The need for planning is a fundamental task required to support large-scale and complex operations. This challenge is in support of the AI for Command AFVentures Technology Focus Areas. Toward this goal, this challenge specifically seeks to engage industry to take the first steps and develop AI agents. The challenge is built on the commercial game Command: Modern Operations. Through a series of scenarios this challenge hopes to excite and foster new innovations in the world of AI for planning.
The Competition: Agents will be submitted to the competition organizers. Your team will submit your model, a readme of how to run the agent, and a short paper documenting your approach.
The agents, selected technical approach, and supporting documentation will be evaluated based on:
- Competition performance against the 4 training and 1 test scenario to include but not limited to the ability of the agent to complete the scenario objectives, completion time, and remaining forces at the end of the scenario;
- Novelty of the technical approach; and
- Generalization – how well the approach can generalize to varying scenarios and complexity.
The Grand Prize for this challenge is $100,000. Finalists will be eligible for a Memorandum of Understanding in support of their proposals for a $750,000 Small Business Innovation Research (SBIR) award.
AI is being used more often to achieve mission success. However, it currently fails to provide benefits in two areas: wargaming and large scale mission planning. The majority of wargames are done manually and are time consuming to complete as adjudication takes hours per turn.
This competition serves as a test on how Command: Modern Operations can be used as a platform for in depth digital wargames with automated adjudication. Additionally, mission planning, notably at the theater level, requires a large amount of data, manpower, and time to create. AI can vastly improve the efficiency of the process by providing a mechanism to process data quickly and simulating battle outcomes. Currently, AI has only been used to create plans in closed systems that are not representative of real-world scenarios. This competition presents an initial foray into AIs that can create mission plans in realistic scenarios.
The Command Platform
The competition will use the Command: Modern Operations game engine for both development and evaluation of AI submissions. To participate each team needs access to a Command Modern Operations Professional Editions license or higher. A small number of licenses will be made available for use by teams with promising proposed approaches and need.
To request a license please do so on the registration form and answer the following questions:
- Please describe the approach you/your team would take to developing a competitive agent for this event.
- What do you consider to be the biggest challenges of this competition?
- How would your approach address those challenges?
Please answer these prompts and submit registration by May 22 to be considered for one of these licenses.
Register today to develop an AI agent! This Challenge is open to all innovators, academics, researchers, start-ups, univeristy teams, etc.
Agents can interact with the game via the Command Lua API. The agents will be tested across five scenarios of varying complexity. Four of these scenarios will be provided prior to the competition for agent training and testing. The final scenario will be similar to previous scenarios, but act as a test case to evaluate the robustness of submitted agents.
April 01 Challenge opens
May 10 Ask Me Anything Session for Interested Participants
May 22 - Registration Deadline: Command License Request (Request must be made on registration form)
June 14 Registration Deadline
August 29 Submission Package due by 11:59 PM (PDT) - Instructions will be sent to registered participants
September 10 Competition winners announced (date subject to change)
First Place: $100,000
Finalists will be eligible for a Memorandum of Understanding in support of their proposals for a $750,000 Small Business Innovation Research (SBIR) award.
AFVentures Technology Focus Area Description
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.
Participants 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 also desire methods that can provide insight into what factors influence decision making, with a long-term goal of decision support for a human player.
Ask Me Anything Session - May 10, 2021
- AFWERX SBIR Open Topic
- Command Modern Operations
- Command’s Manual
- Command’s Lua API
- Lua Language Reference