EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1827. A Rollout Algorithm for Dynamic Stochastic Purchasing Routing With Perfect Information Model Estimation
Invited abstract in session TC-64: Dynamic Vehicle Routing 1, stream VeRoLog - Vehicle Routing and Logistics.
Tuesday, 12:30-14:00Room: S16 (building: 101)
Authors (first author is the speaker)
| 1. | Daniel Cuellar-Usaquen
|
| Universidad de Los Andes | |
| 2. | David Álvarez-Martínez
|
| Industrial Engineering, Universidad de Los Andes | |
| 3. | Marlin Wolf Ulmer
|
| Management Science, Otto von Guericke Universität Magdeburg | |
| 4. | Camilo Gomez
|
| Industrial Engineering, Universidad de los Andes |
Abstract
This paper addresses the dynamic stochastic purchasing routing problem faced by a purchaser in negotiating with multiple suppliers to meet its demand. The purchaser contacts suppliers sequentially and must decide whether to buy based on purchase prices, available quantities, and routing costs. Each supplier can be contacted only once, with uncertain price and quantity information revealed upon contact. The purchaser faces a maximum number of contacts within a planning horizon. This problem is challenging because of the dynamic stochastic components of the negotiation process. The purchaser must weigh the information revealed by the current supplier against the remaining suppliers' behaviors and route consolidation to make decisions. We propose a rollout algorithm that combines dynamic programming with a perfect information model to determine whether to purchase from the current supplier and which supplier to contact next. Our method samples the supplier behavior scenarios and solves independent deterministic problems to estimate the opportunity cost of not purchasing from remaining suppliers or the impact of purchasing from the current supplier. We introduce a reverse discount factor to reduce the optimism in perfect information model estimation. Our method is validated against benchmark policies using real procurement data from an e-commerce platform. Results confirm the superior performance of the rollout algorithm and provide insights into benchmark policy effectiveness.
Keywords
- Vehicle Routing
- Programming, Dynamic
- Logistics
Status: accepted
Back to the list of papers