Unit 5: Robot Tactile Sensors

I. Orientation

Robot tactile sensing is the measurement of physical contact between a robot and its environment. A tactile sensor converts contact-related quantities such as force, pressure, vibration, temperature, or slip into an electrical signal that a controller can interpret. Touch sensing usually answers whether contact exists, while tactile sensing estimates where contact occurs and how strongly or dynamically it acts.

• Governing principle: Mechanical deformation, electrical resistance, capacitance, voltage, or light intensity changes when an external stimulus acts on a sensing element.
• Important variables: Normal force acts perpendicular to a surface; tangential force acts along it; pressure is force per unit area; contact location identifies where interaction occurs.
• Signal convention: Sensor output may be analog, such as 0–5 V, or digital, such as a binary contact state or an encoded array value.
• Robot functions: Tactile feedback supports grasp control, collision detection, object recognition, compliant motion, texture estimation, and slip prevention.
• Measurement conditions: Accuracy depends on sensitivity, range, hysteresis, resolution, response time, repeatability, temperature, mounting, and calibration.

II. Touch Sensors — Contact Detection

A touch sensor is a device that detects the presence or absence of physical contact, normally producing a binary or thresholded output. It is commonly used in robot bumpers, end-effectors, limit switches, and collision-protection systems.

A. Touch sensors

Touch sensors provide the simplest form of tactile information: a contact event. The controller compares the sensor signal with a threshold and changes the robot action when contact is detected.

• Mechanical operation: A normally open microswitch closes when a bumper is pressed, changing the circuit from an open state to a conducting state.
• Electrical operation: A conductive rubber or force-sensitive resistor can reduce its resistance when pressed; the controller classifies contact when the voltage crosses a chosen threshold.
• Binary model: If (V_s) is sensor voltage and (V_T) is the contact threshold, the decision can be written as:

contact = 1, if V_s >= V_T
contact = 0, if V_s < V_T

(V_s) is the measured voltage in volts, and (V_T) is the experimentally selected threshold.

• Typical applications: A mobile robot may stop its drive motors when a front bumper switch closes; a gripper may confirm that an object has reached a mechanical stop.
• Limitations: A basic switch does not measure force magnitude, contact position, surface texture, or slip. Mechanical contacts may also bounce, producing several rapid transitions from one physical press.

III. Tactile Sensors — Force and Contact Measurement

A tactile sensor measures contact more completely than a touch sensor by providing information about force distribution, pressure, deformation, or contact dynamics. It may contain one sensing element or many closely spaced elements.

A. Tactile sensors

The purpose of a tactile sensor is to transform a local mechanical interaction into a measurable signal related to contact force or pressure.

• Transduction principle: A sensor converts an input quantity into another domain. For example, a piezoresistive element changes resistance under force, while a capacitive element changes capacitance as its plates move.
• Pressure relation: For approximately uniform loading, pressure (p) is:

p = F / A

(p) is pressure in pascals, (F) is normal force in newtons, and (A) is the loaded area in square metres.

• Piezoresistive sensing: Force-sensitive resistors and conductive elastomers commonly show decreasing resistance as compression increases. Their response is convenient but often nonlinear and subject to hysteresis.
• Capacitive sensing: If electrode separation changes, capacitance varies approximately as:

C = εA / d

(C) is capacitance, (\epsilon) is effective permittivity, (A) is electrode overlap area, and (d) is electrode separation.

• Piezoelectric sensing: A piezoelectric material generates charge when mechanically stressed. It is highly suitable for changing forces and vibration but is poor for measuring a constant static force because charge leaks over time.
• Optical sensing: Deformation can interrupt or modulate light between an emitter and detector. Optical methods provide electrical isolation and can tolerate electromagnetic interference.
• Performance terms: Sensitivity is output change per input change; resolution is the smallest distinguishable input change; hysteresis is the difference in output for increasing and decreasing force at the same input.

IV. Interfacing of tactile sensors — Signal Conditioning and Acquisition

Interfacing connects the tactile sensing element to a robot controller while converting weak, variable, or noisy sensor signals into usable digital information. The interface includes wiring, excitation, conditioning, analog-to-digital conversion, and software interpretation.

A. Interfacing of tactile sensors

A reliable interface must preserve the relationship between physical contact and measured data while protecting the controller from noise, excessive voltage, and electrical loading.

• Voltage-divider interface: For a resistive tactile element (R_s) and fixed resistor (R_f), the output can be formed as:

V_out = V_cc × R_f / (R_s + R_f)

(V{out}) is the measured output voltage, (V{cc}) is supply voltage, and (R_s) and (R_f) are resistances. The resistor choice determines the useful sensitivity range.

• Bridge circuits: A Wheatstone bridge converts small resistance changes into a differential voltage. An instrumentation amplifier then increases the difference while rejecting common-mode interference.
• Capacitive interface: Capacitance-to-digital converters, charge-transfer circuits, or oscillator-frequency measurement can detect small changes in capacitance. Shielding and a stable reference are important because parasitic capacitance affects the reading.
• Filtering: A low-pass filter reduces high-frequency electrical noise for slowly varying pressure. A moving average may be implemented as:

y[k] = (x[k] + x[k-1] + ... + x[k-N+1]) / N

(x[k]) is the current sample, (y[k]) is filtered output, and (N) is the number of samples.

• Sampling: The sampling frequency must exceed twice the highest signal frequency of interest. For slip vibration near 100 Hz, a sampling rate substantially above 200 samples/s is required in practice to allow filtering and phase margin.
• Calibration: A known load is applied, the raw output is recorded, and a calibration curve maps signal to force. A linear approximation is (F = aV+b), where (F) is force, (V) is voltage, and (a,b) are fitted constants.
• Protection and reliability: Series resistors, input clamps, grounding, debouncing, and cable strain relief help prevent damage and false contact events. Software should also detect sensor disconnection or saturation.

V. Tactile array sensors — Spatial Pressure Mapping

A tactile array sensor combines many sensing elements, called taxels, in a two-dimensional or three-dimensional arrangement. It produces a spatial map of contact rather than a single measurement.

A. Tactile array sensors

The key function of a tactile array is to estimate the distribution and location of contact across a robot finger, palm, foot, or skin-like surface.

• Taxel definition: A taxel is an individual tactile sensing cell. An array with (m) rows and (n) columns contains (m \times n) nominal sensing locations.
• Matrix addressing: Rows and columns can be scanned sequentially to reduce wiring. The controller activates one row, reads the column signals, and repeats for all rows.
• Array representation: Measurements can be stored as a pressure matrix:

P = [p_ij],  i = 1...m,  j = 1...n

(p_{ij}) is the pressure measured by the taxel at row (i), column (j).

• Contact localization: A simple centre-of-pressure estimate is:

x_c = Σ(p_ij x_ij) / Σp_ij
y_c = Σ(p_ij y_ij) / Σp_ij

(x{ij},y{ij}) are taxel coordinates, and (x_c,y_c) locate the pressure-weighted contact centre.

• Ghosting and crosstalk: In multiplexed resistive matrices, current may follow unintended paths and create false readings. Diodes, isolation, suitable scanning sequences, or per-taxel electronics reduce this problem.
• Spatial resolution: Smaller taxels distinguish finer contact patterns but increase wiring, processing load, cost, and sensitivity to manufacturing variation.
• Robot uses: A gripper array can recognize whether an object is centred, detect uneven loading, estimate object orientation, and identify loss of contact on one side.
• Limitation: Array readings are affected by elastomer thickness, curvature, temperature, taxel-to-taxel variation, and nonuniform force. Calibration may be required for each taxel.

VI. Dynamic tactile sensor — Time-Varying Contact

A dynamic tactile sensor is designed to detect changes in contact force, vibration, impact, or slip over time. Its output depends strongly on the temporal variation of the stimulus rather than only on the steady-state load.

A. Dynamic tactile sensor

Dynamic tactile sensing supplies information about events that a static pressure measurement may miss, especially the onset of slip or a short collision impulse.

• Dynamic response: A sensor's response is characterized by rise time, settling time, bandwidth, and frequency response. A fast sensor must respond before the robot loses control of the contact.
• Piezoelectric example: A piezoelectric element produces a signal proportional to changing mechanical stress. A sudden impact creates a short voltage pulse, while a constant force may produce little sustained output.
• Accelerometer or vibration element: High-frequency vibrations generated when two surfaces begin sliding can indicate incipient slip before visible object motion occurs.
• Static versus dynamic sensing:
  1. Static sensing: Measures slowly varying pressure or force, such as the average grip load holding an object.
  2. Dynamic sensing: Measures transitions and fluctuations, such as an impact peak or vibration caused by slip.
• Time-domain features: Useful features include peak amplitude, pulse duration, rise time, root-mean-square vibration, and event count. For sampled signal (x[k]), RMS over (N) samples is:

x_RMS = sqrt((1/N) × Σ(x[k]^2))

(x_{RMS}) is root-mean-square signal magnitude, (x[k]) is the sampled sensor value, and (N) is the window length.

• Frequency-domain analysis: A Fourier transform or short-time Fourier transform can reveal vibration bands associated with sliding. Windowing permits detection of transient events while preserving their timing.
• Control response: When dynamic tactile data indicates slip, a controller can increase normal grip force gradually; when an impact is detected, it can stop or retract the manipulator.
• Limitations: Dynamic sensors may drift, saturate during large impacts, or respond to motor vibration and cable movement. Mechanical mounting, filtering, synchronization, and threshold selection are therefore essential.