Keywords

1 Introduction

Recent technological advances have led to the development of wireless sensor-actuator networks (WSANs) which are capable of observing the physical word, making decisions based on the observations and performing appropriate actions. WSAN is an extension of wireless sensor networks (WSN) [1]. The components of WSAN include sensor nodes, actuator nodes and sink nodes [2]. The sensor nodes sense the events in the physical world and forward the data to actuator nodes or sink nodes. The actuator nodes are responsible to take decisions and react in the event area according to the received data. The function of the sink nodes is to instruct the actuator nodes to perform related operations. WSAN has many merits in applications [3]. For example, wireless sensor nodes are not restricted by cable, which are portable and have low requirements for external environment and facilities [4]. What’s more, low costs, good fault tolerance, good flexibility and scalability are also the reasons it can be widely used.

However, WSAN is also facing many new challenges [5], such as real-time requirement, limited energy, redundant nodes and security problems. One of the main challenges of WSAN is to improve the real-time performance of the network. The real-time performance of the network directly determines the performance of the system. This paper focuses on real-time routing protocol of WSAN.

In this paper, we propose an improved WSAN routing protocol based on SFL-ACA. Our protocol considers the mobile actuator node which may reduce communication delay. We study the influence of different number of actuator nodes to the performance of WSANs, which produces an effective number of actuator nodes configuration scheme.

The remainder of the paper is organized as follows: Sect. 2 introduces the model assumption and clustering algorithm; Sect. 3 further presents the improved WSAN routing protocol based on SFL-ACA; Sect. 4 gives the experimental results and analysis and Sect. 5 finally concludes this paper.

2 Network Model and Problem Definition

2.1 Network Model

Assume that WSAN is a \( Y \times Y \) square event area and the number of sensor nodes and actuator nodes is N and M respectively. The properties of the network are as follows:

  1. (1)

    Each sensor node has a probability \( W_{i} \) (range from 0 to 1), determined by the frequency of the occurrence of the event area.

  2. (2)

    If a sensor node senses an event, it would report the incident to the actuator node in one hop, otherwise it would not report.

  3. (3)

    The actuator nodes are removable. The actuator nodes perform corresponding tasks according to the information uploaded by the sensor nodes.

Each actuator node has its own mobile path and it does not interfere with each other. When the actuator node is close to the sensor node, the message is sent to ask if the event is detected. And if the sensor node detects an event, the data would be uploaded to the actuator node.

2.2 Network Nodes Clustering

In a WSAN, actuator nodes start from a random point, then move to the corresponding node according to the node sequence on its own mobile path, and finally return to the starting point. We assume that mobile path of actuator nodes is a cluster. The actuator nodes are the cluster head node and the sensor nodes passing through the path are cluster members.\( s_{i} \) represents the \( i{\text{ th}} \) sensor node and the set of sensor nodes is \( Sensor = \{ s_{1} ,s_{2} , \ldots ,s_{N} \} \) \( ma_{j} \) stands for the \( j{\text{ th}} \) actuator node and the set of actuator nodes is \( Actuator = \{ ma_{1} ,ma_{2} , \ldots ,ma_{M} \} \) \( M\_P_{j} \) represents the movement path of \( ma_{j} \).The steps of the clustering algorithm are as follows:

Step1: Calculate the node weight \( K \) of the sensor nodes.

$$ K_{i} \approx W_{i} \times M $$

Step2: Assign cluster members.

  1. (1)

    Select the sensor nodes with the same node weight \( K \) to form the set \( S_{K} \), \( K = 1,2, \ldots ,M \);

  2. (2)

    The sensor nodes in \( S_{M} \) are allocated to all paths \( M\_P_{j} \), \( j = 1,2, \ldots ,M \)

  3. (3)

    The sensor nodes in \( S_{M - 1} \) are allocated so that they belong to any \( M - 1 \) path \( M\_P_{j} \), \( j = 1,2, \ldots ,M \). The current path is denoted by \( |M\_P_{j} | \) in the allocation. We select the minimum number of \( |M\_P_{j} | \) to ensure that the number of nodes in each path is approximately the same.

  4. (4)

    Repeat (3) until the allocation is completed (Fig. 1).

    Fig. 1.
    figure 1

    Calculate the value of \( K \) by clustering algorithm

    Full size image

We get the following paths when M = 3 and the results of the distribution are shown in Table 1.

Table 1. Node assignment by clustering algorithm
Full size table
$$ s_{1} = \{ a,e,i\} \quad s_{2} = \{ b,d,g,h,j,k\} \quad s_{3} = \{ c,f\} $$

2.3 Path Selection

The basic framework of the WSAN model has been completed when the sensor nodes are clustered. Path selection is the most important step of the model, which is related to the real-time and energy consumption of the network.

Assume that the mobile path of the actuator node \( ma_{M} \) is \( M\_P_{M} \) and contains \( t(t \le N) \) sensor nodes. Since the path \( M\_P_{M} \) can only pass through each sensor node once and must return to the starting point, so there are \( t! \) paths in the path scheme. Suppose the starting point are randomly selected. According to the previous analysis, it is a closed loop which passes through cluster member only once. Therefore, the shortest path problem can be seen as a Traveling Salesman Problem (TSP) [6].

2.4 Shuffled Frog Leaping and Ant Colony Algorithm

There are many algorithms for solving TSP. In general, TSP algorithm has two parts: approximate algorithms and exact algorithms. Most of the existing methods use approximate algorithms to solve TSP, such as Shuffled Frog Leaping Algorithm (SFLA) [7], Ant Colony Algorithm (ACA) [8], Genetic Algorithm (GA) [9] and so on.

The SFL-ACA node clustering algorithm suitable for WSAN is proposed by combining the local update search ability of the SFLA [10] and the positive feedback mechanism and distributed computing method of ACA, which effectively realizes the real-time information collection and transmission of WSAN.

The SFL-ACA algorithm flow chart is shown in Fig. 2.

Fig. 2.
figure 2

The flow chart of SFL-ACA

Full size image

3 An Improved WSAN Routing Protocol Based on SEL-ACA

The improved WSAN routing protocol based on SFL-ACA is a real-time routing protocol. We set probability of sensor nodes and cluster the nodes by SEL-ACA based on the improved WSAN network model.

3.1 Probability Setting of Sensor Nodes

The sensor nodes are randomly distributed in the improved WSAN network model. Different probability values are set for each sensor node and the probability represents how often the abnormal event may happen in monitored area of each sensor node. In some area, the probability of anomaly detected by sensor nodes are high and the frequencies of sensor nodes uploading information are also high. So an improved WSAN routing protocol based on SFL-ACA is proposed for various situations with different even probabilities.

3.2 Communication Between Sensor Nodes and Actuator Nodes

The network model used in our paper assumes the similar real-time model of WSAN routing protocol with multiple actuators (MA) [11] node models.

In order to minimize the delay of sensor nodes and actuator nodes transmission, we select the mobile actuator nodes as the cluster head and the sensor nodes on the path as cluster members. When the actuator nodes pass through the sensor nodes and the actuator nodes are in propagation range of the sensor nodes, the sensor nodes delivers all the detected information to the actuator nodes in one-hop manner so that the cluster heads aggregate all the information in the cluster.

3.3 Communication Among Actuator Nodes

Ad-hoc communication is adopted under the node assistance of the actuator nodes and the actuator nodes, considering actuator nodes have strong mobility and computation power.

Firstly, the simplified mode of work is to broadcast the information of the node, such as IP address, location coordinate, and so on, to all nodes by Hello messages periodically.

When one of the actuator nodes receives the Hello message from the adjacent actuator nodes, or the broadcast message passing through the other actuator nodes, it can build the topology of the surrounding local actuator distribution. Thus, the routing table of nodes is established, and the routing table is maintained and updated to ensure valid communications among actuator nodes.

4 Experiment and Analysis

4.1 Experiment of Communication Delay Performance

We apply four different routing protocols to simulate the communication delay performance. The Improved WSAN routing protocol based on SFL-ACA (I-SFL-ACA), General WSAN routing protocol based on SFL-ACA (G-SFL-ACA), Improved WSAN routing protocol based on Energy Efficient and QoS Aware Routing protocol (EEQAR) [12] (I-EEQAR) and General WSAN routing protocol based on EEQAR (G- EEQAR).

The parameters setting for the communication delay experiment is shown in Table 2.

Table 2. Parameters for delay experiment
Full size table

The simulation results of four routing protocols are shown in Fig. 3. With the increased number of sensor nodes, the communication delay of the network is increased. The network communication delay of the general WSAN routing protocol is about two times of the improved WSAN routing protocol under the same number of sensor nodes. Comparing G-EEQAR and I-EEQAR, similar results are obtained. The improved WSAN routing protocol reduces the frequency of the sensor nodes reporting information to the mobile actuator nodes with small probability of event occurrence, which thus reduces the average communication delay.

Fig. 3.
figure 3

Simulation diagram of communication delay

Full size image

With the increased number of sensor nodes, the communication delay of the network is increased. With the same number of sensor nodes in the same network, the delay of network communication in G- EEQAR is about 2 times of G-SFL-ACA. A comparative analysis of two simulation models I-SFL-ACA and I-EEQAR produces similar results. Because I-SFL-ACA cluster head and the mobile actuator nodes are set directly. I-EEQAR considers the distance between the hop distance from the actuator nodes to the sensor nodes, and the sensor nodes in the I-SFL-ACA directly upload information to the actuator node in single hop, which greatly reduces the communication delay between sensor nodes and actuator nodes.

Therefore, it is concluded that with the same protocol and the same number of sensor nodes, the improved WSAN routing protocol has less network communication delay than the general WSAN routing protocol. With the same number of sensor nodes in the same network, the delay of the WSAN routing protocol based on SFL-ACA is smaller than that of the WSAN routing protocol based on EEQAR.

4.2 Deployment of Actuator Nodes

We study the performance of the improved WSAN routing protocol based on SFL-ACA by deploying different numbers of actuator nodes. The average time delay in propagation varies as the number of actuator nodes changes. The parameters are shown in Table 3.

Table 3. Parameter setting table for number of actuator experiment
Full size table

The frequency of events in this experiment is divided into low frequency, medium frequency and high frequency. The frequency is set at 0.5 times per second, 2 times per second and 10 times per second respectively. The frequencies of events are different and the probability of event collision between nodes and the network congestion would be different. Therefore, we may study the appropriate deployment scheme of the number of actuator nodes in different network scenarios.

The experiment provides that the acceptable range of average communication delay is where the average communication delay of the network changes very little with the increase of actuator nodes.

From Fig. 4, it can be seen that with the increase of the number of actuator nodes, the trend of the network average communication delay decline is the fastest, faster and slower. Finally, the communication delay of the network will be stable at a lower time. Therefore, under the premise of reducing communication delay and not increasing too much network cost, the number of actuator nodes is proportional to the number of sensor nodes.

Fig. 4.
figure 4

Time delay simulation of different number nodes under low frequency

Full size image

We can achieve an effective number configuration of actuator nodes from the above simulation experiment. The table for the number configuration of actuator nodes under the condition of low frequency, medium frequency and high frequency are shown from Tables 4, 5 and 6.

Table 4. Configuration of the number of actuator nodes under low frequency
Full size table
Table 5. Configuration of the number of actuator nodes under medium frequency
Full size table
Table 6. Configuration of the number of actuator nodes under high frequency
Full size table

Under low frequency and medium frequency condition, when the number of sensor nodes is less than 500 and the number of actuator nodes is about 10% of the number of sensor nodes, the transmission reliability of the network can be guaranteed and the average transmission delay can be limited to an acceptable range. What’s more, under the condition of high frequency, the number of actuator nodes should be about 12% of the number of sensor nodes. The reliability and real-time of the transmission can be guaranteed under low frequency when the number of actuator nodes is greater than 500 and less than 600, the number of actuator nodes is about 15% of the number of sensor nodes. And under medium frequency, the number of actuator nodes should be about 16% of the number of sensor nodes. The number of actuator nodes should be about 18% of the number of sensor nodes when under high frequency.

The simulation experiment of the number configuration of the actuator nodes under three different frequency conditions can draw the following conclusions: for the different number of the sensor nodes in the network, the number of actuator nodes will be different. Too few actuator nodes will affect the communication delay of the whole network while too many actuator nodes will not improve the real-time performance, but it will increase the network construction cost.

5 Conclusion

We proposed an improved WSAN routing protocol based on SFL-ACA in this paper. We assumed the occurrence probability of event in area monitored by each sensor nodes and clustered the nodes according to solution to TSP problem. Different from the general WSAN routing protocol, our protocol considered the mobile actuator node which may reduce communication delay. Through experiments on communication delay performance and number configuration of the actuator nodes, the performance of our WSAN routing protocol and its advantages over other algorithms are verified. It is shown that our protocol can guarantee the transmission reliability of the network while the average transmission delay of the network is limited to a satisfactory threshold. The proposed actuator deployment schemes is able to deploy an appropriate number of actuators in the actual network so as to avoid unnecessary overhead in network configuration.