Reinforcement Learning and Learning-based Control

Prof. Dr. Sebastian Trimpe, Dr. Friedrich Solowjow
Institute for Data Science in Mechanical Engineering(DSME)
rllbc@dsme.rwth-aachen.de

---

The algorithms within this library were developed in the context of the class Reinforcement Learning and Learning-based Control (RLLBC) by the Institute for Data Science in Mechanical Engineering (DSME) at RWTH Aachen University. In this class we use this library in Lectures and Exercises. Students can also use the library to expand their knowledge through self-study. All algorithms are presented via Jupyter notebooks — see the individual folder READMEs for details:

• class_examples/ — lecture and exercise companion notebooks covering MDPs, DP, MC, TD, function approximation, and learning-based control
• tabular_examples/ — tabular RL implementations: Policy Iteration, Value Iteration, MC Control, SARSA, Q-Learning, Dyna-Q
• deep_examples/ — deep RL implementations: DQN, REINFORCE, A2C, TRPO, DDPG, TD3, SAC
• lbc_examples/ — learning-based control notebooks: LQR, Dynamics Learning, MPC, Bayesian Optimisation

You can find installation instructions below.

Installation guide

To install the library, please follow the instructions below.

1. Download the files

2. Install the latest version of Pixi https://pixi.sh/latest/installation/

   • make sure that you install the version for the operating system that you are using

3. Create the uv environment

   pixi install
   

4. Activate the environment

   pixi shell
   

5. Start up JupyterLab from your terminal with

   jupyter-lab
   

→ Now you should be able to browse your file system for the notebooks

Using the library on a local computer:

Once the environment has been successfully installed, the library can be easily accessed via the following steps:

1. Navigate to the project folder and open your terminal there. On Windows, use the powershell.
2. Activate the environment with

   pixi shell
   

3. Start up JupyterLab from your terminal with

   jupyter-lab
   

You are ready to browse the library.