Trajectory planning based on Reinforcement Learning with Hindsight Experience Replay, Prioritized Experience Replay & Dense Reward Engineering to solve openai-gym robotics "FetchReach-v1" environment using PyTorch & Tensorflow2.
-
Dense Reward Engineering: Engineered vector based distance measure to replace sparse rewards.
-
Hindsight Experience Relay (HER): Implemented HER Future Strategy based goal sampling for buffer augmentation.
-
Prioritized Experience Relay (PER): Samples and optimizes the past experiences ended with errors to get better future rewards.
-
Dense Reward Engineering
DDPG Agent 
PER + DDPG Agent 
-
Hindsight Experience Repay
DDPG Agent 
PER + DDPG Agent 
-
Previews from older implementation in TF2.
Dense Rewards HER /Reward Engineering/data/test.gif)
/HER/data/test.gif)
/Reward Engineering/data/test.gif)
/HER/data/test.gif)
Install dependencies using:
pip3 install -r requirements.txt - Additionally install 'mujoco_py' according to 'https://github.com/openai/mujoco-py'
- Name: Kanishk Navale
- Email: navalekanishk@gmail.com
- Website: https://kanishknavale.github.io/
![@dependabot[bot]](https://avatars.githubusercontent.com/in/29110?s=64&v=4){kind=small}