Sensor

1
Sensor

• Heesang Shin

2
What is sensor?

• A sensor is type of transducer
• Convert some form of energy into another form of
  energy
• Antenna
• converts electromagnetic waves into electric
  current and vice versa
• Microphone
• converts changes in air pressure into an
  electrical signal
• LCD
• Converts electrical energy into light

3
What is sensor?

• In robotics
• Acoustic
• Sonar
• Motion
• Turn coordinator
• Light
• CCD
• Etc

4
Passive / Active Sensor

• Passive
• Rely on the environment to provide the medium for
  observation
• e.g. a camera requires a certain amount of
  ambient light to produce a useable picture
• Active
• Put out energy in the environment to either
  change the energy or enhance it
• e.g. A sonar sends out sound, receives the echo,
  and measures the time of flight
• Although a camera is a passive device, a camera
  with a flash is active

5
Active Sensing

• Active sensing connotes the system for using an
  effectors to dynamically position a sensor for a
  "better look"
• Active sensor is not the same as active sensing
• A camera on a pan/tilt head with algorithms to
  direct the camera to turn to get a better view is
  using active sensing

6
Modality

• Sensors which measure the same form of energy and
  process it in similar ways form a sensor modality
• A sensor modality refers to what the raw input to
  the sensor is sound, pressure, temperature,
  light , and so on.
• A modality can be further subdivided
• e.g. vision can be decomposed into visible light,
  infrared light, X-rays

7
Logical Sensors

• Abstraction of sensors
• Introduced by Henderson and Shilcrat
• A unit of sensing or module that supplies a
  particular percept
• Similar like Abstract Data Type in C, e.g.
• There is logical sensor called range_360
• data type
• range_360 provides an object specifying the polar
  plot
• public data / method
• range_360 may uses sonar, a laser, stereo vision
• private data / method
• each different sensor may use different way to
  process signals

8
Logical Equivalence

• Sensor modules would be logically equivalent
• As long as logical sensor returns SAME percept
  data structure
• Wouldn't necessarily be equivalent in performance
  or update rate
• So they falls into same logical sensor

9
Sensing Model
10
Behavioural Sensor Fusion

• Combines information from multiple sensors into a
  single percept
• Redundant (Competing)
• Physical redundancy
• sensors are both returning the same percept
• some sensors may return better percept than the
  other in different situation
• Logically redundant
• returns identical precepts but use different
  modalities or processing algorithms
• Why? overcome false positive and false negative
• False Positive
• when a sensor leads the robot to believe that a
  percept is present, but it is not
• e.g. misses the glass wall with stereo vision
• False Negative
• when robot misses a percept
• e.g. misses thin obstacles with sonar (legs of
  chair)

11
Behavioural Sensor Fusion

• Complementary
• Complementary sensors provide disjoint types of
  information about a percept
• e.g. urban search and rescue robot
• search for survivors by fusing observations from
  a themal sensor for body heat with camera
  detecting motion
• thermal sensor or camera alone neither provides a
  complete view
• Coordinated
• Sequence of sensors
• cue-ing or providing focus-of-attention
• A predator might see motion, causing it to stop
  and examine the scene more closely for signs of
  prey

12
Sensor Fission

• each sensor reading supports specific behaviour
  which leads intermediary action
• intermediary actions combined in resulting action
• mainly competitive method

13
Action-Oriented Sensor Fusion

• sensor data is being transformed into a
  behaviour-specific representation in order to
  support a particular action
• e.g. if a cat hears a noise and sees a movement,
  it will react more strongly than if it has only a
  single stimulus
• covers competing and complementary sensing

14
Sensor Fashion

• changing sensors when circumstances changes
• covers coordinated sensing

15
Sensor Fission, Fusion and Fashion
16
Sensor Suites Reference Frames

• Proprioception
• relative to internal frame of reference
• e.g. shaft encoder
• records wheel revolution
• wheel might slip
• Exteroception
• relative to robot frame of reference
• e.g. camera
• environment supply current position
• camera may heading wrong direction

17
Sensor Suites Reference Frames

• Exproprioception
• relative to external frame of reference
• e.g. GPS
• satellites transmit positioning information
• only available on open sky environment
• Sensor Suite set of sensors for a particular
  robot
• Always have some type of exteroceptive sensor
• otherwise the robot cannot be considered
  reactive there would be no stimulus from the
  world to generate a reaction

18
Sensor Suites Sensor Attributes

• Field of view (FOV)
• exteroceptive sensor has a region of space to
  cover
• usually expressed in degrees A wide angle lens
  will often cover up to 70 degree, while a regular
  lends may only have a field of view around 27
  degree
• distance that the field extends is called the
  range
• Horizontal FOV
• Vertical FOV
• FOV and range are critical in matching a sensor
  to an application

19
Sensor Suites Sensor Attributes

• Accuracy-repeatability
• how correct the reading from the sensor is
• poor accuracy means little repeatability
• sensor resolution how precise sensor reading is
• Responsiveness
• some sensor function poorly in particular
  environments
• sonar is often unusable for navigating in an
  office foyer with large amounts of glass
• normal camera is often unusable in dark room

20
Sensor Suites Sensor Attributes

• Power consumption
• always concern for robots
• less power they consume, the longer they run
• amount of power on a mobile robot required to
  support a sensor package is sometimes called the
  hotel load
• the power needed to move the robot is called the
  locomotion load
• Hardware reliability
• physical limitations on how well they work
• e.g. Polaroid sonar will produce incorrect range
  reading when the voltage drops below 12V

21
Sensor Suites Sensor Attributes

• Size
• the size and weight of a sensor does affect the
  overall design
• Computational Complexity
• Estimation of how many operations an algorithm or
  program performs
• The BIG O problem
• Interpretation reliability
• Interpretation algorithms must be reliable
• Hard to catch errors by untrained human
• robot may "hallucinate" and do the wrong thing if
  sensor is providing incorrect information

22
Sensor Suite Attributes

• Simplicity
• simple sensors are more desirable than complex
  and require costant maintenance
• Modularity
• designer must be able to remove one sensor and/or
  its perceptual schema without impacting any other
  sensing capability
• Redundancy (Physical Logical)
• a faulty sensor can cause the robot to
  "hallucinate"
• sensor suites may offer some sort of redundancy
• Physical redundancy multiple of same sensor
• Logical redundancy another sensor using a
  different sensing modality can produce the same
  percept or releaser
• Fault tolerance
• physical redundancy introduces fault tolerance
• Robots can be programmed in most cases to
  tolerate faults as long as they can identify when
  they occur

23
Proprioceptive Sensors

• Shaft Encoder
• remembers motor activity
• only for estimate, wheel could slip on surface
• Inertial navigation systems (INS)
• accelerometers measure acceleration
• integrate readings to get velocity and position
• not suitable for jerky, bumping motion
• still expensive

24
Exproprioceptive Sensors

• Global Positioning System (GPS)
• selective availability reduces accuracy was
  problem, however
• On May 2, 2000 "Selective Availability" was
  discontinued as a result of the 1996 executive
  order, allowing users to receive a non-degraded
  signal globally.
• no more needs of position averaging and
  differential GPS technique to obtain precise
  position
• however it is still restricted to use GPS signal
  altitude above 18km (60,000 ft) and travelling
  over 515m/s (1,000 knots) out of US

25
Exproprioceptive Sensors

• Radio triangulation
• triangulation is the process of finding
  coordinates and distance to a point by
  calculating the length of one side of a triangle,
  given measurements of angles and sides of the
  triangle formed by that point and two other known
  reference points, using the law of Sines.
• GPS uses Trilateration not triangulation
• triangulation uses angles
• Trilateration uses known reference positions

26
Proximity Sensors

• Sonar / Ultrasonic
• refers to any system for using sound to measure
  range
• different applications operate at different
  frequencies
• active sensors
• time of flight time from emission to bounce
  back
• speed of sound time of flight sufficient to
  compute the range of the object
• speed of sound vary with environments
• Infrared (IR)
• emit near infrared energy and measure whether any
  significant amount of the IR light is returned
• range of inches to several feet
• often fail in practice because the light emitted
  is often "washed out" by bright ambient lighting
  or is absorbed by dark materials

27
Proximity Sensors

• Contact Sensors
• Tactile, or touch, done with bump and feeler
  sensors
• feelers or whiskers can be constructed from
  sturdy wires
• bump sensors are usually a protruding ring around
  the robot consisting of two layers
• useful to protect robot body from colliding

28
Sonar regions and problems

• Sonar (ultrasonic) sensor regions defined as 4
  areas
• Region I area that associated with the range
  reading, width of region I is tolerance
• Region II the area that is empty, if not range
  reading would have been shorter
• Region III the area that theoretically covered
  by the sonar beam
• Region IV outside of the beam and not of
  interest

29
Sonar regions and problems
30
Sonar regions and problems

• Problems
• Foreshortening
• sound broadcast in cone shape, it is possible to
  some sound may bounce back before other sound
  reaches the object
• Specular reflection
• when the wave form hits a surface at an acute
  angle and the wave bounces away from the
  transducer.
• especially glass induces serious specular
  reflection
• Cross-talk
• bounced away wave may returns to another sonar to
  give erroneous reading
• if each sonar uses different frequencies, then
  there is sophisticated aliasing techniques can be
  applied

31
Computer Vision

• This is big chapter alone (machine vision paper
  dedicated for it)
• Let's briefly through the book
• CCD cameras
• Video camera uses CCD (charged couple device)
  technology to detect visible light
• two way to "grab" image from camera
• digital camera (usb / firewire etc)
• frame grabber

32
Computer Vision

• Colour spaces
• Binary (Monochrome)
• on / off (white / black)
• Gray scale
• level of grey
• RGB
• Red / Green / Blue components (triple grey scale)
• HSI / HSV
• Hue / Saturation / Intensity (value)
• Colour histogramming
• generate total pixel of each colour used

33
Computer Vision

• Visual Erosion
• degradation in segmentation quality is called
  visual erosion
• object appears to erode with changes in lighting
• Region segmentation
• identify a region in the image with a particular
  colour
• Thresholding
• separate image with given thresholding value
• Foreground / Background
• separate interested object in foreground and
  others to background

34
Computer Vision

• Range from Vision
• Stereopsis / optic flow
• way that human percept range and depth of an
  object
• reason we have two eyes
• Stereo camera pairs
• mimic human eyes

35
Computer Vision

• Light stripers
• projecting a line, grid or pattern of dots on the
  environment
• regular vision camera observes how the pattern is
  distorted in the image
• Laser ranging
• emit light as sonar emit sound wave
• very accuracy (a range of 30 meters and an
  accuracy of a few millimetres)
• expensive
• Texture
• detects patterns or colour (e.g. carpet)
• if something standing on the carpet, it reads as
  occupied
• strong shadows could create ghost object

36
References

• R. Murphy, Introduction To AI Robotics, MIT Press
  2000.
• Slides from Dr. Chris Messom
• Wikipedia, the free encyclopedia
• J. Allocca and A. Stuart, Transducers Theory and
  Application, Reston 1984.

37
Questions?
38
Thank You!