Displaying discussions with Prof. Vangheluwe here.

13th of July 2012: It had been a while since our last meeting due to other projects and exams. We discussed the status of my first research internship and laid out the basiscs for the second one. The goal now is to have all the work of my first internship done as soon as possible. I have started writing on a text and provided this to prof. Van Gheluwe. so that this can be iterated upon. Cocnerning the second internship: I will start using T-Core to develop the same pac-man proof of concept that I used in ACT-R and SOAR. Once this is done and the necessary comparisons are made we will further discuss where to go.

April 2012: We went over all the work I had done and the papers I read. My ACT-R model is starting to work, but I still have to get started on SOAR. Some additional topics were suggested by prof Van Gheluwe to put in my report such as behavior trees.

26th of March 2012: We discussed SOAR and it's uses. I found a paper that used SOAR to implement NPC behavior which is good news. It was decided that I will implement Pac-Man in Act-R and SOAR to compare usability, efficiency, ... I can find the optimal model for a ghost, by first modelling the user. This is something I will do. It will not be playable by an external player as such, but it will simulate a human. Once this is done, I will add the ghosts and inspect the effect this has on the final game. Prof. Vangheluwe gave me a small explanation about tools to keep a decent bibliography such as Jabref and mendeley. I will be using Mendeley. It is clear the time has come to start using the tools I read about. The first goal is implementing the pac man.

On top of this I have found some new papers that I will have to read. As mentioned before one of these actually used SOAR in a game engine. I will have to contact my colleague Gino to ask about some tools he is using. The goal of his research is the same as mine, except for the fact that he uses statecharts instead of rules to describe behavior. It would be nice if we used the same tools and engines so a decent comparison can be made.

12th of March 2012: I had my second meeting with Professor Vangheluwe. I discussed the contents of the python Act-R paper. It illustrated the functionality of act-r quite well. Prof. Vangheluwe suggested that I implement a project he did using ACT-R (Pac-man project using atom³). We also decided that I have to look into SOAR, an alternative to Act-R

Another thing I will look into is the use of timing in Act-R. In a game engine, time has to look continuous, even if it in fact is discretised. It is unsure of this is possible in ACT-R

5th of March 2012: My first meeting with Professor Vangheluwe. We discussed the goals of this research internship. The plan is to dedicate two research internships and my thesis to one subject, namely rule based generation of NPC behavuior

In my first research internship I will look deeper into cognitive modelling tools, more specifically act-r and SOAR, to see if they are usable in game development. On the other had, I will take a look into different game engines to find out if it is possible to use different tools to adapt the AI of NPC's.

In my second research internship I will provide my own implementation of a rule-based system to specify AI behavior. I will be using T-Core and I might develop a Domain-specific language. Details will follow later, we did not yet come to a conclusion on the specifics.

Finally, in my thesis, this all comes together. I will implement a rule-based AI in an existing game engine to show that it is practically and commercially viable to use these techniques. I can then use this system to do different kinds of analysis such as playability, ... It should also be clear that this would be easier to use than just coding the AI. Maybe I will make a tool that abstracts the details away, time will tell.

Maintained by Kevin Wyckmans. Last Modified: 2012/07/19 16:02:23.