Abstract
This study introduces a novel two-stage approach for robotic grasp detection, addressing the challenges faced by end-to-end deep learning methodologies, particularly those based on convolutional neural networks (CNNs) that require extensive and often impractical datasets. Our method first leverages a particle swarm optimizer (PSO) as a candidate estimator, followed by CNN-based verification to identify the most probable grasp points. This approach represents a significant advancement in the field, achieving an impressive accuracy of 92.8% on the Cornell Grasp Dataset. This positions it among the leading methods while maintaining real-time operational capability. Furthermore, with minor modifications, our technique can predict multiple grasp points per object, offering diverse grasping strategies. This adaptability and high performance suggest substantial potential for practical applications in robotic systems, enhancing their efficiency and reliability in dynamic environments.
Keywords
Robotic grasp detection
Convolutional neural network
Two-stage cascaded system
Particle swarm optimizer
Funding
This work was supported without any funding.
Cite This Article
APA Style
Chu, Z., Hung, M. & Chen, X. (2024). Enhancing Robotic Grasp Detection with a Novel Two-Stage Approach: From Conceptualization to Implementation. IECE Transactions on Intelligent Unmanned Systems, 1(1), 44–54. https://doi.org/10.62762/TIUS.2024.777385
Publisher's Note
IECE stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Institute of Emerging and Computer Engineers (IECE) or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
