Compression Methods for Microclimate Data Based on Linear Approximation of Sensor Data

Abstract

Edge computing is currently one of the main research topics in the field of Internet of Things. Edge computing requires lightweight and computationally simple algorithms for sensor data analytics. Sensing edge devices are often battery powered and have a wireless connection. In designing edge devices the energy efficiency needs to be taken into account. Pre-processing the data locally in the edge device reduces the amount of data and thus decreases the energy consumption of wireless data transmission. Sensor data compression algorithms presented in this paper are mainly based on data linearity. Microclimate data is near linear in short time window and thus simple linear approximation based compression algorithms can achieve rather good compression ratios with low computational complexity. Using these kind of simple compression algorithms can significantly improve the battery and thus the edge device lifetime. In this paper linear approximation based compression algorithms are tested to compress microclimate data.

Sensor Data Stream on-line Compression with Linearity-based Methods

“…It is a lossy method like all the other methods presented in this paper, and it is suitable to be implemented in constrained sensor node due to its computational simplicity. It is very effective and has a high compression ratio for the linearly behaving environmental data [2]. The significant drawback in this method is the latency; hence, it is not well suited for real-time applications [13].…”

“…The linear regression can be calculated from the minimum of three data values; however, also more values can be used. This kind of approach is presented in [2] by the authors and the algorithm is called Linear Regression based Temporal Compression (LRbTC). In [2] 3, 4 and 5 values are used to calculate the regression line and have been tested to compress some environmental data (temperature, air pressure and wind speed).…”

“…This kind of approach is presented in [2] by the authors and the algorithm is called Linear Regression based Temporal Compression (LRbTC). In [2] 3, 4 and 5 values are used to calculate the regression line and have been tested to compress some environmental data (temperature, air pressure and wind speed). For near linearly behaving sensor data like temperature, there is a slight improvement in compression ratio if 4 or 5 values have been used to calculate the regression line.…”

“…LRbTC as presented in [2] is a very simple compression algorithm. In basic form it is presented as a flow chart in Fig.…”

“…These kinds of rather slowly changing and thus linearly behaving magnitudes are for example temperature, air pressure, humidity and wind speed. These kinds of measurements are typical in agricultural applications and in many other different IoT applications [2].…”