Article
Machine Learning in Robotic Grasping Tasks: A Survey
Ali R., Al Akkad M. A.
Article language: English
Abstract. Recently machine learning techniques, including deep learning and reinforcement learning, have been considered as the milestone in the field of computer vision and vision-based robot tasks, such as grasping. This survey presents a set of recent approaches in the field of object pick-and-place tasks and object grasping tasks. These approaches are categorized into two groups, deep-learning-based approaches, and reinforcement-learning-based approaches. Task-oriented grasping decisions for humans, are made intuitively, while it is a big challenge for robots to achieve the grasping tasks as proficient as humans. Several conditions affect the performance of robot grasping such as changes in environment and illumination, existence of a huge number of objects with different properties, complex backgrounds, and occlusion between objects Machine learning techniques are implemented in robotic systems to improve the capability of these robots to handle these conditions and guarantee high performance.
Keywords: robotics, machine learning, deep learning, reinforcement learning, grasping tasks, object detection
Pages: 164–170Total pages: 7
Year of publication: 2020
Цитирование по ГОСТ 7.1-2003:GOST 7.1-2003 citation:
Ali R., Al Akkad M. A. Machine learning in robotic grasping tasks: a survey // Instrumentation Engineering in the XXI Century – 2020. Integration of Science, Education and Production : Proceedings of the XVI Russian National Scientific Conference (December 2–4, 2020, Izhevsk, Russian Federation). – Izhevsk, Russia : Publishing House of Kalashnikov ISTU, 2020. – Pp. 164–170.