Reinforcement Learning

Repo for the RL course at TU Darmstadt.

Algorithms

• Natural Actor Critic
• DDPG

Platforms

• Furuta Pendulum
• Double Cart Pole
• Ball Balancer

Seminar Paper

• Furuta Pendulum Report
• Improvements for NAC (not yet specified)
• Extensions for DDPG and Analysis of it's Components

Challenges

The Challenges are practical RL assignments during the semester. The environments use the gym library by OpenAI and can be found in the following GitLab Repository of Quanser Robots.

1. Dynamic Programming (using the Pendulum)
2. LSPI & DQN (using CartPole)