Learning Based Industrial Bin-Picking Trained with Approximate Physics Simulator

Matsumura, Ryo; Harada, Kensuke; Domae, Yukiyasu; Wan, Weiwei

doi:10.1007/978-3-030-01370-7_61

Cited by 29 publications

(11 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They are a series of algorithms that recognize the underlying relationships in a set of data through a process that mimics the way the human brain operates. Since their first introduction by [1], ANNs have evolved through a broad family of learning algorithms that have advanced the state of the art across multiple domains. Feed Forward Neural Network (FFNN) was the first and has been the most commonly used ANN and [2] were the first to use FFNN for construction crew productivity prediction problem.…”

Section: Literature Reviewmentioning

confidence: 99%

İnşaat Ekibi Üretkenlik Tahmininde K-En Yakın Komşu ve Genelleştirilmiş Sinir Ağının Performansının Karşılaştırılması

Oral¹,

Oral²,

Andaç³

2021

Çukurova Üniversitesi Mühendislik Fakültesi Dergisi

View full text Add to dashboard Cite

Construction crew productivity prediction is one of the most important issues that affect the realistic prediction of construction duration and cost. Use of different search algorithms like Feed Forward Neural Network, Ant Colony, Artificial Bee Colony, Particle Swarm Optimization, Radial Based Neural Networks and Self Organizing Maps for crew productivity prediction problem have been discussed in previous studies. However, the significant effect of the coherence between the nature of the data and the characteristics of the method used in prediction performance has generally been neglected. The aim of the current research thus has been to analyse the prediction performance of two contemporary learning algorithms; K-Nearest Neighbour (K-NN) and Generalized Neural Network (GRNN) when applied to three different crew (formwork, tiling and masonry) productivity related data sets with different distribution characteristics. Performance of both methods varied with the changing coefficient of variation values. K-NN outperformed GRNN for all data sets and both of the methods had their worst performance on the dataset with the highest variance.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

İnşaat Ekibi Üretkenlik Tahmininde K-En Yakın Komşu ve Genelleştirilmiş Sinir Ağının Performansının Karşılaştırılması

Oral¹,

Oral²,

Andaç³

2021

Çukurova Üniversitesi Mühendislik Fakültesi Dergisi

View full text Add to dashboard Cite

show abstract

“…However, deep learning-based methods tend to work well only when the large training dataset is available. The training dataset can be gathered from real-world experiments [54,56], or obtained from synthetic data provided by a physics simulation engine [57][58][59]. These approaches are time-consuming and computationally expensive.…”

Section: Related Workmentioning

confidence: 99%

Deep Learning-based Fast Grasp Planning for Robotic Bin-picking by Small Data Set without GPU

Arai¹,

Zhao²,

Purnomo³

et al. 2021

Preprint

View full text Add to dashboard Cite

<div>This paper proposes a deep learning-based fast grasp detection method with a small dataset for robotic bin-picking. We consider the problem of grasping stacked up mechanical parts on a planar workspace using a parallel gripper. In this paper, we use a deep neural network to solve the problem with a single depth image. To reduce the computation time, we propose an edge-based algorithm to generate potential grasps. Then, a convolutional neural network (CNN) is applied to evaluate the robustness of all potential grasps for bin-picking. Finally, the proposed method ranks them and the object is grasped by using the grasp with the highest score. In bin-picking experiments, we evaluate the proposed method with a 7-DOF manipulator using textureless mechanical parts with complex shapes. The success ratio of grasping is 97%, and the average computation time of CNN inference is less than 0.23[s] on a laptop PC without a GPU array. In addition, we also confirm that the proposed method can be applied to unseen objects which are not included in the training dataset. </div>

show abstract

“…Some researches use CAD model for better recognizing the scene [10], [11], [12]. Recently learning-based method has been widely used to achieve better robustness and generalization [1], [13], [14].…”

Section: Related Workmentioning

confidence: 99%

A Topological Solution of Entanglement for Complex-shaped Parts in Robotic Bin-picking

Zhang

Koyama

Domae

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

This paper addresses the problem of picking up only one object at a time avoiding any entanglement in binpicking. To cope with a difficult case where the complex-shaped objects are heavily entangled together, we propose a topologybased method that can generate non-tangle grasp positions on a single depth image. The core technique is entanglement map, which is a feature map to measure the entanglement possibilities obtained from the input image. We use entanglement map to select probable regions containing graspable objects. The optimum grasping pose is detected from the selected regions considering the collision between robot hand and objects. Experimental results show that our analytic method provides a more comprehensive and intuitive observation of entanglement and exceeds previous learning-based work [1] in success rates. Especially, our topology-based method does not rely on any object models or time-consuming training process, so that it can be easily adapted to more complex bin-picking scenes.

show abstract

Learning Based Industrial Bin-Picking Trained with Approximate Physics Simulator

Cited by 29 publications

References 30 publications

İnşaat Ekibi Üretkenlik Tahmininde K-En Yakın Komşu ve Genelleştirilmiş Sinir Ağının Performansının Karşılaştırılması

İnşaat Ekibi Üretkenlik Tahmininde K-En Yakın Komşu ve Genelleştirilmiş Sinir Ağının Performansının Karşılaştırılması

Deep Learning-based Fast Grasp Planning for Robotic Bin-picking by Small Data Set without GPU

A Topological Solution of Entanglement for Complex-shaped Parts in Robotic Bin-picking

Contact Info

Product

Resources

About