Exploration Robot Based On YOLOv8 Algorithm

Incidents or natural disasters, such as earthquakes, often create hazardous environments that are inaccessible to human rescuers. The extreme-condition exploration robot is the solution to it. The robot is designed to operate in these kinds of extreme conditions to assist in search and rescue operations. Equipped with sensors and utilizing machine learning capabilities, the robot can navigate through debris, detect gases, detect humans, and transmit real-time data to the rescue team. In this work, we propose a prototype extreme-condition exploration robot. The robot is equipped with an ESP32-CAM module. The video can be streamed to the server for object detection tasks. To get faster streaming times, JPEG compression is employed. Then, the YOLOv8 algorithm is employed for object detection. The YOLOv8 model is trained with 180 epochs to be able to detect three different classes. The model reached a precision of 0.951 for all classes and 0.928 mAP50. The model surpasses our target accuracy of 75% for our application.