/Modular_MBRL

This is a Model-Based Reinforcement Learning implementation based on a modular software architecture suitable for extension and easy to understand and use.

Primary LanguagePython

This is a modular architecture for model based reinforcement learning using search.

The components are separated, facilitating the creation of the agents and the extension of the existing components

Architecture

The current implementation learns the hidden-states for planning and an action-mask.

MORE ABOUT THIS: PAPER

to try it, run test.py: it will ask you to choose from different components and then run.