CAREER: Signaling through Motion: A Game-Theoretic Framework for Deception and Cooperation in Mobile Multi-Agent Systems

The goal of this NSF CAREER award is to establish a rigorous scientific understanding of how autonomous robots can communicate using only their physical motion, without relying on explicit verbal or wireless communication. As autonomous systems increasingly operate alongside humans and other machines, their movements are continuously observed and interpreted by others. Successful completion of the project will lead to a transformative improvement in the capability of robots to work alone or in teams by unifying game theory, machine learning, and robotics to enable characterization of when and how motion can influence the decisions of observers, whether they are cooperative partners or adversaries. This will be achieved by modeling interactions between moving robots and observers as strategic games in which some information, such as their intent or capabilities, is hidden. The intellectual merit of the project includes new theoretical tools for analyzing these interactions and integrating them with machine learning to extend these insights to larger and more complex multi-agent systems. The broader impacts of the project include improving the safety and security of autonomous systems in applications such as parcel delivery robots, transportation, and security operations; informing the design of infrastructure and policies that reduce vulnerabilities to adversarial behavior; and advancing engineering education by integrating game theory into hands-on robotics curricula, workshops, and open-source competition platforms that broaden participation in STEM. The research objective of this project is to characterize the feasibility, forms, and effectiveness of motion-based signaling in multi-agent systems. Analyzing interactions between mobile agents and an observing agent is a major challenge because the observer’s interpretation and response to motion are typically unknown. To address this challenge, these interactions will be modeled using incomplete-information games, including dynamic Bayesian games and partially observable stochastic games, which capture both strategic reasoning and uncertainty about other agents’ private information. New solution methods that leverage structural properties of games embedded in physical environments will be introduced, enabling more tractable analysis of signaling behaviors that arise through motion. To extend these methods to more complex multi-agent scenarios, machine learning will be integrated with analytical game-theoretic models to accelerate the discovery of coordinated strategies while maintaining interpretability and performance guarantees. The theory will be validated using simulation platforms and a modular robotic testbed supporting controlled experimentation and human-in-the-loop studies. By bridging formal analysis, learning, simulation, and physical experimentation, the project establishes a rigorous foundation for signaling-aware robot motion control in complex multi-agent systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2540913 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Daigo Shishika | Institution: George Mason University, FAIRFAX, VA | Award Amount: $549,205 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2540913 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2540913.html