| Summary: | 碩士 === 國立清華大學 === 資訊系統與應用研究所 === 94 === In sensor networks, sensors usually do not equipped with global IDs similar to IP addresses or MAC addresses in IP-based networks to be its identifier due to the application-driven usage and the cost of massive production. However, knowledge of physical locations of sensor nodes is essential for many geographic aware applications, such as target tracking, coverage management, and directional flooding, etc. Accordingly, many researchers have proposed several localization algorithms. They often assume some sensors are equipped with GPS devices, and the other sensor nodes can estimate their location through certain of distance computations.
Considering the tradeoff between computation complexity and precision of localization algorithms, in this study, we propose a grid-based approach to fulfill the requirements. In fact, instead of making effort to calculate all sensors’ locations directly, we would rather establish a grid system of partial nodes which are approximately deployed in uniform. Therefore, other sensor nodes can reckon which grid they belong to by some judgment with the distance table maintain by each grid node. Finally, we assign each sensor node a unique ID according to their location information.
The average estimated error of grid sensor node is decreased when there are more nodes involves or with higher grid dimension. The average estimated error is under 0.45 units if the sensors are deployed in a 100´100 unit space. Moreover, the localization and ID assignment of non-grid nodes is efficient once a new sensor node is added in the network. About 80% of non-grid nodes have correct predicted result.
|
|---|
