Overcooked is a great environment to test RL agents. This integration allows to use the versatility and features of Cogment with the gym environment and visualization of the overcooked_ai repository.
Instructions to setup and play experiments are provided under doc/integration/overcooked_ai.md.
Other changes
Added run parameters for render width. Useful for integrations having smaller or larger visual representations.
Fix serialization issue of MultiBinary spaces encountered in environment integrations.
Random actor seed based on actor index, otherwise two random actors in a single environment would take the exact same actions.
Context
Overcooked is a great environment to test RL agents. This integration allows to use the versatility and features of Cogment with the gym environment and visualization of the overcooked_ai repository.
doc/integration/overcooked_ai.md.Other changes
closes #153