1.

2.
3.

1.

2.

3.

Capacitive Sensors

Strong hand-eye
Good geometric
coordination
accuracy
Dexterous (at
Stable and
human scale)
untiring
Flexible and
May be sterilized
Three ways to change the capacity:
adaptable
Resistant to
variation of distance between the plates (d),
Can
integrate
radiation and
variation of the shared area of the plates (A)
extensive and
infection
Variation of the dielectric constant (K).
diverse
Can use diverse
Capacitive sensor:
information
sensors for control
Pressure Profile
Quantum
Good judgment
(e.g. chemical,
Systems (PPS)
Tunneling
Easy to instruct
force, acoustic,
Composite (QTC)
and debrief
etc.)
Limited dexterity
Poor judgment
outside natural
Limited dexterity
scale
and hand-eye
Prone to tremor
coordination
Quantum Tunneling Composite (QTC)
and fatigue
Limited to
smart flexible polymer with extraordinary
Limited
relatively simple
electrical properties used for pressure
geometric
procedures
switching and sensing.
accuracy
Technology influx
Composite material from metal filler particles
Limited ability to
Difficult to
combined with an elastomeric binder,
use quantitative
construct and
typically silicone rubber.
information
debug
In unstressed state the QTC material is a nearSusceptible to
perfect insulator; with any form of
radiation and
deformation material starts to conduct and infection
with sufficient pressure metallic
2.
Human-Centered Robotics
conductivity levels can be achieved.
3.
Prosthetics
Tactile Sensing: Biological &
Engineered
Tactile Mechanoreceptors: Receptive field can
Resistor Sensor:
be defined as the area of the coverage.
CyberGlove Resistive Bend-Sensing
Type I : Small
Type: Huge area
Technology
Area

Joint angle change:

Change in resistance and amplified voltage

ma = kx
where: k = spring constant in N/m; x =
spring extension in m; m = mass in kg; a =
2
Simplest type of displacement sensor involves acceleration in m/s
2
displacement moving wiper of potentiometer. a = (k/m)*xand 1 g = 9.8 m/s
Without
acceleration
Converts linear or angular motion into a
acceleratio
changing resistance
n
Problems: mechanical wear, friction in the
wiper action, limited resolution in the wirewound units and high electronic noise.
Wire is wound around a form, making a wirewound resistor with fixed resistance, R,
between its endpoints, 1 and 2. Wiper
4. Thermal Sensors
assembly is connected in such a way that

motion of an arm causes the wiper to slide

across the wire-wound turns of the fixed
resistor. An electrical connection is made to
this wiper. Therefore, as the arm moves back
and forth, the resistance between the wiper
connection, 3, and either fixed resistor
connection will change in proportion to the
motion.
The
potentiometric sensor is Thermal conductivity is a physical property
simply a three terminal of a material that characterizes the ability of
variable resistor. The that substance to transfer heat.
resolution of this
Q kAdT / dx
sensor is limited to
Q is a measure of the heat flow
the distance, x,
A cross sectional area
between individual
dT / dx temperatu re/thickne ss gradient
turns of wire, with the resulting resistance
change of the single turn.
k thermal conductivi ty
1turn/length = x/work piece motion length
Inductive Sensing: Every new position of the
core produces a different inductance.

Fast Adapting

Slow Adapting

Variable-Reluctance Sensors: two secondary

coils are wired in series opposition, then the
two voltages will subtract

Perform gesture -> process -> capture data

2. Tactile Sensor
Tactile sensing: process of determining
Strain and Stress
physical properties and events through
Effect of applied force is referred to as stress,
contact with objects in the world.
and the resulting deformation as a strain.
Tactile sensor: a device or system that can
Responds at
Responds
measure a given property of an object or
the start of
throughout the
contact event through physical contact
contact and at
contact
between the sensor and the object
the end of
SAI is responsible
Exteroceptors: force sensors
contact
for edge
and on-off contact sensors for
FAI is
detection,
Compressional/Tensile Stress = F / A
sensing interactions on hand
responsible for
SAII is sensitive to
[N/m2]
object environment.
slippage and
stretch of the skin.
Compressional/Tensile Strain = l / l
I. Intrinsic Sensor: Three-axial strain gauge
texture.
[unitless
force
. FAII is sensitive
Stress-Strain Curve
II. Extrinsic Sensor: Distributed contact
to vibrations
sensors
Contact Sensors: sensors in our fingertips can
activate with a stimulus of 5 mN.
3-axis Force Sensors: Three metal-based strain
gauges were attached to obtain quantitative
data on the Fx, Fy and Fz.
Proprioceptors: sensing joint positions and
Modulus of elasticity, E = stress / strain
movements. Using Hall effect principle.
= (F / A) / (l / l)
Hall effect is the production
Strain Gauge Principles: R0 = l0 / A0
of a voltage difference in a
V = l0 A0 = (l0 + l)(A0 - A)
current carrying conductor
R = (l0 + l) / (A0 - A)
perpendicular both current
R = (l0 / A0) (1 + 2 l/ l0)
and the magnetic field.
R ~ 2R0 (l/ l0)
Hall effect sensor is a transducer that varies Lifting Object:
Temperature Effects:
its output voltage in response to changes in
4 mechanoreceptor:1)load force is the force
R(T) = R(T0)[1 + 0T] or R = R00T
magnetic field. In its simplest form, the sensor applied to move the load in a vertical
Sg -> ~ 0.004/C.
operates as an analogue transducer, directly manner.2)Grip force is the force applied to grip Gauge Factor:
returning a voltage. With a known magnetic
the object. Humans control this force such that GF = (R / R) / (l / l)
field, its distance from the Hall plate can be
we increase the force if the object is slipping metal SG, GF is always close to 2.
determined. Using groups of sensors, the
and we reduce this force if the object is being Some special alloys, carbon gauges, GF may
relative position of the magnet can be
crushed.3)Position is the distance of the object large as 10.
deduced.
relative to the table.
Metal Strain Gauges: Gauge sensitivity is
Cable Tension Sensor: Provides information on Requirement of a bio-inspired tactile system,1) often made unidirectional
tension of cable driven by motor that actuates system to determine the contact and release
the PIP and DIP joints (first joint and second
between the object and the fingertip.2)
joint from fingertip)
determine the contact between the object and
Design gap: Sensor density, spatial resolution, the table. 3) determine the slippage between
location and bandwidth. Lacking
the object and the fingertip.
comprehensive theory that is why human
Multi-parametric Design Requirements
MSG Signal Conditioning: (bridge circuit)
hand model is used
Skin Materials: mechanical behavior for
1) Small, fractional changes in resistance that
Architecture:
sensing, grasping or social functions (e.g.
require carefully designed resistance
softness and skin warmth for a handshake
measurement circuits. Good SG system might
scenario), appearance (color, texture).
require a resolution of 2 m/mm strain.
Skin Geometry: external anthropomorphic
2) Provide some compensation for
dimensions and skin thickness, internal or
temperature effects to eliminate masking
external ridges (fingerprints).
changes in strain.
Sensor Type: resistive, conductive,
For, R1 = R2 = RD = R
piezoelectric, piezoresistive, capacitive,
SG-> RA = R [1 + (R / R)]
optical, semiconductor, magnetic and strain
V = Vs [(RD / (RD + R1)) (RA / (RA + R2))]
gauge.
V ~ (-Vs / 4)(R / R) for (R / R) < 1.
Sensor Placement: within the structure
(intrinsic), near-surface, at the bone surface, GF = (R / R) / (l / l)
V = (-Vs / 4) GF (l / l)
anywhere within skin medium or near nail
Skin and Sensor Integration: assembly, spatial
Load Cells: measure deformations produced
resolution, bandwidth, activation threshold,
by the force of weight.
robustness, biomimetic sensing behavior of
compressi
compressive
the system (e.g. contact detection, slippage,
ve
and tensile
texture, shape, force vectors), etc.
Popular topic for Tactile Sensor:
loading.
loading
Other
important
considerations
:
wiring,
ease
1.
Computer Aided Surgery
of manufacturing, maintenance, product cost.
Industrial System
Da Vinci System
Embedded Sensing: To restore lost sensing
3. Mechanical Sensors
Load cells for low loads

Surgery Use
Humans

Potentiometric Sensors
Robots

1)Melting or deterioration of the

semiconductor up to 300C, low temp High
Resistance (-50 C to -100 C)
2) Deterioration of the encapsulation material
3) Encapsulation is typically in plastic, epoxy
or Teflon.
Insensitivity at higher temperatures
Thermocouples: measure temperature on
the basis of thermoelectric principle.

Accelerometers: Motion sensors are

designed to measure the rate of change of
position, location, or displacement of an object
that is occurring.

Accuracy
Sensitivity
Linearity
Response
Time
Signal
Robust
Cost

Solid-state temperature sensors have voltages

that vary linearly with temperature over a
specified range.
Metal Resistance Versus Temperature
Two basic devices used are:
1)
Resistance-temperature detector,
based on the variation of metal
resistance with temperature;
2)
Thermistor, based on the variation of
semiconductor resistance with
temperature.

Linear Approximation: approximates the

resistance vs temperature (R - T) curve over
some specified span.
Line L represents a linear approximation of
resistance versus temperature between T1
and T2

Measuring
Range

RTD

Thermistors

Thermocoupl
es

Best
Goo
d
Goo
d
Slo
w
Mos
t
Hig
h

Medium
Highest

Medium
Medium

Lowest

Fair

Fast

Fast

Least

Low

Low

Average

Narro
wer

Narrowes
t

Others:
RTD: Greater repeatability; Require external
power source; At some temperatures, the
reference voltage can actually heat the sensor
and throw it off.
Thermistors: Lowest stability
Thermocouple: Wide variety of measuring
ranges, including very high limits; Many
physical sizes and configurations; Tiny
measuring point; Very simple configuration
Specific types have to have matching cable
(e.g., type K thermocouple has to have type K
cable.);
Thermal Imaging: technique for converting a
scenes thermal radiation pattern, which is
invisible to the human eye, into a visible
image.
Conduction direct transfer of heat through
solid matter
Radiation the direct transfer of heat across
space.
Convection arises from transfer of mass
arising from differences in density which
depends on temperature

5. Optical Sensors
Electromagnetic (EM) radiation is a form of
energy that is always in motion it propagates
through space.
An object releases, or emits, such radiation
loses energy.
One that absorbs radiation gains energy.
EM radiation propagates through a vacuum at
a constant speed independent of both the
wavelength and frequency. In this case, the
velocity is:

When such radiation moves through a nonvacuum environment, the propagation velocity
is reduced to a value less than c. In general,
the new velocity is indicated by the index of
refraction of the medium. The index of
refraction is a ratio defined by:

Quadratic approx.:

Resistance-Temperature Detectors (RTD)

measure temperature by correlating the
resistance of the RTD element with
temperature.
RTD
RTD
RTD
Elements
Surface
Probes
Elements
Simplest,
Special
Rugged,
used in
Type, As
protect by
limited
thin as
sheath
space
possible
good
measureme
nt

:
The actual energy of one photon is related to
the frequency by

Intensity:

for Round Area

RTD has a response time of 0.5 to 5 s or more.
Thermistor: temperature-sensing element
composed of semiconductor material which
exhibits a large change in resistance
proportional to a small change in temperature.
The resistance of the thermistor decreases as
the temperature increases.
Thermistor
Linear
Thermistor
Elements
Response
Probes
TE
Simplest,
Linearise,
Industrial
compact.
consist of
Environmen
Nonlinear so
TE for
t
need instru
sensing and
to read
resistor for
linearise

The Range Limitation for thermomister:

Organization of a machine vision system: The

frame store is a digital memory with circuitry
to digitize video signals from a camera, display
Computer Vision
images on a monitor, and transfer data
between the frame store and the computer's
main memory.
4 Important Field in software Design: 1.
imaging, 2.Image processing, 3. Pattern
recognition, and 4. Scene analysis
Learning Feature Hierarchy
Learn hierarchy
Binary Images Analysis: (two-valued;
All the way from pixels -> classifier
black/white) images gives better efficiency in One layer extracts features from output of
acquiring, storage, processing and
previous layer
transmission.
Train all layers jointly
Morphological Operators: set-theoretical
Learning Features Approach
approach to image analysis, based on
Supervised Learning
shape.Signals are locally compared with
- End-to-end learning of deep
structuring elements S of arbitrary shape with
architectures (e.g., deep neural networks)
a reference point R
with back-propagation
- Works well when the amounts of labels is
large
- Structure of the model is important (e.g.
convolutional structure)
Unsupervised Learning
Photoconductive Detector
Basic idea: . Scan the image with structuring
- Learn statistical structure or dependencies
(Photoresistor): change in conductivity of a element (SE)
of the data from unlabeled data
semiconductor material with radiation
2. Perform set operations (intersection, union)
- Layer-wise training
intensity.
of image content with structuring element
- Useful when the amount of labels is not
Two basic operations: Dilation and Erosion
large
Several important combinations: Opening,
Closing and Boundary extraction
Dilation(XOR) (A B) - dilation of A by B is
the set of all displacements z, such that (B)z
and A overlap by at least one element.(B)z is
the mirrored version of B, shifted by z.
Effects: (If current pixel z is foreground, set all
pixels under (B)z to foreground)
1)
Expand connected components
Photodiode Detector: used as a variable
2)
Grow features
resistance device controlled by the light
3)
Fill holes
incident on it
Convolutional Neural Networks
Erosion
- erosion of A by B is the set of all
IR Sensors:
Fully
Locally
displacements z, such that (B)z is entirely
Connected
Connected Layer
contained in A.
Layer
Effects: (If not every pixel under (B)z is
foreground, set the current pixel z to
background.)
1)
Erode connected components
2)
Shrink features
3)
Remove bridges, branches, noise
Opening - Sequence of Erosion and Dilation

In (a), IR radiation from the emitter, e, strikes

a dark, highly absorptive object. No signal
returns to detector, d, and the sensor reports
that no obstacle is present.
In (b), the bright reflective object is detected
despite the fact that it is more distant than the
dark object in (a) that was not detected.
Effect: Remove small objects, keep original
In (c), a nearby reflective object in (a) that was shape.
not detected because the object directs most Closing - Sequence of Erosion and Dilation
of the radiation away from the detector
Effect: Fill holes, keep original shape.

The emitter, e, projects a spot of IR light on

the object whose range is to be determined. A
lens in the detector, d, casts an image of the
emitter created spot on internal photosensitive
Pros of Binary Images
components of the detector.
Can be fast to compute, easy to store
These components are sensitive to the
Simple processing techniques available
position (and hence the angle) of the spot's
image. The angle is interpreted as a range. In Lead to some useful compact shape
descriptors
(a), the small angle far, implies a long
Cons of Binary Images
distance while the large angle near in (b),
corresponds to a short distance.The output of Hard to get clean silhouettes, noise common
Can be too coarse of a representation
the sensor may not be linear over the entire
Not 3D
range and the values reported may be
unreliable when objects are too near or too far Applications
Quick Response Code and Medical Image
away.
Processing
Optical Encoders: encoder is a device that
provides a digital output as a result of a linear
or angular displacement. (Position)
1. Incremental: beam of light passes through
slots in a disc and is detected by a suitable
light sensor.
Stereo Vision: determining the 3D location of
When the disc is rotated, a pulsed output is
objects in a scene by comparing images of two
produced by the sensor with the number of
pulses being proportional to the angle through separate cameras.
Pinhole Camera:
which the disc rotates.
Thus the angular position of the disc, and
hence the shaft rotating it, can be determined
by the number of pulses produced since some
datum position.

CASE 1: Parallel Cameras

Z-axis: optical axis (the direction the
cameras are pointing)
b: the distance between cameras
f: focal length
2. Absolute: output in the form of a binary
number of several digits representing a
particular angular position

CASE 2: Non-Parallel Cameras, Rotation

About Y-axis
Z-axis: optical axis for the left camera
Z0-axis: optical axis for the right camera
b: the distance between cameras
f: focal length
Gray Code Vs Binary: 1 bit is change only
from one number to next.
6. Machine Vision
Machine vision is the ability of a computer to
see.
A machine-vision system employs one or more
video cameras, analog-to-digital conversion
( ADC ), and digital signal processing ( DSP ).
The resulting data goes to a computer or robot Deep Learning Method
controller.
Traditional Recognition Approach
The purpose of a general machine vision
system is to produce a symbolic description of
what is being imaged.

Zero drift (DC offset) describe output variation

when input is zero
Sensitivity Drift describe sensitivity of
measurement wrt ambient condition change
Band Stop Filter

Linearity: closeness between calibration curve

and specified curve

3. Digital Fundamental
Resolution: smallest input to detect change
Hysteresis: Difference between two output
values
Systematic Errors: Error
when measuring same
value
Random Error: errors after
systematic errors is
removed
Passive Circuit: Without power source (R, L, C)
DAC:
Active Circuit: Transistors, Ic,
Inverting
Non-inverting:

ADC:

Summing:

Differential

Integrator

Differentiator

Conver
sion- Time Consequences: error if change
during conversion

Convolutional layer
if in = sin
Sample and Hold: Hold value during
conversion

Pooling Layer

Voltage follower:
To isolate
Vo =V-=V+=Vin

Bridge Circuit:

Noise and Interferences

Intrinsic Noise: unwanted noise generated

within the circuit elements
1. White noise: power spectral density over a
wide range of frequencies
2. Thermal Noise:
Extrinsic Noise: Unwanted noise from external
sources.
Capacitive Coupling
Inductive

Motion Field vs. Optical FlowThe apparent motion of brightness

patterns observed when a camera is
moving relative to the objects being
imaged is called the optical flow.Optical
flow can be totally different from
motion field, which depends on the
projection of moving objects on the
image place

Measurement of Noise:

(IN)

(EX)
Noise Reduction:
Instrinsic: Limit Bandwidth, raise input signal
and select low now input amplifier.
Extrinsic: Supress at source(screening),
earthing(balance) and making receiving circuit
less sensitive.
Lead Compensation: Equal length on each side Grounding: Improper grounding causes
of the arm to have effective offset
ground loop and makes measurement
susceptible to noise.
Signal Filtering:
Fix: Thicker cable(reduce voltage drop), single
earth(avoid sharing same ground paths)
Design issue
separate load current return path from signal
Machine vision systems require mixture
reference ground, remove ground loops that
of technologies to build complete
cause inductive coupling, large distance
system. (Mechatronic)
between measurement circuits to reduce
The optical environment must be
mutual inductances.
controlled to make the image
Ground Loops: Rg should be 0, improper
processing easier, i.e. control the
grounding cause it to float:
Low
Pass
Filter
lighting against gross variations, make
judicious use of lenses, mirrors, a
regulated power supply for the lamp
and fiber-optic light guides.
Applications
Shielding: a conductor or circuit consists of
Eliminate defects- inspect 100% of
wholly enclosing by an electrically conductive
products, and ensure that shipments are
material connected to a constant voltage.
defect free
Provides high impedance barrier between
Verify assembly - reduce waste by
noise and measurement system
detecting process errors early.
Electric shielding: reduces infiltration of noises
Automate production - more efficient use High Pass Filter
by localizing noise source to confined area
of manufacturing lines by giving production
(metal case)
equipment the ability to automatically locate
Magnetic shielding by using magnetic material
and identify items.
of high permeability (low freq), thick steel
Track and identify parts - automate
shield(high freq)
tracking of parts and products by reading 1D
Cabling:Largest source of interference,
and 2D identifying codes enables greater
improper cabling causes signal reflection
customization in manufacturing, and verify
which give rise to ringing effect. Also cause
warranties, reduce risk in event of product
standing waves to appear that reduces or
recall.
cancels out original signal
Part 1:
Band Pass Filter
Accuracy: capacity of measuring instrument
to give result close to true value.

Twisted pair: keeps signal loop area small,

cancel out magnetic interference, bandwidth
of few MHz, less prone to interference.
Twin Feeder, used for VHF signal up to 30MHz,
300ohm
Coaxial cable: common for experiment,
Multi-way cable: Connect other wire to ground
to prevent electrostatic
Optical coupler and optical fibers: good
electric isolation, immunity to interference
Communication systems:
Current Telemetry: 4-20mA
Insensitive to voltage drop,
HART: Current loop point to point. New sensor,
new cabling. 4-20mA and 2 way
transmission,up to 15 devices
Fieldbus: Analog Systems, 2 way multidrop,
serves as lan for process control and I/O, new
sensor = software only(fast)
Intelligent Smart sensor = combine sensor and
microprocessor
Frequency Response