Modeling Service Performance and Dynamic Worker Allocation Policies for Order Fulfillment Systems
Type of Degreedissertation
Industrial and Systems Engineering
MetadataShow full item record
In this dissertation, we propose several dynamic worker allocation policies based on the sojourn time distribution in due-date order fulfillment systems. To establish a sojourn-based policy, we must be able to compute the sojourn time distribution for an arriving order and each order in the system. We require the distribution because we must make a probabilistic statement about the completion time of an order. We need two types of sojourn time distribution for the system. The first is a steady-state distribution to estimate the probability of success for an arriving order. The second is the state-dependent sojourn time distribution, which estimates the probability of success for an order in system. First, we develop an approximation model for the sojourn time distribution of customers or orders arriving to an acyclic multi-server queueing network. The model accepts general interarrival times and general service times, and is based on the characteristics of phase-type distributions. Distributions produced by the model agree well with those produced by simulation for a variety of serial and general queueing networks. Second, we develop an approximation model for the state-dependent sojourn time distribution of customers or orders in a multi-server queueing system, when interarrival and service times can take on general distributions. The model is based on the characteristics of phase-type distributions, and uses a Markov process to represent the all-busy period for a waiting time distribution. We also discuss how the model might be used to execute dynamic delivery promises in an order fulfillment system. Third, we propose several dynamic worker allocation policies to maximize service performance of an order fulfillment system. Our policies make use of the state-dependent sojourn time distribution for an order, which we compute with a model based on phase-type distributions. Our approach differs from other work on dynamic worker allocation in that we focus on service performance as perceived by the customer, instead of traditional system performance measures such as cycle time, throughput, and WIP. Our results suggest that order fulfillment systems can significantly improve their service performance by moving the right number of workers to the right place, at the right time.