主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2025, Vol. 44 ›› Issue (2): 181-189.

• • 上一篇    下一篇

室内自主巡检消防灭火机器人设计

李森1, 马小飞1, 冯春勇2, 刘合钊1   

  1. （1.郑州轻工业大学 建筑环境工程学院，河南 郑州 450000； 2.西安建筑科技大学 机电工程学院，陕西 西安 710055）
  • 收稿日期:2024-05-26 修回日期:2024-09-19 出版日期:2025-02-15 发布日期:2025-02-15
  • 作者简介:李 森，郑州轻工业大学建筑环境工程学院副教授，博士，硕士生导师，主要从事建筑SLAM地图构建、火场智能救援方面的研究，河南省郑州市高新区科学大道136号，450007，lisen@zzuli.edu.cn。
  • 基金资助:
    河南省高校青年骨干教师培养计划（2021GGJS094）；河南省科技攻关项目（242102320057）

Design of an autonomous indoor inspection and firefighting robot

Li Sen1, Ma Xiaofei1, Feng Chunyong2, Liu Hezhao1   

  1. (1. College of Architecture and Environmental Engineering, Zhengzhou University of Light Industry, Zhengzhou Henan 450000, China; 2. School of Mechanical and Electrical Engineering, Xi'an University of Architecture and Technology, Xi'an Shanxi 710055, China)
  • Received:2024-05-26 Revised:2024-09-19 Online:2025-02-15 Published:2025-02-15

摘要： 研制了一款具备自主巡检和自动灭火功能的智能化消防机器人，以应对火灾中的复杂环境和紧急情况。在方法上，基于ROS系统，结合SLAM和深度学习目标检测技术，设计了履带式移动载体机器人，通过融合激光雷达、IMU和里程计等传感器数据来构建SLAM地图，并结合热成像与YOLO技术进行火焰识别与定位。在SLAM方面，选择并优化了Cartographer，以解决大规模建筑环境中的构图误差问题。此外，提出了YOLOv4-F火焰检测模型，并将其移植至ROS系统，实现自动瞄准功能。结果表明，优化后的SLAM算法在大规模环境中，有效减少了构图误差，YOLOv4-F模型的mAP@0.5达85.53%，实时检测精度为81%，速度为11 帧/秒。该机器人在试验中，成功实现了自主巡检和自动灭火功能，能够在烟气环境中实时准确识别火焰并迅速扑灭初期火灾，为智能消防机器人的实际应用提供有效的技术支持和新思路。

关键词: ROS系统, 消防机器人, 智能化, SLAM, YOLO

Abstract: This article develops an intelligent firefighting robot with autonomous inspection and automatic fire extinguishing functions to cope with complex environments and emergency situations in fires. In terms of methodology, a tracked mobile carrier robot was designed based on the ROS system, combined with SLAM and deep learning object detection technology. The SLAM map was constructed by integrating sensor data such as LiDAR, IMU, and odometer, and flame recognition and localization were performed using thermal imaging and YOLO technology. In terms of SLAM, a Cartographer was selected and optimized to address the issue of composition errors in large-scale building environments. In addition, the YOLOv4-F flame detection model was proposed and transplanted to the ROS system to achieve automatic aiming function. The results indicate that the optimized SLAM algorithm effectively reduces composition errors in large-scale environments, as demonstrated by the YOLOv4-F model mAP@0.5 reaching 85.53%, with a real-time detection accuracy of 81% and a speed of 11 frames per second. In the experiment, the robot successfully achieved autonomous inspection and automatic fire extinguishing functions, which can accurately identify flames in smoke environment in real time and quickly extinguish initial fires, providing effective technical support and new ideas for the practical application of intelligent firefighting robots.

Key words: ROS system, firefighting robot, intelligence, SLAM, YOLO