Programmer's Interface for the Red Alert 2
Computer Game
|
Ami Berler and Solomon Eyal Shimony.
Departement of Computer Science
Ben-Gurion University, Beer-sheva, 84-105, Israel.
e-mail : ami@cs.bgu.ac.il ,
shimony@cs.bgu.ac.il
July, 2001
|
Overview
Artificial intelligence is fast becoming an important component
in computer games (see "Human-level AI's Killer Application:
Interactive Computer Games", John E. Laird and Michael van Lent,
Proceedings of AAAI-2000, pages 1171-1178, extended version
in AI Magazine, 2001). In order to develop an AI agent for an
existing game, one must have complete flexibility in designing an agent
that can employ existing AI techniques, such as learning, reasoning
under uncertainty, as well as novel techniques.
While some provision
for user-defined AI is made in numerous games, ranging from selection
from canned scripts to scripting in C++, most games do not allow
full programmer access, such as the ability to write agents in an arbitrary
high-level language, or access to the file-system that might be required
for learning across multiple runs of the game. One way to allow the latter is
by writing a piece of code that masquerades as a copy of the game when
the game is run in multi-player mode, and thus allows full access.
|
The project consists of three parts:
- Capture and understand the messages that sent between two server and client
in the Multiplayer mode.
- Build an Interface that takes these messages and has a capability
retrieve data and act in the game.
- Write an example of a intelligent agent for one player
role.
|
|
Note: we do not have any information on message structure. These must either
be fund from the manufacturer(unlikely), or hacked from the data streams.
|