Wireless Sensor Networks During Message Broadcast

Published Date: 02 Nov 2017

Wireless sensor networks (WSNs) consist of sensor nodes that broadcast a message within a wireless network. Well-organized broadcasting is a key requirement in sensor networks and has been a focal point of research over the last few years. Many exigent tasks in the network, together with redundancy control and sensor node localization depend mainly on broadcasting. In this paper, we propose a Broadcasting Algorithm to Control Redundancy and Improve Localization (BACRIL) in WSNs. The proposed algorithm incorporates the benefits of the gossip protocol for optimized message broadcasting within the network. Simulation results show a controlled level of redundancy (up to 57.6% even if the number of sensor nodes deployed in the same 500m x 500m area increases from 50 to 500) besides improving sensor node localization in dense networks.

Keywords: Message Broadcast, Gossip, Localization, Node Density, Redundancy, WSNs

Introduction

WSNs are a bridge between the real and virtual world, they are a collection of sensor nodes with the ability to sense many phenomenon of interest over large chronological scales [1]. Each sensor node in the network is equipped with memory, radio frequency transceiver and a power source to broadcast wirelessly over a specified protocol [2]. Broadcasting is a common means for sensor nodes to efficiently share their data with each other. Broadcasting can be utilized to

initialize the network arrangement for route discovery between a given pair of senor nodes and could serve as an efficient method to localize sensor nodes. The simplest way of broadcasting is flooding [3], under which each sensor node rebroadcasts when it receives a message for the first time. It is attractive for its simplicity but causes high redundancy, packets collision, and bandwidth wastage [4], therefore an efficient broadcast strategy is required to reduce the message broadcast redundancy in sensor networks [5]. As a modification to flooding, various probabilistic broadcast protocols have been proposed [5, 6, 7], the proposed protocols avoid the above cited problems and provide alternative solutions to flooding. In addition, sensor network applications also require broadcast protocols to support different degrees of reliability hence probabilistic protocols are more suitable. One of the basic extensions to flooding is gossiping [8], where each node forwards a message in a probabilistic manner. The extensions to gossiping protocols, [9, 10] are predominantly static in nature and cannot adapt to the changing topology as well as changing application requirements. Therefore, static protocols require the network designer to conservatively pre configure the parameters, on a case by case basis, in order to allow for changes in the network topology (node density and number of duplicate broadcasts). In this paper, we propose a Broadcasting Algorithm to Control Redundancy and Improve Localization (BACRIL) in WSNs. The proposed algorithm infuses gossip protocol and automatically adapts to changing network topology with increasing node density. BACRIL is light weight in view of the limited resources available with sensor nodes and supports localization of sensor nodes in a specified area with less overhead. The rest of the paper is organized as follows: Section 2 provides a review of the gossip protocol and describes few of its preliminaries. Section 3 presents the proposed BACRIL followed by its performance analysis in Section 4. Section 5 highlights the simulation results. Section 6 concludes the paper.

Review of gossip protocol

Gossip is a probability based protocol and its definition states that whenever a sensor node wishes to send a message, it randomly selects a neighboring sensor node, upon receiving the message for the first time the neighboring sensor node repeats this process, if the same message is received twice, it is discarded [11]. In order to achieve this, each sensor node has to keep track of messages it has already received. Besides favorable for message broadcasting, gossip protocol also performs tasks to help inter process interaction for information exchange in networks where sensor node failure is quite frequent. This section summarizes some general preliminaries for gossip protocol in the context of WSNs [12].

Number of Nodes / Node Density: The number of sensor nodes, N determines the level of confidence for the gossip protocol. Gossip protocol relies on the aspect that each sensor node can make its communication based on negotiations with neighboring sensor nodes. For a dense area A, sensor nodes are likely to receive more messages hence it might prove beneficial to listen to messages with very low probability. In some cases the area A might be very sparse, hence it might be beneficial to listen to the messages with a probability, say P=1, therefore the probability with which a sensor node listens to the messages directly depends on the node density D of the sensor nodes deployed in the network. Sensor node density can be calculated using equation (1).

(1)

In equation (1), R denotes the transmission range of the sensor node.

Node Degree: The node degree Ń depends on the values of A, N and R; it can be calculated using equation (2).

(2)

It is to be noted Ń is often variable since A, N and R cannot be always uniform.

Frequency of Received Messages: Sensor nodes are adjacent if the distance between them is less than the defined transmission range. The sensor nodes if set to a very low probability of listening will transmit only when there is a change, however they may send very few messages in some cases. Further, such a sensor node may or may not choose to listen depending upon the initial value of P. Thus a sensor node gossips only if it receives a new message else it continues to be in a passive state.

Optimized and Robust Network Topology: The network topology can change adversely due to node failures and energy depletion even if localized deployment is made. Therefore robust network topology is essential to maintain correct system operation. The performance, functionality and reliability of gossip protocol do not drops rapidly with node failures and robustness is exhibited in the face of largely varying node capabilities in terms of memory, bandwidth and connectivity.

Fan Out (n): It is defined as a configuration parameter to count the number of sensor nodes selected as gossip targets, upon receiving a message for the first time, the sensor node selects

gossip targets to forward the message. The trade off associated with this parameter lies between

desired fault tolerance level and observed redundancy. High value of n guarantees fault tolerance

but also leads to an increase in the network redundancy.

Relative Message Redundancy (RMR): RMR measures the message overhead for a protocol. It is calculated using equation (3).

(3)

In equation (3) m denotes the number of messages broadcasted during a procedure. This metric is applicable if at least two sensor nodes receive the message. Zero value of RMR denotes that there is exactly one message exchange per sensor node and is clearly the optimal value. High value of RMR indicates a poor network usage and for gossip based message exchange, RMR tends to n – 1.

Broadcasting Algorithm to Control Redundancy and Improve Localization (BACRIL)

BACRIL is designed with the goal to obtain satisfactory broadcast performance in high density WSNs. Scalability is a critical issue in sensor networks composed of several densely deployed sensor nodes. The localization of sensor nodes increases with the increase in sensor node density as each sensor node makes the decision to broadcast according to the local information obtained from its neighboring sensor nodes. We assume that BACRIL does not require any topology information thus the overhead remains small; all the sensor nodes have the same characteristics (the same communication and sensing ranges). The position of every sensor node is not known in any arbitrary coordinate system since we assume that the neighbors of a particular sensor node are determined based on the message broadcast. In this way, the sensor nodes decide locally to broadcast message (as an active node) or to disregard the previously received messages. In WSNs, a message broadcast is said to be redundant if each sensor node in

the network has already received the same message at an earlier time. The technical challenge associated with this problem lies in accurate estimation of redundant message counts for varying number of sensor nodes within the network confined to number of broadcasts. BACRIL is based on the assumption that the sensor nodes N are deployed within a specified area A, are allowed to broadcast a message m, intuitively with the increase in number of sensor nodes corresponding to a high density wireless sensor network. Table 1 presents the steps of the proposed algorithm.

Table 1: Algorithm description

Algorithm: Broadcasting Algorithm to Control Redundancy and Improve Localization (BACRIL)

1: BEGIN

2: define A, N and m

3: where m and N ⊂ A

4: Initialize N =50 and m =1,

5: deploy N such that N ∈ A

6: Start

7: message broadcast

8: broadcast new message, m

9: select neighbor node as gossip targets, n

10: If neighbor node receives message for the first time, go to Step 8;

else go to step 13 11: End if

12: neighbor node receives the message twice, then

13: discard message broadcast;

update, if N abandons its attempt to rebroadcast

14: repeat step5 for N = {100,200,300,400 and 500} with m =1

15: check for redundant counts for N = {50,100,200,300,400 and 500} 16: END

We remark that BACRIL incorporates the benefits of gossip protocol to control redundant message rebroadcast and also guarantees sensor node localization with the advantage that sensor node has to be active only for message broadcast and no node has to localize itself with respect to a global coordinate system. This scheme guarantees that sensor nodes with smallest distance from their neighboring sensor nodes will satisfy minimum rebroadcast in order to control the redundancy.

Performance Analysis of BACRIL

4.1 Definitions

A is the area of the entire wireless sensor network.

D is the sensor node density of the network, i.e., the average number of sensor nodes per region.

r is the coverage radius of each node.

N is the total number of sensor nodes deployed.

m is the number of broadcasted messages.

h is the number of sensor nodes that have rebroadcasted the message after its reception.

RB is rebroadcast ratio, i.e., the ratio of number of sensor nodes that have rebroadcasted the message to the number of sensor nodes in the entire network.

P is the probability by which a sensor node can listen to a message.

R is the transmission range of sensor node.

n are the number of sensor nodes selected as gossip targets.

Based on the above definitions, equation (4) and equation (5) are obtained.

(4)

(5)

The rebroadcast ratio RB manifests the efficiency of the gossip protocol since RB is inversely proportional to the broadcast efficiency. Large value of RB results in high redundant rebroadcast

with low broadcast efficiency. Therefore based on equation (4) and equation (5), the efficiency of BACRIL is determined by the minimum value of RB, and can be calculated using equation (6).

(6)

The values of A, D and r are determinate, hence in order to obtain the minimum value of RB, the value of h should be minimized because if the node density D increases then the value of RB decreases. BACRIL is evaluated for different values of N. The sensor node density with different values of N is shown in Figure 1.

Figure 1: Sensor node density with different values of N

Simulation Results

To study the influence BACRIL on WSNs, we simulated a wireless sensor network constituting of different number of sensor nodes. The simulations were performed on SNetSim [13].The simulator has a complete stack for the gossip protocol and also provides a central management with functionality to set the deployment area and other parameters before creating the network

topology. The sensor nodes are randomly placed in an area of 500m × 500m, where each sensor node can communicate with another sensor node. It was observed that BACRIL completes the message broadcast with a satisfying coverage ratio; the message broadcast maintains a controlled level of redundancy with the increase in number of sensor nodes and guarantees the stability of the proposed algorithm in high density sensor networks. The network was simulated individually for different values of N = (50,100,200,300,400 and 500), the received message counts and received redundant message counts for different values of N are shown in Table 2. The simulation results support controlled redundancy within the network with the increase in number of sensor nodes over the same deployment area. Figure 2 provides graphical data between number of sensor nodes and percentage of observed redundancy in the message counts. The evaluation of controlled redundant message counts is promising for networks with large number of sensor nodes as in cases of sensor node failure the network can retain its message broadcast in a fault tolerant manner. The controlled redundancy will also aid in improving sensor node localization as the location knowledge of neighboring sensor nodes is not considered in the algorithm.

Table 2: Received message counts for different values of N

Number of

sensor nodes

Number of message

broadcasts

Received message

count

Received redundant

message count

50

100

30

70

100

200

79

121

200

400

166

234

300

600

248

352

400

800

335

465

500

1000

424

576

80

70

60

Redundancy

50

40

30

20 Percentage

10 Redundancy

During Message

0 Broadcast

50 100 200 300 400 500

Number of Sensor Nodes

Figure 2: Number of sensor nodes Vs percentage redundancy

Conclusions

In this paper, we propose BACRIL for controlling redundancy in WSNs where sensor node density is a dominant factor. The proposed algorithm utilizes the benefits of gossip towards controlling the redundancy and accurately estimates the received redundant message counts without the location knowledge of the neighboring sensor nodes. Although the estimation of least redundancy during broadcast has been done in previous studies [14] but to the best of our knowledge, we are the first to present a theoretical analysis for redundancy estimation. BACRIL is scalable to sensor node density and works well for a small network topology with 50 sensor nodes ranging to a large network topology with 500 sensor nodes over the same deployment area. The analysis of simulation results for BACRIL show that the redundancy maintains a controlled level for increased values of sensor node densities, this approach can be used to improve self localization of irregularly arranged sensor nodes in dense networks as neighborhood knowledge of sensor nodes does not cast much effect on the proposed algorithm. Our results will also benefit the future research on self localization of high density WSNs by cutting down the total energy consumption due to controlled redundancy for maximizing the network life time.