Personalized Large-Scale Queuing Models

University Of Florida, Gainesville FL

Investigators

Abstract

The research objective of this award is to develop understanding (analytical results, numerical algorithms) of the extent information on individual customers/servers can contribute to better system performance. Standard queuing models are based on averages - that is, personalized information is lacking from classical protocols which are oblivious to when exactly will the next arrival happen, or who is the least patient among the customers waiting to be served, or who is the fastest server among those available to serve. However, in many applications, the system operator might have or might acquire information on individual customers/servers. The focus of this project is on a practically relevant case of partial information, where knowledge about individual realizations is noisy. The main paradigm is thus the tradeoff between information availability and performance. The research approach is based on an asymptotic framework that allows one to analyze tradeoffs in many-server systems. If successful, the results of this research provide an opportunity to improve efficiency of various service systems from telephone call centers to complex outpatient clinics. Indeed, existing IT systems can be used to obtain information about individual customers/servers, and incorporate that information into decision-making. While the "personalized" medicine paradigm aims to tailor medical decisions to the individual patient, the results of this research will aid in tailoring operational decisions to specific patients by use of patient-specific information. This will lead to better patient satisfaction, more efficient utilization of resources, and lower cost of medical care. Graduate and undergraduate students will benefit from this award through lectures and involvement in research.