Balancing the costs and benefits of resilience-based decision making
本文提出一个基于连续时间马尔可夫决策过程的随机动态优化模型，用于平衡社会物理系统在中断后的干预成本与恢复时间，通过哥伦比亚运河决堤案例验证，最优策略可减少约40%恢复时间和10%投资。
Most decision models of system resilience use static, deterministic optimization techniques while focusing on resilience assessment. At present, we lack appropriate decision support methodologies and computational tools that can offer dynamic control of resilience and balance the costs of resilience assurance. This paper presents a stochastic dynamic optimization model, based on an infinite horizon Continuous-Time Markov Decision Process , to balance the intervention costs and reduce the total recovery time ensuing a disruption of a social-physical system. We aim to offer a model that can facilitate its application to different disruption scenarios. Our state-space formulation of the recovery process uses discrete performance intervals, whereby actions and resulting rewards/costs are related to investment resources, which govern state transitions. We illustrate the model via a case study based on the 2010 Northern Colombia Dique Canal breach. Our results show that the optimal policy reduced the recovery time and restoration investment by approximately 40% and 10%, respectively, when compared to the efficiency of the government interventions. The proposed model features dynamic control of recovery resources and considers the costs of resilience assurance. The model can inform policymakers of ways to improve system resilience using balanced disruption recovery strategies.