Distributed Control in Transportation and Supply Networks
Abstract:
This talk presents two examples of distributed control in transportation and supply networks. First, a spare parts supply network is considered under the typical goal of minimizing inventory and transportation costs under a service level constraint. We discuss how the optimal stochastic dynamic programming solution can be approximated using parametric dynamic programming, how this yields lower bounds for the optimal cost, and how this compares to the performance of local controllers. The second part addresses online train control in the Swiss railway network. We show how the problem is decomposed geographically into different control regions and present a prototype of an MPC controller for the area of Bern. An idea is outlined how to coordinate the local controllers using an additional global layer with a simpler model of the railway infrastructure and dynamics.
Biography:
Marco Laumanns received the Diploma degree in Computer Science from the University of Dortmund, Germany, in 1999 and the PhD from the ETH Zurich, Switzerland, in 2003. From 2004 to 2006 he was a postdoctoral fellow at the Institute for Operations Research, ETH Zurich. After a stay as visiting assistant professor at Arizona State University he returned to ETH as a senior research associate in 2007. In 2010, he joins IBM Research as a research staff member of the business optimization group in the Zurich Research Lab. His research focuses on optimization and control of logistics systems, in particular optimization under uncertainty.