With increased competition in the marketplace, many sectors of the American economy are acknowledging the need for improved resource planning. By way of example, note the growing acceptance of the Just-In-Time philosophy in many manufacturing firms, the move toward dynamic routing capabilities within the telecommunication industry and adaptive signal control in urban traffic planning. In each of these areas, resources (e.g., productive capabilities and network capacities) are acquired well in advance of their need, and utilization is improved by adaptive allocation as precise system requirements are identified. Since precise information regarding system operation is not available at the time of planning, the resource planning problem involves decision making under uncertainty. This project focusses on modeling and solution of planning problems for systems that allow adaptive resource allocation. These methodologies are based on stochastic decomposition. Computational experience with stochastic decomposition indicates that it is ideally suited to the solution of large scale problems involving optimal planning under uncertainty. It is particularly useful when the variability of system requirements can only be assessed via simulated scenarios. The project involves three phases, involving the mathematical development and verification of these methods, computer implementation of the resulting algorithms, and testing of these algorithms on models based on case-studies which will be developed in conjunction with industrial collaborators.
