Grant: Discovery Grants Program - Individual

We will develop integrated models and solution methods to improve resource planning and scheduling decisions in service and operations management when there are multiple uncertainties (random arrivals, stochastic service times, no-shows and random length of stay) in the system. The objective is to satisfy service targets (priority 2 jobs need to be processed in 3 weeks), reduce wait times, lower costs (overtime, idle time), achieve higher utilization of resources, improve throughput, obtain higher service levels, and reduce resource requirements in an integrated framework in which we optimize at the entire system level rather than optimizing only part of the system. There are many real-world applications for this setting where arrivals are stochastic with multiple job types (advance scheduling), service durations are random (appointment scheduling) and there could be other set of downstream resources with stochastic length of stay. One such example is surgery scheduling in which multiple types of patients arrive to a system, their treatment times are stochastic and then they recover in bed for a random amount of time. The questions are how much capacity to reserve for different type of priorities/types of patients, how much time to book for their appointments and how many beds to operate to achieve service/wait time targets, minimize operational costs and determine number of resources needed. We aim to improve the system with an integrated approach, i.e., optimize all decisions at once with the appropriate interactions. Any other way (disintegrated) such as a sequential approach (i.e., optimizing one decision, fixing it and then optimizing for another one), a local approach (i.e., optimizing for only one subset of the system) does not necessarily improve the overall system and it could make things worse. Furthermore, these problems (advance and combined) are highly dependent each other, i.e., output of one is the input of the other. This inter-dependency suggests that a combined scheduling approach may outperform the optimal solution of each problem solved separately or sequentially. Although an integrated approach is the ideal way, intractability, curse of dimensionality and use of different modeling approaches for both have forced researchers to use a sequential or a local approach. In the short run, we will build on our previous work to study combined advance and appointment scheduling for multi-priority and multi-class patients as well as home healthcare scheduling (which has similar and distinct features). We plan to use a dynamic programming and mathematical programming frameworks complemented with machine learning. We also will consider a few extensions for both problems. In the long term, we plan to study different objective functions (rather than expected cost) and multiple resources for combined scheduling, and synchronous job visits, time windows in job visits and service provider continuity for home healthcare scheduling.