Skip to content

Action jerk penalty #12

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
kristof wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Action jerk penalty #12

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions training/playground/common/rewards.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,6 +79,14 @@ def cost_action_rate(act: jax.Array, last_act: jax.Array) -> jax.Array:
return c1


def cost_action_jerk(
act: jax.Array, last_act: jax.Array, last_last_act: jax.Array
) -> jax.Array:
"""Penalize jerky actions - second derivative of action for smoother motion."""
action_accel = act - 2 * last_act + last_last_act
return jp.nan_to_num(jp.sum(jp.square(action_accel)))


# Other rewards.


Expand Down
7 changes: 6 additions & 1 deletion training/playground/open_duck_mini_v2/joystick.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,7 @@
reward_tracking_ang_vel,
cost_torques,
cost_action_rate,
cost_action_jerk,
cost_stand_still,
reward_alive,
reward_foot_height_tracking,
Expand Down Expand Up @@ -81,7 +82,8 @@ def default_config() -> config_dict.ConfigDict:
tracking_ang_vel=6.0,
torques=-1.0e-3,
action_rate=-0.5, # was -1.5
stand_still=-0.2, # was -1.0 TODO try to relax this a bit ?
action_jerk=-0.1, # smoothness penalty for character-like motion
stand_still=-0.2, # was -1.0 TODO try to relax this a bit ?
alive=20.0,
imitation=1.0,
foot_height_tracking=0.5,
Expand Down Expand Up @@ -648,6 +650,9 @@ def _get_reward(
),
"torques": cost_torques(data.actuator_force),
"action_rate": cost_action_rate(action, info["last_act"]),
"action_jerk": cost_action_jerk(
action, info["last_act"], info["last_last_act"]
),
"alive": reward_alive(),
"imitation": reward_imitation(
self.get_floating_base_qpos(data.qpos),
Expand Down