Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[namer]

Hi everyone,

I'm trying to get an agent to learn to navigate different kinds of 2D maps with reinforcement learning. My group is especially interested in showing that our methods are good at generalizing to new kinds of maps, but one challenge is finding maps that are actually qualitatively different in interesting ways. We have code for generating simple labyrinths and room-based maps, but I figure that if anyone knows about procedural map generation, it's this community

I'd be grateful for any and all pointers to methods for map generation, for 2D grid-based maps that just have (1) a start point; (2) an exit; (3) wall- or non-wall blocks -- we aren't worried about complex terrain for now. Ideally we are after methods where there's open-source code out there, but pseudocode or mathematical descriptions are great too!

Thanks!