Stephan Carpenter (aka GameCarpenter) - Independent Game Creator
Research
Game Research
I have done a few projects in my spare time on games within the last couple of years, focusing on adaptive features and content generation.
The first is an attempt at a form of adaptive gameplay although certain setups ended up being too easy. A simple paddle game designed to modify itself based on the
state at certain key times, taking these as an indication of preferences. The controls are left, right and space, although what that does may vary from game to game.
This can be found here.
The second is an experiment in using color for indicating state instead of a traditional healthbar. The floor moves from white to black as the player looses health,
and the enemies are decorated in colors that roughly follow the positive-negative preferences for americans as well, from blue,to green, to yellow, to orange, to red.
The controls are left and right, with up for punch(high) and down for kick(low). This can be found here.
The last project is actually something I made some time back which fits in well with what I have been researching. It generates rather insane movie plots by using a
standard form and applying different combinatorial choices to it. This can be found here.
Finite State Machines and the Dynamic Programming Algorithm is a paper I wrote for my Stochastic Processes and Control class. It is a fairly
straightforward paper showing the parellels between finite state machines as they are used for games and the Dynamic Programming Algorithm. It's largely a matter of the different
way the concept of state is used, and a little about the way the dynamic programming algorithm could be used as a substitute and possible improvement upon FSMs.
Other CS Research
Living in a Competitive Environment: Studying Swarm Survival Through Genetic Generational Learning is a paper me and my group wrote
for Biomorphic Systems. Working on coding the genetic algorithm AI was a lot of fun. Every time there was a choice to be made about behavior I tried to make it into a variable.
In the end the 'gene sequence' included weighted preferences of many different behaviors (feed, move in a random direction, attack, run, breed) ways the behaviors would be
carried out (how persistantly, at what speed), leadership/followership preferences and so on. The evolved groups seemed to work better than artificially engineered groups entered
to compete with them. I think this points to the likelyhood that a group with many different types of member is more efficient, perhaps more adaptable, or more able to work
together in an emergent way than a group consisting of identical members.
For my computer graphics class, I worked on an implementation of the type of controls present in a normal game. I also added some minor tweaks to simulate a running animation.
The graphics themselves were very simplistic as I spent most of my time on making the interactivity work well. A copy of the '3d' movie I created of the UC logo with OpenFX
designer and animator, a 2d static image created in OpenGL under C++ and the paper I wrote explaining my project are available here.
I'd be interested in getting some feedback on what I've presented here, so feel free to shoot some my way.
-Stephan Carpenter
