Optimization of intelligent sensorization systems with machine learning applied to robotic localization

• Ano de defesa: 2024
• Autor(a) principal: Klein, Luan Carlos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Brasil; Programa de Pós-Graduação em Computação Aplicada; UTFPR
• Palavras-chave em Português: Robótica; Aprendizado do computador; Robótica - Competições; Estimativa de parâmetros; Inteligência computacional; Inteligência artificial; Redes neurais (Computação); Robotics; Machine learning; Robotics - Competitions; Parameter estimation; Computational intelligence; Artificial intelligence; Neural networks (Computer science); CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; Ciência da Computação
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/36015

Efficient localization is an essential feature for autonomous robots. Due to its importance, this topic has been widely studied and debated over the years, with several approaches proposed. This study builds on previous research that suggests the use of artificial intelligence techniques for localization in the context of the RobotAtFactory 4.0 robotics competition. In particular, this study focuses on optimizing the hyperparameters of the Multilayer Perceptron (MLP), evaluating its performance in different situations within the same scenario. The results showed an improvement of up to 60% in the estimates’ precision compared to models without optimization. Another interesting result was the possibility of reusing optimizations between different scenarios, a promising alternative in cases where the computational cost of finding the best configuration is very high. This solution offers an effective approach to reducing the computational cost while improving the performance of machine learning models.