Risk Assessment for Heterogeneous Multi-Agent Systems

Description

Heterogeneous Multi-Agent Systems (MAS) integrate unmanned air, ground, and maritime vehicles into fleets that able to execute complex missions beyond a single platform capabilities. Although this interconnectedness improves operational effectiveness, it also enables failures and attacks to propagate across system layers and affect mission outcomes. Existing approaches typically address safety, cybersecurity, or AI assurance separately and provide limited support for mission-level risk assessment in heterogeneous MAS. This paper proposes a layered risk assessment method that models heterogeneous MAS as interconnected Agent, Coordination, Communication, and Human layers and integrates scenario likelihood, impact, dependency structure, and propagation effects. The method produces risk estimates at the scenario, layer, asset, and mission levels. The mine detection in coastal areas use case demonstrates how localized compromises propagate through system dependencies and contribute to mission-level risk. The proposed method extends previous work on threat assessment by providing a mission-aware quantitative risk estimation for heterogeneous autonomous systems operating in dynamic and adversarial environments.

Additional Description

IEEE MILCOM 2026 (WS3) 3rd Workshop on Security, Resilient, and Robustness of Systems and Software

Period	12 Oct 2026
Event title	IEEE Military Communication Conference 2026
Event type	Conference
Location	Bethesda, United States, MarylandShow on map
Degree of Recognition	International