Data Traffic Management Based on Compression and MDL Techniques for Smart Agriculture in IoT

The sector of agriculture facing numerous challenges for the proper utilization of its natural resources. For that reason, and to the growing risk of changing weather conditions, we must monitor the soil conditions and meteorological data locally in order to accelerate the adoption of appropriate decisions that help the culture. In the era of the Internet of Things (IoT), a solution is to deploy a Wireless Sensor Network (WSN) as a low-cost remote monitoring and management system for these kinds of features. But WSN is suffering from the motes' limited energy supplies, which decrease the total network's lifetime. Each mote collects periodically the tracked feature and transmitting the data to the edge Gateway (GW) for further study. This method of transmitting massive volumes of data allows the sensor node to use high energy and substantial usage of bandwidth on the network. In this research, Data Traffic Management based on Compression and Minimum Description Length (MDL) Techniques is proposed which works at the level of sensor nodes (i.e., Things level) and at the edge GW level. In the first level, a lightweight lossless compression algorithm based on Differential Encoding and Huffman techniques which is particularly beneficial for IoT nodes, that monitoring the features of the environment, especially those with limited computing and memory resources. Instead of trying to formulate innovative ad hoc algorithms, we demonstrate that, provided general awareness of the features to be monitored, classical Huffman coding can be used effectively to describe the same features that measure at various time periods and locations. In the second level, the principle of MDL with hierarchical clustering was utilized for the purpose of clustering the sets of data coming from the first level. The strategy used to minimize data sets transmitted at this level is fairly simple. Any pair of data sets that can be compressed according to the MDL principle is combined into one cluster. As a result of this strategy, the number of data sets is gradually decreasing and the process of merging similar sets into a single cluster is stopped if no more pairs of sets can be compressed. Results utilizing temperature measurements indicate that it outperforms common methods developed especially for WSNs in reducing the amount of data transmitted and saving energy, even though the suggested system does not reach the theoretical maximum.