Presentation is loading. Please wait.

Presentation is loading. Please wait.

Modeling of Robotics Systems (2:30h)

Similar presentations


Presentation on theme: "Modeling of Robotics Systems (2:30h)"— Presentation transcript:

1 Modeling of Robotics Systems (2:30h)

2 Why Geometrical Robotics?
In 3D, especially due to the topology of rotations, geometry is important. Coordinates can bring you to WRONG conclusions, tensors (geometry) NOT ! Complex systems need Powerful Tools to be manageable With the right tools 3D can be as simple as 1D ! 11/20/2018 Semester on Control, Barcelona 2005

3 Semester on Control, Barcelona 2005
Contents 1D Mechanics: as introduction 3D Mechanics Points, vectors, line vectors screws Rotations and Homogeneous matrices Screw Ports 11/20/2018 Semester on Control, Barcelona 2005

4 1D Mechanics

5 Semester on Control, Barcelona 2005
1D Mechanics In 1D Mechanics there is no geometry for the ports: efforts/Forces and flows/velocities are scalars Starting point to introduce the basic elements for 3D 11/20/2018 Semester on Control, Barcelona 2005

6 Semester on Control, Barcelona 2005
Mass Energy Co-Energy where is the momenta the applied force and its velocity. 11/20/2018 Semester on Control, Barcelona 2005

7 The dynamics Equations
The second Law of dynamics is: Integral Form Diff. form 11/20/2018 Semester on Control, Barcelona 2005

8 The Kernel PCH representation
Interconnection port 11/20/2018 Semester on Control, Barcelona 2005

9 Semester on Control, Barcelona 2005
Spring Energy Co-Energy where is the displacement the applied force to the spring and its relative velocity. 11/20/2018 Semester on Control, Barcelona 2005

10 The dynamics Equations
The elastic force on the spring is: Integral Form Diff. form 11/20/2018 Semester on Control, Barcelona 2005

11 The Kernel PCH representation
Interconnection port 11/20/2018 Semester on Control, Barcelona 2005

12 Semester on Control, Barcelona 2005
Mass-Spring System Spring Mass 11/20/2018 Semester on Control, Barcelona 2005

13 Semester on Control, Barcelona 2005
Together…. 11/20/2018 Semester on Control, Barcelona 2005

14 Interconnection of the two subsystems (1 junc.)
Or in image representation 11/20/2018 Semester on Control, Barcelona 2005

15 Semester on Control, Barcelona 2005
Combining… There exists a left orthogonal 11/20/2018 Semester on Control, Barcelona 2005

16 Semester on Control, Barcelona 2005
Finally 11/20/2018 Semester on Control, Barcelona 2005

17 Summary and Conclusions
All possible 1D networks of elements can be expressed in this form Dissipation can be easily included terminating a port on a dissipating element Interconnection of elements still give the same form No geometry In 3D we need a common language for interconnection: Lie groups ! 11/20/2018 Semester on Control, Barcelona 2005

18 The Basics on Space

19 Euclidian Spaces and motions
A 3-dimensional Euclidean space is characterized by a scalar product We can consider the relative position of as a vector : The set of these free-vectors will be indicateds with 11/20/2018 Semester on Control, Barcelona 2005

20 Semester on Control, Barcelona 2005
Notation and concepts Scalar Product Norm (distance) Orthogonality of 11/20/2018 Semester on Control, Barcelona 2005

21 Notation and concepts (cont.)
Angles between vectors Conclusion All concepts defined using only the scalar product. 11/20/2018 Semester on Control, Barcelona 2005

22 Semester on Control, Barcelona 2005
Coordinate Systems A coordinate system is composed of an origin and 3 linear independent free vectors: 11/20/2018 Semester on Control, Barcelona 2005

23 Coordinate Systems (cont.)
A Coordinate System is ortho-normal iff 11/20/2018 Semester on Control, Barcelona 2005

24 Semester on Control, Barcelona 2005
Coordinates Coordinates are real numbers: For (points) For (vectors) We will use often the following notation: 11/20/2018 Semester on Control, Barcelona 2005

25 Semester on Control, Barcelona 2005
Coordinate Mappings 11/20/2018 Semester on Control, Barcelona 2005

26 Semester on Control, Barcelona 2005
Change of coordinates A change of coordinates from to can be expressed with: Which is a mapping like 11/20/2018 Semester on Control, Barcelona 2005

27 Semester on Control, Barcelona 2005
Orientation The vector product operation often used in mechanics, is NOT as often thought an extra structure/operation defined, but it is only a consequence of the Lie group structure of motions: 11/20/2018 Semester on Control, Barcelona 2005

28 Tilde Operator (Notation)
We will often use the following notation In 20-SIM Y=Skew(X) 11/20/2018 Semester on Control, Barcelona 2005

29 Rotations

30 Semester on Control, Barcelona 2005
Approach We will first introduce change of coordinates for rotated frames 2D 3D Relate changes of coordinates to motion of objects 11/20/2018 Semester on Control, Barcelona 2005

31 Semester on Control, Barcelona 2005
Rotations in 2D Given What is the relation of the coordinates of in and in ? 11/20/2018 Semester on Control, Barcelona 2005

32 Semester on Control, Barcelona 2005
We have that Or equivalently With , Similarly, for we have 11/20/2018 Semester on Control, Barcelona 2005

33 Semester on Control, Barcelona 2005
and since , using the expression of in , we have that 11/20/2018 Semester on Control, Barcelona 2005

34 Semester on Control, Barcelona 2005
Conclusion Rotation Matrix 11/20/2018 Semester on Control, Barcelona 2005

35 Semester on Control, Barcelona 2005
Remarks Due to the fact we are considering right-handed orthonormal frames: The columns and row vectors of have length 1 and they are orthogonal 11/20/2018 Semester on Control, Barcelona 2005

36 Semester on Control, Barcelona 2005
3D Rotation In 3D, if the coordinates we use are all right handed, we have for a general rotation around the origin: Basis of espressed in Basis of espressed in 11/20/2018 Semester on Control, Barcelona 2005

37 Semester on Control, Barcelona 2005
Example A rotation around the y axis would be 11/20/2018 Semester on Control, Barcelona 2005

38 Special Orthonormal Group
A square matrix such that is called orthonormal. The group of orthonormal matrices with determinant 1 is called Special Orthonormal group of and indicated as: 11/20/2018 Semester on Control, Barcelona 2005

39 Semester on Control, Barcelona 2005
Theorem If is a differentiable function of time, and are Skew-Symmetric and belonging to : 11/20/2018 Semester on Control, Barcelona 2005

40 Semester on Control, Barcelona 2005
Angular velocities This implies that such that: Remember: 11/20/2018 Semester on Control, Barcelona 2005

41 Semester on Control, Barcelona 2005
is a Lie algebra The linear combination of skew-symmetric matrices is still skew-symmetric To each matrix we can associate a vector such that … It is a vector space It is a Lie Algebra !! 11/20/2018 Semester on Control, Barcelona 2005

42 Semester on Control, Barcelona 2005
SO(3) is a Group It is a Group because Associativity Identity Inverse 11/20/2018 Semester on Control, Barcelona 2005

43 It is a Lie Group (group AND manifold)
where Lie Algebra Commutator 11/20/2018 Semester on Control, Barcelona 2005

44 Common Space thanks to Lie group structure
11/20/2018 Semester on Control, Barcelona 2005

45 Why is the Lie group structure so important?
It makes it possible to talk about motion without knowing the pose of the object! We can get rid of any dependency from configuration: ESSENTIAL to talk about interconnection throw Power Ports of bodies with different poses. 11/20/2018 Semester on Control, Barcelona 2005

46 Semester on Control, Barcelona 2005
Dual Space For any finite dimensional vector space we can define the space of linear operators from that space to co-vector The space of linear operators from to (dual space of ) is indicated with 11/20/2018 Semester on Control, Barcelona 2005

47 Configuration Independent Port !
In our case we have Configuration Independent Port ! 11/20/2018 Semester on Control, Barcelona 2005

48 Other Rotations Representations
Generalized angles: Euler, Bryan etc. (singularity positions and NOT geometric Quaternions: 11/20/2018 Semester on Control, Barcelona 2005

49 Other Rotations Representations
matrix Lie group of unitary matrices of determinant +1 Unit Quaternions and are identifiable with points of a 3-sphere 11/20/2018 Semester on Control, Barcelona 2005

50 Semester on Control, Barcelona 2005
Double Covering of The quaternions and double cover Quaternions and are simply connected is connected but NOT simply connected: see the Dirac Belt Trick ! is one of the few examples of a NON simply connected manifolds WITHOUT holes! 11/20/2018 Semester on Control, Barcelona 2005

51 About the topology of rotations
11/20/2018 Semester on Control, Barcelona 2005

52 General Motions

53 Semester on Control, Barcelona 2005
General Motions It can be seen that in general, for right handed frames where , 11/20/2018 Semester on Control, Barcelona 2005

54 Semester on Control, Barcelona 2005
Homogeneous Matrices Due to the group structure of it is easy to compose changes of coordinates in rotations Can we do the same for general motions ? 11/20/2018 Semester on Control, Barcelona 2005

55 Semester on Control, Barcelona 2005
Projective space Improper hyperplane 11/20/2018 Semester on Control, Barcelona 2005

56 Semester on Control, Barcelona 2005
11/20/2018 Semester on Control, Barcelona 2005

57 Motions versus Coordinate Changes
A rigit motion is a map like: 11/20/2018 Semester on Control, Barcelona 2005

58 Semester on Control, Barcelona 2005
Features of motions must be both: Isometry: it does not change the distance between points. Orientation preserving: it does not change the orientation: swap the z axis. 11/20/2018 Semester on Control, Barcelona 2005

59 Connection with coordinate changes
we want that the moved points in the moved coordinate system have the same coordinates than the original points in the original coordinate system : 11/20/2018 Semester on Control, Barcelona 2005

60 Semester on Control, Barcelona 2005
But is the same as: The representation of the motion from 1 to 2 in the coordinate 1 (1 o h o 1-1) is the INVERSE of the coordinate change from 1 to 2 (2 o 1-1). 11/20/2018 Semester on Control, Barcelona 2005

61 Semester on Control, Barcelona 2005
Theorem If is a differentiable function of time belong to where 11/20/2018 Semester on Control, Barcelona 2005

62 Semester on Control, Barcelona 2005
Tilde operator In 20-SIM Y=Tilde(X) 11/20/2018 Semester on Control, Barcelona 2005

63 Elements of se(3): Twists
The following are vector and matrix coordinate notations for twists: The following are often called twists too, but they are no geometrical entities ! 11/20/2018 9 change of coordinates ! Semester on Control, Barcelona 2005

64 Semester on Control, Barcelona 2005
SE(3) is a Group It is a Group because Associativity Identity Inverse 11/20/2018 Semester on Control, Barcelona 2005

65 SE(3) is a Lie Group (group AND manifold)
where Lie Algebra Commutator 11/20/2018 Semester on Control, Barcelona 2005

66 Common Space independent on H thanks to Lie group structure
11/20/2018 Semester on Control, Barcelona 2005

67 Semester on Control, Barcelona 2005
Intuition of Twists Consider a point fixed in : and consider a second reference where and 11/20/2018 Semester on Control, Barcelona 2005

68 For each of the points in the body
11/20/2018 Semester on Control, Barcelona 2005

69 Semester on Control, Barcelona 2005
Possible Choices For the twist of with respect to we consider and we have 2 possibilities 11/20/2018 Semester on Control, Barcelona 2005

70 Left and Right Translations
11/20/2018 Semester on Control, Barcelona 2005

71 Semester on Control, Barcelona 2005
Remarks The twists are a description independent of H and have a pure geometrical interpretation. Being independent on configuration, they will be the key to interconnection between parts with different configurations 11/20/2018 Semester on Control, Barcelona 2005

72 Semester on Control, Barcelona 2005
Possible Choices and 11/20/2018 Semester on Control, Barcelona 2005

73 Notation used for Twists
For the motion of body with respect to body expressed in the reference frame we use or The twist is an across variable ! Point mass geometric free-vector Rigid body geometric screw + Magnitude 11/20/2018 Semester on Control, Barcelona 2005

74 Chasle's Theorem and intuition of a Twist
Any twist can be written as: 11/20/2018 Semester on Control, Barcelona 2005

75 Calculation of geometrical screw from (,v)
Supposing r perpendicular to omega , it is furthermore possible to see that: and 11/20/2018 Semester on Control, Barcelona 2005

76 Semester on Control, Barcelona 2005
Examples of Twists 11/20/2018 Semester on Control, Barcelona 2005

77 Semester on Control, Barcelona 2005
Examples of Twists 11/20/2018 Semester on Control, Barcelona 2005

78 Changes of Coordinates for Twists
It can be proven that In 20-SIM S=Adjoint(H) 11/20/2018 Semester on Control, Barcelona 2005

79 Semester on Control, Barcelona 2005
… in Matrix form 11/20/2018 Semester on Control, Barcelona 2005

80 Semester on Control, Barcelona 2005
Wrenches Twists belong geometrically to Wrenches are DUAL of twist: Wrenches are co-vectors and NOT vectors: linear operators from Twists to Power Using coordinates: 11/20/2018 Semester on Control, Barcelona 2005

81 Poinsot's Theorem and intuition of a Wrench
Any wrench can be written as: 11/20/2018 Semester on Control, Barcelona 2005

82 Semester on Control, Barcelona 2005
Chasles vs. Poinsot Charles Theorem Poinsot Theorem The inversion of the upper and lower part corresponds to the use of the Klijn form 11/20/2018 Semester on Control, Barcelona 2005

83 Vectors, Screws as “Forces”
Forces and Wrenches are co-vectors, but: Euclidean metric vector interpretation of a Force Klein’s form screw interpretation of a Wrench That is identification of dual spaces. 11/20/2018 Semester on Control, Barcelona 2005

84 Example of the use of a Wrench
Finding the contact centroid 11/20/2018 Semester on Control, Barcelona 2005

85 Transformation of Wrenches
How do wrenches transform changing coordinate systems? We have seen that for twists: 11/20/2018 Semester on Control, Barcelona 2005

86 Changes of coordinates
MTF 11/20/2018 Semester on Control, Barcelona 2005

87 Semester on Control, Barcelona 2005
Implicit relation Dirac Structure! MTF 11/20/2018 Semester on Control, Barcelona 2005

88 Semester on Control, Barcelona 2005
Conclusions Coordinate free interpretation of motion can be related to twists, element of the Lie albebra se(3) Dual elements are called wrenches and they also have a coordinate free interpretation thanks to the Hyperbolic form. The pairing of this two items will be the kern for interconnection 11/20/2018 Semester on Control, Barcelona 2005

89 Semester on Control, Barcelona 2005
Power Port A B belong to vector spaces in duality: such that there exists a bilinear operator 11/20/2018 Semester on Control, Barcelona 2005

90 Finite dimensional case
If is finite dimensional is uniquely defined, namely where indicates the uniquely defined set of linear operators from to Elements of are vectors Elements of are co-vectors 11/20/2018 Semester on Control, Barcelona 2005

91 Semester on Control, Barcelona 2005
In Robotics Is the v.s. of Twists Is the v.s. of Wrenches 11/20/2018 Semester on Control, Barcelona 2005

92 Semester on Control, Barcelona 2005
Remarks A geometrical (coordinate free) way to talk about instantaneous motion is using a Twist and about a system of forces is a Wrench Twists and Wrenches are geometrically a screw (motor): line-bounded-vector (rotor)+pitch. Twists and Wrenches are dual and can be used to define a power-port 11/20/2018 Semester on Control, Barcelona 2005

93 Semester on Control, Barcelona 2005
Power and Inf. Dim Case A B represents the instantaneous power flowing from A to B For inf.dim. systems they belong to k and (n-k) (Lie-algebra-valued) forms 11/20/2018 Semester on Control, Barcelona 2005

94 Dynamics

95 Semester on Control, Barcelona 2005
Contents time derivative Rigid Body dynamics Spatial Springs Kinematic Pairs Mechanism Topology 11/20/2018 Semester on Control, Barcelona 2005

96 Semester on Control, Barcelona 2005
time derivative is function of time It can be proven that where 11/20/2018 Semester on Control, Barcelona 2005

97 Semester on Control, Barcelona 2005
Transformations of If we have ,how does look like? 11/20/2018 Semester on Control, Barcelona 2005

98 It can be shown that in general
11/20/2018 Semester on Control, Barcelona 2005

99 Rigid Bodies Dynamics

100 Semester on Control, Barcelona 2005
Rigid bodies A rigid Body is characterised by a (0,2) tensor called Inertia Tensor: and we can then define the momentum screw: where the Kinetic energy is 11/20/2018 Semester on Control, Barcelona 2005

101 Generalization of Newton’s law
In an inertial frame, for a point mass we had This can be generalized for rigid bodies Where Ni0 is the moment of body i expressed in the inertial frame 0 . That is why momenta is a co-vector !! 11/20/2018 Semester on Control, Barcelona 2005

102 And in body coordinates ?
Using the derivative of AdH 11/20/2018 Semester on Control, Barcelona 2005

103 Semester on Control, Barcelona 2005
….. multiplying on the left for we get 11/20/2018 Semester on Control, Barcelona 2005

104 Semester on Control, Barcelona 2005
and since we have that and we eventually obtain 11/20/2018 Semester on Control, Barcelona 2005

105 Semester on Control, Barcelona 2005
Momentum dynamics which is called Lie-Poisson reduction. NOTE: No information on configuration ! 11/20/2018 Semester on Control, Barcelona 2005

106 Screw Port representation (Bondgraph)
=0 ! stored input 11/20/2018 Semester on Control, Barcelona 2005

107 Semester on Control, Barcelona 2005
Relation to what known Expressing the previous equation in the principal inertial frame k we obtain: 11/20/2018 Semester on Control, Barcelona 2005

108 Semester on Control, Barcelona 2005
Other form Defining which is linear and anti-symmetric 11/20/2018 Semester on Control, Barcelona 2005

109 Port-Hamiltonian form
11/20/2018 Semester on Control, Barcelona 2005

110 Port-Hamiltonian form
Modulation Storage port Interconnection port 11/20/2018 Semester on Control, Barcelona 2005

111 Rotational Dynamics in more details

112 Example: Ellipsoid Dynamics
Semi-axis lengths= 11/20/2018 Semester on Control, Barcelona 2005

113 Example: Elipsoid Dynamics (cont.)
In SI Units: 11/20/2018 Semester on Control, Barcelona 2005

114 If no external torque is applied
Note: in what follows P indicates angular momenta, NOT homogeneous position If no torques are applied: Square of angular momentum in body coordinate is also conserved!! 11/20/2018 Semester on Control, Barcelona 2005

115 Momentum conservation
These are the equation of a sphere in the body coordinate momentum frame 11/20/2018 Semester on Control, Barcelona 2005

116 Semester on Control, Barcelona 2005
Energy conservation Ellipsoid! 11/20/2018 Semester on Control, Barcelona 2005

117 Semester on Control, Barcelona 2005
Orbits of the momentum Body coords!! Conservation of momentum Casimir ! sphere Conservation of energy ellipsoid 11/20/2018 Semester on Control, Barcelona 2005

118 Semester on Control, Barcelona 2005
Orbit Sphere With the intersection of the sphere and ellipsoid we obtain the famous plot: 11/20/2018 Semester on Control, Barcelona 2005

119 Rotation around the biggest inertia
11/20/2018 Semester on Control, Barcelona 2005

120 Rotation around middle inertia
11/20/2018 Semester on Control, Barcelona 2005

121 Rotation around smallest inertia
11/20/2018 Semester on Control, Barcelona 2005

122 Geometric Springs

123 4 cells: 1 stable+3 unstable points
Spatial Springs If, by means of control, we define a 3D spring using a parameterization like Euler angles, we do not have a geometric description of the spring: no information about the center of compliance, instead: Morse Theory 4 cells: 1 stable+3 unstable points 11/20/2018 Semester on Control, Barcelona 2005

124 Semester on Control, Barcelona 2005
Spatial Springs where where 11/20/2018 Semester on Control, Barcelona 2005

125 For Constant Spatial Spring
It could be shown that: Storage port Interconnection port to integrate! 11/20/2018 Semester on Control, Barcelona 2005

126 Kinematic Pairs

127 Semester on Control, Barcelona 2005
Kinematic Pair A n-dof K.P. is an ideal constraint between 2 rigid bodies which allows n independent motions For each relative configuration of the bodies we can define Allowed subspace of dimension n Actuation subspace of dimension n 11/20/2018 Semester on Control, Barcelona 2005

128 Semester on Control, Barcelona 2005
Decomposition of and ! n n 6-n 6-n 11/20/2018 Semester on Control, Barcelona 2005

129 Representations of subspaces
To satisfy power continuity 11/20/2018 Semester on Control, Barcelona 2005

130 And in the Kernel Dirac representation
Interconnection port Actuators ports 11/20/2018 Semester on Control, Barcelona 2005

131 Mechanism Topology

132 Semester on Control, Barcelona 2005
Network Topology Interconnection of q rigid bodies by n nodic elements (kinematic pairs, springs or dampers). We can define the Primary Graph describing the mechanism and than: Port connection graph= Lagrangian tree + Primary Graph 11/20/2018 Semester on Control, Barcelona 2005

133 Semester on Control, Barcelona 2005
Primary Graph 11/20/2018 Semester on Control, Barcelona 2005

134 Semester on Control, Barcelona 2005
Primary Graph The Primary graph is characterised by the Incedence Matrix 11/20/2018 Semester on Control, Barcelona 2005

135 Semester on Control, Barcelona 2005
Lagrangian Tree 11/20/2018 Semester on Control, Barcelona 2005

136 Fundamental Loop Matrix
Lagrangian Tree Primary Graph 11/20/2018 Semester on Control, Barcelona 2005

137 Fundamental Cut-set Matrix
Lagrangian Tree Primary Graph 11/20/2018 Semester on Control, Barcelona 2005

138 Semester on Control, Barcelona 2005
`Power Continuity Power continuity ! 11/20/2018 Semester on Control, Barcelona 2005

139 Mechanism Dirac Structure
Power Ports Rigid Bodies Power Ports Nodic Elements 11/20/2018 Semester on Control, Barcelona 2005

140 Semester on Control, Barcelona 2005
Further Steps… 11/20/2018 Semester on Control, Barcelona 2005

141 Semester on Control, Barcelona 2005
Conclusions Any 3D part can be modeled in the Dirac framework Any interconnection also ! In this case the ports have a geometrical structures: no scalars ! Some steps still to go to bring the system in explicit form A lot of extensions are possible Not trivial to bring everything in simplified explicit form 11/20/2018 Semester on Control, Barcelona 2005


Download ppt "Modeling of Robotics Systems (2:30h)"

Similar presentations


Ads by Google