Research review on damage assessment driven by artificial intelligence technologies

Damage assessment serves as a pivotal link bridging target selection, operational planning, and combat feedback. It encompasses both pre-strike weapon damage effectiveness assessment oriented toward damage probability, target vulnerability, and weapon-target matching, and post-strike battlefield damage effect assessment oriented toward target state, functional degradation, and mission effectiveness. With advances in machine learning, multi-source sensing, numerical simulation, and intelligent computing, damage assessment is transitioning from static evaluation reliant on empirical rules, manual interpretation, and single-source data toward intelligent assessment that integrates perceptual recognition, data-driven modeling, knowledge reasoning, multi-source fusion, and closed-loop decision-making. This paper presents a review of research on artificial intelligence (AI)-driven damage assessment. After clarifying the connotation of damage assessment and the enabling logic of AI, we summarize key technical progress from five perspectives: perception and feature extraction, data-driven assessment models, knowledge-driven assessment methods, multi-source fusion, and intelligent decision-making and optimization. We further discuss the application chain covering pre-strike prediction, post-strike assessment, system-of-systems target assessment, and assessment-driven re-decision-making. The results indicate that AI can substantially improve the efficiency of damage information acquisition, feature extraction, effect prediction, state estimation, and action feedback. Nevertheless, challenges persist, including insufficient real-world samples, simulation bias, unclear functional mapping, multi-source information conflicts, inadequate model generalization and trustworthy verification, and the difficulty of stably incorporating assessment results into decision-making workflows. Future research should concentrate on data credibility, mechanism constraints, knowledge enhancement, multi-source fusion, explainable verification, and human-machine collaboration, so as to construct an intelligent damage assessment system tailored for complex combat environments.