Learning Agents in Board Games

摘要

Artificial Intelligence (AI) in games has become an integral part of game design architecture. A lot has changed since the days of scripted computer games that offered little challenge to human players or were so predictable to be non-immersive. Artificial Intelligent agents have found successful applications in industry, economic and robotics to name a few. In computer games, the intelligent agents known as "Non player characters" (NPCs) are considered to be autonomous entities. Based on scenarios, these agents may be self interested or cooperating to solve a shared problem. The scenario presented in this work is how simplistic learning can be achieved by an agent to exhibit Intelligence using an experimental set up of "TAG DELUXE" between our learning Agent Alpha which is pitted against other agents (or human) with no learning. The Application has been coded in the Java programming language with a graphic user interface for visual purposes and input to the application. The Alpha agent shows learning by using a reinforcement learning approach, it is also a reactive agent and shows an abstraction of the potential of these agents to learn from experience or make cognitive decisions. As computer game players demand better human like cognitive capabilities from NPCs, so will the application of intelligent agents try to bridge the gap to provide adequate Artificial Intelligence to simulate human decision processes.