Probabilistic Assessment of Decentralized Decision-making for Interdependent Network Restoration

Abstract

This study introduces a statistical model that guides decentralized infrastructure restoration processes aligned with field practices. In particular, we make more analytically tractable the previously proposed Judgment Call method to simulate real-world decisions under time and resource constraints. The Judgment Call method explicitly models the largely ignored feature of decentralization in the restoration planning across interdependent networks. The method solves the Decentralized Interdependent Network Design Problem (D-INDP) while acknowledging the lack of proper communication among decision making agents, and hence, the lack of essential information. Here, we use a Bayesian Hierarchical Model (BHM) to simulate the agents practical use of their field expertise and judgments to compensate for essential information shortage. We train the model using synthetic restoration plans that emphasize the local preferences of the agents. The method is applied to the interdependent infrastructure network of Shelby County, TN, and the results show that the performance of BHM-aided restoration plans is close to the conceptual upper bound.