Intro Hello everyone (GSoC-Behavior Trees)

Omar Verduga

New Member
Hi, everyone, some quick details about me
  • Name: Omar Verduga
  • Social: Twitter: @omar13x
  • From: I'm from Mexico , but I'm doing a PhD in London (Cognitive Science, but my backgrounds are in Computer Sciences.)
  • Skills / Tools: Before doing my MSc I had 3-year experience using java for web , standalone but mostly server side projects. (In fact I'm a certified Java Programmer for the old 1.4 version). After that I was involved in videogame industry and interactive arts, using C++ , C# , AS3 and javascript for some projects ( like another 4 years of experience in this kind of projects). At this moment, in my PhD I'm using mostly Python( Deep Learning stuff), so I need to keep my skills to date :).
  • Found via: I found Terasology listed in past GSoC projects. At that time I was unable to apply given some dates with my offer letter, but this time I'm ready.
  • Interests: I'm very interested in the Behavior Tree AI system. I love gameplay and AI most than graphic stuff, for me is in there were the suspension of disbelief resides.
  • Extra: I was also thinking in propose terasology as a test bed for OpenAI environment, and link it some way to use it in the Reinforcement Learning world, but I will wait to know more about OpenAI before compromise myself to something potentially out of the time scope.
 

Cervator

Org Co-Founder & Project Lead
Contributor
Design
Logistics
SpecOps
Hello fellow Java 1.4 certified programmer! :D

Sounds like you've got some impressive background there! AI is a big field and we could use a lot of help with it. I've read a bit about OpenAI just haven't had the time to dig in much. Would love to see some more ideas on that, although yeah I wouldn't necessarily suggest merging that with the BT item, it seems like reinforcement learning could be a whole other item.

We've talked with patham9 from https://github.com/opennars/opennars some now and then about using Terasology as an environment target for OpenNARS agents. Having them learn visually how to play the game. He actually mentioned expecting to have some time this summer to try it out. Probably the tooling of being able to have the agents connect to get a video stream would be among the trickier things to set up, and that links to another GSOC possibility about making a headless client for automatically driving a player around in a world. That would be huge for automated acceptance testing and so on.

Lots of good potential. Let us know if you run into any issues setting up or have questions!
 
Top