Action assignment

  1. Observe
  2. Find the delay
  3. Apply it

The robot follows each action with a delay.

Action Dynamics

Action State
Find the controller delay Robot state follows the action. In the second step, the state curve moves earlier to align with the action; the yellow mismatch shrinks.

State-as-action Dynamics

State-as-action State Desired
State-as-action delay compensation The yellow state-as-action trajectory initially coincides with the orange dashed desired trajectory, which is drawn on top so both remain visible. The charcoal state follows with a delay. Advancing state-as-action by the learned delay makes the robot follow the desired trajectory more closely, with the same controller.