Adversarial Robustness of Autonomous Multi-Agent System

Adversarial Robustness of Autonomous Multi-Agent System

Status:

The project started in June 2026 and is still ongoing.

Researchers:

Prof. Srdjan Capkun

Prof. Aanjhan Ranganathan

Description:

Autonomous multi-agent systems, such as coordinated groups of unmanned aerial vehicles, increasingly rely on decentralized algorithms to achieve collective behaviors without central control. These systems depend on two essential inputs: each agent’s ability to determine its own position, typically through satellite-based navigation, and reliable communication between agents. Both inputs can be disrupted or manipulated, raising open questions about how such systems behave when the information they rely on becomes unreliable.

This project investigates how different classes of coordination algorithms respond to disruptions in localization and communication. Using a simulation environment, we systematically characterize how various forms of interference affect collective behavior, with the aim of understanding which information flows are most critical and why certain algorithms degrade under adverse conditions. The resulting analysis is intended to inform the design of more resilient multi-agent systems in future work.