Intelligent Control Applied to a Mobile Robot for Recognition and Collection of Plastic Bottles

This article presents the control of a mobile robot for the recognition and collection of plastic bottles applying artificial intelligence techniques. The robot was equipped with a mechanical arm with six degrees of freedom, on which two video cameras were mounted, so that the vertical movement of the arm allows changing the orientation of the cameras. In this way, it is possible to locate nearby and distant bottles, and not lose sight of them while the robot moves towards them. To recognize the bottles, a convolutional neural network was trained. To approach the bottles and collect them, a diffuse system was designed. The results show an accuracy of 95.3% in the recognition of the bottles and 87% in the approximation and collection process, with an average time of 43 seconds.