Wireless sensor and event localization
2017-02-17T02:13:38Z (GMT) by
Localization, in the form of pin-pointing a randomly deployed sensor using location-aware anchors beaming beacon signals or confining a random event to a small boundary using sensors having overlapped sensing areas, is a challenging research problem. Different signal processing and estimation techniques are available that vary in terms of localization accuracy and resource efficiency. Accessibility of the region of interest (ROI) and requirement of localization quality govern the choice of techniques applicable in a specific scenario. Considering these factors, this thesis proposes different localization techniques for different scenarios. For impassable ROIs, where mobile anchors cannot be deployed and thus localization is vulnerable to malicious attacks, an attack-resistant decentralized sensor localization scheme is developed. Range-based techniques, where absolute distances are computed from anchors, have to be favoured despite their susceptibility to environmental noises as the alternative range-free techniques cannot achieve acceptable localization accuracy with only static anchors. The potential malicious attacks disguised in unavoidable wireless signal attenuation have been successfully mitigated by exploiting planer geometry and the law of large numbers. For passable ROIs, range-free techniques are preferred for their higher tolerance against environmental/measurement noises and lower hardware cost. In addition, terrain accessibility allows mobile anchors to enjoy the opportunities of using additional spatial information in localization and controlling mobility and beacon transmission pattern to optimize the costs while satisfying user/application demand of localization quality. To avail these opportunities, a computationally-efficient arc-coding based technique is developed to accurately calculate the boundaries of all non-overlapping sectors of a system of overlapping disks to effectively minimize the expected worst-case localization error for uniformly distributed random sensor deployment. For event localization, which is becoming a highly significant problem due to hardware advancement of long sensing range, this thesis proposes the first range-free technique, to the best of our knowledge, using our arc-coding based accurate sectoring technique. This has been possible by developing a novel probabilistic sensing range model that guarantees a user defined event detection probability and efficiently using a small number of mobile sensors in collaboration with static ones to also effectively guarantee user defined event localization quality. Superiority of all the proposed localization techniques against the respective state-of-the-art has been established with analytical models, whenever possible, and comprehensive simulation results in realistic environments.