delay, and increase the access success ratio. Our network selection method and VH-based multi-user access scheme are feasible and effective for VLC-based hybrid networks.

Funding

National Natural Science Foundation of China (NSFC) (61172080, 61771357)

References

1. M. Ayyash, H. Elgala, A. Khreishah, V. Jungnickel, T. Little, S. Shao, M. Rahaim, D. Schulz, J. Hilt, and R. Freund, “Coexistence of WiFi and LiFi toward 5G: concepts, opportunities, and challenges,” IEEE Commun. Mag. 54(2), 64–71 (2016). [CrossRef]

2. J. Vucic and K. D. Langer, “High-speed visible light communications: state-of-the-art,” in OFC/NFOEC (2012), pp. 1–3.

3. L. Feng, R. Q. Hu, J. Wang, P. Xu, and Y. Qian, “Applying VLC in 5G networks: architectures and key technologies,” IEEE Netw. 30(6), 77–83 (2016). [CrossRef]

4. I. C. Lu, C. H. Yeh, D. Z. Hsu, and C. W. Chow, “Utilization of 1-GHz VCSEL for 11.1-Gbps OFDM VLC wireless communication,” IEEE Photonics J. 8(3), 1–6 (2016). [CrossRef]

5. Z. Li, C. Zhang, D. Sun, H. Yang, and J. Song, “A real-time high-speed visible light communication system based on RGB-LEDs,” in 2017 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (2017), pp. 1–4. [CrossRef]

6. J. Kim and H. Park, “A coding scheme for visible light communication with wide dimming range,” IEEE Photonics Technol. Lett. 26(5), 465–468 (2014). [CrossRef]

7. I. Stefan and H. Haas, “Hybrid visible light and radio frequency communication systems,” in 2014 IEEE 80th Vehicular Technology Conference (2014), pp. 1–5.

8. M. B. Rahaim, A. M. Vegni, and T. D. C. Little, “A hybrid radio frequency and broadcast visible light communication system,” in 2011 IEEE GLOBECOM Workshops (2011), pp. 792–796.

9. X. Bao, X. Zhu, T. Song, and Y. Ou, “Protocol design and capacity analysis in hybrid network of visible light communication and OFDMA systems,” IEEE Trans. Vehicular Technol. 63(4), 1770–1778 (2014). [CrossRef]

10. A. Sgora, D. D. Vergados, and P. Chatzimisios, “An access network selection algorithm for heterogeneous wireless environments,” in The IEEE symposium on Computers and Communications (2010), pp. 890–892.

11. Q. Y. Song and A. Jamalipour, “Network selection in an integrated wireless LAN and UMTS environment using mathematical modeling and computing techniques,” IEEE Wirel. Commun. 12(3), 42–48 (2005). [CrossRef]

12. R. K. Goyal and S. Kaushal, “Effect of utility based functions on fuzzy-AHP based network selection in heterogenous wireless networks,” 2015 2nd International Conference on Recent Advances in Engineering & Computational Sciences (RAECS) (2015), pp. 1–5.

13. B. Hu, N. Li, and J. Zhang, “IASS: An intelligence access selection scheme for heterogeneous networks,” 2015 International Conference on Information Networking (ICOIN) (2015), pp. 531–536. [CrossRef]

14. N. N. Sui, D. M. Zhang, W. Zhong, and C. Wang, “Network selection for heterogeneous wireless networks based on multiple attribute decision making and evolutionary game theory,” 2016 25th Wireless and Optical Communication Conference (WOCC) (2016), pp. 1–5.

15. L. Li, Y. X. Zhang, B. Fan, and H. Tian, “Mobility-aware load balancing scheme in hybrid VLC-LTE networks,” IEEE Commun. Lett. 20(11), 2276–2279 (2016). [CrossRef]

16. G. Bianchi, L. Fratta, and M. Oliveri, “Performance evaluation and enhancement of the CSMA/CA MAC protocol for 802.11 wireless LANs,” Personal, Indoor and Mobile Radio Communications, 1996. PIMRC'96., Seventh IEEE International Symposium on (1996), pp. 392–396.

17. J. Ni, B. Tan, and R. Srikant, “Q-CSMA: Queue-length-based CSMA/CA algorithms for achieving maximum throughput and low delay in wireless networks,” IEEE/ACM Trans. Netw. 20(3), 825–836 (2012). [CrossRef]

18. H. Ma, X. Li, H. W. Li, P. Y. Zhang, S. X. Luo, and C. Yuan, “Dynamic optimization of IEEE 802.11 CSMA/CA based on the number of competing stations,” in 2004 IEEE International Conference on Communications (2004), pp. 191–195. [CrossRef]

19. W. Guo, Q. Li, H. Y. Yu, and J. H. Liu, “A parallel transmission MAC protocol in hybrid VLC-RF network,” J. Commun. 10(1), 80–85 (2015). [CrossRef]

20. V. V. Mai, T. C. Thang, and A. T. Pham, “CSMA/CA-based uplink MAC protocol design and analysis for hybrid VLC/Wifi networks,” in 2017 IEEE International Conference on Communications Workshops (ICC Workshops) (2017), pp. 457–462. [CrossRef]

21. D. O’Brien, R. Turnbull, H. L. Minh, G. Faulkner, O. Bouchet, P. Porcon, M. E. Tabach, E. Gueutier, M. Wolf, L. Grobe, and J. H. Li, “High-Speed Optical Wireless Demonstrators: Conclusions and Future Directions,” J. Lightwave Technol. 30(13), 2181–2187 (2012). [CrossRef]

22. T. L. Saaty, “How to make a decision: the analytic hierarchy process,” Interfaces 24(6), 19–43 (1990). [CrossRef]

23. Y. M. Wang, “A method based on standard and mean deviations for determining the weight coefficients of multiple attributes and its applications,” Appl. Statistics Management 22(3), 22–26 (2003).

24. M. Lahby, L. Cherkaoui, and A. Adib, “Performance analysis of normalization techniques for network selection access in heterogeneous wireless networks,” in 2014 9th International Conference on Intelligent Systems: Theories and Applications (SITA-14) (2014), pp. 1–5.

25. Manisha and N. P. Singh, “Optimal network selection using MADM algorithms,” in 2015 2nd International Conference on Recent Advances in Engineering & Computational Sciences (RAECS) (2015), pp. 1–6.

26. P. Chatzimisios, V. Vitsas, A. C. Bouconvalas, D. Kleftouns, and M. Tsoulfa, “Packet delay performance comparison of the IEEE 802.11 and IrDA AIr CSMA/CA protocols in high-speed wireless lans,” in Iasted International Conference on Internet and Multimedia Systems and Applications ACTA Press (2009), pp. 171–176.

27. T. Sakurai and H. L. Vu, “MAC access delay of IEEE 802.11 DCF,” IEEE Trans. Wirel. Commun. 6(15), 1702–1710 (2007). [CrossRef]

28. “IEEE Standard for Information Technology - Telecommunications and information exchange between systems - Local and Metropolitan networks - Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: Higher Speed Physical Layer (PHY) Extension in the 2.4 GHz band,” (IEEE Std 802.11b-1999, 2000).

29. P. Mahasukhon, M. Hempel, H. Sharif, T. Zhou, S. Ci, and H. H. Chen, “BER analysis of 802.11b networks under mobility,” in 2007 IEEE International Conference on Communications (2007), pp. 4722–4727. [CrossRef]

30. H. S. Mewara, M. K. Saini, and R. Kumar, “Throughput and delay analysis of access point in IEEE 802. 11b wireless LAN using opnet simulator,” Rev. Lat. Am. Enfermagem 17(1), 59–65 (2009). [PubMed]

31. P. Chatzimisios and A. C. Boucouvalas, “Packet delay analysis of the advanced infrared (AIr) CSMA/CA MAC protocol in optical wireless LANs,” Int. J. Commun. Syst. 18(3), 307–331 (2005). [CrossRef]

32. “Advanced infrared physical layer specification (AIr-PHY) - Version 1.0” Infrared Data Association (1998).

33. J. M. Kahn and J. R. Barry, “Wireless infrared communications,” Proc. IEEE 85(2), 265–298 (1997). [CrossRef]

34. H. L. Minh, D. O’Brien, G. Faulkner, L. B. Zeng, K. Lee, D. Jung, and Y. J. Oh, “80 Mbit/s Visible Light Communications using pre-equalized white LED,” in 2008 34th European Conference on Optical Communication (2008), Brussels, 2008, pp. 1–2.

35. “IEEE Standard for Local and Metropolitan Area Networks–Part 15.7: Short-Range Wireless Optical Communication Using Visible Light,” (IEEE Std 802.15.7–2011, 2011).

36. H. L. Minh, Z. Ghassemlooy, D. O’Brien, and G. Faulkner, “Indoor gigabit optical wireless communications: Challenges and possibilities,” in 2010 12th International Conference on Transparent Optical Networks (2010), pp. 1–6.

37. K. D. Langer and J. Grubor, “Recent developments in optical wireless communications using Infrared and visible light,” in 2007 9th International Conference on Transparent Optical Networks (2007), pp. 146–151.

38. “TS 22.105 v8.2.0: Services and service capabilities (Release 8)” 3GPP, (2006).

39. “Advanced Infrared (AIr) MAC Draft Protocol Specification–Version 1.0” Infrared Data Association, (1999).

40. V. Vitsas and A. C. Boucouvalas, “Performance analysis of the collision avoidance procedure of the advanced infrared (AIr) CSMA/CA protocol for wireless LANs,” in Vehicular Technology Conference. IEEE 55th Vehicular Technology Conference. VTC Spring 2002 (Cat. No.02CH37367) (2002), pp. 1502–1506. [CrossRef]

41. T. Shang, T. Jiang, Y. T. Yang, P. Wang, and Y. Liu, “Multi-users network model and the corresponding networking scheme for indoor VLC systems,” Opt. Express 23(9), 11600–11618 (2015). [CrossRef] [PubMed]

Intensity of importance	Relative importance of two elements
1	Equal importance
3	Moderate importance
5	Strong importance
7	Very strong importance
9	Extreme importance
2,4,6,8	Intermediate values

Order	RI value
1	0
2	0
3	0.51
4	0.88
5	1.12
6	1.25
7	1.35
8	1.41
9	1.46
10	1.48
11	1.52
12	1.54

Network parameter	WiFi	IR	VLC
Average transmission rate	5.5Mbps	4 Mbps	80 Mbps
Bandwidth	20 MHz	25 MHz	45 MHz
RSS	−35 dBm	−20 dBm	4 dBm
BER	4 × 10⁻⁶	1 × 10-6	9 × 10⁻⁶
Average delay	60 ms	150 ms	40 ms
Normalized throughput	0.8	0.5	0.1
Cost	0.8	0.2	0.1

Conversational service	Averagedelay	Bandwidth	RSS	BER	Transmissionrate	Throughput	Cost
Average delay	1	7	2	5	7	8	4
Bandwidth	1/7	1	2/7	5/7	1	8/7	4/7
RSS	1/2	7/2	1	5/2	7/2	8/2	2
BER	1/5	7/5	2/5	1	7/5	8/5	4/5
Transmission rate	1/7	1	2/7	5/7	1	8/7	4/7
Throughput	1/8	7/8	2/8	5/8	7/8	1	4/8
Cost	1/4	7/4	2/4	5/4	7/4	2	1

Video stream service	Averagedelay	Bandwidth	RSS	BER	Transmissionrate	Throughput	Cost
Average delay	1	3	4	2	3	8	8
Bandwidth	1/3	1	4/3	2/3	1	8/3	8/3
RSS	1/4	3/4	1	1/2	3/4	2	2
BER	1/2	3/2	2	1	3/2	4	4
Transmission rate	1/3	1	4/3	2/3	1	8/3	8/3
Throughput	1/8	3/8	1/2	1/4	3/8	1	1
Cost	1/8	3/8	2/4	1/4	3/8	1	1

Network selection method based on MADM and VH-based multi-user access scheme for indoor VLC hybrid networks

Abstract

1. Introduction

2. VLC-based hybrid network model

3. Network selection method based on MADM

3.1 AHP for subjective weight

3.2 The standard deviation method for objective weight

3.3 Minimization mathematical model for partition coefficients

3.4 SAW for evaluation value

4. VH-based multi-user access scheme

4.1 VH and its related concepts

4.2 VH-based multi-user access method

4.3 Process of VH judgment

4.4 Performance indicators analysis

5. Simulation and analysis

6. Conclusion

Funding

References

Cited By

Figures (9)

Tables (9)

Equations (31)

Optics Express

Parameter	value
Transmission rate	4 Mbps
RTS	244us
TAT	200us
CTS	232us
EOB	232us
EOBC	232us
CA Slot	800us
Payload size	16384bits
Min contention window, CW	8
Max number of attempts, M	20
Threshold of retransmission, k	5
Threshold of MWAD, ∆T	400us

Parameter	Value
Transmission rate	5.5 Mbps
RTS	176 us
CTS	152 us
ACK	152 us
Slot time	20 us
Payload size	16384bits
SIFS	10us
DIFS	50us
Min contention window, CW	32
Max number of attempts, M	7
Threshold of retransmission, k′	3
Threshold of MWAD, ∆T	400us

Parameter	Value
Transmission rate	80 Mbps
PHY header	64bits
MAC header	240bits
CTS	112bits + PHY header + MAC header
ACK	112bits + PHY header
SIFS	16us
DIFS	32 us
Slot time	8us