Aplicação de técnicas de robótica, sistemas computacionais inteligentes e visão computacional no desenvolvimento de um manipulador autônomo

Abstract

The use of robot manipulator for pick-and-place operations is not a new concept in the industry, although it can encompass several applications, such as quality control. However, it has become more accessible according to the evolution of processing power and the use of intelligent systems-based technologies. Thus, the objective of this research is to build a robotic manipulator based on BCN3D Moveo and its implementation in a pick-and-place task. For this, concepts of Artificial Neural Networks (ANNs), computer vision, kinematics and trajectory planning of robotic manipulators were used. The resulting robot recognizes several objects and remove those it considers defective. The manipulator is constituted by 5 degrees of freedom (DOFs) and modeled from divisions in subsystems, each with a specific task. The techniques used to solve the proposed problem were validated through previous work using a manipulator with 3 DOFs. The results indicate that the built manipulator satisfies the proposal of this work satisfactorily, since it presents sufficient precision and repeatability in the operations of object identification, calculation of the inverse kinematics and trajectory planning. Future work addresses the possibility of applying this system, and improvements in the constructive and electromechanical aspects of the robot.