DARPA posts Gamebreaker AIE
On February 3, the Defense Advanced Research Projects Agency (DARPA) posted a new Artificial Intelligence Exploration (AIE) Opportunity. Responses are due by 4:00 p.m. Eastern on March 3.
DARPA is issuing a new AIE opportunity, entitled Gamebreaker, which invites submissions of innovative basic or applied research concepts in the technical domain of automating game balance to explore new capabilities/tactics/rule modifications that are most destabilizing to the game or simulation.
This Gamebreaker AIE Opportunity is being issued under the DARPA Program Announcement for AIE, DARPA-PA-19-03, which details the AIE Program’s overall intent and provides evaluation and submission instructions in addition to those outlined in this solicitation. To view the original DARPA Program Announcement for AIE, visit beta.SAM.gov (formerly FedBizOpps) under solicitation number DARPA-PA-19-03.
All proposals in response to the technical area(s) described herein will be submitted to the Gamebreaker AIE Opportunity, solicitation number DARPA-PA-19-03-05, and if selected, will result in an award of an Other Transaction (OT) for Prototype Projects, not to exceed $1,000,000. This total award value includes Government funding and performer cost share, if required, or if proposed.
The Gamebreaker AIE Opportunity seeks to develop and apply Artificial Intelligence (AI) to existing open-world video games to quantitatively assess game balance, identify parameters that significantly contribute to balance, and explore new capabilities/tactics/rule modifications that are most destabilizing to the game.
For Gamebreaker, game balance is defined as an inherent property of the game that reflects the win/loss ratio of players of equal skill level based on strategies and tactics employed within the game. For example, in a balanced game, if the skill level of both players is equal, each player will win approximately 50 percent of the total games played. Similarly, if the skill level of both players is equal but one player wins disproportionately due to inherent advantages arising from a condition of the game construct, the game is unbalanced.
The commercial gaming industry has a long-standing interest in maintaining game balance as balanced games are typically more entertaining, and market pressures help drive their development. Moreover, the contemporary method for assessing and balancing games is a trialand-error approach, thus representing an opportunity for the application of AI.
Normally, game developers release and observe an initial configuration of the game in largescale play. Then, developers gather high-level win/loss statistics while players provide feedback about elements of the game that are overpowered or imbalanced. Finally, updates to the game are made in which elements are buffed (performance increases) or nerfed (performance decreases) to achieve game balance. To date, little quantitative modeling of game balance exists, and research on the application of AI algorithms to automating game balance assessment (formally referred to as quantitative balance analysis) is extremely limited.
Department of Defense (DoD) applications of automating quantitative balance assessment are plentiful and range from identifying and mitigating adversary capabilities to generating methods for diagnostically assessing the impact of new defense technologies or changes in force design/posture. In future conflicts, DoD investment is designed to maximize imbalance to create an advantage or to seek equilibrium when an adversary is seeking an advantage. New AI algorithms inspired by Gamebreaker could help develop winning warfighting strategies when the adversary’s objectives – i.e. the “rules of the game” – are not clearly known. By exploiting game balance, Gamebreaker addresses an existing gap in AI and data analytics research as applied to current wargaming and simulation.
Complex, multi-domain modeling and simulation (M&S) environments currently under development by several DARPA programs aim to create a useful “Mosaic” model within which to experiment on new warfighting constructs using distributed, adaptive, all-domain force composition, tactics, and strategies – yet these models do not currently exist. It is reasonable to assume, however, that once these simulation environments reach maturity, an “AlphaMosaic” equivalent will be capable of searching for optimal strategies and tactics in the same way AlphaGo and AlphaStar agents have already proven effective in exploiting their respective game environments. Yet this does not take into account modifications to the construct of the game itself. While AI techniques have demonstrated the ability to master games and models of increasing complexity, there has been almost no research in the application of AI to game modification.
Future Mosaic models and wargames will likely include many of the same attributes of current, complex, real-time strategy (RTS) video games. As a result, contemporary RTS video games serve as surrogate engagement models, since they simulate force-on-force engagements with heterogeneous Red/Blue platform and weapon systems, while accounting for geography and imperfect information.
Full information is available here.