Simultaneous Multi-View Object Recognition and Grasping in Open-Ended Domains

Article Properties
Abstract
Cite
Kasaei, Hamidreza, et al. “Simultaneous Multi-View Object Recognition and Grasping in Open-Ended Domains”. Journal of Intelligent &Amp; Robotic Systems, vol. 110, no. 2, 2024, https://doi.org/10.1007/s10846-024-02092-5.
Kasaei, H., Kasaei, M., Tziafas, G., Luo, S., & Sasso, R. (2024). Simultaneous Multi-View Object Recognition and Grasping in Open-Ended Domains. Journal of Intelligent &Amp; Robotic Systems, 110(2). https://doi.org/10.1007/s10846-024-02092-5
Kasaei H, Kasaei M, Tziafas G, Luo S, Sasso R. Simultaneous Multi-View Object Recognition and Grasping in Open-Ended Domains. Journal of Intelligent & Robotic Systems. 2024;110(2).
Journal Categories
Science
Mathematics
Instruments and machines
Electronic computers
Computer science
Technology
Electrical engineering
Electronics
Nuclear engineering
Electronics
Technology
Engineering (General)
Civil engineering (General)
Technology
Mechanical engineering and machinery
Description

Can a robot learn to grasp new objects without retraining? This research proposes a deep learning architecture with augmented memory to handle open-ended object recognition and grasping simultaneously. The robot is able to use multi-views of an object as input and jointly estimates pixel-wise grasp configuration as well as a deep scale- and rotation-invariant representation as output. This approach uses meta-active learning for open-ended object recognition, resolving the issue of catastrophic forgetting when encountering new object categories. The obtained representation is then used for open-ended object recognition through a meta-active learning technique. This enables a robot to continuously acquire knowledge about new object categories. The approach allows the robot to learn new object categories rapidly using very few examples on-site, demonstrating effective grasping of never-seen-before objects. The system shows high object recognition accuracy and grasp success rates in cluttered environments, empowering robots to adapt and acquire new knowledge effectively.

Refrences