Intelligent Neurofuzzy Control of a Robotic Gripper

Domínguez-López, Jorge Axel; Damper, R. I.; Crowder, Richard; Harris, Christopher

doi:10.1007/10992388_5

Cited by 3 publications

(4 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning techniques such as supervised neuro-fuzzy learning, unsupervised reinforcement learning, and unsupervised/ supervised hybrid are used to develop a neuro-fuzzy controller for a two-fingered gripper for grasping the object without slippage, and with the appropriate force for reducing the damage to the object to be grasped. 51,52 The fuzzy logic controller is clubbed with the neural network to form neural-fuzzy for controlling the robot's hand while grasping. The input data and output data are trained with the BP algorithm.…”

Section: Developed Amentioning

confidence: 99%

Application of soft computing methods in robotic grasping: A state-of-the-art survey

Mahanta

Deepak²,

Biswal³

2021

Proceedings of the Institution of Mechanical Engineers, Part E:

View full text Add to dashboard Cite

Robotic grasping has become one of the most important domains of robotics research over the past few decades due to its wide range of applications in industrial automation. The model of grasping objects by robot hand depends on a good number of factors, such as type and size of the object, the morphology of object, type of hand, degree of freedom, etc. Thus, the model sometimes becomes mathematically intractable. With the progress in computational capability, soft computing methods have found a way to address the challenges faced by traditional methods while dealing with the robotic grasping problem. This article aims to summarize the outcome of a systematic study in the field of application of soft computing methods in robotic grasping and manipulation. The key processes of robotic grasping where soft computing methods are applied are identified, and research contributions of all processes are analyzed. This review presents a state-of-the-art survey and attempts to find the research gaps in the area of soft computing applications to address the robotic grasping problem.

show abstract

Section: Developed Amentioning

confidence: 99%

Application of soft computing methods in robotic grasping: A state-of-the-art survey

Mahanta

Deepak²,

Biswal³

2021

Proceedings of the Institution of Mechanical Engineers, Part E:

View full text Add to dashboard Cite

show abstract

“…However, if the applied force and the applied actuator energy (e.g., applied motor voltage) are high, it means that it is not possible to grip tightly without risk of crushing the object or slippage. Accordingly, an approach to reducing the possibility of damaging the object is to restrict the end effector acceleration [9,10]. Therefore, the toolbox trajectory generation function includes a framework to control the maximum end effector acceleration when a gripper is attached to the robot.…”

Section: Graspingmentioning

confidence: 99%

A MATLAB Toolbox for Robotic Manipulators

Vila-Rosado

Domínguez-López

Sixth Mexican International Conference on Computer Science (ENC'05)

View full text Add to dashboard Cite

MATLAB toolboxes have been very useful to teach and train without the need for the physical system. Especially in the robotics field, using simulations avoids the considerable cost of building a full robot manipulator. For robotics, there was only one MATLAB toolbox, which has several limitations. These limitations have been cured in the toolbox reported in this paper. Also, there have been some additions to make the toolbox robust and more complete. Some of these improvements are using a closed-form solution for the inverse kinematics, a tough path and trajectory generation. In addition, an end effector (i.e., a two-fingered gripper) has been added, hence the user can simulate a robot which manipulates objects. The full toolbox and its documentation can be freely downloaded from internet.

show abstract

“…Supervised Learning Network: This is a neurofuzzy controller trained off-line with error back-propagation [11]. We used the labelled training data previously collected [4,5] to train the SLN ab initio. The resulting weights were also used as the initial weights of the ASN.…”

Section: Hybrid Neurofuzzy Controllermentioning

confidence: 99%

“…The gripper was a simple two-fingered, single degree of freedom system with slip and force sensors, based on the experimental system developed in our previous work [6,7]. Full details of the simulation (kinematics equations) can be found in Appendix A of [4]. The simulation was validated against earlier results of the gripper handling a range of weights [5,6,7] The gripper was simulated holding a 0.1 kg object.…”

Section: Introductionmentioning

confidence: 99%