Basic principles for the development of an application to bi-manipulate boxes with a humanoid robot

Logistics is a sector which is continuously growing, due to globalization, as well as the current situation caused by the Covid. In this article, an application to recognize boxes is described. The characteristics are extracted with the goal of identify the opening side of the box by using computer vision techniques. This goal has been achieved considering the dimensions, as well as, the position in the space of the box. Those characteristics were obtained processing 2D and 3D images. Then, this information has been classified by using a decision tree based on the human knowledge. The probability of each of the six faces to be the opening side is obtained. This article is a base to develop in the future an application in which the humanoid robot TEO is capable to learn the optimal way to find the opening of boxes and bimanipulate them to be opened in an automated system.

Automated summary

This article introduces an application that uses computer vision techniques to identify boxes, extracting characteristics and determining the opening side based on a decision tree. The goal is to develop a future application where the humanoid robot TEO can learn to find and manipulate the opening of boxes in an automated system.