Targeted Feature Recognition Using Mechanical Spatial Filtering with a Low-Cost Compliant Strain Sensor.

A tactile sensing architecture is presented for detection of surface features that have a particular target size, and the concept is demonstrated with a braille pattern. The approach is akin to an inverse of mechanical profilometry. The sensing structure is constructed by suspending a stretchable strain-sensing membrane over a cavity. The structure is moved over the surface, and a signal is generated through mechanical spatial filtering if a feature is small enough to penetrate into the cavity. This simple design is tailorable and can be realized by standard machining or 3D printing. Images of target features can be produced with even a low-cost compliant sensor. In this work a disposable elastomeric piezoresistive strain sensor was used over a cylindrical "finger" part with a groove having a width corresponding to the braille dot size. A model was developed to help understand the working principle and guide finger design, revealing amplification when the cavity matches the feature size. The new sensing concept has the advantages of being easily reconfigured for a variety of sensing problems and retrofitted to a wide range of robotic hands, as well as compatibility with many compliant sensor types.