Classificação de visada e mitigação de erros de estimativa de distância em sistemas UWB

Abstract

This master's thesis investigates the classification of line-of-sight (LOS) and non-line-of-sight (NLOS) conditions and their impact on distance measurement accuracy in Ultra-Wideband (UWB) positioning systems. The focus of the research was to develop and validate a machine learning model capable of dynamically classifying LOS and NLOS conditions to adjust the system's parameters for error mitigation. This involved the design of custom hardware and software to conduct extensive tests under various simulated environmental conditions, mirroring real-world complexities. The results demonstrated that the machine learning model significantly enhanced measurement accuracy, reducing average distance errors from over 10 centimeters in baseline conditions to under 3 centimeters in optimized setups. The implications of these findings underscores the potential of adaptive learning models to improve the reliability and operational efficiency of UWB systems, particularly in complex indoor environments. The model's ability to adapt to changing conditions and accurately classify signal disruptions due to physical obstructions provides a critical improvement over traditional static modeling approaches. This research lays the groundwork for future advancements in UWB technology, suggesting that integrating machine learning into UWB systems can lead to more robust and accurate indoor positioning solutions that are crucial for industries like logistics, autonomous navigation, and smart building management.