Simplify Challenging Software Problems with Rocket Science

Aerospace engineering is really no more magical than programming, and yet it's shrouded in such mystery and mysticism that practically no one bothers to study it. I'll show you how to apply concepts from rocket science to challenging software problems, allowing you to spend more time coming up with answers rather than interpreting inputs. (40 minute version presented at EuRuKo 2015)

Transcript

1. Simplify Challenging
   Software Problems with
   Rocket Science
   @bendyworks
   Brad Grzesiak — @listrophy
   

   View Slide

2. –Ashe Dryden
   “Whenever Brad is drinking, you
   wanna make sure you're sitting
   next to him, because instead of
   being obnoxious, he teaches you
   orbital mechanics.”
   

   View Slide

3. Heads up:
   1 Shaky Movie &
   1 Animated GIF
   

   View Slide

4. View Slide

5. View Slide

6. View Slide

7. View Slide

8. Agenda
   • Reduce Cognitive Load
   • Visualize Simulations with Graphics
   • Control and Respond to Real World Processes
   

   View Slide

9. Agenda
   • Reduce Cognitive Load
   • Visualize Simulations with Graphics
   • Control and Respond to Real World Processes
   

   View Slide

10. we
    are
    not
    superheros
    

    View Slide

11. Scalars & Collections
    Primitives & Arrays
    • x, y, z = 2, 3, 5
    • Math.sqrt(x**2 + y**2 + z**2)
    • array.zip(other).map{|(i,j)| i+j}
    • [m[0][0]*v[0]+m[0][1]*v[1],
    m[1][0]*v[0]+m[1][1]*v[1]]
    Vectors & Matrices
    • vector = Vector[2, 3, 5]
    • vector.magnitude # or vector.r
    • vector_1 + vector_2
    • matrix * vector

    

    View Slide

12. “matrix”
    “times”
    “vector?”
    

    View Slide

13. Vector
    • Underlying implementation: Array
    • Usually Numeric, but really anything
    • Native operations are different (c.f. addition)
    • Denoted by: ⟨x,y⟩ or x or x or or
    ⽷x⽹  
    ⽸y⽺  
    ⽷2⽹  
    ⽸0⽺  
    

    View Slide

14. Vector Example
    

    View Slide

15. Vector Example
    ⟨2,0⟩
    

    View Slide

16. Vector Example
    ⟨2,0⟩
    ⟨1,3⟩
    

    View Slide

17. Matrices, or
    “Linear Transformations”
    • Structurally like a 2D Array
    • Used to Scale, Rotate, Skew, Perspective, Translate
    • Common in CSS3
    • Useful for 2D & 3D graphics
    • Common in Kinetics & Material Mechanics
    • Denoted by M or M or or
    ⽷a  b⽹  
    ⽸c  d⽺  
    ⽷1  4⽹  
    ⽸9  0⽺  
    

    View Slide

18. Matrix Times Vector
    1.  ⽷a    b⽹⽷x⽹  
         ⽸c    d⽺⽸y⽺  
    2.  ［x    y］  
         ⽷a    b⽹  
         ⽸c    d⽺  
    3.  ⽷ax+by⽹  
         ⽸cx+dy⽺  
    

    View Slide

19. Matrix Times Vector
    def matrix_times_vector(m, v)
    a, b = m[0]
    c, d = m[1]
    x, y = v
    [
    a*x + b*y,
    c*x + d*y
    ]
    end
    require 'matrix'
    def matrix_times_vector(m, v)
    m * v
    end
    

    View Slide

20. Matrix Times Vector
    def matrix_times_vector(m, v)
    a, b = m[0]
    c, d = m[1]
    x, y = v
    [
    a*x + b*y,
    c*x + d*y
    ]
    end
    require 'matrix'
    def matrix_times_vector(m, v)
    m * v
    end
    NOW
    WITH DIM
    CHECKS!
    

    View Slide

21. Linear Transformation
    Examples
    

    View Slide

22. Hooke’s Law
    for Anisotropic Materials
    σ  =  c  *  ε
    

    View Slide

23. Hooke’s Law
    for Anisotropic Materials
    ⽷σ₁₁⽹  ⽷C₁₁  C₁₂  C₁₃  C₁₄  C₁₅  C₁₆⽹⽷  ε₁₁⽹

    ⾇σ₂₂⾇  ⾇C₁₂  C₂₂  C₂₃  C₂₄  C₂₅  C₂₆⾇⾇  ε₂₂⾇  
    ⾇σ₃₃⾇=⾇C₁₃  C₂₃  C₃₃  C₃₄  C₃₅  C₃₆⾇⾇  ε₃₃⾇  
    ⾇σ₂₃⾇  ⾇C₁₄  C₂₄  C₃₄  C₄₄  C₄₅  C₄₆⾇⾇2ε₂₃⾇  
    ⾇σ₁₃⾇  ⾇C₁₅  C₂₅  C₃₅  C₄₅  C₅₅  C₅₆⾇⾇2ε₁₃⾇  
    ⽸σ₁₂⽺  ⽸C₁₆  C₂₆  C₃₆  C₄₆  C₅₆  C₆₆⽺⽸2ε₁₂⽺
    

    View Slide

24. Hooke’s Law
    for Anisotropic Materials
    ⽷σ₁₁⽹  ⽷C₁₁  C₁₂  C₁₃  C₁₄  C₁₅  C₁₆⽹⽷  ε₁₁⽹

    ⾇σ₂₂⾇  ⾇C₁₂  C₂₂  C₂₃  C₂₄  C₂₅  C₂₆⾇⾇  ε₂₂⾇  
    ⾇σ₃₃⾇=⾇C₁₃  C₂₃  C₃₃  C₃₄  C₃₅  C₃₆⾇⾇  ε₃₃⾇  
    ⾇σ₂₃⾇  ⾇C₁₄  C₂₄  C₃₄  C₄₄  C₄₅  C₄₆⾇⾇2ε₂₃⾇  
    ⾇σ₁₃⾇  ⾇C₁₅  C₂₅  C₃₅  C₄₅  C₅₅  C₅₆⾇⾇2ε₁₃⾇  
    ⽸σ₁₂⽺  ⽸C₁₆  C₂₆  C₃₆  C₄₆  C₅₆  C₆₆⽺⽸2ε₁₂⽺
    

    View Slide

25. Our Vector
    ⟨1,3⟩
    

    View Slide

26. Identity Transform
    

    View Slide

27. Identity Transform
    I      =⽷1  0⽹

             ⽸0  1⽺

    

    View Slide

28. Identity Transform
    I      =⽷1  0⽹

             ⽸0  1⽺

    I*v  =⽷1  0⽹⽷1⽹

             ⽸0  1⽺⽸3⽺

    

    View Slide

29. Identity Transform
    I      =⽷1  0⽹

             ⽸0  1⽺

    I*v  =⽷1  0⽹⽷1⽹

             ⽸0  1⽺⽸3⽺

    I*v  =⽷(1*1)  +  (0*3)⽹

             ⽸(0*1)  +  (1*3)⽺

    

    View Slide

30. Identity Transform
    I      =⽷1  0⽹

             ⽸0  1⽺

    I*v  =⽷1  0⽹⽷1⽹

             ⽸0  1⽺⽸3⽺

    I*v  =⽷(1*1)  +  (0*3)⽹

             ⽸(0*1)  +  (1*3)⽺

    I*v  =⽷1⽹

             ⽸3⽺
    

    View Slide

31. Scale Transform
    

    View Slide

32. Scale Transform
    S      =⽷s  0⽹=⽷2  0⽹

             ⽸0  s⽺  ⽸0  2⽺

    

    View Slide

33. Scale Transform
    S      =⽷s  0⽹=⽷2  0⽹

             ⽸0  s⽺  ⽸0  2⽺

    S*v  =⽷2  0⽹⽷1⽹

             ⽸0  2⽺⽸3⽺

    

    View Slide

34. Scale Transform
    S      =⽷s  0⽹=⽷2  0⽹

             ⽸0  s⽺  ⽸0  2⽺

    S*v  =⽷2  0⽹⽷1⽹

             ⽸0  2⽺⽸3⽺

    S*v  =⽷(2*1)  +  (0*3)⽹

             ⽸(0*1)  +  (2*3)⽺

    

    View Slide

35. Scale Transform
    S      =⽷s  0⽹=⽷2  0⽹

             ⽸0  s⽺  ⽸0  2⽺

    S*v  =⽷2  0⽹⽷1⽹

             ⽸0  2⽺⽸3⽺

    S*v  =⽷(2*1)  +  (0*3)⽹

             ⽸(0*1)  +  (2*3)⽺

    S*v  =⽷2⽹

             ⽸6⽺
    

    View Slide

36. Flip Horizontal Transform
    

    View Slide

37. Flip Horizontal Transform
    H      =⽷-­‐1  0⽹

             ⽸  0  1⽺

    

    View Slide

38. Flip Horizontal Transform
    H      =⽷-­‐1  0⽹

             ⽸  0  1⽺

    H*v  =⽷-­‐1  0⽹⽷1⽹

             ⽸  0  1⽺⽸3⽺

    

    View Slide

39. Flip Horizontal Transform
    H      =⽷-­‐1  0⽹

             ⽸  0  1⽺

    H*v  =⽷-­‐1  0⽹⽷1⽹

             ⽸  0  1⽺⽸3⽺

    H*v  =⽷(-­‐1*1)  +  (0*3)⽹

             ⽸(  0*1)  +  (1*3)⽺

    

    View Slide

40. Flip Horizontal Transform
    H      =⽷-­‐1  0⽹

             ⽸  0  1⽺

    H*v  =⽷-­‐1  0⽹⽷1⽹

             ⽸  0  1⽺⽸3⽺

    H*v  =⽷(-­‐1*1)  +  (0*3)⽹

             ⽸(  0*1)  +  (1*3)⽺

    H*v  =⽷-­‐1⽹

             ⽸  3⽺
    

    View Slide

41. SkewX Transform
    

    View Slide

42. SkewX Transform
    X(θ)          =⽷1  tanθ⽹

                       ⽸0        1⽺

    

    View Slide

43. SkewX Transform
    X(θ)          =⽷1  tanθ⽹

                       ⽸0        1⽺

    X(30°)*v  =⽷1  .58⽹⽷1⽹

                       ⽸0      1⽺⽸3⽺

    

    View Slide

44. SkewX Transform
    X(θ)          =⽷1  tanθ⽹

                       ⽸0        1⽺

    X(30°)*v  =⽷1  .58⽹⽷1⽹

                       ⽸0      1⽺⽸3⽺

    X(30°)*v  =⽷(1*1)  +  (.58*3)⽹

                       ⽸(0*1)  +  (    1*3)⽺

    

    View Slide

45. SkewX Transform
    X(θ)          =⽷1  tanθ⽹

                       ⽸0        1⽺

    X(30°)*v  =⽷1  .58⽹⽷1⽹

                       ⽸0      1⽺⽸3⽺

    X(30°)*v  =⽷(1*1)  +  (.58*3)⽹

                       ⽸(0*1)  +  (    1*3)⽺

    X(30°)*v  =⽷2.73⽹

                       ⽸3      ⽺
    

    View Slide

46. SkewX Transform
    X(θ)          =⽷1  tanθ⽹

                       ⽸0        1⽺

    X(30°)*v  =⽷1  .58⽹⽷1⽹

                       ⽸0      1⽺⽸3⽺

    X(30°)*v  =⽷(1*1)  +  (.58*3)⽹

                       ⽸(0*1)  +  (    1*3)⽺

    X(30°)*v  =⽷2.73⽹

                       ⽸3      ⽺
    

    View Slide

47. SkewX Transform
    X(θ)          =⽷1  tanθ⽹

                       ⽸0        1⽺

    X(30°)*v  =⽷1  .58⽹⽷1⽹

                       ⽸0      1⽺⽸3⽺

    X(30°)*v  =⽷(1*1)  +  (.58*3)⽹

                       ⽸(0*1)  +  (    1*3)⽺

    X(30°)*v  =⽷2.73⽹

                       ⽸3      ⽺
    

    View Slide

48. Rotation Transform
    

    View Slide

49. Rotation Transform
    R(θ)          =⽷cosθ  -­‐sinθ⽹

                       ⽸sinθ    cosθ⽺

    

    View Slide

50. Rotation Transform
    R(θ)          =⽷cosθ  -­‐sinθ⽹

                       ⽸sinθ    cosθ⽺

    R(90°)*v  =⽷0  -­‐1⽹⽷1⽹

                       ⽸1    0⽺⽸3⽺

    

    View Slide

51. Rotation Transform
    R(θ)          =⽷cosθ  -­‐sinθ⽹

                       ⽸sinθ    cosθ⽺

    R(90°)*v  =⽷0  -­‐1⽹⽷1⽹

                       ⽸1    0⽺⽸3⽺

    R(90°)*v  =⽷(0*1)  +  (-­‐1*3)⽹

                       ⽸(1*1)  +  (  0*3)⽺

    

    View Slide

52. Rotation Transform
    R(θ)          =⽷cosθ  -­‐sinθ⽹

                       ⽸sinθ    cosθ⽺

    R(90°)*v  =⽷0  -­‐1⽹⽷1⽹

                       ⽸1    0⽺⽸3⽺

    R(90°)*v  =⽷(0*1)  +  (-­‐1*3)⽹

                       ⽸(1*1)  +  (  0*3)⽺

    R(90°)*v  =⽷-­‐3⽹

                       ⽸  1⽺
    

    View Slide

53. Rotation Transform
    R(30°)*                          =  
    

    View Slide

54. Rotation Transform
    R(30°)*                          =  
    

    View Slide

55. Agenda
    • Reduce Cognitive Load
    • Visualize Simulations with Graphics
    • Control and Respond to Real World Processes
    

    View Slide

56. Agenda
    • Reduce Cognitive Load
    • Visualize Simulations with Graphics
    • Control and Respond to Real World Processes
    

    View Slide

57. Why Graphic Simulations?
    • Complex Systems
    • High Sensitivity
    • Pattern Recognition Required
    • Seeking “Good Enough”
    

    View Slide

58. Why Graphic Simulations?
    • Complex Systems
    • High Sensitivity
    • Pattern Recognition Required
    • Seeking “Good Enough”
    

    View Slide

59. Why Graphic Simulations?
    • Complex Systems
    • High Sensitivity
    • Pattern Recognition Required
    • Seeking “Good Enough”
    Tests?
    

    View Slide

60. Why Graphic Simulations?
    • Complex Systems
    • High Sensitivity
    • Pattern Recognition Required
    • Seeking “Good Enough”
    Tests?
    

    View Slide

61. In Aerospace Engineering
    • Launch Vibration Testing
    • “Kick” tests
    • Thermal Equilibrium
    • Designing to Safety Factor
    

    View Slide

62. In Aerospace Engineering
    • Launch Vibration Testing
    • “Kick” tests
    • Thermal Equilibrium
    • Designing to Safety Factor
    • N-Body Orbital Trajectories
    

    View Slide

63. The n-Body Problem
    • Newton’s Laws + Gravity
    • Solved for N=2
    • For N>2, “no general analytical solution,” except for
    special cases (e.g., Lagrangian Points)
    

    View Slide

64. The n-Body Problem
    Gravity Force = x
    Newton’s 2nd Law Force = mass ⨉ acceleration
    Newton’s 3rd Law Force₁ = -Force₂
    Velocity-Acceleration
    Position-Velocity
    

    View Slide

65. Wait a minute.
    Derivatives?
    

    View Slide

66. Velocity = Derivative of Position
    0 km 60 km
    

    View Slide

67. Velocity = Derivative of Position
    0 km 60 km
    t = 1 hour
    

    View Slide

68. Velocity = Derivative of Position
    0 km 1 km
    

    View Slide

69. Velocity = Derivative of Position
    0 km 1 km
    t = 1 minute
    

    View Slide

70. Velocity = Derivative of Position
    0 km 16.67 meters
    

    View Slide

71. Velocity = Derivative of Position
    0 km 16.67 meters
    t = 1 second
    

    View Slide

72. Velocity = Derivative of Position
    0 km 1.667 centimeters
    

    View Slide

73. Velocity = Derivative of Position
    0 km 1.667 centimeters
    t = 1 millisecond
    

    View Slide

74. Velocity = Derivative of Position
    now 1 hour later
    60 km
    0 km
    

    View Slide

75. Velocity = Derivative of Position
    now 1 hour later
    60 km/hr
    0 km/hr
    

    View Slide

76. Velocity = Derivative of Position
    now 1 hour later
    60 km/hr
    0 km/hr
    1 minute later
    

    View Slide

77. Formal Derivative
    

    View Slide

78. Derivative Applied
    

    View Slide

79. Derivative Applied
    

    View Slide

80. Derivative Applied
    

    View Slide

81. Derivative Applied
    

    View Slide

82. Derivative Applied
    

    View Slide

83. Dimensional Analysis,
    or “Fixing Units”
    

    View Slide

84. Dimensional Analysis,
    or “Fixing Units”
    

    View Slide

85. Our Approximation
    

    View Slide

86. Our Approximation
    

    View Slide

87. Our Approximation
    velocity = [position(now) - position(recent)] / (now - recent)
    

    View Slide

88. Our Approximation
    velocity = [position(now) - position(recent)] / (now - recent)
    acceleration = [velocity(now) - velocity(recent)] / (now - recent)
    

    View Slide

89. Our Approximation
    velocity = [position(now) - position(recent)] / (now - recent)
    acceleration = [velocity(now) - velocity(recent)] / (now - recent)
    position(now) = position(recent) + velocity * (now - recent)
    velocity(now) = velocity(recent) + acceleration * (now - recent)
    

    View Slide

90. Our Approximation
    position(now) = position(recent) + velocity * (now - recent)
    velocity(now) = velocity(recent) + acceleration * (now - recent)
    

    View Slide

91. The n-Body Problem
    Gravity Force = x
    Newton’s 2nd Law Force = mass ⨉ acceleration
    Newton’s 3rd Law Force₁ = -Force₂
    Velocity-Acceleration
    Position-Velocity
    

    View Slide

92. The n-Body Problem
    Gravity Force = x
    Newton’s 2nd Law Force = mass ⨉ acceleration
    Newton’s 3rd Law Force₁ = -Force₂
    Velocity-Acceleration vnow = vprev + anow dt
    Position-Velocity xnow = xprev + vnow dt
    

    View Slide

93. The n-Body Problem
    Gravity Force = x
    Newton’s 2nd Law acceleration = Force ÷ mass
    Newton’s 3rd Law Force₁ = -Force₂
    Velocity-Acceleration vnow = vprev + anow dt
    Position-Velocity xnow = xprev + vnow dt
    

    View Slide

94. Gravity Fnow = x
    Newton’s 2nd Law anow = Fnow ÷ m
    Newton’s 3rd Law (loop over each mass-pair)
    Velocity-Acceleration vnow = vprev + anow dt
    Position-Velocity xnow = xprev + vnow dt
    The n-Body Problem
    

    View Slide

95. The n-Body Problem
    bodies.map do |body|
    [body, gravity(body, bodies-[body])]
    end.each do |(body, force)|
    accel = force / body.mass
    body.velocity += accel * dt
    body.position += body.velocity * dt
    end
    

    View Slide

96. The n-Body Problem
    def gravity(body, others)
    others.reduce(Vector[0,0]) do |memo, other|
    offset = other.position - body.position
    memo + offset * G * body.mass * other.mass / offset.r**3
    end
    end
    

    View Slide

97. (demo)
    

    View Slide

98. Agenda
    • Reduce Cognitive Load
    • Visualize Simulations with Graphics
    • Control and Respond to Real World Processes
    

    View Slide

99. Agenda
    • Reduce Cognitive Load
    • Visualize Simulations with Graphics
    • Control and Respond to Real World Processes
    

    View Slide

100. The World is Complicated
     • “Assume no air resistance”
     • “Assume a frictionless surface”
     • “Assume perfect DC voltage”
     • “Assume no impact from general or special
     relativity”
     

     View Slide

101. You simply cannot predefine a
     specific solution to a problem that
     is perturbed by the real world
     

     View Slide

102. –Helmuth von Moltke the Elder
     “No battle plan ever survives
     contact with the enemy.”
     

     View Slide

103. The Answer: React
     

     View Slide

104. Feedback Controller
     

     View Slide

105. Feedback Controller
     

     View Slide

106. Feedback Controller
     

     View Slide

107. PID Controller
     

     View Slide

108. Controller
     P
     I
     D
     roportional
     ntegral
     erivative
     

     View Slide

109. Feedback Controller
     

     View Slide

110. Feedback Controller
     diff == error
     

     View Slide

111. Proportional
     P  =  error  =  desired  -­‐  actual
     

     View Slide

112. Derivative
     D  =  (error_now  -­‐  error_recent)/dt
     

     View Slide

113. Integral
     I  +=  error*dt
     

     View Slide

114. Our PID Controller
     loop do
     sleep 0.2
     curr_time = Time.now
     dt = curr_time - prev_time
     p = error = desired - actual
     i = i + error * dt
     d = (error - prev_error)/dt
     setThrust(0.1*p + 0.003*i + 0.08*d)
     prev_error, prev_time = error, curr_time
     end
     

     View Slide

115. (demo)
     

     View Slide

116. Agenda
     • Reduce Cognitive Load
     • Visualize Simulations with Graphics
     • Control and Respond to Real World Processes
     

     View Slide

117. Agenda
     • Reduce Cognitive Load
     • Visualize Simulations with Graphics
     • Control and Respond to Real World Processes
     

     View Slide

118. Photo Credits
     • “LPE II” (Bradley Grzesiak)
     • “Zero G Robot” (Bradley Grzesiak)
     • “Outredgeous Lettuce” (Public Domain):

     https://www.nasa.gov/mission_pages/station/
     research/news/meals_ready_to_eat/
     • “ABS/EUTELSAT-1 Launch” (Public Domain):

     http://www.spacex.com/media-gallery/detail/
     127081/4896
     • “Beer Actor” (CC-BY):

     https://www.flickr.com/photos/cogdog/
     21127813750/
     • “Up, up and away!” (CC-BY-SA):

     https://www.flickr.com/photos/kindercapes/
     5694816038
     • “I’m confused” (CC-BY-SA):

     https://www.flickr.com/photos/doctabu/342220423
     • “Cookie Monster Waiting” (no idea):

     https://www.google.com/search?q=cookie+monster
     +waiting
     • “Elmer Pump Heat Equation” (CC-BY-SA):

     https://commons.wikimedia.org/wiki/File:Elmer-
     pump-heatequation.png
     • “Owl Stuck” (CC-BY):

     https://www.flickr.com/photos/eben-frostey/
     7147015007
     • Car (Public Domain):

     https://thenounproject.com/search/?q=car&i=13094
     • Curing Sausage (Used with permission):

     https://www.flickr.com/photos/aaronp/7532277486/
     

     View Slide

119. Thank You!
     Brad Grzesiak
     brad@bendyworks.com
     @listrophy
     gitlab.com/listrophy
     

     View Slide