Towards Robust Long-Term Infrastructure Asset Management under Deep Uncertainty

气候变化的威胁是真实存在的，但当前的基础设施适应计划令人不安。尽管加拿大的几个司法管辖区已经引入了基础设施资产管理条例，但目前的基础设施资产管理实践受到一些因素的阻碍，包括在深度不确定的情况下缺乏一致的长期规划框架和有效的优化工具。与此同时，人工智能的进步正在提供巨大的机会和潜力，可能会改变基础设施资产管理的未来实践。着眼于为可持续和有弹性的基础设施系统开发一种综合的资产管理和气候适应规划方法，拟议的未来五年研究计划希望解决与发展这种综合规划方法有关的三个基本问题。需要解决的三个问题包括：1)气候变化威胁下老化桥梁和公路网基于风险的稳健性评估；2)决策者对气候适应长期投资的风险感知和时间偏好；3)深度不确定性下面向稳健性的最优资产管理和气候适应规划。虽然第二个问题是一个重要的实证问题，将影响IAM的适当投资水平和社区参与，但另外两个问题是方法开发。对于稳健性评估，将使用申请者研究团队最近开发的基于风险的新稳健性指数。进一步的改进包括桥梁和公路网的经验损伤-损失关系；使用深度学习的高效代理建模方法；以及基于稳健性的识别关键部件的技术。将采用面对面调查的方法，调查资产管理决策者对气候适应的长期投资的风险感知和时间偏好。调查将在加拿大的主要资产管理会议上进行，公司资产经理和市政议员通常会参加这些会议。采用基于累积预期理论的风险态度模型和不同的贴现模型对实证数据进行拟合。最后一步将使用最适合的模型。对于面向稳健性的优化，将开发深度强化学习方法，以创建动态、灵活的长期投资计划，既适用于气候适应活动，也适用于常规维护和康复治疗。为了实现这些目标，研究计划被分解为六个相互关联的项目。拟议的计划将为专业资产管理公司提供更好的工具，用于资产管理和气候适应的综合长期规划。该提案还包括一个现实的HQP培训计划，该计划在公平、多样性和包容性的原则下交织在一起。

The threats of climate change are real, but the current infrastructure adaptation plans are unsettling. Although regulations on infrastructure asset management (IAM) have been introduced to several jurisdictions in Canada, current IAM practice has been impeded by a number of factors including the lack of a consistent long-term planning framework and effective optimization tools under deep uncertainty. Meanwhile, advancements in artificial intelligence are presenting huge opportunities and potentials that may change the future practice of infrastructure asset management. With a long-term objective aiming to develop an integrated asset management and climate adaptation planning methodology for sustainable and resilient infrastructure systems, the proposed research program for the next five years wants to address three fundamental issues related to the development of such an integrated planning methodology. The three issues to be addressed include: 1) risk-based robustness assessment of ageing bridges and road networks under climate change threats; 2) decision makers' risk perception and time preference of long-term investment on climate adaptations; and 3) Robustness-oriented optimal asset management and climate adaptation planning under deep uncertainty. While the second issue is an important empirical question that would affect the proper level of investment and community engagement in IAM, the other two issues are of method development. For the robustness assessment, a novel risk-based robustness index that was recently developed by the applicant's research team will be used. Further enhancements include empirical damage-loss relationships for bridges and road networks; efficient surrogate modelling method using deep learning; and robustness-based technique for identifying critical components. An in-person survey method will be used to probe asset management decision makers' risk perception and time preference of long-term investment on climate adaptations. Surveys will be conducted on major asset management conferences in Canada that corporate asset managers and municipal councillors usually attend. Cumulative prospective theory-based risk attitude model and different discounting models will be used to fit the empirical data. The best fitted models will be used for the last step. For the robustness-oriented optimization, deep reinforcement learning methods will be developed to create a dynamic, flexible, long-term investment plan for both climate adaptation activities and conventional maintenance and rehabilitation treatments. To achieve these objectives, the research program is broken down to six interrelated projects. The proposed program will provide professional asset managers with a better tool for integrated long-term planning of asset management and climate adaptation. The proposal also includes a realistic HQP training program that is interwoven under the principle of equity, diversity and inclusion.