Hamilton Institute Seminar

Wednesday, March 18, 2020 - 13:00 to 14:00
Hamilton Institute Seminar Room (317), 3rd Floor Eolas Building

Speaker: ​Dr Ali Forootani, Maynooth University Hamilton Institute

Title: "Stochastic Dynamic Programming for Resource Allocation: Theory and Application"

Abstract: The definition of suitable resource allocation strategies is a crucial business aspect for both manufacturing and service companies. Pricing management systems are fundamental components of their daily operations, since price is one of the most effective variables that managers can manipulate to encour-age or discourage resource demand over time. Resource allocation and price management systems can be actually integrated into comprehensive business processes. However, fnding a proper allocation policy of a fixed amount of resources (aiming at maximizing the profits incurred by such allocation oper-ations) is a notoriously difficult task to be solved. First of all, such systems typically exhibit stochasticity, i.e. resource requests to be processed may arrive randomly according to some stochastic process. Second, if modeled by traditional methods, they suffer from extremely large state and action spaces, making the task of finding an optimal solution infeasible.
In this talk, we introduce a novel model and novel solution of the resource allocation problems via Stochastic Dynamic Programming based framework, which is customizable for different real business contexts. We address both instant (i.e. the customer requires a resource to be allocated immediately) and advance (i.e. the customer books a resource for future use) reservation requests. The above-mentioned scalability issues are addressed by approx-imating the value function via a more compact parametric representation.  In this talk, the term value function can be regarded as the expected to-tal revenue computed over the infinite time horizon for resource allocation problems, when applying a specific pricing policy.
In brief, we will discuss about novel modeling as well as novel theoreti-cal approaches suitable for various areas such as Markov Decision Processes, Control of Queuing Systems, Approximate Dynamic Programming, and Pos-itive Systems. We developed a general MATLAB Toolbox to implement each part of the modeling and simulations.