PR2_Object_Manipulation

The Recurrent Deterministic Policy Gradient(RDPG) now works on the Fetch Reach Gym Env. The network architecture for both Critic and Actor is according to the following Paper: Sim-to-Real Transfer of Robotic Control with Dynamics Randomization

Code will updated soon!

The Learning curve with the Hindsight Experience Replay and the RDPG

Name		Name	Last commit message	Last commit date
Latest commit History 9 Commits
.vscode		.vscode
PR2		PR2
Plots from FF		Plots from FF
cartpole-random		cartpole-random
custom-mujoco-gym		custom-mujoco-gym
ddpg_act_change_sparse_50		ddpg_act_change_sparse_50
ddpg_checkpoints		ddpg_checkpoints
ddpg_checkpoints_1		ddpg_checkpoints_1
ddpg_checkpoints_dense		ddpg_checkpoints_dense
ddpg_checkpoints_episodes		ddpg_checkpoints_episodes
ddpg_cp_dense_500		ddpg_cp_dense_500
ddpg_cp_sparse_500		ddpg_cp_sparse_500
ddpg_l1_loss_200		ddpg_l1_loss_200
ddpg_l1_loss_50		ddpg_l1_loss_50
gym		gym
mujoco-py		mujoco-py
pr2		pr2
README.md		README.md
cartpole_mujoco.py		cartpole_mujoco.py
mujoco_first_test.py		mujoco_first_test.py
shadow_hand_pid_practice.py		shadow_hand_pid_practice.py

Provide feedback