Get Link Transforms

SUMMARY

Get Link Transforms returns a 4×4 homogeneous matrix per frame in the kinematic model. Useful for visualization, attaching coordinate axes, or computing relative transforms between arbitrary frames without calling Forward Kinematics per frame.

UNITS

Returns a dict of link name → 4×4 SE(3) homogeneous transform.

The Skill

transforms = robot.get_link_transforms()

The Code

"""
Read per-link world transforms for a manipulator.

Supports Universal Robots (UR), Epson, and virtual.

Usage:
    python get_link_transforms.py
"""

from loguru import logger

from telekinesis.synapse.robots.manipulators import universal_robots
from telekinesis.synapse import utils


def main():
    """Read every link's world transform at the current joint configuration."""

    #===================== Create Robot ==========================================
    robot = universal_robots.UniversalRobotsUR10E(name='UR10e')

    # ==================== Run Skill ============================================
    transforms = robot.get_link_transforms()
    logger.info(f"Number of frames: {len(transforms)}")

    # Convert each 4x4 transformation matrix to a pose [x, y, z, rx, ry, rz] (m, deg)
    for name, T in transforms.items():
        pose = utils.transformation_matrix_to_pose(T, rot_type="deg")
        logger.success(f"{name}: pose [m, deg] = {pose}")


if __name__ == "__main__":
    main()

Parameter Configuration

This skill takes no input parameters.

Returns

Type	Description
dict[str, np.ndarray]	{frame_name: world_T_link} - each value is a 4×4 homogeneous transform at the current joint configuration. Excludes "universe".

Raises

Exception	Condition
RuntimeError	The robot's state has not been initialized (the derived class did not call its internal state/publisher setup before this method runs).

Where to Use the Skill

• Visualization - Stream every link's pose to a 3D visualizer without running FK per frame.
• Sensor attachment - Look up the world pose of any named frame for calibration of the sensor.
• Relative transforms - Compose world_T_a and world_T_b from the same call to get a_T_b without two extra FK calls.

When Not to Use the Skill

• You only need the TCP pose - call Forward Kinematics or Get Cartesian Pose; both avoid the per-frame overhead.
• You need rendered mesh poses - use Get Visual Mesh Transforms, which composes each link's URDF <visual><origin> offset.