感測網路資料關聯性分群路由協定之研究

Research on Data Correlation Cluster-based Routing Protocols for Wireless Sensor Networks

洪瑛甫(Ying-Fu Hung)

指導教授 ： 林易泉

國立虎尾科技大學/電機資訊學院/資訊工程研究所/碩士(2012年)

https://doi.org/10.6827/NFU.2012.00126

摘要

無線感測網路是由多個電能有限並且具備無線通訊及感測能力的節點所組成的感測系統，此技術能將節點感測到的資料回傳至監控端，因此被廣泛應用於監控環境變化等用途。此項科技的最主要目標是可以從監控的區域獲得完整且精確的資料，但根據感測網路的相關文獻，很多路由協定都著重於省能或延長網路運作時間，反而感測環境變異的考量及感測資料準確性等議題卻很少被提及。在實際的感測網路中，最為精準的路由方式即為各節點皆傳輸資料至監控端，但此方式會導致節點電能迅速耗盡，不適用於長期監控；相對於為了省能而將某些感測資訊壓縮至一固定的封包長度，反而無法讓監控端獲得完整的資料，因此設計出一個高準確性且具備低耗能的路由協定是我們討論的重點。
本論文中，我們將區域性及時間性兩項因素納入考慮，分別是座標位置相近的節點其資料的相似性及隨著時間變化而節點感測到的資料差異性，並且提出一個基於分散式架構的叢集路由協定，此架構特性在於不必等待監控端發佈命令或給予大量感測節點的相關資訊，只需要透過局部性的相關資訊即可自行選出範圍內的叢集管理者，其自我組織的特性使網路具備高延展性，且也可在短時間內收集到相關資訊。
本協定中，在分群階段考量到上述區域性的因素，所以此階段的目標是將感測資料相近的節點歸類成同一叢集，因此各節點採用自我組織的方式利用歷史感測資料計算出該節點與涵蓋範圍內節點的資料關聯性，而資料關聯性最高的節點則擔任叢集首，其功能為資料收集與傳輸。而根據上述時間性的因素，其它非叢集首的節點則加入範圍內歷史資料差異最小的叢集，經由此種分群方式可使各叢集內的節點其資料差異程度盡可能降到最低，但是若擔任叢集首的節點能量即將耗盡，則會導致此節點較早死亡，因此若想延長節點死亡時間，本協定亦可將剩餘能量納入擔任叢集首考量的因素，但是剩餘能量考量的比重越高，叢集內的節點其資料關聯性則越低。但是資料的準確性依然重要，因此在資料傳輸階段時，為了讓監控端得知感測區域有無異常狀況，所以本協定設定了一個門閥值，確保資料精準度能控制在監控端可接受範圍內。假設某叢集內有一節點與叢集首所感測到的資料其差異超過預設的門閥值，則叢集首只需要傳送此節點的相關資訊及其它節點的座標資訊即可，藉此監控端可在不影響資料準確性的情況下推估出感測區域的情形，且叢集首也因不用傳送冗長的封包而達到省能的效果。但是若某一叢集因環境變化的因素而發生差異度超過門閥值的節點增多時，會導致叢集首傳送的負載量變重。因此本協定提出了一個重新分群的機制，主要用意為解決叢集首已不適任的問題，不適任的因素包含負載量過大、剩餘能量過低及叢集首死亡。因此適時的更換叢集首由適當的節點擔任，不但能降低傳送負載量以達到省能的目的同時也能延長網路生存時間。經由模擬運算的實驗結果，本協定在資料精準度、封包長度、叢集變異數及省能效果等效能指標都優於其它類似的協定。且由於重新分群機制也考慮到剩餘能量此因素，因此也能延長節點死亡的時間。

關鍵字

資料關聯性 ； 省能路由 ； 資料精準度 ； 分群

並列摘要

Wireless sensor network consist of a large number of nodes with limited battery power, wireless communication and sensing capabilities, and the data collected by each node can be transmit to the base station through this technology. Therefore, this technology has been widely used to environmental monitoring and other purposes. The most important goal of this technology is to obtain complete and accurate information from the monitored area, however according to the literature of sensor networks, many routing protocols are focused on energy saving or extend network life time, but the sensing of environmental variation and information accuracy issues were rarely mentioned. In actual sensor network, the most accurate routing protocol is each node transmits its own data to base station, but this manner will lead to energy of node depleted rapidly, so this manner does not apply to long term monitoring. On the contrary, the sensory information is compressed to a fixed packet length in order to energy saving, it leads to base station is unable to obtain complete information. Therefore, design a high accuracy and with low energy consumption of routing protocol is the topic in our research.
In this thesis, we take into consideration the factors of the spatial and temporal data; the former is the data similarity between the locations of nodes, and the latter is the difference between the information of a node change over time. We proposed a cluster-based distributed architecture routing protocol. The feature of this architecture is without waiting for commands or obtains information of large number of nodes from the base station; the cluster head can be selected only requires partial information of localized nodes. Its self-organizing feature allows the network with high scalability and also collects data in a short time.
In our research, we take into consideration the factor of spatial in clustering phase; the goal of this phase is classified nodes into the same cluster that has similar information. Therefore, each node calculates the data correlation with the neighboring nodes by use history readings. The node with highest data correlation acts as cluster head that is responsible for data collection and transmission. According to the factor of temporal, the other nodes join the cluster that has smallest differences of history readings between them. Through this clustering manner, the difference of information of all nodes in each cluster can be minimized. If the residual energy of the node is low, the job of cluster head will cause this node dead early. So if we want to extend the survival time of node, we perhaps can take into consideration the factor of residual energy for cluster head selection. However, the higher proportion of the residual energy, it leads the lower data correlation in cluster. The accuracy of the information is still important when the data transfer phase, in order to know any abnormal condition from area, so we set a threshold to ensure data accuracy can be controlled within an acceptable range of base station. Assume the difference exceeds the threshold between data of a node and data of cluster head, and then the cluster head only need to send the information of this node and the coordinate information of other nodes. According to this manner, base station can estimate the information of sensing area under without affecting the accuracy of information, and cluster head can send shorter packet to achieve energy saving. When environment factor changing, the number of nodes which difference defined above exceeds threshold will increase in a cluster. The traffic of cluster head will be overloaded with the increasing. Therefore, we propose a re-clustering mechanism in our protocol; the main goal is to improve the appropriateness of cluster head selection and the considering factors include traffic loading, residual energy of node and the death of cluster head. Therefore, timely replacement of the cluster head not only reduces the traffic loading to achieve energy saving but also extends the network survival time.
By simulation results, the proposed protocol can perform better than other routing protocols in terms of variance of cluster, data accuracy, packet length, and energy-saving. And re-clustering mechanism take into consideration the factor of residual energy, it can extend the time of node death.

並列關鍵字

Data correlation ； energy-saving ； data accuracy ； clustering

參考文獻

[2] K. Romer, F. Mattern, “The Design Space of Wireless Sensor Networks,” IEEE Wireless Communication, vol. 11 no. 6, pp.54-61, 2004 December.

[3] K. Akkaya, M. Youngish, “A Survey on Routing Protocols for Wireless Sensor Networks,” Ad Hoc Network, vol.3, issue.3, pp325-349, 2005.

[4] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam and E. Cayirci, “Wireless Sensor Networks: A Survey,” Computer Network, vol.38, issue 4, pp393-422, 2002

[7] M. A.. Hamza, S. M. Youssef, and S. F. Fayed, “A Distributed Energy Efficient Query Processing in Self-Organized Wireless Sensor Networks,” Proceedings of the World Congress on Engineering 2007, vol. II, pp.1482-1487, 2007.

[11] S. Lindsey, C. S. Raghavendra, “PEGASIS: Power-Efficient Gathering in Sensor Information System,” IEEE Aerospace Conference Proceedings 2002, vol. 3, pp.1125-1130, 2002.