A robot hand that knows when not to squeeze harder

A HAND THAT DOES NOT SQUEEZE BLINDLY

FORTE is not just another attempt to give a robot hand a dramatic potato-chip video. The University of Texas at Austin system targets an older, duller problem: an industrial gripper often knows where its jaws are, but not enough about what it is doing to the object. For a metal bracket, that can be acceptable. For a raspberry, a slipping jar or a thin chip, the same approach quickly becomes a crusher with good public relations.

The name stands for Fragile Object Grasping with Tactile Sensing. Under the acronym is a practical pairing of soft robotics and tactile measurement: 3D-printed fin-ray fingers bend around the object, while internal air channels change pressure as the finger deforms. Small off-the-shelf pressure sensors turn those pressure changes into force and slip signals. The important detail is that the sensor is not merely pasted onto a fingertip. It is built into the compliant structure.

In the published paper materials and project page, the researchers report force estimation from 0 to 8 N with an average error of 0.2 N and slip detection within 100 ms. In the tests described by UT, the gripper handled 31 objects: fragile items such as raspberries and potato chips, slippery items such as jam jars and billiard balls, and everyday items such as soup cans and apples. Overall single-trial grasping success reached 91.9%. The system recognized 93% of slips with 100% precision, meaning it did not unnecessarily tighten its grip because of false slip alarms.

That is a stronger argument than the chip image by itself. Robots can already repeat large motions, stack parts and push boxes through predictable paths. Delicate manipulation is a different category because the object can be soft, wet, uneven, expensive or biologically sensitive at the same time. FORTE is therefore not just a softer finger. It is an attempt to put contact data back into the control loop.

THE DEMO PASSED; DEPLOYMENT HAS NOT

The nearest use case is not a smiling humanoid unpacking groceries. The more useful path is less theatrical: food handling, lab work, medical instruments, delicate electronics and production lines where one universal rigid gripper cannot cover real object variability. UT Austin explicitly points to food processing, health care and manufacturing as areas where a more sensitive grasp could reduce breakage, waste or manual rework.

FORTE is still a research platform. The hardware manual explains why labs will care: the design is modular, uses accessible components and can be 3D-printed in different finger shapes. The same manual does not answer the questions a factory will ask first: how long the fingers survive across shifts, how the signal changes after cleaning, how calibration recovers after an impact and how easy it is to replace a damaged air channel without calling the doctoral student who built the system.

The research record also needs careful wording. The arXiv version of the paper and the linked IEEE DOI describe FORTE as a solution for delicate manipulation, not a finished general-purpose hand. The team lists next steps that include reducing sensor sensitivity to temperature changes and improving the ability to catch objects that are already slipping. Those are not footnotes. A pneumatic channel that reads deformation cleanly on a bench must keep doing so under changing temperature, humidity, dust, motion speed and production rhythm.

The useful conclusion is less spectacular and more durable. FORTE shows that compliant fingers can be sensors, not merely rubber bumpers between a robot and an object. If that approach survives calibration drift and daily maintenance, real-world robots gain a quieter but more important ability: knowing when not to squeeze harder.