Linkerbot adds fingertip touch sensing to its robotic hands
Linkerbot's September 27th post lists a 320 x 240 grid, 0.05 mm spatial resolution and 0.1 N force sensitivity; it reports no manipulation benchmarks.
By RuntimeWire Staff · Published
Primary source: Linkerbot on X
Why it matters
Tactile sensing could give Linkerbot's hands contact information that cameras do not capture reliably. The announcement supplies sensor specifications, while leaving the central commercial question open: whether those measurements improve repeatable manipulation in real tasks.
Linkerbot, founded by Zhou Yong, described vision-based tactile sensing for its LinkerHand robotic fingers in a September 27th post on X. Linkerbot says each fingertip senses at 320 x 240 resolution, with 0.05 mm spatial resolution and 0.1 N force sensitivity. The post also says the hand has more than 380,000 sensing points. These specifications describe the sensor; the post provides no evidence that a robot can reliably pick up, manipulate or release objects.
Zhou has long argued that robotic hands are a central obstacle to useful physical AI. In an April interview with Pandaily, he called the hand "the last great hardware problem in robotics." Linkerbot has focused on dexterous hands rather than building an entire humanoid, supplying hardware to different robot makers and applications while concentrating its engineering effort on the hand.
What the numbers describe
The figures in the X post are Linkerbot's claims. Linkerbot's available materials describe a wider sensing setup that can include fingertip cameras, cameras in the palm and wrist, and optional tactile sensing. The new figures add detail about the tactile layer, which can register local contact and force information where ordinary camera views alone may not reveal exactly how a fingertip is pressing an object.
Resolution is only one measure of whether a tactile sensor helps a robot handle real objects. The post gives no grasp-success rates, slip-detection results, tests across different materials, sensor durability, or calibration stability over repeated use. It also does not show how the tactile readings feed into control software or whether they improve performance on tasks beyond a demonstration. Those measures would help buyers distinguish a finer sensing grid from a hand that performs better outside a controlled setup.
Tactile sensing is an active area of robotics development. GelSight develops optical tactile sensors, while XELA Robotics markets distributed tactile sensors integrated into robotic hands. Linkerbot's post gives headline numbers for its approach; it does not compare them with competing systems under a shared test.
Zhou's bet on the hand
Zhou's route to Linkerbot was unusual for a robotics founder. Pandaily's profile reports that he entered Huazhong University of Science and Technology's Youth Class at 14 and later built an internet company that, by his account, reached about 300 million users, almost all outside mainland China. He told the publication that work on robotics began around 2018 or 2019; Linkerbot was founded in 2023.
Linkerbot's product work spans more than hand mechanics. The company sells dexterous hands and describes work on teleoperation, data collection and manipulation software. A hand that senses contact can provide information about how an object responds during a task. Turning that information into repeatable manipulation skills depends on the sensor, the hand's control system and data gathered across real tasks. The X post establishes none of those results, but the sensing claim sits alongside Linkerbot's broader effort to connect hardware with manipulation data.
Linkerbot frames tactile feedback as another input for its LinkerHand line. The practical test is whether it helps the hand adjust its grip or complete tasks that vision alone cannot handle reliably, and whether those gains hold across different objects and repeated use. Until Linkerbot publishes task-level results, the 320 x 240 grid and force-sensitivity figure describe what the sensor is designed to measure. They do not show how much more capable the robot becomes.
For Zhou, the product advances the bet he made when founding Linkerbot: that better robotic hands can help make physical AI useful beyond staged demos. The announcement adds a measurement layer to that bet. Its value will be clearer when Linkerbot connects the measurements to demonstrated manipulation performance.