This force-sensitive robot system gives robotics and AI researchers direct access to low-level control mechanisms and sensor data for advancing machine learning and manipulation research. The hardware targets top research institutions that need both sophisticated sensing capabilities and programmatic flexibility.
The arm operates with seven degrees of freedom, matching human-like dexterity for complex manipulation tasks. Each joint contains an integrated torque sensor, providing real-time external contact force estimation that's crucial for safe human-robot interaction experiments and compliant manipulation research. These sensors feed into a high-frequency 1 kHz motion control loop, allowing researchers to implement fine-grained control algorithms that respond to environmental forces within milliseconds.
Physical specifications matter for lab deployment. The arm reaches 855 mm with a 3 kg payload capacity while maintaining position repeatability under 0.1 mm. That repeatability level supports precision experiments where small deviations corrupt results. The system achieves 94.5% workspace efficiency, meaning researchers can access nearly the entire theoretical reach volume without singularity issues or awkward configurations.
Hand-guiding capabilities let researchers physically move the arm to desired positions, which accelerates demonstration-based learning workflows. Instead of programming waypoints, you can grab the arm and show it trajectories directly. The robot records these movements through its torque sensors, capturing both position and force profiles.
The Desk API provides programmatic administration and operation outside pre-built interfaces. Researchers can write custom control loops, deploy learning algorithms, and integrate the system into larger experimental setups. Integration support includes ROS 2 for robotics middleware, MATLAB for algorithm prototyping, and MUJOCO for physics simulation. These connections let teams move between simulated experiments and physical validation without rewriting control code.
Software updates arrive over-the-air without requiring technician visits or manual installation procedures. Torque calibration happens in the field rather than requiring factory returns, reducing downtime when sensors drift. Joint-space jogging gives direct control over individual joint angles, useful for debugging kinematics or reaching specific configurations. The FR3 Arm V2 adds intuitive joint rotation markings and X/Y axis indicators, helping operators understand the current pose at a glance.
The system gets engineered and manufactured in Germany with actuators selected for research-grade performance. The hardware optimization focuses on both computational performance and ergonomic integration into lab environments where space and accessibility matter.
Researchers working on contact-rich manipulation, compliant control, learning from demonstration, or human-robot collaboration get the sensor fidelity and control access these applications demand. The torque sensing at every joint distinguishes this from position-controlled arms that can't measure interaction forces directly.
