Recording Cumulative Energy in Isaac Lab with SKRL

[amirhnourian]

Hello everyone,

I’m using SKRL to train my rover robot in Isaac Lab, and I’d like to measure the total energy consumed over each episode. By default, SKRL doesn’t expose energy‐usage metrics from Isaac Lab, so I’m looking for:

• A reliable way to compute and log the cumulative energy during an episode
• How to pull that energy data from Isaac Lab into my SKRL loop
• A method to reset the energy counter at the start of every new episode

Any tips or code examples on how to implement this would be greatly appreciated. Thanks!

[Toni-SM]

Hi @amirhnourian

Whether scalar value you put on the self.extra dict will be logged to TensorBoard

IsaacLab/source/isaaclab_tasks/isaaclab_tasks/direct/franka_cabinet/franka_cabinet_env.py

Lines 459 to 467 in 3d5ea25

	self.extras["log"] = {
	"dist_reward": (dist_reward_scale * dist_reward).mean(),
	"rot_reward": (rot_reward_scale * rot_reward).mean(),
	"open_reward": (open_reward_scale * open_reward).mean(),
	"action_penalty": (-action_penalty_scale * action_penalty).mean(),
	"left_finger_distance_reward": (finger_reward_scale * lfinger_dist).mean(),
	"right_finger_distance_reward": (finger_reward_scale * rfinger_dist).mean(),
	"finger_dist_penalty": (finger_reward_scale * finger_dist_penalty).mean(),
	}