Panopticon AI is a web-based software simulation platform compatible with Gymnasium , the API standard for reinforcement learning. Our project is open source under the Apache 2.0 license, with code available to military modeling and simulation researchers. We hope to provide an open platform for military modeling and simulation and AI research.

Applications

Realistic Simulations

We're building a realistic red and blue forces simulation at the unclassified level for AI researchers both with the Department of Defense and outside of it. Our backend engine allows researchers to train algorithms and plug in custom agents, while our web app allows users to edit and play back scenarios.

Wargaming

While our platform is still under construction, we hope to enable users to plug in agents that can control red and blue forces, allowing them to create and play through custom scenarios.

Getting started

Our source code is available here . Information for interested code contributors is also on our github. To understand the features we have available, click Play or read the docs.

Complete the OODA loop

Reinforcement Learning

Panopticon AI is inspired by the technologies behind AlphaStar, the AI program which powered the computer players in the real-time-strategy game Starcraft. Our system provides a platform for researchers to train and run reinforcement learning algorithms for wargaming simulations. By training agents, military researchers will be able to explore more powerful models and strategies for training AI in an ecosystem of researchers. We're hoping to foster the creation a network of people who can catalyze additional advancements in military AI.