An Affordable Deep Learning Based Solution to Support Pick and Place Robotic Tasks
Mahmoodpour, M., Lobov, A., Hayati, S., Pastukhov, A.
Article language: English
Abstract. In the current competitive market, the lack of financial resources is a major challenge for Small and Medium Enterprises (SMEs). As a result, SMEs seek low-cost technologies to be employed in their enterprises. Moreover, robotics together with vision systems have become an indispensable part of the current production systems regarding their capabilities to improve the productivity and performance of manufacturing processes. The aim of this paper is to describe an affordable solution for the pick and place robot operation using computer vision technology. The application is designed to identify, pick and place the objects with different arrangements on the palette. For the fulfillment of the vision system, the latest deep learning techniques and image processing software is used to detect the parts. Moreover, the proposed application is dynamic and scalable, created for different use cases so that the operation for various parts with diverse shapes can be handled. The application is tested and validated at the Tampere University Robotic Laboratory (Robolab).
Keywords: computer vision, Convolutional neural networks, deep learning, robot vision
Pages: 66–75Total pages: 10
Year of publication: 2019