In response to urban challenges, particularly in the context of mobility and pollution, this research addresses innovation in lightweight electric vehicles, with a special focus on electric bicycles. The recent regulation in Brazil, National Traffic Council Resolution 996 (June 2023), underscores the importance of these vehicles, establishing essential parameters such as nominal power and maximum speed. The thesis focuses on enhancing the Battery Management System (BMS) of electric bicycles, emphasizing the application of TinyML for the estimation of critical battery parameters. Additionally, it introduces a comprehensive methodology for continuous monitoring, deviation detection, and dynamic updating of TinyML models embedded in microcontrolled systems. 

Incremental Learning techniques are explored to identify usage patterns, diagnose battery health and autonomy, and detect unusual scenarios. The research proposes the creation of a robust experimental environment to validate and improve these approaches, thereby contributing to the development of a more efficient and adaptable BMS. At the core of the proposal is the introduction of a parametric estimation mechanism for batteries, specifically applied to electric bicycles, guided by TinyML algorithms. This approach not only enhances precision in energy management but also establishes a dynamic system capable of adapting to changes in operational conditions, promoting optimized battery usage and extending its lifespan. Thus, the research not only addresses current urban mobility challenges but also significantly contributes to the viability and sustainable efficiency of lightweight electric vehicles, particularly electric bicycles.