Solving Inverse Kinematics with LULA Kinematic Solver

You can use Lula kinematic solver to solve FK and IK for your robot. Just like the RMPFlow, Isaac Sim supports Franka Panda and several other manipulators. You can also add your robots by making configurations for them.

All Lula algorithms, including RMPFlow, requires a description file for the robot. Isaac Sim supports four manipulators by default, so you can directly use them. To support your robot, use the Lula robot description editor.

from omni.isaac.motion_generation import ArticulationKinematicsSolver, LulaKinematicsSolver, interface_config_loader

kinematics_config = interface_config_loader.load_supported_lula_kinematics_solver_config("Franka")
_kine_solver = LulaKinematicsSolver(**kinematics_config)
_art_kine_solver = ArticulationKinematicsSolver(_franka, _kine_solver, "right_gripper")

You can also check all the frames available in the robot:

all_frame_names = _kine_solver.get_all_frame_names()

To solve the inverse kinematics, you need to specify the robot base pose as well as the desired target pose. The solver will generate a motion action to reach the target pose.

# Desired end effector pose
target_pos = np.array([0.0, 0.0, 1.0])
target_rot = np.array([1.0, 0.0, 0.0, 0.0]) # wxyz quaternion

# Solve IK
robot_base_translation, robot_base_orientation = franka.get_world_pose()
kine_solver.set_robot_base_pose(robot_base_translation, robot_base_orientation)
action, success = art_kine_solver.compute_inverse_kinematics(target_pos, target_rot)

# Apply action
if success:
    franka.apply_action(action)
else:
    print("IK failed")

Of note, if you use Tasks and creates multiple environments with offsets, add the offset to target_pos here.