Using Ryzom for a research project in AI

Hello everyone,

I am a CS PhD student at Georgia Tech, and I am working on a project that is attempting to create tools for authoring AI agents through demonstration. We have an existing framework for doing this in RTS games, and this year we are interested in applying it to RPG's.

I went on the irc channel, and a very nice person answered some questions for me, to help me clarify things... but I would like to get some more info.

Does the default system maintain a trace of a users gameplay? If so, how detailed is it, and is there any way that one could get their hands on a large collection of these? Or if we could ask for volunteers to submit those traces to us in some way/shape/form?

Primarily, we would like to use Ryzom to test our approach in as real an environment as possible. The way we have done this in the past (in Second Life) is to hook in our generated AI agent into the server as a client, and have it control the character in the same way as a person would. It seems that this is not possible with the live Ryzome server, since it forbids modified clients to connect.

Our other option would be to create our own server, but here we run into the overhead of having to come up with a quest structure, maps, etc., as well as finding some way to get people to come interact with our agents. So, would there be any way that we could either 1. find some way to test our AI on the live server, or 2. find an alternate map/quest structure that we could put up on our own server, and then invite people to interact with.

Thank you for reading, and thanks in advance for any answers! Also, I'd be happy to answer any questions about what we're planning on doing in our research!

Replies (3)

RE: Using Ryzom for a research project in AI - Added by ragnar-gd over 7 years ago

virtuousD,

are you still with us?

RE: Using Ryzom for a research project in AI - Added by bristle over 7 years ago

its getting harder now to hooked into second life because of their 3rd party clients rule. people are afraid of clients doing certain wierd stuff and there fore dont allow unauthorized clients. ryzom servers would be the same way, unless they allow some sort of softbot connect and experimentation. the best bet is opensim, which is almost the same as second life, or have a ryzom server and then build your ai routines there. the problem is having all your need and getting some other players on. it almost impossible to keep anyone on.

for years ive argued that the ai game group have a server but they are not interest either.

right now, you have opensim and ryzom. that it. the rest are all dead or nearly dead.

RE: Using Ryzom for a research project in AI - Added by ragnar-gd over 7 years ago

bristle,

that was some interesting info. I was ignorant of opensim, as, probably, i'm not interested in second life either. I just ran through their wiki.

The good thing of opensim - and worldforge, in this regard - is the inherited ability to add persistent content online and dynamically, like, adding shacks to a city when it's population grows.

Unfortunately, although you can have their servers up and running, and connect to them, and both have awesome technical features (some much beyond ryzom core), neither opensim nor worldforge have the stability (opensim) nor maturity (worldforge) to make a game with it within a reasonable time that people actually can play.

Thus, only ryzom remains for my purpose.

It's just that the ai-bot of Georgia Tech would have been very interesting for me for testing purposes.
Perhaps i have to make contact externally, as they seem to have left the ryzom forums.

(1-3/3)