Document Type
Dissertation
Degree
Doctor of Philosophy
Major
Business Administration
Date of Defense
12-16-2016
Graduate Advisor
Keith Womer, Ph.D.
Committee
Cipriano Santos, Ph.D.
Douglas Smith, Ph.D.
Abstract
This research addresses the problem of sequential decision making in the presence of uncertainty in the professional service industry. Specifically, it considers the problem of dynamically assigning resources to tasks in a stochastic environment with both the uncertainty of resource availability due to attrition, and the uncertainty of job availability due to unknown project bid outcome. This problem is motivated by the resource planning application at the Hewlett Packard (HP) Enterprises. The challenge is to provide resource planning support over a time horizon under the influence of internal resource attrition and demand uncertainty. To ensure demand is satisfied, the external contingent resources can be engaged to make up for internal resource attrition. The objective is to maximize profitability by identifying the optimal mix of internal and contingent resources and their assignments to project tasks under explicit uncertainty. While the sequential decision problems under uncertainty can often be modeled as a Markov decision process (MDP), the classical dynamic programming (DP) method using the Bellman’s equation suffers the well-known curses-of-dimensionality and only works for small size instances. To tackle the challenge of curses-of-dimensionality this research focuses on developing computationally tractable closed-loop Approximate Dynamic Programming (ADP) algorithms to obtain near-optimal solutions in reasonable computational time. Various approximation schemes are developed to approximate the cost-to-go function. A comprehensive computational experiment is conducted to investigate the performance and behavior of the ADP algorithm. The performance of ADP is also compared with that of a rolling horizon approach as a benchmark solution. Computational results show that the optimization model and algorithm developed in this thesis are able to offer solutions with higher profitability and utilization of internal resource for companies in the professional service industry.
Recommended Citation
Simon Solomon, Stanislaus, "Multi-Period Stochastic Resource Planning: Models, Algorithms and Applications" (2016). Dissertations. 29.
https://irl.umsl.edu/dissertation/29