International Conference

Do-Hyun Chun, Myung-Il Roh, Hye-Won Lee, Seung-Ho Ham, "A Crane Movement Control for Stability of Block Erection Based on Deep Reinforcement Learning", MIM 2019, Berlin, Germany, 2019.08.28-30

by SyDLab posted Jul 03, 2019
?

단축키

Prev이전 문서

Next다음 문서

ESC닫기

크게 작게 위로 아래로 댓글로 가기 인쇄
Extra Form
Abstract The block erection using a crane such as a gantry crane and a floating crane is one of the most important processes in the production of ships and offshore structures. To mount the block with the correct position and angle, it is important to control the block accurately not to occur the unexpected movement like rotation. However, it is difficult to apply the existing control theory to the block lifting operation because the movement of the block is controlled indirectly with the control of various objects such as the crane and wire ropes. To solve this problem, a block control method based on deep reinforcement learning is proposed in this study. The proposed method is easier to control the block with wire ropes and to consider irregular external force than existing control theory. In this study, the angle and angular velocity of the lifting block and the hoisting speed of each wire rope that can affect the motion of the block are set as states of reinforcement learning, and the hoisting speed that is the control object is set as an action of reinforcement learning. The reward function of reinforcement learning is designed to increase when the angle of the block decrease and the speed of the block is close to the target speed. In this study the policy gradient method which is a kind of policy-based methods of deep reinforcement learning is used to solve the problem with continuous states and action. To check the applicability and feasibility of the proposed method, The block lifting simulation is performed using the existing control theory and the proposed method. We compared the proposed method with the existing control theory. The result shows that the proposed method can minimize the motion of the lifting block more effectively than the existing control theory.
Publication Date 2019-08-28

Do-Hyun Chun, Myung-Il Roh, Hye-Won Lee, Seung-Ho Ham, "A Crane Movement Control for Stability of Block Erection Based on Deep Reinforcement Learning", MIM(International Federation of Automatic Control) 2019, Berlin, Germany, 2019.08.28-30