search

MultiAgentLearning / playground

PlayGround: AI Research into Multi-Agent Learning.
https://www.pommerman.com
Apache License 2.0
765 stars 217 forks source link

V1 environment step should apply collapse before and not after V0 step #174

Open suphoff opened 5 years ago

suphoff commented 5 years ago

Logically wall collapse happens before the next V0 environment game step - as observations, game done state and winners are all determined by the V0 environment game step.

However rendering currently sees an updated board - that is inconsistent with the observations. ( Example render shows no agents (all caught by collapsing wall) - but game is not done.

As this only affects rendering I suggest fixing it after the competition submission deadline. ( Just move the code collapsing the wall before the V0 step )

cinjon commented 5 years ago

Aye. Thanks, will wait like you suggest.