Model-based and model-free reinforcement learning for visual servoing

Farahmand, Amir-massoud; Shademan, Azad; Jägersand, Martin; Szepesvári, Csaba

doi:10.1109/robot.2009.5152834

Cited by 27 publications

(13 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Regression-based neural networks (NNs) have been considerably used to approximate the non-linear image-Jacobian. [25][26][27][28] They suffer from local minima which are complex to avoid. 29 It is worth mentioning that using NN in approximating the hybrid interaction matrices is not widely investigated in the literature.…”

Section: Related Workmentioning

confidence: 99%

Optimized hybrid decoupled visual servoing with supervised learning

Rastegarpanah

Aflakian

2021

Proceedings of the Institution of Mechanical Engineers, Part I:

View full text Add to dashboard Cite

This study proposes an optimized hybrid visual servoing approach to overcome the imperfections of classical two-dimensional, three-dimensional and hybrid visual servoing methods. These imperfections are mostly convergence issues, non-optimized trajectories, expensive calculations and singularities. The proposed method provides more efficient optimized trajectories with shorter camera path for the robot than image-based and classical hybrid visual servoing methods. Moreover, it is less likely to lose the object from the camera field of view, and it is more robust to camera calibration than the classical position-based and hybrid visual servoing methods. The drawbacks in two-dimensional visual servoing are mostly related to the camera retreat and rotational motions. To tackle these drawbacks, rotations and translations in Z-axis have been separately controlled from three-dimensional estimation of the visual features. The pseudo-inverse of the proposed interaction matrix is approximated by a neuro-fuzzy neural network called local linear model tree. Using local linear model tree, the controller avoids the singularities and ill-conditioning of the proposed interaction matrix and makes it robust to image noises and camera parameters. The proposed method has been compared with classical image-based, position-based and hybrid visual servoing methods, both in simulation and in the real world using a 7-degree-of-freedom arm robot.

show abstract

Section: Related Workmentioning

confidence: 99%

Optimized hybrid decoupled visual servoing with supervised learning

Rastegarpanah

Aflakian

2021

Proceedings of the Institution of Mechanical Engineers, Part I:

View full text Add to dashboard Cite

show abstract

“…Traditionally, model predictive control (MPC) relies on known or learned dynamics robot models, and it is used to generate advanced motor controls. Once the model of the effect of the robot's actions on its environment (e.g., workspace with objects) is known or learned, it can be used to plan a sequence of optimal actions to fulfill a task: this can be done with MPC, often with a receding horizon approach, but also with reinforcement learning [25], [26], [27]. For planning sequence of manipulation actions, this was often done with reduced visual models computing features of the objects in the scene, to lower the dimensionality of the problem.…”

Section: Related Workmentioning

confidence: 99%

VP-GO: a "light" action-conditioned visual prediction model

Ma¹,

Fleytoux²,

Mouret³

et al. 2021

Preprint

View full text Add to dashboard Cite

Visual prediction models are a promising solution for visual-based robotic grasping of cluttered, unknown soft objects. Previous models from the literature are computationally greedy, which limits reproducibility; although some consider stochasticity in the prediction model, it is often too weak to catch the reality of robotics experiments involving grasping such objects. Furthermore, previous work focused on elementary movements that are not efficient to reason in terms of more complex semantic actions. To address these limitations, we propose VP-GO, a "light" stochastic action-conditioned visual prediction model. We propose a hierarchical decomposition of semantic grasping and manipulation actions into elementary end-effector movements, to ensure compatibility with existing models and datasets for visual prediction of robotic actions such as RoboNet. We also record and release a new open dataset for visual prediction of object grasping, called PandaGrasp. Our model can be pre-trained on RoboNet and fine-tuned on PandaGrasp, and performs similarly to more complex models in terms of signal prediction metrics. Qualitatively, it outperforms when predicting the outcome of complex grasps performed by our robot.

show abstract

“…The second ingredient for UVS seems to be available and easy to implement (for example the Broyden update approach), thanks to extensive research and evaluation in control [10], [13], [14]. However the first one, which is the crucial one, despite a lot of research [15], [16], [17], [18], is still hard to synthesize for reliable tracking of everyday objects as shown by Fig.…”

Section: Visual Servoing In Practicementioning

confidence: 99%

Towards Practical Visual Servoing in Robotics

Fomena

Quintero

Gridseth

et al. 2013

2013 International Conference on Computer and Robot Vision

View full text Add to dashboard Cite

Abstract-Visual servoing was introduced in robotics nearly 4 decades ago. However until now, there are still only a handful of known examples of application of this technique in addressing real word robotics problems such as disaster response, assistance for elderly or handicapped people, etc. As the world is moving towards the use of robotics to improve quality of life, it is time to assess the challenges involved in applying visual servoing to solve real world problems. This paper presents an overview of these challenges, by asking the question what are the missing components for practical visual servoing? and by providing practical possible solutions for these components. Illustration of these challenges and our current practical solutions are given using our 7-DoFs Barrett WAM Arm.

show abstract

Model-based and model-free reinforcement learning for visual servoing

Cited by 27 publications

References 26 publications

Optimized hybrid decoupled visual servoing with supervised learning

Optimized hybrid decoupled visual servoing with supervised learning

VP-GO: a "light" action-conditioned visual prediction model

Towards Practical Visual Servoing in Robotics

Contact Info

Product

Resources

About