Generalized coordinates and constraints Basilio Bona DAUIN - - PowerPoint PPT Presentation

Generalized coordinates and constraints

Basilio Bona

DAUIN – Politecnico di Torino

Semester 1, 2016-17

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 1 / 25

Coordinates

Each rigid body is defined by 6 coordinates (called d.o.f. or dof), 3 for the position χ, 3 for the orientation α of a body frame RB attached to it; the position and orientation stacked together are called the pose p of the body. p(t) def = x(t) α(t)

• =

       p1(t) p2(t) p3(t) p4(t) p5(t) p6(t)        =        x1(t) x2(t) x3(t) α1(t) α2(t) α3(t)        A multibody system is a composition of several rigid bodies connected to each other and satisfying some physical or geometrical constraints.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 2 / 25

Coordinates

A discrete (or “atomic”) rigid body B is composed by a finite set of N geometrical points Pi, each one defined in the 3D space by its position vector with reference to some reference frame (usually the body frame)

xi(t) = xi1(t) xi2(t) xi3(t)

• i = 1,...,N

The body B is globally characterized by M = 3N quantities.

χ(t) =        x1(t) . . . xk(t) . . . xN(t)        =                   x11(t) x12(t) x13(t) . . . xk1(t) xk2(t) xk3(t) . . . xN1(t) xN2(t) xN3(t)                   =          χ1(t) χ2(t) . . . χj(t) . . . χM(t)          ∈ X ⊆ RM X = configuration space

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 3 / 25

Coordinates

We can express the χi coordinates in many different ways; between the two representations we can define a transformation χ′ = f(χ). The transformation f(·) must be non singular almost everywhere, i.e., the transformation jacobian ∂fi ∂χj

• must be full rank ∀χj, with a possible

exception of a countable set of configurations.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 4 / 25

Coordinates

Example Consider a point P described by cartesian coordinates x = x1 x2 x3 T

• r by polar/spherical coordinates x′ =

x′

1

x′

2

x′

3

T, where x1 = x x2 = y x3 = z x′

1 = ρ

x′

2 = θ

x′

3 = φ

In this case the transformations between x′ and x are defined as follows f :      x1 = x′

1 sinx′ 2 cosx′ 3

x2 = x′

1 sinx′ 2 sinx′ 3

x3 = x′

1 cosx′ 2

f−1 :          x′

1 =

• x2

1 +x2 2 +x2 3

x′

2 = arctan(

• (x2

1 +x2 2)/x3)

x′

3 = arctan(x2/x1)

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 5 / 25

Coordinates

The Jacobian Jx of the transformation f is Jx =    sinx′

2 cosx′ 3

ρ cosx′

2 cosx′ 3

−ρ sinx′

2 sinx′ 3

sinx′

2 sinx′ 3

ρ cosx′

2 sinx′ 3

ρ sinx′

2 cosx′ 3

cosx′

2

−ρ sinx′

2

   with the determinant detJx = ρ2 sinx′

2

If ρ = 0 the determinant goes to zero only for θ = x′

2 = 0±2kπ; this

configuration is called a singular configuration. For example, the spherical coordinates are useful to model a satellite orbit motion around Earth.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 6 / 25

Constraints

In general, kinematic constraints active on the body elements are defined by implicit function of the M = 3N coordinates, and possibly also of time t, as: ψ(χ1,...,χ3N,t) = 0 If the constraint are nc, a system of nc equalities arises ψ1(χ1,...,χ3N,t) = 0 ψ2(χ1,...,χ3N,t) = 0 . . . ψnc(χ1,...,χ3N,t) = 0 that is equivalent to the following matrix equation Ψ(χ(t),t) = 0. where Ψ is a nc ×1 matrix containing the nonlinear functions of the coordinates.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 7 / 25

Constraints

It is also possible to write constraints involving the time derivative ˙ χ(t) Ψ′(χ(t), ˙ χ(t),t) = 0. and, in general, some constraints can be expressed as inequalities Ψ′′(χ(t), ˙ χ(t),t) ≤ 0. A direct time dependency is present when some constraints are varying according to an external time law, otherwise the constraints depend from time only through the coordinates χi(t). The various types of constraints can be written as: a) Ψ(χ(t),t) = 0 b) Ψ′(χ(t), ˙ χ(t),t) = 0 c) Ψ′′(χ(t), ˙ χ(t),t) ≤ 0 The constraints that directly depend on time are called rheonomic, while the time-independent ones are called sclerononomic.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 8 / 25

Constraints - An example

Example The rigid system is composed by N = 4 point masses, with x1 = 0T x2 = 1 0T x3 = 1 0T x4 = 1

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 9 / 25

Constraints - An example

The rigid constraints are expressed as (x1 −x2)T(x1 −x2)−d2

12 = 0

(x1 −x3)T(x1 −x3)−d2

13 = 0

(x1 −x4)T(x1 −x4)−d2

14 = 0

(x2 −x3)T(x2 −x3)−d2

23 = 0

(x2 −x4)T(x2 −x4)−d2

24 = 0

(x3 −x4)T(x3 −x4)−d2

34 = 0

where dij is the distance between the point masses. There are 3N = 12 configuration variables and N(N −1)/2 = 6 constraint equations, all independent. The three oriented segments xi −x1 form a basis of mutually orthogonal vectors, and are the ideal representation of a cartesian reference frame, the most simple example of a rigid body. The system has therefore only 3N −nv = 12−6 = 6 free parameter; this number is the maximum number of dof of a rigid body in space.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 10 / 25

Generalized Coordinates

From now on, we assume that all nc constraints are independent. The implicit function theorem guarantees that it is always possible to express nc variables as functions of the n = M −nc remaining ones. We can therefore identify n = M −nc independent variables q1, q2,..., qn. These variables are called generalized coordinates q(t) =   q1(t) . . . qn(t)   ∈ Q They univocally represent the motion of a multibody system, implicitly taking into account the kinematic constraints acting on the system. All the other nc configuration variables can be computed from them, using the constraint equations.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 11 / 25

Generalized Coordinates

The set of independent generalized coordinates is not unique: many other sets of coordinates may represent the system motion. The set must be independent (no generalized coordinates qi shall exist that can be

• btained as linear combinations of other generalized coordinates)

complete (the motion of the constrained set is completely determined by the generalized coordinates included in the set) If the set is complete and independent, it is also minimal. The number n defines the dimension of the generalized coordinate space Q.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 12 / 25

Generalized Coordinates

It is always possible to express each position vector xi, with i = 1,...,M, as a function of the n generalized coordinates xi = hi(q1,q2,...,qn,t) = hi(q(t),t) where hi is a generic nonlinear vector function, whose derivatives with respect to its arguments exist up at least to the second order. Similarly, if we consider the configuration variables χ, we can set the following transformation between q and χ: χ = g(q1,q2,...,qn,t) = g(q(t),t)

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 13 / 25

Generalized velocities

The generalized velocities are defined as ˙ q(t) = dq(t) dt = ˙ q1(t) ... ˙ qn(t)T The configuration velocities ˙ χ are defined as ˙ χ(t) = dχ(t) dt = ˙ χ1(t) ... ˙ χM(t)T with M = 3N.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 14 / 25

Generalized velocities and accelerations

The relation between ˙ χ and ˙ q is ˙ χ(t) = J(t)˙ q(t)+b(t) where J ∈ RM×n and b ∈ RM×1 are defined as [J]ij = ∂fi(t) ∂qj(t) [b]i = ∂fi(t) ∂t J is called the transformation Jacobian; b is non zero only if χ directly depends from time. The generalized accelerations are ¨ χ(t) = J(t)¨ q(t)+ ˙ J(t)˙ q(t)+ ˙ b(t)

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 15 / 25

Generalized constraints

The constraints obtained considering the generalized coordinates and velocities are called generalized constraints. If the constraints depend directly on time they are written as Φ(q(t), ˙ q(t),t) = 0

• r

Φ(q(t), ˙ q(t),t) ≤ 0 If they do not depend directly on time they are written as Φ(q(t), ˙ q(t)) = 0

• r

Φ(q(t), ˙ q(t)) ≤ 0

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 16 / 25

Virtual Displacements and Constraints

Virtual displacements or admissible variations δr are a small (i.e., virtual, not real) displacements of body points, allowed by the kinematic constraints. Virtual displacements can take place independently from time, i.e., are not subject to the law of physics. We assume δt ≡ 0. For every generalized coordinate qi there is a virtual displacement δqi. The number ndof of independent and complete virtual displacements δqi defines the degrees-of-freedom of the multibody system. Usually the number n of independent and complete generalized coordinates qi is equal to the degrees-of-freedom n = ndof. However, this is not always the case, and depends on the type of constraints; when the constraints are non-holonomic ndof < n. The term holonomic come from the Greek oλoς and means integer, integrable.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 17 / 25

Non-holonomic constraints

Let us consider first the equality constraints, and, in particular, those that depend only on the positions Φ(q(t),t) = 0 The equality constraints that depend only on the positions are always holonomic constraints. Non-holonomic constraints belong to two classes of constraints: Inequality constraints: Φ′(q(t), ˙ q(t),t) ≤ 0 Equality constraints that depend also on velocities: Φ′′(q(t), ˙ q(t),t) = 0 but are not exactly integrable.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 18 / 25

Non-holonomic constraints

For simplicity and without loss of generality, we restrict our attention to the constraints that depend only on the generalized velocities, Φ′′(˙ q(t),t) = 0 When these differential equations do not provide an exact integral, they represent non-holonomic constraints.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 19 / 25

Non-holonomic constraints

Take a generic i-th holonomic constraint φi(q(t),t) = 0, and derive it with respect to time to obtain the corresponding constraint expressed as a function of the velocities d dt φi(q(t),t) = 0 ⇔ φ ′′

i (˙

q(t),t) = 0 ⇔ a(q)T ˙ q+b(q) = 0 where a(q) =       ∂φi(t) ∂q1(t) . . . ∂φi(t) ∂qn(t)       b(q) = ∂φi(t) ∂t The two constraints φi(q(t),t) = 0 and φ ′′

i (˙

q(t),t) = 0 are equivalent, since φi can be obtained integrating φ ′′

i .

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 20 / 25

Non-holonomic constraints

The constraint written in differential form: a(q)Tdq+b(q)dt = 0 represents the so-called Pfaffian form. If the Pfaffian form is integrable, i.e., if it represent an exact differential, it can substituted by its integral: in this case the constraint is holonomic. If on the contrary the Pfaffian form is not an exact differential, it cannot be integrated and the corresponding constraint is non-holonomic.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 21 / 25

Exact differentials

Given the differential form dφ = a(q)Tdq it is an exact form in Rn if

dφ does not depend on the integration path.

This is true when dφ = (∇φ)Tdq where (∇φ)T = (grad φ)T = ∂φ ∂q1 ··· ∂φ ∂qn

• Therefore the coefficients a(q) must satisfy the relation

ai(q) = ∂φ ∂qi , i = 1,...,n

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 22 / 25

Exact differential form

These identities between the second partial derivatives hold ∂ 2φ ∂qjqi = ∂ 2φ ∂qiqj , ∀i,j = 1,...,n and this implies ∂ai(q) ∂qj = ∂aj(q) ∂qi , ∀i,∀j = 1,...,n If the coefficients ai’s satisfy all the above relations, the differential form is integrable and the constraint is holonomic. Otherwise the form is not exactly integrable and the constraint is non-holonomic. From a physical point of view, the classical examples are those of a wheel rolling on a plane without slippage between wheel and plane at the contact point, or certain kind of sliding object, such as the ice skates.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 23 / 25

Virtual work

Virtual displacements are important for the definition of the virtual work δW . Given a system consisting of N point masses, each defined by a position vector ri, on which acts a system of N forces fi, applied on the system and having their application point in ri, the virtual work δW is defined as: δW =

N

∑

i=1

fi ·δri ≡

N

∑

i=1

fT

i δri

The system is said to be in static/dynamic equilibrium if the virtual work

• f the static/dynamic forces is zero, i.e., if

δW =

N

∑

i=1

fi ·δri = 0

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 24 / 25

Virtual work

If both Nf linear forces and Nτ rotational moments act on the system, it is necessary to distinguish the relative contributions, as follows δW =

Nf

∑

i=1

fi ·δri +

Nτ

∑

i=1

τi ·δαi = 0 where now we have introduced the moments τi and the virtual angular displacements δαi.

• B. Bona (DAUIN)

Generalized coordinates and constraints Semester 1, 2016-17 25 / 25