Adaptive control of stochastic queueing networks (2013-2015)

Stochastic queueing network models are well suited for analysis, design and control of manufacturing, road traffic, customer service, and computer systems. Many variations are useful in both theory and practice, yet a key drawback is the fact that current theory and control methodology assumes parameters are pre-specified and known. This project will devise and analyse adaptive control and parameter estimation methods for queueing networks yielding stable, tractable and sensible policies that can be implemented in view of parameter uncertainty. The approach is based on approximate decomposition of queueing networks utilising structural results of departure processes as well as stability analysis of systems in time-varying environments.