Isaac Lab Reach Task: UR10 with Gripper

Overview

In this tutorial, we create a custom USD file by configuring the UR10 robot with the Robotiq 2F-140 gripper in Isaac Sim. We’ll then train our robot to perform a “reach” task where the objective is to align the robot’s end-effector pose with a dynamically changing target pose, represented in the simulation by a hovering coordinate frame widget. This tutorial is based on “Train Your Second Robot in Isaac Lab” on Nvidia Deep Learning Institude, covering the “Reach” task provided with Isaac Lab.

Our training pipeline will:

1. Configure the robot in Isaac Sim → Set up the customized UR10’s model with Robotiq 2F-140 gripper.
2. Train the control policy in Isaac Lab → Learn a policy that commands the robot to match its current pose to the target pose.
3. Validate in Isaac Sim → Test the trained policy in simulation for accuracy and robustness.

By the end, the policy will enable the robot to smoothly and accurately match its end-effector pose to the commanded goal pose.