AFRL-AFOSR-VA-TR-2017-0028 Dynamic Decision Making under Uncertainty and Partial Information

The researchers made significant progress in all of the proposed research areas. The first major task in the proposal involved duality in stochastic control and optimal stopping. In support of this task, the researchers developed new methods for efficiently solving optimal stopping problems of partially observable Markov processes and optimal stopping problems under jump-diffusion processes. The researchers also studied the information relaxation approach and established duality for controlled Markov diffusions and weakly coupled dynamic programs. In the second major task aiming at solving difficult global optimization problems, the researchers proposed and developed a new framework that integrates the idea of model-based randomized optimization with gradientbased optimization, and further extended this method to simulation optimization problems. All the developed methods have been tested through numerical experiments and demonstrated excellent performance. The methods have also been applied to problems in revenue management, option pricing, and power allocation in communication networks. A-1 DISTRIBUTION A: Distribution approved for public release.