Optimal/near-optimal static and dynamic distribution policies for an integrated logistics system
Abstract
This dissertation develops optimal/near-optimal allocation and replenishment policies for a multi-level, periodic-review, fixed-route logistics system which is comprised of a Logistics Center and several non-identical retailers located along a fixed route. The Logistics Center coordinates the allocation and replenishment function for the system in a centralized manner. For determining the allocations, it employs one of the two types of allocation policies: (i) a "static" allocation policy which determines the allocations to all the retailers as the delivery vehicle commences the route or (ii) a dynamic allocation policy which determines the allocations sequentially at each retailer as the delivery vehicle negotiates the route. The later policy allows greater amount of risk-pooling due to the delayed allocations and hence results in lower system-variance and lower system expected cost. Associated with each allocation policy is the corresponding replenishment policy that the Logistics Center pursues. The retailer's model also could be one of the two types: (i) the conventional or "unconstrained" type in which the retailer faces an independent, stationary, stochastic demand (assumed normally distributed for the most part) and incurs a proportional holding and shortage cost at the end of each period or (ii) the "constrained" type in which the service-level and holding capacity constraints are superimposed over the unconstrained retailer's model. In addition, the constrained retailer also incurs an extra component of cost (demurrage type) for the amount of overflow, if any, beyond the maximum holding capacity. The system objective is to minimize the average expected inventory cost per period for the system over an infinite horizon. The salient contributions of this dissertation are: (i) Development of a near-optimal, quickly computable dynamic allocation and replenishment policy for the logistics system comprised of the unconstrained retailers. The policy is contingent upon an assumption called the 'dynamic allocation assumption' which is empirically/analytically shown to hold with substantial frequency for systems with low ($\leq$ 0.4) coefficient of variations of the retailer's demand. For such systems, the dynamic distribution policies are analytically shown to incur 30-60% lower costs than the corresponding static counterparts analyzed in the literature. Parameterizations that specifically favor deployment of dynamic policies in lieu of the static policies are identified. (ii) Development of two methods for determining the optimal allocations to the retailers. The methods are applicable to the systems comprised of either constrained or unconstrained retailers. (iii) Development of a heuristic method that determines the replenishment quantity of a logistics system comprised of the constrained retailers.
Degree
Ph.D.
Advisors
Schwarz, Purdue University.
Subject Area
Management|Operations research
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.