Sensors & Transducers 203 by IFSA http://www.sensorsportal.co A New Localization and Tracking Algorith for Wireless Sensor Networks Based on Internet of Things, 2 Zhang Feng, Xue Hui-Feng, 2 Zhang Yong-Heng, 2 You Fei School of autoation, Northwestern Polytechnical University, Xi an 70072, China 2 School of Inforation Engineering, Yulin University, 79000, Yulin, China Tel.: +863899566 E-ail: tfnew2@sina.co Received: 4 May 203 /Accepted: 9 July 203 /Published: 3 July 203 Abstract: To iprove the racking accuracy for wireless sensor networks and solve the proble of network overhead, reduce the energy consuption of nodes, analyzed existing localization and tracking algoriths, ensure that one of the core technology for wireless sensor network perforance. In this paper, based on the existing localization and tracking echaniss, the sensor node localization and target tracking application technology in WSN were analyzed fro the views of iproving precision, prolonging the network life based on the Rang-free theory, coordination theory, particle filter and other coputing ethods. Cobining sensor activation algorith and dynaic clustering algorith, a parallel extended particle filter algorith was proposed for target tracking. Considering Internet of Things target tracking, specific regions of applications such as intrusion detection, a localization and tracking prototype syste based on WSN was built. For the localization proble in WSN, the localization algorith based on obile beacon was designed and ipleented. And cobined with target tracking collaborative algorith, construct a prototype syste for target tracking, which can lay the foundation for further application of localization and tracking technology in WSN. Copyright 203 IFSA. Keywords: Target tracking, Wireless sensor network, Internet of things, Localization, Particle filter.. Introduction Wireless sensor networks (WSNs) as a new inforation acquisition and processing technology, positioning in areas relevant to a wide range of applications in environental onitoring, target tracking, intelligent transportation, and intrusion detection []. In these applications, access to onitoring inforation often need to know the location inforation of sensor nodes in WSNs, otherwise, the collected data is eaningless or invalid [2]. Gets the sensor node location inforation is very iportant, for exaple, only bind the location inforation of the onitoring data position to elaborate on what happened in what position event, in order to achieve the goal of localization and tracking [3]; Another exaple, access to the sensor nodes position distribution can iprove the efficiency of network routing, in order to achieve the load balancing of the network [4] and the network topology auto-configuration [5], to iprove the quality of the coverage of the entire network [6], can also network naespace [7]. Therefore, scientific and effective positioning algorith design is one of the key technologies in WSNs. Usually, WSNs positioning ethod is the ost siple is loaded with the global positioning syste (GPS) receiver for each node, to deterine the location of the node. However, due to the cost, 56 Article nuber P_25
energy of nodes and GPS has certain requireents for the deployent of environental constraints, not actually ay load the GPS receiver to all nodes. Therefore, WSNs generally only a few nodes by loading the GPS or pre-deployent in a particular position to get their coordinates, and a large nuber of unknown node location inforation ust pass a certain location algorith to obtain. In the design of WSNs algorith for node location, need to consider any factors, including the proportion of anchor node, network scale, network fault tolerance, ranging and error, ranging and error, positioning accuracy [7]. Therefore, WSNs location proble is a very challenging task. Target tracking as a scientific and technological developent in one direction, the first can be traced back to the eve of World War II; the world's first tracking radar station SCR-28 [8]. After that, with the continuous developent of sensor technology, such as radar, infrared, sonar, laser and satellite-based target tracking syste eerging and aturing, the theories and ethods of target tracking has been greatly developed. With the developent of wireless sensor network technology in the wireless sensor network environent aneuvering target tracking has becoe an iportant research target tracking direction. Target tracking in wireless sensor networks as a ultidisciplinary subject, it involves including control, signal processing, network architecture, distributed coputing and optiization algorith [9] to achieve target tracking, wireless sensor networks has becoe a popular research direction [0]. In this paper, fro the target tracking theory, indepth analysis based on the principle of particle filter, based on dynaic clustering, this paper proposes a parallel extended particle filter algorith. Through the expansion filter to generate the proposal distribution, a reduction in the required nuber of particles to iprove the particle filtering accuracy at the sae tie, reduce the calculation of target tracking algorith coplexity, thus reducing the energy consuption. In the particle filter stage, the cluster head will particle set into subsets, and assign it to run in parallel cluster eber nodes in each cluster, and inforation fusion in the cluster head, get the target state estiation. Between each eber node cluster does not exist the exchange of inforation, only the inforation exchange between the sensor nodes and cluster head, reducing the energy consuption of network counication. Meanwhile, since the particle filter is assigned to each sensor node in parallel to iprove the efficiency of particle filtering, avoid single node excessive energy consuption, the equalization network energy consuption. 2. WSN Target Tracking Syste Based on Internet of things target tracking, the specific area of intrusion detection applications, often with the onitoring area, the poor onitoring the environent and the lack of the necessary counications infrastructure characteristics, and wireless sensor networks with large-scale rando deployent, and so on ad hoc networks, wireless sensor networks used in these occasions can give full play to the advantages of wireless sensor networks, location tracking has broad application prospects. Target tracking syste based on wireless sensor network usually consists of sensor nodes, the base station and reote onitoring center. Sensor nodes by aircraft or other rando dispenser in the onitoring area, the eergence of positioning sensor nodes detect the target, once the target appears, the sensor nodes easuring a oving target the essage sent by the application of the proposed target tracking algorith, the target tracking and forecasting, and real-tie status inforation sent to the gateway target. The target state is estiated by the WSN gateway, the inforation is sent through the GPRS network to the onitoring center, and real-tie display of the trajectory of the target position in the onitoring center to achieve real-tie tracking of the target. WSN-based target tracking network architecture is shown in Fig.. The overall syste is divided into three layers: the perception layer, network layer and application layer. Sensing layer, using the 802.5.4 protocol, sensor nodes target perception and detection. This layer is ainly responsible for the perception of the detected target inforation and self-organizing clusters strategy for the easureent of target inforation for efficient data co-processing and data fusion, to calculate the target position status inforation to predict the target trajectory tracking of the erged transission of inforation to the gateway device. The network layer ipleents the transission, routing and control of inforation, and to convert between the different protocol stacks and access, and to guarantee the security of data transission process in the data network and the transission reliability. Application layer inforation interaction through reote counication network and field station, real-tie onitoring of target track and display, and can be real-tie network state inforation to the target query. 3. Tracking Filter Algorith 3.. Bayesian Filtering The principle of Bayesian filtering is the a priori probability density and the state observer state variables of the syste function to construct the posterior probability density, i.e. with the predicted state of the syste odel a priori probability density using a cobination of the recent systeatic observation value correction, thereby obtaining the posterior probability density. 57
Fig.. Syste architecture of target tracking in IoT. Variable syste probles in reasoning and analysis of target localization, tracking, dynaic properties, often use the state space odel to describe the tie-varying syste. The state-space odel can describe the unknown aount of dynaic or tense and observed signal with an unknown aount of the relationship, it is suitable to solve the ulti-variable data and non-linear, non-gaussian process. The state space odel typically includes a description of the state variables change over tie syste odel and observation odel state variables associated with easureent noise. The state variable contains inforation that describes the syste state variables for target tracking, target position, speed and otion characteristics. The easureent vector typically eans that the noise associated with the state vector of observations. Based on Bayesian fraework, based on the discrete state-space odel, recursive obtain the probability density distribution. Control input is ignored; the discrete state space odel of the syste is described as follows: State-space odel: x f x, q () Observation odel: z h x, v, (2) where fx, q and, h x v are x n R as the n bounded by nonlinear functions, state vector of the syste at tie, z R is the observation vector x of the syste state, q, v respectively, for the process noise and observation noise, The process noise and easureent noise probability density distributions of ships are known. Definition:,,...,, x x, x,..., x z z z z 2, 2 were all observed values of k points in tie, the set of values of the state variables. Assues that the syste state of a priori probability distribution is known, that is p x0 z0 px0, p x z filtering distribution can be obtained through the prediction and update of a two-step. Assuing the probability of tie k- p x z is known, where adhesion, p x x p x Z dx p x Z p x x Z p x Z dx k k, (3) k where p x x v is a first-order Markov process, prior probability density distribution for the state variables not get the latest observations by the syste of equations obtained. 3.2. Trilateral Localization Algorith In WSNs positioning algorith, the triangular positioning algorith based positioning algorith, and on its basis also spawned a lot of iproved algorith. The principle of trilateral positioning algorith is shown in Fig. 2. In Fig. 2, the location of the three anchor nodes is known, their coordinates are xi, y i, i, 2,3. Fig. 2 white dots represent unknown nodes, its location xu, y u, d, d2, d3 denote the unknown nodes to easure the distance of three anchor nodes, then the following relationship: 58
2 2 2 u i u i i x x y y d ( i,2,3) (4) Fig. 2. Principle of trilateral localization. In order to solve this equation, it can be converted into the on unknown node coordinates x u and y u linear equations. Therefore need to eliinate the quadratic ter; to solve the above equations can calculate the coordinates for the unknown node: 3 3 xu 2 x x 2 y y y u 2 x2 x3 2 y2 y3 x x y y d d 2 2 2 2 2 2 3 3 3 x x y y d d 2 2 2 2 2 2 2 3 2 3 3 2 (5) It is not difficult to see, three side positioning algorith is extended to high diension space is quite convenient. 4. Extended Particle Filter Algorith The particle filter is recursive Bayesian filtering ethod based on Monte Carlo siulation (Monte- Carlo), the key idea is to use rando saple weighted with the value of the related rights and to represent the posterior probability. When the saple size is very large, this probability estiates will be equivalent to the posterior probability density. Assuptions can be independent fro the state of the posterior probability distribution px z rando saple of N independent saples X () i, i,2,..., N state probability density distribution can be approxiated as: N p x Z dx (6) ( i ) x N i dx x as a Dirac pulse function, the corresponding function is expected to where ( ) i I g g X p X Z dx N k N () i gxk N i (7) It can be guaranteed on convergence by the law of large nubers, and convergence is not dependent state diension, can easily be applied to the case of the high-diensional. Because of sapling a probability to estiate distribution is often very difficult or even ipossible, the iportance sapling ethod avoids direct sapling difficulty, fro the distribution of another ore easily in saples randoly saple. If the probability density p x Z directly fro the saple is difficult to obtain particles directly fro sapling, here introduce a probabilistic easier sapling and sapling fro the distribution of q x Z, that is called the iportance distribution. In this case, the above forula becoes: p X Z I g g X qx Z dx q X z g X w XqX Zdx E ( ) q Z g Xk w Xk k One of the iportant weights: () i p X Z wx ( ) wx q X Z p X Z p( X) p Z q X Z p X Z p( X) pz ( ) q X Z 5. Siulation Testing and Analysis (8) (9) The ost iportant paraeters to easure the perforance of target tracking accuracy of target tracking, target tracking algorith based on wireless sensor networks should protect high tracking accuracy under the preise as uch as possible to save energy consuption of sensor nodes and to eet short track reaction tie. In this paper, considering the target tracking accuracy and network energy consuption of two perforance etrics, the experient will be presented in this chapter iproved particle filter algorith with the traditional distributed particle filter were copared, the ain difference between the two lies in the selection of the iportance density function and the particle filter node to select two aspects, two aspects of the target 59
tracking accuracy and network energy consuption siulation coparison. In the experient of Fig. 3, the 26 Rules unknown node deployed in the area of 200 00 2 of, respectively, in the anchor nodes 6, 9, 2, 5 case, all the unknown nodes to locate the positioning error respectively: 2.93, 4.0, 3.06, 3.2. Fig. 4 copares the tracking error target tracking in the case of different particles. The figure shows, with the increase in the nuber of particles, the higher the accuracy of target tracking, but the increase in the nuber of particles increases the aount of coputation of the sensor node, the network power consuption increases. these algoriths built for wireless sensor networks targeting tracking prototype syste achieved a certain theory and applied research in the relevant fields, and foring characteristics. The ain contribution of the features include: research target tracking in sensor nodes collaboration strategy to solve the proble of target detection and tracking process node collaboration. A parallel expansion of the particle filter algorith. Algorith using the extended filter to generate the iportance density function of the particle filter, aking the iportant density function saple of closer inspection the probability density of the saple, to iprove the efficiency of particle filter weakened particle degradation. Considering Internet of Things target tracking, specific regions of applications such as intrusion detection, localization and tracking prototype syste based on WSN was built. For the localization proble in WSN, the localization algorith based on obile beacon was designed and ipleented. And cobined with target tracking collaborative algorith, construct a prototype syste for target tracking, which can lay the foundation for further application of localization and tracking technology in WSN. Acknowledgeents Fig. 3. Randoly deployed sensor nodes and network connectivity graphs. This work is partially supported by the Defense basic research projects in 203 #A0420350, Funding Project for Departent of Education of Shaanxi Province in 203 #203JK67, Research and Cooperation Project for Departent of Yulin city in 202 #20SKJ09. Thanks for the help. References 6. Conclusions Fig. 4. Different particle nuber. In this paper, the theory and application of technology based on wireless sensor network target tracking conducted in-depth research, around of target tracking in wireless sensor networks in recent years carried out research work, three algoriths, and []. R. Wade, Mitehel W. M, F. Petter, Ten Eerging Technologies that will Change the World, Technology Review, 06,, 2003, pp. 33-49. [2]. N. Patwari, J. N. Ash, S. Kyperountas, Locating the nodes: cooperative localization in wireless sensor networks, IEEE Signal Processing Magazine, 22, 4, 2005, pp. 54-69. [3]. Reddy, Yenuula B. Secure packet transfer in wireless sensor networks, Sensors & Transducers, Vol. 4, Special Issue, No. 2, 202, pp. 308-320. [4]. Yang J, Xu M, Zhao W, et al., A ultipath routing protocol based on clustering and ant colony optiization for wireless sensor networks, Sensor, 0, 5, 200, pp. 452 4540. [5]. Aledar H, Ersoy C., Wireless sensor networks for healthcare: A survey, Coputer Networks, 54, 5, 200, pp. 2688-270. [6]. Du J., Chao Sh. W., A study of Inforation Security for M2M of IoT., in Proceedings of the 3 rd International Conference on Advanced Coputer Theory and Engineering, Chengdu, August 20-22, 200, pp. v3-576 v3-579. [7]. Kivelä, Ilkka, Design of networked low-cost wireless noise easureent sensors, Sensors &Transducers, Vol. 0, Special Issue, February 20, pp. 7-90. 60
Sensors & Transducers, Vol. 54, Issue 7, July 203, pp. 56-6 [8]. Jing Y, Mai X, Wei Zh et al., A ultipath routing protocol based on clustering and ant colony optiization for wireless sensor networks, Sensor, 0, 200, pp. 452-4540. [9]. Liu Q, Zh G H., Key technologies and applications of Internet of Things, Coputer Science, 37, 6, 200, pp. -4. [0]. Rolfh W., Internet of things-new security and privacy challenges, Coputer LAW&Security Review, 26, 200, pp. 23-30. 203 Copyright, International Frequency Sensor Association (IFSA). All rights reserved. (http://www.sensorsportal.co) 6