Skip to content

Latest commit

 

History

History
10 lines (10 loc) · 553 Bytes

README.md

File metadata and controls

10 lines (10 loc) · 553 Bytes

RL

reinforcement learning demos

NarrowDoor

Use Q-leraning/SARSA/Policy Gradient to train a differental drive robot with only two ultrasonic sensors 
to go through a narrow door which is just a little wider than the robot.

Follower

Use Deep Reinforcement Learning(DQN) to train a differential drive robot with a kinect to follow a worker.
To speed up training, supervised learning is used.