Collaborative Aided Target Recognition Using Next-Generation Foundation Models for Multi-Domain Unmanned Systems

This Department of Defense (DoD) SBIR solicitation (Topic OSW26TZ04-NV001) seeks to develop and demonstrate a collaborative Aided Target Recognition (AiTR) capability for heterogeneous, multi-domain unmanned systems (UAS, UGV, USV). The initiative aims to leverage next-generation pre-trained foundation models—including Vision-Language Models (VLMs), Vision-Language-Action models (VLAs), and modern State-Space Models (e.g., Mamba)—to achieve robust, predictive target disambiguation and multi-modal sensor fusion in contested, cluttered, and PNT-denied operational environments. The project has multi-service relevance across the Army, Navy, USMC, and Air Force. Phase I focuses on conceptual design, architecture selection, and demonstrating technical feasibility using simulated or limited real-world datasets. Phase II requires developing, integrating, and validating a fully functional software prototype (TRL 6) across at least three heterogeneous unmanned platforms, demonstrating performance against specific key performance parameters (KPPs) such as high classification accuracy, low latency, and low communication bandwidth. Phase III targets dual-use military and commercial applications, including collaborative combat aircraft, search-and-rescue, and autonomous vehicle fleets. The technology is restricted under ITAR/EAR export control laws, and the projected Cybersecurity Maturity Model Certification (CMMC) requirement is Level 2 (Self).