Parallel Funding of R&D Tasks with Probabilistic Outcomes

This paper addresses the problem commonly faced by R&D managers of funding redundant R&D tasks across several stages or components of a project. In the proposed methodology it is assumed that task outcomes are random, but that their distribution can be determined from engineering inputs. In order to account for individual preferences a utility function is used to measure overall project results, and hence guide the decisions at each stage. The problem is formulated as a probabilistic network and solved by means of a heuristic comprising simulation and dynamic programming. Although computational experience is limited, the results obtained from a number of examples associated with the development of a photovoltaic solar module were quite promising; that is, the heuristic always found the optimal investment plan. In general, computational times were minimal when compared to data collection efforts, with the latter perhaps being the principal hurdle to full implementation. The sponsoring organization, however, felt that the unified approach offered by the methodology represented a major improvement over current analytic techniques.