Accidents due to undetected fires have caused huge losses in various contexts in the world, such as in offices, manufacturing workplaces, residential areas, even in forest areas. The evacuation process in the firefighting system requires an effective and fast response. The source of fire must be found quickly to prevent it from spreading. Image processing based on an object detection system, it is believed, can circumvent this problem. The method for detecting objects is performed by the You Only Look Once (YOLO) algorithm in real-time. The most suitable YOLO model for the system is the Tiny YOLO VOC model which was built with the Darkflow framework. Besides detecting the fires, the robot can be used to ease the search process and firefighting efforts. However, it needs planning to fit the robots with the system optimized to do the best in the minimum time possible. In robotics, finding the fastest path can be solved by embedding the A* algorithm in the robot. A* algorithm is one of the artificial intelligence methods for finding the fastest path. Not many studies have applied this algorithm to the wheel robots to find the shortest path. Therefore, this A* method will be used for finding the fastest path to find the fire with accuracy by finding the best distance and route with the coordinate system. In this study, the wheel robots had three missions, finding the fire, extinguishing the fire, and returning to the starting points. The system accuracy for those missions was 100%, 83.33%, and 50%, respectively.

Chapter Contents:

• 17.1 Introduction
• 17.2 Robot design
• 17.3 Fire detection using YOLO
• 17.4 Shortest path using A* algorithm
• 17.4.1 A* algorithm
• 17.4.2 Map and node design
• 17.4.3 How the robots work
• 17.5 Results and discussions
• 17.5.1 Position of robots and fire
• 17.5.2 Collecting data for YOLO
• 17.5.3 Training YOLO
• 17.5.4 Evaluation of the shortest path
• 17.5.5 Receiving data
• 17.5.6 Evaluation of the entire system
• 17.6 Conclusions
• References

Inspec keywords: augmented reality; control engineering computing; haptic interfaces; occupational safety; emergency services; mobile robots; fire extinguishers; artificial intelligence; object detection; computer based training

Other keywords: darkflow framework; fire extinguishing; firefighting system; residential areas; forest areas; evacuation process; AR technologies; haptic technologies; smart manufacturing workplace safety; YOLO VOC model; wheel robots; coordinate system; virtual training; search process; image processing; object detection system; artificial intelligence methods

Subjects: Human resource management; Control in industrial production systems; User interfaces; Robot and manipulator mechanics; Robotics; Health and safety aspects; Optical, image and video signal processing; Knowledge based systems; Education and training; Computer-aided instruction; Mobile robots; Control engineering computing; Computer vision and image processing techniques; Virtual reality