OPTIMAL SELECTION OF RESEARCH-AND-DEVELOPMENT PROJECTS

Abstract

A model for the selection process of research and development (R&D) projects belonging to some common area and submitted to a funding agency is presented. Every project is evaluated, the result of this procedure providing the agency with some useful information about the distribution of the project's payoff. The uncertainty due to the unknown perspectives of the whole area of R&D is incorporated in a Bayesian way, so that the agency learns about the `'area value'' from the projects already handled. After specifying the model assumptions, adaptive dynamic programming techniques are applied to develop an optimal funding strategy for a given number of submitted projects. Some qualitative properties of the optimal strategy are derived, and the asymptotical behavior of the maximum expected reward is determined.