An online and unsupervised algorithm for automatic data compression on edge computing under IoT scenarios
Data compression, IoT, Intelligent Vehicles, TinyML, edge computing, machine learning, unsupervised learning
With the advance and mass adoption of solutions in the Internet of Things (IoT) and connected cities fields, the number of devices and sensors connected to the network tends to grow exponentially. In this scenario, the transmission and storage of the growing volume of data brings new challenges. When devices transmit potentially irrelevant or redundant data, there is a greater expenditure of energy and processing, in addition to the unnecessary use of the communication channel. In this way, data compression solutions on the edge computing devices becomes increasingly attractive, enabling the elimination of samples that would have little or no contribution to the application, in order to significantly reduce the volume of data needed to represent the information. However, such devices present in the market today have serious limitations on storage and processing power. In order to circumvent such limitations, the TinyML field appears, looking for ways to implement machine learning models in low power devices. Given this context, one of the sectors that can most benefit from these new technologies is the automotive industry, as currently all cars produced must be instrumented with a series of sensors. Thus, when connecting an edge device to the vehicle, it is possible to do a local processing of the data and transmit it to a remote server later. In this context, the present work proposes the development of a new unsupervised, online, and automatically adapting data compression algorithm for IoT applications. The proposal was embedded on a edge device based on an ESP microcontroller currently available on the market. The device with the embedded algorithm was connected to a car and preliminary results show that it is possible to achieve significant compression rates without interfering in the processing time of the primary operations of the system.