The paper proposes a new decision-making framework for partially observable systems in continuous time with discrete state and action spaces. The framework uses approximation methods for the filtering and control problems that scale well with an increasing number of states. The high-dimensional filtering distribution is approximated by projecting it onto a parametric family of distributions. This is integrated into a control heuristic based on the fully observable system to obtain a scalable policy. The effectiveness of the approach is demonstrated on several partially observed systems including queueing systems and chemical reaction networks.
Publication date: 5 Feb 2024
Project Page: N/A
Paper: https://arxiv.org/pdf/2402.01431
