Solicitation Opportunities

MxD seeks input from industry partners to support a project advancing digital twin and reinforcement learning capabilities for batch energetics manufacturing at a government research and development facility in the Northeast. Building on prior work at Picatinny Arsenal, this effort will extend an existing descriptive digital twin — supported by an on-premise sensor network and data historian in a secure, isolated environment — into a physics-informed, predictive twin capable of representing process state and correlating multi-source variation over time. A reinforcement learning model will be developed and trained to monitor batch state, recommend parameter adjustments, and adapt its policy based on historical performance, driving toward a repeatable “Golden Batch” outcome, with human operators retaining decision authority throughout.

The project aims to establish a repeatable implementation and transition framework that can be replicated across additional energetics processes and facilities, helping modernize legacy, isolated manufacturing environments more broadly. Key deliverables include a quantified definition of the target Golden Batch (covering key process parameters, quality metrics, and measurement methods) and validation of model performance against historical and available real batch data, supplemented by offline simulation with defined evaluation metrics. Together, these efforts are intended to lay the groundwork for safer, more efficient, and more consistent production of critical defense materials.