Exam 2 ME503

  1. What is Localization
    Determining the position & orientation of the robot in a desired reference frame
  2. Is an essential component in an autonomous system
    Mobile robot localization
  3. Localization Challenges
    • Accuracy & Precision
    • Pose not just position
    • Sensor noise
    • Error Propagation
  4. Accuracy & Precision
    How well does pose need to be known
  5. Pose not just position
    Heading, velocity, acceleration may be needed
  6. Sensor noise
    Price increases inversely with noise characteristics
  7. Error Propagation
    How is the error expected to develop
  8. There are several (not mutually exclusive) approaches to localization (4)
    • Odometry 
    • Dead reckoning
    • Active beacons
    • Landmarks
  9. What is odometer
    use the kinematic model and wheel velocity measurements to estimate pose
  10. What is dead reckoning
    use the wheel velocity and heading sensors (e.g. compasses, IMU) to estimate pose
  11. Example of a active beacon
  12. Example of landmarks
    Stereo odometry
  13. What is an encoder used for
    An encoder is used to measure angular position or velocity
  14. What is the principle of operation for an encoder
    generate pulses as the shaft rotates
  15. For an encoder the angular position is
    proportional to the number of pulses
  16. Incremental Encoders and what phase are the channels out
    • Generate pulses and use two output channels (A and B) to sense position(Two channels or 4 states). 
    • Ch A and Ch B are 90 degrees out of phase
  17. IN single ended or open collector encoders
    only two channels are used (Ch A & Ch B)
  18. In differential encoders
    At least four channels are used (Ch A, Ch A’, Ch B, Ch B’) with Ch A’ 180± out of phase with Ch A and Ch B’ 180± out of phase with Ch A.

    Channels are compared to reduce noise
  19. Absolute encoders
    Generate a unique word pattern for every position of the shaft. Number os signal lines determines the resolution.
  20. Where would you put an encoder and why
    Would put on the output shaft because of gear ratio
  21. Why use gray code
    Only one bit changes at a time. Good because only have to monitor one bit change at a time and it has fewer channels (more rigid)
  22. With odometry what causes error and where do the biggest errors come from
    errors are the result of wheel slip, propagate, leading to more uncertainty the longer the systems operates. Biggest error comes when turing so from theta
  23. Why is error propagation in dead reckoning generally less than odometry
    Due to in inclusion of heading sensors
  24. What are two different sensor types used for heading estimation
    • Magnetic compass
    • inertial sensores (Accelerometers, gyroscopes, IMU)
  25. Magnetometers
    measure strength and or direction of magnetic field
  26. compass sensor
    magnetometer that determines heading angle of a vehicle using earth's magnetic field
  27. Two types of magnetic compasses
    • fluxgate compass (large mobile robotics)
    • MEMs compass (small robotics)
  28. The strength of earth's magnetic field is the _______ vector
  29. Magnetic compass sensors experience distortion due to
    metal objects or artificial magnetic fields
  30. Compass may not always provide a
    true absolute reference
  31. Magnetic compass can still be used in the ____ reference and why
    local reference, in all situations where the external influences are constant over time
  32. Accelerometers are used to estimate and most use what type of sensors
    mobile robot position and velocity and MEMS
  33. Accelerometer readings and specification are reference to
    gravitation acceleration
  34. An accelerometer sensor detects acceleration in
    one axis, is insensitive to accelerations in orthogonal axes
  35. Accelerometers are used to measure
    • inclination (if stationary or constant speed)
    • vibration
  36. Gyroscopes are the key components of
    inertial navigation units or systems
  37. gyroscopes provide _____
    they measure _____
    inertial platforms for accelerometers and measure angular displacement and velocity
  38. most well known gyroscope is the ____ and it consists of
    fly wheel gyroscope and of a wheel whose angular momentum is such that the axis of rotation tends to remain fixed in space
  39. Vibrating structure gyroscopes use
    MEMs devices, most common today and operate on Coriolis force principle.
  40. Coriolis force principle
    to travel in a straight line, lateral velocity must be increased/decreased
  41. Laser ring gyroscope
    Highest accuracy and does not suffer from noise issues of MEMs gyroscope. Difference in wavelengths tells you the rate of rotation
  42. IMUs refer to
    3-,6-, or 9 axis units with each axis referred to as a DOF
  43. These units incorporate a combination of 3-axis
    • accelerometers
    • gyroscopes
    • magnetometers
  44. Inertial measurements are either in terms of ______ and _____. And must be
    acceleration or velocity, must be integrated in order to give positions
  45. What is an Active beacon?
    Signal must be ?
    multiple bacons are used for ?
    common types?
    • emit a signal or energy that is read by an onboard sensor. 
    • signal is interpreted to determine distance 
    • triangulation
    • IR and GPS
  46. What is GPS
    method of finding the position of an object with reference to a fixed global coordinate frame
  47. Inertial Navigation System
    Measures 6 DOF, IMU + GPS, uses sensor fusion through kaman filter (accelerometers,gyros,gps), and found in high cost applications
  48. Difference in
    • Does not accrue error over time
    • slow update rate
    • gps position jumps

    • high short term accuracy
    • fast update rate
    • smooth changes in position
  49. Examples of Landmarks (4 examples)
    • Passive buoys
    • visual odometry
    • LIDAR edge detectors
    • Star trackers
  50. Kalman filtering
    used to estimate dynamic variables in the presence of noise
  51. Kalman dynamic case components
    • process model 
    • measurement model
    • noise characteristics
  52. Process model
    • state space representation of dynamics
    • linear discrete model
  53. measurement model
    • relatives measurements to process states
    • linear discrete model
  54. noise characteristics
    contain information about uncertainty
  55. Extended kalman filter
    enables basic kaman filter equations to be applied by linearizing the system at each sample instant about the input and the state estimate
  56. Pressure sensing can provide
    altitude and depth
  57. MEMs pressure sensor use either a
    resistive or capacitive approach
  58. resistive approach
    a piezoresistive diaphragm displaces and the resulting resistance is measured with a wheatstone bridge
  59. capacitive approach
    the pressure difference changes the capacitance between electrodes
  60. Resistive sensors are :
    • small 
    • linear relationship 
    • temperature sensitive
  61. Capacitive sensors are :
    • bigger
    • lower cost
    • nonlinear 
    • good for high temp
  62. What can still occur using inertial sensors and why
    dead reckoning and do to major drift errors
  63. Mapping in autonomous robots
    assumes position and pose of the robot are known, sensors are used to locate and measure landmark positions, kaman filtering is often used to better estimate landmark positions
  64. What is SLAM
    robot is placed in an unknown location in an unknown environment. robot incrementally builds a map of the environment while simultaneously using this map to compute its location
  65. What is active ranging. Is Active or Passive more accurate ?
    Active ranging is the process of determining the distance to an object via active sensors

  66. what qualities of phase measurement and one application
    • extreme active
    • suffers from range ambiguity
    • often used for relative motion

    Application:Laser Vibrometers
  67. what qualities of time-of-flight and an application
    • limited resolution
    • provides absolute range
    • typically has a minimum range

    Application: Laser rangefinders
  68. Infrared sensors to measure distance also experience errors due to:
    • hysteresis nonlinearity
    • Low signal to noise
  69. Time of flight sensors
    • ultrasonic 
    • and laser
  70. For use in mobile robotics, LIDAR data is typically treated in one of two ways:
    • Occupancy grid
    • Object segmentation
  71. Trade off of Occupancy grid
    Object segmentation
    • occupancy grid - more memory 
    • object segmentation -more computational
  72. Occupancy Grid
    • Simple representation
    • Well suited to graph search algorithms

    • Inherently loses information
    • Resolution/Memory trade off
    • Deals poorly with motion
  73. Object Segmentation 

    • Keeps object information
    • Enables object identification
    • Provides ability to track objects
    • More memory efficient than grid

    • More complicated to implement
    • More difficult to search
  74. Line Segmentation Algorithms
    • Split-and-Merge
    • RANSAC
    • Hough Transform
  75. Split and merge is not robust to
  76. Split and merge recursive algorithm
    • A line is fit to the data
    • The line is split if the variance is too high
    • Adjacent lines are merged if close enough
    • Iterate until all thresholds are achieved
  77. RANSAC is highly
    robust to outliers
  78. RANSAC uses what method
    non-deterministic method, relies on random sampling
  79. Hough transform is a what scheme
  80. hough  transform is often used in ?
    Good for what type of sensors ?
    Disadvantage ?
    • vision processing
    • vision base sensors 
    • end points are hard to compute
  81. Algorithm Comparison:
    Split and merge 
    • Speed     False Positives   Precision
    • 1500              10               +++
    • 400                10               +++
    • 30                  30              ++++
    • 10                  30              ++++
  82. Sonar 
    Sends and receives acoustic pulse 

    Listens for ambient acoustic energy
  83. Sonar sensor uncertainties
    exact position of detected object is unknown 

    sonar distortion
  84. when do specular reflections occur
    when the sonar beam hits a smooth surface at a shallow angle, and is therefore not reflected back to the robot, but outwards.
  85. time-of-flight cameras
    capture the whole 3D scene and is without moving parts
  86. Example of a passive triangulation
    stereo camera
  87. Triangulation and active ranging
    structured light sensor
  88. Structured light sensors:
    Sensor resolution is limited by ?
    Sensor range is limited by ?
    • Image resolution 
    • laser strength
  89. Structured light is best used for
    give objects profiles at high resolution
  90. Doppler Effect is used with ____ to measure ______. One problem?
    • active ranging and velocities
    • may only detect moving objects
  91. radar can be used as a ____ or _____? Radar can eight be _____ or ______
    • ranging or speed sensor
    • active or passive
  92. ranging is accomplished using?
    speed estimation occurs through ?
    • time-of-flight
    • doppler effect
Card Set
Exam 2 ME503
Exam 2 ME503