Research on Auxiliary Decision-Making System for Manned Underwater Vehicle Damage Management Based on Deep...

Area of Science:

• Robotics and Autonomous Systems
• Artificial Intelligence
• Marine Engineering

Background:

• Manned underwater vehicles (MUVs) face significant risks from obstacles and equipment damage, impacting operational continuity.
• Current damage management systems struggle with dynamic, unpredictable underwater environments.
• Effective damage control is crucial for MUV functionality recovery and mission success.

Purpose of the Study:

• To propose an Auxiliary Decision-Making System (ADMS) for MUV damage management using Deep Reinforcement Learning (DRL).
• To address the critical issue of cabin flooding and enhance MUV survivability in emergency situations.
• To develop adaptive strategies for dynamic MUV damage scenarios.

Main Methods:

• Development of an ADMS leveraging Deep Reinforcement Learning (DRL) for real-time decision support.
• Construction and digitization of a comprehensive States-Actions cluster for diverse damage scenarios.
• Design and implementation of novel reward functions to ensure safe and effective DRL strategies.
• Definition and analysis of MUV buoyancy and stability vitality evaluation criteria.

Main Results:

• Simulations demonstrate the ADMS provides effective and rational damage management measures.
• The system achieved buoyancy vitality exceeding 38% and stability vitality surpassing 92% in simulated damage states.
• Optimal stability vitality values exceeding 99% were observed, indicating high system performance.

Conclusions:

• The proposed DRL-based ADMS significantly improves MUV damage management capabilities, particularly for cabin flooding.
• The system offers adaptive decision-making crucial for navigating dynamic and hazardous underwater environments.
• The defined vitality criteria provide a robust framework for evaluating MUV resilience and system effectiveness.