火力与指挥控制2026，Vol.51Issue(1)：133-141,147,10.DOI:10.3969/j.issn.1002-0640.2026.01.017
面向火炮模拟器人机交互的大模型长时记忆机制
A Long-term Memory Mechanism Powered by Large Language Models for Human-computer Interaction in Artillery Simulator Systems
邓呈泽 1李世兴 1武亮明 1张红梅1
作者信息
- 1. 北方自动控制技术研究所,太原 030006
- 折叠
摘要
Abstract
Aiming at the problems of low training data utilization rate in the traditional human-computer interaction(HCI)module of artillery simulation training systems and insufficient interactivity during the training process,we propose to construct a HCI application system based on large language models(LLMs).We build a vector database that integrates professional knowledge in the artillery field and historical Q&A data,and enable the connection between LLMs and the vector database via LangChain.Additionally,we optimize the text processing algorithm to enhance the system's ability to recognize professional terms and improve word segmentation performance.Experimental results demonstrate that the system exhibits long-term memory capabilities,can provide professional responses to training-related questions,and effectively improves both interaction efficiency and training data utilization.关键词
记忆机制/大模型/向量数据库/火炮模拟训练/文本处理Key words
memory mechanism/LLMs/vector database/artillery simulation training/text processing分类
军事科技引用本文复制引用
邓呈泽,李世兴,武亮明,张红梅..面向火炮模拟器人机交互的大模型长时记忆机制[J].火力与指挥控制,2026,51(1):133-141,147,10.
