PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK

Transcription

1 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER 2016 PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK Yuqiang Qin *,1 and Hui Ying 2 1 Department of Computer Science and Technology,Taiyuan University of Science and Technology Taiyuan, , P.R. China 2 The Affiliated Middle School of Taiyuan Normal University Taiyuan, Shanxi, , P.R. China s: qinyuqiang@126.com Submitted: Mar. 25, 2016 Accepted: July 10, 2016 Published: Sep. 1, 2016 Abstract- This paper proposes a novel localization algorithm for wireless sensor network (WSN). Accurate localization is very important for WSN. WSN localization problem is sometimes regarded as an optimization problem. Plant growth simulation algorithm (PGSA) is a kind of new intelligent optimization algorithm, which is intelligent simulation of plant growth in natural way. In addition to the common characteristics of intelligent algorithms, PGSA show robustness and provides a global optimal solution, etc. In this paper, further enhancement of the algorithm by adding the plant root of adaptive backlight function to effectively improve the computing speed and localization precision has been reported. Comparing this algorithm with simulated annealing algorithm (SAA), simulation results show that this algorithm has a higher and more consistent localization precision and faster computational speed. Index terms: wireless sensor network (WSN), localization, PSGA, simulated annealing. 1287

2 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK I. INTRODUCTION WSN refers to a large group of spatially dispersed and dedicated sensors for monitoring and recording the physical conditions of the environment and organizing the collected data at a central location [1]. In general, the sensor nodes should be low cost and small size that allows them to be deployed in large numbers. Moreover, the power consumption should be small enough to enhance the network lifetime as much as possible. Based on these characteristics, WSN has been applied in many fields, such as environmental measurement, medical, industrial and military applications[2]. Therefore, many researchers have proposed a series of localization algorithms[3][4]. The common feature of all localization algorithms is using such a priori knowledge about the location information of the beacon nodes to estimate the unknown node location. WSN localization is divided into two phases. The first phase is the ranging phase that measures the distance between the unknown node and the beacon nodes. Measurement methods include: received signal strength (RSS), Angle of concatenated (AOA), Time difference of concatenated (TDOA), or Round trip Time (RTT). The second phase is to estimate the location of the unknown nodes through the measured information in ranging phase. But either ranging method, with the effect of environmental noise, it will exist certain range error values. In order to improve the location precision, it must minimize the location error. And this problem could be regarded as optimization problems. This paper proposes a WSN localization method based on plant growth simulation algorithm (PGSA). PGSA is a bionic random algorithm that characterizes the growth mechanism of plant phototropism.at present, research on PGSA is still in its infancy, such as integer programming problems [5], transmission network planning problemsand some intelligent optimization computing[6], butno previous method of spatial localization PGSA has yet been found. Compared to other similar optimization algorithm, PGSA is simple, fast convergence and robustness, which is more suitable for the large-scale environment [7]. 1288

3 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER 2016 II. ENHANCED PLANT GROWTH SIMULATION ALGORITHM PGSA is a bionic random algorithm that characterizes the growth mechanism of plant phototropism. The dynamic characteristics of PGSA are derived from plant growth phototropism. Based on the inherent power of the plant growth and phototropism force, it establishes the dynamic mechanism of plant growth and reproduction, regards the plant throughout growing space as the solution of the feasible region, and regards the light source as a global optimal solution, so the use of PSGA for solving optimization problems is the simulation process of plant to grow towards the light source (global optimum solution). Here is the probability growth model of plant phototropism simulation, and morphactin concentration Equations of growing point in trunk and branches respectively. (The probability of a branch to grow from a node depends on the morphactin concentration of a node. The node with higher morphactin concentration is preferred for growth.) Assuming that the growing points (Better growth environment than the root ) in the trunk M and branches q,, satisfy. Their morphactin concentration are computed as follows. Where is the root of plant, is the backlight function of growing point, when the light is smaller, the greater the value of, and therefore it s inversely proportional to the amount of received light of the growing point. Obviously, Equation (1), (2) shows that. Trunk and branches have p + q growing point totally, and their morphactin concentration are(p 1, P q+p ) as shown in Fig.1. A number is randomly generated between [0,1] and thrown into the morphactin concentration state space to find new basis points and grow a new branch and delete the corresponding growth point, as shown in Fig

4 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK Figure 1. MORPHACTIN CONCENTRATION STATE SPACE Taking into account the growth of plants tends to grow in the trunk and the new branches, the fastest growth of plants must be outermost, so the probability of growth of around the root will become smaller. In order to better simulate the growth morphology of plant, this paper proposes adaptive improvement of the root as follows. Where α, β, γ is determined according to the environment situation. Update according to Equation (3) and update step, search growing point at the basis point In accordance with the Equation (1),(2), recalculate morphactin concentration of the growing point, and get new basis to grow a new branch, and delete the corresponding growth point. Similarly, repeat update according to Equation (3) and update step, search growing point at the basis point In accordance with the Equation (1),(2),recalculate morphactin concentration of the growing point, and get new basis to grow a new branch, and delete the corresponding growth point. After these steps, the growth model grows rapidly to the global optimum in the feasible region until no new branches give birth. At this time gets global optimal solution to obtain the minimum of. 1290

5 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER 2016 III. LOCALIZATION ALGORITHM BASED ON ENHANCED PLANT GROWTH SIMULATION ALGORITHM The goal of the wireless sensor network Localization is to estimate the location of N unknown nodes distributed according to the priori information from M beacon nodes[8]. Considering the WSN localization system, this paper makes the following reasonable assumptions. 1) N unknown nodes and M beacon nodes randomly deployed in a 2-dimensional or 3- dimensional space. Each unknown nodes and beacon nodes have the same communication range R. 2) When the unknown node has been located, then be automatically into a beacon node. 3) When there are less than 3 beacon nodes in the unknown node communication range, the unknown node can t be located. 4) In the ranging phase it will be blurred by the noise, so the ranging value is, where is the actual distance between unknown node and beacon node, is the Gaussian white noise from unknown node ambient. is the detected distance between unknown nodes and beacon nodes by using RSSI, AOA, TDOA or RTT ranging technology. This paper doesn t consider any particular ranging technology.[9][10] Based on the above reasonable assumptions, the localization algorithm process is as follows [11][12]. 1) The beacon nodes set up a network, the unknown nodes communicate with the beacon node around to get ranging value. 2) Through ranging correction method further determine the ranging values between the unknown node and beacon nodes around. 3) Each unknown nodes set its root, could be processed by Centroid Algorithm as follows. Where is the actual location of the i-th beacon nodes. 4) Through the PGSA, each unknown nodes minimizes the objective function, the objective function is minimum Localization error as follows. 1291

6 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK Until there is no new branches gave birth, each unknown nodes could obtain the minimum of with global optimal solution. The omission node in the initial localization phase needs assisted localization as follows [13,14]. 1 The slave beacon node is determined from the one-hop neighbor beacon nodes. The slave beacon node receives communication packets, reads RSSI and calculates distance information. Then, binary linear regression is used to calibrate the parameters P 0 and n for a certain environment. And then, the slave beacon node sends these parameters to other beacon nodes in its communication range. 2 Blind node sends request to the neighbor beacon nodes. The assisted localization request contains the feedback times. 3 The beacon nodes receive the assisted localization request and then give response. The transmit power will enhance accordance with feedback times. 4 Blind node receives information and converts the assisted average RSSI into distance using equation (1). The parameters P 0 and n has calibrated in step 1. The maximum likelihood estimation method is used in localization. The maximum likelihood estimation method is as follows [15,16]. ( x i, y i )(i =1,2,, n ) is the coordinates of the i th beacon node. M is the blind node and its coordinates are ( x m, y m ). r i (i =1,2,, n ) is the distance between the i th beacon node and M. According to the equation of the distance between two points, equation (6) can be got. m m x x y y r x x y y r x x y y r m 2 m n m n m 2 (6) Using the 1 to n 1th equations subtract the n th equation, equation (7) can be got. 1292

7 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER 2016 x x y y 2 2 : : y m 2 xn 1 xn 2 yn 11 yn 1 n 1 n xm x1 xn y1 yn rn r1 x x y y r r n1 n n1 n n n1 (7) Equation (7) can be solved by standard minimum mean square error estimation (MMSE). So, the coordinates of blind node M ( x m, y m ) can be expressed as equation (8). T 1 T X A A A b Where, (8) x x y y 2 1 n 2 1 A : : 2 2 xm X y m x x y y n1 n n11 n x1 xn y1 yn rn r1 B x x y y r r n1 n n1 n n n1 n (9) IV. SIMULATION RESULTS AND ANALYSIS WSN localization simulation and performance analysis of the proposed scheme were processed in Matlab with a total of 60 nodes. The unknown nodes and beacon nodes were randomly deployed in a 100m 100msensor filed. Each node had the same communication radius R = 30m. Do not consider any specific ranging technology. As previously mentioned, it was assumed that the ranging value was only blurred by the additive white Gaussian noise (AWGN), such as, where was assumed to be a zero mean Gaussian noise with variance. Every experiment results were based on an average of 30 tests.[17] In order to measure the average localization error, define the mean localization error as follows. 1293

8 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK Where is the actual location of the i-th unknown nodes, is the estimated location of the i-th unknown nodes, L is the number of the nodes located.[18] a. PGSA VS SAA(Ideal Situation) In the ideal situations, there are no noises during measuring the distances between sensor nodes.in Fig.2, it shows the localization performance between PGSA and SAA when there is no noises. Both algorithms have a good localization performance.[19] Location of beacon node Actual location of unknown node Estimated location of unknown node based on PGSA 80 Y-axis X-axis 1294

9 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER Location of beacon node Actual location of unknown node Estimated location of unknown node based on SAA 80 Y-axis X-axis Figure 2. LOCATION ESTIMATED BY PGSA AND LOCATION ESTIMATED BY SAA However, the localization error of each node is shown in Fig.4. 8 x PSGA SAA Distance from actual and estimated node location Node Figure 3. COMPARISON OF DISTANCE ESTIMATED BY PGSA AND SAA 1295

10 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK From Fig.3, PSGA has higher precision than SAA obviously. In fact, by the Equation (6) their mean localization errors are the following Table 1. They have almost computational complexity (Actually, the computational complexity of PSGA is lower than SAA), so the PSGA localization performance is better than the SAA. TABLE1: COMPARISON OF PSGA AND SAA PERFORMANCE Computing Time/s Mean Localization error/unit PSGA e-005 SAA e-004 b. PGSA VS SAA(Non-ideal Situation) In this situation, there are noises in ranging between sensor nodes.in many WSN applications, the actual environmental noise often occurs, which could cause search time and localization error increase. Therefore, based on this situation, this paper shows some localization tests under the influence of the noise with different variance. From Fig.4, the mean localization error is proportional to the noise variance. And in the same noise environment, PSGA has a better performance.[20] 1296

11 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER PGSA SAA 20 Mean localization error Variance of noise Figure 4. COMPARISON OF MEAN LOCALIZATION ERROR BY PSGA AND SAA In order to verify the correctness and validity of the proposed algorithm, simulation has done using MATLAB. MATLAB is one of the most popular mathematic software. It has been widely used for algorithm analysis, research and teaching. Using MATLAB, the maximum and minimum absolute error of x coordinate, y coordinate and location have been calculated. The comparison between the estimated location and the actual location of blind node has also simulated by MATLAB. The number of beacon nodes is 121 and they are evenly distributed in an area of 100 r *100r. r is the unit length of the area and location error is measured by it. The amount of blind node is 5, 10, 15, 20, 25, 30, 35, 40, 45 and 50 respectively, and the blind nodes are distributed at random. The communication distance of nodes is 12.5 r. The side length of a grid is 10 r. 1297

12 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK Beacon nodes around a certain blind node are divided into two groups, one-hop beacon nodes and two-hop beacon nodes. Figure 5-7 shows the situations of that, there are 10, 30 and 50 blind nodes in the simulation area respectively. Those beacon nodes within the communication range of a blind node are called one-hop beacon nodes. While other beacon nodes outside of the communication range but within two-hop communication range are called two-hop beacon nodes. Fig blind nodes in the simulation area 1298

13 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER 2016 Fig blind nodes in the simulation area Generally speaking, the location error has a tendency to increase with the number of blind nodes. When there are less blind nodes in the simulation area, the localization error is smaller. While, when the blind nodes are distributed intensively, the location error is quite large. Even individual blind node localization would be failure. The right balance is somewhere in the middle, and that is the next phase of work to be done. The location error is taken as the evaluation criterion of localization algorithm performance. The simulation results are shown in the figure

14 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK Fig blind nodes in the simulation area 1300

15 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER 2016 Fig. 7. Location error of 10 blind nodes Fig. 8. Location error of 20 blind nodes Fig. 9. Location error of 30 blind nodes Fig. 10. Location error of 40 blind nodes Fig. 11. Location error of 50 blind nodes V. CONCLUSIONS This paper proposes a novel WSN localization algorithm based on PGSA, which could effectively escape from local optimal solution of the problem that often appeared in nonlinear optimization problem. From the analysis of simulation results, the proposed algorithm has better localization precision and computing speed than the SAA. Although the papers were considered only 2D space, but the algorithm could be easily expanded to 3D space. Recently mobile wireless sensor localization 1301

16 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK and target tracking require higher localization precision and faster computing speed, so the PGSA may be suitable to this field. Therefore, this is an important direction for future research work. ACKNOWLEDGEMENTS This work is sponsored by 2015 Key Project of Jincheng Office of Science and Technology( ), The PhD Start-up Foundation of Taiyuan University of Science and Technology (W ). REFERENCES [1] Yuqiang Qin, Yudong Qi. Ensemble-SVM-based Model of Credit Rating System in Electronic Commerce, Bio Technology : An Indian Journal 2013, 8 (9): [2] B. Hoadley, L. E. Rosenberger and A. Flint, Algorithm for explaining credit scores, U.S. Patent 8,001,041, Aug 16, [3] Bellotti, T., & Crook, J., ``Support vector machines for credit scoring and discovery of significant features". Expert Systems with Applications, 36(2), ,2009. [4] Yuqiang Qin, Xueying Zhang. ``Fuzzy Support Vector Machine-Based Emotional Optimal Algorithm in Spoken Chinese". Journal of Computational and Theoretical Nanoscience, vol.9, no.10, pp ,2012. [5] J. A. Feinstein, Method and system for modeling future action impact in credit scoring, U.S. Patent 7,970,676, Jun 28, [6] Yuqiang Qin, Multiplex Temperature Monitoring System of Transformer, CN Patent , October 6,

17 INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS VOL. 9, NO. 3, SEPTEMBER 2016 [7] Chen, W., Ma, C., & Ma, L. ``Mining the customer credit using hybrid support vector machine technique''. Expert Systems with Applications, 36(4), , [8] Elliott, R. J., & Filinkov, A. ``A self tuning model for risk estimation''. Expert Systems with Applications, 34(3), , [9] Yuqiang Qin, Xueying Zhang. EEMD-based Speaker Emotional Analysis for Speech Signal[J]. Applied Mechanics and Materials.vol ,p ,2012. [10] Quah, J. T. S., & Sriganesh, M. ``Real-time credit card fraud detection using computational intelligence''. Expert Systems with Applications, 35(4), ,2008. [11] J. P. Milana, Data transaction profile compression, U.S. Patent 7,853,526, Dec 14, [12] Yu, L., Wang, S. Y., & Lai, K. K. ``Credit risk assessment with a multistage neural network ensemble learning approach''. Expert Systems with Applications, 34(2), , [13] Yu, L., Wang, S. Y., Lai, K. K., & Zhou, L. G.. ``Bio-inspired credit risk analysis computational intelligence with support vector machines''. Berlin: Springer-Verlag,2008. [14] Yuqiang Qin, Xueying Zhang. ``MSF-Based Speaker Automatic Emotional Recognition in Continuous Chinese Mandarin''. Procedia Engineering, vol.15, no.11, pp , [15] DeS. Bowman and C. Duane, Detecting, Classifying, and Tracking Abnormal Data in a Data Stream, U.S. Patent , November6, [16] Yuqiang Qin, and Yudong Qi. ``EEMD-Based Speaker Automatic Emotional Recognition in Chinese Mandarin''. Appl. Math. Inf. Sci. 8, No. 2, 1-8,2014. [17] Bashir Muhammad and Syed Abd Rahman Abu-Bakar. FACE DETECTION IN PROFILE VIEWS USING FAST DISCRETE CURVELET TRANSFORM (FDCT) AND SUPPORT 1303

18 Yuqiang Qin and Hui Ying, PGSA-BASED LOCALIZATION ALGORITHM FOR WIRELESS SENSOR NETWORK VECTOR MACHINE (SVM), International Journal on Smart Sensing and Intelligent Systems, vol. 9, no. 1, pp , [18] Nuzaihan Mhd Yusof, Norlela Ishak, Ramli Adnan, Yahaya Md. Sam, Mazidah Tajjudin and Mohd Hezri Fazalul Rahiman. International Journal on Smart Sensing and Intelligent Systems, vol. 9, no. 1, pp , [19] Yan He and Benxian Xiao. RESEARCH ON POWER CHARACTERISTIC OF THE ELECTRIC FORKLIFT EPS SYSTEM, International Journal on Smart Sensing and Intelligent Systems, vol. 8, no. 3, pp , [20] Xing Haihua, Yu Xianchuan, Hu Dan1 and Dai Sha, SENSITIVITY ANALYSIS OF HIERARCHICAL HYBRID FUZZY - NEURAL NETWORK. International Journal on Smart Sensing and Intelligent Systems, vol. 8, no. 3, pp ,