Despite multi-stage decision problems being common in production planning, there is a class of such problems for which a general solution framework does not exist, namely problems with endogenous uncertainty. Methods from decision analysis and stochastic programming can be used, but both require significantly constraining assumptions. In order to overcome the current challenges, Decision Programming combines approaches from these two fields, making it possible to acquire optimal strategies for different decision problems.
Decision Programming is strictly limited to problems in which uncertainty events and decisions are taken from a finite discrete set, reducing its applicability to problems with continuous decision spaces. Discretizing a continuous decision space increases the problem size and can lead to computational intractability.
This thesis presents a problem decomposition approach extending the Decision Programming framework. The decomposition approach allows for considering continuous decision and uncertainty spaces in problems with a suitable structure. The proposed framework is applied to three different problems, including a large-scale production planning problem from literature. The main example in this thesis is a novel approach on climate change mitigation cost-benefit analysis, where R&D is carried out simultaneously with the emissions abatement decisions. The R&D projects provide information on the climate damage severity and decrease the price of abatement. Problems with similar structure have not been discussed in the literature, and the extended Decision Programming framework is able to solve the problem to optimality.