Triboelectric Sensors Advance Soft Robotics and Digital Twins

Triboelectric nanogenerator (TENG) sensors are emerging as powerful tools for tactile and motion sensing in soft robotics, with applications extending into digital twin systems. Two complementary designs—length TENG (L-TENG) and tactile TENG (T-TENG)—operate in single-electrode mode, each optimized for distinct sensing tasks.

In the L-TENG configuration, a nickel-fabric coated gear serves as the positive triboelectric element, while a polytetrafluoroethylene (PTFE) film backed by a copper electrode forms the negative layer. Stretching a soft strip drives the gear into intermittent contact with the PTFE, producing electrical pulses. A spring on the rotation shaft provides self-recovery, maintaining tension and preventing buckling. This arrangement yields displacement sensing with a minimum resolution of 5 mm and an error rate near 8.3%, with potential for refinement by increasing gear teeth or reducing shaft diameter.

The T-TENG sensor employs five nickel-fabric electrodes patterned on a polyethylene terephthalate substrate, coated with silicone rubber as the negative triboelectric material. Four short electrodes (E1–E4) are spaced at 20 mm intervals, while a long electrode (EL) runs along the sensor’s length. Contact position is determined by voltage ratios between short electrodes, sliding motion is tracked via sequential peaks, and EL measures total contact area to distinguish point versus surface contact. This flexible, thin sensor integrates seamlessly with soft grippers.

Mounted on a nitrile glove, L-TENG and T-TENG sensors form a self-powered human–machine interface (HMI) capable of controlling a robotic hand. The L-TENG on the index finger encodes bending angle through peak counts, while a T-TENG segment on the thumb switches bending direction via taps. This setup enables real-time mirroring of human finger motions in the robotic counterpart, with discrete peaks corresponding to 30°, 60°, and 90° bends.

A soft actuator with a hollow-bellows structure demonstrates L-TENG’s compatibility in pneumatic systems. Corrugated upper surfaces deform more under inflation, producing bending motions quantified by L-TENG output peaks. Experiments show a near-linear relationship between air pressure and bending angle, confirming feasibility for closed-loop control.

Integrating T-TENG into a three-finger soft gripper allows detection of object contact position, contact area, and motion phases. Signals from E1–E4 vary inversely with distance to the contact point, while EL output scales with curvature-induced contact area. Tests with apples, oranges, eggs, and turnips verify the sensor’s ability to discern subtle differences in grip location and object geometry.

Machine learning (ML) enhances these sensing capabilities. A support vector machine (SVM) classifier processes multi-channel voltage data from L-TENG and T-TENG sensors to recognize gripped objects. In a six-channel configuration (one L-TENG, five T-TENG), recognition accuracy reaches 97.1% across six object types. Expanding to 15 channels—equipping all fingers with both sensors—boosts accuracy to 98.1% for 16 diverse objects. Comparisons show that more channels yield better performance, particularly for similar-shaped items, though at increased cost and complexity.

Sphere-shaped objects such as baseballs, tennis balls, apples, oranges, and tangerines highlight the value of combined sensing. Pure L-TENG data (3 channels) achieves 77.0% accuracy, pure T-TENG (12 channels) reaches 92.0%, and the full 15-channel system delivers 95.0%. T-TENG’s ability to capture contact position and area provides richer features than L-TENG’s bending angle alone.

A digital twin demonstration underscores the system’s potential in unmanned warehouses. Real-time signals from the gripper’s sensors feed into the trained SVM model, projecting recognized objects into a virtual environment without cameras. As physical objects are gripped and sorted, their digital counterparts are manipulated in sync, enabling automated sorting and monitoring in industrial settings.

These triboelectric sensors, combining mechanical ingenuity with advanced signal processing, offer a robust pathway toward intelligent, self-powered soft robotic systems. Their integration with ML and digital twin frameworks positions them as enabling technologies for next-generation automation in manufacturing, logistics, and beyond.