Energy-Efficient Moving Objects Tracking In Wireless Sensor Networks

In recent years, wireless sensor networks (WSN) have permeated a plethora ofapplication domains, due to the ability of the constituent nodes to self-organize in awireless network in addition to simply sensing and performing local calculations. This,in turn, enables their deployment in various environments, where they can observe andgather data of interest for scientific, traffic management, environmental safety/hazards,infrastructure, health-care and military applications. One of the canonical problems inWSN settings is the tracking of mobile objects and its related issues, includingimproving the accuracy of the tracking process and trading off the quality of thetracking information for energy savings.This paper analyzes and evaluates the emerging approaches proposed foraddressing various aspects of moving object tracking in WSN, especially the algorithmsbased on particle filter. To overcome the problems in existing methodologies, weproposed a series of solutions in targets tracking. Along these lines, the maincontributions of this work are as follows:1. To address the localization accuracy problem in single target tracking, wepresented an Unscented particle filter based Moving Object Tracking algorithm, UMOT,in WSN settings, where the sensor nodes are clustered dynamically to provide sensingand data fusion tasks. The key idea of UPF is to capture accurately the posterior meanand covariance of non-linear Gaussian variable up to the second order, throughpropagating a set of sample points in the state system.A particle filtering based method, which we will refer to as Dead-Reckoning basedParticle Filtering (DRPF) is presented for recursively calculating target’s angulardeflection, which, by leveraging the dead-reckoning prediction within permissibleerror-bound, can significantly reduce the computation and communication overhead ofstate estimation.2. Another problem considered in this work is the issues related to selectingtracking principals–which is, the nodes with two special tasks:(1) coordinating theactivities among the sensors that are detecting the tracked object’s locations in time; and (2) selecting a node to which the tasks of coordination and data fusion will be handedoff when the tracked object exits the sensing area of the current principal. We observethat in many WSN applications in which sensing/sampling needs to be combined withmulti-hop transmission and, possibly, in-network aggregation, the typical processing isorganized in synchronized intervals, called epochs. We postulate that taking thesemantics of the epoch into consideration is important when selecting trackingprincipals and we present SLS (Sampling Look-ahead Selection), an efficientalgorithmic solution towards this goal.Extending the existing results which based the respective principal selectionalgorithms on the assumption that the target’s trajectory is approximated with straightline-segments, we consider more general settings of (possibly) continuous changes ofthe direction of the moving target. We developed DAS, Deflection Aware trackingprincipals Selection, an approach based on particle filters to estimate the target’sangular deflection at the time of a hand-off, and we considered the trade-offs betweenthe expensive in-node computations incurred by the particle filters and theimprecision-tolerance when selecting subsequent tracking principals.3. This work presented approaches towards addressing the problem of managingthe sensor-coverage and organizing the epochs in a manner that balances the trades-offsbetween the accuracy and energy consumptions during target tracking in WSN. Whilethe typical target tracking approaches are based on movement prediction, we onlyassume a knowledge of some maximal speed of the target during certain time-intervals.This, in turn, restricts its whereabouts to a disk-bound area throughout such intervals. Insuch settings, we seek to determine a sensor cover, a subset of all the nodes that need tobe awake, which ensures that the target can be detected during the given epoch.Towards this, we propose two sensor-cover selection methodologies, Greedy UncertainMoving Object coverage sensor set selection (GUMO) and PAttern Based coveragesensor set selection (PAB). GUMO is a heuristic maximizing the coverage gain at eachstep, while PAB is an approach based on optimal deployment pattern of sensor nodesachieving coverage of the disk area bounding the target’s whereabouts.4. Designing energy efficient scheduling mechanism is a challenge in WSNtracking scenarios due to the limitations on target’s movement prediction, and lack ofglobal network knowledge. Another observation of this work is that task conflicts and channel congestion preclude the utilization of the nodes shared by common trackingtasks, which may result in poor Quality of tracking (Qot) and/or increasing targetambiguity. In order to tackle this problem, we propose a lightweight sensor schedulingpolicy–Synchronization based Sampling Reduction (SSR), which explicitly prunes theredundant measurements in the conflicting nodes without decreasing Qot, throughsynchronizing the tracking tasks. In addition to conserving the energy by reducing thesamplings, SSR also is capable of mitigating the data associating problem in multipleobjects tracking, without requiring any global knowledge about the network.5. We considered some of the approaches used by the moving objects databasesand computational geometry communities, and we demonstrate that with appropriateadaptation, they can yield significant benefits in terms of energy savings and,consequently, lifetime of a given WSN. Towards that, we developed distributedvariations of three approaches for spatio-temporal data reduction–two heuristics,Dead-Reckoning and the Douglas-Peuker algorithm, and a variant of a computationalgeometry based optimal algorithm for polyline reduction. In addition, we examinedifferent policies for managing the buffer used by the individual tracking nodes forstoring the partial trajectory data. Lastly, we investigated the potential benefits ofcombining the different data-reduction approaches into “hybrid” ones during tracking ofa particular object’s trajectory. Our experiments demonstrate that the proposedmethodologies can significantly reduce the network-wide energy expenses due tocommunication and increase the network lifetime.We present the details of design and implementation of the proposed algorithms.Extensive experimental evaluations of our approaches were conducted, demonstratingthat combing these methods can yield a significant improvement during the trackingprocess in WSN.