Prediction-based channel assignment for minimizing channel switching in mobile WBANs

Abstract

As the world’s population rises, the healthcare system experiences significant changes. Wireless body area network (WBAN) is an emerging technology that has considerable impact on medical and non-medical applications. However, two crucial challenges in WBANs are interference minimization and channel assignment. High interference may increase collision probability, transmission delay, and energy consumption. Multichannel schemes are proposed to reduce the data transmission latency and improve the system throughput by allowing simultaneous transmission of sensors in coexisting WBANs. When WBAN users move, they need to switch the channels frequently to avoid potential channel conflicts and to maintain the Quality of Service (QoS). However, frequent switching may raise energy consumption and aggravate delay. Existing multichannel assignment schemes failed to perform well in highly dynamic and densely deployed WBANs environments. In contrast to existing studies, this paper proposes a Prediction-based Channel Assignment (PCA) algorithm that selects the channels for WBANs to remain valid for future time instances and thus minimizes the delay and number of channel switches for dynamic and coexisting WBANs. When a WBAN needs to switch a channel, the proposed method predicts the future neighbors of that WBAN based on its history. It explores the channel information of present and future neighbors to select a suitable channel with higher resilience in a dynamic environment. Thus, our algorithm minimizes channel interference by avoiding unnecessary channel switching. We have used machine learning algorithms to predict the future neighbors of a WBAN. Experiment results show that the proposed algorithm performs better than an existing algorithm and random channel assignment in delay and throughput.

Keywords

WBAN co-channel interference multichannel future neighbor prediction channel switch

1. Introduction

Over the past decade, the rapid development of wireless communication and integrated circuit design has changed human life and engineering. Many wireless technologies have been developed to enhance wireless communication quality. The wireless sensor network is considered one of the enabling technologies in many applications, including agriculture, monitoring, surveillance, smart cities, smart grid, etc. A wireless body area network (WBAN) is a special kind of sensor network, mainly used in the healthcare domain [1,17,26,29]. The WBAN primarily consists of tiny, intelligent, resource constraint, and low-power biosensors that can be placed or implanted in the human body. These nodes monitor various physiological parameters of the body continuously or periodically and then send them to the gateway or coordinator (for example, Smartphones, PDA). The communication within a WBAN is known as intra-WBAN communication. In another kind of communication, i.e., inter-WBAN communication, the coordinator transmits the collected data to the base station or access point for further processing or analysis [36]. WBANs are used in remote and health monitoring and in many non-medical applications like sports, entertainment, movement detection, etc. Medical applications require high throughput, low delay, and high reliability. However the network performance degrades when more WBANs come closer and start transmitting. Interference is an inevitable problem among WBANs when they come within a short range. For example, when people with wearable body sensors gather in an area such as a room or a hospital, the coexisting WBANs start interfering. This kind of interference is known as inter-WBAN interference, and as a result, the signal strength gets degraded [18]. High interference decreases the network throughput and raises the delay in data transmission, which may further cause a potential risk to patients. Moreover, transmission failure due to interference can lead to power depletion of body sensors by increasing traffic load, re-transmission, and congestion. Hence, minimizing interference is one of the crucial challenges in WBAN design.

Several schemes like TDMA-based slot allocation [27], beacon shifting [15], MAC design [12] have been proposed to mitigate interference among WBANs. But they mainly considered a single channel, resulting in larger delay, energy depletion, re-transmission, and collisions when the network becomes dense. Furthermore, single channel-based schemes are ineffective in body area networks when WBANs are subjected to mobility. Multiple channel allocation schemes provide efficient solutions for reducing interference, as discussed in [5,14,31,39]. A two-tier multi-channel-based MAC protocol (2TM-MAC) was proposed in [39] to alleviate intra and inter-WBAN interference. In this work, the access point is responsible for allocating interference-free multiple channels to coexisting WBANs. In the second step, multiple channels are assigned to body sensors to reduce intra WBAN interference. In 2L-MAC proposed in [5], multiple coexisting WBANs contended to access the channel, leading to higher collision and latency. In another work, the authors [14] proposed an adaptive channel estimation and selection-based scheme (ACESS) which maintains a history table for free channels. However, it consumes extra energy to update the table in real-time. Although the existing channel assignment techniques mitigated the co-channel interference, they assumed that the WBANs were static. However, in a dynamic environment, multiple channel assignment raises several challenges. Firstly, when WBANs move, they must switch to different channels frequently to reduce channel conflicts. Hence, movement prediction and channel selection are two critical issues that should address to reduce channel switching costs. Secondly, channel switching needs calibrating and synchronization, which increases switching delay and expense. Finally, frequent channel switching increases power consumption, which indirectly minimizes the lifetime of a sensor and the network.

Motivated by the challenges mentioned above, this paper proposes a Prediction-based Channel Assignment (PCA) algorithm that minimizes the number of channel switching in dynamic and coexisting WBANs scenarios. At a glance, the PCA algorithm assigns the channels to the WBAN coordinators during switching, so the allocated channels should be valid in future instances. PCA is a centralized method consisting of three phases: (a) initial channel assignment, (b) predicting future neighbors, and (c) channel switching and allocation. Since WBAN users move, the received signal strength (RSS) at a WBAN fluctuates with time. Our PCA predicts the future neighborhood of a WBAN by analyzing the RSS from other WBANs based on the history. It uses two machine learning algorithms, i.e., auto regressive model (AR) and moving average (MA) model, to predict the future neighbors. During channel assignment, PCA considers channel information of current and future neighborhood and select a channel with higher resilience in a dynamic environment. We evaluate the performance of the proposed algorithm by simulations under semi-dynamic and dynamic scenarios. Experimental results show that PCA performs better than random channel assignment and 2TM-MAC [39] against different performance measures.

The main contributions of our work are listed below.

We formally define our channel assignment algorithm for coexisting WBANs. The objective is to minimize the delay in WBANs data transmission.

Mobility of WBANs are considered in studying channel assignment.

We propose prediction-based channel allocation that reduces the channel switching of WBANs by predicting future neighbors based on history. Hence, the proposed algorithm is more resilience to dynamic interference.

AP allocates the channels to the WBAN coordinators, thus reducing the complexity and overhead at each WBAN.

Extensive experiments have been conducted to validate the performance of the proposed algorithm. Moreover, we compare our method with some existing channel assignment algorithms on total network delay, throughput, and successful assignment per channel in semi-dynamic and dynamic scenarios.

In the rest of the paper, Section 2 reviews some existing channel assignment schemes for WBANs and Section 3 describes various models and problem definition. In Section 4, we describe some primitives. Section 5 presents the proposed channel assignment algorithm in details. In Section 6, the performance of the proposed algorithm is evaluated and results are analyzed and discussed. Finally, in Section 7 we conclude the paper.

2. Related works

Co-located WBANs may cause many interference-related problems. Several research works have been done in the past years to avoid interference in WBANs [18]. This section discusses several single-channel-based interference minimization techniques and then explores multi-channel-based channel allocation techniques.

2.1. Single channel assignment techniques

Single channel based interference mitigation schemes have been widely studied in the literature. We can categorize them as power control schemes, medium access schemes, transmission scheduling, etc. For example, in [41], a Bayesian game theory-based power control technique was developed to avoid inter-BAN interference. In [12], the authors designed an adaptive CSMA/CA-based MAC protocol by adjusting the length of each polling period/MAC frame based on its perceived interference level. The transmission medium was accessed on a timely basis. In some works, the authors have addressed transmission scheduling of the sensors to avoid medium access collision. AIM [27] considered node level interference and assigned orthogonal time slots to interfered sensors. Similarly, ITLS [19] allocated non-interfering time slots to the coexisting sensors. Nevertheless, non-interfering nodes of the conflicting WBANs can share the same time slots. IPC [13] allowed simultaneous transmissions from sensor nodes of non-interfering WBANs. It considered traffic priority for the sensors during scheduling. CWS [32] is a clique-based scheme where sensors of different WBANs are grouped to reduce node level interference. After that, groups are mapped to other time slots by using a random coloring algorithm.

Authors in [38] analyzed the interference of coexisting WBANs by considering WBAN mobility and traffic priority. They adopted the asynchronous IEEE 802.15.6-based Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) protocol to access the channels by taking traffic priority and WBAN severity. However, this protocol worked on a sparse network with limited coexistence. To overcome the problem, the authors in [40] developed another CSMA/CA-based IEEE 802.15.6 Medium Access Control (MAC) mechanism, which is more scalable and efficient for dealing with multiple densely deployed mobile WBANs. Moreover, by assigning efficient channels, it improved the quality of service (QoS), like delay and throughput. Another QoS [6] aware channel allocation protocol was developed for mobile WBANs by taking into account both user priority and bandwidth requirements. Nevertheless, it reduces delay and improves throughput; the use of mobility models to capture mobility is unspecified. Considering WBAN mobility, the author proposed nest-based WBAN scheduling (NBWS) algorithm [35] to cluster the sensors of similar types in a single or multiple WBANs into different groups to prevent interference.

2.2. Multi-channel assignment techniques

In a dynamic environment, single-channel-based schemes are ineffective as the WBANs expose to mobility. Moreover, this may lead to increased latency, higher energy consumption, and sensor collision. Multi-channel allocation algorithms have been proposed to overcome the said problems. Multi-channel algorithms deal with either sensor channel assignment, coordinator assignment, or both. In this subsection, we highlight some existing multi-channel algorithms.

Some algorithms use hybrid solutions to minimize interference among coexisting WBANs. In [21], the authors developed a hybrid multi-channel media access protocol, HM-MAC, to alleviate interference between WBANs. In HM-MAC, multiple communications are slotted simultaneously on different channels, which may decrease collision and maximize throughput. In addition, CSMA/CA and TDMA are used in the super frame to transmit data packets. However, HM-MAC reduces collision; it increases control channel overhead in the network for maintaining synchronization with the hubs. HM-MAC may lead to higher energy consumption and more significant delay. In another work, the authors [15] introduced TDMA-based beacon interval shifting to minimize the coexistence in heterogeneous networks. The protocol senses the transmission medium before beacon transmission, improving network performance. But beacons from different WBANs may collide when WBANs are unaware of the shifting patterns of others. Authors [16] presented an adaptive and energy-efficient multi-channel MAC protocol called isMAC based on IEEE 802.15.4. The proposed MAC uses a collision prevention technique to reduce collision. In addition, they employed transmission power adjustment to minimize energy consumption. However, isMAC reduces latency, the priority of the WBAN was not taken into account in channel allocation, which causes starvation of low-priority nodes.

Distributed algorithms have gained popularity in the context of channel assignment in wireless communications. Random Incomplete Coloring algorithm was proposed in [7] to evade interference among WBANs. An unsupervised coloring algorithm for spectrum allocation for multiple WBANs [25] groups the nodes based on the distance. Then the channel group is assigned to each cell by coloring. The above-cited protocols mainly focused on interference avoidance, but they have yet to consider the priority of the WBANs while giving channels. Authors [33] considered the node priority and proposed incomplete coloring for assigning channels to the WBANs. Several protocols have applied distributed game theory for interference mitigation in WBANs. In [4], the authors proposed a proactive power update algorithm and two adaptive pricing mechanisms to increase channel capacity and reduce interference. A similar kind of work is found in [11] where channel selection of the coexisting WBANs is done as a potential game.

Several works focus on multi-channel and time slot adjustment to reduce co-channel interference among WBANs. In [3], the authors proposed a spectrum allocation technique called DAIL to mitigate co-channel interference among non-cooperative high density coexisting WBANs. DAIL employed Latin rectangles for channel and time-slot allocation to sensors, enabling automatic medium access for each WBAN. The main drawback of DAIL is that it incurs transmission overhead due to frequent channel hopping. Consequently, the energy consumption of WBANs increases even when some of their corresponding sensors do not experience any collision. To overcome it, the authors proposed another predictable channel hopping scheme called CHIM using Latin rectangles in [3]. Although CHIM has less overhead, its performance in dense networks still needs to be determined. In [8], a channel hopping scheme with a backup mechanism was proposed. An exchange-free resource scheduling (EFRS) was developed in [9] to reduce the channel hopping overhead. EFRS allocates the channels and time slots dynamically based on traffic demand. The disadvantage of EFRS is when interfering sensors are assigned the same channel and slot, the packets are retransmitted. Hence, the crash may lead higher latency and energy wastage.

Several works focus on prediction-based multi-channel allocation techniques to avoid interference in WBANs. For example, DCSSS [31] is a multi-channel protocol based on IEEE 802.15.6 to allocate channels to WBANs based on channel state prediction. Distributed prediction-based game theory techniques for channel management were explored in [37] and [24]. Authors in [5,39] have addressed two-layer MAC protocol to alleviate interference. Multiple WBANs in 2L-MAC [5] contend to access the same channel, leading to high collision and starvation. Two Tier Multi-channel MAC Protocol (2TM-MAC) [39] has been designed to minimize intra and inter-WBAN interference. The above protocol assigns multiple interference-free channels to the coexisting WBAN coordinators.

The previous studies cited above mainly focus on co-channel interference minimization. These approaches improve the network performance where the number of coexisting WBANs is small. However, they do not suit when WBANs are highly mobile and densely deployed because of the frequent channel switching. In such situations, predicting future locations of WBANs for channel assignment becomes vital for minimizing the number of channel switching. In our work, we have addressed these issues and proposed a prediction-based channel assignment algorithm that reduces the number of channel switches in dynamic and coexisting WBANs environments.

3. Models and problem definition

Fig. 1.

System overview.

3.1. System model

We consider a scenario where a group of people wearing body sensors moves in a hospital room S. An access point (AP) or base station (BS) is also located in S. We can model such a scenario as a network with N coexisting WBANs, denoted as $w_{i}$ , $1 ⩽ i ⩽ N$ . Each WBAN $w_{i}$ is composed of one coordinator (c) and up to K number of sensor nodes. The sensors can generate different user priority levels of traffic and send the measured data to the coordinator using CSMA/CA, in which, high-user priority sensors are scheduled to deliver their data packets first, followed by low-priority sensors. Further, the coordinator transmits data, collected from the different sensors to the AP for further analysis. We consider that the WBANs are working on M-number of channels and the access point manages the channels for the WBANs in S. Let $A = {a_{1}, a_{2}, a_{3}, \dots, a_{M}}$ be a set of channels where $a_{m}$ denotes m-th channel in A being such that $a_{m} \neq a_{n}$ , $1 ⩽ m$ , $n ⩽ M$ . However, in our work, at most one channel can be assigned to a WBAN at a time. Moreover, for each WBAN, we define the transmission range of the coordinator by $R_{c}$ , the transmission power of the coordinator by $P_{t, c}$ , and the transmission power of the sensor node by $P_{t, s}$ . We consider that the WBANs are mobile, and their mobility follows the popular Random Waypoint (RWP) model [13].

For the sake of simplicity, we have made the following assumptions in our work:

The transmission range and transmission power of all WBANs are the same.

The transmission power of all the sensors is equal.

Each WBAN follows star topology to send data to the coordinator during intra-WBAN communication.

AP is responsible for selecting and managing the channels to the WBANs.

Within each WBAN, the sensors adopt TDMA to transmit the measured data to the coordinator to avoid intra-WBAN interference. However, inter-WBAN interference can still occur between the neighboring WBANs due to the broadcasting nature of the wireless medium. Let

N_{i}^{t}

be the set of neighbors of WBAN

w_{i}

during time t. Two WBANs

w_{i}

and

w_{j}

are said to be neighbors of each other when their Euclidean distance

d_{i, j}

satisfies

d_{i, j} < R_{c}

. In general, the distance between WBANs should be large enough not to cause inter-WBAN interference. However, due to WBAN mobility, the distance between WBANs may change, and their communication range may overlap. The WBANs must be assigned distinct channels to avoid inter-WBAN interference; otherwise, this could result in a significant degradation of network performance. The system overview is illustrated in Fig. 1.

3.2. Communication model

We adopt the free space path loss model (FSPL) [20] to calculate the receiving power at a WBAN based on the sender and receiver distance. The received power $P_{r} (i, j, t)$ at the coordinator of $w_{i}$ during time t is measured as: $\begin{matrix} (1) & P_{r} (i, j, t) = P_{t, c} \times G_{t x} \times G_{r x} \times {(\frac{λ}{4 π d_{i, j} (t)})}^{η} \end{matrix}$ where $P_{t, c}$ is the transmit power of the coordinator of another WBAN, η is the path loss exponent, λ is the wavelength, $G_{t x}$ is the transmitter gain, and $G_{r x}$ is the receiver gain. In our work, we consider $P_{t, c} = 10 mW$ , $G_{t x} = 31.2 mW$ and $G_{r x} = 27.3 mW$ [20] and $η = 2$ . Therefore, the received signal strength, $RSSI (i, j, t)$ , from WBAN $w_{j}$ at $w_{i}$ during time t in terms of dBm [34] can be calculated as $\begin{matrix} (2) & RSSI (i, j, t) = 10 \times log (\frac{P_{r} (i, j, t)}{P_{r f}}) \end{matrix}$ where $P_{r} (i, j, t)$ is the received power at WBAN $w_{i}$ from WBAN $w_{j}$ and $P_{r f}$ is the reference power which is equal to 1 mW.

3.3. Interference

It is assumed that each WBAN can choose at most one channel at a time. Let channel $a_{m} \in A$ is assigned to WBAN $w_{i}$ . We define a binary variable $α_{i, a_{m}}^{t}$ to indicate the channel assignment of WBAN $w_{i}$ at time t. $\begin{matrix} (3) & α_{i, a_{m}}^{t} = \{\begin{matrix} 1, & if channel a_{m} is assigned to w_{i} during t \\ 0, & otherwise \end{matrix} \end{matrix}$

Definition 1.
When $w_{j} \in N_{i}^{t}$ uses the same channel as that of WBANs $w_{i}$ , WBAN $w_{j}$ interferes with $w_{i}$ and we call $w_{j}$ is an interfering WBAN.

We use a binary variable $σ_{i, j}^{t}$ to capture the interference between two WBANs; where $σ_{i, j}^{t} = 1$ means two neighbor WBANs are assigned the same channel. $\begin{matrix} (4) & σ_{i, j}^{t} = \{\begin{matrix} 1, & if α_{i, a_{m}}^{t} = α_{j, a_{m}}^{t} \forall j \in N_{i}^{t} \\ 0, & otherwise \end{matrix} \end{matrix}$ Therefore, the signal-to-interference plus noise ratio (SINR) at the coordinator of $w_{i}$ in presence of a set of interfering WBANs at time t, denoted as ${SINR}_{i}^{t}$ , is calculated as $\begin{matrix} (5) & {SINR}_{i}^{t} = \frac{P_{t, s} d_{i}^{- 2}}{N_{0} + \sum_{j \in N_{i}^{t} j \neq i} σ_{i, j}^{t} P_{t, c} d_{i, j}^{- 2}} \end{matrix}$ where $N_{0}$ is the additive white noise [22], $P_{t, c}$ is the transmit power of the interfering WBAN $w_{j}$ , and $d_{i}$ is the average distance between coordinator and sensor nodes in i-th WBAN. Moreover, the term $\sum_{j \in N_{i}^{t} j \neq i} σ_{i, j}^{t} P_{t, c} d_{i, j}^{- 2}$ in Eq. (5) represents the amount of interference experienced by WBAN $w_{i}$ due to interference from other interfering WBANs. Various notations used in this paper are summarized in Table 1 for better readability.

Table 1
Description of notations

Notations and Meanings

$w_{i}$ WBAN i

K Number of body sensors

$R_{c}$ Transmission range of the coordinator

$d_{i j}$ Euclidean distance between WBAN $w_{i}$ and WBAN $w_{j}$

M Number of available channels

A Set of channels

$a_{m}$ m-th channel

$N_{i}^{t}$ Set of neighbors of WBAN $w_{i}$ at time t

B Maximum data transmission rate

$P_{t, c}$ Transmit power of the coordinator

$P_{t, s}$ Transmit power of the sensor

$P_{r} (i, j, t)$ Received power at $w_{i}$ from $w_{j}$ at time t

$RSSI (i, j, t)$ Received signal strength at $w_{i}$ at time t

${TD}_{i}^{t}$ Transmission delay of $w_{i}$ at time t

${SD}^{t}$ Sensing delay at time t

${WD}^{t}$ Channel switching delay at time t

$D_{i}^{t}$ Delay of $w_{i}$ at time t

$D_{tot}$ Total network delay

${LAC}_{i}^{t}$ List of available channels of ith-WBAN at time t

3.4. Delay analysis

Notations and	Meanings
$w_{i}$	WBAN i
K	Number of body sensors
$R_{c}$	Transmission range of the coordinator
$d_{i j}$	Euclidean distance between WBAN $w_{i}$ and WBAN $w_{j}$
M	Number of available channels
A	Set of channels
$a_{m}$	m-th channel
$N_{i}^{t}$	Set of neighbors of WBAN $w_{i}$ at time t
B	Maximum data transmission rate
$P_{t, c}$	Transmit power of the coordinator
$P_{t, s}$	Transmit power of the sensor
$P_{r} (i, j, t)$	Received power at $w_{i}$ from $w_{j}$ at time t
$RSSI (i, j, t)$	Received signal strength at $w_{i}$ at time t
${TD}_{i}^{t}$	Transmission delay of $w_{i}$ at time t
${SD}^{t}$	Sensing delay at time t
${WD}^{t}$	Channel switching delay at time t
$D_{i}^{t}$	Delay of $w_{i}$ at time t
$D_{tot}$	Total network delay
${LAC}_{i}^{t}$	List of available channels of ith-WBAN at time t

Delay is one of the crucial factors that affect WBAN performance. When fewer channels are available in the network, some neighboring WBANs may use the same channel, increasing their interference. This will reduce the data rate and may lead to latency. The WBAN can dynamically scan and switch among the channels to avoid coexisting interference. However frequent switching increases data transmission delay and energy consumption.

Definition 2.
The delay of a WBAN is defined as the amount of time taken by a packet to reach the coordinator. The following factors determine the delay:
Sensing delay ( $SD$ ): Each WBAN senses the channel for a fixed duration of time before the transmission of data. The coordinator sets the channel sensing time for each WBAN. It divides the total sensing period ( $T_{sense}$ ) into M number of time slots of fixed duration, one for each channel. A channel can be sensed by N number of nodes, so the channel sensing delay ${SD}^{t}$ for each WBAN at time t is $\frac{T_{sense}}{N * M}$ .

Transmission delay ( $TD$ ): Transmission of a sensor k to send a packet of length $l_{k}$ to the coordinator. Transmission delay of WBAN $w_{i}$ during time t is denoted as ${TD}_{i}^{t}$ and it is given by $\begin{matrix} (6) & {TD}_{i}^{t} = \sum_{k = 1}^{K} \frac{l_{k}}{B log (1 + {SINR}_{i}^{t})} \end{matrix}$ where B is the data transmission rate.

Channel switching delay ( $WD$ ): This delay is due to switching from one channel to another channel and it is represented by ${WD}^{t}$ .

Thus, the delay experienced by WBAN $w_{i}$ at time t, $D_{i}^{t}$ , is $\begin{matrix} (7) & D_{i}^{t} = {SD}^{t} + {TD}_{i}^{t} + β_{i}^{t} {WD}^{t} \end{matrix}$ where $\begin{matrix} (8) & β_{i}^{t} = \{\begin{matrix} 1, & if w_{i} switches to a channel \\ 0, & otherwise \end{matrix} \end{matrix}$ where $β_{i}^{t}$ stands a binary variable representing the channel switching of WBAN $w_{i}$ . The value of $β_{i}^{t}$ becomes 1 when WBAN $w_{i}$ switches to another channel during time t, otherwise it is 0. Definition 3.
The total network delay, denoted as $D_{tot}$ , is defined as the cumulative delay of all the WBANs over a time period T and is calculated as $\begin{matrix} (9) & D_{tot} = \sum_{i = 1}^{N} \sum_{t = 1}^{T} D_{i}^{t} \end{matrix}$

3.5. Problem definition

In a network with M channels and N coexisting WBANs, we aim to determine the channel assignment profile of the WBANs to minimize the total network delay. Therefore, the problem can be formulated as follows: $\begin{matrix} (10) & Minimize D_{tot} = min \sum_{i = 1}^{N} \sum_{t = 1}^{T} D_{i}^{t} \end{matrix}$ Subject to:

$\sum_{a_{m} \in A} α_{i, a_{m}}^{t} = 1$ , $\forall i \in N$

$α_{i, a_{m}}^{t} \neq α_{j, a_{m}}^{t}$ , $\forall i, j \in N$ , $\forall j \in N_{i}^{t}$

$P_{t, c} ⩽ P_{max}$

$P_{t, s} ⩽ P_{max}$

$D_{i}^{t} ⩽ D_{max}$

$β_{i}^{t} ⩾ α_{i, a_{m}}^{t} - α_{i, a_{m}}^{t - 1}$ , $\forall i \in N$ , $a_{m} \in A$

Channel assignment constraint C1 guarantees that a WBAN can assign only one channel at a time. Interference constraint C2 elaborates that when the same channel is assigned to two neighboring WBANs $w_{i}$ and $w_{j}$ , interference may arise between them. Power constraints C3 and C4 respectively impose transmit power of the coordinator and the sensor does not exceed its maximum power. Delay constraint C5 imposes an upper limit $D_{max}$ on $D_{i}^{t}$ to meet the delay requirement of the data traffic. Channel switching constraint C6 states that for any WBAN $w_{i}$ and channel $a_{m}$ , when $α_{i, a_{m}}^{t}$ takes the value of 1 at time t and the value of $α_{i, a_{m}}^{t - 1}$ is 0 at time $t - 1$ , then the value of $β_{i}^{t}$ must be 1, which represents a channel switching.

4. Primitives

In this section, we will discuss the autoregressive (AR) model and simplified moving average (MA) model used in the proposed algorithm.

4.1. Autoregressive model

The autoregressive (AR) model of order p [2,30] is a linear model used to predict the current value $X_{n}$ based on the weighted sum of the last p value and a stochastic term. We can define the AR model as the following: $\begin{matrix} (11) & X_{n} = c + \sum_{i = 1}^{p} ϕ_{i} (X_{n - i}) + ϵ_{n} \end{matrix}$ where $X_{n}$ represents the current value, $ϕ_{i}$ represents the autoregression coefficient, c is constant, and $ϵ_{n}$ indicates prediction error (also known as white noise) [30]. For the sake of simplicity, we omit c. In AR model, determining the value of coefficients $ϕ_{i}$ for $i \in [1, \dots, p]$ is essential to predict the accurate value. The model must be updated continuously with new samples to ensure reasonable accuracy or maybe recomputed when the predicted and actual value’s error difference becomes significant.

4.2. Simplified moving average model

Moving Average (MA) is another approach for modeling time-series data, where the output value linearly depends on the current and past values. We can calculate the MA model of order p by taking the average of p recent samples as $\begin{matrix} (12) & X_{n} = \frac{1}{p} \sum_{i = 1}^{p - 1} (X_{n - i}) \end{matrix}$

5. Proposed channel assignment with reduced channel switching

The frequency of channel switching is not cost effective in terms of delay and energy. Therefore it is important to design an efficient channel switching and assignment algorithm by exploiting the coexistence capability effectively that brings benefits on coexisting WBANs. This section presents a Prediction-based Channel Assignment (PCA) algorithm that minimizes the delay of the WBAN data transmissions by assigning channels in a conflict–avoiding way and by minimizing channel switching based on neighborhood prediction.The proposed method is designed with three phases: (a) initial channel assignment, (b) predicting future neighbors, and (c) channel switching and assignment. In the first phase, conflict-free initial channel allocation is executed based on the priority of the WBANs and their neighborhood. In the channel switching phase, due to movement, if another neighbor occupies the currently used channel, WBAN switches its channel based on neighborhood prediction and channel information of current and predicted neighbors so that the new channel assignment would remain valid in next-time instances. We present the initial channel assignment and prediction-based channel switching in the following subsections in detail.

Table 2
User priority mapping in IEEE 802.15.6

User priority Packet type

1 Background (BK)

2 Best effort (BE)

3 Excellent effort (EE)

4 Video (VI)

5 Voice (VO)

6 Low-priority medical data (e.g., body TMP)

7 High-priority medical data (e.g., body SBP)

8 Emergency medical data (e.g., body ECG)

User priority	Packet type
1	Background (BK)
2	Best effort (BE)
3	Excellent effort (EE)
4	Video (VI)
5	Voice (VO)
6	Low-priority medical data (e.g., body TMP)
7	High-priority medical data (e.g., body SBP)
8	Emergency medical data (e.g., body ECG)

5.1. WBAN priority

In this paper, we define WBAN priority ( $WP$ ) based on the user priority of the sensors. Let the sensing data generated by the sensors in a WBAN be classified into L-different user priorities, which are denoted by $θ_{1}, θ_{2}, \dots, θ_{L}$ . Based on IEEE 802.15.6 [5], we consider eight different user priorities i.e. $L = 8$ and $θ_{1} < θ_{2} < θ_{3} \dots < θ_{L}$ . The user priority table can be found in Table 2. Higher user priority indicates emergency medical data which needs to be transmitted in a lower delay. We consider that each node in the WBAN only collects one kind of user priority traffic. The user priority of sensor k in ith WBAN is represented as ${SP}_{i, k} = θ_{l}$ if sensor k collects data of $θ_{l}$ – priority where $l \in L$ . Hence the priority of ith WBAN $({WP}_{i})$ can be obtained through Eq. (13) $\begin{matrix} (13) & {WP}_{i} = \frac{1}{K} \sum_{k \in K} {SP}_{i, k} \end{matrix}$ In our work, the WP values vary from 1 to 8. A WP is less than 6 indicates WBAN with non-medical data, while WP greater than equal to 6 indicates WBAN with medical data. Moreover, high priority WBANs are more competitive to select channels in multi-WBANs with interference.

5.2. Initial channel assignment

An AP does the channel assignment in an interference-avoiding way by considering the priority of WBAN, as shown in Algorithm 1 . Each WBAN knows its neighbors and periodically broadcasts the neighbor information to AP. Based on the WBAN priority and neighbor information, AP assigns a single channel to each WBAN. At first, all the WBANs are sorted in descending order of their priorities. Let the ordering is $w_{1}, w_{2}, \dots, w_{N}$ and $a_{1}, a_{2}, \dots, a_{M}$ are the available channels. The highest priority WBAN, $w_{1}$ , is selected, and AP assigns the first available channel, say $a_{1}$ to it. AP broadcasts this information so that the neighbors of $w_{1}$ have information about the channel assignment. After that, the next WBAN, $w_{2}$ is selected and is assigned the next available channel if the channel is unused in its neighbors, i.e., if $w_{1}$ and $w_{2}$ are neighbors. Otherwise, they can share the same channel, i.e., $a_{1}$ . AP allocates a channel to the WBAN using random fit if no available channel exists. The channel allocation process is repeated until all the WBANs get the channel for data transmission.

Algorithm 1:

Initial channel assignment

5.3. Predicting future neighbors

At time t, AP checks whether $N_{i}^{t - 1}$ of $w_{i}$ is the same as $N_{i}^{t}$ of $w_{i}$ . If yes, $w_{i}$ does not switch the channel. Otherwise, the channel allocation at $w_{i}$ at $t - 1$ is no longer qualified in $N_{i}^{t}$ . AP switches the channel of $w_{i}$ so that the new channel remains valid the next time. Moreover, when two interfering WBANs want to switch channels, the WBAN (say $w_{i}$ ) with the highest priority ( $WP$ ) is chosen. To switch channels, AP predicts the future neighbors i.e., $N_{i}^{t + 1}$ of $w_{i}$ based on the history of signal strength. For each WBAN $w_{i}$ , AP maintains a matrix $H_{i}$ of size $(N - 1) * p$ where each row represents a WBAN and the column values represent the corresponding received signal strength ( $RSSI (i, j, t)$ ) between $w_{i}$ and remaining $(N - 1)$ WBANs over past $t - p$ times including t. Here p is the window size and predefined parameter, significantly affecting the prediction accuracy discussed in the next section. Moreover, the zero-entry in the matrix indicates that the corresponding WBAN may not have been neighboring with $w_{i}$ during the past instances. On the other hand, the non-zero row values of the matrix are the signal strength between the WBANs over time, represented as time-series data.

Table 3
Example of $H_{1}$ matrix of $w_{1}$ at t

WBAN ID $t - 4$ $t - 3$ $t - 2$ $t - 1$ t

$w_{2}$ 0 −2.341 0 −2.097 6.923

$w_{3}$ 0 0 −2.341 0 2.928

$w_{4}$ −2.097 2.869 0 0 −2.766

$w_{5}$ −0.087 0 0 −0.087 0.228

$w_{6}$ 2.198 0.196 0 0 −3.63

$w_{7}$ 0 0 −0.664 0 1.66

$w_{8}$ 2.619 0 0 −1.178 −2.25

$w_{9}$ 0 0 −1.178 0 2.6

$w_{10}$ −0.747 −1.296 0 0 2.91

WBAN ID	$t - 4$	$t - 3$	$t - 2$	$t - 1$	t
$w_{2}$	0	−2.341	0	−2.097	6.923
$w_{3}$	0	0	−2.341	0	2.928
$w_{4}$	−2.097	2.869	0	0	−2.766
$w_{5}$	−0.087	0	0	−0.087	0.228
$w_{6}$	2.198	0.196	0	0	−3.63
$w_{7}$	0	0	−0.664	0	1.66
$w_{8}$	2.619	0	0	−1.178	−2.25
$w_{9}$	0	0	−1.178	0	2.6
$w_{10}$	−0.747	−1.296	0	0	2.91

Therefore, based on this matrix, we apply an autoregressive (AR) model of order p and moving-average (MA) model. After modeling, we might predict $RSSI (i, j, t)$ between $w_{i}$ and other $(N - 1)$ WBANs at time $t + 1$ . Table 3 shows matrix $H_{1}$ of $w_{1}$ , which contains RSSI values of $w_{1}$ at beginning of time t. The equations used by the AR and MA models to predict $RSSI (i, j, t + 1)$ between two WBANs $w_{i}$ and $w_{j}$ based on past received signal strength are the following. $\begin{array}{l} (14) & AR : RSSI (i, j, t + 1) = \sum_{i = 1}^{p} ϕ_{i} RSSI (i, j, t - i) + ϵ_{t} \\ (15) & MA : RSSI (i, j, t + 1) = \frac{1}{p} \sum_{i = 1}^{p - 1} RSSI (i, j, t - i) \end{array}$

There are several approaches for determining the appropriate value of coefficients $ϕ_{i}$ for $i \in [0, \dots, p - 1]$ such as Yule–Walker equations, forward-backward method etc. In this work, we have estimated the value of ϕ and $ϵ_{t}$ in Eq. (14) using the Yule–Walker equations [30]. The WBANs for which the predicted $RSSI (i, j, t + 1)$ is higher than a given threshold value become the neighbors of $w_{i}$ at $t + 1$ . Furthermore, when WBANs reach at time $t + 1$ , the actual neighbors are recomputed by broadcasting control messages, and the actual signal strength measurements are used to update the prediction model.

5.3.1. Prediction accuracy evaluation

Fig. 2.

Prediction accuracy with respect to varying number of WBANs.

Fig. 3.

Prediction accuracy with respect to varying speed.

Fig. 4.

Error with respect to WBAN ID for (a) window size = 5 (b) window size = 10.

We have used prediction accuracy to assess the performance of the prediction models (both AR and MA). Prediction accuracy defines the percentage of actual neighbors detected by a model. We have tested the AR and MA models’ prediction accuracy for varying WBANs and the speed of WBAN. Here, we have considered 50-coexisting WBANs. Figure 2 shows the prediction accuracy of each model for the different number of WBANs, where the accuracy value decreases with increasing number of WBANs. The figure shows that the AR model has better performance and improved accuracy from 48% to 74% than the MA model.

On the other hand, Fig. 3 illustrates the effect of speed on the prediction accuracy, and we can observe that the accuracy falls when the speed increases. However, the AR model has achieved better accuracy than the MA model. In another experiment, we evaluate the impact of p on prediction accuracy. We fix the number of coexisting WBANs to 50 and vary the speed from 0.2 m/s to 2 m/s.

Figure 4(a) and (b) show the error (average deviation from the actual neighbors) of predicted values for coexisting 50-WBANs when $p = 5$ and $p = 10$ respectively. We can notice that when $p = 5$ , the error is much higher in both models. This high error value may be due to under-fitting. However when $p = 10$ , the error value is reduced, and the AR model produces a lower error than the MA model.

5.4. Channel switching and assignment

During movement, a WBAN may experience channel conflict due to topology changes. To switch a channel during conflict, the proposed PCA requires each WBAN ( $w_{i}$ ) to maintain a list of channels being used by the neighbor set of $w_{i}$ at the current time t. This allocated channel information is sent periodically to AP. After receiving the list, AP records the set of channels that are not being used in $N_{i}^{t}$ and evaluates a list of available channels for $w_{i}$ at t, and it is given by ${LAC}_{i}^{t}$ . At the current time t, if channel conflict occurs between neighboring WBANs $w_{i}$ and $w_{j}$ , high priority WBAN immediately switches its assigned channel. Channel switch is executed based on the allocated channels of the predicted neighbors at $t + 1$ and the available channels ${LAC}_{i}^{t}$ at current time t. If ${LAC}_{i}^{t}$ is empty, $w_{i}$ does not switch its channel. If ${LAC}_{i}^{t}$ is not empty, WBAN $w_{i}$ chooses those channels from ${LAC}_{i}^{t}$ , which would remain valid at both the instances t and $t + 1$ , thus giving better channels for transmission. If more than one suitable channel is found, it randomly selects a channel and switches to the selected channel.The proposed channel switching and assignment algorithm is shown in Algorithm 2.

Algorithm 2:

Channel switching algorithm for a WBAN ( $w_{i}$ )

5.5. Example of proposed PCA

To illustrate PCA, we present an example. We consider 10-coexisting WBANs and 3-channels: $a_{1}$ , $a_{2}$ , $a_{3}$ . Initial network topology at time $t_{0}$ is shown in Fig. 5(a). Let each WBAN has a priority, given in Table 4. Each WBAN is assigned a channel marked above the node. At time $t_{0}$ , WBAN $w_{1}$ may use the channel $a_{3}$ , $w_{2}$ , $w_{3}$ are using channel $a_{2}$ and $a_{3}$ , respectively, and so on. Initial channel assignment is given in Table 5. At time $t_{1}$ , WBANs move, and topology changes. Node mobility follows the Random Way Point model. WBAN $w_{1}$ enters into the range of $w_{6}$ , and $w_{6}$ moves out of the range of $w_{10}$ , shown in Fig. 5(b). Unfortunately, there are some channel conflicts between pairs of WBANs $⟨ w_{5}, w_{4} ⟩$ , $⟨ w_{6}, w_{3} ⟩$ , $⟨ w_{1}, w_{6} ⟩$ and $⟨ w_{2}, w_{9} ⟩$ , marked with stars in Fig. 5(b). To solve this conflict, WBANs need to switch channels. Let us consider a conflict pair $⟨ w_{5}, w_{4} ⟩$ . Since the priority of $w_{5}$ is higher than that of $w_{4}$ , so $w_{5}$ will switch the channel. The free channels for $w_{5}$ are $a_{2}$ , $a_{3}$ . The random-CA technique selects a free channel randomly and thus chooses $a_{2}$ for $w_{5}$ . In PCA-AR, we adjust the channel of $w_{5}$ to $a_{3}$ . When the topology changes at time $t_{2}$ , shown in Fig. 5(c), channel interference is rechecked. In random-CA, $w_{5}$ will again experience channel conflict on $a_{2}$ and will switch to $a_{3}$ and increase switching overhead. As a result, the total delay is about 0.06 s. On the other hand, according to the proposed method, $w_{5}$ does not need to switch the channel during this movement, thus lowering the switching delay. Our PCA-AR minimizes the switching while ensuring conflict-free channel assignment during movement. Now, the total delay under PCA-AR is 0.02 s, which is lower than that of random-CA. Channel assignment using PCA-AR and Random-CA at $t_{1}$ ,and $t_{2}$ are shown in Table 6 and Table 7, respectively.

Fig. 5.

Coexisting WBAN topology at three different times (a) $t_{0}$ (b) $t_{1}$ (c) $t_{2}$ .

Table 4

Priority of WBANs

WBANs	$w_{1}$	$w_{2}$	$w_{3}$	$w_{4}$	$w_{5}$	$w_{6}$	$w_{7}$	$w_{8}$	$w_{9}$	$w_{10}$
WP	2	5	1	6	7	4	3	2	1	2

Table 5

Initial channel assignment at $t_{0}$

WBANs	$w_{1}$	$w_{2}$	$w_{3}$	$w_{4}$	$w_{5}$	$w_{6}$	$w_{7}$	$w_{8}$	$w_{9}$	$w_{10}$
channels	$a_{3}$	$a_{2}$	$a_{3}$	$a_{1}$	$a_{1}$	$a_{3}$	$a_{2}$	$a_{3}$	$a_{2}$	$a_{2}$

Table 6

Channel assignment (CA) at $t_{1}$ and $t_{2}$ using PCA-AR

WBANs	CA at time $t_{1}$	CA at time $t_{2}$
$w_{1}$	$a_{3}$	$a_{3}$
$w_{2}$	$a_{2}$	$a_{2}$
$w_{3}$	$a_{3}$	$a_{3}$
$w_{4}$	$a_{1}$	$a_{1} \to a_{2}$
$w_{5}$	$a_{1} \to a_{3}$	$a_{3}$
$w_{6}$	$a_{3} \to a_{1}$	$a_{1}$
$w_{7}$	$a_{2}$	$a_{2}$
$w_{8}$	$a_{3}$	$a_{3}$
$w_{9}$	$a_{2} \to a_{1}$	$a_{1}$
$w_{10}$	$a_{2}$	$a_{2}$

Table 7

Channel assignment (CA) at $t_{1}$ and $t_{2}$ using random-CA

WBANs	CA at time $t_{1}$	CA at time $t_{2}$
$w_{1}$	$a_{3}$	$a_{3}$
$w_{2}$	$a_{2}$	$a_{2} \to a_{1}$
$w_{3}$	$a_{3}$	$a_{3} \to a_{2}$
$w_{4}$	$a_{1}$	$a_{1} \to a_{2}$
$w_{5}$	$a_{1} \to a_{2}$	$a_{2} \to a_{3}$
$w_{6}$	$a_{3} \to a_{1}$	$a_{1}$
$w_{7}$	$a_{2}$	$a_{2} \to a_{1}$
$w_{8}$	$a_{3}$	$a_{3} \to a_{2}$
$w_{9}$	$a_{2} \to a_{1}$	$a_{1}$
$w_{10}$	$a_{2}$	$a_{2} \to a_{1}$

6. Performance evaluation

In this section, we evaluate the performance of the proposed algorithm (PCA) through extensive simulations using MATLAB. We have applied auto-regressive and moving average models to PCA and called them PCA-AR, PCA-MA. PCA-AR and PCA-MA are compared with random channel assignment (Random-CA) and 2TM-MAC [39]. All the algorithms are evaluated in terms of total network delay, throughput, switching overhead, and the number of conflicting channel allocations. We run each simulation for each algorithm 25 times and then take the average. Our experiments consider a $10 m \times 10 m$ network topology, where N-WBANs, each with eight sensors, are randomly located. The maximum transmit power of each WBAN is set to 10 mW [22], and the transmission rate is 240 kbps. The transmission range for each WBAN is set to 2 m, whereas the average distance between sensors and the coordinator in a WBAN may vary from 0.1 m–0.2 m as referenced in [23]. The packet size of each sensor depends on its user priority and increases from 50 bytes to 350 bytes for priority 1 to priority 8. Furthermore, each WBAN follows the Random Waypoint (RWP) model during movement. In the simulation studies, we consider two-movement scenarios. Scenario 1 represents a dynamic scenario where WBANs move fast and the moving speed of a WBAN varies from 1.2 m/s to 2 m/s, while scenario 2 indicates the semi-dynamic scenario where WBANs move slowly and the moving speed varies between 0.2 m/s to 1 m/s. The pause time is 1.5 s. Moreover, we set the total sensing period for all the channels to 0.5 s, and the channel sensing delay for each WBAN to 0.001 s when the network has 10 available channels and 50-WBANs. Other parameters used in the simulations are summarized in Table 8.

Table 8
List of simulation parameters

Parameters Value

Simulation area 10 m × 10 m

Number of WBANs 10–50 (default value 50)

Number of sensors in a WBAN 8

Average distance between coordinator and the sensors 0.1 m–0.2 m [23]

Mobility model Random waypoint model

WBAN speed in semi-dynamic scenario 0.2 m/s to 1 m/s

WBAN speed in dynamic scenario 1.2 m/s to 2 m/s

Transmission range 2 m

Transmit power of the coordinator 10 mW

Total sensing period 0.5 s

Sensing delay 1 ms

Channel switching delay 1.3 ms [28]

Maximum delay ( $D_{max}$ ) 0.125 s (for priority 8, 7, 6) and 0.25 s (priority 5, 4, 3, 2, 1)

Number of channels 2 to 10 (default value 10)

Size of data packet 50 bytes to 350 bytes

Data transmission rate 250 kbps

Frequency 2.4 GHz

Simulation time 50 s

Parameters	Value
Simulation area	10 m × 10 m
Number of WBANs	10–50 (default value 50)
Number of sensors in a WBAN	8
Average distance between coordinator and the sensors	0.1 m–0.2 m [23]
Mobility model	Random waypoint model
WBAN speed in semi-dynamic scenario	0.2 m/s to 1 m/s
WBAN speed in dynamic scenario	1.2 m/s to 2 m/s
Transmission range	2 m
Transmit power of the coordinator	10 mW
Total sensing period	0.5 s
Sensing delay	1 ms
Channel switching delay	1.3 ms [28]
Maximum delay ( $D_{max}$ )	0.125 s (for priority 8, 7, 6) and 0.25 s (priority 5, 4, 3, 2, 1)
Number of channels	2 to 10 (default value 10)
Size of data packet	50 bytes to 350 bytes
Data transmission rate	250 kbps
Frequency	2.4 GHz
Simulation time	50 s

6.1. Simulation results and discussion

A. Analysis of total network delay

Figure 6 (a) and (b) present the total network delay for varying number of WBANs under different channel assignment techniques in dynamic and semi-dynamic scenarios, respectively. Here we vary the number of WBANs (N) from 10 to 50 with an increment of 10 and fix the number of channels to 10. We see that when N increases, the total delay also increases. The figures show that the delay in semi-dynamic scenario is comparatively lower than that of dynamic scenario. However, in both scenarios, PCA-AR has achieved the lowest total delay while Random-CA has the highest delay. The reason can be explained as follows: PCA-AR predicts the neighborhood of WBAN during switching and assigns the channel to a WBAN based on the prediction, which helps to minimize the number of channel conflicts and switches, and as a result, the delay decreases.On the other hand, 2TM-MAC is a multi-channel assignment scheme, but it did not consider prediction and channel negotiation – a reason behind increasing channel switching and latency. Compared to 2TM-MAC and Random-CA, the delay of PCA-AR is decreased by 43% and 75%, respectively in dynamic scenario. We also observe that PCA-MA has obtained a higher delay than PCA-AR; however, it is much lower than the other two techniques.

In Fig. 7, we investigate the impact of the number of channels on total network delay in both dynamic and semi-dynamic scenarios. In this experiment, we fix the number of WBANs to 50 and change the number of channels from 2 to 10 with an increment of 1. We see that the total delay decreases when the number of channels grows up. Adding more channels minimizes channel switching overhead, and contention among WBANs; as a consequence, the latency is lowered. The results show that PCA-AR minimizes the network delay more compared to Random-CA and 2TM-MAC as shown in Fig. 7(a). On the other hand, PCA-MA has a comparatively higher delay than PCA-AR in semi-dynamic scenario, but the delay gap between them gets closer when the number of channels is more than 7, as seen in Fig. 7(b).

Fig. 6.

Total network delay with respect to varying number of WBANs in (a) dynamic scenario (b) semi-dynamic scenario.

Fig. 7.

Total network delay with respect to varying number of channels in (a) dynamic scenario (b) semi-dynamic scenario.

B. Analysis of throughput

The relationship between the throughput and different numbers of WBANs is evaluated in Fig. 8 for dynamic and semi-dynamic scenarios. In this experiment, we set the number of channels is 10 but vary the number of WBANs from 10 to 50. With an increasing number of WBANs, the channel conflict becomes more prominent, increasing network interference and consequently reducing the throughput. We find that the throughput in semi-dynamic scenario is much better than in dynamic scenario. Although in both scenarios, PCA-AR achieves higher throughput than other algorithms. From Fig. 8 (a), we see that PCA-AR increases throughput by 19% to PCA-MA and by 43% and 75% compared to 2TM-MAC and Random-CA, respectively. In dynamic scenario, PCA-AR outperforms 2TM-MAC and Random-CA, whereas it is marginally better than 2TM-MAC in semi-dynamic scenario when the number of WBANs grows, as shown in Fig. 8 (b). Further, we see that the throughput of PCA-MA is marginally better than 2TM-MAC in both scenarios.

C. Impact of WBAN priority on total delay and throughput

Figure 9 (a) shows the impact of WBAN priority on total network delay in different numbers of coexisting WBANs. Here, we have considered two types of networks. A network with 25 coexisting WBANs and a network with 50 coexisting WBANs and then we compare our results with the 2TM-MAC protocol. It is observed that the delay decreases with increasing WBAN priority. The proposed method PCA-AR performs better than the existing 2TM-MAC. In PCA-AR, high-priority WBANs are assigned the channels before to satisfy emergency data requirements, and channel prediction is used to reduce unwanted channel switching, which minimizes the delay. On the other hand, 2TM-MAC performs multi-channel assignments based on WBAN priority without considering channel switching, leading to delay and switching overhead. The variation of throughput with WBAN priorities for varying numbers of WBAN is shown in Fig. 9(b). The throughput increases with WBAN priority. Higher throughput results in higher packet delivery. The figure shows that PCA-AR performs better than 2TM-MAC with increasing WBAN priority.

Fig. 8.

Network throughput with respect to number of WBANs in (a) dynamic scenario (b) semi-dynamic scenario.

Fig. 9.

Impact of WBAN priorities on (a) total network delay (b) throughput.

Fig. 10.

Impact of user priorities on packet delay.

Fig. 11.

Number of conflicting channel allocations with respect to varying number of WBANs in (a) dynamic scenario (b) semi-dynamic scenario.

D. Packet delay analysis for different user priorities

Figure 10 shows the packet delays of the proposed PCA-AR, PCA-MA and other two schemes with different user priorities. We consider 50-coexisting WBANs and 10-available channels. Hence, channel sensing delay experienced by each WBAN is 0.001 s. We see that with increasing user priorities, the packet delays of all the schemes decrease, and PCA-AR yields the lowest delay. As the figure displays, all the schemes have satisfied the required time latency of the medical data (i.e., data with user priority greater than equal 6), which should be less than 125 ms or 0.125 s [10]. On the other hand, the required delay for the non-medical data is higher, which is less than 250 ms [10], and the proposed method satisfies that requirement when user priority is below 6.

E. Analysis of the number of conflicting channel allocations

Figure 11 and Fig. 12 reveal the number of conflicting channel allocations performed by PCA-AR, PCA-MA, 2TM-MAC, and Random-CA for varying numbers of WBAN and channels, respectively. As shown in Fig. 11, channel conflicts grow in the network with an increasing number of WBANs. The results highlight that the performance of 2TM-MAC and Random-CA was initially better than PCA-AR and PCA-MA, but PCA-AR improves the performance significantly when the network becomes dense. However, PCA-MA is marginally better than 2TM-MAC. On the other hand, PCA-AR in semi-dynamic scenario has better performance than others, as shown in Fig. 11 (b). We observe from Fig. 12 that the number of channel conflicts in PCA-AR reduces with the increasing number of channels, and much lower than in other techniques.

Fig. 12.

Number of conflicting channel allocation with respect to number of available channel in (a) dynamic scenario (b) semi-dynamic scenario.

Fig. 13.

Switching overhead vs varying speed of WBANs in (a) dynamic scenario (b) semi-dynamic scenario.

Fig. 14.

Total network delay w.r.t varying speed of WBANs (a) dynamic scenario (b) semi-dynamic scenario.

Fig. 15.

Throughput vs varying speed in (a) dynamic scenario (b) semi-dynamic scenario.

F. Impact of moving speed of WBAN

Figures 13, 14 and 15 evaluate the impact of the moving speed of WBAN on switching overhead, total network delay, and throughput under all the schemes, respectively. Here we set the number of WBANs to 50 and the number of channels to 10. We vary the moving speed of a WBAN from 1.2 m/sec to 2.0 m/sec with a step size 1.2 m/sec. The figures show that the total delay and switching overhead rise with increasing WBAN’s speed, but the throughput decreases. PCA-AR achieves lower delay, lower switching overhead, and higher throughput compared to 2TM-MAC and Random-CA. When the speed increases, the network topology frequently changes, leading more WBANs to switch their channels often to lessen the channel contention. As a result, the delay rises, and throughput drops. Compared with 2TM-MAC and Random-CA, PCA-AR predicts the future neighbors and selects the channels accordingly to remain valid in next-time instances, thus reducing channel contention and switching overhead. Although PCA-MA is not superior to PCA-AR, but it can give better performance in terms of delay, throughput, and switching overhead compared to other existing channel assignment techniques.

7. Conclusion and future work

Channel assignment in coexisting WBANs is challenging when WBAN users move. As a result, the channels get conflicted frequently. Proper channel assignment minimizes the number of frequent channel switches and thus reduces the total network delay in data transmission. In this paper, we proposed a channel assignment scheme in a dynamic and coexisting WBANs environment, which selects the free channels for WBANs when channel conflict occurs. The proposed method assigned the channels to WBANs during switching based on neighborhood prediction, mobility, and current channel assignment. We applied RWP model to capture the mobility of the WBANs and machine learning algorithms for neighborhood prediction. Simulation results highlighted that the proposed method is more resilient in a dynamic environment and minimizes the number of channel switches and network delay than existing channel assignment algorithms. In the future, we will investigate the impact of energy efficiency and link quality during channel switching in mobile coexisting WBANs.

Footnotes

Acknowledgements

The first author would like to thank Ministry of Human Resource Development (MHRD), India for providing the facilities and support to carry out this work.

Conflict of interest

The authors have no conflict of interest to report.

References

Adhikary,

Chattopadhyay,

Ghosh,

Choudhury,

S.B.

Nath and

Garain, Reliable routing in wireless body area network using optimum number of relay nodes for enhancing network lifetime, Journal of Ambient Intelligence and Smart Environments (2022), 1–9.

Aguiar and

Wolisz, Channel prediction heuristics for adaptive modulation in WLAN, in: IEEE 65th Vehicular Technology Conference-VTC Spring, 2007, pp. 1091–1095.

Ali,

Moungla,

Younis and

Mehaoua, Efficient medium access arbitration among interfering WBANs using Latin rectangles, Ad Hoc Networks 79 (2018), 87–104. doi:10.1016/j.adhoc.2018.04.001.

Kazemi,

Vesilo,

Dutkiewicz and

Fang, Inter-network interference mitigation in wireless body area networks using power control games, in: IEEE International Symposium on Communications and Information Technologies, 2010, pp. 81–86.

Chen,

Chen and

Shen, 2L-MAC: A MAC protocol with two-layer interference mitigation in wireless body area networks for medical applications, in: IEEE International Conference on Communications (ICC), 2014, pp. 3523–3528.

Cheng,

Huang and

Tu, RACOON: A multiuser QoS design for mobile wireless body area networks, Journal of medical systems (2011), 1277–1287. doi:10.1007/s10916-011-9676-3.

S.H.

Cheng and

Ching, Coloring based inter-WBAN scheduling for mobile wireless body area networks, IEEE Transactions on parallel and distributed systems 24(2) (2012), 250–259. doi:10.1109/TPDS.2012.133.

M.N.

Deylami and

Jovanov, A distributed scheme to manage the dynamic coexistence of IEEE 802.15.4-based health-monitoring WBANs, IEEE journal of biomedical and health informatics 18(1) (2013), 327–334. doi:10.1109/JBHI.2013.2278217.

Fan,

Xuxun,

Huan,

Victor,

Jian and

Alex, Efficient resource scheduling for interference alleviation in dynamic coexisting WBANs, IEEE Transactions on Mobile Computing (2021).

10.

George and

Jacob, Multi-class delay sensitive medical packet scheduling in inter-WBAN communication, in: IEEE International Conference on Advanced Networks and Telecommunications Systems (ANTS), 2019, pp. 1–5.

11.

George and

Jacob, Interference mitigation for coexisting wireless body area networks: Distributed learning solutions, IEEE Access 8 (2020), 24209–24218. doi:10.1109/ACCESS.2020.2970581.

12.

Huang and

Quek, Adaptive CSMA/CA MAC protocol to reduce inter-WBAN interference for wireless body area networks, in: IEEE International Conference on Wearable and Implantable Body Sensor Networks (BSN), IEEE, 2015, pp. 1–6.

13.

Khan,

Nawaz,

Ahmad,

Ahmed,

Alam and

Drieberg, Interference and priority aware coexistence (IPC) algorithm for link scheduling in ieee 802.15. 6 based wbans, IEEE Access 7 (2019), 168736–168751. doi:10.1109/ACCESS.2019.2955395.

14.

Kim,

Jinsung,

Kim and

Lee, ACESS: Adaptive channel estimation and selection scheme for coexistence mitigation in WBANs, in: Proceedings of ACM International Conference on Ubiquitous Information Management and Communication, 2016, pp. 1–4.

15.

Kim,

J.W.

Kim and

D.S.

Eom, A beacon interval shifting scheme for interference mitigation in body area networks, Sensors 12(8) (2012), 10930–10946. doi:10.3390/s120810930.

16.

Kirbas,

Karahan,

Sevin and

Bayilmis, IsMAC: An adaptive and energy-efficient mac protocol based on multi channel communication for wireless body area networks, KSII Transactions on Internet and Information Systems (TIIS) 7(8) (2013), 1805–1824. doi:10.3837/tiis.2013.08.004.

17.

Ko,

Lu,

M.B.

Srivastava,

A.J.

Stankovic,

Terzis and

Welsh, Wireless sensor networks for healthcare, Proceedings of the IEEE 98(11) (2010), 1947–1960. doi:10.1109/JPROC.2010.2065210.

18.

T.T.

Le and

Moh, Interference mitigation schemes for wireless body area sensor networks: A comparative survey, Sensors 15(6) (2015), 13805–13838. doi:10.3390/s150613805.

19.

T.T.

Le and

Moh, An interference aware traffic priority based link scheduling algorithm for interference mitigation in multiple wireless body area networks, Sensor 16(12) (2016), 2190. doi:10.3390/s16122190.

20.

T.T.

Le and

Moh, Link scheduling algorithm with interference prediction for multiple mobile WBANs, Sensors 17(10) (2017), 2231. doi:10.3390/s17102231.

21.

T.T.

Le and

Moh, Hybrid multi channel mac protocol for WBANs with inter-WBAN interference mitigation, Sensors 18(5) (2018), 1373. doi:10.3390/s18051373.

22.

T.T.T.

Le and

Moh, A spectrum-aware priority-based link scheduling algorithm for cognitive radio body area networks, Sensors 19(7) (2019), 1640. doi:10.3390/s19071640.

23.

Li,

Zhang,

Yuan,

Ullah and

A.V.

Vasilakos, MC-MAC: A multi channel based MAC scheme for interference mitigation in WBANs, Wireless Networks 24(3) (2018), 719–733. doi:10.1007/s11276-016-1366-0.

24.

Liu,

Jing and

Hui, Power control mechanism based on hybrid access schemes in coexisting WBANs, in: IEEE International Conference on Wireless Communications and Signal Processing (WCSP), 2016, pp. 1–5.

25.

Ma and

Mu, Improved unsupervised coloring algorithm for spectrum allocation in multiple wireless body area networks, Ad Hoc Networks 111 (2021), 102326. doi:10.1016/j.adhoc.2020.102326.

26.

Movassaghi,

Abolhasan,

Lipman,

Smith and

Jamalipour, A wireless body area networks: A survey, IEEE Communications surveys and tutorials 16(3) (2014), 1658–1686. doi:10.1109/SURV.2013.121313.00064.

27.

Movassaghi,

Mehran,

David and

Abbas, AIM: Adaptive inter network interference mitigation amongst co-existing wireless body area networks, in: IEEE Global Communications Conference, 2014, pp. 2460–2465.

28.

N.Q.

Nhan,

Gautier and

Berder, Asynchronous MAC protocol for spectrum agility in wireless body area sensor networks, in: IEEE International Conference on Cognitive Radio Oriented Wireless Networks and Communications (CROWNCOM), 2014, pp. 203–208.

29.

Senniappan and

Subramanian, Biogeography-based krill herd algorithm for energy efficient clustering in wireless sensor networks for structural health monitoring application, Journal of Ambient Intelligence and Smart Environments 10(1) (2018), 83–93. doi:10.3233/AIS-170468.

30.

Shaman and

R.A.

Stine, The bias of autoregressive coefficient estimators, Journal of the American Statistical Association 83(403) (1988), 842–848. doi:10.1080/01621459.1988.10478672.

31.

H.W.

Tseng,

Yu-Bin and

Yi, An adaptive channel hopping and dynamic superframe selection scheme with QoS considerations for emergency traffic transmission in IEEE 802.15.6 based wireless body area networks, IEEE Sensors Journal 20(7) (2019), 3914–3929. doi:10.1109/JSEN.2019.2958936.

32.

Wang,

Wei,

Lin and

Wang, A clique base node scheduling method for wireless sensor networks, Journal of Network and Computer Applications 33(4) (2010), 383–396. doi:10.1016/j.jnca.2010.03.002.

33.

Wu,

Hong and

Sheu, Coloring based channel allocation for multiple coexisting wireless body area networks: A game theoretic approach, IEEE Transactions on Mobile Computing 21(1) (2020), 63–75.

34.

Xi,

Han,

Li,

Jiang and

Ding, Location inferring in Internet of things and big data, in: Big Data: Principles and Paradigms, 2016, pp. 309–335.

35.

Xie,

Huang,

Zarei,

Ji,

Ye and

He, A novel nest-based scheduling method for mobile wireless body area networks, Digital Communications and Networks 6(4) (2020), 514–523. doi:10.1016/j.dcan.2020.06.006.

36.

Yi and

Cai, Transmission management of delay-sensitive medical packets in beyond wireless body area networks: A queuing game approach, IEEE Transactions on Mobile Computing 17(9) (2018), 2209–2222. doi:10.1109/TMC.2018.2793198.

37.

Yong,

ke.

Mengya,

Fen and

Qianming Zha, A self adaptive power control algorithm based on game theory for inter-WBAN interference mitigation, in: IEEE International Conference on Computer and Communications (ICCC), 2016, pp. 2873–2877.

38.

Yuan,

Changle,

Qiang,

Kuan,

Nan,

Ning and

Shen, Performance analysis of IEEE 802.15. 6-based coexisting mobile WBANs with prioritized traffic and dynamic interference, IEEE Transactions on Wireless Communications 17(8) (2018), 5637–5652. doi:10.1109/TWC.2018.2848223.

39.

Yuan,

Jiaxin,

Jun,

Kuan,

Changle and

Qiang, 2TM-MAC: A two-tier multi channel interference mitigation mac protocol for coexisting wbans, in: IEEE Global Communications Conference (GLOBECOM), IEEE, 2019, pp. 1–6.

40.

Yuan,

Li,

Zhang,

Ye,

Cheng,

Zhang and

Shen, Dynamic interference analysis of coexisting mobile WBANs for health monitoring, in: IEEE International Conference on Communications (ICC) 2018, pp. 1–6.

41.

Zou,

Liu,

Chen and

C.W.

Chen, Bayesian game based power control scheme for inter-WBAN interference mitigation, in: IEEE Global Communications Conference, 2014, pp. 240–245.

Prediction-based channel assignment for minimizing channel switching in mobile WBANs

Abstract

Keywords

1. Introduction

2. Related works

2.1. Single channel assignment techniques

2.2. Multi-channel assignment techniques

3. Models and problem definition

3.2. Communication model

3.3. Interference

4. Primitives

4.1. Autoregressive model

4.2. Simplified moving average model

5. Proposed channel assignment with reduced channel switching

Table 2 User priority mapping in IEEE 802.15.6 User priority Packet type 1 Background (BK) 2 Best effort (BE) 3 Excellent effort (EE) 4 Video (VI) 5 Voice (VO) 6 Low-priority medical data (e.g., body TMP) 7 High-priority medical data (e.g., body SBP) 8 Emergency medical data (e.g., body ECG)

5.2. Initial channel assignment

A. Analysis of total network delay

B. Analysis of throughput

C. Impact of WBAN priority on total delay and throughput

D. Packet delay analysis for different user priorities

E. Analysis of the number of conflicting channel allocations

F. Impact of moving speed of WBAN

Footnotes

Acknowledgements

Conflict of interest

References