Optimal capacity allocation under local load redistribution

Determine an optimal allocation of a fixed total capacity budget across the nodes of a given network and initial-load configuration under local load redistribution, maximizing robustness against overload-driven cascading failures.

Background

The paper studies overload-driven cascading failures in which a failed node redistributes its load to active neighboring nodes, potentially causing further failures. Node capacities are decision variables subject to a fixed total free-space budget, and robustness is measured by the aggregate survivor fraction over initial failure sets of different sizes.

Unlike global-redistribution models, which admit closed-form characterizations and optimality guarantees because failed load is shared across all surviving nodes, local redistribution couples cascade outcomes to network topology, initial loads, capacities, and multi-round failure propagation. The paper introduces TANGCO as a simulation-based learning method that seeks high-quality allocations, but it does not establish an exact optimum; consequently, the optimal allocation for the local-redistribution problem remains unresolved.