The End Is Near

Well as the post headline suggests I am now into the final few days of this project but fortunately things are going well and should be finished with plenty of time to spare, well not plenty but enough.

I have pretty much finished my dissertation with just the abstract to do and then touching up a bit. Not sure about my second appendix but may just add it in anyways since as I have discovered the domain knowledge in a learning Algorithm is very important so my rules base may help someone else should they try to do something similar.

Showcase is on Friday 22nd May and anyone that has seen this blog and wants to see the end result should come along and have a look. I have enjoyed the experience of writing it and hope that someone manages to find a use for it some day.

Home Stretch

Well I was able to get a first draft finished for Friday as I had planned and I was able to get it looked over. Its still a bit short of the word count but with the improvements that I plan on making there should be no problem making the count for my final copy.

With the maths exam on Wednesday I won't be maintain my dissertation until after then. But when I do I should be able to get another draft finished, including references, before the end of the week giving the weekend for it to be checked by Dr Ellison before I have to hand it in for marking.

Reporting

Well with the conclusion of testing it is now time to get back to writing up my Dissertation.

So far I have completed a few first drafts of some of the sections which have been read over by my project supervisor, Dr Dave Ellison, and some comments made. I now therefore have to move on to continue writing while preparing for the Project Presentation on Wednesday.

Fortunately the two tie in together so I am planning to work on my testing and results sections of the dissertation so that I will be able to add them into my Presentation. The secondary plan is too make good progress overall and have a first draft of the complete dissertation ready by Friday. This is not an impossible task but needs to be juggled with math revision as well.

Testing Continued

Well fortunately things are back on track after performing some plenipotentiary tests.

After tweaking the maximum reward and penalty and by correctly implementing a weight distribution function, as well as changing the code for the script so that when a build force rule is called its shown immediately in the output and not just after the force has been completed the problem of the states scores getting stuck at the same amount have been removed so after completing the initial tests of 10 runs of DS for each of the 3 different static AI's on the three different maps will be able to produce some final results in the next day or so.

Testing

Well finnaly time to start running my test, unfortunatly things not going to plan.

It started off well and the Dynamic Script Player only took 3 matches to achive victory over the static player, however the same could not be said when a new static tactic was introduced. Along with that I noticed that the scripts being used were the old versions so I have been forced to restart my testing with the new scripts and updated Dynamic Scripting code, and so far I have not been able to match the initial success I saw with the broken files.

But I supose it is better to fail doing it right than to succeed doing it wrong.

Im going to continue testing and try find a more improved rules base if neccisary as this would be a key factor I think in why its not developing properly.

Dissertation Writeup

The past two weeks since the start of the Easter break has allowed me the time to start writing my dissertation properly.

So far I've been able to cover the major sections of my literature review including the introduction to Learning AI and Dynamic Scripting, What makes good learning AI and how I implemented Dynamic Scripting into WARGUS.

So now what I'm going to do is finish my AI coursework due at the beginning of next month and then do my testing of my application which will allow me to complete my main methodology section as well as get started on my Introduction and conclusion.

Update

So my application is finally finished, not quite to the most optimal position as the STRATAGUS engine somehow conflicts with normal LUA functions. This problem was solved by writing the Dynamic Scripting algorithm in a C++ program which can be run after a WARGUS match to deliver a updated script.

Therefore I am now quite a bit ahead of my projected progress so the plan is to catch up on the writing over the Easter break and to try have a first draft of my dissertation done as quickly as possible.

Here is an initial proposal for the content of my dissertation:

Contents
1. Abstract
2. Introduction
3. Literature Review
3.1. Current methods of learning Artificial Intelligence
3.2. Problems with current techniques
3.3. What defines good learning Artificial Intelligence?
3.4. Introduction to Dynamic Scripting
3.5. Why Dynamic Scripting is better than current methods?
4. Application
4.1. Explain how Real Time Strategy games work and introduce WARGUS
4.2. Why is WARGUS suitable testing environment?
4.3. How Dynamic Scripting is implemented into WARGUS
5. Testing
5.1. How tests were performed
5.2. What were the results of the tests?
5.3. Conclusion of the results in relation to evolutionary requirements
6. Future Study
7. Conclusions
8. References
9. Bibliography
10. Appendices