Sage Journals: Discover world-class research

Abstract

Cognitive radio sensor networks offer a promising means of meeting rapidly expanding demand for wireless sensor network applications in new monitoring and objects tracking fields. Several challenges, particularly in terms of quality of service provisioning, arise because of the inherited capability-limitation of end-sensor nodes. In this article, an efficient resource allocation scheme, improved Pliable Cognitive Medium Access Protocol, is proposed to tackle multilevel of heterogeneity in cognitive radio sensor networks. The first level is the network’s application heterogeneity, and the second level is the heterogeneity of the radio environment. The proposed scheme addresses scheduling and radio channel allocation issues. Allocation-decision making is centralized, whereas spectrum sensing is distributed, thereby increasing efficiency and limiting interference. Despite the limited capabilities of the sensor’s networks, the effectiveness of the proposed scheme also includes increasing the opportunity to utilize a wider range of the radio spectrum. improved Pliable Cognitive Medium Access protocol is quite appropriate for critical communications that gain attention in the next 5G of wireless networks. Simulation results and the comparison of the proposed protocol with other protocols indicate the robust performance of the proposed scheme. The results reveal the significant effectiveness, with only a slight trade-off in terms of complexity.

Keywords

Heterogeneous network quality of service resource allocation scheduling critical communications

Introduction

Scholarly and commercially interests in wireless sensor networks (WSNs) are increasing day by day. The development of WSN technology enables many emerging applications in new domains in addition to the traditional domains. Potential WSN applications include emerging monitoring and objects tracking applications such as smart cities, e-Health, industrial control and monitoring, asset tracking, and supply chain management. Implementing these applications requires quality of service (QoS) provisioning while coping with the capability constraints of the end-sensor node.^1,2 Low data rate, low power consumption, low transmission range, and low computational capabilities are some of these constraints in addition to the hostile radio environment that the WSNs are operating in.^3,4

A new radio spectrum management paradigm is proposed, offering a solution to the wireless spectrum scarcity problem while improving the efficiency of licensed spectrum bands utilization. Cognitive radio (CR) is an emerging technology that mitigates the overutilization situation in the license-free spectrum by enabling an opportunistic spectrum access (OSA) mechanism to the licensed spectrum for unlicensed devices. The unlicensed devices are called the secondary users (SUs), and the licensed devices are called the primary users (PUs), which have priority in using the spectrum. In other words, the SUs have many obligations toward the PUs to prevent the harmful mutual interference between signals. The absence of licensed-user activity is called a spectrum hole or a white space, as shown in Figure 1. Typically, an SU performs spectrum sensing (SS) processes in the targeted radio spectrum and detects periods during which the channel is free of PU activities.^5,6 The SU is permitted to access the medium in those periods of time and utilize the radio channel.^7,8

Figure 1.

Opportunistic spectrum access.

The energy detection (ED) is the common SS technique. A channel is considered busy if the energy of the detected signal exceeds a predefined threshold value.^9–11 Let $S (n)$ denote the PU’s transmitted signal. $H_{0}$ and $H_{1}$ indicate the hypothesis of the absence and the presence of the PU signal, respectively. The detected signal can be expressed as

E (n) = {\begin{matrix} W (n); H_{0} \\ W (n) + S (n); H_{1} \end{matrix}

(1)

where $W (n)$ represents the additive white Gaussian noise (AWGN). The channel’s occupied and idle times are denoted by $T_{on}$ and $T_{off}$ , respectively. Accordingly, the probability of the channel being free is expressed as

P_{off} = \frac{T_{off}}{T_{on} + T_{off}}

(2)

The detection probability, $P_{de}$ , determines the probability that the channel is sensed as busy whereas it is actually busy. In contrast, the false alarm probability, $P_{fa}$ , determines the probability that the channel is detected busy whereas it is actually idle. The determining of the channel status is conducted according to a comparison between the value of the received signal’s energy, $E$ , and a predefined threshold value, $T_{r}$

P_{de} = P_{r} {D = 1 | H_{1}} = P_{r} {E > T_{r} | H_{1}}

(3)

and

P_{fa} = P_{r} {D = 1 | H_{0}} = P_{r} {E > T_{r} | H_{0}}

(4)

Cognitive radio sensor network (CRSN) is an emerging technology promising a widespread deployment of the wireless sensing applications. In the CRSN, some or all nodes are equipped with radio and processing capabilities to perform opportunistic and dynamic utilization of the spectrum holes in the licensed frequencies. In a typical CRSN, ideal sensing nodes coexist with higher capability devices such as smart phones, laptops, and cordless phones. The CRSN consists of many sensor nodes and at least one sink node. Some or all of those nodes are equipped with the CR capabilities.

Resource allocation schemes for wireless networks can be classified into two main types: central-based, where the sink node is responsible for making the allocation decision, and distribute-based. In the latter type, each unit decides what resources it can utilize. The decision can be taken either with or without the cooperation of other units.^12–14

The emerging networks for monitoring and objects tracking seem to be heterogeneous, meaning that each network comprises more than one application. This heterogeneity necessitates the provisioning of multilevel QoS. In general, QoS assurance is correlated with the resource allocation scheme, which must be efficient enough to fulfill the QoS requirements. Although the resource allocation elements may vary from one scheme to another, the key elements are power, scheduling, and channel assignment. In wireless networks, particularly in cellular networks, there are two main schemes for centralized channel allocation: fixed channel allocation (FCA) and dynamic channel allocation (DCA).¹⁵ The allocation is static in FCA and cannot be changed, whereas channel reallocation is frequently occurred in DCA. Furthermore, most of the existing studies impose work only in homogeneous radio environments, meaning that exploited channels have similar radio conditions.

This study proposes an efficient resource allocation scheme to tackle twofold of CRSN heterogeneity: traffic and radio resource heterogeneity. The main contributions of this study are summarized as follows:

It presents a novel scheduling and radio allocation scheme that is treating high important data and reducing intra-network and inter-network interference in the same time of maintaining the conventional end-sensor characteristics.

It enables targeting heterogeneous radio environments by allowing utilization of more channels belonging to different radio environments.

The main motivation behind this study is to reinforce the feasibility of implementing CR technique in the emerging wireless monitoring and tracking applications. Most of the existing studies either employ improper architecture of the network, for example, the ad hoc-based topology, or require significant changes for the power supply and the computational capability of the end-sensor node.^16–18 Furthermore, these works do not efficiently address the heterogeneous nature of the new wireless networks. In contrast, the improved Pliable Cognitive Medium Access (iPCMAC) protocol proposed in this study maintains the conventional characteristics of end-sensor node, copes with the required QoS for emerging wireless sensor applications, and allows more efficient utilization of the radio spectrum.

The remainder of this article is organized as follows: Section “Related work” reviews related previous works. Section “System model” introduces the proposed scheme and describes the proposed improvements through the system model. Section “Mathematical model” provides the mathematical model. Section “Performance evaluation in heterogeneous radio environment” presents the results of the performance evaluation. Finally, section “Conclusion” concludes this study.

Related work

Existing resource allocation schemes designed for conventional WSNs are not suitable for CR-based WSNs. In general, the main concern of WSN protocols is to maintain low levels of power consumption throughout the duty cycle. WSN protocols are usually designed for homogeneous-type networks with only one application.¹⁹ On the other hand, most studies addressing resource allocation schemes in cognitive radio networks (CRNs) focus only on one aspect and pay less attention to others. These aspects include energy efficiency, latency decrease, interference avoidance, and channel utilization efficiency. Moreover, the proposed algorithms for CRN are also inappropriate for CRSN networks because they do not take into account the limited capabilities of the end-sensing nodes.

To increase the efficiency of power consumption in a CRSN, Han et al.²⁰ proposed a channel management scheme that also took the protection of the PU signal into account. Similarly, Li et al.²¹ exploited an R-coefficient to estimate the residual energy in the sensors and to predict the energy consumption, which helped in assigning channels. Among several studies aiming to maximize the throughput in ad hoc CRSNs via efficient resource allocation schemes, Piran et al.²² presented a networking paradigm for vehicular communication through TV white spaces in two mobile and stationary modes. In the same way, several studies have concentrated on QoS provisioning as the main factor when proposing their resource allocation schemes. To minimize the energy consumed in SS and provide a given throughput while satisfying reliability constraints, Ejaz et al.²³ examined the energy-throughput trade-off for cooperative SS. Deng et al.²⁴ proposed a framework to obtain periodic scheduling by adopting both non-linear and linear programming.

The study in Lin and Chen²⁵ developed an algorithm to enable the simultaneous transmissions in the available spectrum to decrease the delay and fulfill the QoS requirements. Likewise, Liang et al.¹³ introduced a framework for CRSN to support real time (RT) traffic using two types of channel switching: periodic and triggered switching. The goal was to decrease the average delay time for RT traffic. A previous study²⁶ improved and modeled the performance of the dynamic open spectrum sharing (DOSS) MAC protocol, which is carrier sensing multiple access (CSMA)-based and incorporated a multiple channel access in an ad hoc CRSN network.

A new CSMA-based protocol for cluster-based CR sensor networks is proposed in Al-Medhwahi et al.²⁷ Pliable cognitive medium access control protocol (PCMAC) maintains traditional WSN features such as the low complexity and the low power consumption at the same time it addresses a variety of node traffic criticality in homogeneous radio environments. The latter environment is assumed to have similar radio parameters, such as signal-to-noise ratio (SNR) and PU behavior, for all targeted radio channels. Despite its simplicity, PCMAC protocol is attributed with flexibility and it introduces a robust performance. On the other hand, restricting the targeted radio environment limits the efficiency of the protocol.

System model

This study investigates a CRSN consisting of multiple clusters $N = {N_{1}, N_{2}, \dots, N_{i}, \dots, N_{n}}$ , each of which contains one cluster head (CH) and tens of end-sensing nodes distributed over a square area of size $A$ , as shown in Figure 2. The available licensed channels are assumed to be $M$ , each of which is utilized by only one licensed PU. The idle time of those channels differs from channel to another. The saturated situation is assumed which means that each CH in the network always has data to transmit to the sink node. The embraced cluster-based topology is more appropriate than the ad hoc-based topology for heterogeneous networks and also mitigates the effect of the idle listening problem, thus decreasing power consumption.^27,28 Each cluster has a CH node that is responsible for collecting end-sensor data and is equipped with CR capabilities.

Figure 2.

Cluster-based CRSN.

The communication inside the cluster between the sensor nodes and the CH is performed via the unlicensed spectrum; however, the communication between the CH and the sink node is performed opportunistically via the licensed spectrum. The sink node, which acts as a network gateway, has $Q$ transceivers while each CH has only two: one to communicate with the end-sensing nodes using the industrial, scientific and medical (ISM) band and the other to communicate with the sink node using either a common control (CC) or licensed channel. The targeted licensed spectrum consists of $M$ radio channels that differ in their PU behaviors and SNRs.

Given the variety of radio characteristics, that is, the heterogeneous radio environment, channels are dynamically ranked and placed into two groups $M_{rt} = {M_{11}, M_{12}, \dots, M_{1 j}, \dots, M_{1 q}}$ and $M_{nrt} = {M_{21}, M_{22}, \dots, M_{2 j}, \dots, M_{2 s}}$ . Channels in the $M_{rt}$ group are dynamically allocated to the RT traffic whereas those in the other group are allocated to non-real time (NRT) traffic. Each group’s size is proportional to the number of requesting channel allocations of the CHs. Radio resource allocation is dynamic, for example, node $N_{i}$ is allocated channel $M_{1 j}$ in the current transmission cycle, but it might be allocated channel $M_{2 j}$ in the following transmission cycle.

CH nodes use the CC channel to send their transmission requests to the sink node using CSMA with collision avoidance (CSMA/CA) scheme. The proposed scheme employs an enhanced request to send/clear to send (RTS/CTS) handshake mechanism. The RTS message not only carries the submission request but also encloses the size of the requesting node buffer and type of the traffic, that is, either RT or NRT. Moreover, the CTS message encloses the number of the allocated channel. Once the requesting node receives the CTS message, it tunes its transceiver into the specified radio channel at the same time as the sink node.

The CH node starts sensing the channel in a continuous manner until it finds it free and thereafter commences data transmission. The transceiver of the sink node is “ON” for a period of time sufficient to transmit the volume of data declared in the RTS message. Consequently, the CH node can utilize the allocated channel throughout the predefined period of time regardless of the PU’s behavior since the latter has actually been taken into consideration when defining the time.

If the allocated time has elapsed and the node still has data, it has to evacuate the channel immediately and contend again in the CC channel. The former policy decreases the number of handoff processes since they usually followed by contention processes in the CC channel, which require enormous energy. The other case in which the transmitting node has to evacuate the allocated channel is when it receives no acknowledgment (ACK) message for a packet submitted three times, indicating a failed transmission attempt. Packets transmission is performed sequentially in a burst-based manner which means that the CH node will continue the transmission until its buffer becomes clear of the declared packets.

Note that the algorithm proposed in this study, iPCMAC, is an improved version of the PCMAC protocol proposed in the study.²⁷ Unlike its predecessor PCMAC, iPCMAC is designed to treat the heterogeneity of the radio environment and improve the efficiency with which the heterogeneity of the traffic is treated. The traffic verity is treated with additional techniques that give priority for the most critical traffic as will be illustrated in the following sections.

Traffic heterogeneity

Considering that the assumed network is heterogeneous, multiple levels of QoS requirements have to be provided. In this study, it is proposed that the network produces two types of traffic: RT and NRT. The former is intolerant to delay, meaning that the RT traffic has to be scheduled earlier than the NRT traffic. In CSMA-based algorithms, nodes desiring to submit their data have to contend to access the medium and to avoid collisions, they have to perform carrier sensing (CS) operations before proceeding with data submission. When the medium is sensed busy, the node has to backoff for a while before it senses the medium again. The backoff time is chosen by the random selection of a value drawn uniformly between zero and a minimum contention window size ( $C W_{\min}$ ).

In the proposed algorithm, there are two values for $C W_{\min}$ : one for the nodes having RT traffic, ${CW}_{\min}^{rt}$ , and one for those having NRT traffic, ${CW}_{\min}^{nrt}$ . The value of the latter is greater than that of the former, thus enabling the node with RT traffic to access the medium earlier and send its transmission request.

Radio environment heterogeneity

In CRSN’s homogeneous radio environment, the sink node considers only the current allocation status of the channel, whereas in heterogeneous radio environment, the sink node has to utilize the knowledge it has obtained in the previous transmissions for the upcoming channel allocation. Depending on the statistics of the current transmissions, the sink node updates its database of the targeted channels. Consequently, the channels are ranked according to one or more factors such as the spectrum hole’s average length, the bit error rate (BER), and the bandwidth.

A slight computational capabilities, that is, in memory and processor, must be added to the CH node architecture to address this issue. In our case, it is assumed that channels only vary in their average lengths of the spectrum holes and have similar values for other radio characteristics. Thus, the longer the length of the idle time, the higher the rank in the list of channels. This feature allows the utilization of a wider range of radio channels, even though they do not share similar radio characteristics.

The channels in the sink node’s database are placed into two groups according to their rank. Channels in the first group, that is, those with a longer mean idle time $τ_{off}^{rt}$ , will be allocated to the RT traffic, whereas channels in the second group will be allocated to the NRT traffic. The channel’s chosen operation in each group is fair and once the allocated node accomplishes its transmission, the sink node can reallocate the channel to another node. The sink node has to evaluate the previous transmission regarding the received throughput and the time it takes, enabling the sink node to update the rank of the utilized channel.

The algorithm’s simple pseudo code can be written as follows:

1: sort packets in the CH node’s buffer according to their priorities (two parts)
2: for each packet in the CH node’s buffer do
3: sense the CC channel
4: if

S_{rc} < S_{Tc}

//the received signal is less than a threshold value: the medium is free
5: send RTS message
6: else
7: backoff and repeat the sensing process
8: end if
9: electing an unallocated channel among

M

channels //by the sink node
10: receive CTS message
11: tune the transceiver to the elected frequency
12: start timer for losing the connection with the sink node
13: sense the data channel
14: if

S_{rd} < S_{Td}

//the power of the received signal is less than a threshold value
15: send the packet
16: ifreceive ACK message
17: send next packet
18: else
19: sense the channel
20: else
21: continue sensing process
22: end if
23: end for

Most of the statements in the algorithm are simple, and the running time of each included operation is either constant or linear. In the worst cases, for example, excessive sensing processes of the CC channel and data channels, numbers of operations are proportional to the size of inputs. The inputs for the sensing operations include density of RTS messages and the length of the idle time. Accordingly, the algorithm’s complexity that can be estimated according to the worst cases is linear.

The efficiency of the proposed scheme is proven by two common performance metrics, namely delay and throughput. However, several other metrics that are implied in the employed designing principles reinforce the efficiency of the scheme as illustrated in Table 1.

Table 1.

Additional performance metrics.

Performance metric	Description
Power consumption	Assigning high consumption CR’s processes to a limited number of nodes
Scalability	Adopting the cluster-based architecture
Reliability	Assuring the delivery for all data packets with unlimited number of transmission attempts

CR: cognitive radio.

Mathematical model

PU behavior

The PU behavior is modeled using a two-state Markov chain to capture the temporal dependency between two consecutive states.^29,30 The presence and absence of the PU signal in a channel are represented by the busy and idle states, respectively. The transitions are between adjacent timeslots in the same channel, and they are independent of other channels. Thus, for a channel $M_{i}$ , let $α_{i}$ denote the probability that the channel will be busy in the next timeslot, that is, the PU will be present given that it is idle in the current timeslot. In the same way, let $β_{i}$ denote the probability that the channel will be idle, that is, that the PU will be absent in the next timeslot given that the channel is busy in the current timeslot. Thus, the probability that channel $i$ is idle in timeslot $x$ can be obtained regarding the steady-state probabilities as shown in Figure 3

{P_{ii}}^{x} = \frac{β_{i}}{α_{i} + β_{i}}

(5)

Figure 3.

Two-state Markovian model for PU behavior.

The probability that the channel is busy in the same timeslot is estimated as

{P_{ib}}^{x} = \frac{α_{i}}{α_{i} + β_{i}}

(6)

Algorithm model

This study proposes an enhanced $M / G / 1$ priority queuing to model the average delay, where $M$ stands for Markovian, that is, the system has exponentially distributed intervals for the arrival packets. $G$ stands for the general distribution of the service time, and 1 represents the used server, the CC channel. In addition to the CC channel employed to serve the data packets in the proposed algorithm, licensed data channels also contribute in this mission. The enhanced non-preemptive $M / G / 1$ queue model has two virtual queues: high queue (HQ) and low queue (LQ). However, the HQ packets are prioritized, that is, scheduled first, they are not permitted to interrupt any current service for LQ packets. HQ and RT packets are referring to the same type; LQ and NRT packets are also of the same type of packets.

The end-to-end average delay time consists of two parts: the waiting time, $W$ , and the service time, $S$ , where the waiting time for a packet in the CH node consists of two parts, $W 1$ and $W 2$ . While $W 1$ represents the waiting time caused by the preceding packets in the queue, $W 2$ denotes the waiting time because of a current transmission from the same node, even though it belongs to the other queue, as illustrated in Figure 4. The service time is estimated as

S = T_{c} + T_{ss} + T_{tr}

(7)

where $T_{c}$ represents the contention time in the CC channel, $T_{ss}$ is the data channel’s observation time, and $T_{tr}$ is the packet’s transmission time

{\bar{T}}_{c} = DIFS + {\bar{T}}_{b} + T_{rts} + SIFS + T_{cts}

(8)

where DIFS and SIFS are the distributed and the short inter-frame spaces, respectively. While $T_{rts}$ represents the time at which an RTS message is sent, $T_{cts}$ denotes the time of the CTS message. The mean backoff time, ${\bar{T}}_{b}$ , is estimated as in Zhao et al.³¹

{\bar{T}}_{b} = C W_{\min} \frac{1 - p - 2^{N} p^{N + 1}}{2 - 4 p} - \frac{1}{2}

(9)

where $C W_{\min}$ is the minimum contention window size, $p$ represents the conditional collision probability, and $N$ is the number of contender nodes. In the steady state

p = 1 - (1 - τ)^{N - 1}

(10)

where $τ$ is the probability that the node transmits in a random timeslot. Similar to Bianchi³² and for the simplification

τ = \frac{2}{C W_{\min} + 1}

(11)

Figure 4.

Virtual queues for packets.

The mean observation time can be expressed as

{\bar{T}}_{ss} = \frac{τ_{on}}{2} P_{on}

(12)

$P_{on}$ is the occupancy probability of a licensed channel

P_{on} = \frac{τ_{on}}{τ_{on} + τ_{off}}

(13)

While $τ_{on}$ represents the channel’s busy time, $τ_{off}$ expresses the idle time. In the proposed model, an accurate SS is assumed; hence, the probability of detection $P_{de} \approx 1$ and the probability of false alarm $P_{fa} \approx 0$ . The estimated time for a packet transmission is calculated as

T_{tr} = CCA + \frac{L}{D_{lc}} + SIFS + T_{ack}

(14)

where $CCA$ denotes the clear channel assessment mid time, $L$ is the packet size, and $D_{lc}$ is the data rate of the data channels. According to equation (7), the mean service time is

\bar{S} = {\bar{T}}_{c} + {\bar{T}}_{ss} + T_{tr}

(15)

The arrival of the packets in each virtual queue is assumed to be as a Poisson stream. $λ_{HQ}$ and $λ_{LQ}$ represent the arrival rates of the HQ and LQ, respectively. The occupation rate of the CC channel is the sum of the occupation rates for both the queues is given as

ρ_{C} = ρ_{HC} + ρ_{LC}

(16)

where

ρ_{HC} = λ_{HQ} {\bar{T}}_{c} and ρ_{LC} = λ_{LQ} {\bar{T}}_{c}

To maintain the system stability, the occupation rate for the CC channel must be less than one. Similarly, the occupation rate of the data channels after considering the number of utilized channels can be estimated as

ρ_{HD} = N λ_{HQ} \frac{{\bar{T}}_{ss} + T_{tr}}{C}

(17)

ρ_{LD} = N λ_{LQ} \frac{{\bar{T}}_{ss} + T_{tr}}{C}

(18)

It is also assumed that the duty cycle time is 1 s. The mean waiting time for a packet from the HQ is estimated as

{\bar{W}}_{HQ} = {\bar{L}}_{HQ} \bar{S} + ρ \bar{R}

(19)

where $\bar{R}$ is the mean residual service time of a current served packet and $ρ$ represents the general occupation rate of the entire traffic in the CC and the data channels. The model is non-preemptive. Thus

\bar{R} = \frac{{\bar{S}}^{2}}{2 \bar{S}} = \frac{σ_{S}^{2} + {\bar{S}}^{2}}{2 \bar{S}}

(20)

Using Little’s law and when ${\bar{L}}_{HQ}$ represents the mean length of the HQ

{\bar{L}}_{HQ} = λ_{HQ} {\bar{W}}_{HQ}

(21)

This former equation can be rewritten as

{\bar{W}}_{HQ} (1 - ρ_{H}) = ρ \bar{R}

(22)

where, $ρ_{H}$ represents the sum of the HQ packets’ occupation rate in the CC channel and the data channels. Thus, the mean waiting time for a packet from the HQ is estimated as

{\bar{W}}_{HQ} = \frac{ρ \bar{R}}{1 - ρ_{H}}

(23)

Consequently, the mean throughput time, that is, the average delay time, can be estimated as

T_{HQ} = {\bar{W}}_{HQ} + \bar{S}

(24)

Considering that the new packet may arrive at any time in an ongoing service with a mean service time, $\bar{S}$ , the mean residual service time can be estimated as

\bar{R} = \frac{1}{2} (c_{s}^{2} + 1) \bar{S}

(25)

where $c_{s}$ is the coefficient of the service variation time and it is equal to $σ_{s} / S$ . Note that in case of an exponential service time, $c_{s}^{2}$ is equal to 1 and in case of a deterministic service time, it equals 0. For simplicity, $c_{s}^{2} = 0.5$ . Thus

\bar{R} \approx \frac{3}{4} \bar{S}

(26)

The mean delay time for an NRT packet, that is, belongs to the LQ, can be estimated as

\frac{{\bar{W}}_{HQ}}{{\bar{W}}_{LQ}} = 1 - ρ

(27)

{\bar{W}}_{LQ} = \frac{{\bar{W}}_{HQ}}{1 - ρ}

(28)

Let $W_{T}$ represent the wasted time between sequent transmissions that is equal to the sum of the gaps between the packets transmissions and the related contention periods in addition to the contention periods in which the channel is not chosen for submission. A uniform distribution is used to provide the fairness property for the channels. Thus

W_{T} = P_{i} \sum_{j = 1}^{Max F_{n}} | T_{cj} - T_{trj} | + (1 - P_{i}) ⌊ \frac{τ_{off}}{{\bar{T}}_{c}} ⌋ {\bar{T}}_{c}

(29)

where $P_{i}$ is the probability that the sink node will choose the channel for the next submission and $Max F_{n}$ is the maximum number of transmissions that can be achieved in the $τ_{off}$ course of time after deducting the time in which the channel was not chosen for submission, $T_{cm}$

Max F_{n} = \frac{τ_{off} - T_{cm}}{Max (T_{c}, T_{tr})}

(30)

W_{T} \approx P_{i} Max F_{n} | T_{ci} - T_{tri} | + (1 - P_{i}) τ_{off}

(31)

Each channel has the same probability of selection among the $M$ channels. Thus

P_{i} = \frac{1}{M}

(32)

The available transmission time equals ( $τ_{off} - W_{T}$ ) and the achievable number of transmissions in frames are

N T_{tr} = ⌊ \frac{τ_{off} - W_{T}}{T_{tr}} ⌋

(33)

The net transmission time is estimated as

N T_{tr} (T_{tr} - SIFS - T_{ack}) = N T_{tr} L

(34)

Considering that the probability of a node to submit in a specific idle time is $P_{s}$ and the CC channel’s non-blocking probability is $P_{b}$ that can be expressed as $P_{s} = \overset{- N / M}{\exp}$ and $P_{b} = 1 - {\bar{T}}_{c} / \bar{S}$ , respectively, the achievable throughput is estimated as

D_{off} = (1 - P_{s} P_{b}) N T_{tr} L D_{lc}

(35)

Regarding the case of the heterogeneous radio environment described in section “Related work,” there are two values for the $C W_{\min}$ which give priority for CH nodes that have RT data to access the CC channel and send their RTS messages: ${CW}_{\min}^{rt}$ and ${CW}_{\min}^{nrt}$ . Consequently, there are two values for the mean backoff time, $T_{b}$ obtained from equation (9). For channels allocation, channels are divided into two groups: G1 and G2 and the size of each group is proportional to the portion of the related requests as

\frac{M_{rt}}{M} = \frac{N_{rt}}{N}

(36)

where $M = M_{rt} + M_{nrt}$ and $N = N_{rt} + N_{nrt}$ . $M_{rt}$ denotes the number of the channels that allocated to the submissions of the RT data $N_{rt}$ , that is, G1, while $M_{nrt}$ denotes the size of the G2 group that is allocated to the NRT data, $N_{nrt}$ . Consequently

M_{rt} = M (N_{rt} / N_{rt} + N_{nrt})

(37)

Within each group, channel selection probability is fair for all channels

P_{i}^{rt} = \frac{1}{M_{rt}} and P_{i}^{nrt} = \frac{1}{M_{nrt}}

Imperfect SS

Perfect sensing was originally assumed; however, imperfect sensing can be included in the model by considering values of $P_{fa}$ and $P_{de}$ . In consequence, the new value of the probability that a channel will be sensed idle will be given by $P_{off} (1 - P_{fa})$ , whereas the old value, that is, for the perfect sensing $P_{off}$ , can be obtained by equation (2). In several applications, the acceptable values for $P_{fa}$ and $P_{de}$ are 0.1 and 0.9, respectively.

Performance evaluation in heterogeneous radio environment

To evaluate the performance of the iPCMAC, a simulation is performed using MATLAB. The parameters are similar to those defined in IEEE 802.11 MAC and are listed in Table 2. The duty cycle is 1 s.

Table 2.

Simulation parameters.

Parameter	Value
CW minimum size for RT traffic ( ${CW}_{\min}^{rt}$ )	8
CW minimum size for NRT traffic ( ${CW}_{\min}^{nrt}$ )	32
Common channel data rate (Mbps)	1
Data channel data rate (Mbps)	1
Slot time in CC channel (ms)	0.02
$CCA$ (ms)	0.02
$DIFS$ (ms)	0.05
$SIFS$ (ms)	0.01
$T_{rts}$ (ms)	0.352
$T_{cts}$ (ms)	0.304
$T_{ack}$ (ms)	0.304

CW: contention window; RT: real time; NRT: non-real time; CC: common control.

Figure 5 describes the relationship between nodes density increase and the induced delay for both RT and NRT packets. This figure also illustrates a performance comparison between the previous protocol designed for the homogeneous radio environment, PCMAC, and the improved version of the protocol designed for the heterogeneous radio environment, iPCMAC. The obtained results indicate that iPCMAC significantly outperforms PCMAC. The average delay of iPCMAC packets is less than that of PCMAC packets, and the delay values start with $140$ and $190 ms$ for the iPCMAC and PCMAC packets, respectively. Furthermore, the trend toward the exponential behavior of iPCMAC packets begins at a higher nodes density than that of PCMAC packets. The system with the iPCMAC protocol maintains its stability as the nodes density increases, whereas the system with the PCMAC protocol loses stability at medium density, $N = 12$ .

Figure 5.

iPCMAC delay and a comparison with PCMAC; $τ_{off} = 0.5; M = 10$ .

Similar to the performance in the PCMAC protocol, the higher priority RT packets exhibit less sensitivity against the exponential behavior than NRT packets in iPCMAC protocol. This is attributed to the impact of assigning a shorter contention window size for the RT traffic, thus giving it the privilege of access to the medium earlier than the NRT traffic.

The relationship between nodes density increase and the induced delay is illustrated in Figure 6. The scenario includes multiple busy times for the G1 group of channels while G2’s busy time is fixed at $0.8 s$ . When the busy time is doubled from $0.2$ to $0.4 s$ , the induced delay increases from $32$ to be $93 ms$ and becomes $194 ms$ when the busy time is $0.6 s$ . The average delay increases as the busy time increases, and the trend toward the exponential behavior in long busy times predates that in short busy times. The red curve describes the trend of the average delay for NRT packets when $τ_{on}^{rt} = 0.6 s$ . NRT packets suffer from a longer delay time that begins with $0.33 s$ when $N = 1$ and reaches $0.35 s$ when $N = 14$ at the beginning of its exponential behavior.

Figure 6.

Relationship between N and the average delay in case of various values of $τ_{on}^{rt}; M = 10$ .

Figure 7 shows a comparison between the performance of the proposed iPCMAC, PCMAC,²⁷ and DOSS²⁶ protocols in terms of their induced delay. Although the DOSS MAC presents a shorter delay for packets during shorter lengths of busy time, it expands exponentially at longer busy times starting at $0.4 s$ . The PCMAC presents a more stable performance for both the short and long periods of channel’s idle time, whereas iPCMAC presents the best scheduling performance for RT packets with linear behavior. While the system with DOSS protocol losses its stability when the channel’s busy time exceeds $0.7 s$ , iPCMAC protocol prevents the maximum induced delay from exceeding $230 ms$ when the channel’s busy time is 0.9 s.

Figure 7.

Average delay comparison between iPCMAC, PCMAC, and DOSS protocols; M = 10.

The enchantments ensured by the iPCMAC protocol for the system’s throughput are illustrated in Figure 8. The assumed average busy times are $0.3$ and $0.7 s$ for the G1 and G2 channels, respectively. When seven data channels are available, the aggregated throughput of G1 channels is greater than that of G2. The throughput increases as the nodes density increases until $N$ equals nine nodes when the G2 throughput starts to decrease slightly while the throughput increasing curve of G1 channels starts to slow. This behavior is correlated with the impact of the contention in the CC channel, which is tougher for nodes allocated channels from G2. The contention time of nodes allocated these channels becomes longer than that of nodes allocated channels from G1. Accordingly, the proposed algorithm alleviates the contention impact for both types of nodes, and the node with RT traffic has better chances to transmit more data.

Figure 8.

Aggregated throughput relationship with nodes density; M = 7.

The outperformance of iPCMAC protocol is also demonstrated in Figure 9. This figure shows a comparison between iPCMAC and PCMAC protocols in terms of the achievable throughput and the efficiency of channel utilization. The latter measures the ratio between the achieved throughput and the maximum throughput. The utilization ratio of iPCMAC represents the average value of G1 and G2 channels that their idle times are $0.4$ and $0.6 s$ , respectively, whereas the idle time of PCMAC channels is $0.5 s$ . iPCMAC protocol achieves higher ratios than PCMAC protocol starting from $71.44 %$ when the available channels are two. G1 Channels have higher utilization ratios because of the short contention time, $T_{c}$ , of RT traffic, regardless of the fact that RT and NRT traffic have the same size. Channel utilization ratios decline as the number of channel increases since the throughput is divided between the available channels according to the aforementioned policies. These results justify the motivation of our proposed work where adopting an efficient resource allocation scheme can significantly enhance the system performance.

Figure 9.

Relationship between channel utilization ratio and number of channels; N = 10.

Conclusion

Given that data transmission in CR-based networks is performed opportunistically, adopting an efficient resource allocation scheme becomes an essential task, particularly in cases requiring multilevel of QoS. The proposed scheme in this study tackles scheduling and radio channel allocation in a cluster-based CRSN. Our objective is to improve resource allocation for heterogeneous traffic in heterogeneous radio environments and to enable the best resources for the most critical data. RT traffic is intolerant of latency and deserves priority in transmission. Moreover, in emergency cases, transmissions become in burst-nature and require longer transmission times. This paradigm is treated efficiently by the proposed resource allocation scheme and the iPCMAC algorithm. Furthermore, the proposed algorithm offers a chance of utilizing a wider radio spectrum than one using homogeneous environment. Simulation results demonstrate the effectiveness of the proposed protocol, which has been proven to outperform existing ones.

Footnotes

Handling Editor: Daming Zhou

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Mohammed Al-Medhwahi

References

Becker

Services in wireless sensor networks: modelling and optimisation for the efficient discovery of services. Wiesbaden: Springer, 2014.

Yang

Wireless sensor networks principles, design and applications. London: Springer, 2014.

Khattab

Perkins

Bayoumi

Cognitive radio networks: from theory to, practice. New York: Springer, 2012.

FCC. Report of the spectrum efficiency working group. FCC Spectrum Policy Task Force, Washington, DC, 15 November 2002.

Zahmat

Hussain Fernando

Grami

. Cognitive wireless sensor networks: emerging topics and recent challenges. In: IEEE Toronto international conference science and technology for humanity (TIC-STH), Toronto, ON, Canada, 26–27 September 2009. New York: IEEE.

Ghosh

. Cognitive radios in television white spaces. In: Gerbarg

(ed.) Television goes digital. New York: Springer, 2009, pp.187–207.

Kaabouch

. Handbook of research on software-defined and cognitive radio technologies for dynamic spectrum management. Hershey, PA: IGI Global, 2014.

Holland

Bogucka

Medeisis

. Opportunistic spectrum sharing and white space access: the practical reality. Hoboken, NJ: John Wiley & Sons, 2015.

Cabric

Mishra

Brodersen

. Implementation issues in spectrum sensing for cognitive radios. In: Conference record of the thirty-eighth Asilomar conference on Signals, systems and computers, Pacific Grove, CA, 7–10 November 2004. New York: IEEE.

10.

Lee

Akyildiz

. Optimal spectrum sensing framework for cognitive radio networks. IEEE T Wirel Commun 2008; 7: 3845–3857.

11.

Atapattu

Tellambura

Jiang

. Energy detection for spectrum sensing in cognitive radio. New York: Springer, 2014.

12.

Ahmad

Rehmani

et al . A survey on radio resource allocation in cognitive radio sensor networks. IEEE Commun Surv Tut 2015; 17: 888–917.

13.

Liang

Feng

Zhao

et al . Delay performance analysis for supporting real-time traffic in a cognitive radio sensor network. IEEE T Wirel Commun 2011; 10: 325–335.

14.

Byun

Balasingham

Liang

. Dynamic spectrum allocation in wireless cognitive sensor networks: improving fairness and energy efficiency. In: IEEE 68th vehicular technology conference, Calgary, BC, Canada, 21–24 September 2008. New York: IEEE.

15.

Singh

Srinivas

Das

. Dynamic channel assignment in TDMA mobile cellular systems using hybrid discrete binary differential evolution algorithm. In: 7th Iberian conference on information systems and technologies (CISTI), Madrid, 20–23 June 2012. New York: IEEE.

16.

Huang

PXL

Soltani

Mutka

et al . The evolution of MAC protocols in wireless sensor networks: a survey. IEEE Commun Surv Tut 2013; 15: 101–120.

17.

Zhang

. Performance analysis of cluster-based and tree-based routing protocols for wireless sensor networks. In: International conference on communications and mobile computing, Shenzhen, China, 12–14 April 2010. New York: IEEE.

18.

Sinha

Lobiyal

. Performance evaluation of data aggregation for cluster-based wireless sensor network. Hum Centric Compu Inform Sci 2013; 3: 13.

19.

Collotta

Gentile

Pau

et al . Flexible IEEE 802.15.4 deadline-aware scheduling for DPCSs using priority-based CSMA-CA. Comput Ind 2014; 65: 1181–1192.

20.

Han

Jeon

Jeong

. Energy-efficient channel management scheme for cognitive radio sensor networks. IEEE T Veh Technol 2011; 60: 1905–1910.

21.

Wang

McNair

et al . Residual energy aware channel assignment in cognitive radio sensor networks. In: IEEE wireless communications and networking conference (WCNC), Cancun, Mexico, 28–31 March 2011. New York: IEEE.

22.

Piran

Cho

YYJ

Ali

et al . Cognitive radio-based vehicular ad hoc and sensor networks. Int J Distrib Sens N. Epub ahead of print 5 August 2014. DOI: 10.1155/2014/154193.

23.

Ejaz

Shah

Kim

. Energy and throughput efficient cooperative spectrum sensing in cognitive radio sensor networks. T Emerg Telecommun T 2015; 26: 1019–1030.

24.

Deng

Chen

et al . Energy-efficient spectrum sensing by optimal periodic scheduling in cognitive radio networks. IET Commun 2012; 6: 676–684.

25.

Lin

Chen

. Improving spectrum efficiency via in-network computations in cognitive radio sensor networks. IEEE T Wirel Commun 2014; 13: 1222–1234.

26.

Shah

Akan

. Performance analysis of CSMA-based opportunistic medium access protocol in cognitive radio sensor networks. Ad Hoc Netw 2014; 15: 4–13.

27.

Al-Medhwahi

Hashim

Ali

et al . Pliable cognitive MAC for heterogeneous adaptive cognitive radio sensor networks. PLoS ONE 2016; 11: e0156880.

28.

Joshi

Kim

. A survey on node clustering in cognitive radio wireless sensor networks. Sensors 2016; 16: 1465.

29.

Jayaweera

Bkassiny

et al . Optimal myopic sensing and dynamic spectrum access in cognitive radio networks with low-complexity implementations. IEEE T Wirel Commun 2012; 11: 2412–2423.

30.

Adas

. Traffic models in broadband networks. IEEE Commun Mag 1997; 35: 82–89.

31.