Skip to content
ART — Advanced Realtime Tracking
GermanEnglish
← Blog & Guides & Papers

Direct Pose Control of Industrial Robots with ART Tracking — TU Berlin IWF

Closed-loop control of a FANUC robot using ARTTRACK5 / AT7-80 and DTRACK: position and orientation errors reduced by around 90% — a TU Berlin IWF and ART collaboration.

  • Robotics
  • HRI
  • Motion Tracking
  • 6DOF
  • Industrial robots

Optical tracking for precise industrial robotics

Industrial robots are fast and flexible — yet their absolute pose accuracy often falls short for machining tasks such as milling, additive manufacturing, or forming. In a joint project with ART, the Institute for Machine Tools and Factory Management (IWF) at Technische Universität Berlin implemented direct pose control: measurements from an ART tracking system correct the robot pose in real time.

Results: position and orientation errors were reduced by more than 90% in the static case and by around 90% in the dynamic case — including under external load at the end effector.

The project shows how optical 6DOF motion tracking closes the gap between commanded and actual pose in human–robot interaction and industrial manufacturing. See also our Robotics & Human-Robot Interaction solutions.

Closed-loop control: commanded pose, ART measurement, and robot correction (from Uhlmann et al., see below)
Closed-loop control: commanded pose, ART measurement, and robot correction (from Uhlmann et al., see below)

Setup: FANUC robot and ART cameras

Industrial robot: FANUC M-900iB/700

  • Payload 700 kg
  • Reach up to about 2.8 m
  • Dynamic Path Modification (DPM) option on the R-30iB controller

ART tracking:

  • Four ARTTRACK5 or AT7-80 cameras with DTRACK
  • Three-sided target combining flat and spherical passive markers, mounted on the robot actuator
Passive target with flat and spherical markers for robot pose (from Uhlmann et al., see below)
Passive target with flat and spherical markers for robot pose (from Uhlmann et al., see below)

The redundant camera layout delivers stable 6DOF end-effector pose data — the basis for real-time correction.

ART AT7-80 tracking camera
ART AT7-80 tracking camera

FaRoC system architecture

The FANUC controller is connected via the Python/Karel library FaRoC (Fanuc Robot Control — developed at TU Berlin IWF) over TCP/IP and UDP/IP. Compared with ASCII interfaces, FaRoC uses a binary data format:

  • faster communication cycles
  • larger feasible data streams
  • no round-off errors from text encoding
FaRoC architecture: tracking, control loop, and FANUC controller (from Uhlmann et al., see below)
FaRoC architecture: tracking, control loop, and FANUC controller (from Uhlmann et al., see below)

FaRoC supports trajectory corrections with a cycle time of about 8.5 ms. The pose error (commanded pose minus measured pose) drives offset values sent to the controller through the DPM interface.

Experimental results

CasePosition errorOrientation errorLoad
Static−98%−93%up to 1500 N
Dynamic (50 mm/s)−92%−87%up to 2300 N

Reductions are relative to operation without direct pose control. Correction remained effective under significant external force at the end effector.

FANUC robot with ART target on the actuator (from Uhlmann et al., see below)
FANUC robot with ART target on the actuator (from Uhlmann et al., see below)

Publication

Eckart Uhlmann, Mitchel Polte, Julian Blumberg, and Oliver Sobiak:
Real-time direct pose control of industrial robots with proprietary robot controller based on a redundant camera tracking system
18th CIRP Conference on Intelligent Computation in Manufacturing Engineering, 2024, Naples
DOI: 10.1016/j.procir.2026.01.049

Next steps