Balancing Robot Seminar

Embed Size (px)

Citation preview

  • 8/9/2019 Balancing Robot Seminar

    1/14

    10/27/2003 1

    Gyrobota Two Wheel Balancing Robot

    Larry Barello

    http://www.barello.net/Robots/Gyrobot

  • 8/9/2019 Balancing Robot Seminar

    2/14

    10/27/2003 2

    How Balancing Works

    Easier to balance high GC object.

    The Devil is in the details.

    Balanced Tilted

    Sense tilt and drive wheels to make robot erect.

  • 8/9/2019 Balancing Robot Seminar

    3/14

    10/27/2003 3

    How Balancing Works

    Restoring Torque = * K

    = Angular rate (deg/sec)

    K = some factor based upon robot mass & moment arm.

  • 8/9/2019 Balancing Robot Seminar

    4/14

    10/27/2003 4

    How Balancing Works

    Balance torque = M * g * sin(angle) M is moment arm (center of mass & distance from pivot).

    g = acceleration of gravity

    Angle is deviation from balance point

    T = M * g * sin(0)

    M

    Motion

    g

    M

    T = M * g * sin(90)

    M

  • 8/9/2019 Balancing Robot Seminar

    5/14

    10/27/2003 5

    Controlling Drive Torque As velocity increases, DC motors have less torque

    (back-EMF)

    Need to correct this so balance equation can work. Positive velocity feedback

    Drive = drive + Kv * velocity

    Kv = factor based upon motor specification.

  • 8/9/2019 Balancing Robot Seminar

    6/14

    10/27/2003 6

    The robot tilts in the direction of movement.How to move

    To move robot, need to adjust balance angle

    By adding an error to the angle, the robot will drive alittle backwards, then forwards trying to maintain the

    balance angle

    Error can be simply a negative drive value.

  • 8/9/2019 Balancing Robot Seminar

    7/14

    10/27/2003 7

    Tilt Sensor Gyroscope measures Rate Of Turn.

    Integrate for angle

    Gyroscopes drift

    Drift changes withtemperature

  • 8/9/2019 Balancing Robot Seminar

    8/14

    10/27/2003 8

    Tilt Sensor Accelerometers measure gravity.

    Affected by movement.

    Very accurate.

  • 8/9/2019 Balancing Robot Seminar

    9/14

    10/27/2003 9

    Fusing Gyro and Tilt sensors Weighted Average Filter. Use fraction of error between tilt and gyro

    Add to integrated Gyro output.

    Results are fast & drift free The weighted average IIR filter technique can be

    used to smooth out any noisy signal.

  • 8/9/2019 Balancing Robot Seminar

    10/14

    10/27/2003 10

    Fusing Gyro and Tilt sensors Weighted Average Filter. Tiltgyro = Sum(Gyro Offset)

    Tiltaccel = sin-1(x/g)

    Tiltgyro = Tiltgyro + (Tiltaccel Tiltgyro) * K

    Gyro

    Accelerometer

    +

    +

    +-

    K

    Tilt

    Offset -

    K = weighting factor (1%)

  • 8/9/2019 Balancing Robot Seminar

    11/14

    10/27/2003 11

    Alternate to Tilt SensorF = g * sin(theta)F = A * m

    A * M = g * sin(theta)

    Theta = sin-1(A * m/g)

    Presumably we can use the acceleration of therobot to infer the angle of tilt and correct forgyro drift. (Note: Moment arm not included, formulas must bewrong)

    m = mass

    A = Acceleration M/S2

    g = Gravity

    F = force

  • 8/9/2019 Balancing Robot Seminar

    12/14

    10/27/2003 12

    Motion Sensor Encoders measure position and velocity of wheels.

    Dead reckoning can be derived from encoders andthe robots position determined in X, Y and Theta.

  • 8/9/2019 Balancing Robot Seminar

    13/14

    10/27/2003 13

    Current Status Motion control not implemented.

    Navigation disabled.

    Remote control (Radio link) not implemented.

    Using acceleration to correct tilt not tried.

    In short, it doesnt do anything interesting yet.

  • 8/9/2019 Balancing Robot Seminar

    14/14

    10/27/2003 14

    Additional readingRobots & controls

    www.barello.net/Papers/Motion_Control

    Pendulum Joe http://leiwww.epfl.ch

    Data sheetshttp://www.systron.com/prodinfo/AQRS.html

    http://products.analog.com/products/info.asp?product=ADXL202