An Adaptive Energy-Efficient Data Collection System for ZigBee Wireless Sensor Networks

Abstract

Wireless sensor networks (WSNs) are deployed in large areas to monitor a number of events in an area of interest. Monitoring environmental events by distributed sensor networks faces the challenge of high power consumption requirement over time, due to the large number of packets required for multihop data collection. To overcome the scalability issue of large scale WSNs, a proof-of-concept implementation demonstrates that integrating a mobile robot (MR) system with a clustering system for ZigBee WSNs will significantly increase the lifetime of the system, by conserving energy that the sensor nodes otherwise would use for communication. In this paper, two energy-efficient systems have been proposed: clustering and MR systems. The former divides the ZigBee WSN into smaller regions, allocates a cluster-head for each region, and aggregates the collected data, whereas the latter collects the sensed data from cluster-heads. The effectiveness of the proposed system has been demonstrated via simulation and experimental studies and verified that, using a single robot for data collection, the lifetime of the network can be extended by 2.3 times in average.

1. Introduction

Wireless sensor networks (WSNs) have emerged as an essential technology for monitoring and exploring remote, hostile, and hazardous environments. WSNs have been deployed widely in several applications including military, medical, safety, and environment monitoring [1–6]. A WSN is a collection of sensor nodes organized into a cooperative network. Sensor nodes are small in size, are low in cost, and have short communication range. Usually, a sensor node consists of four subsystems which are as follows: (i)

a computing subsystem: it is responsible for main functions such as processing the communication protocols and control of onboard sensors;

(ii)

a sensing subsystem: environment characteristics are sensed through a wide range of sensors (temperature, humidity, light, gas, etc.);

(iii)

a communication subsystem: it is a short radio range used to communicate with neighboring nodes;

(iv)

a power supply subsystem: it includes a battery source that feeds computing, onboard sensors, and communication subsystems.

In environment monitoring application, sensor nodes are deployed to sense and collect data from surrounding environment and forward them to a sink node. With sensor networks technology, sensor devices can be deployed close to the phenomenon which must be observed and measurements collected from nodes’ onboard sensors. Since the sensor nodes are physically small, are battery-operated, and contain a tiny wireless radio, deploying such a WSN disturbs the environment minimally and reduces the installation and maintenance costs. Furthermore, the inexpensive nature of these devices attracts scientists to place a high resolution node grid in the field and obtain frequent measurements, providing an extremely rich data set.

In most environmental monitoring applications, sensor nodes collect information from the monitored site and transmit data through energy-expensive multihop wireless communications to a single base station, in which the base station makes the collected data available for offline data analysis. It is unrealistic to deploy end-to-end environmental monitoring WSN over a large geographic area, as a large number of packets required to multihop data collection. Sending high-fidelity sensed data in a multihop manner would prohibitively consume energy of the static sensor nodes in the relaying path which would in turn, over time, cause exhausting sensor nodes. This is because every single packet has to be forwarded multiple times on the routing path to reach the sink node.

In order to address the energy consumption issue, a diverse number of approaches have been proposed. The most widely used sensor network architecture consists of a single sink node and distributed sensor nodes placed over the environment, in which the sink node may be placed at the edge or center of the network and initiates data collection. Each sensor node collects sensed data from the monitored site and transmits data through energy-consuming multihop wireless communications. However, this approach requires high forwarding of sensed data to the nodes in the vicinity of the sink node, which in turn consumes energy more frequently and may reduce the sensor nodes’ lifetime. Another approach includes deploying a mobile robot (MR) for data gathering; robotic assistance reduces the energy consumption on the data collection paths by eliminating the need to transmit large number of sensed data through multihop communication, where this approach offers low power consumption by eliminating the need for multihop data communication; however, an additional effort is required on the design and implementation of MR system; in addition, MR systems are inadequate for some applications.

This paper addresses the area of energy-efficient data gathering in WSNs consisting of sensor nodes deployed in a large area. A new architecture has been proposed to reduce the power consumption, hence extending the ZigBee WSN lifetime, which in turn provides greater network usability. The proposed architecture is based on adopting a new ZigBee based clustering algorithm, and a MR system, in which the sink node serves as an entry point of the ZigBee WSN that enables users of the network to interact with collected data from static sensor nodes. The sink node keeps the data for offline study and analysis in addition to logs of event occurrence. The main contributions of this paper are as follows: (a)

A new ZigBee based clustering algorithm is proposed, in which router nodes in ZigBee network are clustered to minimize the number of exchanged packets.

(b)

A MR system is proposed and implemented; MR collects the sensed data from static nodes, based on two energy threshold systems.

(c)

Unlink the existing approaches which mainly focus on simulation experiments; a new energy-efficient approach is introduced practically using XBee sensor nodes.

The rest of the paper is organized as follows. The relevant works are presented in Section 2. In Section 3, the system model is presented and discussed. Section 4 presents the simulation and experimental test-beds used to test the efficiency of the proposed model, in addition to results obtained from simulation and real experiments. And finally, Section 5 presents a conclusion and future works.

2. Related Work

In this section, the related works in the areas of tiered architecture (clustering) and mobile sink and a combination of these two approaches (tiered architecture and mobile sink) are discussed.

First, the tiered WSNs are designed to facilitate the operation of large sensor systems. In tiered WSNs, sensor nodes are grouped and organized in a hierarchical manner. Nodes at the top layer generally have more resources (i.e., computing power, storage, and processing) and assist the network to perform coordinating functions for data gathering, while nodes at lower layers mainly perform sensing and data forwarding functions. A number of clustering approaches have been designed and implemented recently, driven by the need to achieve low power consumption for tiny sensor nodes with the best delivery rate. A various number of clustering algorithms are summarized in [7, 8].

In this paper, clustering through ZigBee network is considered, when the three roles exist (coordinator, router, and end-device). The implementation of clustering algorithms through ZigBee network has received less attention. In [9], authors focused on the performance of a clustered ZigBee WSN with data fusions. Authors of [10] developed a clustering algorithm for ZigBee WSN, where a hybrid routing algorithm based on weighted clustering for load-balancing network energy was proposed. The work presented in [11] includes a clustering method for ZigBee sensor nodes, which performs wide range data transferring depending on the signal strength of sensor nodes, to transfer data reliably to the sink node.

Second, the mobile sink (mobile data collector) approach is considered. This approach includes the use of mobile elements for data collection. Mobile elements have been deployed in a number of WSN applications, to enhance the coverage and operation of the network. A mobile element may move through the distributed sensor nodes in the area of interest and collect the sensed data from stationary sensor nodes instead of forwarding the collected data to a sink node. A survey on sink mobility models in WSNs is presented in [12].

In [13], authors proposed a two-tiered multihop hybrid WSN architecture, where static sensor nodes collect high-fidelity data and robots act as mobile data collectors traversing the area of interest. This architecture reduces the amount of energy consumed from wireless communications. In [14], the authors explored cooperation among MRs and WSNs in environmental monitoring in which robotic data mules collect measurements gathered by sensor nodes.

A novel sink mobility model derived from De Bruijn graph was proposed in [15]; the proposed model combines the use of single hop and multihop communication to collect data from static sensor nodes. The performance was studied in terms of end-to-end delivery and data success rate. In [16] authors investigated the effect of using mobile sinks for data gathering through WSNs. Authors first studied the improvement in network lifetime, and then a distributed and localized solution was proposed to decide sink's movement when the movement path is not predetermined. Authors of [17] proposed a new framework for using a mobile sink to improve the network lifetime and its effectiveness in applications that can tolerate a certain amount of delay in data delivery.

Third, the integration of the aforementioned schemes (cluster and MR) is considered, in which clustering is required at the first stage, and then a MR is deployed to collect the observed data from subsink nodes. A novel data collection algorithm using MR was proposed in [18], in which two control approaches were proposed for identifying the locations of partitioned WSNs: global-based and local-based. On the other hand, the work presented in [19] includes a framework consisting of transmit-only sensors which are low in cost and require less energy than regular sensors due to the absence of receiver circuit and a mobile element for data collection. However, transmit-only sensors cannot receive or forward data, and therefore they are inadequate in homogenous networks.

As noticed above, a few number of clustering systems have been implemented through ZigBee protocol and most of the existing works focused on simulation studies. In this paper, a new energy-efficient gathering system for ZigBee WSN is proposed and tested through both simulation and real experimental studies.

3. Energy-Efficient Data Collection Architecture

In this section, the communication protocol deployed in the real experiments is discussed, and then the proposed system is overviewed.

3.1. ZigBee Communication Protocol

ZigBee is a low power, low data rate, and low cost wireless communication standard and aims to be used in home automation, remote control, and sensor applications. ZigBee network standard performs three main roles: coordinator, router, and end-device. A single coordinator is required for each ZigBee network; it has a unique Personal Area Network Identification (PAN ID) and a channel number. ZigBee coordinator initiates the network formation and may act as a router once a network is formed. ZigBee network may have more than one router, in which router may associate with ZigBee coordinator or/and with other ZigBee routers. ZigBee router associates in multihop routing of messages, and, finally, ZigBee end-device is an optional network component which is utilized for low power operations and does not allow association nor participation in routing [20]. Figure 1 presents the architecture of a ZigBee network and presents the three roles discussed above.

Figure 1

An example of a ZigBee WSN with three roles presented.

3.2. System Overview

The architecture of the system under consideration is presented in Figure 2. The root denoted that the sink is the base station of the WSN and is the coordinator of the ZigBee network. Each node (router and end-device) excluding the root is a sensor node deployed in the field of interest. The locations of sensor nodes are predetermined using Global Positioning System (GPS) or other location aware systems. Router nodes have a large memory size and energy compared to end-device nodes in order to collect and process sensed data. In this architecture, ZigBee WSN is divided into small groups (clusters), in which each cluster consists of a number of sensor nodes, and each sensor node has the ability to detect the desired information from the surrounding environment and forward the collected data to a cluster-head (CH).

Figure 2

The partitioned ZigBee WSN.

The system definition is given as an illustration as presented in Figure 2. Nodes $\{B, C, D, \dots, O, P\}$ are router nodes, whereas node A is denoted as a drain router, since all messages may pass through router node A to reach the sink node, which results in four portioned WSNs as follows: $P_{1} = \{B, C, D, E\}$ , $P_{2} = \{F, G, H, I\}$ , $P_{3} = \{J, K, L\}$ , and $P_{4} = \{M, N, O, P\}$ . Each router node will become a CH for t period of time, where each CH collects the sensed data from its child nodes and aggregates and transmits the aggregated data to the sink node.

In normal situations, as soon as the network starts, each router node collects the sensed data from its end-device nodes and forwards them using multihop communication to a sink node. Through experiments discussed later in this paper, it was demonstrated that router nodes located close to the sink node will drain their energy fast, compared to the router nodes placed away from a sink node. In the above example, router node $A$ will drain first, since all sensed data will pass through router node A to reach the sink node. For this reason, router node A is considered as a drain router. After that, all other router nodes will become drain routers since they participate in forwarding the sensed data to the sink node and cannot enter a sleep mode.

As a result, a new clustering algorithm for ZigBee WSN is proposed in this paper, in which the sensor nodes are distributed into clusters; a CH collects the sensed data from sensor nodes in its range and then forwards them to a sink node. This approach will minimize the power consumption for sensor nodes placed away from a sink node; however, the proposed clustering system does not trim down the energy consumption for sensor nodes placed close to the sink, because they participate in forwarding and routing sensed data to the sink node. For that reason, to minimize the power consumption for such nodes, an additional energy saving approach may be adopted, in which the sensed data may be gathered using MR system in order to reduce the power consumption for the sensor nodes located close to the sink node.

3.3. Clustering Approach

Through clustering approach, sensor nodes are divided into a number of clusters, in which each cluster consists of a number of router nodes (2–4), where each router node may have a number of end-device nodes connecting to itself in one hop communication. A router node collects the sensed data from its end-device nodes and aggregates and transmits it to a CH. The CH collects the sensed data from other routers in its cluster and further aggregates the collected data and then forwards them to a sink node. The clustering approach consists of four main phases as follows: (1)

Initialization: this includes dividing the WSN into groups (usually, each group consists of 2–4 routers and a number of end-device nodes).

(2)

Selection of a CH: in a cluster c, router node i which poses the shortest distance to sink, the large residual energy, and the large number of end-devices connecting to i in single hop communication will be elected to play the role of CH for a cluster c, for t period of time.

(3)

Data collection: CH will request the sensed data from the sensor nodes in its cluster and aggregate the collected data.

(4)

Transmission to sink node: the aggregated sensed data will be transmitted from a CH to the sink node through multihop communication.

The initialization phase is quite simple, since each group of router nodes placed close to each other will consist a single cluster. This phase is repeated till each router in the ZigBee network is assigned to a single cluster. Algorithm 1 presents the division process.

Algorithm 1: Dividing the WSN into clusters.

(1) let M be the nodes in the ZigBee network $M > 0$

(2) let R be the router nodes $0 < R < M$

(3) let sink be the sink node

(4) let (0, 0) be the start point

(5) let $r_{j}$ be the nearest router node to the start point

(6) $r_{j}$ checks router nodes in its vicinity and groups them in a single cluster

(7) $r_{j}$ sends cluster information to the sink node (location, # of nodes, nodes’ IDs)

(8) system checks the first router node set close to the start point and has not been assigned to any cluster

(9) repeat (6–8) till each router node belongs to a single cluster

(10) end

In the second phase, the selection process has taken place, in which a CH is elected based on three factors, as follows: (1)

Residual energy ( $R_{e_{i}}$ ): the sensor node i with a large residual energy has a high probability to be a CH.

(2)

Number of hops (h) between the router and the sink node: the router node with the least number of hops to the sink node will be given reasonable probability to be a CH.

(3)

Total number of end-device nodes (T) connecting to the router: the router node with a large number of child nodes (end-devices) has a reasonable probability to be a CH.

In order to enhance the selection process, a weight value is assigned to each factor, which presents its significance. The weight values were carefully calculated through real experiments, to obtain the impact of each factor on the power consumption.

The number of hops (h) has a significant impact on power consumption. Increasing the number of hops between CH and coordinator node will significantly increase the power consumption required to transmit a single packet. For evaluation purposes, a number of real experiments were conducted to assess the impact of the number of hops on the power consumption. A number of 30 packets (size of 22 bytes) were transmitted from a router node to a coordinator using different number of hops (1, 2, and 3) asynchronously. In each experiment, the power consumption was estimated for the whole network. Figure 3 presents the total energy required through different number of hops.

Figure 3

Estimating the power consumption for transmitting 30 packets through (1, 2, and 3) hops.

On the other hand, the weight value for the total number of end-device nodes (T) presents the lowest impact on the power consumption. This includes the total number of end-device nodes connecting to a router node $r_{i}$ , since each end-device node requires two bytes to be added to the router packet (the aggregated packet required to be transmitted to the sink node); that is, when a router node $r_{i}$ is connected to a total number of p end-device nodes, then router node $r_{i}$ requires $E_{s}$ energy to transmit this packet. Therefore, if router node $r_{i}$ is connected to $p + 1$ end-device nodes, then router node $r_{i}$ will require $E_{s} + E_{u}$ energy to transmit this packet to the coordinator, where $E_{u}$ is the extra energy needed to transmit an additional 2 bytes. Hence, the router node with the maximum number of end-device nodes will be given reasonable priority to be a CH. Figure 4 presents the energy required to transmit and receive 30 packets with different sizes (22, 24, 26, and 28 bytes).

Figure 4

Estimating the power consumption for transmitting 30 packets with different packets sizes (22, 24, 26, and 28) bytes.

In order to balance energy in all router nodes, a weight value for residual energy ( $R_{e_{i}}$ ) was given the highest value $= 0.6$ , due to its significance in balancing the remaining energy for the whole WSN. As presented in Figure 3, the average increment ratio in power consumption when increasing the number of bytes is $= 0.125 \approx 0.1$ , whereas in Figure 4 the average increment ratio in power consumption when increasing the number of hops is $= 0.311 \approx 0.3$ . Table 1 presents the weight values for the energy ( $w_{E}$ ), number of hops ( $w_{H}$ ), and number of end-devices ( $w_{D}$ ).

Table 1

Weight values for the three factors.

Weight of factor	Value
$w_{E}$	0.6
$w_{H}$	0.3
$w_{D}$	0.1
	1.00

The energy (E), number of hops (H), and number of end-device nodes (D) parameters are estimated through the following, respectively:

\begin{matrix} E = \frac{R_{e_{i}}}{100 (Full)} \times w_{E}, \end{matrix}

(1)

\begin{matrix} H = \frac{1}{h} \times w_{H}, \end{matrix}

(2)

\begin{matrix} D = T \times w_{D} . \end{matrix}

(3)

Each router node estimates its own selection value (

S_{i}

) (4) and transmits it to the CH. Based on the received S values

(S_{1}, S_{2}, \dots ., S_{k})

, where k is the number of router nodes in a cluster c, the existing CH selects the new CH for the next round:

\begin{matrix} S = E + H + D . \end{matrix}

(4)

In the data collection phase, each CH requests the sensed data from its child nodes (end-devices and routers). Algorithm 2 presents the process of CH data collection. After a time t, a new CH will be elected based on comparing the obtained S values from routers in that cluster. Algorithm 3 shows the process of selection of a new CH.

Algorithm 2: Data collection through CHs.

(1) let M be the nodes in the ZigBee network $M > 0$

(2) let R be the router nodes $0 < R < M$

(3) let $R_{k}$ be the router nodes in a cluster k

(4) let $r_{k_{i}}$ be the CH for a cluster k

(5) let sensed data be the information detected by a sensor node

(6) let sink be the sink node

(7) for every router node (lets start with $r_{k_{j}}$ ) in a cluster k

(8) $r_{k_{j}}$ collects the sensed data from its end-device nodes

(9) $r_{k_{j}}$ aggregates the collected sensed data

(10) $r_{k_{j}}$ transmits its aggregated data to a CH (i.e. $r_{k_{i}}$ )

(11) end for

(12) $r_{k_{i}}$ transmits the aggregated data to the sink

(13) end

Algorithm 3: The selection process for the next CH.

(1) let M be the nodes in the ZigBee network $M > 0$

(2) let R be the router nodes $0 < R < M$

(3) let $R_{k}$ be all the router nodes in a cluster k

(4) let $r_{k_{i}}$ be the CH in a cluster k

(5) $r_{k_{i}}$ collects the S values from each router node in a cluster k

(6) $r_{k_{i}}$ compares the received values from all routers in a cluster k

(7) the router node (for instance $r_{k_{j}}$ ) in a cluster k with the best S value will be elected as a CH

(8) $r_{k_{j}}$ will transmit the information of a new CH to all routers in the cluster k

(9) end

3.4. Mobile Robot System

Clustering methods may reduce the power consumption for router nodes in the ZigBee network; however, router nodes located close to the sink node will drain their energy first compared to the router nodes located away from the sink node. In order to address this issue, two threshold-energy systems have been proposed and implemented, in which each CH has two thresholds defined in advance: the values of the remaining energy of near-drain $E_{n d}$ and drain $E_{d}$ . These two thresholds must meet the condition of $0 < E_{d} < E_{n d} < E$ , where E is the total energy of a router node. The status of a router node r is based on the remaining energy $E_{r}$ ; whenever any router node exceeds the threshold energy predetermined formerly, it will broadcast a status message about its energy level and the number of child nodes connecting to itself. Sink node will inform MR to navigate the location of the drain router node and its child nodes. The energy thresholds categorize the sensor nodes into different roles, which are defined as follows: (i)

$n_{n d}$ (near-drain node): when a CH (for instance $r_{k_{i}}$ ) becomes a near-drain node, it stops forwarding the sensed data (from child nodes and other router nodes in its range) to the next hop, and Ndrain packet is triggered. $r_{k_{i}}$ will wait for $M R$ to visit.

(ii)

$n_{d}$ (drain node): when a CH $r_{k_{i}}$ becomes a drain node, then $r_{k_{i}}$ will stop forwarding the sensed data from itself, its child nodes, and other router nodes in its range, to the next hop, and a drain packet is triggered. $r_{k_{i}}$ will wait for MR to visit.

(iii)

$s_{n d}$ (the child nodes of $n_{n d}$ ): $s_{n d}$ forwards the sensed data to its parent (e.g., $r_{k_{i}}$ ) and stops forwarding the sensed data to the sink node.

(iv)

$s_{d}$ (the child nodes of $n_{d}$ ): $s_{d}$ will stop forwarding sensed data including control packets to their CH (e.g., $r_{k_{i}}$ ); instead they store the sensed data locally and wait for MR to visit.

To discuss the energy thresholds more precisely, in the first case when a CH (e.g.,

r_{k_{i}}

) becomes a near-drain node, then

r_{k_{i}}

will send a Ndrain packet along with its location

(x_{r_{k_{i}}}, y_{r_{k_{i}}})

to the sink node, where

x_{r_{k_{i}}}

and

y_{r_{k_{i}}}

are the x-coordinate and y-coordinate for local-sink

r_{k_{i}}

, respectively. The sink node sends this information to MR in order to schedule visiting

r_{k_{i}}

, collects the sensed data, and brings it back to the sink node. Algorithm 4 illustrates the data collection assisted by a MR through energy threshold 1.

Algorithm 4: Data collection using MR through energy threshold 1.

(1) let M be the nodes in the ZigBee network $M > 0$

(2) let R be the router nodes $0 < R < M$

(3) let $E_{i}$ be the energy of router node i

(4) let sensed data be the information detected by a sensor node

(5) let sink be the sink node

(6) let MR be a mobile robot

(7) M detects ambient environment

(8) M transmits sensed data to local-sink (for instance $r_{k_{i}}$ )

(9) if a CH ( $r_{k_{i}}$ for instance) exceeds the pre-assigned energy threshold 1

(10) if partitioned WSNs exist

(11) MR obtains information of partitioned WSN from the sink node

(12) $r_{k_{i}}$ collects and aggregates the sensed data from its child-nodes

(13) $r_{k_{i}}$ stops forwarding the received packets to the sink node

(14) $r_{k_{i}}$ waits for MR to visit

(15) MR collects sensed data from near-drain router node $r_{k_{i}}$

(16) MR brings these data back to the sink node

(17) end if

(18) end if

(19) end

In the second case, when any near-drain CH (e.g., $r_{k_{i}}$ ) becomes a drain CH, then $r_{k_{i}}$ will stop receiving the sensed data from its child nodes and transmits a drain packet to the sink node along with the number of its child nodes ( $s_{d}$ ) and their locations $\{(x_{e_{1}}, y_{e_{1}}), (x_{e_{2}}, y_{e_{2}}), \dots, (x_{e_{n}}, y_{e_{n}})\}$ , where $(x_{e_{a}}, y_{e_{a}})$ and n are the x-coordinate and y-coordinate for a child node a and the total number of child nodes connecting to CH $r_{k_{i}}$ , respectively. Sink node will send this information to MR, to navigate that CH and visit its child nodes and collect the sensed data. Algorithm 5 illustrates the data collection approach using MR through energy threshold 2.

Algorithm 5: Data collection using MR through energy threshold 2.

(1) let M be the nodes in the ZigBee network $M > 0$

(2) let R be the router nodes $0 < R < M$

(3) let N be the child-nodes $0 < N < M$

(4) let $s_{d}$ be the child nodes for a local sink $r_{k}$

(5) let $E_{i}$ be the energy for node i

(6) let $r_{k_{i}}$ be a drain local-sink

(7) let sensed data be the information detected by a sensor node

(8) let sink be the sink node

(9) let MR be a mobile robot

(10) if any local-sink (lets say $r_{k_{i}}$ ) exceeds energy threshold 2

(11) $s_{d}$ connecting to local-sink $r_{k_{i}}$ will stop forwarding sensed data

(12) $s_{d}$ waits for MR to visit and collect sensed data

(13) As soon as MR visits $s_{d}$ , $s_{d}$ will transmit the sensed data to MR

(14) MR moves to a sink node and offloads the collected data

(15) end if

(16) end

4. Experimental Results

In order to test the efficiency of the proposed system in this paper, the experimental test-bed was carried out through two stages. The first stage involves testing the proposed approach using NS2 network simulator, whereas in the second stage real experiments were conducted using XBee modules.

4.1. Simulation Experiments

A number of simulation experiments were carried out to evaluate the proposed system. For evaluation purposes, relevant parameters have been collected from actual sensor devices and inserted into simulation runs. The simulation environment of the area of the geographical region is 140 m × 160 m as presented in Figure 5, in which it consists of 71 nodes distributed as follows: 1 coordinator, 12 routers, and 58 end-device nodes. Table 2 presents the simulation parameters.

Table 2

Simulation test-bed parameters.

Field size	140 m × 160 m
Total nodes	71
Average number of hops	3
Experiment time	40 minutes
Communication protocol	IEEE 802.15.4
Routing protocol	AODV
Sampling rate	120 seconds
$E_{nd}$	400 mA
$E_{d}$	300 mA

Figure 5

Simulation test-bed.

4.2. Simulation Results

In order to evaluate the efficiency of the proposed system, four main scenarios were considered, as follows: (1) end-to-end scenario, in which each sensor node transmits its sensed data directly to a sink node; (2) cluster-based scenario where each local-sink (CH) collects and aggregates the sensed data from its child nodes and transmits it to a sink node; (3) MR scenario a, in which MR travels along the sensor field and collects the sensed data from local-sinks (CHs); and (4) MR scenario b, in which MR collects the sensed data directly from child nodes (router nodes) of a drain local-sink (CH).

For the above four scenarios, the total number of transmitted packets, the power consumption for CHs, and the delivery rate are estimated. The total numbers of messages transmitted in end-to-end, cluster, and MR systems a and b are represented in the following, respectively:

\begin{matrix} n_{sink-to-sink} = (r_{chlds} + r_{rtrs}) \times h \times \frac{t}{f}, \end{matrix}

(5)

\begin{matrix} n_{cluster} = r_{chlds} \times \frac{t \times h}{4 f}, \end{matrix}

(6)

\begin{matrix} n_{MR_a} = r_{chlds} \times \frac{t}{f}, \end{matrix}

(7)

\begin{matrix} n_{MR_b} = r_{rtrs} \times \frac{t}{f}, \end{matrix}

(8)

where

r_{chlds}

, t, f,

r_{rtrs}

, and h are the number of child nodes, the total experiment run time, the sampling rate, the number of routers, and average number of hops, respectively.

The total number of packets transmitted in the whole network is evaluated in Figure 6, in which the total number of packets transmitted by local-sinks in Scenario 1 is the highest, since every single packet transmitted by child nodes must be forwarded to the sink node via multihop communication. However, in Scenario 2, the number of transmitted packets is less than that in Scenario 1, as the received packets by each CH must be grouped and aggregated before being transmitted to a sink node.

Figure 6

The total number of packets transmitted in the ZigBee WSN for the four scenarios.

In MR scenario a, the total number of packets is less than that in Scenarios 1 and 2, since MR has to collect the sensed data from each local-sink (CH) and brings it back to the sink node, whenever the selected CH reaches the energy threshold 1. In MR scenario b, as soon as the elected CH reaches the energy threshold 2, then MR will manage to visit that CH and its child nodes. However, this only occurs when all CHs in a cluster reach the energy threshold 2.

According to the obtained simulation results in Scenarios 1 and 2, local-sinks (CHs) drain their energy first, and therefore the WSN will be inaccessible shortly. However, in Scenarios 3 and 4 local-sinks will last for long time compared to Scenarios 1 and 2. This is because, in Scenario 3, local-sinks participate only in gathering the sensed data from child nodes and then transmit the sensed data to MR, whereas, in Scenario 4, local-sinks may only transmit the sensed data to MR.

The total number of packets transmitted by each local-sink through the four scenarios is presented in Figure 7. Local-sinks ( $R_{9}$ , $R_{10}$ , $R_{11}$ , and $R_{12}$ ) placed close to the sink node consumed the highest amount of energy through Scenarios 1 and 2, as they participate in routing message from other local-sinks to sink node, as presented in Figure 8. However, when Scenarios 3 and 4 were adopted, the power consumption was reduced by more than a half, therefore improving the lifetime of the ZigBee WSN.

Figure 7

Total number of packets transmitted from local-sink in four scenarios.

Figure 8

The remaining energy for local-sinks (initial energy = 500 mA).

On the other hand, the delivery rate was evaluated for the above four scenarios, in order to assure the feasibility of the proposed system. Figure 9 presents the delivery rate for the aforementioned four scenarios. End-to-end and cluster scenarios offer the worst delivery rate, since a large number of packets are required to be exchanged in the ZigBee WSN. However, when MR and clustering systems are adopted, the dropping rate will be less, and hence the delivery rate will be enhanced.

Figure 9

Delivery rate through four scenarios.

4.3. Real Experiments

In this section, a proof-of-concept and results obtained from a small experiment are presented, which proves the feasibility of using a clustering algorithm and a MR as data collector in ZigBee sensor field.

4.3.1. Static and Mobile Robot Models

Through real experiments, XBee series 2 module has been employed for static nodes. XBee module offers cost-effective wireless connectivity in ZigBee mesh networks, and it allows for a very reliable and simple communication between microcontrollers, computers, and systems.

The proposed system targets the environmental monitoring applications, in which temperature, humidity, carbon monoxide, and light values are significant and can be used to determine hazards in any environment [21]. The experiment test-bed includes the design and implementation for two nodes: (1) end-device nodes where each end-device node collects the sensed from onboard sensors (temperature, humidity, and carbon monoxide) as presented in Figure 10, and (2) router nodes where each router node gathers sensed data from its child nodes (end-devices and routers) in its cluster and has the ability of processing, aggregating, and electing a CH for the next round. Figure 11 presents the implemented router node.

Figure 10

End-device node.

Figure 11

Router node platform.

For the data collection function, a MR system was designed and implemented. The proposed robotic prototype is based on Arduino Uno board which can navigate the terrain and collect the sensed data from static nodes distributed over the area of interest. Figure 12 presents the designed prototype robotic system. XBee router node is attached to the robotic system in order to allow communication between robotic system and static sensor nodes. The designed robotic system has information about the static sensor nodes’ locations, in order to schedule visits to near-drain nodes, drain nodes, and child nodes.

Figure 12

Mobile robot model.

4.3.2. Nodes Architecture

A total number of 12 nodes were deployed in the experiment test-bed: a single coordinator, 7 router nodes, and 4 end-device nodes, as depicted in Figure 13. A single MR system was deployed to gather the observation collected by stationary sensor nodes.

Figure 13

Real experiment architecture.

In order to obtain the sensed data frequently from static sensor nodes in environment monitoring applications, an efficient sampling rate must be configured. According to [22] the optimal sampling rate ( $S R$ ) in environment monitoring applications is 120 seconds; therefore, in the experiment test-bed, the $S R$ was set to 120 seconds. Table 3 presents the real experiment parameters.

Table 3

Experimental test-bed parameters.

Number of nodes	12
Average number of hops	2
Experiment time	20 minutes
Sampling rate	120 seconds
Communication protocol	ZigBee
Tx energy	0.46 mA
Rx energy	0.47 mA
Transmission range	Very low (about 7 meters)
Carbon monoxide (MQ-7)	<900 mW
TMP 36 sensor	10 mV
$E_{nd}$	2800 mV
$E_{d}$	2600 mV

4.4. Real Experiments Results

In this section, the total number of packets transmitted in the four scenarios is evaluated and discussed. The following assumptions have been made. First, each end-device node can communicate only with its direction neighbors on the grid. Second, each packet can be lost with probability p.

Given these assumptions, the sum of the expected number of transmissions (ETX) is estimated, which are required to enable each of the 11 nodes to deliver a single packet to the sink node successfully. Through 20 minutes experiment time, the ETX for Scenario 1 was equal to 226 and $p = 0$ . In Scenario 2, the ETX value was equal to 131 and $p = 0$ , because each child node transmits its sensed data to its local-sink, where the sensed data is aggregated and retransmitted to a sink node. The ETX value in MR scenario a was equal to 67 and $p = 0$ , as each local-sink gathers the sensed data from its child nodes and waits for a MR to visit; as soon as a MR visits the local-sink, then the local-sink offloads the collected data. And finally, in MR scenario b, the ETX value was equal to 42 and $p = 0$ , since MR has to collect the sensed data directly from each child and local-sink nodes in the WSN field.

Figure 14 presents the total number of packets transmitted in four scenarios through 20 minutes experiment time, and Figure 15 shows the total number of packets transmitted by each local-sink ( $R_{1}$ , $R_{2}$ , $R_{3}$ , $R_{4}$ , $R_{5}$ , $R_{6}$ , and $R_{7}$ ), throughout the four scenarios. Figure 16 presents the power consumption for the entire WSN. As presented, the power consumption for the MR threshold 2 was the minimum.

Figure 14

The total number of packets transmitted in four scenarios.

Figure 15

Total number of packets transmitted by each local-sink through 4 scenarios.

Figure 16

The average power consumption in 4 scenarios for the entire WSN.

On the other hand, the time required for a MR to collect sensed data from a router node and bring data back to sink nodes was estimated. As presented in Figure 17, the average access time is presented for each router node. The average access time will increase when the proposed system is deployed in large environment, and hence integrating multiple MRs will solve the problem.

Figure 17

Average access time required to collect sensed data using MR.

5. Conclusion and Future Work

In WSN data collection, sensor nodes placed close to the sink node tend to consume more energy than those far away from the sink. This is because, besides transmitting their own packets, they forward packets of other nodes to the sink node. However, WSN lifetime can be significantly improved if the energy consumed in the transmission is reduced.

In this paper, an energy-efficient data collection system is proposed through adopting two systems: a clustering and a MR system. The implementation of the proposed energy-efficient system offers three main advantages as follows: saving energy, minimizing data collision, and reducing computational load on the router and sink nodes. Throughout experiments, the power consumption rate was improved by about 33% through adopting energy threshold system 1 and was improved by 56% through applying energy threshold system 2. For future work, the proposed system is aimed at being deployed in large field with an extra number of nodes and then testing its efficiency in terms of power consumption and dropping rate, in addition to studying the efficiency of adopting multiple MRs for data gathering.

Footnotes

Conflict of Interests

The author declares that there is no conflict of interests regarding the publication of this paper.

References

Pasca

Rosu

A new approach for healthcare monitoring

Proceedings of the E-Health and Bioengineering Conference (EHB ′13)

November 2013

Iași, Romania

IEEE

1 6

10.1109/EHB.2013.6707293

Sghaier

Mellouk

Augustin

Amirat

Marty

Khoussa

M. E. A.

Abid

Zitouni

Wireless sensor networks for medical care services

Proceedings of the 7th International Wireless Communications and Mobile Computing Conference (IWCMC ′11)

July 2011

Istanbul, Turkey

IEEE

571 576

10.1109/iwcmc.2011.5982596

2-s2.0-80052444151

Alippi

Camplani

Galperti

Roveri

A robust, adaptive, solar-powered WSN framework for aquatic environmental monitoring

IEEE Sensors Journal 2011 11 1 45 55

10.1109/JSEN.2010.2051539

2-s2.0-78149270870

Alhmiedat

Omar

Abu Taleb

Alsswey

Road safety and energy saving proposed system: a ZigBee WSN approach

International Journal of Online Engineering 2015 11 2

Navarro

Davis

T. W.

Liang

ASWP: a long-term WSN deployment for environmental monitoring

Proceedings of the 12th International Conference on Information Processing in Sensor Networks (IPSN ′2013)

April 2013

ACM

351 352

10.1145/2461381.2461443

2-s2.0-84876749096

Louise

Toulgoat

Deziel

Patterson

Tiered wireless sensor network architecture for military surveillance applications

Proceedings of the 5th International Conference on Sensor Technologies and Applications

2011

Nice, France

288 294

Olutayo

Takizawa

A survey on clustering algorithms for wireless sensor networks

International Journal of Space-Based and Situated Computing 2011 1 2 130 136

Abbasi

A. A.

Younis

A survey on clustering algorithms for wireless sensor networks

Computer Communications 2007 30 14-15 2826 2841

10.1016/j.comcom.2007.05.024

2-s2.0-34548850872

Razi

Afghah

Abedi

Binary source estimation using a two-tiered wireless sensor network

IEEE Communications Letters 2011 15 4 449 451

10.1109/lcomm.2011.020311.102378

2-s2.0-79954632834

10.

Yan

Yuan

Liu

Huang

Cui

A weight cluster-based hybrid routing algorithm of ZigBee network

International Journal of Future Generation Communication and Networking 2014 7 2

11.

Woo

S.-J.

Shin

B.-D.

Efficient cluster organization method of zigbee nodes

International Journal of Smart Home 2013 7 3 45 56

2-s2.0-84878658428

12.

Abu Taleb

Alhmiedat

Hassan

Turab

A survey of sink mobility models for wireless sensor networks

Journal of Emerging Trends in Computing and Information Sciences 2013 4 9

13.

Tinta

S. P.

Zhou

Wong

J. L.

Robot-assisted energy-efficient data collection from high-fidelity sensor networks

Proceedings of the IEEE International Conference on Technologies for Practical Robot Applications (TePRA ′09)

November 2009

Woburn, Mass, USA

101 106

10.1109/tepra.2009.5339637

2-s2.0-71849104650

14.

Tekdas

Isler

Lim

J. H.

Terzis

Using mobile robots to harvest data from sensor fields

IEEE Wireless Communications 2009 16 1 22 28

10.1109/mwc.2009.4804365

2-s2.0-63449111312

15.

A. Abu Taleb Alhmiedat

T. A.

R. Abu Taleb O. Al-Haj Hassan

Sink mobility model for wireless sensor networks

Arabian Journal for Science and Engineering 2014 39 3 1775 1784

10.1007/s13369-013-0811-y

16.

Marta

Cardei

Improved sensor network lifetime with multiple mobile sinks

Pervasive and Mobile Computing 2009 5 5 542 555

10.1016/j.pmcj.2009.01.001

2-s2.0-70349771047

17.

Yun

Xia

Maximizing the lifetime of wireless sensor networks with mobile sink in delay-tolerant applications

IEEE Transactions on Mobile Computing 2010 9 9 1308 1318

10.1109/tmc.2010.76

2-s2.0-77954852265

18.

Chen

T.-C.

Chen

T.-S.

P.-W.

On data collection using mobile robot in wireless sensor networks

IEEE Transactions on Systems, Man, and Cybernetics Part A: Systems and Humans 2011 41 6 1213 1224

10.1109/tsmca.2011.2157132

2-s2.0-80054820990

19.

Baris

Tosun

Data collection using transmit-only sensors and a mobile robot in wireless sensor networks

Proceedings of the 21st International Conference on Computer Communications and Networks (ICCCN ′12)

August 2012

IEEE

10.1109/icccn.2012.6289264

2-s2.0-84867776390

20.

Smith

S. L.

Schwager

Rus

Persistent robotic tasks: monitoring and sweeping in changing environments

IEEE Transactions on Robotics 2012 28 2 410 426

10.1109/tro.2011.2174493

2-s2.0-84859746737

21.

Mainwaring

Polastre

Szewczyk

Culler

Anderson

Wireless sensor networks for habitat monitoring

Proceedings of the 1st ACM International Workshop on Wireless Sensor Networks and Applications

September 2002

Atlanta, Ga, USA

88 97

10.1145/570738.570751

2-s2.0-0036986465

22.

Ingelrest

Barrenetxea

Schaefer

Vetterli

Couach

Parlange

SensorScope: application-specific sensor network for environmental monitoring

ACM Transactions on Sensor Networks 2010 6 2, article 17

10.1145/1689239.1689247

2-s2.0-71149086750