Real-Time Visual Grasp Synthesis Using Genetic Algorithms and Neural Networks

Chella, Antonio; Dindo, Haris; Matraxia, Francesco; Pirrone, Roberto

doi:10.1007/978-3-540-74782-6_49

Cited by 7 publications

(9 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Neural networks are well-fitted for such tasks, as they are computationally simple, have the ability to map complex and non-linear data relationships and have the ability to learn and then predict in real-time the displayed behavior. This explains the interest of researchers from both the deformable object modeling [5,6] and the grasping and manipulation research fields [7][8][9][10][11] into such techniques.…”

Section: Related Workmentioning

confidence: 98%

Learning and Prediction of Soft Object Deformation Using Visual Analysis of Robot Interactions

Creţu

Payeur

Petriu

2010

Advances in Visual Computing

View full text Add to dashboard Cite

The paper discusses an innovative approach to acquire and learn deformable objects' properties to allow the incorporation of soft objects in virtualized reality applications or the control of dexterous manipulators. Contours of deformable objects are tracked in a sequence of images collected from a camera and correlated to the interaction measurements gathered at the fingers of a robotic hand using a combination of unsupervised and supervised neural network architectures. The advantage of the proposed methodology is that it not only automatically and implicitly captures the real elastic behavior of an object regardless of its material, but it is also able to predict the shape of its contour for previously unseen interactions. The results obtained show that the proposed approach is fast, insensitive to slight changes in contrast and lighting, and able to model accurately and predict severe contour deformations.

show abstract

Section: Related Workmentioning

confidence: 98%

Learning and Prediction of Soft Object Deformation Using Visual Analysis of Robot Interactions

Creţu

Payeur

Petriu

2010

Advances in Visual Computing

View full text Add to dashboard Cite

show abstract

“…Incremental Shaping pseudocode. The algorithm executes a hill climber [1][2][3][4][5][6][7][8][9][10][11][12][13][14] (see text for description). If the current genome fails [15,16], the task environment is eased [17,18]; while it is successful [19][20][21][22][23], the task environment is made more difficult [24,25].…”

Section: Trainingmentioning

confidence: 99%

“…Place target object at (x, z) and let simulation run for 10,000 time steps 6) If Success() [see Fig. 2…”

Section: Scaffolding Schedulesmentioning

confidence: 99%

“…[3], [4]) in which individual motor commands combine in a nonlinear fashion to produce behavior, thereby making analytical derivations of optimal controllers infeasible. However, evolutionary algorithms alone are often not sufficient to evolve multiple dynamic behaviors: to date most reported efforts have primarily focused on realizing a single behavior, such as locomotion [3], [4] or grasping [5], [6].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

How robot morphology and training order affect the learning of multiple behaviors

Auerbach

Bongard

2009

2009 IEEE Congress on Evolutionary Computation

View full text Add to dashboard Cite

Abstract-Automatically synthesizing behaviors for robots with articulated bodies poses a number of challenges beyond those encountered when generating behaviors for simpler agents. One such challenge is how to optimize a controller that can orchestrate dynamic motion of different parts of the body at different times. This paper presents an incremental shaping method that addresses this challenge: it trains a controller to both coordinate a robot's leg motions to achieve directed locomotion toward an object, and then coordinate gripper motion to achieve lifting once the object is reached. It is shown that success is dependent on the order in which these behaviors are learned, and that despite the fact that one robot can master these behaviors better than another with a different morphology, this learning order is invariant across the two robot morphologies investigated here. This suggests that aspects of the task environment, learning algorithm or the controller dictate learning order more than the choice of morphology.

show abstract

“…One work that uses high-level shape primitives, and is similar to ours in terms of learning, but by using an SVM approach, is [9]. Another approach to learning from 2D grasp qualities, using neural networks and genetic algorithms, is presented in [10]. This paper builds on the work of Huebner et al .…”

Section: Introductionmentioning

confidence: 99%

Learning of 2D grasping strategies from box-based 3D object approximations

Geidenstam

Huebner

Banksell

et al. 2009

Robotics: Science and Systems V

View full text Add to dashboard Cite

Abstract-In this paper, we bridge and extend the approaches of 3D shape approximation and 2D grasping strategies. We begin by applying a shape decomposition to an object, i.e. its extracted 3D point data, using a flexible hierarchy of minimum volume bounding boxes. From this representation, we use the projections of points onto each of the valid faces as a basis for finding planar grasps. These grasp hypotheses are evaluated using a set of 2D and 3D heuristic quality measures. Finally on this set of quality measures, we use a neural network to learn good grasps and the relevance of each quality measure for a good grasp. We test and evaluate the algorithm in the GraspIt! simulator.

show abstract

Real-Time Visual Grasp Synthesis Using Genetic Algorithms and Neural Networks

Cited by 7 publications

References 12 publications

Learning and Prediction of Soft Object Deformation Using Visual Analysis of Robot Interactions

Learning and Prediction of Soft Object Deformation Using Visual Analysis of Robot Interactions

How robot morphology and training order affect the learning of multiple behaviors

Learning of 2D grasping strategies from box-based 3D object approximations

Contact Info

Product

Resources

About