Research on the Selection Strategy for Optimal Anchor Nodes Based on Ant Colony Optimization

Transcription

1 Sensors & Transucers 2014 by IFSA Publishing, S. L. Research on the Selection Strategy for Optimal Anchor Noes Base on Ant Colony Optimization Baoxin ZHAO, Bi ZENG Guangong University of Technology, Guangzhou , China Receive: 6 November 2013 /Accepte: 9 December 2013 /Publishe: 31 January 2014 Abstract: With the progress of research on Wireless Sensor Networ, location in Wireless Sensor Networs has become a very important research topic. There must be unable location area that is calle blin zone in this paper ue to the ranomness of the noes eployment. In orer to improve the localization accuracy of the blin noe in blin zone, a new anchor noes selection algorithm was propose base on ant colony algorithm an location constraints. The length of the path an the number of the hop count are all the constraints of the evaluation function, then through the ant colony algorithm we can fin the shortest an the fastest route, which is use to the blin noe s localization. The new algorithm is easy to achieve, the experimental results show that the algorithm can obviously improve the accuracy of the localization in blin zone. Copyright 2014 IFSA Publishing, S. L. Keywors: Ant colony algorithm, Location constraint, Anchor noe selection, Wireless localization. 1. Introuction Wireless sensor networ (WSN) consist of plenty of sensor noes an partial anchor noes [1]. These noes have little volume, low power consumption an can be arrange in large scale. Anchor noes coorinate can be locate via GPS or artificially setting, the other noes whose coorinate is unnown, blin noes can be locate through anchor noes [2]. Localization technology is one of the supporting technologies in wireless sensor networs. It is not only use to report the incient location, but also can be use to assist routing, networ management, networ loa balancing an target tracing [3] etc. To reuce the number of anchor noes an the cost of wireless sensor networs, generally we transform the locate blin noes into potential anchor noes for a new blin noe s location for lac of anchor noes. This paper propose a selection strategy for optimal anchor noes base on ant colony algorithm. If the neighbor of the blin noes inclues three of more anchor noes, it can be locate irectly, at the same time, we can select the nearest an the smallest number of blin hop noes as the potential anchor noes base on ant colony algorithm for the blin noe s location. Simulation results show that the algorithm can improve the localization accuracy, higher networ coverage for selforganizing wireless sensor networ scenarios. The simulation results show that this algorithm can improve the localization accuracy an the networ coverage for self-organizing wireless sensor networ. 2. Factors Affecting Positioning In WSN, the noe localization mechanism is establishing the spatial relationships mechanism via the limite number of anchor noes to etermine the blin noes position. In the receive signal strength inication (RSSI) positioning mechanism, the Article number P_

2 electromagnetic waves reflection an iffraction phenomenon in the actual process maes the RSSI values change ranomly an affecting the positioning accuracy RSSI Ranging s Impact on Positioning In the RSSI location, the receive noes can calculate the istance among its neighbors base on the RSSI value an then figure out its position. The RSSI ranging statistical moel calculations are as follows. where P is the RSSI value on the istance of. Then calculate the erivatives, where is given by Eq. 3 an Fig. 2. P0 P 10n 1 p 10 ln10, (3) 10 n p P ( ) P0 10np log, (1) where P ( ) is the RSSI value at the istance of ; n p is the path loss factor ranging from 2 to 4; P 0 is the RSSI value at the istance of 0 (Bm). Taen 0 1, P0 40, the relationship between RSSI value an the istance is shown in Fig Fig. 2. Relationship between RSSI value an ranging error. Through experimental tests, with the incensement of the istance, the ranging error is also became larger [6] as shown in Fig. 3. Fig.1. The relationship between RSSI value an the istance. It is obvious that the absolute value of the slope of the curve graually ecreases i.e. the RSSI resolution is graually reuce as the istance increases. So the greater the istance, the more effect on the ranging taen by RSSI value. Accoring to the formula (1), taen 0 1, the relationship of the istance an the RSSI value is as follows. P0 P 10np 10, (2) Fig. 3. Ranging error/m in ifferent time an istance/m The Impact of Location Taen by the Number of Anchor Noes In WSN, the blin noe can be locate through three anchor noes RSSI information at least. While within the blin noe s coverage, with the increase in the number of anchor noes, the epenence of each noe is reucing, which can improve the location accuracy [7]. 162

3 2.3. The Impact of Location Taen by Cumulative Position Error DV-Hop algorithm [8-9] is propose by Dragos Niculescu at the Unite States Rutgers University, which is a no-ranging positioning algorithm using istance vector routing an GPS positioning principle. This positioning algorithm iea is to use the hop istance between noes instea of the actual istance. When the unnown noe gets three or more jump istance to the reference noe, the coorinates of the unnown noe can use trilateration or maximum lielihoo estimation metho calculate its position. DV-istance algorithm is similar with DV-Hop. The ifference is that the istance between noes is calculate through RSSI measurement, an then obtains the cumulative istance between blin noe an anchor noe. When three or more anchor noes istance, the blin noe s coorinate can be calculate using trilateration positioning algorithm. However, DV-istance algorithm is only applicable to ense isotropic networ. With the number of the hop times increasing, a sharp increase position error will occur. 3. The Selection Strategy for Optimal Anchor Noes Base on Ant Colony Optimization 3.1. Ant Colony Algorithm Description Ant colony optimization (ACO) [4] is a natural algorithm prouce through real ant foraging behavior stuies. It is one of the most successful group intelligent systems [5, 6]. Ant can leave pheromones on path for information transmission. Meanwhile with time going over, the pheromone will graually evaporate. So the length of the path an the number of ants will mae an impact on the strength of pheromone. In turn, the strength of pheromone can also guie the other ants action. Thus, the more ants through the path, the greater probability the later will choose. It is the reason why ant can fin the nearest foo source through this exchange of information. ACO solves the TSP problem with a goo performance. In orer to reuce the impact on the positioning ranging error, this paper applies Ant colony algorithm on the selection strategy for optimal anchor noes (SSAN). What we shoul consier is that it will trap into local optimal solution via basic ACO [7, 8]. Seconly, we shoul select the appropriate objective function an pheromone function base on the characteristics of WSN Algorithm Moeling This paper improves the basic ACO for the using in WSN with new characteristics of ants. We use Ant-Net algorithm an ivie ants into forwar an bacwar ants. All the forwar artificial ants have memory an can ientify the path gone through to avoi multiple searches on the same path or cycle into an infinite loop. Meanwhile the forwar ants also have limite life cycle T. If the ant fin an anchor noe, its life cycle will be irectly ene an bring the message to the bacwar ant. On the contrary, if the ant oesn t fin an anchor noe, the ant will ranomly sip the next noes until its life cycle ene. In the algorithm, forwar ants gather ajacent noes information, upate pheromone ensity an bring message to bacwar ants. Bacwar ants will return to the blin noe accoring to original path an change the characteristic of ants. We now escribe the algorithm in etail. The number of ant is m, the number of noes is n, an calculate the transition probability of ant from noe i to noe j, which is given by Eq. 4. h ( t) 1 () t () t, j allowe h ( t) 1 () () () sallowe P t t t 0, others (4) where () t present the pheromone on the path from noe i noe j at the time t; () t represent the heuristic information at the time t; an are the relative coefficient of pheromone an heuristic information respectively, which represent the relative importance between the past an the present ata. h ( t)( i, j 1,2,..., n, 1,2,..., m) represents the hop times of ant from noe i to noe j at the time t; allowe {1,2,..., m} tabu represent the set of noes that ant can select, in which tabu is contrainicate table, which recors the set of noes that the ant has gone through. The amount of pheromone will upate when the entire ant complete one step, which is given by Eq. 5. (, tt1) (1 ) () t () t, (5) where 0 1 m pheromone. () t () t represent the passage rate of 1 means the sum of the amount of pheromone in the path from noe i to noe j. In WSN, we efine heuristic factor () accoring the characteristics an Eq. 3, which is given by Eq. 6. t 163

4 a j () t, (6) where ( i, j 1,2,..., n) is the istance between noe i an noe j. 0 a j 1 is the inspire proportion of noe j, which is efine by Eq. 7 1 h 2 min a j 1 h 2 min 1 when noe j is anchor noes, others (7) h where min represents the minimum of hop time from noe i to noe j, which is upate by the bacwar ant. In orer to measure the potential value of noe j that can be use as an anchor noe, this paper use Floy algorithm to figure out the best noes. The weight of path D is given by Eq. 8. D 1, (8) Then we can figure out the length L from the noe i to the other noes an the Transfer matrix using Floy algorithm. An we efine the reciprocal of L as the potential value, which is given by Eq Algorithm Design 1 V, (9) L life cycle is en, if yes, go to step 4, if no, continue step 2. 4) Upate information. Bacwar ants upate each noe neighbor state accoring to forwar ants path an calculate each noe s potential value. 5) Determine the termination conition. If there is no information upating or the loop iterations has been reache, then stop; otherwise, assign t=t+1 an go to step Performance Analysis an Simulation 4.1. Performance Analysis In this algorithm, each ant has a life cycle T, thus the complexity of the algorithm can be measure in the worst case search. We efine the times each ant search is T. So the search time complexity for one cycle is O(2 mt ). We assign the max search times is N, so the total time complexity of this algorithm is C O(2 mtn C ) Simulation an Analysis To evaluate the performance of the improve ant colony algorithm in WSN location, we use Matlab to measure an compare localization error rates for DV-istance an SSAN-istance on accuracy. The simulation environment is as follows: With a transmission raius of 60 m, all sensor noes are ranomly place in a 400 m 400 m topology. 200 sensors are ranomly place in a rectangular sensing fiel, in which there are 40 anchor noes remare as re star. The noes istribution can be seen form Fig. 4. In orer to fin the optimal potential anchor noes quicly, this paper uses improve ant colony optimization for localization (IACO-L) searching path in parallel way. The major steps of the algorithm are as follows. 1) Select the initial value. Initialize colony size m an put them all on the blin boe S. Assign all ants hop times h ( 1,2,..., m) 0, (0) 0.1, (0) 0. 2) Parallel search. Each forwar ant searches the next hop noe accoring to Eq. 4 an upates the taboo list an pheromone ensity on each path accoring to Eq. 5 3) Determine the ant s life cycle. Juge the next hop noe whether it is anchor noe or not. If yes, then en the forwar ant s life cycle an transfer the path information to the bacwar ants an go to step 4. If not, etermine whether the Fig. 4. Noes istribution. Because of both of the istance an hop times can affect the location accuracy; we can efine the value 164

5 of an through experiments in ifferent values. So tae the noe number 148 for example shown as Fig. 6. With the passage rate of pheromone 1, ant life cycle T 10 an total cycle times N 30, the result can be seen from Fig. 7. C The neighbor of the number 148 noe can be seen from Fig. 6. Compare Fig. 5 an Fig. 6, we can fin that number 85 noe has the shortest istance but three hop times. So the potential value of number 85 is low. Fig. 5. Simulation results. Fig. 6. Optional anchor noes. Fig % ensity simulation result. 165

6 In Fig. 5(a), we can see that number 36, 111, 120 an 127 are the better potential anchor noes, but number 111 with the shortest istance has the best potential value. In Fig. 5(b), we can see that number 36, 111, 120 an 127 are the better potential anchor noes an the original anchor noe number 36 has the best potential value. In Fig. 5(c), only number 111 an 127 are the better choice because the pheromone coefficient 2 represent the istance has a more significant effect. In Fig. 5(), also number 111 an 127 are the better choice but potential value is not same. So in the case of 1 an 2, the algorithm has a better performance. This paper applies the algorithm in wireless localization. We have simulate two anchor noes ensity. First, the anchor noes ensity is 20 %, the result can be seen from Fig. 7. Where the blue line represents the result calculate via improve algorithm. Its mean is 9.7 an its stanar eviation is While the re line represents the result calculate via DV-istance. Its mean is 11.4 an its stanar eviation is Secon, the anchor noes ensity is 10 %, the result can be seen from Fig. 8. Where the blue line represents the result calculate via improve algorithm. Its mean is 37.1 an its stanar eviation is While the re line represents the result calculate via DV-istance. Its mean is 42.6 an its stanar eviation is Thus, the improve algorithm has a better performance in accuracy an stability. Fig % ensity simulation result. 5. Summary As a new evolutionary algorithm, ant colony optimization has characters of istribute computing, information positive feebac an heuristic search an so on. This paper improves the basic ant colony optimization an avoi falling into local optimal solution an resulting in premature convergence problem. This paper applies it into WSN for the Selection Strategy for Optimal Anchor Noes, which has a goo balance between istance an hop times. In this paper, this algorithm an the classic DVistance algorithm are compare. This algorithm is better than the classic DV-istance algorithm in the localization accuracy an stability. In practical applications, it can also reuce networ traffic positioning ata an has a high application value. References [1]. Kerem Kucu, Anan Kava, Halil Yigit, et al., A novel localization technique for wireless sensor networs using aaptive antenna arrays, in Proceeings of the IEEE Raio an Wireless Symposium, January 2008, pp [2]. Shuang Tian, Xinming Zhang, Xinguo Wang, et al., A selective anchor noe localization algorithm for wireless sensor networs, in Proceeings of the International Conference on Convergence Information Technology, November 2007, pp [3]. Jan M. Rabaey, Josie Ammer, Julio L. a Silva Jr., Danny Patel, an Rouny Sha, Picoraio supports a hoc ultra-low power wireless networing, IEEE Computer, Vol. 33, Issue 7, July 2000, pp [4]. M. Gunes, U. Sorges, I. Bouaziz, IARA the ant colony base routing algorithm for MANETs, in Proceeings of the International Conference on Parallel Processing Worshops (ICPPW 02), 2002, pp [5]. M. Dorigo, V. Maniezzo, Ant system: optimization by a colony of cooperating agents, IEEE Transactions on Systems, Man, an Cybernetics, Part B: Cybernetics, Vol. 26, Issue 1, 1996, pp [6]. E. Bonabeau, M. G. Dorigo, Swarm intelligence: from natural to artificial systems, Oxfor University Press, New Yor, [7]. J. S. Baras, H. A. Mehta, Probabilistic emergent routing algorithm for mobile a hoc networs, in Proceeings of the International Conference on Mobiling an Optimization in Mobile, A Hoc an Wireless Networ (WIOPT 03), Sophia-Antipolis, France, March [8]. V. Shnayer, M. Hempstea, B. Chen, Simulating the power consumption of large-scale sensor networ application, in Proceeings of the International Conference SEN-SYS 04, Baltimore, 2004, pp Copyright, International Frequency Sensor Association (IFSA) Publishing, S. L. All rights reserve. ( 166