A Framework for Hybrid Agent-Network Analysis for Mission Assurance

Modern complex systems, including power grids, aerospace infrastructures, and other mission-critical domains, require robust approaches to ensure mission assurance: the ability of a system to maintain operational functionality under adverse conditions. Traditional methods, which rely on static reliability assessments and abrupt failure analysis, fall short when real-world systems experience gradual, dynamic degradation and cascading failures. To address these challenges, we present a hybrid framework that integrates graph-based dependency modeling, a genetic algorithm (GA) for network optimization enhanced by reinforcement learning (RL), and agent-based modeling (ABM) for degradation and catastrophic failure analysis. The framework begins with a dependency network model that captures both the functional and structural interdependencies within a system. This model is then iteratively refined through network optimization using a GA, with RL dynamically adjusting the optimization parameters prior to execution to improve performance. Once the network is optimized, resilience and robustness are evaluated using a rule-based ABM module. This module simulates real-world degradation scenarios, ranging from gradual performance declines to sudden component failures, providing a comprehensive assessment of the system's ability to maintain and recover functionality under stress. By combining these methodologies, the proposed framework offers mission designers, engineers, and system analysts a powerful and integrated tool for enhancing mission assurance in complex systems. This is demonstrated through an electrical grid case study, which showcases the framework's ability to identify network vulnerabilities and inform effective mitigation strategies.