考慮有限誤差的無線感測網路資料壓縮與聚合

Abstract

感測節點在實際應用中常常布置於無法供給電力甚至無法進行更換電池的環境，因此如何節省能量消耗以延長感測網路時間成為重要的研究議題。因此，不少失真資料壓縮演算法被提出來，以減少資料傳輸的能量消耗，但這些失真資料壓縮演算法並無法控制誤差大小，難以用在有限誤差的實用問題上，因此本論文將提出有限誤差的資料壓縮演算法，以同時維持良好的資料壓縮率以及有限誤差控制。本研究針對無線感測網路的省電需求，利用資料相關性、時間相關性以及空間相關性等資料特性進行編碼，提出BEDCA (Bounded Error Data Compression and Aggregation in Wireless Sensor Network)。傳統的失真資料壓縮VQ (Vector Quantization)演算法跟DCT (Discrete Cosine Transform)演算法無法控制誤差大小，當容許誤差設定為0.5%，我們的方法可以較其節省至少70%的能量消耗；當容許誤差設定為1.0%，則可以較其節省超過90%的能量消耗。此外，跟 VQ相比，我們方法可以改進20% 壓縮率跟節省62%能量；與DCT相比則可以改進38%壓縮率跟52%能量。經實驗發現，若在傳送過程中就先進行壓縮以及聚合，可比在目的端節點才進行處理節省超過80%以上的能量消耗。本論文針對四種不同屬性的資料進行查詢評估，發現我們的方法之執行結果都較傳統方法為佳。然而時間相關性高的資料，其壓縮率可能會因為誤差的增加反而變差，原因可能歸因於我們所採取的codebook是靜態的，採取動態codebook應可解決此問題。

In this paper, an efficient data compression and aggregation method, called BEDCA, is proposed to reduce the size of transmission data under the given bounded error. We first apply the observed transmission data to construct a codebook which is related to the data correlation of the monitoring environment. Given a bounded error, the proposed method determines whether the new sensed data should be compressed or not by comparing it with the reference data such as the previous sensed data (for temporal correlation), the neighboring sensed data (for spatial correlation), and the codebook encoded data (for data correlation). Thus, the total size of transmission data can be minimized for energy saving. We use a real dataset to evaluate the performance of our mechanism. Even if the bounded error is set as a small value (under 0.5%), the proposed method can reduce a lot of the transmission data (over 70%) to cut down the total energy consumption. Our improvement exceeds 90% in the total energy consumed when bounded error is more than 1%. Compared to VQ, our proposed methods can enhance 20% better compression ratio and save 62% energy at least. Our method improve 38% compression ratio and retain 52% energy at least than DCT. Moreover, in our simplified hierarchy model of architecture, more than 20% energy can be saved in any aggregation function in SN than aggregated in sink. Experiment results show that the proposed method can make WSNs more efficient in energy consumption.