Summary
This DoD SBIR/STTR topic solicits technologies for Collaborative Aided Target Recognition using next-generation foundation models to enable multi-domain unmanned systems to detect, identify, and disambiguate targets under appearance ambiguity, viewpoint/scale variation, sensor heterogeneity, and degraded PNT. It seeks model-based, technology-agnostic approaches built around modern foundation models (vision-language, vision-language-action, state-space models) and emphasizes cybersecurity, supply-chain provenance of pre-training data, adversarial robustness, and explainability. The topic calls out tri-service relevance (Army, Navy/USMC, Air Force/Space Force) and requests solutions suitable for collaborative, multi-platform military Aided Target Recognition rather than legacy CNN-only systems.