google-deepmind / google-deepmind/lab
How can I make new map with Capture The Flag model?
- Dominant language
- C
- Stars
- 7.4k
- Forks
- 1.4k
- PR merge metrics
- No merged PRs in 30d
Description
I want to try replicate the environment in the paper "Human-level performance in first-person multiplayer games with population-based deep reinforcement learning". But I‘m stuck in creating capture-the-flag maps.
1) Using the Lua API, I can create maps based on Text as the docs says.
https://github.com/deepmind/lab/blob/master/docs/developers/creating_levels/text_level.md
2) And I I can turn on the game model to CTF using following code.
----
function api:gameType()
return game_types.CAPTURE_THE_FLAG
end
----
But I don't know how to import the model Red Flag and Blue Flag into the level I create, since the flag is different from other pick-up models like apple or lemon, which needs to interact with agent more complexly. For now, I can play CTF mode in only one build-in map named "ctf_bounce.map"
I find there is no docs about CTF. So I want to ask if there is any way to create maps with flags using Lua-api or C++ methods.
3) For other methods like using third party tools like GtkRadiant to make maps, I failed in install the game engine Quake III Arena because it's hard to install and run in Linux system. And I guess using GtkRadiant can't create different maps automaticly.
Does anyone build maps with capture-the-flag mode successfully or know how to deal with this problem? Thank you.
Contributor guide
Assessment
This issue has not been assessed yet.