A comparative study of chain based routing algorithms in wireless sensor networks

Many routing algorithms for Wireless Sensor Networks (WSNs) are designed and presented in the literature. The main target of these algorithms is to improve the performance of WSNs. In this paper, a comparative study of chain-based routing algorithms has been done. In this study, we compare the performances of PEGASIS [3], COSEN [6], IECBSN [8], CHAIN-BASED [15], CRA [12] and EAPHRN [9] algorithms when the mobile base station (BS) is used. We perform analytical simulations in terms of network lifetime and average energy consumption


INTRODUCTION
A WSN consists of a number of sensor nodes that are randomly deployed and transform physical data into a form that would make it easier for the user to understand [1]. The direct approach to collect data from sensor nodes is that each sensor node transmits the data directly to the BS. However, this approach consumes a lot of energy to transmit data from each sensor node to the BS, thus, nodes die very quickly and as a result reducing the network lifetime. Therefore, a few transmissions as possible is desirable for efficient energy utilization.
Many routing protocols have been proposed [3,4,6,7,8,9] in which consider reducing the amount of data transmissions in a WSN by fusing these sensing data and then transmit fused data to BS. A mobile BS can follow different types of mobility patterns in the sensor field, such as random mobility, or fixed path mobility, such as random mobility, or fixed path mobility, which has consequences with respect to energy efficiency and data collection strategies [13].
In this paper, a comparative study of chain based routing algorithms, PEGASIS [3], COSEN [6], IECBSN [12], CHAIN-BASED [15], CRA [12] and EAPHRN [9], is conducted by adding the concept of BS mobility in the algorithms and then compare the performance of these algorithms in terms of network lifetime and average energy consumption. The rest of the paper is organized as follows: Section 2 briefly reviews related work. In Section 3, we introduce our problem statement. The experimental simulation is proposed in Section 4. In Section 5, we conclude our work.

RELATED RESEARCH
Routing algorithms and data-gathering raises an important topic in WSNs due to the power limitation of sensor node [11]. There are two categories for the existing routing/data-gathering algorithms: hierarchy algorithms and non-hierarchy algorithms. In non-hierarchy algorithms, sensor nodes have the same role and collaborate to perform the sensing task and multi-hop communication. On the other hands, in the hierarchical algorithms, the network divided into several logical groups called clusters within a fixed area and nodes perform different tasks. The hierarchical algorithms include cluster based, tree-based and chain based routing protocols. In this paper, we focus on chain based routing protocols. The key in chain-based routing, however, is to form chains among the nodes so that each node will receive and transmit only to one pre-determined one-hop neighbor. Data is thereby aggregated through the chain until it reaches the chain leader, which transmits directly to the BS [14]. Many types of research in the last few years have explored chain-based routing in WSN from different perspectives. A variety of protocols has been proposed for prolonging the life of WSN and for routing data to the sink. PEGASIS (Power-Efficient gathering in Sensor Information Systems) [3] is the first and most popular energy-efficient chain based algorithm for WSNs that was proposed for reducing power consumption. PEGASIS introduces only a routing protocol that is near optimal for a data-gathering problem in WSNs. The main idea of the PEGASIS protocol is the formation of a chain among the sensor nodes so that each node will receive from and transmit to a close neighbor. Gathered data moves from node to node, are fused, and eventually, a designated node transmits it to the sink. In [4], the authors proposed an improved energy efficient PEGASIS based protocol (PEGASIS-E). PEGASIS-E uses average distance among the sensor nodes as the criteria for chaining, thereby providing better performance in terms of energy dissipation and amount of information sent to the sink. The simulation results obtained show that PEGASIS-E gives an increase in the network lifetime as compared to PEGASIS. Y.Song [5] proposed an Energy-Efficient Chain-Based routing protocol (EECB) that is an improvement over PEGASIS. EECB uses distances between nodes and the base station and remaining energy levels of nodes to decide which node will be the leader that takes charge of transmitting data to the base station. In [6], the authors proposed a Chain Oriented Sensor Network (COSEN) protocol, which is a two-layer protocol where a number of chains are formed in order to cover the whole region. COSEN reduce the latency associated with the original PEGASIS protocol by forming multiple smaller chains instead of a single longer chain. At first several small, fixed-length chains constructed using the same distance-based greedy heuristic of the original PEGASIS protocol and a chain leader is selected in each chain based on the highest remaining energy. Leader nodes are selected for a certain number of rounds. These chain leaders again create a higher-level chain and among those leaders, one is chosen depending on the distance to the sink and remaining energy level; to accumulate all data from the network and send that data to the sink at each round whereas other nodes only communicate with its neighbor in the chain .
Energy Efficient Chain Based Sensor Network (ECBSN) [7] overcomes several problems of PEGASIS. ECBSN overcomes the problem of excessive delay; instead of one long chain in the network, numbers of short chains are formed.
In ECBSN, every first node in a chain becomes a leader node. Thus, it ignores the suitable proportion of nodes energy and distance between node and BS, which optimize the leader selection according to the various application environments. ECBSN protocol has certain deficiencies like the non-optimal selection of leader nodes in rounds, aggregation, and transmission of data by head nodes that leads to unbalanced energy consumption [8]. An improved ECBSN protocol called IECBSN is proposed in [8]. IECBSN adopts a new method of selection of leader nodes based on selection value (SV) parameter, which outperforms ECBSN. EAPHRN (Energy Aware PEGASIS Based Hierarchical Routing) [9] is a chain based routing protocol in which node does not connect to the next closest node but connects to a random node that is not far than the Distance Threshold (DT). EAPHRN is divided into two phases, Chain Setup phase, and Leader election phase. A Chain-Based Routing Protocol to Maximize the Lifetime of Wireless Sensor Networks (CHAIN-BASED) proposed in [15]. CHAIN-BASED is based on constructing multiple chains in the direction of the BS. The first node of each chain sends data to the closest node in the same chain. This latter collects aggregates and transmits data to the next closest node. This process repeats until reaching the last node, which aggregates and transmits data directly to the BS. The work in [10] is similar to our work, the authors compare five hierarchal cluster-based routing protocols and introduce the mobility of BS in the five protocols and compare their performance. The main difference between our work and [10] is that our work provides analytical simulations in terms of network lifetime and average energy consumption for different chain based routing protocols.

PROBLEM STATEMENT
Our objective is to analyze the performance of the chain routing algorithms when using a mobile BS. We compare chain based routing algorithms such as CRA, PEGASIS, CHAIN-BASED, COSEN, IECBSN, and EAPHRN, by implementing the concept of BS mobility in them. Their performances are observed and compared. The position of BS is varied at different locations within the network area according to different scenarios.
In the proposed scenarios, we consider data collected by a single mobile BS traveling through proposed mobility scenario in the monitored region. Initially, BS starts motion from the initial position of the bounded services area. BS changes its relative position according to the different mobility scenarios (random, and predetermined (Fixed path)).

3.1Random mobility scenario
In random mobility scenario, the BS initially is placed randomly on the edge of the area. Then a destination is randomly chosen and the BS moves towards this destination. Upon arrival, the BS pauses for a time and again chooses a new destination for the next movement. As a result, the BS is staying in a location for a certain period of time (pause time). While the BS is staying in a location for a certain period, it broadcasts a start message to the network nodes. After receiving the start message each head node sends the data packets to the BS until receives the stop message. Before the BS changes its position, it broadcasts another message to reset the nodes and stop the transmission, to reduce the packet drop. After that, BS changes to a new position and follow the same steps every time [16,17]. We use pause time because BS needs to collect the data packets before change its position and we have taken long pause time equals to round's time.
 Rectangle mobility scenario: in rectangle mobility scenario, the BS moves along rectangle trajectory along the boundary of the network and pauses to collect data during round period then moves to next point.  Circle mobility scenario: in circle mobility scenario, the BS circles along the boundary of the network and pauses to collect data during round period then moves to next point in the next round.

EXPERIMENTAL SIMULATION
We modified the WSN simulator developed by David J. Stein [20]. The simulation consists of two stages: deploying the network and running simulations. The simulator is written in C# using Microsoft Visual Studio.NET.
In our simulation, 100 sensor nodes are randomly deployed in a region of size 200m x 200m on a two-dimensional plane and are uniformly distributed with BS (static and mobile).

Energy Model
In order to measure the energy consumption of sensor nodes, we use the same energy parameters and radio model as discussed in [2], wherein energy consumption is mainly divided into two parts: receiving and transmitting messages. The transmission energy consumption needs additional energy to amplify the signal depending on the distance to the destination. Thus, to transmit a k-bits message a distance d, the radio power consumption will be, and to receive this message, the radio expend will be ( ) where is the distance between sender and receiver, ( ) is the cost of transmitting an k-bit message for a distance d, ( ) is the cost of receiving and l-bit message for a distance d, is the electronics energy that depends on the circuit itself, is the energy consumed by the transmitter amplifier for longer distance, the energy consumed by the transmitter for shorter distance, the threshold distance √ and is the energy for data aggregation.
The simulated model parameters are given in Table 1.

Performance Metrics
In two different scenarios, Static BS and Mobile BS (random, rectangle, circle, and middle path) we use the following performance metrics to indicate the performance of algorithms (COSEN, EAPHRN, IECBSN Table 2 and Table 3 show the simulation results for static BS at different locations (100,100), (100,200), and (100,300).  . IECBSN and COSEN perform significantly better than the proposed CRA if the BS is located far away from the network. This is because in CRA there is long distance between the BS and the leader nodes that want to send data, however IECBSN and COSEN have two levels of leaders (lower level with many leaders send their data to higher level leader node), which save the energy comparing with the CRA, PEGASIS, CHAIN-BASED, or EAPHRN. Figure 2 presents the average energy consumption in COSEN, EAPHRN, IECBSN, PEGASIS, CHAIN-BASED and CRA algorithms.   Table 4 and Table 5 show the simulation results for mobile. Figure 3 and Figure 4 show the lifetime and average energy consumption for PEGASIS, CHAIN-BASED, COSEN, IECBSN, CRA and EAPHRN in the case of mobile BS.

Table3: Average energy consumption in each algorithm for static BS scenario
As shown in Table 4 and Figure 3, COSEN gives better network lifetime in case of random mobility scenario and IECBSN gives better network lifetime in case of circle mobility scenario, moreover, PEGASIS and EAPHRN gives better network lifetime in case of rectangle mobility scenario. While CRA and CHAIN-BASED give better network lifetime in the case of middle path scenario.
As shown in Table 5 and Figure 4, COSEN gives minimum average energy consumption in case of random mobility scenario, IECBSN gives minimum average energy consumption in the case of circle mobility scenario, and EAPHRN gives minimum average energy consumption in case of rectangle mobility scenario, while PEGASIS, CRA, and CHAIN-BASED give minimum average energy consumption in case of middle path scenario.
From Figure 1 and  Figure 2 and Figure 4, we can notice that the average energy consumption for COSEN decreased in case of random mobility scenario compared to the BS at (100,100) and decreased in case of all mobility scenarios compared to the BS at (100,300), while the average energy consumption for COSEN increased only in case of rectangle mobility scenario compared to the BS at (100, 200). The average energy consumption for EAPHRN and IECBSN is increased in all mobility scenarios compared to the BS at (100,100), (100,200), (100,300). In addition to that the average energy consumption for PEGASIS and CHAIN-BASED increased in all mobility scenarios compared to the BS at (100,100), (100,200), while decreased in all mobility scenarios compared to the BS at (100, 300).
Moreover, the average energy consumption for CRA increased in all mobility scenarios compared to the BS at (100, 100) and decreased compared to the BS at (100,300). However, the average energy consumption for CRA increased in the case of circle and rectangle mobility scenarios and decreased in case of random and middle path mobility scenarios decreased compared to the BS at (100,200).
In summary, by comparing the network lifetime and average energy consumption of all algorithms in both cases, mobile and static BS (by considering the average values of different BS locations) scenarios; it is concluded that the network lifetime of CRA, CHAIN-BASED is increased in the case of middle path and random, while decreased for PEGASIS, COSEN, IECBSN and EAPHRN in the case of all mobility scenarios of the BS. In addition to that the average energy consumption is increased in all mobility scenarios for PEGASIS and IECBSN, while decreased for COSEN in the case of random mobility, for PEGASIS decreased slightly in the case of middle path, for CHAIN-BASED decreased in case of middle path and random mobility, and for CRA decreased in all cases of mobility scenarios.

CONCLUSION
In this paper, we have introduced the mobility of sink in all proposed protocols, COSEN, EAPHRN, IECBSN, PEGASIS, CHAIN-BASED and CRA to compare their performances. We have introduced two scenarios are discussed to compare the performances of chain based algorithms; in the first scenario, static BS is used and in the later one mobile BS is used. We have proposed different mobility scenarios, random, circle, middle path and rectangle mobility. In our future work, we aim to find the optimal path for BS movement and incorporate multiple BSs in the network.