CE39 - Sécurité globale, résilience et gestion de crise, cybersécurité

Robust and Scalable Prescriptive Analytics for the Resilience of Critical Infrastructure Networks – RoScaResilience

Submission summary

This project advances the body of knowledge on critical infrastructure network resilience by proposing robust and scalable prescriptive analytics to guide actions and decisions for effective resilience improvement. With this aim, we develop a novel interdisciplinary approach that integrates risk and resilience engineering, mathematical optimization, and (deep) reinforcement learning to improve 1) the ability to make robust planning of ex-ante strategic mitigation and preparedness actions under deep uncertainty of future disruptions caused by natural extremes, and 2) the ability to plan for effective ex-post response and recovery, emphasizing on the computational efficiency of the method when applied to large-scale real-world systems.
The research approach is composed of three scientific and one technical components. The first component develops a new optimization technique, the distributionally robust optimization, for ex-ante strategic mitigation and preparedness planning against natural hazards, which integrates all available but ambiguous information of uncertain disruptions via ambiguity sets. It can find robust ex-ante resilience strategies that perform relatively well across a wide range of potential disruptions. The second component proposes a computationally efficient decision-aid framework for optimal post-disruption emergency response and recovery planning. It exploits deep reinforcement learning and takes its advantage to overcome the curse of dimensionality faced by traditional methods. The application of the first two integrated components to electric power networks and railway systems constitutes the third component, offering two application perspectives on demonstrating the effectiveness of the developed techniques. Lastly, the technical component develops a GIS-based visualization platform for empowering the trust of human decision-makers in the proposed algorithmic approaches and promoting their applications in real-world system designs and operations.

Project coordination

Yiping Fang (CentraleSupélec)

The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.

Partner

DYOGENE Institut national de la recherche en informatique et automatique
LGI CentraleSupélec
University of Edinburgh Business School

Help of the ANR 271,200 euros
Beginning and duration of the scientific project: September 2022 - 48 Months

Useful links

Explorez notre base de projets financés

 

 

ANR makes available its datasets on funded projects, click here to find more.

Sign up for the latest news:
Subscribe to our newsletter