Data-Driven Object Pose Estimation in a Practical Bin-Picking Application.

Viktor Kozák,Roman Sushkov,Miroslav Kulich,Libor Přeučil

doi:10.3390/s21186093

Viktor Kozák, Roman Sushkov + Show 2 more

Open Access

https://doi.org/10.3390/s21186093

Copy DOI

Abstract

This paper addresses the problem of pose estimation from 2D images for textureless industrial metallic parts for a semistructured bin-picking task. The appearance of metallic reflective parts is highly dependent on the camera viewing direction, as well as the distribution of light on the object, making conventional vision-based methods unsuitable for the task. We propose a solution using direct light at a fixed position to the camera, mounted directly on the robot’s gripper, that allows us to take advantage of the reflective properties of the manipulated object. We propose a data-driven approach based on convolutional neural networks (CNN), without the need for a hard-coded geometry of the manipulated object. The solution was modified for an industrial application and extensively tested in a real factory. Our solution uses a cheap 2D camera and allows for a semi-automatic data-gathering process on-site.

Highlights

For several decades, production lines in various industries have relied heavily on the use of robotic manipulators
We propose an object localization pipeline based on a combination of feature descriptorbased image segmentation and the use of an end-to-end trained convolution neural network (CNN) for pose estimation
For the task of precise pose estimation, we propose an end-to-end neural network based on regression

Summary

Introduction

Production lines in various industries have relied heavily on the use of robotic manipulators. Traditional applications, such as welding automation, part manipulation, or assembly rely on precise predefined movements and defined positions of items in the robot’s workspace. Recent advancements in robotic automation allow for the deployment of intelligent manipulators in tasks previously unsolvable by robots. The introduction of collaborative robots allowed robots to share the workspace with humans and perform cooperative tasks, while the advancements in machine learning in combination with fine force control in robotic manipulators enabled the automation of fine motor tasks, which could previously only be performed by a human operator. The increased affordability of robotic technology and the rising costs of human employees represent incentives to increase automation in production

Objectives

Results

Conclusion