基於漸進式資料趨勢分析之無線感測網路節能技術

Adaptive Power-Saving Techniques for WSN based on Incremental Data Trend Analysis

林冠榮(Guan Rong Lin)

指導教授 ： 陳良弼

國立清華大學/電機資訊學院/資訊工程學系/碩士(2007年)

https://doi.org/10.6843/NTHU.2007.00305

摘要

無線感測器是使用電池當作它的電源，由於受限於所處環境，例如森林生態的監測，它們通常無法再充電。當在無法開源的狀況下，如何節流(省電)是相當重要的議題。

傳統的感測器資料管理系統，如TinyDB與TAG，收集資料的方式為定期地偵測環境變數並且回傳到資料管理系統。在一般的感測應用中，如溫度光度監測，一天內的溫度變化，可能會有週期性。傳統的省電機制著重於節省傳輸的耗電，並沒有利用此資料重複出現的特性。有鑑於此，我們提出一個針對歷史資料進行趨勢分析的演算法，並利用找出的週期特性，來動態的調整探測時間點，節省不必要的感測能源消耗，達到節省探測的耗損，同時也能節省傳輸的耗電。

我們的系統先收集一段時間資料，之後開始計算其週期性，利用此週期將歷史資料收集在一起，並分析資料的趨勢當作未來估計資料的依據。建立好歷史資料趨勢模型後，我們假定每個時間點上的資料分佈呈現常態分布，過去資料有些時間點的資料分佈太過離散，因此該時間點上的資料將無法有效的估算資料，所以未來相對應時間點必須要實際去探測資料；有些時間點的資料比較集中在平均值的附近，因此我們可以回報過去資料該時間點的平均值當作估計值，藉以代替實際探測的耗費。利用Chebyshev’s Inequality我們可以保證每一筆回報的估計值，在一定的信心水準之下，會小於一個誤差範圍。

實驗證明，在將近30%的省電率中，錯誤率僅不到3%，証明了我們的系統在省電率與錯誤率之間有非常好的平衡。因此，我們的系統應用在無線感測器網路上，將可有效的節省探測的耗電，同時也避免了不必要的傳輸。

關鍵字

感測器 ； 資料趨勢 ； 省電 ； 週期

並列摘要

Wireless sensor networks have received considerable attentions in recent years and played an important role in monitoring applications. Sensor nodes usually have limited supply of energy. Therefore, one of major design considerations for sensor network applications is to conserve the energy for sensor nodes.

In most sensor applications, communication is considered as the factor requiring the largest amount of energy. Therefore, existing approaches for conserving energy mainly focus on reducing the communication of sensor networks. However, as the applications of sensor networks continue to expand, we find that in some sensor applications sensing operations dominate the use of the energy, making existing approaches are not good for use.

Therefore, in this study, we propose a novel energy-conserving approach for sensor networks. Our approach builds on the observation that the values of the collected sensor data exhibit periodical patterns over time. We exploit the periodical patterns to construct prediction models for sensor data and use the constructed models to approximately answer queries over sensor networks.

In addition, we provide theoretical analyses for the use of the proposed approach, and show that a tight bound of the accuracy of reported value is guaranteed. Finally, we conduct a comprehensive experiment to validate the proposed approach. The experiment results show that our approach significantly reduces the sensing cost as well as the communication cost.

並列關鍵字

sensor ； trend ； power-saving ； period

參考文獻

[3]. Antonios Deligiannakis, Yannis Kotidis, and Nick Roussopoulos, "Compressing Historical Information in Sensor Networks" In Proceedings of the SIGMOD Conference, pages 527–538, 2004.

[5]. Amol Deshpande, Carlos Guestrin, and Samuel Madden, "Using Probabilistic Models for Data Management in Acquisitional Environments" In Proceedings of the Conference on Innovative Data Systems Research, page 317-328, 2005.

[7]. Mohamed G. Elfeky, Walid G. Aref, and Ahmed K. Elmagarmid "WARP: Time Warping for Periodicity Detection" In Proceedings of the International Conference on Data Mining, page 138-145, 2005.

[8]. Mohamed G. Elfeky, Walid G. Aref, and Ahmed K. Elmagarmid, "STAGGER: Periodicity Mining of Data Streams using Expanding Sliding" In Proceedings of the International Conference on Data Mining, page 188-199, 2006.

[9]. Johannes Gehrke and Samuel Madden, "Query Processing in Sensor Networks" In Proceedings of International Conference on Pervasive Computing, 2004.