Auctions for resource allocation and decentralized restoration of interdependent networks

This paper prescribes an auction-based methodology to efficiently allocate resources for the restoration of disrupted interdependent networks in a decentralized fashion. Auctions entail no communication among decentralized decision-makers implying lack of coordination during the decision-making process. We build upon the Judgment Call (JC) method, which models the distributed environment in which restoration decisions are made, and whose main challenge is to cope with resource allocations. We focus on static, private-value auctions with truthful bidders, and explore several well-known multi-unit auction mechanisms to address the resource allocation challenge. Results show that the combinatorial auction yields the most efficient allocations in terms of the deviation from a centralized optimal allocation, called gap, and associated restoration plan for enhanced resilience, called relative performance. Our sensitivity analysis reveals that allocation gap and relative performance are highly influenced by the number of layers and the initial disruption level. Also, the computational time of JC with auction-based resource allocation is generally lower than its centralized counterpart. Two applications showcase our methodology: A database of synthetic, ideal networks, and the realistic interdependent infrastructure network of Shelby County, TN. The latter application shows that our method enhances community resilience to natural hazards in terms of economic loss and serviceability, while being consistent with distributed decisions during contingencies.