Welcome to the online Design Document for the Acquire Project. We have set up the site as a series of documents which describe conceptually how we intend to implement our system. There is one document for each major subsystem. Use the navigation bar on the left, or follow the links at the end of each section to the next section. The flow of the document should be smooth, beginning from a high-level overview of the entire system, and then moving to a discussion of each subsystem in turn. We end the Design Document with a brief list of System Evolution points.
Our project grew from the idea that building intelligent computer programs to play games could be applicable to real-world problems such as emulating the stock market, or solving complex engineering problems. Researchers in artificial intelligence have coined the term "agent" to describe these intelligent computer programs, which use any number of techniques, including searching, planning, or knowledge-based learning, to solve problems on their own. Many researchers who study game theory and artificial intelligence have applied their efforts toward building distributed systems of agents much like ours. There are many open and interesting problems in AI related to game playing; we have chosen to address the problem of strategizing (i.e., formulating a plan) in a world where only partial information is available. Our goal can be stated in this way:
We propose to design and construct a system with components that incorporate the concepts of intelligent agents, collaboration, and game theory to model a specific domain. These agents will communicate openly with an engine that defines the domain. The domain chosen for this project is Avalon Hill (now Hasbro)'s strategy game Acquire; however, the system will be open to the addition of new engines in the form of extensible modules. During play, the agents will be assigned one of a set of heuristics to use in evaluating the world and determining how to act upon it. The heuristics are comprised of a number of methods that can evaluate the state of the world and compare the outcomes of several actions. The system is intended to play many games of Acquire in one game session with a given set of agents and heuristics and outputting a simple statistical analysis of which agent-heuristic was the most successful (i.e., won the most games in that session).
Because we wanted to build an easily extensible system that could be applied to many domains other than the one we specifically build for it (Acquire), the design had to take this into account and be general enough in certain areas and specific enough in others. The choice of Acquire was based on the fact that the game is an interesting strategy game in and of itself. However, the research behind this project is intended to be applicable to nearly any strategy game which is similar to Acquire, so that other engines could be built using the same agent architecture we incorporate into this project.