Environments

The Curiosity Gym package provides multiple Gymnasium-based grid environments to test and benchmark curiosity-driven RL algorithms. The dynamics of the grid are implemented in the abstract base class GridEngine. A concrete environment can be created by inheriting these dynamics and providing custom EnvironmentSettings, RenderSettings and EnvironmentObjects for the environment.

Grid engine

GridEngine

Abstract grid-based environment class that implements the gymnasium api.

Environment types

DistractiveEnv	Defines the structure of the curiosity-gym distractive rewards environment.
MultitaskEnv	Defines the structure of the curiosity-gym multitask environment.
SparseEnv	Defines the structure of the curiosity-gym sparse reward environment.