Dave on Dave:
I'm the Engine Lead and AI specialist for Ensemble Studios. I have a terrific
wife of one year by the name of Kristen. I'm a 25 year old graduate of the
University of Arizona with a computer engineering degree. I spend most of my
time working; free time is spent relaxing with my wife, going to movies, and
working on our house.
In most real-time games, the emphasis has been on multi-player play versus the
single player experience, and thus, the AI has suffered. What has Age of
Empires done to improve the single player experience in terms of the AI?
The simplest answer is that we have spent almost a year of development on the
AI and given it a very high weight (in terms of time and effort) compared to
other tasks. We've also picked a couple of key areas such as tactics and varied
strategies to focus on. This focus allows us to concentrate on making those
things jump right out and grab the player's attention (hopefully as a crushing
herd of war elephants rumbles through a human player's town:).
Did you address the computer AI and the unit AI separately?
Yes. The UnitAI is exactly the same for the human and computer units. We did
this to help reinforce the concept that we didn't want to cheat. The computer
player interacts with the game in exactly the same way as the human player
does. The UnitAI was completed first in order to let us play multiplayer games
(for design and balance reasons); the computer player AI was started after the
UnitAI was stable. The computer player also went through a comprehensive design
process that has paid us huge dividends.
How sophisticated in the AI, does it learn?
It's really pretty sophisticated. It does learn about the scenarios it plays
and also learns player tendencies (so it should improve over time). The tactics
were probably the hardest thing to get right; we had to make the tactical AI
module a lot more complicated that we originally thought in order to get it up
to the level of the human player (we ended up going through three totally
different models before finding one that clicked).
A major weakness with the computer AI in games is that they cheat, is this the
case with Age of Empires?
This is a tough question; we've debated the issue extensively. Our primary
focus for the AI is to have it create the best single player experience on the
market. As such, cheating is secondary to that. Our goal is not to cheat, but
we will if we need it to create a better game. We are getting closer to our
goal everyday, though. The only way that we have to cheat right now is to give
the computer player a resource boost at the beginning of the game. As we make
improvements to the AI, we are able to reduce that boost. We should be able to
remove the resource handicap on all but the hardest difficulty levels.
Does you computer controlled units maintain formation in combat? Also, do
individual units cooperate with each other?
Yes, we have what we call a playbook. The computer player has a ton of plays
that he can choose to run (simple frontal assaults, multi-front attacks,
formation attacks, protection formations, etc.) from all directions on the map.
It's a fantastic thing to see the computer attacking with catapults and
protecting those catapults with infantry. The unit cooperation is achieved with
a hierarchical combat model. Grouped units have a unit commander that runs the
combat for that group. If a unit runs into trouble, he'll report to his
commander for help. If the commander's group needs help, he'll ask the computer
player to send more forces over.
What else about the AI do you find worth noting? What makes it different or
We've attempted to do a complete AI that handles all phases of the game. I
think the major differentiations will be the tactics, strategies, and the
emphasis on not cheating. To play a game and have the computer build completely
different units than the last time and use them in a different way is a great
thing to see.
What's the most challenging part of what you do?
At this point, it's time management and execution. There are so many areas that
I'd like to spend my hours on: there are always a ton of features to add to a
game or an engine, the engine is never quite fast enough, the game designs
change as the product evolves, the AI can always handle one more situation,
etc. Deciding which items are most important and which deserve time is
critical. Even more critical is an accurate, successful execution of our
As technology advances, multiplayer games are more and more involving, where do
you feel that computer-based games are headed?
I see a big shift in two areas: the real expansion of the massively multiplayer
games and the resurrection of the adventure and role playing genres. Massively
multiplayer games have been around for a long time, but recent technology is
allowing us to do a lot of fun things (persistent games with a 100K players,
etc.) that will really open up the games to huge audiences. I think the
industry rotation is also coming around to role playing and adventure games
again. New technology like speech recognition/synthesis and improved AI will
play a big role in these, I think.
Was there a particular point in your life where you realized that you were
definitely headed towards game development?
I did some "time" at a large computer company -- not MS:). I hated every minute
of it and decided to do something fun; so I started a game company. That never
really took off, so I joined Ensemble Studios last year and have been happily
busy ever since.