Representation and Analysis of Multi-Stage Problems in R & D

The paper describes a method of representing and analysing the multi-stage, dynamic nature of many types of research and development projects. The method makes use of a number of well-known techniques, decision trees, mathematical programming and simulation in combination. Each project which could form part of a laboratory portfolio is represented in a modified decision tree format called a "project tree," and the project trees are then considered in one analysis based on heuristics. The approach is demonstrated using a small illustrative example, and an industrial case study is described, giving attention to computational problems of size.