Research On Robust Tracking Methods Of Mobile Targets In Wireless Sensor Networks

Mobile target tracking is an important technique in Wireless Sensor Networks(WSNs), which can get the mobile target’s location and trace by multiple sensors’respective sensing and coordinated computing. Because of the uncertainty ofenvironment, robustness is the key problem of tracking system. Based on the previoustracking methods, this paper tries to improve the robustness of tracking algorithms,propose more robust and high efficient methods and pursue the balance betweenrobustness and optimization.According the characteristics of all the tracking methods, this paper divide thetracking algorithms into two categories—tracking on real-time localization and trackingon filtering. The former methods use current information of targets to localize and trackthem and suit for the tasks with high requirement of real-time; the later methods useboth current and previous information to tracking targets and suit for the tasks with highrequirement of accuracy. Based on the deep research of the two kinds of methods andcombining the requirement of robustness, this paper improves the former trackingmethods and proposed three algorithms with good robustness. They are robustlocalization based on grids, localization based on adaptive grids division and robustparticle filter.(1) Robust localization based on grids, based on the WSNs deployed in grid pattern,improves the low fault-tolerant single grid-level localization and raises an highfault-tolerant localization method using double grid-levels. This method is not onlyreal-time, but also decreases the location error caused by uncertainty.(2) Localization based on adaptive grids division, based on the nodes sequencematching algorithm, improves the traditional centroid computing methods which are oflarge computing quantity and complexity and come up with a new centroid computingmethods with changeable grid granularity to localization targets. In addition to the lowcomputing complexity, this algorithm can also be used to computing the centroid ofirregular targets estimated region, which shows stronger robustness.(3) Robust particle filtering develops some robust measures to improve thedegeneration in traditional particle filtering. On the one hand, it determines thesuggested distribution and chooses the particles based on the target localization results.On the other hand, it can restrain the degeneration of filtering by regular the particlenumber using particle degeneration index.At last, based on the sensor nodes, gateway nodes, computing centre and otherhardware devices, the paper designs and realizes the robust tracking prototype systembased on WSNs, and validates the three robust tracking methods under different scenes.Based on the experiments’ results, the paper concludes the feathers and suitable scenes for each tracking method, which lays emphasis on the implementation of robust trackingmethods in WSNs.