Research On Target Tracking Under Uncertainty In Wireless Sensor Networks

Mobile targets’ tracking in wireless sensor networks plays a very important role inmany military and civilian areas. However, uncertainty, such as environmental noise,environmental terrain, sensing resolution, the randomness and suddenness of targets’mobility as well as the dynamic error tolerance of queriers are all different appearancesof uncertainty. Therefore, under the impact of ubiquitous uncertainty, increasing thereliability and robustness of target tracking algorithm, optimizing target tracking taskscollaboration and data management strategies and improving the performance of targettracking sensor networks have been a concern. Traditional target tracking methods aredesigned to be tolerant of the uncertainty by calibrating or filtrating the lowranging-quality sensing result, or compromising the flexibility and practicality, orconsuming more network energy. However, these methods failed to address the positiveeffects of the uncertainty on target tracking, losing sight of the fact that uncertaininformation can also promote the tracking efficiency.In order to improve the target tracking quality under uncertainty and enhance theoverall performance of sensor networks, this paper proposed an efficient and practicaltarget tracking methods under uncertainty in wireless sensor networks from the localaspect of target tracking algorithm and global aspect of target tracking managementstrategy. The main contributions of our work are as follows:(1) Survey the related works deeply and comprehensively. As there are a largenumber parallel approches about target tracking in sensor networks, we firstlysummarize the local target tracking methods by comparing them from four aspects: thecomplex and performance, deployment requirements, flexibility and practicality as wellas the tolerance of uncertainty. Besides, global target tracking management strategiesare also surveyed, including task management protocols with sleep/wake-up mechanismor efficient data aggregation, and data management protocols with push-pull balance ornetwork cost compromising. This clarifies the objective of our work.(2) Relate models and research framework of target tracking underuncertainty are introduced. Firstly, the related target detection model, target mobilityand sensor network structure are introduced. Then according to the structural features ofheterogeneous sensor networks as well as the Voronoi diagram based clustering method,a cluster-based framework of double-level heterogeneous sensor networks is proposed,which is expected to be the overall network architecture of this paper. On this basis, abrief description of local intra-cluster target tracking algorithm and the globleinter-cluster target tracking tasks collaboration and data management protocols is given.(3) A target tracking method with pairwise uncertainty of RSSI (TPU-RSSI) isproposed. By analyzing unreliability as it occurs in the uncertain area of pairwise RSSI sensors, intersection of areas’ boundaries divides the monitor area into faces withuncertain characteristics, and the uncertainty of target sensing can also be obtained bymultiple grouping samplings, so that the target tracking method is transformed into avector matching process. This improves the flexibility of the target tracking and reducesthe impact of environmental factors to improve tracking accuracy. Besides, consideringthe potential problems (high computational complexity, difficult geometric division andnone/multiple matching results), the heuristic matching algorithm, the approximate griddivision,the fault tolerance of TPU-RSSI and its quantitative extension are furtherdiscussed. Finally, the simulation results show that TPU-RSSI has outstanding trackingperformance under uncertainty.(4) A target tracking method with the uncertainty of nonideal binaryproximity (NBP) is proposed. Tracking area is divided by the nonideal binarydetection model into several faces with a signature vector, which contains theuncertainty of binary proximity sensing. Then finding the face which has the maximumsimilarity with the binary sensing vector can locate the target effectively. Furthermore,the problems such as, the uniqueness of the face with same signature, obtaining trackingtrace, impact of false positive detection, is discussed. Hereafter, the tracking error ofNPB methods is analyzed and evaluated theoretically, which clarify the deploymentrequiments, burst sensing conditions and other influencing parameters. Finally, theavailability of NBP method has been validated by expanded simulation experiments.(5) A prediction-based energy-conserving approximate management strategywith querier’s error-tolerance (P-EASE) is proposed. P-EASE strategy utilize thespatial-temporal correlation of targets cleverly, sector whose central angle is2/3isadopted as the approximate area instead of circle to adapt to the linear trend predictingmethod, which can reduce the location error effectively without increasing the messagecomplexity of the network combined with the user’s dynamic error tolerance.Furthermore, the optimal query schema is introduced to reduce the additional costcaused by choosing improper destination of querying request, and reduce the hotspot’soverhead at a certain extent as well. Theoretical analysis and simulation experimentresults show that P-EASE’s target location precision, energy efficiency and load balanceshould be significantly greater than parallel works.(6) According to the aforementioned methods or technologies, a prototype oftarget tracking system under uncertainty is designed and applied, which validatethe feasibility and effectivity of the system in real environment with various uncertainty.