Dynamic Flexible Queueing Network Models for the Design and Control of High Performance Operational Systems.

Resource Allocation;Stochastic Programming;Humanitarian Logistics;Markov Decision Processes;Simulation;Cross-training;Industrial and Operations Engineering;Engineering;Industrial and Operations Engineering

Parvin, HodaRomeijn, Edwin ;

University of Michigan

关键词: Resource Allocation;    Stochastic Programming;    Humanitarian Logistics;    Markov Decision Processes;    Simulation;    Cross-training;    Industrial and Operations Engineering;    Engineering;    Industrial and Operations Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/93888/hoda_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

In this dissertation, we study the impact of efficient resource allocation policies on the performance of a variety of systems including service centers, manufacturing systems, and pharmaceutical distribution centers.In Chapter 2, we investigate the optimal server scheduling policy in service industries such as call centers and off-line information technology service centers. We model the system as a Markov Decision Process and analytically characterize the optimal server allocation and scheduling policy. We also propose an efficient heuristic to improve server scheduling with no need to solve the MDP formulation. Our computational results confirm the effectiveness of our heuristic as compared to other well-studied routing algorithms from the literature.In Chapter 3, we propose a new production line design framework for a U-shaped production system consisting of several stations and cross-trained workers. We address efficient line design principles to enhance the system;;s throughput while keeping the number of required skills per worker significantly lower. We design an extensive test suite and use simulation to show that the system we designed can achieve nearly the same level of throughput as a fully cross-trained system.In Chapter 4, we present two-stage and three-stage stochastic network flow formulations to address the problem of deploying disease treatment in developing countries when the demand is uncertain. We find efficient distribution strategies that improve access to treatments at a minimum cost. We use demand data on the facility-based malaria treatment distribution provided by the Malawian Ministry of Health. We show that the proposed stochastic approaches can effectively reduce shortages and lower transportation costs.Finally in Chapter 5, we address the problem of routing incoming calls in an information technology service center with cross-trained servers and heterogeneous demand. Another important characteristic of the studied model is the fixed task completion deadline associated with each incoming service request. If the deadline is not met, a relatively large deadline violation penalty will be charged. To design and implement the proposed routing algorithm, we use real data from an industrial research partner. The simulation results confirm the effectiveness of the proposed routing heuristic in improving customer satisfaction by avoiding deadline violation.