Hi, I have a question about relative control in shadowhand environment, why in each "pre_physical_step", we use the prevous target as current target? Since the action we set for last step can not make sure the target can be reached, why not get the current shadow hand joint position to calculate the new target, otherwsie will have mismatch to the observation of the joint position?
Hi, I have a question about relative control in shadowhand environment, why in each "pre_physical_step", we use the prevous target as current target? Since the action we set for last step can not make sure the target can be reached, why not get the current shadow hand joint position to calculate the new target, otherwsie will have mismatch to the observation of the joint position?