Fuzzy based optimal and traffic-aware restricted access window mechanism for dense IoT networks

Abstract

IEEE 802.11ah defines amendments to IEEE 802.11 to support the Internet of Things (IoT). IEEE 802.11ah implements restricted access window (RAW) mechanism to reduce the contention and energy consumption in dense IoT networks. The RAW mechanism is a group-based MAC protocol that partitions the devices into various groups and confines the channel access of a group of devices to the restricted time interval known as the RAW slot. However, the standard does not specify, grouping mechanism, duration of RAW slots, and the number of RAW slots while configuring the RAW mechanism. In an IoT network, each device has distinct transmission requirements. Thus, it is necessary to find the optimal number of RAW slots that can maximize the network performance, to group the devices with similar transmission requirements and to assign a RAW slot that adaptively varies with the traffic requirements of the respective group. In this paper, we exploit fuzzy logic to find the optimal number of RAW slots by considering network size, collision probability, and modulation and coding schemes. Further, we propose a traffic-aware adaptive RAW slot allocation (TARA) scheme that uses fuzzy c-means clustering algorithm to group the devices with similar traffic requirements and to assign each group with a RAW slot whose duration adaptively varies with the transmission requirements of the devices. We have also presented a simple yet accurate analytical model to evaluate the performance of the RAW mechanism. Results show that the optimal number of RAW slots found using fuzzy logic significantly enhances the performance of the RAW mechanism in terms of throughput and energy consumption. Further, it is observed that the TARA scheme can effectively meet the traffic requirements of different group of devices when compared to the uniform grouping scheme. Finally, extensive simulations are conducted using ns-3 to validate the analytical results.

Keywords

Internet of Things (IoT)restricted access window (RAW)IEEE 802.11ah fuzzy logic,Wi-Fi HaLow,fuzzy c-means clustering

1 Introduction

Internet of Things (IoT) is an emerging trend and a new revolution in technology. It interconnects every living and non-living things together through the Internet to enrich the quality of human life.It seamlessly connects myriads of things and evolves as a vast and powerful heterogeneous network. Things in the IoT network includes not only people but also trees, gadgets, vehicles and everything from large industrial equipment to small household objects. Everyday objects in IoT are equipped with sensors, actuators, processors, and communication interfaces to communicate with each other without human intervention [1]. According to a forecast, 50 billion devices will be interconnected through the Internet by 2020 [2]. This paradigm has numerous applications in environmental and disaster monitoring, vehicular transportation system, smart grids, clinical aids, farming, home automation, and many others [3].

The standardization of IoT is a challenging task. Because, IoT requires a networking technology that can provide large-scale connectivity, broad coverage, and high energy efficiency. Technologies like RFID, Bluetooth, Zigbee etc. are recommended because of their higher data rates, lower energy consumption, but they have limited coverage and scalability. On the other hand, WiMAX and LTE have extensive coverage and high data rates, but consumes more energy. Hence, these technologies are not suitable for IoT [4].

Despite being a popular WLAN standard, IEEE 802.11 suffers from severe contention in the dense networks. To get around the trade-off among the existing wireless networking technologies, IEEE 802.11ah is introduced by the IEEE 802 LAN/MAN standards committee (LMSC) to support IoT applications [5, 6]. IEEE 802.11ah operates at sub 1 GHz ISM frequency bands. The standard provides a broad coverage up to 1 km, a data rate of 78 Mbps and connectivity to 8000 devices [7].

IEEE 802.11ah standard introduces restricted access window (RAW), a group-based MAC protocol, to reduce the channel contention and the energy consumption in dense networks [8]. However, the access point (AP) faces a few challenges while configuring the RAW mechanism. First, the RAW mechanism partitions the network into several groups. Unfortunately, the IEEE 802.11ah draft standard does not specify any grouping schemes. Thus, the design of the grouping schemes is still an open research issue. Most of the existing literature uses the uniform grouping (UG) scheme as the default grouping scheme [9 –11], because the UG scheme is simple to implement as it divides the network into an equal number of groups and does not consider any constraints of the devices. Second, the channel time is partitioned into several RAW slots and each RAW slot is assigned to a group of devices. One of the most challenging tasks is to find the optimal number of RAW slots that can maximize the network performance. Because choosing a lesser number of RAW slots increases the number of devices per group and rises the contention in the network. On the other hand, choosing a higher number of RAW slots results in wastage of channel resources. Next is the duration of each RAW slot. Because every device in the network has distinct transmission requirements. For example, a temperature sensor monitoring a hot surface plate or a heartbeat sensor monitoring the patient health has different sampling rates and channel access requirements. So both the nodes cannot be assigned with a RAW slot of equal duration. Thus, it is necessary to adaptively vary the duration of RAW slots according to the transmission requirements of the devices rather than providing them with a RAW slot of equal duration [12].

In this article, we exploit the fuzzy logic system (FLS) to find the optimal number of RAW slots. The FLS considers network size, collision probability, and modulation and coding schemes (MCSs) as inputs to find the optimal number of RAW slots1. Further, we propose a traffic-aware adaptive RAW slot allocation (TARA) scheme that uses fuzzy c-means (FCM) clustering algorithm to group the devices with similar transmission requirements and to allocate each group with a RAW slot whose duration adaptively varies according to the transmission requirements of the devices. The throughput and energy consumption of the RAW mechanism is assessed using the proposed scheme with a simple yet accurate analytical model. The analytical model presented in this paper is based on [13, 14]. The performance of RAW mechanism is evaluated using the TARA scheme and compared with the UG scheme. The results show that the proposed scheme outperforms the UG scheme and significantly enhances the performance of the RAW mechanism.

Rest of the paper is organized as follows: Section 2 presents the related work done. Section 3 discuss the salient features of IEEE 802.11ah protocol. Section 4 presents an analytical model to evaluate the performance of the RAW mechanism. Finding the optimal number of RAW slots using fuzzy logic is proposed in Section 5. TARA scheme is proposed in Section 6. Section 7 discusses the analytical and simulation results. Finally, Section 8 concludes the paper. The major contributions of this paper are summarized as:

Motivated by soft computing techniques, a novel method is proposed to find the optimal number of RAW slots using fuzzy logic. This scheme provides the optimal number of RAW slots by considering the network size, collision probability, and MCSs.

We exploit fuzzy logic to design TARA scheme that classifies the devices into several groups and adaptively vary the duration of a RAW slot according to the channel requirements of each group of devices.

A simple yet accurate analytical model is presented to evaluate the performance of the RAW mechanism in terms of saturation throughput and energy consumption per bit.

Using this analytical model the performance of RAW mechanism is evaluated with the optimal number of RAW slots and results are compared with the non-optimal number of RAW slots. Further, this model is used to evaluate the performance of the TARA scheme and compare it with the UG scheme.

Results show that the optimal number of RAW slots found using the fuzzy logic approach significantly increases the throughput and decreases energy consumption. It is also observed that the TARA scheme can effectively meet the traffic requirements of different group of devices.

The analytical results are corroborated by extensive ns-3 simulations.

2 Related work

Many works have been proposed to assess the performance of IEEE 802.11ah draft standard. Tian et al. in [8], found throughput, latency, and energy consumption to assess the performance of the RAW mechanism. In [15], authors considered non-cross slot boundary case and proposed various holding schemes to assess the performance of the RAW mechanism. Authors in [16], evaluated the range and bit error rate for the outdoor urban environment under multi-path fading in IEEE 802.11ah network. To reduce RTS collisions in large networks and increase the performance of the system, a sector-based RTS/CTS access scheme is introduced in [17]. Hidden matrix-based regrouping algorithm is proposed in [18] to elevate hidden nodes in the network.

Duration of RAW plays a vital role in network performance. Park et al. in [19] found the optimal RAW parameters by relating the number of devices and the size of the RAW period to maximize the success probability. In [20], authors extended the connectivity using relay nodes and found the duration of the RAW period by using traffic intensity. In [21], the authors provided the guidelines to choose the number of groups based on the traffic arrival rate and duration of the beacon interval. Khorov et al. in [22] discussed the process of estimating the number of devices to select an appropriate RAW slot by finding the distribution of time required by the devices to transmit their frames.

Scalability is the other important feature in IEEE 802.11ah. RAW divides a large network into several groups and allows each group to contend in their respective RAW slot. Unfortunately, IEEE 802.11ah standard does not provide any grouping schemes in the draft. The authors in [23] proposed a simple grouping scheme to mitigate contention by dividing the network into various groups. The scheme divides the network into several groups and selects a group head to minimize the channel contention. Tung et al. in [24], proposed a load-balanced grouping for a saturated network to improve the channel utilization. A simple spatial grouping based on geographic division strategy to decrease the collision probability by reducing hidden node pairs is proposed in [25]. In [26], authors proposed a grouping model based on the measurement of received signal strength.

In [27], authors hasten the association between AP and the network devices by proposing a linear dependence of link set-up time to the number of associating devices. Authors in [28] used probability theory to propose an access window algorithm to optimize the uplink communication energy consumption in 802.11ah networks.

Many works have been published in the literature based on fuzzy logic and its application in wireless networks. Most of the works present clustering algorithms, routing protocols, techniques to improve the energy consumption and lifetime of the network. In [29], authors presented a fuzzy-based clustering architecture to improve the energy consumption and network lifetime. Similarly, authors in [30] proposed a clustering method to extend the network lifetime. Authors in [31] predicted the residual energy of the devices and clusters them using the fuzzy logic. A novel energy efficient routing protocol is presented in [32] based on fuzzy logic. In [33], authors presented power allocation scheme for cooperative and delay constrained energy harvesting nodes.

Many schemes were developed to evaluate the performance of 802.11ah and its RAW mechanism. In this paper, a novel scheme is proposed to find the optimal number of RAW slots using fuzzy logic. Further, TARA scheme is proposed to group the devices and adaptively allocate the RAW slot to each group-based on their transmission requirements. Then, the throughput and energy consumption of the IEEE 802.11ah RAW mechanism is assessed using the optimal number of RAW slots found using the fuzzy approach in a dense IoT network. It is observed that the optimal number of RAW slots can significantly improve the throughput and reduce energy consumption. It is also shown that the proposed TARA scheme can effectively meet the traffic requirements of different group of devices.

3 IEEE 802.11ah: A wireless LAN standard

IEEE 802.11ah IEEE 802.11ah defines amendments to IEEE 802.11 to support IoT [34, 35]. The emerging wireless standard provides better performance in terms of connectivity, throughput, coverage, and energy consumption in dense networks. This section emphasizes the notable features of the IEEE 802.11ah standard.

3.1 Physical layer

The IEEE 802.11ah PHY layer inherits the salient features of IEEE 802.11ac standard [36]. The standard operates in sub 1 GHz (S1G) license-exempt spectrum and supports 1 MHz, 2 MHz, 4 MHz, 8 MHz, and 16 MHz channels. The availability of S1G spectrum differs from country to country. The standard adopts orthogonal frequency division multiplexing (OFDM), multiple-input multiple-output (MIMO) and downlink multi-user MIMO (DL MU-MIMO). IEEE 802.11ah inherits ten different modulation and coding schemes (MCS0-MCS9) with different data rates that are one-tenth of IEEE 802.11ac. Among all the MCSs, MCS0 is the most robust coding scheme that can provide an extended transmission range up to 850 m [37]. Data-rates of various MCSs are shown in Table 1 for 2 MHz channel.

Table 1
Data rates of various MCSs for 2 MHz Channel

MCS Modulation CR Bits Per OFDM Symbol Data Rate Mbps

MCS0 BPSK 1/2 52 0.65

MCS1 QPSK 1/2 104 1.30

MCS2 QPSK 3/4 104 1.95

MCS3 16QAM 1/2 208 2.60

MCS4 16QAM 3/4 208 3.90

MCS5 64QAM 2/3 312 5.20

MCS6 64QAM 3/4 312 5.85

MCS7 64QAM 5/6 312 6.50

MCS8 256QAM 3/4 416 7.80

MCS	Modulation	CR	Bits Per OFDM Symbol	Data Rate Mbps
MCS0	BPSK	1/2	52	0.65
MCS1	QPSK	1/2	104	1.30
MCS2	QPSK	3/4	104	1.95
MCS3	16QAM	1/2	208	2.60
MCS4	16QAM	3/4	208	3.90
MCS5	64QAM	2/3	312	5.20
MCS6	64QAM	3/4	312	5.85
MCS7	64QAM	5/6	312	6.50
MCS8	256QAM	3/4	416	7.80

Table 2

Parameters used to obtain analytical and simulation results

Payload	256 bytes
MAC header	14 bytes
PHY header	156 bits
Ps_Poll	20 bytes
ACK	14 bytes
basic_datarate	650 Kbps
delta	1 μs
T_sym	40 μs
sigma	52 μs
SIFS	160 μs
DIFS	264 μs
P_Tx	255 mW
P_Rx	135 mW
P_idle	1.3 mW
RAW duration	500 ms
Beacon interval	equal to RAW duration
Rate control algorithm	constant
Wi-Fi mode	all MCSs

3.2 MAC layer

IEEE 802.11ah standard proposes multiple enhancements in MAC layer. This section describes critical improvements relevant to this context. A unique 13-bit association identifier (AID) is assigned by the AP for every device in the network. Thus, it extends the length of the traffic indication map (TIM) up to 8192 bits [38]. The bits in the AID represent the hierarchical organization of the devices into pages, blocks, sub-blocks, and devices.

Every bit in the TIM bitmap corresponds to a device if the AP has a data packet destined to it. The delivery-TIM beacon is the bitmap of the TIM beacons that are broadcasted periodically by the AP as shown in Fig. 1. The duration between two consecutive TIM beacons is divided into multiple RAW periods and contention access period [39]. Further, each RAW period has several RAW slots. The AP broadcasts the beacon with RAW parameters set information element (RPS-IE) that specifies the RAW start time, duration of RAW and AIDs of the devices. Accordingly, a device chooses a RAW slot using, $χ_{slot} = ({AID}_{N} + offset) \mod L,$ (1) where χ_slot is the index of the RAW slot, AID_N is the association identifier of N^th device, offset is a numerical value in the mapping function, and L is the number of RAW slots in the RAW period [7].

Fig.1

Beacon interval in IEEE 802.11ah.

Further, IEEE 802.11ah introduces the RAW mechanism to reduce contention among the devices. RAW mechanism divides the devices into several groups, partitions the channel time into several RAW slots and assigns each group with a RAW slot. Then, each group of devices is allowed to contend in their respective RAW slots using enhanced distributed channel access (EDCA) as shown in Fig. 1. The device who wins the channel contention sends a data frame followed by PS_Poll frame. The AP acknowledges each received packet after SIFS duration.

4 Analytical modeling of IEEE 802.11ahand its RAW mechanism

We assume all the devices in the network are completely associated and always have a packet for transmission i.e., saturation condition. We consider an error-free channel and assess the uplink performance of the network. The channel is divided into mini-slots of duration σ. We consider a network of size N which is divided into K groups, each of size g and divides the duration of RAW period T_R into K RAW slots, each of duration T_slot such that $\sum_{j = 1}^{K} T_{slot, j} = T_{R}$ . Then each group of devices is assigned with a RAW slot in which they contend for the channel access using legacy DCF mechanism. DCF mechanism implements the binary exponential back-off algorithm. A device in the j^th RAW slot senses the channel to be idle for DIFS duration before initiating the back-off counter. The back-off counter is uniformly chosen from [0, W₀ - 1], where W₀ is the minimum contention window. For every successful transmission, the contention window is reset to W₀ and for every unsuccessful transmission, the contention window is doubled up to the maximum contention window (CW_max). The contention window for the i^th back-off stage is given by,

$W_{i} = {\begin{matrix} 2^{i} \times W_{0}; & 0 \leq i \leq m - 1, \\ 2^{m} \times W_{0}; & m \leq i \leq R, \end{matrix}$ (2) where m is the number of back-off stages and R is the maximum number of retries made by a device to transmit a packet before discarding2. A device gets the transmission opportunity when the back-off counter decrements to zero. In between two consecutive transmissions, the device initiates the back-off counter followed by DIFS duration [40].

Let s (t) and b (t) be the stochastic processes representing the back-off stage and back-off counter in a j^th RAW slot. A bi-dimensional discrete time Markov chain (DTMC) model illustrating the back-off mechanism is shown in the Fig. 2.

Fig.2

A bi-dimensional discrete time Markov chain model for back-off mechanism.

Let $b_{i, k} = lim_{t \to \infty} P {s (t) = i, b (t) = k}$ be the stationary distribution of the Markov chain, where i ∈ [0, m] and k ∈ [0, W_i - 1]. We have, $b_{i, 0} = p_{c, j}^{i} b_{0, 0} .$ (3) Due to the chain regularities we have, $b_{i, k} = \frac{W_{i} - k}{W_{i}} b_{i, 0} i \in [0, m], k \in [0, W_{i} - 1] .$ (4) All the b_i,k in Equation (4) can be expressed as a function of b_0,0. Imposing the normalization condition, we have, $\sum_{i = 0}^{m} \sum_{k = 0}^{W_{i} - 1} b_{i, k} = 1,$ (5) $\begin{matrix} b_{0, 0} = \frac{\begin{matrix} 2 (1 - p_{c, j}^{R + 1}) (1 - 2 p_{c, j}) \end{matrix}}{(\begin{matrix} (1 - 2 p_{c, j}) (1 - p_{c, j}^{R + 1}) \\ + W_{0} (1 - (2 p_{c, j})^{m + 1}) (1 - p_{c, j}) \\ + W_{0} 2^{m} p_{c, j}^{m + 1} (1 - 2 p_{c, j}) (1 - p_{c, j}^{R - m}) \end{matrix})} . \end{matrix}$

The probability that a device transmits a packet in a randomly chosen mini-slot of a j^th RAW slot is given by, $τ_{j} = \sum_{i = 0}^{m} b_{i, 0} = \sum_{i = 0}^{m} p_{c, j}^{i} b_{0, 0} .$ (6)

Then, the conditional collision probability that a transmitted packet encounters collision is given by, $p_{c, j} = 1 - (1 - τ_{j})^{g - 1} .$ (7)τ_j and p_c,j can be obtained by solving Equations (6) and (7). Let P_tr,j be the probability that any one of the g devices transmits in a mini-slot of the j^th RAW slot, $P_{tr, j} = 1 - (1 - τ_{j})^{g} .$ (8)

P_s,j is the probability that a device successfully transmits a packet and (g - 1) devices differ the transmission conditioned that there is at least one transmission in the mini-slot of a j^th RAW slot,

$P_{s, j} = \frac{g τ_{j} {(1 - τ_{j})}^{g - 1}}{1 - (1 - τ_{j})^{g}} .$ (9)

4.1 Saturation throughput

The saturation throughput S_j of a j^th RAW slot can be calculated as,

$\begin{matrix} S_{j} & = & \frac{(\begin{matrix} Average information \\ transmitted in a mini - slot \end{matrix})}{Average duration of a mini - slot} \begin{matrix} \end{matrix}, \\ = & \frac{P_{tr, j} P_{s, j} E [P]}{(1 - P_{tr, j}) σ + P_{tr, j} P_{s, j} T_{s} + P_{tr, j} (1 - P_{s, j}) T_{c}}, \end{matrix}$ (10) where E [P] is the average payload size, T_s and T_c are the average time for successful transmission and collision, respectively.

$\begin{matrix} T_{s} & = & T_{PS_Poll} + T_{E [P]} + 2 T_{ACK} + 3 SIFS \\ + DIFS + 3 δ, \\ T_{c} & = & T_{PS_Poll} + DIFS + δ . \end{matrix}$ (11)

Here δ is the propagation delay, T_{PS_Poll} is the duration of PS_Poll frame, T_ACK is the duration of ACK frame, and T_E[P] is the duration of the payload.

The time taken to transmit the payload T_E[P] is a function of data rate corresponding to the MCSs, that can be calculated using Equation (12). Similarly, the duration of the control frames (PS_Poll, ACK) is calculated by Equation (13). It is noteworthy to point out that, MCS0 (basic _ datarate) is used to transmit the control frames and PHY header. $L_{sym}^{basic_datarate}$ is the bits per OFDM symbol when the basic data rate is used.

$\begin{matrix} T_{E [P]} (Rate) = & ⌈ \frac{8 \times (E [P] + MAC)}{\frac{Rate}{basic_datarate} \times L_{sym}^{basic_datarate}} ⌉ \\ \times T_{sym} + T_{PHY}, \end{matrix}$ (12)

$\begin{matrix} T_{control} & = & ⌈ \frac{8 \times ControlFrame}{L_{sym}^{basic_datarate}} ⌉ \times T_{sym} \\ + T_{PHY} . \end{matrix}$ (13)

4.2 Energy consumption

The energy consumption is defined as the ratio of total energy consumed during a transmission of a device to the successfully transmitted data. In DCF mechanism, a device can be in either a back-off state, freezing state, or a transmission state. Thus, each device consumes energy in four parts: E_b is the energy consumed during the back-off process. E_f is the energy consumed when a device freezes its back-off counter. E_s and E_c are the energies consumed due to a successful transmission and collision. Therefore, the energy consumption is defined as [41]: $η_{j} = \frac{E_{b} + E_{f} + E_{s} + E_{c}}{P_{tr, j} P_{s, j} E [P]} .$ (14)

The average energy consumed during the back-off process is given by, $E_{b} = E [B] σ P_{idle},$ (15) where E [B] is the average number of back-off slots which is given by, $E [B] = \sum_{i = 0}^{R} p_{c, j}^{i} (1 - p_{c, j}) \sum_{j = 0}^{i} \frac{W_{j} - 1}{2} .$ (16)

In a RAW slot, among the g devices, a node overhears a transmission when one of g - 1 devices is successfully transmitting conditioned that there is at least one transmission in the j^th RAW slot. Therefore, the success probability is given by, $P_{s, j}^{'} = \frac{(g - 1) τ_{j} {(1 - τ_{j})}^{g - 2}}{1 - (1 - τ_{j})^{g - 1}} .$ (17)

The average number of transmissions overheard by a device during the back-off process is given by, $N_{0} = \frac{E [B] p_{c, j}}{1 - p_{c, j}} .$ (18)

Therefore, the energy consumed by a device due to overhearing the other devices during the back-off process is given by, $E_{f} = N_{0} [P_{s, j}^{'} T_{s} + (1 - P_{s, j}^{'}) T_{c}] P_{idle} .$ (19)

Similarly, the average number of transmission attempts before a successful transmission is given by, $N_{t} = \sum_{i = 0}^{R} {ip}_{c, j}^{i} (1 - p_{c, j}) .$ (20)

Then the average energy consumed due to successful transmission and collision is given by,

$\begin{matrix} E_{s} & = & P_{Tx} (T_{PS_Poll} + T_{E [P]}) + P_{Rx} (T_{s} - T_{PS_Poll} \\ + T_{E [P]}), \\ E_{c} & = & N_{t} [P_{Tx} (T_{PS_Poll}) + P_{Rx} (T_{c} - T_{PS_Poll})] . \end{matrix}$ (21)

Therefore, the average energy consumed to successfully transmit a packet in the j^th RAW slot is given by, $E_{T} = E_{b} + E_{f} + E_{s} + E_{c} .$ (22)

Finally, the energy consumption per bit η_j is given by, $η_{j} \approx \frac{E_{T}}{P_{tr, j} P_{s, j} E [P]} .$ (23)

5 Fuzzy logic-based approach to find theoptimal number of RAW slots

In this section, we exploit fuzzy logic to find the optimal number of RAW slots rather than using the conventional analytical approach. Generally, analytical approaches introduce computational overhead that can reduce the performance of the AP. Hence, soft computing techniques would be a feasible solution to reduce computational overhead [33]. Although many soft computational techniques exist, the motivation behind selecting fuzzy logic is its efficiency in decision-making [42]. We briefly introduce the architecture of fuzzy logic system (FLS) before discussing the proposed scheme.

Fuzzy logic is used to model the uncertainties of a system to get desired results [42]. According to boolean logic, an element belongs to a set with a degree of either 0 or 1. In fuzzy logic, every set has a mathematically represented membership function, and every membership function is defined by linguistic terms. Let X be a set with elements denoted by x. Then, the fuzzy set $R$ in X is given by, $R = {(x, φ_{R} (x)) x \in X} .$ (24) where $φ_{R} (x)$ is the membership function and x is an element that belongs to the fuzzy set $R$ with a degree of membership in the interval [0, 1]. The membership functions may be piece-wise linear functions (triangular and trapezoidal), Gaussian distribution function, quadratic or polynomial curves [33]. Figure 3 illustrates the structure of FLS. The FLS maps each input value also called as crisp value to their respective fuzzy sets. This process is known as fuzzification. The fuzzy inference system (FIS) combines the input membership functions with fuzzy rules to derive an output. The fuzzy rules are framed based on human intuition according to the given problem [42]. This process is called as defuzzification. Generally, center of gravity (COG) or min-max methods are used for defuzzification.

Fig.3

Proposed scheme to find optimal number of RAW slots.

Figure 3 shows the block diagram of the FLS used to find the optimal number of RAW slots. The FIS takes network size, collision probability, and MCS as inputs and finds the optimal number of RAW slots based on the fuzzy rules.

5.1 Fuzzification

Fuzzification is defined as the process of mapping the input variable to their respective fuzzy set. As shown in Fig. 3, we have the following three inputs to the FIS to find the optimal number of RAW slots. (i) network size N, (ii) collision probability calculated using Equation (29), (iii) MCSs listed in Table 1. Each input variable has multiple fuzzy sets defined by certain linguistic terms, and each fuzzy set is represented by a membership function. Due to computational efficiency and less complexity, we consider trapezoidal membership function $φ_{R} (x; ν_{1}, ν_{2}, ν_{3}, ν_{4})$ for all the inputs as given below. $φ_{R} (x) = {\begin{matrix} 0; & for x \leq ν_{1} \\ \frac{x - ν_{1}}{ν_{2} - ν_{1}}; & for ν_{1} \leq x \geq ν_{2} \\ 1; & for ν_{2} \leq x \geq ν_{3} \\ \frac{ν_{4} - x}{ν_{4} - ν_{3}}; & for ν_{3} \leq x \geq ν_{4} \\ 0; & for x \geq ν_{4} \end{matrix}$ (25)

Figure 4 illustrates the membership functions of both input and output variables. For example, the input variable N is represented by four fuzzy sets { $R_{Less}$ , $R_{Medium}$ , $R_{High}$ , $R_{Dense}$ } with membership functions {φ_Less (N) , φ_Medium (N) , φ_High (N) , φ_Dense (N)} defined by the linguistic variable ’Less’, ’Medium’, ’High’ and ’Dense’. Every membership function of both input and output variables follows the same trapezoidal function according to Equation 25. Therefore, ${\begin{matrix} φ_{R} (N) & = φ_{R} (N; L_{1}, L_{2}, L_{3}, L_{4}) \\ φ_{R} (P_{c}) & = φ_{R} (P_{c}; L_{1}, L_{2}, L_{3}, L_{4}) \\ φ_{R} (MCS) & = φ_{R} (MCS; L_{1}, L_{2}, L_{3}, L_{4}) \\ φ_{R} (L_{opt}) & = φ_{R} (L_{opt}; L_{1}, L_{2}, L_{3}, L_{4}) \end{matrix}$ (26) where $R$ is the fuzzy set of input or output variables represented by their respective linguistic variables. L₁, L₂, L₃, and L₄ are the limits of the trapezoidal function that are chosen intuitively.

Fig.4

Membership functions for input and output variables of FLS.

5.2 Fuzzy rules

The core of FLS is fuzzy rules. Fuzzy rules are used to evaluate the degree of membership of input and output variables. To find the optimal number of RAW slots, we define hundred and forty four fuzzy rules. For every set of input variables, the AP finds the output based on the fuzzy rules. Some of the fuzzy rules are defined as follows:

IF N=LOW, and Pc=LOW, and MCS=MC1 THEN L_opt = Opt1;

IF N=Medium, and Pc=Medium, and MCS=MC2 THEN L_opt = Opt2;

IF N=High, and Pc=High, and MCS=MC4 THEN L_opt = Opt4;

IF N=Dense, and Pc=Dense, and MCS=MC8 THEN L_opt = Opt8;

5.3 Defuzzification

Each fuzzy rule is evaluated for every set of input variables and mapped to the corresponding fuzzy set of output variable. FLS computes the output using the COG method from the output of each rule. For example, if the fuzzy rule is mapped to a fuzzy set given by Equation (24), then the output is calculated using the following equation. $ouptut = \frac{\int^{X} x φ_{R} (x) dx}{\int^{X} φ_{R} (x) dx} .$ (27)

6 Traffic-aware adaptive RAW slot allocationscheme

In this section, we propose a traffic-aware adaptive RAW slot allocation scheme using FCM to classify the devices according to their channel access requirements. We systematically formulate the duration of each RAW slot as a function of the channel access requirements of the devices. Initially, every device in the network listens to the periodically broadcasted beacons and undergoes association procedure with the AP. Prior to the association, each device estimates the required channel access time during its full functionality using, $Δ_{n} = \sum_{i = 0}^{R} W_{i} σ + E [T_{s, n}] + Δ_{f, n} .$ (28)

The first term in Equation (28) is the duration of the back-off counter. E [T_s,n] is the duration between transmission of a packet till reception of the acknowledgment frame and is calculated by using Equation (11). Δ_f,n is the duration in which the back-off counter freezes due to overhearing of other transmission. Therefore,

$Δ_{f, n} = R (E [T_{s, n}] + SIFS + T_{ACK_timeout}) .$ (29)

Here, T_{ACK_timeout} is the duration of acknowledgment timeout. Having estimated the transmission times, each device communicates Δ_n with the AP using the additional field of short MAC header in the association request frame [43]. The AP finds the optimal number of RAW slots by using the scheme presented in the previous section. The optimal number of RAW slots (L_opt) is equal to the optimal number of groups (K_opt), as each group is assigned with a RAW slot. According to Algorithm 1, the AP sorts all the transmission requirements of the devices into an array A (n). Then, the AP uses FCM to classify the devices into K_opt groups according to their channel access requirements. Having grouped the devices into K_opt groups, the AP assigns each group with a RAW slot whose duration is a function of the channel requirements of the respective group. Thus, the duration of the k^th RAW slot is given by,

$T_{slot, k} = β_{k} \times T_{RAW},$ (30) where β_k is the scaling factor that depends on the traffic requirements of each group. Therefore, $β_{k} = \frac{C_{k}}{\sum_{i = 1}^{K} C_{i}} .$ (31)

Algorithm 1

TARA scheme

1: Inputs:

2: ∈ = 0.01; \\ Convergence value

3: U; \\ Randomly initialized partition matrix of size K_opt × N

4: l = 0; \\ Number of iterations

5: z = 2; \\ Fuzziness parameter

6: for each device do

7: A (n) = Δ_n;

8: end for

\\Classifying A (n) using FCM

9: while |U^(l+1) - U^(l)| ≤ ∈do

10:

C_{k} = \frac{\sum_{i = 1}^{n} (u_{ki})^{z} d_{i}}{\sum_{i = 1}^{n} (u_{ki})^{z}}, k \in [1, K_{opt}]

; \\ Calculate K-centroids corresponding to K-groups

11:

u_{ki} = \sum_{j = 1}^{k} {[(\frac{d_{i} - C_{k}}{d_{j} - C_{k}})^{\frac{2}{z - 1}}]}^{- 1}

\\ Each element of partition matrix U

12: l=l+1;

13: end while

14: AP forms each row of the matrix U into a group with average transmission requirements given by C;

15: T_slot,k = β_kT_R; \\ The duration of the RAW slot of each group

16:

β_{k} = \frac{C_{k}}{\sum_{i = 1}^{K} C_{i}}

;

where C_i is the average transmission requirements of the i^th group. Using Equation (10), the data transferred in each RAW slot is calculated using,

$D_{j} = S_{j} \times T_{slot}, j \in [1, K_{opt}] .$ (32)

7 Analytical and simulation results

Analytical and simulation results are presented in this section. The analytical model presented in Section 4 is evaluated using MATLAB. The analytical results are validated using the open source network simulator ns-3 [8]. In this paper, we consider a network of size N uniformly deployed around the AP. The AP employs TARA scheme and divides the network into K groups based on their transmission requirements according to Algorithm 1. Let the duration of the RAW period be 500 ms. According to the TARA scheme, the duration of each RAW slot is adaptively varied according to the transmission requirements of the respective group. The AP always runs FIS to find the optimal number of RAW slots (L_opt). As mentioned earlier, the L_opt is equal to the optimal number of groups (K_opt) in the network, as each group is assigned with a single RAW slot. Table 2 lists the parameters used to obtain the analytical and simulation results.

Figure 5 illustrates the performance of the RAW mechanism for various groups using different MCSs. We consider a network of size N = 256 devices that is divided into K ∈ {8, 16, 32, 64, 128, 256} groups. It is observed from the figure that the performance of RAW mechanism increases in terms of throughput and energy consumption with an increase in the number of groups. Because an increase in K reduces the number of devices per group and decreases the contention among them. Meanwhile, it is observed for larger values of K, there is a slight increase in energy consumption and a gradual decrease in the throughput. The main reason behind this phenomenon is the wastage of channel resources. Although the contention is minimized for larger values of K, the devices in their respective groups access the channel using the conventional DCF, wasting the channel time in the back-off process. This unwanted phenomenon reduces the saturation throughput. But during the back-off time, the device will be idle and consumes less power. Hence, there is no significant increase in energy consumption. The feasible solution to get the best performance of the RAW mechanism is to divide the RAW period with an optimal number of RAW slots.

Fig.5

Saturation throughput and energy consumption performance for different number of groups using different MCSs.

Figure 6 shows the optimal number of RAW slots found using the fuzzy technique. The L_opt obtained using the proposed scheme is used to assess the performance of the RAW mechanism. Figure 7 shows the performance of RAW mechanism using the proposed fuzzy technique. Since the IEEE 802.11ah standard [2] supports up to 8192 devices, the performance of RAW mechanism is evaluated for a network of size N = 8192 devices. The optimal number of RAW slots L_opt that can maximize the networkperformance is found using the fuzzy based proposed scheme. In Fig. 7, the performance of RAW mechanism is evaluated using L_opt and the results are compared with the non-optimal value L_non-opt = 32. It is observed that there is a significant increase in throughput and decrease in energy consumption for the L_opt found using fuzzy logic. The results justify the optimal number of RAW slots found using fuzzy logic can significantly increase the performance of RAW mechanism in contrast to the conventional analytical technique.

Fig.6

Optimal number of RAW slot found using fuzzy logic.

Fig.7

Saturation throughput and energy consumption performance over different network sizes using MCS1.

Table 3 compares the performance among optimized RAW mechanism, non-optimized RAW mechanism and legacy DCF mechanism in terms of throughput and energy consumption for a dense IoT network. A network of size N up to 8192 devices is considered and the performance of RAW mechanism is evaluated with L_non-opt = 4 and L_opt using MCS0. It is clearly shown from the Table 3 that the performance of the RAW mechanism is significantly improved using fuzzy logic approach. Because in a dense IoT network, numerous devices contend for the channel which increases the contention and collisions among the devices in the network. Because of the increased collisions, the throughput is decreased and the energy consumption is increased. Since the RAW mechanism partitions the network into several groups and allows the channel access in their respective RAW slots, it outperforms the legacy DCF in a dense network, promoting IEEE 802.11ah as a prevalent standard for IoT. Results also show that the throughput is improved and energy consumption is reduced by using the optimal number of RAW slots found by fuzzy based approach.

Table 3

Performance comparison between optimal RAW mechanism, non-optimal RAW mechanism and legacy DCF in a dense IoT network

Number of Devices	Throughput (Mbps)						Energy Consumption per bit (nJ/bit)
	RAW				DCF		RAW				DCF
	L _non-opt		L _opt				L _non-opt		L _opt
	Ana.	Sim.	Ana.	Sim.	Ana.	Sim.	Ana.	Sim.	Ana.	Sim.	Ana.	Sim.
1024	0.3066	0.3034	0.3397	0.3397	0.2241	0.2169	706.85	699.43	519.56	520.11	5364.1	5361.7
2048	0.2865	0.2816	0.3377	0.3380	1.1505	1.1550	842.45	828.54	536.01	537.03	5218.8	5213.1
3072	0.2676	0.2610	0.3354	0.3359	0.0330	0.0304	999.76	977.06	548.18	549.59	5259.2	5257.8
4096	0.2481	0.2400	0.3334	0.3341	0.0095	0.0085	1199.3	1164.2	558.17	559.91	5247.0	5246.6
5120	0.2278	0.2181	0.3317	0.3326	0.0024	0.0021	1463.4	1410.2	566.81	568.86	5244.1	5244.0
6144	0.2063	0.1952	0.3301	0.3311	0.0006	0.0005	1822.5	1742.6	574.55	576.87	5243.5	5243.5
7168	0.1838	0.1715	0.3287	0.3299	0.0001	0.0001	2320.6	2201.0	581.63	584.20	5243.4	5243.4
8192	0.1607	0.1476	0.3273	0.3286	0.0000	0.0000	3023.3	2843.7	588.20	591.02	5243.4	5243.4

Ana.=Analytical, Sim.=Simulation.

Table 4 illustrates the variation of the amount of data transferred according to the transmission requirements of the devices and compares it with the UG scheme. We consider N = 32 devices which are divided into K_opt = 4 groups. Here group-1 has highest average transmission requirements and group-4 with the lowest transmission requirements. The duration of each RAW slot is found according to the Algorithm 1 and is listed in Table 4. We use MCS8 with 2 MHz channel at the PHY layer. Unlike the UG scheme, the proposed scheme classifies the devices according to their transmission requirements and assigns each group with a RAW slot whose duration varies adaptively according to Equation (30). Thus, the group-1 devices with longer duration of RAW slot access the channel for a long time and can transfer more amount of data when compared to group-4 devices which has shorter duration of RAW slot. Whereas the UG scheme equally divides the network into several groups and assigns each group with RAW slots of equal duration. Hence, same amount of data is transferred by all the groups in their respective RAW slot. From the results given in Table 4, it is observed that significant amount of data is transferred by group-1 which is having highest transmission requirements compared to the groups with lower transmission requirements. Hence, the TARA scheme can effectively meet the traffic requirements of different group of devices.

Table 4

Data transferred in a RAW slot

Number of Groups	β	Duration of RAW slot (ms)	Data transferred in a RAW slot (Kb)
			TARA Scheme		UG Scheme
			Ana.	Sim.	Ana.	Sim.
group-1	0.3308	0.1654	229.88010	229.8319	173.7304	173.6940
group-2	0.2951	0.1475	205.0714	205.0284	173.7304	173.6940
group-3	0.2281	0.1140	158.5116	158.4784	173.7304	173.6940
group-4	0.1460	0.0730	101.4586	101.4373	173.7304	173.6940
Aggregate throughput			694.9216	694.7760	694.9216	694.7760

Ana.=Analytical, Sim.=Simulation.

8 Conclusion

In this paper, a novel scheme is proposed to find the optimal number of RAW slots using fuzzy logic that can significantly improve the throughput and energy consumption. The proposed scheme provides the optimal number of RAW slots (optimal number of groups) by considering network size, collision probability and MCSs. Further, TARA scheme is proposed to classify the devices with similar transmission requirements and allocates each group with a RAW slot whose duration adaptively varies with the traffic requirements of a group of devices. We have presented a simple yet accurate analytical model to evaluate the throughput and energy consumption of IEEE 802.11ah RAW mechanism, by modifying Bianchi’s model for DCF. The RAW mechanism is evaluated using the TARA scheme and compared with the default UG scheme. From the results, it is clear that the TARA scheme significantly enhances the performance of the RAW mechanism and outperforms the UG scheme. Finally, extensive simulation studies have been conducted to validate the analytical findings.

References

Miorandi

, Sicari

, Pellegrini

F.D.

and Chlamtac

, Internet of things: Vision, applications and research challenges, Ad Hoc Networks 10(7) (2012), 1497–1516.

Man

L.A.N.

, Standards Committee, and IEEE Computer, IEEE Standard 802.11ac-2013: Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, Amendment 4 Enhancements for Very High Throughput for Operation in Bands below 6 GHz, 2013.

Zanella

, Bui

, Castellani

, Vangelista

and Zorzi

, Internet of things for smart cities, IEEE Internet of Things Journal 1(1) (2014), 22–32.

Palattella

M.R.

, Accettura

, Vilajosana

, Watteyne

, Grieco

L.A.

, Boggia

and Dohler

, Standardized protocol stack for the internet of (Important) things, IEEE Communications Surveys Tutorials 15(3) (Third 2013), 1389–1406.

Banos

, Afaqui

M.S.

, Lopez

and Garcia

, Throughput and Range Characterization of IEEE 802.11ah, IEEE Latin America Transactions 15(9) (2017), 1621–1628.

Corrales Madueño

, Stefanović

Č.

and Popovski

, Reliable and efficient access for alarm-initiated and regular M2M traffic in IEEE 802.11ah systems, IEEE Internet of Things Journal 3(5) (2016), 673–682.

Khorov

, Lyakhov

, Krotov

and Guschin

, A survey on IEEE 802.11ah: An enabling networking technology for smart cities, Computer Communications 58(2015), 53–69. Special Issue on Networking and Communications4 for Smart Cities.

Tian

, Famaey

and Latré

, Evaluation of the IEEE 802.11ah Restricted AccessWindow mechanism for dense IoT networks, In 2016 IEEE 17th International Symposium on A World of Wireless, Mobile and Multimedia Networks (WoWMoM), 2016, pp. 1–9.

Justin Gopinath

and Nithya

, Mathematical and simulation analysis of contention resolution mechanism for IEEE 802.11ah networks, Computer Communications 124 (2018), 87–100.

10.

Kim

, Hwang

, Um

, Yoo

, Jung

and Park

, Throughput performance optimization of super dense wireless networks with the renewal access protocol, IEEE Transactions on Wireless Communications 15(5) (2016), 3440–3452.

11.

Khorov

, Lyakhov

and Yusupov

, Two-Slot Based Model of the IEEE 802.11ah Restricted Access Window with Enabled Transmissions Crossing Slot Boundaries, 2018 IEEE 19th International Symposium on “A World of Wireless, Mobile and Multimedia Networks” (WoWMoM), 2018, pp. 1–9.

12.

Chang

, Lin

K.C.

and Chen

, Traffic-aware sensor grouping for ieee 802.11ah networks: Regression based analysis and design, IEEE Transactions on Mobile Computing 18(3) (2019), 674–687.

13.

Bianchi

, Performance analysis of the IEEE 802.11 distributed coordination function, IEEE Journal on Selected Areas in Communications 18(3) (2000), 535–547.

14.

Chatzimisios

, Boucouvalas

A.C.

and Vitsas

, IEEE 802.11 packet delay-a finite retry limit analysis, In Global Telecommunications Conference, 2003 GLOBECOM ’03 IEEE, volume 2, 2003, pp. 950–954.

15.

Qutab ud din

, Hazmi

, Badihi

, Larmo

, Torsner

and Valkama

, Performance analysis of IoT-enabling IEEE 802.11ah technology and its RAW mechanism with non-cross slot boundary holding schemes, In 2015 IEEE 16th International Symposium on A World of Wireless, Mobile and Multimedia Networks (WoWMoM), 2015, pp. 1–6.

16.

Domazetović

, Kočan

and Mihovska

, Performance evaluation of IEEE 802.11ah systems, In 2016 24th Telecommunications Forum (TELFOR), 2016, pp. 1–4.

17.

Aust

, Prasad

R.V.

and Niemegeers

I.G.M.M.

, Sectorbased RTS/CTS access scheme for high density WLAN sensor networks, In 39th Annual IEEE Conference on Local Computer Networks Workshops, 2014, pp. 697–701.

18.

Yoon

S.-G.

, Seo

J.-O.

and Bahk

, Regrouping algorithm to alleviate the hidden node problem in 802.11ah networks, Computer Networks 105 (2016), 22–32.

19.

Park

C.W.

, Hwang

and Lee

T.J.

, Enhancement of IEEE 802.11ah MAC for M2M communications, IEEE Communications Letters 18(7) (2014), 1151–1154.

20.

Ahmed

and Hussain

M.I.

, Relay-based IEEE 802.11ah network: A Smart City solution, In 2016 Cloudification of the Internet of Things (CIoT), 2016, pp. 1–6.

21.

Seo

J.O.

, Nam

, Yoon

S.G.

and Bahk

, Group-based contention in IEEE 802.11ah networks, In 2014 International Conference on Information and Communication Technology Convergence (ICTC), 2014, pp. 709–710.

22.

Khorov

, Krotov

and Lyakhov

, Modelling machine type communication in IEEE 802.11ah networks, In 2015 IEEE International Conference on Communication Workshop (ICCW), 2015, pp. 1149–1154.

23.

Abichar

and Chang

J.M.

, Group-based medium access control for IEEE 802.11n wireless LANs, IEEE Transactions on Mobile Computing 12(2) (2013), 304–317.

24.

Chang

T.C.

, Lin

C.H.

, Lin

K.C.J.

and Chen

W.T.

, Load-Balanced Sensor Grouping for IEEE 802.11ah Networks, In 2015 IEEE Global Communications Conference (GLOBECOM), 2015, pp. 1–6.

25.

Dong

, Wu

, Gao

and Zhao

, An efficient spatial group restricted access window scheme for IEEE 802.11ah networks, In 2016 Sixth International Conference on Information Science and Technology (ICIST), 2016, pp. 168–173.

26.

Ghasemiahmadi

, Li

and Cai

, RSS-Based Grouping Strategy for Avoiding Hidden Terminals with GS-DCF MAC Protocol, In 2017 IEEE Wireless Communications and Networking Conference (WCNC), 2017, pp. 1–6.

27.

Bankov

, Khorov

and Lyakhov.

, The Study of the Centralized Control Method to Hasten Link Set-up in IEEE 802.11ah Networks, In Proceedings of European Wireless 2015; 21th European Wireless Conference, 2015, pp. 1–6.

28.

Wang

, Li

, Chai

K.K.

, Chen

and Schormans

, Energy-aware adaptive restricted access window for IEEE 802.11ah based smart grid networks, In 2015 IEEE International Conference on Smart Grid Communications (Smart GridComm), 2015, pp. 581–586.

29.

Akila

I.S.

and Venkatesan

, A fuzzy based energy-aware clustering architecture for cooperative communication in WSN, The Computer Journal 59(10) (2016), 1551–1562.

30.

Arjunan

and Sujatha

, Lifetime maximization of wireless sensor network using fuzzy based unequal clustering and ACO based routing hybrid protocol, Applied Intelligence (2017).

31.

Lee

J.S.

and Cheng

W.L.

, Fuzzy-logic-based clustering approach for wireless sensor networks using energy predication, IEEE Sensors Journal 12(9) (2012), 2891–2897.

32.

Shen

, Wang

, Hung

P.C.K.

and Lai

C.F.

, An efficient centroid-based routing protocol for energy management in WSN-assisted IoT, IEEE Access 5 (2017), 18469–18479.

33.

Yousaf

, Ahmad

, Ahmed

and Haseeb

, Fuzzy power allocation for opportunistic relay in energy harvesting wireless sensor networks, IEEE Access 5 (2017), 17165–17176.

34.

Park

, IEEE 802.11ah: Sub-1-GHz license-exempt operation for the internet of things, IEEE Communications Magazine 53(9) (2015), 145–151.

35.

Wang

and Fapojuwo

A.O.

, A survey of enabling technologies of low power and long range machine-to-machine communications, IEEE Communications Surveys Tutorials 19(4) (2017), 2621–2639.

36.

IEEE Standard for Information technology– Telecommunications and information exchange between systems– Local and metropolitan area networks– Specific requirements– Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications– Amendment 4: Enhancements for Very High Throughput for Operation in Bands below 6 GHz. IEEE Std 802.11ac(TM)-2013, (Amendment to IEEE Std 802.11-2012, as amended by IEEE Std 802.11ae-2012, IEEE Std 802.11aa-2012, and IEEE Std 802.11ad-2012), 2013, pp. 1–425.

37.

Sun

, Choi

and Choi

, IEEE 802.11ah: A long range 802.11 WLAN at sub 1 GHz, Journal of ICT Standardization 1 (2013), 1–26.

38.

Zheng

and Lei

, TIM encoding for IEEE 802.11ah based WLAN, In 2014 IEEE International Conference on Communication Systems, 2014, pp. 559–563.

39.

Lei

and Rhee

S.H.

, Performance improvement of sub 1 GHz WLANs for future IoT environments, Wireless Personal Communications 93(4) (2017), 933–947.

40.

Zheng

, Ni

, Cai

, Pan

, Ghosh

and Doppler

, Performance analysis of group-synchronized DCF for dense IEEE 802.11 networks, IEEE Transactions on Wireless Communications 13(11) (2014), 6180–6192.

41.

Kuo

W.-K.

, Energy efficiency modeling for ieee 802.11 dcf system without retry limits, Computer Communications 30(4) (2007), 856–862. Nature-Ispired Distributed Computing.

42.

Zadeh

L.A.

, Soft computing and fuzzy logic, IEEE Software 11(6) (1994), 48–56.

43.

Tian

, Camtepe

and Tian

Y.C.

, A deadline-constrained 802.11 mac protocol with QoS differentiation for soft real-time control, IEEE Trans Ind Informat 12(2) (2016), 544–554.