Generalized Network Implementations

Abstract : Generalized networks are an important class of optimization models, with uses in a wide variety of fields. This report describes the development and implementation of a generalized network algorithm. Jarvis et. al. recommend a generalized network model, system for closure optimization and planning (SCOPE), for crisis action deployment planning. In SCOPE, large generalized networks must be repeatedly solved. These networks have special structure, which results in computational advantages. In this report, a generalized network implementation is developed for solving very large generalized networks. This implementation includes new data structures for storing the basis, in-core/out-of-core handling of the arcs, and special handling of pure network structure. In this report, a detailed examination of the SCOPE model is provided and its effect on implementation issues is discussed. The SCOPE model is highly structured. This report demonstrates how this structure can be used to advantage. In a companion report extensive testing is presented which addresses the question 'what affects the computation time for a SCOPE model?' Keywords: Linear programming; Data processing; Data storage systems.