Research on the Collaborative Perception and Decision Mechanism of Vision and Force Perception for Industrial Robot Grasping Tasks
DOI: https://doi.org/10.62517/jes.202602108
Author(s)
Ding Cong
Affiliation(s)
Zhengzhou Technical College of Commerce, Henan, China
Abstract
Aiming at the problem of low grasping accuracy and insufficient robustness of industrial robots in complex scenes, a visual and force perception collaborative perception and decision-making mechanism is proposed. By integrating the hardware of "binocular stereo vision+structured light 3D camera" and six axis force sensor, a multimodal feature fusion network based on attention mechanism (AMFF Net) is constructed to achieve deep data fusion; Design a reinforcement learning grasping point decision model and a force position hybrid dual closed-loop control strategy to form a "perception decision execution" closed-loop system. The verification results in three typical scenarios of automobile manufacturing, 3C electronic assembly, and logistics sorting show that this mechanism can control the grasping positioning error within ± 0.3mm, and improve the success rate of grasping in complex working conditions to over 95%, meeting the high-precision and high robustness requirements of flexible production.
Keywords
Industrial Robots; Collaborative Perception; Visual Sensory Fusion; Reinforcement Learning; Force Position Hybrid Control
References
[1] Wang Yaonan, Jiang Yiming, Jiang Jiao, etc Key Technologies of Robot Perception and Control and Their Intelligent Manufacturing Applications [J]. Automation Expo, 2023, 40 (10): 50-66
[2] Gu Xin Research on Multi modal Perception and Grasping Detection of Robots Based on Spatiotemporal Attention Mechanism [D]. Guangdong University of Technology [2021-12-14]
[3] Xue Songdong Research on Coordinated Control and Simulation of Target Search Oriented Swarm Robots [D]. Lanzhou University of Technology [2021-12-14]
[4] Jie Yinggang, Lanjiang Rain A review of research on collaborative robots and their motion planning methods [J]. Computer Engineering and Applications, 2021, 57 (13): 16
