An Evaluation System for Ground Warfare Reconnaissance and Strike Links Based on Reinforcement Learning: Designing a Multidimensional Evaluation Framework for Intelligent Combat Systems

As the complexity of modern ground warfare environments continues to increase, traditional combat evaluation methods struggle to meet the rapidly changing demands of the battlefield. This paper proposes a novel evaluation system for the ground warfare reconnaissance and strike link based on a reinforcement learning framework. The system centers on intelligent agents and introduces four types of agents-reconnaissance, decision-making, strike, and logistics-to simulate and evaluate their contributions to combat effectiveness during operations. First, reinforcement learning models were designed for each type of agent based on their roles and functions, and training and testing were conducted across various combat environments. Second, leveraging the training results of the reinforcement learning algorithms, a multidimensional evaluation metric system is proposed to comprehensively assess the performance and collaborative effectiveness of the agents. Experimental results demonstrate that the proposed evaluation system can effectively reflect the operational outcomes of the ground warfare reconnaissance and strike link, providing theoretical support for the design and optimization of future intelligent combat systems. Finally, the paper discusses the challenges of applying this evaluation system in practice and outlines potential directions for future research.