Beta 1


Title General Game Playing Systems
Author Holt, Andreas
Supervisor Villadsen, Jørgen (Algorithms and Logic, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Master's thesis
Year 2008
Abstract General Game Playing (GGP) is a field in artificial intelligence (AI) that deals with systems that, provided with only the rules of an arbitrary game, can play the game in an intelligent way. This problem is much harder than making a computer play a specificc game since you cannot rely on predefined evaluation functions or any other domain specific knowledge. It is also more interesting from an AI point of view since the computer needs to show some intelligent behaviour in order to come up with a good move instead of just following a predefined formula. In this report we will investigate different methods of making such a system. We will look at what others have done, and an actual implementation of our own general game player is presented. This implementation can either use the minimax algorithm with a simulation based evaluation mechanism or the UCT algorithm based on Monte Carlo simulations. In the end of this report these two techniques are compared by playing against each other in different games. These comparisons shows that the minimax algorithm with an evaluation function is a good choice in GGP but the UCT algorithm can be very strong when given enough time or computational resources to make a suitable amount of simulations.
Series IMM-M.Sc.-2008-117
Fulltext
Original PDF ep08_117_web.pdf (1.09 MB)
Admin Creation date: 2008-11-17    Update date: 2008-11-17    Source: dtu    ID: 228660    Original MXD