| Abstract | Ships and offshore structures are constructed in a way that several units of blocks are assembled. In the block lifting process, a block is controlled indirectly through the interaction of the crane, hooks, equalizers, and wire ropes (simply, wires). Therefore, it is difficult to control the unexpected movement motions of the block accurately with several connected objects. Furthermore, it is important to construct a robust control model to cope with the modeling uncertainty of the block and the change of lug arrangement. In this study, we proposed the automatic control of the block lifting process with deep reinforcement learning (DRL), which can provide a robust control under uncertainties. The state of a block and wires, such as position, orientation and angular velocity of the block, and the lifting speed of each wire, were used as the input of DRL. Then, the lifting speed of each wire was obtained as the output action of DRL. The reward was applied to reduce the roll and pitch angle of the block and to stabilize the speed of block lifting. The proposed method was applied to various simulation examples of block lifting and compared with traditional control algorithms. As a result, it was confirmed that the proposed method could effectively control the block with unexpected motions due to the modeling uncertainty and the change of the lug arrangement. |
|---|---|
| Publication Date | 2022-10-09 |
Do-Hyun Chun, Myung-Il Roh, Hye-Won Lee, Seung-Ho Ham, "A Method for Automatic Control of Cranes for Block Lifting in Shipyard", Proceedings of International Symposium on PRADS(Practical Design of Ships and Other Floating Structures) 2022, Dubrovnik, Croatia, pp. 64, 2022.10.09-13