Collaborative Aided Target Recognition Services for Multi-Domain Unmanned Systems

A government authority in Washington, DC is seeking information on collaborative aided target recognition services for multi-domain unmanned systems. The effort focuses on developing and demonstrating an Aided Target Recognition (AiTR) capability for heterogeneous unmanned platforms, including air, ground, and sea systems, using next-generation pre-trained foundation models. Relevant model approaches include Vision-Language Models (VLMs), Vision-Language-Action models (VLAs), and modern State-Space Models such as S4, S5, Mamba, and Mamba-2.

The requested capability is intended to support robust, accurate, and predictive multi-platform, multi-modal target recognition across at least three heterogeneous unmanned platforms. The work includes development, integration, and validation of a fully functional prototype, with software delivered through documented APIs. Expected deliverables also include a comprehensive technical report covering system architecture, training data provenance, and known limitations, as well as a transition plan addressing at least two of the three military services.

The development scope should address collaborative multi-platform target disambiguation, including foundation-model-based representations that enable cross-platform target correspondence under conditions such as appearance ambiguity, viewpoint changes, scale variation, and clutter. It should also address geometry-consistent multi-modal sensor fusion and culminate in a demonstration in a relevant operational environment at Technology Readiness Level 6.