Compressão de dados em redes LoRa: Um compromisso entre desempenho e consumo de energia

• Autor(a) principal: Oliveira Junior, Javan Ataíde de
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações do UNIOESTE
• Texto Completo: https://tede.unioeste.br/handle/tede/6832
• topic: Internet of things; Compressão de dados; Consumo de energia; Desempenho; Internet of things; Data compression; Power consumption; Performance; Ciência da Computação

Most IoT (Internet of Things) devices have major limitations, mainly related to hardware and their energy autonomy. In general, the largest energy overhead is related to communication, reaching up to 60% depending on the application. Several methods can be found in the literature to optimize the energy consumption, for instance by modifying the transmission hardware, the communication modulation or edge computing to reduce the amount of data to be sent. Among the various approaches to edge computing are the data compression methods. Currently, most compression algorithms are designed for use on personal computers and thus often need to be adapted to the IoT context, facing memory and runtime constraints. Given these limitations, this work adapted the classical algorithms (LZ77, LZ78, LZW, Huffman and Arithmetic) and analyzed the performance and energy variables of the algorithms. The study was carried out in an ESP32 processor device with LoRa modulation and C language. The work evaluated as case studies a set of real data from an IoT application in the field of monitoring the heating of concrete blocks in large buildings and GPS data. The results obtained showed compression rates of 80%, the number of messages sent increased by 200%, and a 22% reduction in device energy consumption. The LZW algorithm achieved the highest compression rates in most scenarios, but was 8 times slower than some other algorithms. In addition, the Huffman and Arithmetic algorithms showed a more stable compression rate compared to other algorithms evaluated in this work.