R&D project selection and scheduling with a filtered beam search approach

Abstract

Models developed for selecting an optimal portfolio of R&D projects from among a pool of available projects do not generally include project scheduling as part of the selection criteria. Typically after a portfolio of projects is selected according to various selection criteria, they are subsequently scheduled. If it is not possible to schedule the selected projects through various research facilities and/or stages given the available resources, projects may be replaced with others requiring less time, or resources may be increased, which can result in a suboptimal portfolio. In this paper, project scheduling is included in the selection process, with a heuristic filtered beam search approach. Projects are selected on the basis of traditional selection criteria such as expected profits, as well as the makespan of the portfolio. The heuristic search procedure is demonstrated by an in-depth example and computational experimentation.