JOURNAL OF APPLIED COMPUTER SCIENCE Vol. 19 No. 2 (2011), pp. 73-88 Hybrid Data Gathering Scheme in Wireless Sensor Networks Swarup Kumar Mitra 1, Ayon Chakraborty 2, Mrinal Kanti Naskar 2 1 MCKV Institute of Engineering Department of Electronics and Communication Engineering 243 G. T. Road (N) Liluah, Howrah-711204, India swarup.subha@gmail.com 2 Advanced Digital and Embedded Systems Lab Department of Electronics and Telecommunication Engineering Jadavpur University, Raja SC Mullick Road, Kolkata-700052, India jucse.ayon@gmail.com, mrinalnaskar@yahoo.co.in Abstract. For time-sensitive applications from a remote wireless sensor network, demands to design an efficient routing scheme that can enhance network lifetime and also offer an optimized performance in energy efficiency and reduced delay. In this paper, we propose an improved clustered-hop data gathering scheme which is a amalgamation of clustering and nearest neighborhood selection of the sensor nodes in each hop. The cluster heads and the super leader are altered every round for ensuring an uniformly distributed energy consumption among all the nodes. We have implemented the proposed scheme in nesc and performed simulations in TOSSIM. Successful packet transmission rates have also been analyzed using the interferencemodel. Compared with the existing popular schemes such as PEGASIS, BI- NARY, LBEERA and SHORT, our scheme offers an improved "energy delay" performance and has the capability to achieve a very good symmetry among different performance metrics. Keywords: Network Lifetime, Interference model, Energy Delay, TOSSIM.
74 Hybrid Data Gathering Scheme in Wireless Sensor Networks 1. Introduction Wireless Sensor Networks (WSNs) consists of large number of sensor nodes with limited memory, power supply and capabilities in signal processing and wireless communication [1], which are randomly deployed in a large field usually organized within itself to form a wireless adhoc networks Applications of sensor networks vary widely from climatic data collection, seismic, acoustic and underwater monitoring to surveillance and national security, military and health care. The sensor networks are required to transmit gathered data to the base station (BS) or access point where the data can be processed. It is often undesirable or infeasible to replace or recharge sensor nodes. Network lifetime thus becomes an important parameter for efficient design of sensor networks. In case of WSNs, the lifetime can be taken as the time from inception of the nodes to the time when the network becomes non-functional. A network may become non-functional when a single node dies or when a particular percentage of nodes die depending on requirement.each node is provided with transmit power control and omni directional antenna and therefore can vary the areas of its coverage[2].since communication requires significant amount of energy compared to computations, sensor nodes must collaborate in an energy-efficient manner for transmitting and receiving data so that lifetime can be enhanced and also a better "energy delay" performance is achieved.we propose and analyze in this paper a new routing scheme called Hybrid Data gathering Scheme(HDS), which can ensure the best "energy delay" performance while, at the same time, achieve a good balance among other performance metrics such as energy efficiency. We divide the network into clusters and subsequently [3]is applied for data gathering in each cluster as well as among the cluster heads. In our work, the data gathering scheme HDS is coded in [4] for [5]software platform and next simulation is conducted in [6] environment. This not only signifies the coding feasibility of the scheme, but also verifies it for running in real hardware platforms (like Micaz or Mica2). A radio interference model was introduced in simulating the packet losses and packet delivery rates. The rest of this paper is organized as follows. Section 2 deals with related works. Section 3 gives the system model and formulates the data gathering problem in wireless sensor networks, Section 4 illustrates the proposed algorithm for HDS together with an in-depth analysis of some important performance metrics. Simulation results for different routing schemes and network sizes are compared and discussed in Section 5, followed by conclusions in Section 6.
S. K. Mitra, A. Chakraborty, M. K. Naskar 75 2. Related Works Several cluster based and chain based algorithms have been proposed for efficient data gathering. The Power Efficient Gathering in Sensor Information Systems (PEGASIS) scheme proposed in[2] is based on a chain, which starts from the farthest node from the BS. By connecting the last node on the chain to its closest unvisited neighbor, PEGASIS greatly reduces the total communication distance and achieves a very good energy and lifetime performance for different network sizes and topologies. CDMA capable and non- CDMA-capable sensor nodes, the chainbased BINARY and 3-Level Hierarchy schemes were proposed respectively in [7] to achieve better "energy delay" performance than PEGASIS. In LBEERA[8], a cluster-based Load Balance and Energy Efficient Routing Algorithm (LBEERA) is presented. LBEERA divides the whole network into several equal clusters and every cluster works as in PEGASIS. To reduce energy consumption, a new algorithm to construct the lower chain in each cluster is also proposed. A tree-structure routing scheme called Shortest HOP Routing Tree (SHORT)[3]offers a great improvement in terms of "Energy Delay" with a good performance for network lifetime.leach [9]rotates the roles of cluster heads among all the sensor nodes. In doing so, the energy load is distributed evenly across the network and network lifetime (in unit of data collection round) becomes much longer than the static clustering mechanism. Hybrid Indirect Transmission (HIT)[10] is another novel routing protocol for data gathering in remote sensing sensor networks that increases network lifetime. WSN provides wide area of application which requires radio connectivity, geographic coverage and reliable communication between sensor nodes. The abstract regions [11] are provided by a family of spatial operator that captures resource usage of local communication. A range of common idioms in sensor network are used to define the regions in network programming. These include identification of neighboring nodes, data sharing, and data reduction within local neighborhoods. These operators allow nodes to query the state of neighboring nodes and implement efficient aggregation, compression, and summarization of local data. The impact of three sensing model Boolean sensing model, Shadow-fading sensing model and Elfes Sensing model [12] is dealt here with for monitoring of deployed nodes in Wireless Sensor Network. We also observe the network coverage based on Possion node distribution. A comparative study of random and regular node deployment is also dealt with. The energy balanced and reduced cost is also established regarding hierarchical grid structure [13]. The significant reduc-
76 Hybrid Data Gathering Scheme in Wireless Sensor Networks tion in delay with good network coverage in different environmental conditions is also preserved. A proposed energy efficient data gathering scheme is implemented using ARM LPC 1248 microcontroller based sensor node [14].The scheme also adopts temporal correlation to gather compressed data by using Huffman coding and prediction algorithm to the sink. 3. System Model and Problem Formulation Our aim in this paper is to maximize network lifetime by minimizing the total energy usage of the individual sensor nodes by formation of an unique datagathering technique. For justification of the efficiency of our protocol and fair comparison with previous works[2],[3],[8],[9], we choose to follow the similar assumptions for the system modeling. We briefly discuss some important features of our model in the following points. 3.1. The Network Model The foundation of HDS relies on the realization of a powerful Base Station which is connected to an adequate source of energy supply. Some of the important features of the sensor network are:(i) The Base Station is fixed and located far away from the sensor nodes.the sensor nodes are static, energy constrained and homogeneous with a uniform initial energy allocation.(ii)the nodes are equipped with power control capabilities and omni directional antenna to control the direction and magnitude of transmitted power. (iii) Each node senses its vicinity at a fixed rate and always has data to send to the Base Station in every data gathering round.(iv) The inter-nodal distances are smaller compared to distance between the nodes and the Base Station. The two key elements considered in the design of HDS are the sensor nodes and the Base Station. The sensor nodes are capable of operating in two modes: The Sensing Mode and The Leader Mode. In the Sensing Mode, the nodes perform sensing tasks and relay the sensed data to the Leader node through a multihop routing chain. In the Leader Mode a node gathers data from the other nodes in the chain performs the final data fusion tasks and sends them to the Base Station. The Base Station on the other hand performs some of the crucial tasks like formation of the data gathering chain and selection of the Leader Node.
S. K. Mitra, A. Chakraborty, M. K. Naskar 77 3.2. The Radio Model We considered the first order radio model for calculation of the energy dissipation for data communication operations like transmission and reception. This is one of the most widely accepted and used models in literature for sensor network simulations and theoretical analysis. The energy spent by a node in transmitting a k-bit packet to another node d meters away, is given by: and that for receiving the packet is, E T X (k, d) = (ξ elec + ξ amp d n ) k (1) E RX (k) = ξ elec k (2) Here ξ elec (50nJ/bit) is the energy dissipated per bit to run the radio electronics and ξ amp is the energy required by the transmit amplifier to maintain an acceptable signal to noise ratio (SNR) in order to transfer data messages reliably. n is called the path loss exponent, whose value enhances with increasing channel non-linearity (usually, 2.0 n 4.0). In our approach, we have used both the free space (distance 2 power loss) and the multipath fading (distance 4 power loss) channel modes. It is also assumed that the channel is symmetric so that the energy spent in transmitting a packet from node i to j is the same as that from node j to i for any given value of SNR. For communication among sensor nodes we take n=2, and that between the leader and Base Station, we take n = 4, in (1). Value of ξ amp =10pJ/bit/m 2 for n = 2 and 0.0013pJ/bit/m 4 for n=4. Now for all practical purposes, we can assume that the computational energy is much less than the communication energy and thus can be neglected. 3.3. Problem Formulation The figure 1 below shows a 100-node fixed sensor network in a field of size 50m 50m with the base station(bs) fixed and far away from all the sensor nodes at (x=50m,y=150m)further, the sensor nodes are assumed to be homogeneous and energy constrained with uniform energy. An important operation in a sensor network is systematic collection of data from the field, where each node has a packet of information in each round of communication[7]. For the sake of comparison with previous works we choose to follow the similar assumptions for the system model. Data aggregation is performed at intermediate nodes to combine the locally
78 Hybrid Data Gathering Scheme in Wireless Sensor Networks Figure 1. A random deployment of 100 sensor node in a (50m 50m) field divided into five clusters sensed data with multiple[15] incoming k-bit data packets and generate a single k- bit outgoing packet. The position information of all the nodes is known to the BS by using the Global Positioning Systems (GPS) or other techniques[9]. Once deployed, all the nodes are stationary and cannot replace or recharge their limited batteries. But they have the capability of adjusting their transmission power to control the radio coverage range, such that a node can transmit its k-bit packet directly to the BS (onehop) or, alternatively, through several intermediate nodes (multi-hop). For wireless communication, the simple first-order radio model, as in figure 2 is used to calculate the energy consumption for transmitting and receiving data packets. Defined in[7], the performance metric "energy delay" characterizes the trade off relationship between energy efficiency and delay in each data collection round. The factor of packet loss during transmission was observed while simulating our scheme which might encounters delay. Hence by considering the interference model as de-
S. K. Mitra, A. Chakraborty, M. K. Naskar 79 Figure 2. First order Radio Model Figure 3. The mean packet loss rate versus distance is shown, with error bars indicating one standard deviation from the mean. The model is highly variable at intermediate distances TOSSIM radio loss model based on empirical data picted in figure 3 we overcome the losses. The Simulation process in [6] considers the TOSSIM radio loss model, which is based on the empirical data.the loss probability captures transmitter interference using original trace that yielded the model.
80 Hybrid Data Gathering Scheme in Wireless Sensor Networks More detailed measurements would be required to simulate the exact transmitter characteristics; however experiments have shown the model to be very accurate. 4. Proposed Algorithm The key idea of our approach is to divide the whole field into a number of Clusters as in LBEERA[8]. Then applying shortest hop data gathering [3] scheme in each of the cluster. The scheme adopts centralized algorithm which requires a powerful BS, rather than the sensor nodes with limited resources and it take the responsibility to manage the network topology and calculate the routing path [16] and time schedule for data collection. 4.1. Phases for the Cluster Formation The cluster-head in each of the cluster acts as a leader. HDS operates in three phases: Cluster and group formation phase In each round one leader for each cluster will be elected based on the residual energy of the cluster members and their distances from the BS. Leader and super leader selection phase Initially in each cluster the nearest node to the BS is selected as the cluster-head and among the cluster-heads the nearest to the BS is selected as the super leader. From the 2 nd round to select cluster-heads as well as the super leader, BS considers two important parameters. The first parameter is the distance between the node and BS, denoted by D. The second parameter taken into account is the remaining energy of a node denoted by E residual. P i = E residual /D i 2 (3) P is a function of D and E residual. For a particular round the cluster member with the maximum P will be selected as the cluster-head and the cluster-head with the maximum P as the super leader. Super leader and leader are rotated in every round according to criterion for evenly distributing the energy load among all the nodes. Data transmission phase After the creation of the clusters and selection of cluster-heads and super leader, sensors start data gathering and transmission operation.
S. K. Mitra, A. Chakraborty, M. K. Naskar 81 4.2. Calculation of Delay, Message Complexity, Energy Delay product and Mean Delay We consider a field containing N randomly deployed sensor nodes, divided into M geographic clusters. Without any loss of generality, we assume that Cluster 1 contains N 1 nodes, Cluster 2 contains N 2 nodes and so on, Cluster M containing NM nodes. Delay calculation In each cluster delay for data gathering in the individual cluster heads is log 2 N i, for the i th cluster. After the data is accumulated in the M cluster heads, it takes anotherlog 2 M+1 time slots to gather the data in the base station. The plus one factor is for transmitting the final data packet from the super leader to the base station. Since, the algorithm is applied in parallel among all the clusters, data gathering tasks to the cluster heads occur simultaneously. Thus the delay for data gathering in each cluster is lower bounded by,log 2 N max where N max = max of(n 1, N 2,... N M ). So overall delay will be, log 2 N max + log 2 M + 1 log 2 N max N log 2 N + k (4) k is a constant which decreases as the distribution of the sensor nodes is more uniform, and becomes zero when N 1 = N 2 =... = N M = N max. Here x is the ceiling of x, denoting the least integer, greater than x. So, we see the delay to have a complexity be O(log N). Message Complexity In the HDS scheme, in each round, there is a single cluster and group formation phase and a single leader and super leader election phase. Assuming, each node has a packet to send in every round, total number of messages passed is N. Thus message passing complexity is linear, i.e. O(N). Energy Delay product As we can, reasonably say that the radio transmission and reception energy is greater than CPU or processing energy by several orders of magnitude, we take message passing as a rough measure of energy consumption in the nodes. Thus "energy delay" product have a complexity of O(N log 2 N). Mean Delay We define the mean delay as the average of the delay to the BS from each of the nodes. Mean delay (MD) is formulated as MD = log N i=1 { log M j=1 (N j[log N i i + 2]/2 i )} N (5)
82 Hybrid Data Gathering Scheme in Wireless Sensor Networks Figure 4. Comparison of network lifetime of algorithm vs number of nodes 5. Simulation Results Our proposed Hybrid Data gathering Scheme was evaluated by extensive computer simulations. We implemented HDS, in NesC [4], hosted by the TinyOS software platform, simulated in the TOSSIM (a simulator of TinyOS) environment and also in MATLAB and finally compared the results with the existing schemes like the PEGASIS [2], SHORT [3] and LBERRA [8]. Apart from it, our HDS code written in nesc can run directly on sensor motes also.the figure 4 above shows minimum cost(energy Delay) comparision results of HDS with different Data gathering Schemes.Our proposed Hybrid Data gathering Scheme was evaluated by extensive computer simulations. We implemented HDS, in nesc [4], hosted by the TinyOS software platform, simulated in the TOSSIM (a simulator of TinyOS) environment and also in MATLAB and finally compared the results with the existing schemes like the PEGASIS [2], SHORT [3] and LBERRA [8]. Apart from it, our HDS code written in nesc can run directly on sensor motes also.the figure 4above shows Minimium cost(energy Delay) comparision results
S. K. Mitra, A. Chakraborty, M. K. Naskar 83 Figure 5. Comparison of Energy Delay of different schemes vs Number of Nodes of HDS with different Data gathering Schemes.In the TOSSIM environment [6], we have simulated a network of 20 nodes, divided into 4 clusters. TOSSIM builds up an interference model by simulating lossy-links and consequently enable us to study the packet losses as if in a real life situation. The figure (table 1 and figure 6) shows, the maximum and minimum fraction of the total of 20 packets that reached successfully, for various number of retransmission attempts. TOSSIM environment requires a huge amount of memory to simulate larger networks, and is not suitable for our purpose hence we have used MATLAB for larger networks. The nodes were distributed randomly in a "100m 100m" field, i.e., the x and y coordinates of the node were independent and uniformly distributed in the interval [0,100]. The Base Station was located at (50,150). All the nodes had initial energy of 1 Joule. Data packets considered were of constant size of 2000 bits each, and the first order radio model (refer equation 1 and 2) was used to calculate energy consumption in the nodes.
84 Hybrid Data Gathering Scheme in Wireless Sensor Networks Figure 6. Fraction of Packets successfully reaching the base station with retransmission Attempts The upper dark portion of the bar shows the range of the fraction, the tips indicating maximum and minimum fractions (simulated in TOSSIM). Table 1. Number Retransmission Max Min Total of Simulation Attempts Packets 10 1 13 9 20 10 2 17 9 20 10 3 18 13 20 10 4 20 17 20 10 5 20 19 20 For generality and accuracy the following results are the average of the 5 similar simulations for different randomly distributed network topologies of the same dimensions.the rate of packet loss was simulated using the radio interference model discussed earlier. The number of successful packet transmissions was studied in TOSSIM by varying the number of retransmission attempts, shown in table 1.
S. K. Mitra, A. Chakraborty, M. K. Naskar 85 Figure 7. Variation in network lifetime with number of dead nodes,for a 100-node network 5.1. Performance Comparison SHORT develops the shortest hop routing tree, which optimizes energy consumption with a minimized delay. Our HDS performance comparison with other scheme is shown in table 2 and in table 3. As shown in table below, Hybrid Data gathering Scheme can achieve much better "energy delay" performance than other routing schemes, e.g. the performance gains over BINARY, SHORT, PEGASIS and LBEERA are about 20%,10%, 80% and 90%, respectively, for the network of 50 nodes, with a good network lifetime. The mean delay (D) defined as the average of the delay to the BS from each of the nodes computed from equation 5 and simulation results are shown in table 2. This demonstrates that Hybrid Data gathering Scheme can make a good balance between energy efficiency and delay. It is seen from the above obtained results for different network sizes that the denser network of 100 nodes has more resources (e.g. energy and nodes), shorter average communication distances between the nodes, and is more efficient in energy consumption
86 Hybrid Data Gathering Scheme in Wireless Sensor Networks Table 2. Network of 50 nodes FND: First Node Dies, HND: Half Node Dies, LND: Last Node Dies Performance PEGA BIN SHORT LBEERA HDS Metrics -SIS -ARY Network FND 705 510 1482 1143 1511 Lifetime HND 2182 1432 2169 2160 2177 (no of rounds) LND 2575 1897 2413 2533 2613 Energy Consumption is 11.5 15.54 12.31 13.96 12.07 (10 3 Joule per round) Delay in Mean 31.20 6.30 6.82 8.42 6.85 (slots per Round ) Energy Delay 0.36 0.097 0.084 0.117 0.082 Joule slot Table 3. Network of 100 nodes FND: First Node Dies, HND: Half Node Dies, LND: Last Node Dies Performance PEGA BIN SHORT LBEERA HDS Metrics -SIS -ARY Network FND 849 514 1427 1377 1455 Lifetime HND 2587 1744 2533 2400 2583 (no of rounds) LND 2945 2271 2992 2504 2989 Energy Consumption is 20.12 25.87 19.85 24.42 19.46 (10 3 Joule per round) Delay in Mean 66.02 7.38 7.71 17.82 7.74 (slots per Round ) Energy Delay 1.3282 0.19 0.153 0.435 0.15 Joule slot and data collection. As an example, compared with the 50-node network, the energy consumption per round increases about 60%-75% while throughput is almost doubled in the 100-node network, and thus the energy consumption per packet is reduced in the denser network. Performance comparison of different schemes tabled below consequently for 50 nodes as well as 100 nodes.
S. K. Mitra, A. Chakraborty, M. K. Naskar 87 6. Conclusion In this paper, we have proposed an algorithm called Hybrid Data Gathering Scheme for efficiently collecting useful data from a remote wireless sensor network to the BS. The network lifetime dominates throughout with the increment of nodes always except with slight variation at certain stages. But in overall the performance requirements is good in comparison with other algorithms. Our approach overcomes the losses incurred from all other data gathering schemes, hence the network lifetime override its performance. The fertility of our work lies that we can embedded the features of packet transmission in sensor motes like Micaz or Mica2. References [1] Clare, P. and Agre, Self-Organizing Distributed Sensor Networks, In: In SPIE Conference on Unattended Ground Sensor Technologies and Applications, 1999. [2] Lindsey, S. and Raghavendra, C., PEGASIS: Power Efficient Gathering in Sensor Information Systems, In: Proceedings of IEEE ICC 2001, 2001. [3] Yang, Y. and Wu, H.H.and Chen, H., SHORT: Shortest Hop Routing Tree for Wireless Sensor Networks, In: IEEE ICC 2006 proceedings, 2006. [4] Gay, D., Levis, P., Culler, D., and Brewer, E., NesC1.1 Language Reference Manual, 2003. [5] Levis, P., TinyOS Programming, 2006. [6] Levis, P., Lee, N., Welsh, M., and Culler, D., TOSSIM, In: Accurate and Scalable Simulation of Entire TinyOS, www.eecs.berkeley.edu/ pal/pubs/tossimsensys03.pdf. [7] Lindsey, S., Raghavendra, C., and Sivalingam, K., Data Gathering in Sensor Networks using energy delay metric, In: Proceedings of the Fifteen International Parallel and Distributed Processing Symposium, 2001. [8] Yongchang, Y. and Gang, W., Energy Aware Routing Algorithm Based on Layered Chain in Wireless Sensor Network, In: International Conference
88 Hybrid Data Gathering Scheme in Wireless Sensor Networks on Wireless Communications, Networking and Mobile Computing, 2007. WiCom 2007 IEEE, 2007. [9] Heinzelman, W., Chandrakasan, A., and Balakrishnan, H., Energy- Efficient Communication Protocol for Wireless Microsensor Networks, IEEE Proceedings of the Hawaii International Conference on System Sciences, In: Thirtythird Hawaii International Conference on System Sciences, 2000. [10] Culpepper, J., Dung, L., and Moh, M., Design and analysis of hybrid indirect transmissions (HIT) for data gathering in wireless micro sensor networks, In: ACM SIGMOBILE Mobile Computing and Communications Review, 2004. [11] Welsh, M. and Mainland, G., Programming Sensor Networks Using Abstract Regions, 2003, Harvard University. [12] Asraf, H., Biswas, P., and Chakrabarti, S., Sensing Model and Its Impact on Network Coverage in Wireless Sensor Network, In: proceedings of the Tenth Colloquium and the third ICIIS, Kharagpur,India, 2008. [13] Hwang, S.-F., Lu, K.-H., Chang, T. s.-h., and Dow, C.-R., Hierarchical data gathering schemes in wireless sensor networks, International Journal of Pervasive Computing and Communications, Vol. 4, No. 3, 2008, pp. 299 321. [14] Tharini, C. and VanajaRanjan, P., An Efficient Data Gathering Scheme for Wireless Sensor Networks, European Journal of Scientific Research ISSN 1450-216X, Vol. 43, No. 1, 2010, pp. 148 155. [15] Miao, Z. and Yuanyuan, Y., Data Gathering in Wireless Sensor Networks with Multiple Mobile Collectors and SDMA Technique Sensor Networks, In: IEEE Communications Society in the WCNC 2010 proceedings, 2010. [16] Sengottaiyan, N., Somasundaram, R., and Balasubramanie, A Hybrid Routing Protocol for Wireless Sensor Network, Information Processing and Management Communications in Computer and Information Science, Vol. 70, 2010, pp. 188 193.