A Localized Event Boundary Detection Algorithm forNon-Location-Aware Wireless sensor networks

• Main Authors: Chi-Lun Hung, 洪啟倫
• Other Authors: 林偉
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/75407757290191558317

碩士 === 國立中興大學 === 資訊科學與工程學系 === 96 === Event detection is a crucial issue is the research area of wireless sensor network. It deals with the deployment of sensor nodes for observing some events of interest in the environment under monitoring. Each sensor node is equipped with a sensing mechanism for detecting the occurrence of event. When an event of interest is detected, a sensor node should be able to rapidly report it through message transmission to the user. Upon receiving messages, the user can take proper action to put event development under control in the early stage. In this thesis, we propose a self-detecting algorithm for the nodes of a wireless sensor network to determine the event boundary in the presence of faulty nodes. The nodes exchange messages and execute the algorithm in parallel to determine whether they are located on the boundary of an event. Our algorithm allows the nodes to rapidly locate the event boundary and report it with minimum message transmission. In this thesis, we take into account the presence of faulty nodes and their impact on accuracy of boundary detection. Faulty nodes may spread error messages and affect neighboring nodes to give rise to incorrect report of event boundary. To circumvent this problem, we equip the proposed algorithm with a self-checking capability. Neighboring nodes periodically check on each other through calculating neighbor coefficient of variation difference (NCD). As soon as nodes are found faulty, they are shut down and messages originated from them are purged. This helps minimize the impact of faulty nodes on event boundary detection. We have conducted simulations to evaluate the performance of the proposed algorithms. We employed two metrics in the measuring the performance: false alarm rate and degree of fitting. False alarm rate is a probability of mistaking a node on the boundary of event. Degree of fitting indicates the completeness of detected boundary. Compared with Ding’s approach, the proposed algorithm improves degree of fitting by 15.4% and minimizes false alarm rate by 43.2%. Our simulation results show that the proposed algorithm performs well with low false alarm rate and high degree of fitting.