A Rugged Tactical Perception System for AI-Driven Casualty Registration and Medical Scene Awareness

We propose the development and evaluation of a rugged, edge-capable perception system for autonomous medical scene understanding in austere environments. Designed for both helmet-mounted and unmanned ground vehicle (UGV) configurations, the system enables real-time detection of casualties, upper-body skeletal tracking, medical object recognition, and automated scene documentation during Tactical Combat Casualty Care (TCCC). All perception and inference operations are performed locally on embedded hardware with no reliance on cloud connectivity or active illumination, making the system fully passive and suitable for outdoor deployment.The integrated hardware-software stack combines passive stereo RGB, thermal imaging, and inertial measurement (IMU) sensors fused through a modular AI pipeline. The core perception stack includes multi-subject human registration with persistent IDs, upper-body pose classification, body part segmentation, and detection of 10+ TCCC-critical tools such as tourniquets, IV bags, and syringes. These outputs are structured into time-stamped, open-format logs compatible with autonomous documentation systems like AutoDoc.All sensing and computation runs on a shared edge platform powered by the NVIDIA Jetson Orin Nano. The helmet-mounted system is designed to weigh under 500 grams, operate for over 2 hours on a swappable battery, and meet IP65 environmental ratings. The UGV-mounted variant shares the same sensor and compute core, enabling streamlined development, training, and sustainment across configurations. The system is engineered to perform reliably under operationally relevant conditions including direct sunlight, low light, motion blur, occlusion, camouflage, and rough terrain.Phase II will deliver fully integrated prototypes, validate system performance in realistic multi-casualty scenarios, and complete ruggedization testing (drop, ingress, thermal) in preparation for evaluation at the Telemedicine and Advanced Technology Research Center (TATRC). The effort also includes dataset expansion using DoD-supplied video and real-world data collection from special operations medics and EMS partners, supported by synthetic augmentation pipelines. The system's outputs are designed to support broader human-machine teaming applications and future integration into tactical documentation and autonomy platforms.With dual-use applications in both military medicine and civilian EMS, this effort will establish a scalable foundation for AI-driven battlefield perception, triage automation, and autonomous medical documentation under real-world conditions.