1 Introduction

WSN is used in vast application domain such as military application, health-care, smart home and environmental monitoring [1–4]. In WSN, numerous small sensor nodes are randomly deployed in the sensing area. The sensor nodes are battery-powered, constrained with limited energy [5, 6]. Since, WSNs are widely used in application area such as target tracking, battlefield monitoring, environmental monitoring, it is impossible to replace the battery power of sensor nodes. Thus the power consumption of sensor node is the main constraint in WSN, which decides the network lifetime. The main functionalities of sensor node comprises of sensing, computation and wireless communication. In sensor nodes, the communication utilizes more power than sensing and computation. Thus, many routing protocols are introduced in WSN to reduce the amount of transmission for subsequent power reduction. In WSN, the data are flooded from the sensor node to the base station located at the center. If the sensor node is located far away from the base station, then the data packet has to traverse long distance which leads to more power consumption. Further, the sensor located around the base station referred as bottle neck zone experience heavy traffic flow and depletes the node’s battery energy faster. Thus the nodes in bottle neck zone die faster which affects the network lifetime and network connectivity.

In order to avoid the power consumption due to long relaying of data packet, sink mobility based data gathering are introduced in WSN [7, 8]. The sink mobility significantly minimizes the number of hops the data packet traverses to reach the sink, which reduces the power consumption in data gathering. The sink mobility delays the transfer of data packet from any particular node to the sink, whereas the critical data requiring immediate action can be broadcasted by the sensor node to travel in multi-hop communication towards the sink. Due to the latency in data transfer, the sink mobility scheme is generally used in non-critical applications such as environmental monitoring [9–11]. The sink mobility achieves significant network lifetime improvement based on the relationship between mobility and data transfer delay [12–15].

2 Related Work

There are many recent work available on mobile sink based WSN. In directed diffusion [16], data centric routing based on attribute-based naming is used, where the mobile sink broadcast interest for data-queries. In Two-Tier Data Dissemination (TTDD) [17] uses grid based approach; the whole network is split to small grids. The sensor node close to the grid point act as dissemination node has the location information for routing. The mobile sink sends the data queries to the nearest grid and it is routed to other grid points, the node reply through reverse path. The TTDD based approach has less control overhead compared to other protocol. In [18], an efficient Data Driven Routing Protocol (DDRP) provides route learning by exploiting the wireless medium broadcast feature, where the data packet has an option called recording for each listener to learn the route to mobile sink. In [19], a Mobility-Assisted Data Collection model (MADC) provides a theoretical model for throughput capacity and lifetime based on parameters such as sink velocity, number of mobile sinks and travelling path. In [20], a Hierarchical Cluster-based Data Dissemination scheme (HCDD) is proposed where routing is performed without expensive GPS device for location information Further several types of mobile sink are used to achieve the mobility for efficient coverage of sensor network. For Random movement of mobile sink, entities such as human, vehicles and animal are used. In [21], humans are used as mobile entity. In [22], the vehicles such as buses which move in fixed paths are used as mobile entity. In [23], whales are used as mobile entities used for data communication in WSN.

3 Proposed Work

The proposed method uses a novel data gathering based on clustering approach using multiple mobile sink for WSN shown in Fig. 1. The proposed method has several stages in data gathering they are (1) sink location trace path (2) Cluster formation stage (3) Data aggregation. If a single mobile sink is used in data gathering, the retrace of the same path is delayed for more time which increases the latency in data transfer. In the proposed method, multiple mobile sink is used in data gathering which retraces the same path with minimum delay. The sensor region is covered by several trace paths and the trace paths are across the region, a single mobile sink is used for pair of trace path.

Fig. 1
figure 1

Proposed architecture

Full size image

The proposed method uses cluster based data gathering approach. The proposed method adopts Heterogeneous WSN, such that the high energy nodes are used as Cluster Head (CH). The CH receives the message from all the nodes in cluster, buffers the message and transmits the message to the mobile sink once it comes into contact. Since CHs experience heavy traffic flow, its energy gets depleted faster which results in loss of connectivity. In order to reduce the amount of traffic, Data Aggregation (DA) [24, 25] is used in the sensor nodes. If multiple nodes send data to the CH at the same time in the cluster, then data collide at the cluster. This can be overcome by allocating a TDMA slot to all nodes in the cluster by the CH. The heterogeneous node located along the trace path not selected as CH directly communicate with the mobile sink while it come into contact. This is achieved by proper moving speed of the mobile sink.

3.1 Sink Location Registration

Initially, the sink moves along the trace path and it halts at various location to initiate route discovery process. In each location, the sink broadcast the route discovery message called INIT message. The heterogeneous nodes which receive the INIT message, having energy level greater than the threshold level respond to mobile sink through a reply message having sender ID and its energy level. The mobile sink selects the node with high energy level as CH and sends the start message with the receive node ID to initiate the cluster formation. Further, the mobile sink evenly distributes the CH selection among the possible heterogeneous nodes at each location. This helps to avoid the energy depletion of particular heterogeneous node. When the mobile sink retraces the same path, the CH transmits the data of all sensor nodes to mobile sink.

The mobile sink contains the information of the previous CHs. Let Wi is the weight of the heterogeneous node Si based on the past election as CH. Such that the heterogeneous node which is recently elected as CH has low weight and the heterogeneous node not recently elected as CH has more weight. Let ERCH denotes the CH election record which contains the information of previous CH. Ei is the energy level of the heterogeneous node. The heterogeneous node having maximum weight is selected as Cluster Head.

$$ W_{i} \alpha \left( {ER_{ch} \left( {S_{i} } \right)/Max\left( {E_{i} } \right)} \right) $$
(1)
$$ CH_{SN\_Curr} = \hbox{max} \left( {W_{i} } \right) $$
(2)

3.2 Cluster Formation Stage

When the heterogeneous node receives the start message from the mobile sink, it becomes CH and starts the cluster formation process. The CH transmits its ID as the sender ID in small beacon message to the subsequent nodes. The Cluster is formed based on K-coverage scheme; due to the cluster contains k sensors to cover the region. Since the K-coverage scheme has better coverage, thus the proposed algorithm covers the sensor region efficiently. The sensor node which receives the message, access the channel potential using the message based on parameters such as Received Signal Strength Indicator (RSSI). Since the channel is symmetric, the received message with good RSSI indicates the channel is in good condition. Thus, the sensor node having high signal strength will join the cluster by sending the ACK message to the Cluster Head. The ACK message contains sender ID same as node ID along with the CH ID of the corresponding cluster, since the same sensor node will receive REQ messages from multiple CH. In order to avoid the data collision in CH, as data has been received from multiple sensor node. In the proposed method, the CH employs TDMA time slot to all sensor nodes in the cluster. Figure 2 shows the cluster architecture. Figure 3 shows the structure of data packet.

Fig. 2
figure 2

Cluster Formation

Full size image
Fig. 3
figure 3

Data packets

Full size image

The CH uses equal time interval τ for all the sensor nodes. The CH informs the time schedule to the sensor node by transmitting a unique message, which contains the ID of sensor node in the order of time schedule, the schedule time interval and initial start time of TDMA time slot. The first sensor node in the order starts to transmit the data in the initial start time and completes within time interval. All other node begins the transmission in the time schedule calculated from the initial start time using the order as offset.

3.3 Data Aggregation

In the proposed method, the data from the sensor node are forwarded to the mobile sink only in its allocated time slot. Since the cluster contains more number of sensor nodes, it creates overhead in CH to transmit a large amount of data to the mobile sink. This overhead is avoided by employing data aggregation on the sensor nodes. The data aggregation also called as data fusion is application dependent. The sink mobility based approach is mostly used for non critical application such as environmental studies. In such case, aggregate functions such as average function can be effectively while collecting the environmental data. The proposed method employs an average function to aggregate the data in the sensor node.

In most of the methods, the Data Aggregation (DA) is performed in centralized manner such as CH, where each sensor node has to transmit all data to the CH. Since the data are aggregated in the cluster head, the sensor node has unnecessary power consumption in order to transmit the data. In the proposed method, each sensor node employs aggregation function to aggregate the data before it sends to the CH to reduce the amount of traffic. Such that the sensor node collects and aggregates the data other than its TDMA time slot. The aggregated data are transmitted to the CH in its TDMA time slot.

Let τ represents the time schedule for all the sensor nodes in the cluster \( \left\{ {\tau_{1} , \tau_{2} , \ldots ,\tau_{k} } \right\} \). τ i denotes the time schedule for a particular sensor node, such that the time interval for the schedule is {tst, tfh}, where tst and tfh is the start time and finish time of the schedule.

Let k be the number of sensors connected to the Cluster Head. Sensor node Si aggregates the data sample set X {X1, X2, X3, … Xn} taken in the time period T given as \( {\text{T}}\forall \tau | \tau_{i} \). The Aggregation function is given as,

$$ P_{avg} = Avg \left( {\mathop \sum \limits_{j = 1}^{n} Xj} \right) $$
(3)

The CH stores the aggregated data and transmits to the mobile sink. In order to differentiate the data of the sensor node, the CH uses separate buffer for each sensor node. The data received in particular time slot is stored in the corresponding buffer. The CH transmits the data in each buffer along with the corresponding node ID to the mobile sink.

4 Energy Analysis

Figure shows the proposed data gathering architecture implemented in a square sensor region. The sensor region is covered by six trace paths, where a single mobile sink is used for two trace paths. Let N is the total number of sensor node present in whole sensor region, where the separated region between the trace paths has ‘n’ number of sensor node. The separated sensor region is similar to the rectangular region. The cluster formed has k number of sensors, such that the number of clusters in the separated region is n/k.

The sensor nodes equally transmit M bits of data. The energy required to transmit and receive ‘l’ bits of data for distance dn based on the Radio Energy Dissipation model of sensor node as shown in Fig. 4 is \( E_{Tx} \left( {l,d} \right) \) and \( E_{Rx} \left( {l,d} \right) \) respectively, such that

$$ E_{Tx} \left( {l,d} \right) = \left\{ {\begin{array}{*{20}c} {lE_{elec} + l\varepsilon_{fs} d^{2} , \quad d < d_{0} } \\ {lE_{elec} + l\varepsilon_{mp} d^{4} , \quad d \ge d_{0} } \\ \end{array} } \right. $$
(4)
$$ E_{Rx} \left( {l,d} \right) = l E_{elec} $$
(5)
Fig. 4
figure 4

Radio dissipation model

Full size image

The energy required in transmission depends upon the path loss exponent n, where n is 2 for free space and 4 for multipath interference. E elec is the energy consumed by transmitter/receiver electronics upon single bit. ‘d’ is the distance between the transmitter a receiver. The energy consumption of transmitter amplifier is given by \( \varepsilon_{fs} \), \( \varepsilon_{mp} \) based on the path loss exponent. Since the sensor nodes perform aggregation, the ‘m’ bits of data are aggregated into fewer bits of ‘h’. The energy consumption of the model with centralized data aggregation (i.e. Cluster Head) is given by \(E_{C\_DA}\) . The energy consumption of the proposed method for Data Aggregation performed in the sensor node is given by \(E_{S\_DA}\). \(E_{C\_DA}\) and \(E_{S\_DA}\) denotes energy used by cluster for both transmitting and receiving the data. The energy gain Egain for a cluster achieved by proposed method is,

$$ E_{S\_DA} = m(E_{elec} + \varepsilon_{fs} d^{2} ) + mE_{elec} $$
(6)
$$ E_{C\_DA} = h(E_{elec} + \varepsilon_{fs} d^{2} ) + hE_{elec} $$
(7)
$$ E_{gain} = E_{S\_DA} - E_{C\_DA} $$
(8)
$$ E_{gain} = \left( {2m - h} \right)E_{elec } + \left( {m - h} \right)\varepsilon_{fs} d^{2} $$
(9)

Since, there are (N/k) clusters in the sensor region. The Total energy gain of the proposed method is given by,

$$ E_{gain\_Tot} = \left( {\frac{\text{N}}{\text{k}}} \right)E_{gain} $$
(10)

5 Performance Evaluation

In order to evaluate the efficiency, the proposed algorithm is simulated in a 500 × 500 m square region with 500 nodes. Table gives the parameter value of radio model. The heterogeneous nodes are 10 % of the total nodes, whereas its energy level is three times of the normal nodes. The nodes are uniformly distributed in the sensor region, where the heterogeneous nodes are distributed along the trace path. Simulation parameters are shown in Table 1.

Table 1 Parameter of the radio model
Full size table

The proposed algorithm is compared with the normal mobile sink based data gathering approach, which uses the same trace path for data collection. The data reception capabilities are analyzed with varying the mobile speed as 100 and 200 cm/s. Comparison of the amount of data packets received for sink mobility speed of 100 cm/s is shown in Fig. 5. Due to the ordered data gathering approach along with data aggregation reduces the number of packets to sense the same region. With sink mobility speed of 100 cm/s, the mobile sink covers a single trace path in 500 s, such that the pair of trace path is covered by 1000 s. Comparison of the amount of data packets received for sink mobility speed of 200 cm/s is shown in Fig. 6. With sink mobility speed of 200 cm/s, the mobile sink covers a single trace path in 250 s, such that the pair of trace path is covered by 500 s. When the speed of the mobile sink is increased, the mobile sink covers the sensor region with increase in number of times, receiving more packets.

Fig. 5
figure 5

Number of data packet received with sink speed of 100 cm/s

Full size image
Fig. 6
figure 6

Number of data packet received with sink speed of 200 cm/s

Full size image

Since the sensor nodes aggregates the data before sending to the Cluster Head, reduces the amount of traffic maintain the residual energy of nodes. Further the heterogeneous nodes are only allowed to act as Cluster Head, which helps to retain the total residual energy of the network. This helps the proposed algorithm to achieve network lifetime increase compared to other data gathering approach as shown in Fig. 7. The proposed method has more overall residual energy compared to other algorithm as shown in Fig. 8. When the node density (i.e. increase in number of nodes), the residual energy of the overall network is increased (Fig. 8).

Fig. 7
figure 7

Comparison of network lifetime by varying the number of nodes

Full size image
Fig. 8
figure 8

Comparison of energy usage by varying the number of nodes

Full size image

Figure 9 shows that the delay performance by varying the number of nodes. Compared to the existing scheme, proposed technique has lesser delay. Figure 10 shows that the throughput analysis. Channel bandwidth is assigned for the proposed scheme is 2 Mbps and the proposed technique achieves 1.4 Mbps throughput by varying the number of nodes. Figure 11 shows that the delivery rate performance. Proposed technique has 40–60 % packets are delivered by varying the number of nodes. Remaining packets are dropped, that packet drop rate analysis is shown in Fig. 12 by varying the number of nodes from 100 to 500 nodes.

Fig. 9
figure 9

Comparison of delay by varying the number of nodes

Full size image
Fig. 10
figure 10

Comparison of throughput by varying number of nodes

Full size image
Fig. 11
figure 11

Comparison of delivery rate by varying number of node

Full size image
Fig. 12
figure 12

Comparison of packet drop rate by varying number of nodes

Full size image

6 Conclusion

In the proposed algorithm, data gathering based on the clustering architecture with TDMA time slot helps to achieve an efficient data gathering approach. Further the data aggregation on the nodes within the cluster reduces the data traffic. The efficiency of the proposed model is evaluated based on number of data packets received, network lifetime and residual energy. The proposed algorithm achieves network lifetime increase with low energy utilization compared to other algorithm. The aim is not to highlight ourself but give a solution for network researchers and motivate to work in a new direction.