A Discussion of Energy-Efficient Data Transmission with Data Compression
Written by Wei Zhang, Andrew Keong Ng, and B. Sivaneasan
Many real deployments of smart city systems do not have the luxury of unlimited resources. Energy supply is often among the critical system constraints that either set an upper bound of the system’s usability or force the system to embark on new designs and upgrades. Consider a typical scenario that data is collected by sensors with limited resources and pending data transmission from sensors to a much more resourceful host, e.g., server or the cloud, for further processing and analysis.
This brings some immediate questions to mind. Do the sensors have sufficient energy for data transmission? When should the data be transmitted, and at what speed? Should raw data be transmitted, or would certain data pre-processing before transmission be beneficial? Questions are not limited to the above ones, but they offer us some potential paths to optimize the energy efficiency of such a typical data transmission scenario. In this article, we will follow up those questions and present our research findings accordingly. Reflection and outlook will also be presented to inspire future design and upgrade of energy-efficient data transmission for smart city applications.
There are some general observations for data transmission and compression . For data transmission, energy dissipation accelerates with 1) fast transmission, i.e., transmitting much data during a time unit, and 2) poor network conditions, such as when the signal-noise-ratio is low. For data compression, energy usage normally is linearly proportional to how much data is to be compressed. Imagine we have some data in a sensor pending for transmission and intuitively we would like to know how urgent the transmission is and what the network’s condition is. If it is urgent with fortunately good network conditions, data shall embark on the journey of fast transmission without causing much energy panic. When it is not urgent and the network is not satisfactory, we may wait for something good to happen. However, “wait” is not always feasible; at least it cannot be too long. A feasible choice is to trade the expensive transmission energy with compression energy. By spending some energy to compress large raw data into compressed data of smaller size, the transmission workload can be reduced with less transmission energy usage. Considering the fact that compression can be done almost instantly and many simple yet efficient compression algorithms are available, saving transmission energy with compression energy seems quite promising. Following the idea of balancing compression energy and transmission energy, research problems, such as how to measure or estimate the system/operational condition precisely, how much data to compress (compressing all or none may not be optimal), and how to react to the changing network condition by adjusting transmission speed, have been discussed in many recent references where interested readers can gain more in-depth understanding.
Figure 1: Compression Fraction
Figure 1: The optimal total energy consumption for data transmission with data compression under the parameters specified in reference . With different compression, the fraction varies between 0% and 100%. The compression energy increases with compression fraction, while the transmission energy decreases. Neither compressing none nor compressing all can achieve the best total energy. The dot in the total energy curve corresponds to the optimal compression fraction.
Despite such an innovative idea, joint compression-transmission is yet prevalent in real deployments and industrial applications. Indeed, there are real concerns that are sometimes overlooked by the academic community. First, in the application layer, real-world applications may not be so energy sensitive for data transmission. They may have an abundant renewable energy supply, and data transmission energy is nearly negligible for the whole system, like in smart grid  and smart agriculture. Second, in the system layer, introducing additional factors, like compression in this case, adds complexity to the whole system. This incurs additional costs to the system for development and operation. Also, system reliability is a concern, and the industry often favors simplicity. The idea of saving some transmission energy through updating existing systems may not be sufficiently justified for many companies. Thirdly, in the hardware layer, not all sensors are sophisticated. Many of them cost just a few dollars or even a few cents each and expecting sufficient computing resources for compression and intelligent network interface for transmission is not that realistic. Even if both are available, jellying them together with appropriate algorithms and control units is challenging (e.g., design and implementation).
So, in any scenario, is it worth exploring joint compression-transmission for energy efficiency? The answer is yes. For urban traffic monitoring in smart cities, accessing energy supply and installing sensors isn’t easy, in some cases, like along the highway. Relatively the budget is not stringent, and each sensor can integrate hardware worth ten dollars or even more, offering necessary resources for computing, networking, control, and so on. It has the need of saving data transmission energy and meanwhile is capable to do so with data compression. On the consumer electronics side, wearables like Apple Watches or Microsoft Bands are another example. Elegant design matters greatly for commercial wearables where small size and long battery life are key design considerations. Advanced energy-efficient data transmission with compression is part of the technical profile of these wearables. This helps commercialize products and build up the brand.
Overall, technology development and adoption need to be attached to applications, and the technology discussed in this newsletter is no different. Data transmission with compression is promising but is only feasible and meaningful with the support of hardware, systems, and applications. As such, we also advocate industrial-relevant research to bring academic research, which is still dominant in nowadays publications, closer to real-world applications and challenges to make a real impact.
- Nguyen TT, Ha VN, Le LB, Schober R. Joint data compression and computation offloading in hierarchical fog-cloud systems. IEEE Transactions on Wireless Communications. 2019 Oct 4;19(1):293-309.
- Zhang W, Fan R, Wen Y, Liu F. “Energy optimal wireless data transmission for wearable devices: A compression approach.” IEEE Transactions on Vehicular Technology. 2018 Jul 25;67(10):9605-18.
- Chowdhury MR, Tripathi S, De S. Adaptive multivariate data compression in smart metering Internet of Things. IEEE Transactions on Industrial Informatics. 2020 Mar 17;17(2):1287-97.IEEE Transactions on Smart Grid 12, no. 4 (2021): 3570-3580.
This article was edited by Mehrdad Boloorchi.
Wei Zhang is currently an Assistant Professor with the Information and Communications Technology Cluster, Singapore Institute of Technology (SIT). Before joining SIT, he was a Scientist at the A*STAR, Singapore. From 2015 to 2017, he was a Research Fellow at NTU, Singapore. Dr. Zhang received a Ph.D. degree in computer science from the School of Computer Science and Engineering, Nanyang Technological University (NTU), Singapore, in 2015. Dr. Zhang serves as the Programme Leader of the Bachelor of Science with Honours in Computing Science programme. He also serves as an Associate Editor for the IEEE Internet of Things Journal (IoT-J) as well as the Area Chair, Track Chair, Review Committee Member, and TPC members of top conferences such as IJCAI, GLOBECOM, and ICME. His research interests lie in utilizing artificial intelligence (AI) for smart cities, especially smart grid and smart buildings, towards improved user experience and enhanced sustainability.
Andrew Keong Ng is an Associate Professor and a Programme Leader with the Engineering Cluster, Singapore Institute of Technology. He is also a Chartered Engineer with the UK Engineering Council and serves on the committees of various professional engineering institutions such as IEEE, IES, IET, and IRSE. He is a Senior Member of IEEE and IES; a Consultant and Advisor to startups and multinational corporations; and a Topic Editor of MDPI journals. Moreover, Dr. Ng is a Principal Investigator of several grants amounting to more than SGD 1.5 million. He holds one international patent and has over 30 publications as both first and corresponding author. His research and development innovations have garnered him several prestigious awards like the Outstanding Researcher Award, Teaching Excellence Award, Amity Researcher Award, Young Investigator Award, and Editor’s Choice Engineering Impact Award. Furthermore, Dr. Ng has been a keynote and an invited speaker at various international conferences.
B. Sivaneasan received B.Eng. and Ph.D. degrees in Electrical and Electronic Engineering from Nanyang Technological University, Singapore, in 2007 and 2012 respectively. In 2019, he joined Singapore Institute of Technology as an Assistant Professor. He is registered as Chartered Engineer with the UK Engineering Council. He has published over 25 technical papers and co-authored a scholarly book chapter on Vehicle-to-Grid (V2G) for a book titled "Energy Storage for Smart Grids: Planning and Operation for Renewable and Variable Energy Resources (VERs)". In addition, he also won the Best Innovation in Renewable Energy award at NI ASEAN Graphical System Design Competition for his work on a functional smart grid prototype. His work on IIoT based electrical asset management system at a local shipyard won the Best Paper Award at the IEEE ICPEA 2022 conference.