WhiteDolphin: A TAC travel agent

Abstract

In this paper, we detail our WhiteDolphin agent that was designed for the Trading Agent Competition (TAC) Travel game. Specifically, we employed the multi-layered 1KB framework to design our strategy, and describe the intricate cogs involved at the different layers in this complex decisionmaking process. We focus, in particular, on WhiteDolphin's strategic behaviour when bidding in the different types of auctions involved in the game, and how the information and knowledge required to support the complex decisions made is gathered and inferred respectively. Finally, we empirically analyse our agent by considering its performance in the 2006 competition where it ranked third. Copyright © 2007, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.