Cognitive Robotics Lab: State Space Models for Reinforcement Learning

This repository contains information on relevant background, the provided code and exercises of the cognitive robotics lab on state space models for reinforcement learning.

Installation

I recommend creating a new environment for this project using Conda. Then, clone this environment and change into the base directory by running

git clone https://github.com/maltemosbach/cognitive-robotics-lab-2021.git
cd cognitive-robotics-lab-2021

After activating the environment created for the cognitive robotics lab, running

pip install -e .

should install the code in this project and required dependencies.

Training models

To start a run of a PlaNet agent execute the following line.

 python -m lab.run algo=planet env=Pendulum-v0

Exercises

You find an overview of the problems and details on how to solve them in exercises.