Inputs from those who knows programming

[Zaphling]

I'm planning to have an Expert Systems as my solo mini-project in my AI course. I was planning on making Encounter Maker for DMs.

First of all, I need to make a database for all monsters available. Then the database for traps, hazards, and what-nots for an encounter.

My initial idea is a series of questions for the DM to take before the program generates the battle map with the monsters. I'll provide a special question for themed battles. It might be "What monster theme would you like?: reptiles, demons, undead, immortals, feys, etc."

The AI part of the software/program will be reflected on the very first question I will provide for the DM. "How many PCs are there?" then "How many defenders/strikers/leaders/controllers?"
These questions will help the program determine how to counter the PCs best. Like if there are more melee warriors, then the program will generate a map with impassable terrain to give the melee PCs a challenge/threat paired by monsters that have advantages with those terrains.

Then the monster themes question will follow, followed by questions about threat level. "easy, medium, or hard encounter."

To summarize, the Encounter Maker gives the DM the best suggestion on how to counter the PCs. Then generates a map with the monsters that is best in it.

Note: This is a four month project and a full running program is not expected. This is a research-based project on Expert Systems.

Thanks to those who can give inputs.

Few problems: acquiring monster data.

 

[Fast Learner]

Definite problem. Monster data (even in database form) from 3.0/3.5/Pathfinder is easy, but 4.0 is protected IP.

 

[dvvega]

AI is something I used to teach before becoming a full-time stay @ Home father.

An expert system would be an adequate solution for what you are doing, however there is so much to Artificial Intelligence that could take your research to new levels.

I would love to see the program actual run a predictive execution to work out the percentage of loss by round to be expected and throwing away the encounter if it is greater than a base number for each level (easy/medium/hard). This would require the system to run a quick tactical simulation (non-graphical so it is internal and pretty fast) which involves monsters using cover, how that affects the PC to-hit rates and so forth.

The system would also take into account that the PCs are not meant to die either and that their losses are also a factor.

However I'm assuming that is probably going a little "crazy" with the possibilities. To answer your expert system questions:

Monster data can be generated by yourself using the rules - use the average damage by level and to-hit numbers by level information supplied for building your own monsters. You can create "average" monsters with that data and then add a small margin of change to each of the stats. Their attacks would be simply Artillery Single Target, Artillery Burst, Brute Single Target, etc. You could write an application to generate those monsters for you in the data structure you need for your expert system and away you go.

In addition, by doing it that way you are homogenising the data set and thus variance is kept to a minimum and your expert system will not be quibbling over little differences.

How, in the end, will this be useful to a real DM running a real game? Well you spit out the "Average Encounter" data and the DM can then try and fit monsters into the system from their own Monster Manuals ... OR ... they could go traditional Traveller style and go with these "fake" monsters and become a true storyteller and make up their appearances etc to fit their campaign themes.

D