The Defense Advanced Research Project Agency’s (DARPA) Strategic Technology Office (STO) recently awarded Northrop Grumman [NOC] a contract under the agency’s Gamebreaker program, the company said on Aug. 12.
Gamebreaker aims to use artificial intelligence (AI) to develop future warfighting strategies that gain U.S. advantage during a military deadlock or regain balance when an adversary seeks the advantage.
Susan Wilson, the director of Northrop Grumman intelligent mission capabilities and advanced technology laboratory, said in a statement that “using AI to exploit engagement models can help to enable intelligent systems that could in turn enhance military strategy.”
“We are exploring how we may be able to use this methodology in the future,” she said.
Northrop Grumman’s Gamebreaker team includes Hazardous Software and Slitherine Software’s Matrix Games.
“Working closely together, this partnership will use advanced AI techniques to model and break balance within a highly complex simulator environment called ‘Command: Modern Operations,'” according to Northrop Grumman.
Christopher Hazard, the CEO of Hazardous Software, said that the partnership will leverage his company’s “13 years of history modeling dynamic adversarial scenarios” and “the Diveplane machine learning platform.”
“Command Professional Edition is the only wargame being analyzed by DARPA in the Gamebreaker program,” Iain McNeil, CEO of Slitherine Software, said in the Northrop Grumman statement. “We are very interested to see how the AI behaves and if it manages to identify loopholes that need assessment, or it comes up with innovative strategies that are applicable to the real world.”
DARPA held a virtual kickoff meeting for the Gamebreaker program on May 5 and chose 9 teams–each employing two video games–for the project:
- Northrop Grumman with Hazard Software and Matrix Games using Command: Modern Operations and TORCS;
- Boeing‘s [BA] Aurora Flight Sciences with the Massachusetts Institute of Technology using StarCraft II and Google Research Football;
- Lockheed Martin [LMT] with Cycorp Inc. using Multi-agent Particle Environment and SpringRTS: 1944;
- BAE Systems with the University of California Santa Barbara and AIMdyn, Inc. using StarCraft II and the AFRL Strategem Wargame;
- Blue Wave AI Labs using SpringRTS: 1944 and OpenRA;
- EpiSci using miniRTS and StarCraft II;
- Heron Systems Inc. using DeepRTS and StarCraft II;
- Purdue University using microRTS and StarCraft II; and
- Radiance Technologies with BreakAway Games using FreeCiv and Zero-K.
Each of the 9 teams received $1 million contracts for 18 months. Phase 1 feasibility demonstrations for Gamebreaker will occur in January, while Phase 2 final demonstrations are to be in August and September 2021 next year, per DARPA.
“If we can figure out a generic method to assess and then manipulate balance in commercial video games, my hope is that we might then apply those AI algorithms to create imbalance in DoD simulated war games used to train warfighters for real-world battle,” Air Force Col. Dan “Animal” Javorsek, the Gamebreaker program manager in DARPA’s Strategic Technology Office, said in DARPA’s Gamebreaker kickoff announcement in May.
“The commercial gaming industry has a long-standing interest in maintaining game balance since balanced games are more entertaining, and market pressures drive their development,” Javorsek has written. “Instead, the military has the opposite goal whereby strategists deliberately explore technologies that maximize imbalance.”
“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,” according to Javorsek. “By exploiting game balance, Gamebreaker addresses an existing gap in AI and data analytics research as applied to current wargaming and simulation.”