Leveraging big data for grasp planning

Kappler, Daniel; Bohg, Jeannette; Schaal, Stefan

doi:10.1109/icra.2015.7139793

Cited by 230 publications

(235 citation statements)

References 23 publications

Supporting

Mentioning

231

Contrasting

Unclassified

Order By: Relevance

“…Kappler et al [18] created a database of over 700 object instances, each labelled with 500 Barrett hand grasps and their associated quality from human annotations and the results of simulations with the ODE physics engine. The authors trained a deep neural network to predict grasp quality from heightmaps of the local object surface.…”

Section: Related Workmentioning

confidence: 99%

“…Grasp planning considers the problem of finding grasps for a given object that achieve force closure or optimize a related quality metric, such as the epsilon quality [35], correlation with human labels [2], [18], or success in physical trials. Often it is assumed that the object is known exactly and that contacts are placed exactly, and mechanical wrench space analysis is applied.…”

Section: Related Workmentioning

confidence: 99%

“…To measure grasp similarity in the Dex-Net 1.0 algorithm, we embed each grasp g = (x, v) of object O in Dex-Net in a feature space based on a 2D map of the local surface orientation at the contacts, inspired by grasp heightmaps [15], [18]. We generate a depthmap d i for contact c i by orthogonally projecting the local object surface onto an m × m grid centered at c i and oriented along the line to the object center of mass, a i = z − c i .…”

Section: Depthmap Gradient Featuresmentioning

confidence: 99%

“…Can analogous scaling effects emerge when datasets of 3D object models are applied to learning robust grasping and manipulation policies for robots? This question is being explored by others [12], [18], [27], [32], [33], and in this paper we present initial results using a new dataset of 3D models and grasp planning algorithm.…”

Section: Introductionmentioning

confidence: 98%

See 3 more Smart Citations

Dex-Net 1.0: A cloud-based network of 3D objects for robust grasp planning using a Multi-Armed Bandit model with correlated rewards

Mahler

Pokorny

Hou

et al. 2016

2016 IEEE International Conference on Robotics and Automation (ICRA)

331

237

View full text Add to dashboard Cite

This paper presents the Dexterity Network (DexNet) 1.0, a dataset of 3D object models and a sampling-based planning algorithm to explore how Cloud Robotics can be used for robust grasp planning. The algorithm uses a MultiArmed Bandit model with correlated rewards to leverage prior grasps and 3D object models in a growing dataset that currently includes over 10,000 unique 3D object models and 2.5 million parallel-jaw grasps. Each grasp includes an estimate of the probability of force closure under uncertainty in object and gripper pose and friction. Dex-Net 1.0 uses Multi-View Convolutional Neural Networks (MV-CNNs), a new deep learning method for 3D object classification, to provide a similarity metric between objects, and the Google Cloud Platform to simultaneously run up to 1,500 virtual cores, reducing experiment runtime by up to three orders of magnitude. Experiments suggest that correlated bandit techniques can use a cloud-based network of object models to significantly reduce the number of samples required for robust grasp planning. We report on system sensitivity to variations in similarity metrics and in uncertainty in pose and friction. Code and updated information is available at http://berkeleyautomation.github.io/dex-net/.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Depthmap Gradient Featuresmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 98%

See 2 more Smart Citations

Dex-Net 1.0: A cloud-based network of 3D objects for robust grasp planning using a Multi-Armed Bandit model with correlated rewards

Mahler

Pokorny

Hou

et al. 2016

2016 IEEE International Conference on Robotics and Automation (ICRA)

331

237

View full text Add to dashboard Cite

show abstract

“…Although force closure implies the existence of an equilibrium, this is not sufficient for ensuring grasp stability; 13,14 as it was shown in recent works, physics-based dynamical simulations are a more reliable way to rate a grasp success. 30,31 The need for further studying grasp dynamics and developing analytical models that better resemble reality is identified in Bohg et al [10]. An approach to bridging the gap between reality and models, is the design of model free grasp controllers that dynamically achieve a stable grasp equilibrium state.…”

Section: Introductionmentioning

confidence: 99%

Stable pinching by controlling finger relative orientation of robotic fingers with rolling soft tips

et al. 2017

View full text Add to dashboard Cite

SUMMARYThere is a large gap between reality and grasp models that are currently available because of the static analysis that characterizes these approaches. This work attempts to fill this need by proposing a control law that, starting from an initial contact state which does not necessarily correspond to an equilibrium, achieves dynamically a stable grasp and a relative finger orientation in the case of pinching an object with arbitrary shape via rolling soft fingertips. Controlling relative finger orientation may improve grasping force manipulability and allow the appropriate shaping of the composite object consisted of the distal links and the object, for facilitating subsequent tasks. The proposed controller utilizes only finger proprioceptive measurements and is not based on the system model. Simulation and experimental results demonstrate the performance of the proposed controller with objects of different shapes.

show abstract

Physics‐based assistive grasping for robust object manipulation in virtual reality

Nasim

Kim

2018

Computer Animation & Virtual

View full text Add to dashboard Cite

In this paper, an effective, interactive grasping algorithm is proposed to provide physically realistic interactions in the virtual world. We have provided a solution to lacking dynamic control resulting in penetrations and frictionless contact due to kinematic virtual hand provided by hand-tracking devices, by including a proxy hand in the system. The introduced proxy hand not only provides correct visual feedback but also is used to simulate dynamics between the hand and virtual objects. Our method also provides semiautonomous assistive grasping, coupled with physics-based grasping, which makes the system help the user in achieving a grasp by identifying grasp pose and orienting the object in the hand to be grasped robustly. We have implemented and evaluated our technique with various benchmarks, conducted a user study to compare our method against the state-of-the-art, pinch-based grasping mechanism, and showed that our technique provides physically realistic, stable, and time-efficient real-time interactions.

show abstract

Leveraging big data for grasp planning

Cited by 230 publications

References 23 publications

Dex-Net 1.0: A cloud-based network of 3D objects for robust grasp planning using a Multi-Armed Bandit model with correlated rewards

Dex-Net 1.0: A cloud-based network of 3D objects for robust grasp planning using a Multi-Armed Bandit model with correlated rewards

Stable pinching by controlling finger relative orientation of robotic fingers with rolling soft tips

Physics‐based assistive grasping for robust object manipulation in virtual reality

Contact Info

Product

Resources

About