Toward Reliable Bin-Picking: Collision-Aware Robotic Design and Control Strategy for Heavily Cluttered Environment

Reliable robotic grasping is a fundamental capability for automating tasks traditionally performed by humans. Many attempts have been made, but existing systems that utilize single modality grasping -only suction or finger grasping- have had difficulty solving this problem in environments where objects with various types and characteristics are mixed. In addition, many variables can occur when complex objects are cluttered or there are external constraints, such as bins. To configure a reliable system, we propose a network that combines each modality previously studied, along with a strategy to protect the robot's hardware, ultimately ensuring reliability. We participated in the Robotic Grasping and Manipulation Competition (RGMC) held at the International Conference on Robotics and Automation (ICRA) 2024 with various functions designed in the laboratory. During this process, the presented functions were verified once again at the competition site, and the applicability in actual industrial sites was confirmed.