Skip to content

Latest commit

 

History

History
13 lines (8 loc) · 577 Bytes

README.md

File metadata and controls

13 lines (8 loc) · 577 Bytes

Solving CartPole using Distributional RL

In this project,I implemented an agent that can learn to control a cartpole using C51 algorithm which is introduced by "A Distributional Perspective on Reinforcement Learning".Also I used double q-learning instead q-learning to keep stability during training.

graph

Final score(testing the agent)

It could achieve score of 195.19 over 100 episodes!

Run on a single episode(reward=200)