In This Story
A multidisciplinary George Mason team is developing algorithms using artificial intelligence to improve how teams of autonomous robots operate in complex and uncertain environments. Their work, funded by a $1.7 million grant from the Army Research Lab, could benefit both military and civilian applications.
The team’s four researchers—Greg Stein and Xuesu Xiao from the Department of Computer Science, Xuan Wang from the Department of Electrical and Computer Engineering, and Daigo Shishika from the Department of Mechanical Engineering— are each using unique expertise to address a technical challenges at varying levels of abstraction, including decision-making under uncertainty, game-theoretic engagement with adversaries, and safe traversal of rugged terrains.
“The project is about how to gain an operational advantage, or at least mitigate operational disadvantages, when a team of robots are moving through territory that they do not themselves control," said Stein. While Shishika, Wang, and Xiao have been working on this topic since 2022, Stein, who focuses on machine learning and long-horizon decision-making, is a recent addition. Shishika applies game theory to team behavior in dynamic environments. Wang works on how to carry out the best coordinated team behavior, and Xiao focuses on preparing autonomous vehicles for off-road mobility and vertically challenging terrains.
Set to continue for four years, with an optional fifth year for integration, the team aims to develop a comprehensive framework of algorithms unifying their research, from high-level team coordination to low-level vehicle control. The project emphasizes the need for robust algorithms to handle incomplete information and dynamic environments.
While the primary focus of the project is on military applications, the research has broader implications for various non-military scenarios, such as search and rescue operations, said Shishika.
"The thing that we are really interested in is the decision-making and coordination component, to cooperate better and to make longer decisions over these extended periods of time while also behaving safely and mitigating risk," said Stein. "The framework that we're all collectively working towards is this thing that unifies all those ideas, to have a system that's capable of all of these novel capabilities together.”