Related references
Note: Only part of the references are listed.
Article
Automation & Control Systems
Shengqi Duan et al.
Summary: This paper presents a modular intelligent robot architecture using multi-task convolutional neural network for specific object grasping and manipulation in a stacked and cluttered environment. The proposed end-to-end multi-task semantic grasping convolutional neural network (MSG-ConvNet) can simultaneously detect grasps and perform semantic segmentation in cluttered scenarios. The experimental results show that the proposed method improves the recognition accuracy and detection speed compared to the benchmark, and achieves state-of-the-art performance on single-object datasets.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Liangdong Wu et al.
Summary: This study proposes an efficient push-grasping method based on reinforcement learning for multiple target objects in cluttered environments. The key is to consider the states of all targets, expanding the graspable space of each target through pushing actions to achieve minimal total number of actions and improve system efficiency. Experimental results show that the proposed method outperforms other methods for multiple target objects and a single target in clutter.
FRONTIERS IN NEUROROBOTICS
(2023)
Article
Computer Science, Information Systems
Yuan Wei
Summary: In this paper, a solution called Intelligent Human-like Motion Planner (IHMP) is proposed to address the human-like motion issue for anthropomorphic arms. The IHMP integrates Movement Primitive (MP), Bayesian Network (BN), and Coupling Neural Network (CPNN) to generate human-like arm movements. The approach includes decoupling the arm motion model, utilizing Bayesian network to choose suitable motion models, and using Coupling Neural Network to obtain inverse kinematic solutions. The IHMP aims to improve the accuracy of human-like motion and enhance the independent consciousness of robots. Experimental validation on humanoid robot Pepper demonstrates the effectiveness of the IHMP.
Article
Robotics
Jianglong Ye et al.
Summary: This research proposes using implicit functions to learn how to generate grasping motion for manipulation with a dexterous hand. The proposed model, named Continuous Grasping Function (CGF), can generate a continuous and smooth grasping plan with continuous time inputs. CGF is learned through generative modeling with a Conditional Variational Autoencoder using 3D human demonstrations. The CGF is trained by converting large-scale human-object interaction trajectories to robot demonstrations via motion retargeting. During inference, sampling is performed with CGF to generate different grasping plans in the simulator, and successful plans are selected for transfer to the real robot. By training on diverse human data, our CGF allows generalization to manipulate multiple objects. Compared to previous planning algorithms, CGF is more efficient and achieves significant improvement on success rate when transferred to grasping with the real Allegro Hand.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Computer Science, Information Systems
Jiaming Zhang et al.
Summary: This paper presents a sampled data control method for a flexible robotic manipulator with unknown internal dynamics. The method considers the fuzzy logical system and addresses the limited network bandwidth issue. By utilizing the Lyapunov stability theory, the signals of the flexible robotic manipulator are guaranteed to be ultimately bounded.
Article
Computer Science, Information Systems
Salvatore Vitabile et al.
Summary: CGA is a new geometric computation tool that is widely used in robotics, and this paper presents a novel embedded coprocessor to accelerate CGA operations, which has demonstrated significant performance improvements through two robotic algorithms.
Proceedings Paper
Computer Science, Theory & Methods
Malte Mosbach et al.
Summary: This study proposes a reinforcement learning framework that learns the interactive grasping of various geometrically distinct real-world objects by controlling a humanoid robotic hand. The results show that the framework is capable of learning and performing targeted grasping as well as utilizing environmental constraints under different conditions.
2022 SIXTH IEEE INTERNATIONAL CONFERENCE ON ROBOTIC COMPUTING, IRC
(2022)
Proceedings Paper
Automation & Control Systems
Lars Berscheid et al.
Summary: This research extends the use of fully-convolutional neural networks to estimate grasp rewards by parametrizing the remaining degrees of freedom of primitives for further optimization of robot learning in planar manipulation. By introducing a model-based controller, the robot is able to learn collision avoidance, maximize grasp quality, and keep uncertainties small during training, significantly improving grasp success rate and speed.
2021 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2021)
(2021)
Proceedings Paper
Automation & Control Systems
Shirin Joshi et al.
2020 IEEE 16TH INTERNATIONAL CONFERENCE ON AUTOMATION SCIENCE AND ENGINEERING (CASE)
(2020)
Article
Robotics
Yang Yang et al.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2020)
Article
Robotics
Cosimo Della Santina et al.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2019)
Proceedings Paper
Automation & Control Systems
Michael Hegedus et al.
2019 INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA)
(2019)
Article
Robotics
Sergey Levine et al.
INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH
(2018)
Article
Robotics
Changhyun Choi et al.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2018)
Proceedings Paper
Computer Science, Artificial Intelligence
Gao Huang et al.
30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017)
(2017)
Article
Computer Science, Artificial Intelligence
Roland Hafner et al.