Projeto e implementação de um robô com sistema de rastreamento para monitoramento de nadadores

Abstract

Swimming has established itself as one of the most practiced sports globally, valued for its recreational, therapeutic, and highly competitive nature. In high-performance scenarios, enhancing athlete performance increasingly relies on a scientific approach based on rigorous biomechanical metrics, such as instantaneous velocity, stroke length, and stroke frequency. However, training routines still face a critical bottleneck: the dependence on slow, post-session evaluation processes. This paper presents the development and validation of an autonomous mobile robot designed for monitoring and supporting high-performance swimmers. The system aims to automate the collection of biomechanical metrics in real-time, overcoming the limitations of traditional manual analysis. The platform utilizes a Raspberry Pi 4 for computer vision processing, employing a YOLOv8n face detection model, which achieved an average inference time of 171 ms (approximately 8.26 FPS). For real-world control system validation, straight-track trials were conducted using the detection model to simulate swimmer-tracking dynamics. The mechanical design was developed in SolidWorks and manufactured via 3D printing (FFF) using PLA and ABS, resulting in a stable chassis equipped with EVA wheels to reduce oscillations during image capture. Motor control for the JGY-370 DC motors is managed by an ESP32 microcontroller, executing a closed-loop PID algorithm to keep the target centered in the camera's field of view. System modeling was performed through step response identification, characterizing the plant as an integrator system with an average static gain of 1.011. Experimental results demonstrated that the robot reaches an average speed of 1.22 m/s, compatible with the critical swimming speed sustained in long-duration training. In closed-loop operation, a maximum tracking error of 1.5 m was recorded at constant speed, ensuring stability despite trajectory variations. The system also features a Flask-based web interface for remote operation and obstacle sensors for fall prevention. This work concludes that the platform is a promising tool for modernizing sports training, providing stable footage and automated instantaneous velocity data.