Force-torque sensor

A surgeon stitching a wound does not stare at a ruler. She feels, through her fingers and the needle driver, how much tension the suture can bear before tearing. She adjusts her grip, her pull, her angle — all in real time, guided by sensation rather than calculation. A robot performing the same task cannot see this information in a camera feed; it has to feel it.

The instrument that gives robots this sense of feel is the force-torque sensor.

What it measures

A force-torque sensor (often abbreviated F/T sensor) measures all six components of mechanical loading acting at a point: three forces (push or pull along the X, Y, and Z axes) and three torques (twisting moments about each of those same axes). Together, these six numbers completely describe the mechanical interaction between a robot's tool and whatever it is touching.

Most F/T sensors use strain gauges — tiny resistors bonded to a precision-machined metal structure. When force deforms the structure, even by micrometres, the strain gauges change resistance in proportion. Electronics convert those resistance changes into the six force/torque readings, typically hundreds or thousands of times per second.

Some sensors use piezoelectric elements instead, which generate a voltage when mechanically stressed. These respond faster and handle higher loads, but drift over time and are less suited to sustained static forces.

Force control: the key capability

Knowing force unlocks force control — a robot behaviour that adjusts motion based on what it feels rather than just what it was told. The simplest version is compliance: if the robot pushes against something and the force reading exceeds a threshold, it backs off. A more sophisticated version is impedance control, where the robot's controller makes the arm behave as if it were a spring: it yields to external forces proportionally, so a human can guide it by hand without fighting the motors. This is the same principle behind cobot safety — those joints are effectively running force control continuously.

Real-world examples

Universal Robots ships an optional F/T sensor with its cobots for assembly tasks — inserting a bearing into a housing, for instance, requires sensing the exact moment the part seats correctly rather than jamming in at the wrong angle. The Intuitive Surgical da Vinci system (the dominant surgical robot worldwide) has faced persistent criticism for lacking force feedback, meaning surgeons cannot feel tissue tension; competitors like Avatera and CMR Surgical have positioned force sensing as a differentiator. In research, MIT's manipulation group and Stanford's AI Lab routinely use ATI and Robotiq F/T sensors as fundamental tools for studying how robots handle deformable objects.

The gap between feeling and understanding

An F/T sensor solves the data problem but not the interpretation problem. Getting six numbers is easy; knowing what to do with them in a complex manipulation task is still one of robotics' hardest research challenges. Teaching a robot to crack an egg — firm enough to break the shell, gentle enough not to squeeze the yolk — requires not just F/T sensing but the control logic to act on it wisely.

Could a robot ever learn to judge the ripeness of a mango by touch alone — and would that be easier or harder than learning to see it?