Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Design and implementation of an automated bridge playing agent

Asiminakis Marios

Full record

---

Summary

The game of bridge is one of the most challenging card games, making it a mental sport like chess. Building an automated bridge player is a great challenge in the field of artificial intelligence because the players lack perfect information. A game of bridge consists of the bidding phase and the playing phase. In this thesis we mainly addressed the playing phase. We implemented an agent that solves the perfect information variant of bridge, called double-dummy bridge. In order to achieve that, we had to theoretically prove the correctness of our methods. Afterwards, we designed and implemented methods for overcoming the problem of imperfect information from the declarer's point of view. Three different agents were implemented, one based on the widely used Monte-Carlo approach, while the other two are based on a novel method which gathers information more effectively. We also implemented two agents that played as defenders. While they both play with perfect information, one of those tries to exploit the lack of information for the declarer. We studied the effectiveness of our agents by testing them in a wide range of game scenarios, with very encouraging results.

Available Files

Services