Đang tải... (xem toàn văn)
JOURNAL OF SCIENCE & TECHNOLOGV • No 95 2013 USING ENERGY EFFICIENTLY WITH REGIONAL MONITORING MODEL IN WIRELESS SENSOR NETWORKS SU'''' DUNG HIEU QUA N A N G LUONG \ti\ MO HINH GIAM S A T THEO VUNG TRONG[.]
JOURNAL OF SCIENCE & TECHNOLOGV • No 95-2013 USING ENERGY EFFICIENTLY WITH REGIONAL MONITORING MODEL IN WIRELESS SENSOR NETWORKS SU' DUNG HIEU QUA N A N G LUONG \ti\ MO HINH GIAM S A T THEO VUNG TRONG MANG C A M BIEN K H O N G DAY Trung Dung Nguyen, Van Due Nguyen, Ngoc Tuan Nguyen, Tien Dung Nguyen Hung Tin Trinh, Tien Dat Luu Ha Not University of Science and Technology Received June 18, 2012; accepted January 16, 2013 ABSTRACT Nowadays object tracking application is a very hot topic on the world But what we have to to use energy efficiently in this application? Normally, to track an object we often tum on all sensorsio the networi< So It makes energy consumption and more redundant data This monitoring model is called entire network monitonng model in using energy efficiently why we don't tum off unnecessgit sensor nodes It means only nodes around the object are turned on to collect Information ofobjectml other nodes are turned off to save energy It reduces total of energy used in netwodt This modelis called regional monitonng model In this paper we will describe two models In detail; simulate ami evaluate efficiency in using energy between two models Keywords' Regional Monitoring, LEACH-C, Wireless Sensor Network, Energy Efficiency, Particle Filter Wireless Sensor Networks (WSN) - m^ng cim biin khdng diy, dang din Ii6 thinh mit cing cu hOu hiiu cho cic ling dung giim sit Met s6 d6 phii ki din Cmg dung theo vit dii tuong, chu di dang duac quan tim nghien cCru nhiiu thiri gian gin diy Bii v6i bii toin theo idl dii tuong cich dan giin nhit li tram gic (Base Station) thu thip dir liiu cua toin ba cic node ch biin mang sau d6 sir dung thuit toin dw doin di xie dinh vi tri cOa dii tu? hinh niy so v6i md hlnh giim sit theo toin bd mfng ciia m?»J cim biin khdng diy hong theo vit dii tugng INTRODUCTION Wireless sensor network is a network that includes many sensor nodes Sensor nodes use wireless link to transmit data to each other and sink node They sense, collect information about environment and calculate based on the received information to make decision what we have to In wireless sensor network a large amount of sensor nodes are expanded on the large geography All nodes have to sense and send sensed information to the sink node The sink node will analyze this information and show result So every device in this network operates continuously It makes energy anii bandwidth consumed In our model only some nodes in regional monitoring have to operate to collect the information and transmit to the sink node So energy is only used on these nodes And the number of packets transmitted is less So we use energy and bandwidth link more efficiently We will analysis our model clearly later The rest of this paper is organized as follows Section II shortly discusses particle filter algorithm [12], a predictable algorithm to JOURNAL OF SCIENCE & TECHNOLOGV « No 95 - 2013 stermine current and next posidon of object In ^e next section we discuss about LEACH-C rotocol [1] [11], a routing protocol in wireless ''^nsor network In the section IV, we propose ;giona! monitoring model Section V discusses ntire network monitoring model And finally, Ve present simulation results, comments and valuations to compare between the two models nd suggest the future works for more optimal esults ; PARTICLE FILTER ALGORITHM Before the Particle Filter (PF) algorithm vas released to solve the problem of object racking in WSN, some algorithms such as Caiman Filter (KF) algorithm [14], Extended Caiman Filter algorithm [14] had been egarded as the key for this problem ^ In essence, the aim of Kalman Filter algorithm is to solve the problems in linear Ksystem which has noise in accordance wilh the Gauss distribudon And this algorithm solves that problem very well But when the system has a non-linear or non-Gauss distribution iinoise, the Kalman Filter algorithm is not able 'do well So other algorithms were proposed to •solve those problems One of them is Extended 'Kalman Filter algorithm (EKF) [14] It is '•improved based on Kalman Filter algorithm In l^the Extended Kalman Filter algorithm, before ;:applying the Kalman Filter formulations, the system equations are linearized by a Taylor lexpansion to convert a nonlinear function into a ,isequence of linear functions In other words, Taylor expansion is the key to apply Kalman -lalgorithm in non-linear and non-Gauss distribution systems This method solved the limitations of traditional Kalman Filter initially (However, it also has some restrictions, that are i difference in the results caused by approximate calculation in the system linearization process, and at the same time it requires more complex calculations than traditional Kalman Filter due to the linearization non-linear functions Until , the Particle Filter algorithm has been taken into ; use, these disadvantages can be solved Because i most of the motion process of an object is a , non-linear process and the noise of the sensors measurement is not always in accordance with , the Gauss distribution, the Particle Filter will be very useful when applied to object tracking problem To compare PF to EKF, we can see the graph in Figure That is the result of the simulation the tracking process of EKF and PF on MATLAB In which, the red dots depict the real state of an object in accordance with a prior non-linear equation, the green line represents the tracking (estimate the change state process) of object using EKF, the blue line represents the tracking of object using PF Through this graph, we can see the tracking of PF is more accurate than the tracking of EKF Time step Fig Estimation results ofPF and EKF Particle Filter is an algorithm which is used to solve the estimation problems, such as in the object tracking applications The basic operation of PF is based on maintaining a set of "particles" - these particles are considered as candidate points in the description of the object's position, to estimate the position ofthe object The set of particles is in fact a set of samples which are generated by a prior probability density function More specifically, assuming that after estimating the position of an object at moment t, we will have to continue to estimate its position at moment t+1 Based on the position estimated at moment t, the algorithm will generate a set of particles in some points which can be the estimated position of the object at moment t+1 The particles will be bom in a certain area, so we could zone an area where the object can move to at the moment t+l that limits the region tracking of object (Figure 4) JOURNAL OF SCIENCE & TECHNOLOGY * No 95-2013 reduces data transmission distance, therefore, energy can be saved estimated position at moment t + limited region to monitoring estimated position at moment t Fig Description of object position estimation by Particle Filter Instead of turning all the sensor nodes over the network on, sink node just turns on some sensor nodes around the region containing the set of particles Then based on the data measured by these nodes Particle Filter algorithm calculates and assigns a weight to each particle If these weights are smaller than a threshold, a new set of particles will be generated and recalculated the weights Finally, based on the positions ofthe particles and their weights, the Particle Filter algorithm calculates and gives an average position which is the estimated position ofthe object at moment t+1 The process will be repeated so we can track the object continuously LEACH-C PROTOCOL When WSN is deployed on a large area, the sensor nodes that transmit direcdy to the BS will consume more energy, and can cause redundancy of the information in the nodes in the same area Therefore, Modeling Data Centric [13] and LEACH (Low Energy Adaptive Clustering Hierarchy) protocol [13] was developed to improve this problem LEACH operation is based on clustering algorithm and has the characteristics as follows' nodes can self-configure cluster formation [13], nodes themselves will decide to be in which cluster and select a Cluster Head (CH), CH controls data transmission from the nonnal nodes to CH within its cluster Then it gathers the data and sends to the BS So, we can avoid redundant data And the nodes from far the BS \ \ Owtw; - Fig Modeling Data - Centric Figure represents modeling Data Centric In this figure the sensor network is divided into four clusters which are represented by four dashed circles Each cluster has a Cluster - Head (CH) which gathers data from the normal nodes and sends to the BS LEACH-C (LEACH Centralized) is developed from LEACH protocol [1] Unlike LEACH, when forming networks, BS will require the nodes within the network to send information about its location and eneigy Then, BS determines its monitoring area from the location and energy of the nodes BS executes an algorithm to form clusters and selects CH within each cluster CH implements gathering data in each cluster Finally, BS collects and processes data from the CH The CH nodes have the more tasks and roles so it would be the nodes with higher energy Ene^ consumption will be distributed evenly for better energy saving and longer netwoA lifetime Clearly, in LEACH-C, BS forms cluster and selects CH to reduce the power used for processing and calculating to form cluster for nodes On die other side, when BS uses an object tracking algoridim, its proactive control over cluster is very convenient for object tracking Therefore, LEACH-C is an optimal solution Next section, we present the detailed JOURNAL Oil SCIENCE & IECHNOLOGY • No 95 - 2013 )peration of LEACH-C protocol in regional nonitoring model I REGIONAL MONITORING MODEL Based on the results from tracking object dgorithm (Particle Filter algorithm), BS will turn on an area with minimal sensor nodes to track object Therefore, this method will help to reduce energy consumption across the network as well as improving the lifetime of the sensor network because we not need to tum on all the sensor nodes The protocol which is used to centralize data from nodes is LEACH-C protocol The steps of regional monitoring mechanism as follows: At the initial time, all sensor nodes are in idle mode After receiving an advertising (ADV) message from BS, this message includes information about the position of region that needs to tum on to track object - monitoring region, nodes will compare its current position to the position ofthe monitoring region If node is inside the monitoring region, it will change to active mode; whereas, if node is outside the monitoring region, it will remain the idle mode After changing its mode, nodes will send STATUS message to BS, this message includes information about its position, current energy and ratio of energy over distance to BS After receiving sufficiently number of STATUS messages corresponding to the number of nodes in the monitoring region, BS divides the region into smaller clusters, selects the cluster head node for each cluster A sensor node whh the highest ratio of energy to distance from it to BS will be selected as the cluster head After dividing the cluster, selecting cluster head, BS sends CLUSTERHEAD message to all sensor nodes in monitoring region, this message includes cluster ID, cluster head ID and number of nodes in a cluster After receiving a CLUSTERHEAD message from BS, node compares its information to the information in CLUSTERHEAD message and then it gives the appropriate treatments If node is a cluster head, it will broadcast TDMA message to all sensor nodes in its cluster TDMA message includes information about time slot for all sensor nodes in its cluster TDMA message will help other nodes in the cluster can send their data in appropriate time slot in order to avoiding collision in cluster If node is not a cluster head, it will wait until receiving a TDMA from its cluster head node After receiving a TDMA message, node will find its time slot in TDMA message and send DATA2CH message with its sensed information to cluster head in its time slot After cluster head node receiving sufficiently number of DATA2CH corresponding to number of nodes in cluster, cluster head node will centralize data from DATA2CH messages into one message that is DATA2BS message DATA2BS message includes all sensed information of all nodes in a cluster After that Cluster head node sends DATA2BS message to BS After receiving sufficiently number of DATA2BS message corresponding to the number of cluster, BS will process received data and predict the current position and the next position of the object by object tracking algorithm (e.g Particle Filter) Then, BS will give treatment about new position of monitoring region for the next round and send ADV message to all sensor nodes in the network Base on new ADV message node will know it is in the monitoring region or not If it is in the monitoring region it will continue to sense and send back the information to the BS using LEACH-C routing protocol as step by step presented before And the process continues like that to track the object in the network ENTIRE NETWORK MONITORING MODEL For the entire network monitoring mechanism, all sensor nodes will always in active mode They sense the information about object and send back to the BS using LEACHC protocol The BS will process received data and predict the current position and the next position of the object by object tracking algorithm After amount of time, all nodes continue to sense and send back the sensed information to the BS using LEACH-C protocol again The steps in the entire network monitoring model and in the regional monitoring model are similar But in the entire JOURNAL OF SCIENCE & TECHNOLO network monitoring model the BS does not determine monitoring region as in the regional monitoring model So all nodes in the network sense the information about object and send back to the BS The BS runs object tracking algorithm to determine the current position and the next position ofthe object Therefore, it will lead to energy waste and lifetime of network will be reduced very fast EiW iH