DogTrot_RL

Caution: Beware of Dogs!

Implementation of Augmented Random Search Algorithm to teach a loyal dog to go and fetch a stick :P

ARS is actually up to 15 times faster than other algorithms with higher rewards in specific applications! Which is pretty fascinating!

This algorithm uses the method of finite differences to adjust its weights and learn how to perform a task.

Cloning

$ git clone https://github.com/7enTropy7/DogTrot_RL.git

Dependencies

$ pip3 install -r requirements.txt

Testing Instructions

$ python3 main.py

Note :

• Make sure you have licensed MuJoCo physics simulator set up on your machine.

Output

HalfCheetah-v2 in MuJoCo

HalfCheetahBulletEnv-v0 in PyBullet

Author

• Unnikrishnan Menon