标题: R2D-EQ: a two-stage workflow for risk reasoning and decision-making in earthquake emergency scenarios
作者: Yao, LW (Yao, Liwei); Ren, F (Ren, Fu); Du, QY (Du, Qingyun)
来源出版物: EXPERT SYSTEMS WITH APPLICATIONS 卷: 299 文献号: 130178 DOI: 10.1016/j.eswa.2025.130178 Published Date: 2026 MAR 1 子辑: C
摘要: Effective decision-making is crucial in earthquake emergency response, where rapid and well-informed actions are vital to saving lives and minimizing losses. However, current methods often struggle to rapidly acquire and effectively analyze localized historical data under specific disaster scenarios, achieve real-time risk identification, and seamlessly connect these insights to emergency decision-making. This study proposes R2D-EQ, a two-stage workflow designed to enhance risk reasoning and decision-making in earthquake emergencies. The framework integrates real-time web intelligence, event evolutionary graph construction, and retrieval-augmented generation (RAG) to dynamically uncover risk propagation chains and generate context-specific response recommendations. By combining symbolic reasoning with neural generation, R2D-EQ offers an interpretable, adaptive, and scalable solution for complex disaster management scenarios. System evaluations show that the framework achieves strong performance in risk perception, risk reasoning, and response generation. Notably, causal risk chain modeling enhances interpretability, while LightRAG-based decision generation improves precision and domain-level professionalism. Despite these advances, the framework faces limitations, including high computational costs, challenges in managing event generalization, limited geographic and contextual coverage in the knowledge base, and difficulties in detecting unknown risks. Future work will focus on incorporating multimodal data sources, embedding geospatial computation tools as agent-accessible modules, and developing multi-agent collaboration mechanisms, aiming to strengthen the system's adaptive and cooperative potential in nextgeneration disaster management.
作者关键词: EEG; RAG; LLM; Workflow; Decision-Making; Risk Reasoning
地址: [Yao, Liwei; Ren, Fu; Du, Qingyun] Wuhan Univ, Sch Resources & Environm Sci, 129 Luoyu Rd, Wuhan 430079, Peoples R China.
[Ren, Fu; Du, Qingyun] Wuhan Univ, Key Lab GIS, Minist Educ, Wuhan 430079, Peoples R China.
[Ren, Fu; Du, Qingyun] Wuhan Univ, Collaborat Innovat Ctr Geospatial Technol, Wuhan 430079, Peoples R China.
通讯作者地址: Du, QY (通讯作者)，Wuhan Univ, Sch Resources & Environm Sci, 129 Luoyu Rd, Wuhan 430079, Peoples R China.
电子邮件地址: [email protected]; [email protected]; [email protected]
影响因子：7.5
