Time Consistency, Risk-Mitigation and Partially Observable Systems (2018–2023)
This project considers controlling partially observable systems that are dependent on a random background environment comprised of multiple resources evolving in parallel over time. The controller aims to harvest these resources optimally by attempting to maximize safety in a time consistent manner, while meeting throughput targets. That is, the probability of a catastrophe at any such resource is consistently minimized. A key feature of such systems is the trade-off between exploration and exploitation. The intended outcomes of the project are novel control policies and effective numerical techniques, which can be applied to fishery management, communication networks, power systems and social resource allocation scenarios.