A flexible pressure-sensor array made of velostat and aluminum tape
A soft pneumatic actuator can grasp an object without knowing how much force it applies or where it touches. This project looked for a thin, flexible sensor that covers the gripping surface and tells where there is pressure. Two were built. The first, made of polyimide, aluminum and velostat, works and shows a pressure map, although with faults in the connector. The second, made entirely of silicone, remained a demonstration of the molding process: no silicone conductive enough for its electrodes was found.
Context
The usual sensors in soft robotics measure the pressure of a fluid chamber or deformation by optical means. They report that there is a force, but not where it is applied, and their readout electronics are bulky. A piezoresistive material allows a different strategy. Velostat is a carbon-loaded polymer sheet whose electrical resistance drops when it is compressed. It can be cut into small squares and placed at each crossing of a matrix of rows and columns.
One cell
Each cell is a square of velostat between two electrodes. It is read with a voltage divider that uses the 20 kΩ internal resistor of an Arduino Mega: the velostat resistance is solved from the measured voltage. To obtain the full map, the cells must be read one by one.
ZUCC: polyimide, aluminum and velostat
The first version was assembled to check the principle. It measures 104 × 32.5 × 1 mm. Rows and columns are aluminum tape, and polyimide tape (Kapton) serves as substrate and insulator.
Layers of the first version and the finished sensor
All layers are cut on a desktop vinyl cutter from drawings exported from CAD. Each layer is lifted with transfer tape and aligned with the previous one using the edges as a guide. A laser cut opens the windows through which the rows contact the output traces. The thesis documents the process step by step, with the cutting parameters.
The second version reduces the size of each cell to increase resolution and ends in traces with a 1 mm pitch, compatible with an FPC connector. Two soldered adapters take those traces to standard pins.
Columns, velostat squares, rows and output traces
SPABS: the same principle in silicone
The silicone version aims for a sensor that stretches together with the actuator. A 3D-printed mold holds the velostat squares while the silicone is poured. The cured base has channels on both sides, rows on one and columns on the other, which are then filled with conductive silicone.
Silicone base with the velostat squares and empty channels
Channels filled with silicone and graphite; the bubbles are visible
Results
ZUCC 2.0 was tested by pressing each cell with a finger, with a Processing program that draws the pressure map. The sensor responds. In some cases a neighboring cell reports pressure without being touched, and some rows or columns give no reading. Continuity tests showed that all rows and columns of the sensor conduct, so the fault is attributed to the adapters, whose tiny connectors led to manufacturing errors.
The resistance of loose velostat squares was measured with a multimeter, pressing by hand:
| Size | No pressure | At 750 Pa | At 3000 Pa |
|---|---|---|---|
| 5 × 5 mm | 9 to 13 MΩ | 740 to 820 Ω | 196 to 323 Ω |
| 1 × 1 cm | 9 to 14 MΩ | 180 to 210 Ω | 87 Ω |
| 2 × 2 cm | 9 MΩ | 78 to 83 Ω | – |
Resistance drops four orders of magnitude with very little pressure, and the larger squares give more stable readings.
Silicone mixtures with carbon fiber and with graphite
For the SPABS electrodes, silicone mixtures with chopped carbon fiber and with graphite were tested. Below 2% by mass none conducted. With 2.75% fiber, between 5.6 and 21.4 kΩ per centimeter were measured, and with 3.85% graphite, hundreds of megaohms per millimeter. More filler prevented curing or made the silicone stiff. The carbon black delivered by the supplier turned out not to be conductive.
What is missing
There is no calibration. The table comes from pressures applied by hand and the sensor map is qualitative, with no established relation between reading and force. Cell-by-cell readout is slow for controlling a grasp. The polyimide sensor bends but does not stretch, and its fabrication is manual.
The silicone sensor does not work. Besides conductivity, the available vacuum chamber, at 15 inHg, did not remove the bubbles from the channels. The author proposes inserting carbon fiber threads along the channels, using conductive carbon black, and heat- or UV-cured silicones that give more working time. Neither sensor was mounted on an actuator: the gripper molds were printed, but no silicone was poured into them.
How it fits in Robiolab
The lab works with soft pneumatic actuators, and knowing what force they apply and where is a recurring need, as shown by the characterization of the impedance between an actuator and a finger. The lab's hand prostheses also needed to measure grip force. This project contributes a manufacturing method that can be reproduced with desktop tools for thin tactile arrays, and it bounds with data the problem that remains open for a fully soft sensor: the conductive, stretchable electrode.