AASP: An HPC-centric Interactive Mission Planning Framework
Recent advancements in Artificial Intelligence (AI) have demonstrated promising results in game theory. However, in critical real-world warfare domains, human planners are still necessary to ensure strategies consistent with national security and geopolitical objectives. The Maui High Performance Computing Center’s (MHPCC) AI-Assisted Strike Planning (AASP) project worked in close coordination with mission planners to develop an interactive platform that enabled mission level planners to collaborate with AI algorithms within an HPC environment. Missions that reflected potential real-world scenarios were first defined within the AFSIM modeling framework. Sophisticated algorithms, well suited for HPC environments, then rapidly searched massive state-action spaces in parallel and identified actions leading to the highest probability of mission success. This information was then presented to planners via an intuitive and interactive web-based user interface, where objectives and constraints could subsequently be adjusted to reflect relevant changes on the battlefield. The model could then be rapidly retrained with new information and iteratively assessed within the predefined AFSIM environment. Initial results demonstrated a promising HPC-centric framework for mission level planners to develop and assess adjustable strategies using an interactive user interface.
PRESENTER
Trevino, Rob
rtrevino@aicollaborations.com
480-628-7347
MHPCC
CO-AUTHOR
Roe, Kevin
kevin.roe.1.ctr@us.af.mil
Webber, Fred
frederick.webber@infinitetactics.com
Harris, Jack
jack@infinitetactics.com
CATEGORY
Artificial Intelligence / Machine Learning usage for HPC Applications
SYSTEMS USED
Reef, MHPCC Portal
SECRET
No