基于DDPG算法的耙吸挖泥船耙头活动罩控制策略研究  

Research on control strategy of draghead visor of TSHD based on DDPG algorithm

作　　者：张红升 衣凡 邓伍三 ZHANG Hong-sheng;YI Fan;DENG Wu-san(National Engineering Research Center of Dredging,Shanghai 200082,China)

机构地区：[1]中交疏浚技术装备国家工程研究中心有限公司,上海200082

出　　处：《中国港湾建设》2022年第9期7-10,80,共5页China Harbour Engineering

基　　金：工业和信息化部高技术船舶科研项目(MC-202033-P05);交通运输部交通运输行业重点科技项目(2020-MS2-079)。

摘　　要：针对耙吸挖泥船挖泥装舱作业过程产量优化问题,提出了基于强化学习理论中的DDPG(深度确定性策略梯度,Deep Deterministic Policy Gradient)算法,设计了主动耙头活动罩控制器。控制器根据当前施工状态判断耙头活动罩的最优对地角度,通过耙头活动罩对地角度控制吸入密度,进而控制原状土瞬时产量,实现产量优化。在“新海虎8”轮进行了算法验证,与操耙手人工操作耙头活动罩对地角度进行对比,结果表明基于DDPG算法的主动耙头活动罩控制器可以有效提高产量。In this paper,an active draghead visor controller is proposed based on DDPG(Deep Deterministic Policy Gradient)algorithm to optimize the output of loading process of TSHD.The controller will autonomously control the optimum visor angle relative to the ground according to the current working status of the draghead,and control the density of suction through the visor angle relative to the ground,thus controlling the instantaneous production of the undisturbed soil to optimize the production.The controller was verified on XINHAIHU8.Compared with human control,the active draghead visor controller based on DDPG algorithm can improve production effectively.

关 键 词：耙吸挖泥船 耙头活动罩 强化学习 产量优化 DDPG算法 

分 类 号：U616[交通运输工程—船舶及航道工程]