ONR is seeking white papers and proposals in support of advancing artificial intelligence for future naval applications. Work under this program will consist of basic and applied research with the overall S&T effort conducted at TRLs 1 – 5.
This topic call-out falls under Long range BAA N00014-23-S-B0001 and encompasses the following ONRG technical areas:
- machine learning, reasoning and intelligence
- computational methods for decision making – automated image understanding
- command decision making
- cognitive science for human machine teaming
- tactical AI for Marine Corps
Topic 1 – Human-Inspired Computational Models of Vision-Language Interactions for Agents – develop a principled computational framework and architecture for vision-language interaction, informed by human performance, that is open-domain and capable of strong compositional generalization. Interactions that enable agents to learn and reason about the real world with high levels of complexity in a transparent manner, result in multimodal dialogue for human-agent collaboration performing challenging tasks, and more.
Topic 2 – Mission-focused AI (AI fundamental & applied research) – investigate and develop techniques to support mission planning and execution activities that are dynamic, uncertain, and require coordination across different areas. Solutions are sought that will provide applications or foundational knowledge that enable the generation and evaluation of Courses of Actions (COA), transfer of learning across mission areas, countering mission focused AI, and interactive machine learning applications. Researchers may use game datasets such as from StarCraft or Supreme Commander: Forged Alliance Forever or unclassified datasets generated from mission simulators to support research.
Topic 3 – Collaborative AI – create collaborative agents capable of working with humans towards common goals and support data intensive tasks under resource and time constraints in real-world settings. While using simulated, non-military applications is acceptable and expected, approaches must be capable of shifting from simulation environments to real world settings with humans, using real world data, dealing with real world complexities, and supporting goal-directed task feedback loops. The resulting agents must be incentivized to optimize on satisfying task objectives in data-driven environments with minimal data.
– while papers due 7 November 2022; full proposals due 4 January 2023