A Mobile Robot Framework in Industrial Disaster for Human Rescue

In this paper, we develop a rescue robot framework for first-step rescue work that can help rescue teams. We model a disaster environment in detail and updated the layered map and divided the topography as per the reachability of the robot using memory-efficient 3D Octomap. And we construct an autonomous driving algorithm for efficient and fast calculation by projecting it into a 2D map. Also, unlike previous studies that detect a moving person with an onboard sensor, we propose a system that detects a person’s posture by converting the 2D keypoint coordinates extracted through pose estimation using a depth camera into 3D. Keypoints converted to 3D distinguish a person’s posture and show the current state with a 3D map. And a person’s pose is discriminated by using the coordinates and the posture is sent to the team. Therefore, the whole process was experimented, and we confirmed that we can provide the first stage of an efficient human structure framework.