Skip to content
/ drlnd Public

Projects for Udacity's Deep Reinforcement Learning Nanodegree

License

Notifications You must be signed in to change notification settings

tomtung/drlnd

Repository files navigation

DRLND

This repo contains project code for Udacity's Deep Reinforcement Learning Nanodegree. Each project consists of a self-contained Jupyter notebook solving a (modified) environment from Unity's ml-agent example learning environments.

Demo

The agent needs to navigate a 3D space to collect as many yellow bananas as possible while trying to avoid blue bananas.

The environment is solved with Deep Q-Learning, with several "Rainbow" extensions, including prioritized experience replay, noisy network, double Q-learning, and dueling network.

Demo

The agent needs to control a double-jointed arm to track a moving target in a 3D environment.

The environment is solved with Deep Deterministic Policy Gradient (DDPG), with extensions from Twin Delayed DDPG (TD3) and also utilizes prioritized experience replay.

Demo

A pair of agents needs to control rackets to play toy tennis with each other.

The environment is solved with Multi-agent DDPG (MADDPG), while incorporating extensions from TD3 and utilizing prioritized experience replay as well.

About

Projects for Udacity's Deep Reinforcement Learning Nanodegree

Topics

Resources

License

Stars

Watchers

Forks