[Paper Review] Power-efficient Hierarchical Data Aggregation using Compressive Sensing in WSN

Xi Xu, Rashid Ansari+1 more|arXiv (Cornell University)|Oct 15, 2012

Sparse and Compressive Sensing Techniques9 references4 citations

TL;DR

This paper proposes a hierarchical compressive sensing (HDACS) architecture for wireless sensor networks that reduces energy consumption by adapting compression thresholds to local cluster sizes rather than global network size. By applying compressive sensing within multi-level clusters, the method achieves 37–77% energy savings over prior CS-based and non-CS aggregation methods, with significant gains in higher-level cluster heads due to localized data compression and reduced transmission overhead.

ABSTRACT

Compressive Sensing (CS) method is a burgeoning technique being applied to diverse areas including wireless sensor networks (WSNs). In WSNs, it has been studied in the context of data gathering and aggregation, particularly aimed at reducing data transmission cost and improving power efficiency. Existing CS based data gathering work in WSNs assume fixed and uniform compression threshold across the network, regard- less of the data field characteristics. In this paper, we present a novel data aggregation architecture model that combines a multi- resolution structure with compressed sensing. The compression thresholds vary over the aggregation hierarchy, reflecting the underlying data field. Compared with previous relevant work, the proposed model shows its significant energy saving from theoretical analysis. We have also implemented the proposed CS- based data aggregation framework on a SIDnet SWANS platform, discrete event simulator commonly used for WSN simulations. Our experiments show substantial energy savings, ranging from 37% to 77% for different nodes in the networking depending on the position of hierarchy.

Motivation & Objective

• To address the inefficiency of fixed, global compression thresholds in existing compressive sensing (CS)-based data aggregation in WSNs.
• To reduce energy consumption in large-scale dense sensor networks by leveraging hierarchical clustering and localized data compression.
• To improve data aggregation efficiency by basing compression on cluster size rather than global network size N.
• To validate the theoretical energy savings through simulation on the SIDnet-SWANS platform.
• To enable scalable, energy-efficient data gathering in correlated sensor data fields such as temperature and sound.

Proposed method

• The proposed HDACS architecture organizes sensors into a multi-resolution hierarchy of clusters across multiple levels, with cluster heads aggregating data from child nodes.
• Leaf nodes transmit raw data directly to their cluster heads, while intermediate cluster heads apply Discrete Cosine Transform (DCT) to achieve sparse signal representation.
• Random measurements are taken at each cluster head based on local cluster size $N_i$, with $M_i = K \log N_i$ measurements required for signal recovery.
• The DCT-based CoSaMP algorithm is adapted for signal recovery, exploiting sparsity in the DCT domain to reconstruct original data at the sink.
• Energy cost is modeled as proportional to the number of transmitted frames, with frame size $m$ factored into transmission cost comparisons.
• The framework is implemented and evaluated on the SIDnet-SWANS discrete-event simulator for 2D randomly deployed sensor networks of varying sizes.

Experimental results

Research questions

• RQ1 How does hierarchical clustering with localized compressive sensing reduce energy consumption compared to global CS-based aggregation in WSNs?
• RQ2 What is the impact of cluster size on the number of required random measurements and transmission energy in compressive data gathering?
• RQ3 To what extent can adaptive compression thresholds based on local data field characteristics improve energy efficiency in hierarchical data aggregation?
• RQ4 How does the proposed HDACS framework compare in energy efficiency to non-compressive sensing (NCS) and hybrid compressive sensing (HCS) methods?
• RQ5 What is the trade-off between signal recovery accuracy and energy savings when using DCT-based sparse representation and CoSaMP recovery in hierarchical clusters?

Key findings

• The proposed HDACS framework achieves up to 77% energy savings for cluster heads in the highest level compared to non-compressive sensing (NCS) aggregation.
• Energy savings range from 37% to 70% when compared to hybrid compressive sensing (HCS) methods, with the highest gains observed in higher-level cluster heads.
• Transmission energy consumption ratio (HDACS vs. NCS) is less than 0.5, indicating at least 50% energy reduction across all nodes.
• For a 400-sensor network, the ratio of transmission energy between HDACS and HCS is ≤ 0.633 for non-leaf nodes, confirming substantial savings in intermediate clusters.
• Signal recovery performance remains stable across different network sizes (300 to 700 sensors), with SNR unaffected by network size due to robust DCT-based CoSaMP recovery.
• The energy savings are robust to MAC layer frame size variations, with 50% savings preserved even when data is segmented into multiple frames, due to localized $\log N_i$ scaling.