Universidad Complutense de Madrid :: NILS mobility project
Abel Extraordinary Chair - Profiles
Gregorio Tirado
Venue: Norwegian University of Science and Technology, Norway. From: April 7, 2010 to: October 6, 2010
| Gregorio Tirado is assistant professor at the Department of Statistics and Operations Research, Universidad Complutense de Madrid, Spain, His project, together with Kjetil Fagerholt, consists of tto inspect whether the long term effects of planning can be improved by incorporating information about uncertain (stochastic) future events in the tactical planning phase. To do this we will build a simulator to evaluate the planning methods: based on data from real shipping companies we will be able to run simulations of their operations over several months (up to one year may be required to achieve realistic valuations). During the simulations decisions regarding routing and scheduling are determined using either current state-of-the-art methods or using novel methods that incorporate stochastic information. The simulations can thus be used to evaluate the expected value of adding stochastic information. The methods that incorporate stochastic information are as follows. About his work: "Initially we will focus on methods that have been successfully applied for other transportation problems: Stochastic information is built into scenarios. These scenarios correspond to deterministic planning problems that can be solved by the current methods. The scenarios are solved repeatedly, while gradually making decisions to be implemented in the simulation. Two versions of this method can be evaluated: a dynamic stochastic hedging heuristic and a branch-and-regret heuristic, with the latter being more sophisticated. However, both of these variations were developed for a land-based application where response time was crucial. Thus, they can be described as construction heuristics aiming to produce solutions as quickly as possible. Despite showing improved results compared to methods that do not include stochastic information, we believe that they can be improved further. Part of the proposed research will be to add local search heuristics to the solutions produced by the construction heuristics." |
Universidad Complutense de Madrid
Facultad de Ciencias Matemáticas - Instituto de Matemática Interdisciplinar
Tel. +34913944385
imi@mat.ucm.es