Planning and Autonomous Control | Artificial Intelligence for Theater Multi-Domain Operations
What is the objective?
To develop a machine learning approach that can competitively operate in a strategic and tactical conflict in a real-time, adversarial simulation. This includes explicit, symbolic representation for the selected strategy and actions that will provide human observers with insight into what factors influenced decisions.
What problem are we trying to solve?
Recent advances in Artificial Intelligence have shattered expectations on how effectively machines can learn and plan within competitive game environments. These impressive demonstrations have relied on thousands of hours of example expert play combined with hundreds of years of computational effort. In order to realize these benefits for complex military operations, we will need new learning methods that can operate with limited or no training data, with time and cost constraints on available compute. We challenge participants to develop competent AI players for theater scale, multi-domain operations using a commercial gaming platform. Success will pave the way for a new generation of AI at the speed and scale of military operations, unleashing new opportunities for analytics, wargaming, and warfighter training.
What outcome do we hope to achieve?
Demonstrate the feasibility of training and executing AI in complex action and state spaces, motivating additional work to generalize capability across multiple scenarios and operating conditions.
What resources could the lab provide?
The lab could provide the commercial game application, subject-matter expertise and evaluation criteria by which to determine the effectiveness of the proposed learning approach.
What would success look like?
Participant algorithms will be evaluated against human-designed opponents and scored based speed and effectiveness over one or more scenarios of interest. While not directly scored, we desire methods that can provide insight into what factors influence decision making, with a long-term goal of decision support for a human player.
What types of solutions would we expect?
Solutions will include AI training methodologies, trained models for the scenarios of interest, and concept-based descriptions of strategies being employed by the AI.
What's in it for industry?
There will be a cooperative research agreement where both the industry/academic team and the government will benefit from the solution. The technologies developed have large applicability to both civilian and military applications. This effort can be leveraged as part of a future greater submission to an Air Force program or DoD effort.
The Request for Partnership Submission Period Has Now Ended.