HIMAR: Human-Informed Map Augmentation with Panoramic LiDAR Imaging for Enhanced Robotic Search-and-Rescue

This paper introduces HIMAR, a semi-autonomous robotic system designed to enhance search and rescue (SAR) operations in challenging environments. The system integrates autonomous navigation for systematic exploration, human-informed map augmentation to localize survivors, and real-time teleoperation for situational insights. Central to HIMAR is a 3D object detection framework that utilizes fixed-resolution, 360°panoramic LiDAR images, where each pixel value quantifies the intensity of the returned light, effectively encoding photon detection data. A CNN-based detector (YOLO) processes these panoramic images to extract object-level information, which is then combined with corresponding 3D point clouds to achieve precise survivor localization within the map. As a semi-autonomous first responder, the robot enhances situational awareness and minimizes risks to the SAR team by providing a human-augmented map of hazardous zones, enabling informed decision-making before deployment. Experimental validation demonstrated the system's effectiveness, delivering improved accuracy and inference performance compared to previous methods. Additionally, the system exhibits enhanced generalization capabilities compared to conventional approaches, enabling the generation of high-precision semantic maps specifically optimized for SAR tasks.