A system and method for blocking potential risks in naval combat systems utilizing reinforcement learning

Technology for controlling risks in reinforcement learning-based naval combat systems through AI mode definition

This technology relates to a system and method for blocking potential risks in naval combat systems that utilize reinforcement learning. Specifically, it is a technology that controls the AI's level of intervention during combat simulation and operational execution through the AI terminal's tactical inference module and AI mode provision module.

Existing reinforcement learning-based combat systems, if trained solely on achieving objectives, could not sufficiently reflect rules of engagement, operational guidelines, and the necessity of human intervention, potentially leading to dangerous decision-making. To address this, this technology proposes a method that provides definitions of intervention levels, such as AI Mode 0 to 5, and performs tactical inference and constraint reflection according to the corresponding mode before the agent acts.

Accordingly, this technology can progressively manage the autonomy level of military AI and secure the supervisory capability of operational personnel, thereby contributing to ensuring the safety of defense AI, naval combat systems, and high-risk reinforcement learning systems.

Key Features

• Defining the AI's level of intervention in reinforcement learning-based naval combat systems in AI Mode stages
• Performing tactical selection and constraint reflection before action through the tactical inference module and AI mode provision module
• Supports reflecting rules of engagement, operational guidelines, and the necessity of human intervention into the reinforcement learning model
• Applicable to the safety control of defense AI, naval combat systems, and high-risk autonomous systems