Ph.D. Exam
Topic
Algorithms For Procurement In Transportation Networks
Abstract
With the advent of the Internet, companies are getting interested in e-procurement to reduce cost and time of the procurement operations. In this research work, we address the procurement problem in the context of transportation networks. We consider two types of network procurement problems. For some industries, e.g., oil, electricity or natural gas supply chain, links of the networks (namely pipelines, transmission lines) are the object of procurement. We call this scenario as edge procurement, because edges are procured to facilitate formation of a network. For other types of businesses, there is a predefined underlying network specified by the sources and destinations of transportation of goods or services. A route in such a network is a set of edges leading from the source node to the destination node. For these industries, the procurement process includes selection of transportation agents and routes. We refer to this as transportation service procurement.
In the context of edge procurement, our research considers a distribution network model where each edge is owned by a selfish owner and a customer wants to supply a product or a service to each of the nodes in the network using a set of edges that results in the lowest possible distribution cost, i.e., a minimum spanning supply tree. The owners incur costs when their edges are used by the customer, and this cost information is privately known only to respective edge owners. We characterize the equilibrium price space of such an economy and analyze a descending price auction mechanism that discovers one of the Walrasian equilibrium prices satisfying certain conditions. We then study the strategic behavior of the owners in such an auction and show that following a greedy strategy is a Nash equilibrium for the owners.
Transportation service procurement auctions are typically characterized by limited number of rounds and low profit margin. In this context, our research addresses the bundle selection and bid determination problem on behalf of truckload carriers for single-lane and combinatorial bids in the successive rounds of an auction, which maximizes the profit for the carriers. First, we develop a framework for calculating the cost of providing transportation service for a lane or a set of lanes. We also develop a methodology for calculating winning probabilities of a bundle of lanes, given the bid price and historical information about the same lane. Based on the cost framework and the methodology for calculating winning probability, we develop a model using constraint programming which generates bundles with high probability of winning. We then use an MIP formulation to select a set of bundles which maximizes the expected profit, given a specific set of profit multipliers. After we select the set of bundles to bid on at a particular round, we compute a profit factor for each of the bundles using the industry average price for a lane or lane bundle from the historical figures and the winning bids for the past round of an auction. We develop two methods to update the estimate of the present market rate of a lane or a lane bundle, and two methods to evaluate a profit factor which will maximize the expected profit of the carrier. We perform numerical experiments to validate our results, and also to determine which strategy combination works best for the carrier.
We also study an extension of the above process where we relax the assumption of a carrier winning all the truckloads associated with the lane. A simple modification of the above models is used to analyze this case.
Full Thesis
Full thesis will soon be available here as pdf.
Final presentation will soon be available here as pdf.
Please contact me by email if you need these documents in the meantime.
|
|
