A single force-torque sensor at a robot's wrist only ever returns one aggregate number: the total load it is carrying. It cannot tell you whether that load came from one sharp point or was spread evenly over a wide area — a firm poke and a gentle full-hand press can sum to the exact same reading.
- Distributed tactile array — hundreds of individually-addressable piezoresistive/capacitive taxels tiled across the skin each report their own local pressure, hundreds of times a second. Together they form a live 2D pressure-image of the contact patch, the same way a camera sensor forms an image from individual photosites.
- Shape from the pressure map — a fingertip leaves a small rounded high-pressure blob; a palm leaves a broad, low-pressure ellipse; a stylus leaves a tiny, very intense point; an edge leaves a thin elongated ridge. A simple classifier reads contact area, peak pressure and the footprint's aspect ratio straight off the taxel image to tell them apart — no prior knowledge of what is touching is required.
- Single-point sensor — toggle it off to see the same contact reduced to one blind number. The spatial pattern — and with it, the ability to localize contact or recognize what is touching — disappears entirely.