Franka Research 3 is a 7-axis research robot with built-in torque sensors at every joint, giving labs human-like dexterity and precision control.
Photo source:
Franka
Try grabbing an egg without crushing it while
wearing oven mitts, and you'll understand the core challenge roboticists have
faced for decades: machines move with force, but rarely with feel. Franka
Robotics built Franka Research 3 (FR3) to close that gap. FR3 is the
reference world-class, force-sensitive robot system tailored for robotics and
AI, empowering researchers with easy-to-use robot features as well as low-level
access to the robot's control and learning capabilities.
So, what makes it stand out? The arm offers
seven degrees of freedom, giving it human-like dexterity that allows motion in
tight spaces and around obstacles, the kind of maneuverability a simpler four
or five-axis arm simply can't replicate. Therefore, researchers working on
delicate manipulation tasks, whether that's grasping irregular objects or
navigating cluttered environments, get a physical range of motion that mirrors
how a human arm actually moves.
Sensors are where this robot earns its name,
and FR3 builds its entire identity around them. Torque sensors are integrated
at each of the seven joints, finely estimating external contact forces, while a
fast control loop enables prompt reaction to collisions. That high sensitivity
unlocks advanced torque or force control possibilities and provides smooth,
intuitive hand-guiding, letting a researcher physically move the arm and have
it respond naturally rather than fighting back.
Precision backs up that sensitivity with hard
numbers: the robot carries a 3kg payload, reaches 855mm, achieves 94.5%
workspace efficiency, and holds position repeatability under ±0.1mm. In
addition, the Franka Control Interface, or FCI, gives roboticists who need
fine-grained control direct, real-time access to the robot without internal
transformation, including joint-level torque, position, and velocity commands,
plus Cartesian pose and velocity control. Furthermore, FCI streams 1 kHz
measurements of joint data, external force estimation, and collision
information back to researchers, while libfranka, the platform's open-source
C++ interface, and franka_ros connect the whole system into the broader ROS 2
ecosystem, complete with URDF models, RViz visualization, Gazebo simulation,
and MoveIt! motion planning.
A robot that ships once and never changes again
quickly falls behind in a field moving this fast, which is why FR3 takes a
different approach. The platform continues to grow through free over-the-air
software updates. The most recent release, System Image v5.9, introduced
joint-space jogging that lets researchers control individual joints directly
from the Desk sidebar for precise, intuitive pose adjustments, along with FCI
control enhancements that now support asynchronous joint-position updates and
full access to datasheet-level joint limits for smoother, more compliant
control workflows. Earlier updates added a Desk API for programmatically
administrating the robot, on-field torque sensor recalibration for long-term
accuracy, and community-driven MuJoCo support for simulation work.
Please subscribe to have unlimited access to our innovations.