Fusão sensorial e otimização de trajetória aplicadas a robôs seguidores de linha

Abstract

This paper describes the development of a high-performance Line Follower Robot (LFR) designed for mobile robotics competitions, aimed at overcoming the logical and structural limitations identified in the historical record of reactive and hybrid prototypes from UTFPR’s PatoBots team. The research is motivated by the need to mitigate deficiencies in dynamic stability, traction loss during cornering, and sensor reading inaccuracies observed in the group’s previous projects, which lacked kinematic smoothing mechanisms and maneuver anticipation. To solve these bottlenecks and elevate the vehicle’s competitive standing in compliance with national and international tournament standards, an embedded ecosystem focused on intelligent track mapping and predictive trajectory planning was implemented. The project’s motivation lies in the need to optimize the path and increase dynamic stability at high speeds, overcoming adherence and sensory precision deficiencies of prior prototypes. To achieve this, a systemic approach was adopted, integrating mechanics, dedicated electronics, and PID control algorithms. Regarding hardware, a custom Printed Circuit Board (PCB) was designed, incorporating an active suction system (downforce) to increase grip and traction capacity in curves. The central processing unit consists of an STM32 microcontroller, responsible for executing PID control loops and fusing sensor data from an optical array, Hall effect sensors (odometry), and an Inertial Measurement Unit (IMU). The project’s technical differentiator is the computational optimization pipeline developed in Python, which utilizes Gaussian filters to smooth the mapped trajectory, allowing the robot to anticipate maneuvers and execute optimized paths predictively. Experimental results confirmed the platform’s viability, demonstrating high repeatability in recording track geometry and achieving a more stable, smooth, and fluid navigation under competition conditions. The paper concludes that diagnosing the limitations of sensory fusion with static weights, combined with the theoretical proposition of Adaptive Weighting, consolidates the scientific maturity of the research. The integration between high-rigidity integrated hardware and the predictive planning pipeline solves the bottlenecks of previous models, delivering a robust and validated technological foundation for high-performance competitions to the robotics team.