Desenvolvimento de uma aplicação Android para mapeamento de uma mesa com obstáculos e robôs E-Puck utilizando visão computacional

Abstract

The increasing need for high precision technologies, coupled with high repeatability and speed in performing various tasks, has led to the accelerated growth of robotics over the years, and in particular mobile robotics. Among the numerous challenges in this area it is the need to calculate the shortest path to be traveled from one point to another, free of obstacles, in order to achieve a defined goal. Based on this, the path planning is fundamental in robotics. Several theories have been applied in the study of path planning, such as visibility graphs and cellular automata, the latter based on cellular decomposition. In this context, several approaches are used in the treatment of problems involving path planning. One of the possible approaches is its integration with computer vision techniques. In this way, this work proposes the development of an Android application based on computer vision techniques for mapping a testing table, or testing bench, with obstacles, E-Puck robots and a goal. Thus, a real testing environment has been proposed, in which path planning studies can be performed with mobile robots in future works. Such a testing environment is quite advantageous in a field of studies that usually involves high cost equipment or only available in the form of simulation environments which, while providing good results, still requires verification in a real environment for validation of results. At the end of this work, a flexible and configurable Android application is presented, developed with the use of color filtering techniques, noise filtering, morphological operations, identification of objects and simplification by polygons. In addition, the OpenCV development library has been integrated into the application, as it contains a mature and optimized implementation of mentioned techniques. As it is a first study, improvements are also proposed, as well as suggestions for future works.