Truly Agent-Based Strategy Selection for Transportation Simulations

Bryan Raney
Dept.Computer Science, ETH Zürich, CH-8092 Zürich, Switzerland
Telephone: +41 1 632 0892
Fax: +41 1 632 1374
E-mail: raney@inf.ethz.ch

and

Kai Nagel
Dept.Computer Science, ETH Zürich, CH-8092 Zürich, Switzerland
Telephone: +41 1 632 2754
Fax: +41 1 632 1374
E-mail: nagel@inf.ethz.ch

Submission date: Nov 15, 2002

Corresponding author: Kai Nagel

Number of words: 7500

Abstract:

Multi-agent transportation simulations represent travelers as individual ``agents,'' who make independent decisions about their actions. We are implementing such a simulation for all of Switzerland, which is composed of modules that model those decisions for each agent, such as: (i) Route planner: Generates routes. (ii) Micro-simulation: Executes routes simultaneously; computes agent interactions, leading to congestion. (iii) Feedback: Iterates the above modules, resolving interdependencies. We discuss the operation of these modules, and focus on improvements made to the feedback system, such as an agent ``memory'' that allows agents to choose among previously used routes based on their past performance.