Replanning for Evasive Autonomy to Counter Threats

Objective: The objective of this topic is to develop applied research toward an innovative capability that generates and executes game-theoretically optimized Courses of Action (COAs) for both Blue (swarm) and Red (adversary) forces. The contractor will co-develop realistic operational scenarios with the Government, encode Commander’s Intent into a formal utility function, and iteratively refine the autonomy using simulation environments integrated with a Tactical Assault Kit or another system. The final capability will enable a 200-agent UAS swarm to operate in degraded environments—such as those with EW interference, kinetic threats, or weather events—while achieving mission objectives in alignment with AFSOC’s provided Adaptive Airborne Enterprise (A2E) vision for tiered ISR. Description: As a part of this feasibility study, the proposers shall address all viable system design options with respective specifications. This capability will integrate AI-enabled multi-agent planning and adversarial reasoning into a software system that supports resilient, decentralized UAS swarm coordination. The autonomy system must produce and execute game-theoretically optimized Courses of Action (COAs) for both Blue (UAS swarm) and Red (adversarial threat) forces, accounting for the interdependence of their strategies’., Blue’s optimal COA depends on Red’s, and vice versa. The Government seeks scalable autonomy software that produces AI-generated, game-theoretically optimized Courses of Action (COAs) for both Blue (UAS swarm) and Red (adversary threat) forces in contested environments. The capability must be developed using a government-approved modeling framework (e.g. Markov decision process application programming interface) and demonstrated in a simulation environment chosen by the contractor. The final solution will: Incorporate the provided Commander's Intent into a mission utility function. Support realistic operational scenarios aligned with AFSOC's Adaptive Airborne Enterprise (A2E) and the FANTOM Futures line of effort. Scale to control no fewer than 200 UAS simultaneously. Function in degraded conditions, including EW interference, kinetic threats, and adverse weather. Include decentralized decision-making, communication-aware planning, and peer-to-peer fallback behaviors for denied environments. Be suitable for integration into a maritime Find, Fix, Track, Target, Engage, assess (F2T2EA) demonstration >100 NM offshore against targets defended by notional integrated air defense systems. Keywords: Drone swarm; planning against an adversary; computational game theory; artificial intelligence CMMC Level: Level 2 (Self)