AI-guided laser carves micro-pyramids to give robots a touch so gentle that can hold a jelly without crushing it

Robots can lift car engines with brute force, yet they routinely crush a soft berry or drop a slippery glass because they cannot feel the delicate boundary between a secure grip and destructive pressure. In both surgical robotics and wearable diagnostics, pressure sensors face an enduring dilemma: devices sensitive enough to register a faint arterial pulse usually max out under moderate contact, while sensors built for heavy loads fail to detect gentle touch.

Crafting rubbery surfaces with microscopic pyramids can resolve this compromise by funneling mechanical forces through sharp tips before spreading across the pyramid walls as pressure increases. However, sculpting these delicate three-dimensional microstructures has long been a manufacturing bottleneck. Conventional laser carving operates on a flat plane; as the beam cuts deeper into transparent silicone, the light defocuses and loses intensity, creating jagged steps rather than crisp micro-pyramids. Chemical cleanroom etching avoids this optical blurring, but it relies on rigid silicon templates, toxic reagents, and multi-week processing cycles that stall rapid customization.

Publishing in the International Journal of Extreme Manufacturing, Prof. Chunjin Wang at the Hong Kong Polytechnic University and his co-workers paired an ultrafast picosecond laser with a recurrent neural network that acts as an automated optical sculptor to address this bottleneck.

Instead of relying on guesswork or exhaustive trial-and-error runs, the algorithm calculates the exact scanning speeds and focal adjustments needed to carve target microstructures within 5% of their intended dimensions. Operating like a high-speed autofocus camera lens synchronized with agile steering mirrors, the optical unit shifts its focal plane downward layer-by-layer as material vaporizes away, keeping the cutting beam sharp and uniform from apex to base.

Once the silicone pyramid arrays are carved, the team coats them with an atomized mist of carbon nanoparticles and water-soluble polymer, producing a flexible conductive skin filled with tiny porous channels.

The resulting piezoresistive skin bridges extreme sensitivity with a remarkably wide operating window. It registers faint touches down to 5.8 Pa, equivalent to the weight of a single snowflake resting on skin, while maintaining a high linear sensitivity of 104 kPa⁻¹ up to 400 kPa (R² = 0.996), an upper limit matching the gripping force of heavy industrial machinery.

This performance yields a linear sensing factor (LSF) of 41,600, outperforming conventional square-frustum designs by more than a hundredfold. The device responds in just 21 milliseconds, nearly twice as fast as natural human skin receptors (30 to 50 milliseconds), and sustains over 40,000 compression cycles without physical degradation or baseline drift.

In proof-of-concept trials, a robotic gripper equipped with a single sensor performed closed-loop handling of soft, slippery gelatin cubes (~120 grams), dynamically calculating contact stiffness and fine-tuning its grip in milliseconds to prevent slippage without puncturing the fragile surface. Worn directly on the human wrist, the sensor resolved fine cardiovascular pulse waveforms, including the distinct percussion, tidal, and diastolic peaks, and tracked respiratory cycles across resting and exertion states.

Because the neural-guided laser technique cuts with equal fidelity across flexible polymers as well as hard ceramics like silicon carbide and tungsten carbide, it offers an agile and mold-free blueprint for surface micro-manufacturing. Moving forward, translating this laboratory advance into commercial manufacturing will require scaling up dynamic laser ablation for continuous roll-to-roll production lines and assessing how these micro-structured sensor arrays hold up against twisting shear forces, sweat, and variable environmental humidity over years of daily use.

International Journal of Extreme Manufacturing (IJEM, IF: 25.1) is devoted to publishing articles of the highest quality and significance to pushing the limits of scales, precision, performance and environments in manufacturing.

• Maintained #1 in Engineering, Manufacturing for 4 consecutive years
• Average time to First Decision after Peer Review: 34 days
• Open Access Publishing 

Visit our webpage, like us on Facebook, and follow us on Twitter and LinkedIn.