Abstract

Purpose

     This paper aims to save time spent on manufacturing the data set and make the intelligent grasping system easy to deploy into a practical industrial environment. Due to the accuracy and robustness of the convolutional neural network, the success rate of the gripping operation reached a high level.

Design/Methodology/Approach

     The proposed system comprises two diverse kinds of convolutional neuron network (CNN) algorithms used in different stages and a binocular eye-in-hand system on the end effector, which detects the position and orientation of workpiece. Both algorithms are trained by the data sets containing images and annotations, which are generated automatically by the proposed method.

Findings

     The approach can be successfully applied to standard position-controlled robots common in the industry. The algorithm performs excellently in terms of elapsed time. Procession of a 256 × ۲۵۶ image spends less than 0.1 s without relying on high-performance GPUs. The approach is validated in a series of grasping experiments. This method frees workers from monotonous work and improves factory productivity.

Originality/Value

     The authors propose a novel neural network whose performance is tested to be excellent. Moreover, experimental results demonstrate that the proposed second level is extraordinary robust subject to environmental variations. The data sets are generated automatically which saves time spent on manufacturing the data set and makes the intelligent grasping system easy to deploy into a practical industrial environment. Due to the accuracy and robustness of the convolutional neural network, the success rate of the gripping operation reached a high level.

Introduction

     With the development of digital manufacturing technology, the component assembly production based on industrial robots becomes increasingly efficient (Yang et al., 2016). For assembly production, industrial robots free workers from monotonous, duplication work. Nevertheless, for most industrial robots, if there is a small-scale change, the industry must redesign work process and programming, which will greatly affect the economic benefits of the plant. Therefore, a smart system with forceful adaptability is particularly essential (Qiao et al., 2014).

     The object detection algorithm is the core of the intelligent system (Hua et al., 2019). The traditional object detection algorithms are designed to reduce the amount of calculation as much as possible on the premise of manually extracting rich feature points, thereby improving the calculation efficiency and the recognition speed. However, although manual feature extraction is easy to understand and straightforward and intuitive, it cannot cope with the identification of a large number of categories. When the target recognizer changes, it needs to perform complex feature design and extraction again. The target detection algorithm based on deep learning uses neural networks to extract the bottom- and high-level features of the image, which not only can extract more abundant and expressive features but also do not require manual participation in feature extraction and can also achieve end-to-end training and prediction.

Conclusion

     In this paper, an industrial intelligent grasping system based on convolutional neural network and binocular eye-in-hand visual system is presented. This grasping system is able to adapt to various positioning errors including angle and distance under various environmental circumstances. Notably, the neural network is trained automatically by the proposed method without manual labeling, which saves time spent on manufacturing the data set and makes the intelligent grasping system easy to deploy into a practical industrial environment. From the above experiments and analysis, the system has two properties that help to achieve high grasping effectiveness.