Progress In Electromagnetics Research Symposium, Hangzhou, China, March 24-28, 2008
883
Cluster Head Selection Using Evolutionary Computing in Wireless Sensor Networks
G. Ahmed1 , N. M. Khan1 , and R. Ramer2
1
Mohammad Ali Jinnah University, Pakistan 2 University of New South Wales, Australia
Abstract— Wireless Sensor Network (WSN) comprises of micro sensor nodes with limited
energy and processing ability. It is used in military as well as civil applications. In order to enhance the network life time by the period of a particular mission, many routing protocols have been devised. One of these are network clustering, in which network is partitioned into small clusters and each cluster is monitored and controlled by a node, called Cluster Head (CH). The CH should be powerful, closer to the cluster-centroid, less vulnerable and has low mobility, so that it can aggregate the data from its own cluster nodes and then send it directly to the Base Station (BS). In this paper, we are using the method of evolutionary computing for the selection of the CHs. The BS periodically runs the proposed algorithm to select new CHs after a certain period of time. Results show that network life time is drastically increases by the help of evolutionary computing. 1. INTRODUCTION
Wireless Sensor Network (WSN) comprises of micro sensor nodes with limited energy and processing ability. It is used in military as well as civil applications. In order to enhance the network life time by the period of a particular mission, many routing protocols have been devised. One of these is network clustering, in which network is partitioned into small clusters and each cluster is monitored and controlled by a node, called Cluster Head (CH). These cluster heads can communicate directly with the base station (BS). Other nodes send the data, sensed from the environment to these CHs. CHs first aggregates the data from the multiple sensor nodes, and then finally send it directly to the BS. Hence the CH should be powerful, closer to the cluster-centroid, less vulnerable [5] and has to have low mobility. These factors on which CH selection should depend, are described in Section 2.
Base Station
Cluster Head
Sensor Node
Cluster
Figure 1: Wireless Sensor Network.
Several analyses of energy efficiency of sensor networks have been realized [2, 3, 7, 8] and several algorithms that lead to optimal connectivity topologies for power conservation have been proposed [1, 6, 9–12]. But most of them do not provide long lifetime system in a dynamic environment, where the sensors are unattended and it is very difficult to recharge their battery. Therefore we need more energy-efficient solution. In this paper, we are using Evolutionary Computing (EC), also called Genetic Algorithm (GA) for CH selection in a cluster-based WSN. Many aspects of such an evolutionary process are stochastic [4].
884
PIERS Proceedings, Hangzhou, China, March 24-28, 2008
The rest of the paper is organized as follows: Section 2 describes the four factors that influence the network lifetime; Section 3 describes our approach to CH selection in a sensor network. Finally, Section 4 concludes the paper.
2. FOUR FACTORS THAT INFLUENCE NETWORK LIFETIME
A brief discussion of four factors is given below: 1. Distance of a node from the cluster centroid: The BS calculates the distance of each node to its cluster centroid. The lesser the distance, the higher the probability that the node will become CH. 2. Remaining battery power: Obviously, the higher the battery power, the higher the probability that the node will become CH. 3. Degree of mobility: The mobility of the node has great impact on the network lifetime. The topology of the network will be change very frequently due to the high mobility of nodes, which leads to reselection of CHs rapidly. 4. Vulnerability index: This factor tells us how much vulnerable a node is. If it is high, then the node will not be selected as the CH. A detail discussion of this factor can be found in [5].
3. CLUSTER HEAD SELECTION USING GA
Our goal is to search best sensor nodes among hundreds of nodes, so that they can act as CHs. Conventional search methods are not robust, while the GA is a search procedure that uses random choice as a tool to guide a highly exploitative search through a coding of a parameter space. According to Goldberg in [4], the GA has 4 major characteristics: 1. GAs with a coding of the parameter set, not the parameters themselves. 2. GAs search from a population of points, not a single point. 3. GAs use payoff (objective function) information, not derivatives or other auxiliary knowledge. 4. GAs use probabilistic transition rules, not deterministic rules. In many optimization methods, we move carefully from a single point in the decision space to the next using some transition rule to determine the next point. This point-to-point method is dangerous because it is a perfect prescription for locating false peaks in multi modal (many peaked) search spaces. By contrast, GA works from a rich database of points simultaneously (a population of strings), climbing many peaks in parallel; thus, the probability of finding a false peak is reduced over methods that go point-to-point [4]. A GA starts with a population of strings and thereafter generates successive populations of strings. A simple GA consists of three operators: 1. Reproduction 2. Crossover 3. Mutation The chromosome of the GA contains all the building blocks to a solution of the problem at hand in a form that is suitable for the genetic operators and the fitness function. Each individual sensor node is represented by a 4-bit binary number called ‘gene’. These four-bit genes which define the feature of the node are called “allele” and are represented as follows: X1X2X3X4 X1: distance of a node from the cluster centroid, X2: its vulnerability index, X3: its degree of mobility, and X4: its remaining battery power. The possible values of these 4 attributes are shown in Table 1. Let’s take an example. To start off, select an initial population at random. Here, we select a population of size 4. Each string (node) has some fitness value. This value can be evaluated from a fitness function, f (x) = f (x1, x2, x3, x4). In the case of WSN, the fitness function depends upon the four factors, discussed in Section 2. Nodes with higher energy, low mobility, closer to the cluster centroid and low in vulnerability have high fitness values and can be declared as CHs. For our example, let f (x) = x2 . A generation of the GA begins with reproduction. We select the mating pool of the next generation by spinning the weighted roulette wheel four times. From this, the best string get more copies, the average stay even, and the worst die off.
Progress In Electromagnetics Research Symposium, Hangzhou, China, March 24-28, 2008 Table 1: Possible values for the attributes. X1 X2 X3 X4 Digital Value Near Low Low High 1 Far High High Low 0
885
Table 2: CH selection through GA.
No. Node* x f(x) 2 =x 36 144 16 81 277 69.25 Pselecti (fi/Σf) 0.12 0.51 0.05 0.29 1.00 0.25 Expected Count (fi /avg (f)) 0.51 2.07 0.23 1.16 4.00 1.00 Actual Count** 1 2 0 1 4.0 1.0 Mating Pool*** 011|0 110|0 11|00 10|01 Mate **** 2 1 4 2 CrossOver Site**** 3 3 2 2 New Nodes 0110 1100 1101 1000 X f(x)=x2
1 2 3 4 Sum Avg.
0110 1100 0100 1001
6 12 4 9
6 12 13
36 144 169
8
64 413 103.25
*Initial population (Randomly Generated) **From Roulette Wheel ***After Reproduction (Cross site shown) ****Randomly selected From Table 2, we can see that the CHs have improved in the new population. Nodes which are not copying in the next generation will not be elected for CH in the future. In case if the new nodes (offspring) are not present in current WSN setup (have become die off), then we will exclude it for further processing.
4. CONCLUSIONS
The improvement that is provided by GA is not a fluke (a lucky or unusual thing that happens by accident). In our example Table 2, the best string of the first generation (1100) receives two copies because of its high, above average performance. When this combines at random with the next highest string (1001) and is crossed at location 2 (again at random), one of the resulting strings (1101) proves to be a very good choice for CH indeed.
ACKNOWLEDGMENT
The authors would like to thanks to Mr. Badaruddin, PhD Scholar at Institute of Business Administration (IBA), Karachi for his help and guidance.
REFERENCES
1. Chmielewski, D. J., T. Palmer, and V. Manousiouthakis, “On the theory of optimal sensor placement,” AlChE J., Vol. 48, No. 5, 1001–1012, 2002. 2. Slijepcevic, S. and M. Potkonjak, “Power efficient organization of wireless sensor networks,” Proc. IEEE Int. Conf. on Communications, 472–476, Helsinki, Finland, 2001. 3. Krishnamachari, B. and F. Ordo nez, “Analysis of energy-efficient, fair routing in wireless sensor networks through non-linear optimization,” Proc. IEEE Vehicular Technology Conference — Fall, 2844–2848, Orlando, FL, 2003. 4. Goldberg, D., Genetic Algorithm in Search, Optimization and Machine Learning, AddisonWesley Publishing Company, Inc, 1989. 5. Khalid, Z., G. Ahmed, N. M. Khan, and P. Vigneras, “A real-time energy-aware routing strategy for wireless sensor networks,” accepted for presentation in The 2007 Asia-Pacific Conference on Communications, Bangkok, Thailand, 2007. 6. Zhou, C. and B. Krishnamachari, “Localized topology generation mechanisms for wireless sensor networks,” IEEE GLOBECOM’03, San Francisco, CA, December 2003.
886
PIERS Proceedings, Hangzhou, China, March 24-28, 2008
7. Trigoni, A., Y. Yao, A. Demers, J. Gehrke, and R. Rajaraman, “Wave Scheduling: energyefficient data dissemination for sensor networks,” Proc. Int. Workshop on Data Management for Sensor Networks (DMSN), in conjunction with VLDB, 2004. 8. Mhatre, V., C. Rosenberg, D. Kofman, R. Mazumdar, and N. Shroff, “A minimum cost heterogeneous sensor network with a lifetime constraint,” IEEE Trans. Mobile Comput., Vol. 4, No. 1, 4–15, 2005. 9. Ghiasi, S., A. Srivastava, X. Yang, and M. Sarrafzadeh, “Optimal energy aware clustering in sensor networks,” Sensors, Vol. 2, 258–269, 2002. 10. Rodoplu, V. and T. H. Meng, “Minimum energy mobile wireless networks,” IEEE J. Select. Areas Commun., Vol. 17, No. 8, 1333–1344, 1999. 11. Heinzelman, W. R., A. Chandrakasan, and H. Balakrishnan, “Energy-efficient communication protocol for wireless microsensor networks,” Proc. 33rd Hawaii Int. Conf. on System Sciences, Maui, Hawaii, 2000. 12. Chang, J.-H. and L. Tassiulas, “Energy conserving routing in wireless ad-hoc networks,” Proc. IEEE INFOCOM’00, 22–31, Tel Aviv, Israel, 2000.