Navegação autônoma em robôs de baixo custo com smartphone como sensor e cérebro

Abstract

This work presents a low-cost robotic platform, based on Arduino, which integrates an Android smartphone as the only visual sensor and processing unit, to detect and pursue an object through Computer Vision. The robot, controlled by Arduino, performs navigation based on commands received via Bluetooth, generated by the Android application that uses OpenCV and color-based object detection (Blob Detection). 40 executions in two standardized scenarios compared the solution to a traditional line-following robot, equipped with infrared sensors, revealing equivalent performance, but with much more operational flexibility. Additionally, new experiments were carried out with a second smartphone, more modern and, consequently, with greater computational power, which made it possible to explore these new characteristics to obtain more accurate results. This includes the test of distance estimates, where estimates were made for distances between 5 and 210 centimeters and the errors for the largest and smallest estimated distances were calculated. Based on this, 4 approaches were developed with the aim of reducing these errors. This work presents as a contribution the demonstration of a smartphone combined with a robotic kit, replacing dedicated optical sensors and autonomous navigation using open-source libraries in the mobile environment.