Scholarlink

Smart cities face critical challenges in disaster response due to latency, privacy concerns, and heterogeneous data from multi-modal sensors (e.g., LiDAR, drones, IoT). Traditional centralized AI models struggle with real-time processing while preserving privacy. This paper proposes a federated learning (FL) frame- work [1] to optimize disaster response through decentralized, privacy-preserving model training across edge devices. By integrating multi-modal data (environmental, seismic, video) with adaptive coordination algorithms, the system achieves sub-500 ms latency and 92% earthquake prediction accuracy. Case studies show 40% faster flood detection and 35% reduced wildfire response time versus centralized methods. Transformer-based fusion and lightweight models address data heterogeneity and resource constraints, advancing resilient smart cities via privacy-aware real-time analytics.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging Federated Learning for Real-Time Disaster Response Optimization in Smart Cities Using Multi-modal Sensor Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Lakshmojee Koduru,
  • Sri Ram Chandra Polisetty,
  • Mohana Naga Vamsi Thalatam,
  • Shubneet,
  • Anushka Raj Yadav,
  • Navjot Singh Talwandi

摘要

Smart cities face critical challenges in disaster response due to latency, privacy concerns, and heterogeneous data from multi-modal sensors (e.g., LiDAR, drones, IoT). Traditional centralized AI models struggle with real-time processing while preserving privacy. This paper proposes a federated learning (FL) frame- work [1] to optimize disaster response through decentralized, privacy-preserving model training across edge devices. By integrating multi-modal data (environmental, seismic, video) with adaptive coordination algorithms, the system achieves sub-500 ms latency and 92% earthquake prediction accuracy. Case studies show 40% faster flood detection and 35% reduced wildfire response time versus centralized methods. Transformer-based fusion and lightweight models address data heterogeneity and resource constraints, advancing resilient smart cities via privacy-aware real-time analytics.