Untying cable by combining 3D deep neural network with deep reinforcement learning

Fan, Zheming; Shao, Wanpeng; Hayashi, Toyohiro; Ohashi, Takeshi

doi:10.1080/01691864.2022.2126729

Cited by 2 publications

(3 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Simulation Open Framework Architecture (SOFA) [40] physics engine is developed for medical simulation, which uses FEM to model DO and can model the interaction between rigid and deformable objects and allows customised solvers for different mechanical objects, constraints and collision geometries. Unity with Obi [41] has been used by Weng et al (2021) [42] for generating synthetic data for bag manipulation and it could provide a good commercial environment for rope manipulation [43]. ARCsim [44] is a cloth-specific simulation environment which has been employed in many perception-focused applications [45,46].…”

Section: Simulation Environmentsmentioning

confidence: 99%

“…There are many applications of IL in the KT/KU literature, but it is difficult to frame the task as MDP to develop an RL controller-one of the challenges is reward shaping. Fan et al (2022) [43] use DRL algorithm Deep Q-learning (DQN) [126] that takes embedded states as input and discretised grasping points and moving directions as action. They achieve a 54% success rate on single-arm knot untying knots on a table.…”

Section: Data-driven Manipulation In Knot Tying/untyingmentioning

confidence: 99%

See 1 more Smart Citation

Data-Driven Robotic Manipulation of Cloth-like Deformable Objects: The Present, Challenges and Future Prospects

Kadi

Terzic

2023

Sensors

View full text Add to dashboard Cite

Manipulating cloth-like deformable objects (CDOs) is a long-standing problem in the robotics community. CDOs are flexible (non-rigid) objects that do not show a detectable level of compression strength while two points on the article are pushed towards each other and include objects such as ropes (1D), fabrics (2D) and bags (3D). In general, CDOs’ many degrees of freedom (DoF) introduce severe self-occlusion and complex state–action dynamics as significant obstacles to perception and manipulation systems. These challenges exacerbate existing issues of modern robotic control methods such as imitation learning (IL) and reinforcement learning (RL). This review focuses on the application details of data-driven control methods on four major task families in this domain: cloth shaping, knot tying/untying, dressing and bag manipulation. Furthermore, we identify specific inductive biases in these four domains that present challenges for more general IL and RL algorithms.

show abstract

Section: Simulation Environmentsmentioning

confidence: 99%

Section: Data-driven Manipulation In Knot Tying/untyingmentioning

confidence: 99%

Data-Driven Robotic Manipulation of Cloth-like Deformable Objects: The Present, Challenges and Future Prospects

Kadi

Terzic

2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Using P&P action primitives. Fan et al (2022) [93] use discrete-action DRL algorithm Deep Q-learning (DQN) [255] that takes the embedded states as input and discretised grasping points and moving directions as action. They achieve a 54% success rate using single-arm for untying knots on the table.…”

Section: Data-driven Control In Rope Manipulationmentioning

confidence: 99%

Data-Driven Robotic Manipulation of Cloth-like Deformable Objects: Present, Challenges, and Future Prospects

Kadi¹,

Terzic²

2022

Preprint

View full text Add to dashboard Cite

Manipulating cloth-like deformable objects (CDOs) is a long-standing problem in the robotics community. CDOs are flexible (non-rigid) objects that do not show a detectable level of com- pression strength while two points on the article are pushed towards each otherand include objects such as ropes (1D), fabrics (2D) and bags (3D). In general, CDOs’ many degrees of freedom (DoF) introduce severe self-occlusion and complex state-action dynamics as significant obstacles for perception and manipulation systems. These challenges exacerbate existing issues of modern robotic control methods such as imitation learning (IL) and reinforcement learning (RL). This review focuses on the application details of data-driven control methods on four major task families in this domain: cloth-shaping, rope manipulation, dressing and bag manip- ulation. Furthermore, we identify specific inductive biases in these four domains that present challenges for more general IL and RL algorithms, and summarise the future direction for the development of the field.

show abstract

Untying cable by combining 3D deep neural network with deep reinforcement learning

Cited by 2 publications

References 12 publications

Data-Driven Robotic Manipulation of Cloth-like Deformable Objects: The Present, Challenges and Future Prospects

Data-Driven Robotic Manipulation of Cloth-like Deformable Objects: The Present, Challenges and Future Prospects

Data-Driven Robotic Manipulation of Cloth-like Deformable Objects: Present, Challenges, and Future Prospects

Contact Info

Product

Resources

About