[Paper Review] To Compress or Not To Compress: Processing vs Transmission Tradeoffs for Energy Constrained Sensor Networking

Davide Zordan, Borja Martinez+2 more|arXiv (Cornell University)|Jun 11, 2012

Underwater Vehicles and Communication Systems25 references19 citations

TL;DR

This paper investigates whether lossy compression reduces energy consumption in energy-constrained wireless and underwater sensor networks by comparing processing vs. transmission energy costs. It finds that lightweight temporal compression (LTC) saves energy in wireless networks, while DCT-based methods are better for underwater networks due to higher transmission energy, with analytical formulas provided to assess tradeoffs based on hardware and signal parameters.

ABSTRACT

In the past few years, lossy compression has been widely applied in the field of wireless sensor networks (WSN), where energy efficiency is a crucial concern due to the constrained nature of the transmission devices. Often, the common thinking among researchers and implementers is that compression is always a good choice, because the major source of energy consumption in a sensor node comes from the transmission of the data. Lossy compression is deemed a viable solution as the imperfect reconstruction of the signal is often acceptable in WSN. In this paper, we thoroughly review a number of lossy compression methods from the literature, and analyze their performance in terms of compression efficiency, computational complexity and energy consumption. We consider two different scenarios, namely, wireless and underwater communications, and show that signal compression may or may not help in the reduction of the overall energy consumption, depending on factors such as the compression algorithm, the signal statistics and the hardware characteristics, i.e., micro-controller and transmission technology. The lesson that we have learned, is that signal compression may in fact provide some energy savings. However, its usage should be carefully evaluated, as in quite a few cases processing and transmission costs are of the same order of magnitude, whereas, in some other cases, the former may even dominate the latter. In this paper, we show quantitative comparisons to assess these tradeoffs in the above mentioned scenarios. Finally, we provide formulas, obtained through numerical fittings, to gauge computational complexity, overall energy consumption and signal representation accuracy for the best performing algorithms as a function of the most relevant system parameters.

Motivation & Objective

• To evaluate whether lossy compression reduces overall energy consumption in energy-constrained sensor networks.
• To compare processing and transmission energy costs across different compression algorithms and network types (wireless vs. underwater).
• To identify which compression techniques are most effective under varying hardware and signal statistics.
• To derive analytical models for computational complexity, energy consumption, and reconstruction accuracy as functions of key system parameters.

Proposed method

• The authors analyze multiple lossy compression techniques, including Lightweight Temporal Compression (LTC), DCT-based low-pass filtering (DCT-LPF), and others, using real-world sensor datasets.
• They model energy consumption as a function of transmission energy per bit ($E_{Tx}^\prime$), micro-controller energy per cycle ($E_0$), and algorithm-specific computational cost ($N_c$).
• A key analytical inequality is derived: $\frac{E_{Tx}^\prime}{E_0} > \frac{N_c(\eta)}{1-\eta} = \frac{\alpha\eta + \beta}{1-\eta}$, determining when compression saves energy.
• Numerical fitting is used to derive parameters ($\alpha$, $\beta$) for $N_c(\eta)$, enabling prediction of computational cost and energy savings.
• The framework is validated across two scenarios: wireless (CC2420 radio) and underwater (AquaModem acoustic modem), with hardware-specific energy ratios.
• The method enables system designers to assess whether a given compression algorithm is energy-efficient for a specific hardware and signal type.

Experimental results

Research questions

• RQ1 Under what conditions does lossy compression reduce total energy consumption in sensor networks?
• RQ2 How do processing and transmission energy costs compare across different compression algorithms in wireless and underwater networks?
• RQ3 Which compression techniques—LTC or DCT-based—provide better energy savings in each network scenario?
• RQ4 How can computational complexity, energy consumption, and signal accuracy be quantitatively related to system parameters like compression ratio and hardware characteristics?

Key findings

• In wireless sensor networks using CC2420 radio, LTC provides energy savings only for compression ratios up to $\eta \leq 0.6$, while DCT-LPF is never energy-efficient due to high processing cost.
• In underwater networks with high transmission energy ($E_{Tx}^\prime = 10$ mW), both LTC and DCT-LPF provide energy savings for almost all values of $\eta$, due to the much higher $E_{Tx}^\prime / E_0$ ratio.
• The energy-saving threshold is determined by the inequality $\frac{E_{Tx}^\prime}{E_0} > \frac{\alpha\eta + \beta}{1-\eta}$, which depends on hardware and algorithm parameters.
• For LTC, the fitting parameters are $\alpha = 0.17$, $\beta = 0.08$, enabling prediction of computational cost and energy savings.
• For DCT-LPF, the parameters are $\alpha = 0.58$, $\beta = 0.32$, showing significantly higher computational cost.
• The derived formulas allow designers to evaluate compression efficiency for new hardware architectures by comparing $E_{Tx}^\prime / E_0$ with the algorithmic threshold.