An applied complex systems framework for risk-based decision-making in infrastructure engineering

• Infrastructure is viewed as interacting natural, physical and social systems.
• Distributed risk-based decision-making is applied to socio-technical systems.
• Decision problems are analyzed across component, network and agent levels.
• Modules of network modeling, assessment and intervention comprise a control strategy.
• Illustrative examples show decision support capabilities and computational efficiency.

Infrastructure networks are complex, distributed, socio-technical systems whose study requires the integration of several domains of knowledge and techniques to coherently capture key features of their constituent components (e.g., fragility), the dynamics of such networked components, and their behavior as affected by human use and operation. This paper introduces an integrative, applicable framework of Complex Distributed Agent Networks (CoDAN) for infrastructure risk assessment and management. CoDAN detects sub-systems at multiple geographical scales and associates decision units (agents) to them in the context of a decision problem. Agents use advanced optimization and risk assessment methods for decision support, and allow to model decentralized yet coordinated decisions about performance and risk management (e.g., maintenance or restoration), mimicking behavior and system response observed in practice. The application of CoDAN to illustrative examples shows the potential of the framework to unravel the complex evolution of infrastructure network performance under hazard exposure, and to reduce the computational burden of tasks associated to risk assessment and management, as compared to traditional centralized formulations.