search

geek-ai / MAgent

A Platform for Many-Agent Reinforcement Learning
MIT License
1.68k stars 332 forks source link

more details about the observation #35

Closed woaipichuli closed 5 years ago

woaipichuli commented 5 years ago

I consider the information provided in the doc about the observation of agents may be not sufficient enough. I wander what the group 2 or group 2 minimap stands for in the spatial observation. Besides, how many numbers the embedding ID in the non-spatial observation takes up in the sequence（【ID embedding, last action, last reward and normalized position】)？ I consider these will be helpful for me to better understand the game. Thank you.

woaipichuli commented 5 years ago

Is the last action in the non-spatial sequence encoded in a one-hot style?