ON IMPROVING RELIABILITY IN MULTICAST ROUTING PROTOCOL FOR WIRELESS SENSOR NETWORK

Multicast routing becomes the most challenging problem in Wireless Sensor Networks (WSN). Multicasting is an effective way to facilitate group communication in which the multicast data need to be sent from a source node to multiple receivers. In this paper, a simple and efficient algorithm Minimum Connected Dominating Set (MCDS) is used to form a virtual backbone as forwarding group of the network. The MCDS aims at minimizing the number of nodes, where few nodes should be dominated, which are responsible for forwarding the multicast packets by applying Random Linear Network Coding (RLNC). RLNC has great potential to improve the performance of multicast routing protocol. The objective of this paper is to improve the performance of On-Demand Multicasting Routing Protocol (ODMRP) with respect to reliability using RLNC over MCDS for WSN, so that bandwidth utilization can be increased in the network. The proposed approach is named as


INTRODUCTION
Wireless Sensor Network (WSN) is a wireless network consisting of relatively large number of sensor nodes to monitor environmental or physical conditions [1]. WSN  operations to multicast tactical information [2].
The multicast routing protocol is mainly classified into three categories: reactive, proactive and hybrid [3]. The reactive routing protocol is called as on-demand routing protocol. It creates routes only when desired by the source node.
Example for reactive multicast routing protocol is: ODMRP [4]. The proactive routing protocol is called as table-driven. In which, the route for other node is maintained in the routing table. The hybrid routing protocol is a combination of both reactive and proactive multicast routing protocol.
In WSN, network backbone formation and channel capacity are some networking issues [5].
To solve these issues two most popular techniques were used, they are, (1) Minimum Connected Graph (UD Graph) based on computation of Convex Hull (CH) of sensor nodes [10]. MCDS improves the reliability of the network, because limited number of sensor nodes are engaged in multicast message transmission.
Network coding is a technique where forwarding nodes mix the packets using mathematical operations, which reduces the number of transmissions and save the bandwidth in wireless network [11]. Network coding can be classified as either inter or intra-session. In the inter session network coding, the coded packets are received from different sources to be mixed to solve the bottleneck problem. In the intra-session network coding, the coded packets are received and mixed from same source to address the packet loss problem [12]. Network coding also can be classified into XOR (binary) coding, Reed-Solomon and Random Linear Network coding (RLNC).

Motivation and justification
In this work, MCDS and RLNC techniques are used in ODMRP to send code updates or other data from a sink node to a group of sensor nodes for WSN. Finding MCDS of the network is a promising approach. Recently, some researchers have proposed MCDS alone to construct a virtual backbone for multicast operation and to improve performance of multicast routing protocols in WSN [13,14,15]. In general, MCDS can be constructed and calculated by using either global or local network information and centralized or distributed way respectively. However, due to the characteristics of WSN, it is hard to obtain and maintain global network information also MCDS calculation in a single node is not efficient [16].
Therefore, the proposed multicasting routing protocol focuses on local information and distributed way to construct and calculation of MCDS in WSN.
Javad A.T et al. [5] [11] illustrated this through famous "butterfly network". Therefore, RLNC is essential to communicate a source to multiple receivers at a Thus, the proposed protocol is essential to develop efficient multicast routing protocol for WSN.

Outline of the paper
This paper first presents a comprehensive investigation of MCDS and RLNC, also discusses details of their operations. Second, implementation of the proposed protocol has two phases as shown in Figure 1. In the first phase, the

Figure1
Outline of the paper

Organization of the paper
The rest of the paper is organized as follows: Proposed methodology is given in section 2.
Section 3 discusses about the experimental results of proposed approach. Finally, conclusion about the proposed approach is given in section 4.

PROPOSED METHODOLOGY
In this paper, two most popular techniques were used, they are, (1) Minimum Connected Dominating Set (MCDS), (2) Random Linear Network Coding (RLNC).

Dominating Set
The concept of the MCDS comes from the graph theory [41]. It defines a set of nodes for a given connected graph (network). The CDS network is shown in Figure 2. i.e. G and repeat the process from step 1 to step 5.
By above process, remove the vertex G and go to

Multicast routing protocol for WSN
In this work, the selected reactive or ondemand routing protocol is ODMRP. Because, most of the researchers show that reactive method is better than the proactive method in many aspects such as nodes movement, network life time, self-organizing network model also states that the major strength of ODMRP are its simplicity and scalability [4].

On-Demand Multicast Routing
Protocol ODMRP is a state-of-art on-demand multicast routing protocol [4]. It is a mesh based and a source initiated protocol. Forwarding Group (FG) concept is used to establish a mesh structure in a given network also "soft state" approach is followed to maintain a mesh.  This work extends the ODMRP algorithm by adding further routing information as shown in

Random Linear Network Coding
Recently, RLNC is emerged promising technique for various applications in wireless networks, which has been applied in multicast routing to increase the capacity of a network for maximum multicast flows and reduce the multicast traffic.

Random Linear Network Coding for Unicast
In RLNC, the output data of a given node is obtained as a linear combination of its input data.
The coefficients selected for this linear combination are completely random in nature, hence named Random Linear Network Coding.
The forwarding node combines a number of packets it has received or created into one or several outgoing coded packets. Typically, RLNC performs three different operations [43], they are 1. Encoding, 2. Re-encoding, 3. Decoding  The data X1, X2 and X3 are given to node V1 as input then node (V2) received two coded packets: aX1+ bX2+ cX3 and dX1+ eX2+ fX3 as output of node V1. In order to perform the re-encoding operation on the two received coded packets, the node (V2) generates two random coefficients (g, h) for the two coded packets to be re-encoded. The coding vector of the new reencoded packet can be calculated as following: where, (ga+ hd), (gb+he) and (gc +hf) are the new coefficients of the re-encoded packet. The decoding operation is performed at the node V4 by collecting the coded packets. The coded packets are decoded by forming a matrix from linear coefficients. The matrix is referred to as decoding matrix or transfer matrix [43].

Multicast
In this section, multicast network is considered with multiple independent messages, when a source node wants to send multi message to a set of destination nodes, cutset bound is tight and is achieved error-free using random linear network coding [21]. Distributed RLNC has been applied to multicast routing in WSN, in which destination nodes decode the output data by taking random linear combinations of input data.
RLNC process for multicast is depicted in the  Xe2= dX1+ eX2+ fX3 Xe1= aX1+ bX2+ cX3 Information along the edges: Sink output: Where Y(e) is coded packets on the outgoing edges from node V1, which are linear combination of the sources X(v,1), X(v,2), X(v, 3). The information at the destination nodes can be calculated by equation (2), which is received from forwarding nodes, Where Z(v, j) is re-encoded packet at destination nodes, which are received from forwarding nodes. Linear combinations of coded packets y (e1), y (e2), y (e3) on the edges e1, e2, and e3 can be expressed as For the destination node V5, Re-encoded data on edges e8, e9 and e10 denoted by Y(e8), Y(e9) and Y(e10). They are linear combinations of Y(e3) and Y(e4) and can be expressed as, Destination node V5 recover the original packets from the received re-encoded packets [y(e8), y(e9), y(e10)] T and obtain, where x is the vector of input processes, z is the vector of output processes and M is the transfer matrix, which is obtained from solving the following matrix, where and The square matrices β′.α and κ. α are invertible and unicoding is possible. Each destination node wants to decode the vector data z . This implies that det(β.α) ≠ 0 and det(κ. α) ≠ 0 det(Mi) ≠ 0  i , therefore the product of determinant is non-zero.
Determinant is non-zero means that it has some data for the particular destination node.

RLNMCDS-ODMRP
The main contributions of this work can be summarized Implementation of the proposed protocol has two phases as shown in Figure 19 and Figure 20. In the first phase, the source node discovers the route and constructs the MCDS using convex hull, in the second phase, the source node transmitting the data by applying RLNC through the constructed MCDS in ODMRP to its receivers.
As shown in Figure 19,  the re-encoded packet to its neighbour node until the re-encoded packet reached to its destination node, now the source node is ready to multicast next packet to its receivers.

Experimental setup
In the simulation experiment, nodes were  Table 1.

Reliability
Reliability is defined as the successful end-to-end data delivery ratio [46,47].
is distance between the nodes ri ri snr 1 − is the transmitted signal-to-noise power

Experimental results and analysis
In this section, simulation results of the

Scenario-I -By varying the Terrain Size
In the scenario-I, the performance of proposed protocol is measured for the reliability considered

3.3.2.Scenario-II -By varying the Arrival Rate
In the scenario-II, the performance of proposed protocol is measured for the reliability