nmelone/reinforcement-learning-quad-copter
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Create README.md

Browse Source
This commit is contained in:
Nico Melone
2018-08-03 14:50:55 -05:00
committed by GitHub
parent fabec6a2dd
commit 8d31fc3c79
1 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
README.md Normal file
View File

@@ -0,0 +1,4 @@
# reinforcement-learning-quad-copter
A reinforcement learning task to have a simulated quad copter learn to achieve a particular goal
This project was part of the Machine Learning Engineer nanodegree from Udacity. The assignment was to create a task and agent to train a quad copter in simulation. The task was left open ended as well was the model but the simulation environment was provided. I chose to have the quad copter try to learn to fly to a designated location and hover there. This is seen in the task.py file under the get_reward method. I employed a DDPG agent through the Keras-RL infrastructure. DDPG is a model that uses two networks that cooperate to acheive a task. One is the Actor and performs a task while the other is a Critic that judges how well the actor did a task. Both networks start out untrained and learn in tandum.