【摘要】 潜艇水下破损进水严重威胁其生命力，及时的决策响应至关重要。现有抗沉决策大多基于规则推理或案例匹配，因受限于规则库或案例库，对复杂动态工况的损伤应对能力明显不足。针对此问题，本文提出基于双重深度Q网络（Double Deep Q-Network,DDQN）的抗沉辅助决策方法。首先，通过破损进水机理构建多维状态空间，整合静力与动力抗沉措施并映射为离散化损管动作集合，创新设计考量生存能力核心要素的复合加权奖励函数；然后，结合随机经验回放与贪婪衰减探索机制，采用DDQN算法框架实现抗沉决策；最后，构建生命力评价指标，将所提方法与深度Q网络（Deep Q-Network,DQN）进行仿真对比。仿真结果表明，典型工况下DDQN抗沉决策使潜艇生命力指标更优，验证了方法的有效性和优越性。更多还原

【Abstract】 Underwater damage and flooding of submarines pose a severe threat to their vitality, making timely decision-making responses crucial. Existing anti-sinking decision-making methods are mostly based on rule-based reasoning or case matching. However, limited by rule bases or case bases, their capability to respond to damage under complex dynamic operating conditions is significantly insufficient. To address this issue, this paper proposes an auxiliary antisinking decision-making method based on the Double Deep Q-Network(DDQN). First, a multi-dimensional state space is constructed based on the damage and flooding mechanism; static and dynamic anti-sinking measures are integrated and mapped into a discretized damage control action set; and a composite weighted reward function that takes into account the core elements of viability is innovatively designed. Second, by combining random experience replay and the epsilon-greedy exploration with decay mechanism, the DDQN algorithm framework is adopted to realize anti-sinking decision-making. Finally, vitality evaluation indicators are established, and a simulation comparison is conducted between the proposed method and the Deep Q-Network(DQN). The simulation results show that the DDQN-based anti-sinking decision-making yields better submarine vitality indicators under typical operating conditions, which verifies the effectiveness and superiority of the proposed method.更多还原