Home Page

The Double Pendulum Problem

Franziska von Herrath & Scott Mandell May 19, 2000

Title Page

Contents

Abstract The double pendulum will be analyzed using Lagrangian and Hamiltonian methods. The systems behavior is then visualized using Matlabs ODE45 routine. Finally, the dierent behaviors are categorized.
Page 1 of 20

Go Back

Full Screen

Close

Quit

Contents
1 Introduction
Home Page

3 4 4 4 6 7 9 9 10 12 12 13 14 14 16 16

Title Page

Contents

2 Derivation 2.1 Position Equations . . . . . . . . 2.2 Energy Equations . . . . . . . . . 2.3 Lagrangian Function . . . . . . . 2.4 Lagranges Dierential Equations 2.5 Generalized Momenta . . . . . . 2.6 Hamiltonian Function . . . . . . 2.7 Hamiltons Equations of Motion .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

3 Finding a Numerical Solution 3.1 Creating a Function M-File . . . . . . . . . . . . . . . . . . . . . 3.2 Using ODE45 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Behaviors of the System 4.1 Periodic Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Quasiperiodic Motion . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Chaotic Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Page 2 of 20

Go Back

Full Screen

Close

Quit

1. Introduction
In order to analyze the system by using Newtons second law, all the forces acting on the system have to be known and be able to be expressed. Because the double pendulum problem is quite complex, Lagrangian and Hamiltonian equations are used to simplify the analysis. This approach is more suitable because it uses energy, a scalar quantity, as a basis for analysis rather than force, a vector quantity, and generalized coordinates of any suitable quantity, rather than the limiting rectangular, polar, or spherical coordinates. The Lagrangian formalism will render two second-order dierential equations that are functions of the generalized coordinates. These equations will be rewritten in Hamiltonian formalism which generates four rst-order dierential equations that are functions of the generalized coordinates and momenta. Furthermore, the use of Lagrangian and Hamiltonian formalism to analyze the system is justied as two important conditions are met:
Page 3 of 20

Home Page

Title Page

Contents

Go Back

1. The energy of the system is conserved, as the double pendulum rotates in the conservative force eld of the earth 2. Angular momentum is conserved as dissipative friction and air resistance forces are ignored

Full Screen

Close

Quit

2. Derivation
A double pendulum consists of two pendulums with the rst suspended from a point in space and the second suspended from the end of the rst. Consider the double pendulum shown in Figure 1 with masses m1 and m2 attached by rigid massless rods of lengths l1 and l2 . To simplify the analysis of the system, the systems motion will be constricted to a plane. Let the angles the two rods make with the vertical be denoted as 1 and 2 , and let gravity be given by g.

Home Page

Title Page

Contents

2.1. Position Equations

Then the positions of m1 and m2 in Cartesian coordinates are described by (x1 , y1 ) and (x2 , y2 ), respectively, and are given by

x1 = l1 sin 1 y1 = l1 cos 1
Page 4 of 20

(1) eq1 (2) eq1b (3) eq1c (4) eq1d

x2 = l1 sin 1 + l2 sin 2 y2 = l1 cos 1 l2 cos 2 .

Go Back

2.2. Energy Equations

The potential energy of the system is found by combining the potential energy of each mass as follows

Full Screen

Close

Quit

Home Page

y x1 1 l1 m1 y 1 2 y 2 l2 m2 x2 x

Title Page

Contents

Page 5 of 20

Go Back

Full Screen

Close

Figure 1: A double pendulum

Quit

V = m1 gy1 + m2 gy2
Home Page

(5)

= m1 g(l1 cos 1 ) + m2 g(l1 cos 1 l2 cos 2 ) = (m1 + m2 )gl1 cos 1 m2 gl2 cos 2 . (6)

Title Page

Contents

The kinetic energy of the system is given by 1 1 2 2 T = m1 v1 + m2 v2 2 2 1 1 = m1 (x2 + y1 ) + m2 (x2 + y2 ) 1 2 2 2 2 2 1 1 2 2 2 2 2 2 = m1 l1 1 + m2 [l1 1 + l2 2 + 2l1 l2 1 2 cos(1 2 )]. 2 2 (8)

(7)

Page 6 of 20

Go Back

2.3. Lagrangian Function

Full Screen

Since the motion of the system takes place in a conservative force eld, V is a function of the generalized coordinates and not the generalized velocities V =
Close

Quit

V (1 , 2 ). The Lagrangian function is dened as L T V.

Home Page

(9)

Thus the Lagrangian function of the generalized coordinates (1 , 2 ) is 1 1 2 2 2 2 L = (m1 + m2 )l1 1 + m2 l2 2 + m2 l1 l2 1 2 cos(1 2 ) 2 2

Title Page

+ (m1 + m2 )gl1 cos 1 + m2 gl2 cos 2 . (10) eq10

Contents

2.4. Lagranges Dierential Equations

Lagranges dierential equation for 1 is d dt L 1 L = 0. 1 (11) eq11

Page 7 of 20

From Equation 10, L 2 2 = m1 l1 1 + m2 l1 1 + m2 l1 l2 2 cos(1 2 ) 1 d dt L 1

2 = (m1 + m2 )l1 1 + m2 l1 l2 2 cos(1 2 )

(12)

Go Back

m2 l1 l2 2 sin(1 2 )(1 2 )
Full Screen

(13)

Close

L = l1 (m1 + m2 )g sin 1 m2 l1 l2 1 2 sin(1 2 ), 1

(14)

Quit

and substituting these in Equation 11, we obtain

2 (m1 + m2 )l1 1 + m2 l1 l2 2 cos(1 2 ) 2 + m2 l1 l2 1 sin (1 2 ) + l1 (m1 + m2 )g sin 1 = 0. (15) eq15

Home Page

Similarly, Lagranges dierential equation for 2 is

Title Page

d dt Once again from Equation 10,

L 2

L = 0. 2

(16)

Contents

L 2 = m2 l2 2 + m2 l1 l2 1 cos(1 2 ) 2 d dt
Page 8 of 20

(17)

L 2

= m2 l2 2 + m2 l1 l2 1 cos(1 2 ) m2 l1 l2 1 sin(1 2 )(1 2 ) (18)

Go Back

L = m2 l1 l2 1 2 sin(1 2 ) l2 m2 g sin 2 , 2 and substituting these in equation (16) gives

(19)

Full Screen

2 2 m2 l2 2 + m2 l1 l2 1 cos(1 2 ) m2 l1 l2 1 sin(1 2 ) + l2 m2 g sin 2 = 0. (20) eq20

Close

Equation 15 and Equation 20 are coupled second-order ordinary dierential equations that describe the motion of the system. Note that these two equations

Quit

depend on four unknowns. In order to solve the system numerically using Matlabs ODE45 routine, a system of four rst-order ordinary dierential equations is needed.
Home Page

2.5. Generalized Momenta

Since the system is closed, the equations of motion can also be written in the Hamiltonian formalism. This introduces the generalized momenta p1 and p2 , so the motion of the system depends on four initial conditions (1 , 2 , p1 , p2 ). Computing the generalized momenta gives p1 = L 2 = (m1 + m2 )l1 1 + m2 l1 l2 2 cos(1 2 ) 1 L 2 = m2 l2 2 + m2 l1 l2 1 cos(1 2 ). 2 (21) eq21

Title Page

Contents

p2 =

(22) eq22

2.6. Hamiltonian Function

Page 9 of 20

The Hamiltonian function of the generalized coordinates(1 , 2 ) and generalized momenta (p1 , p2 )is dened as H T +V The Hamiltonian can also be written in terms of the Lagrangian (23)

Go Back

Full Screen

Close

1 1 2 2 2 2 H = i pi L = (m1 + m2 )l1 1 + m2 l2 2 + m2 l1 l2 1 2 cos(1 2 ) 2 2 (m1 + m2 )gl1 cos 1 m2 gl2 cos 2 (24) eq24

Quit

Solving the generalized momenta, Equation 21 and Equation 22, for 1 and 2 and substituting back into equation Equation 24 gives H=
2 2 l2 m2 p2 + l1 (m1 + m2 )p2 2m2 l1 l2 p1 p2 cos(1 2 ) 2 2 2 2 2l1 l2 m2 [m1 + m2 sin2 (1 2 )]

Home Page

Title Page

(m1 + m2 )gl1 cos 1 m2 gl2 cos 2 (25)

Contents

2.7. Hamiltons Equations of Motion

Since the Hamiltonian is now a function of all the initial conditions 1 , 2 , p1 , and p2 , the equation can be separated into four rst-order dierential equations by taking partial derivatives with respect to the appropriate variables l2 p1 l1 p2 cos 1 2 H 1 = = 2 p1 l1 l2 [m1 + m2 sin2 (1 2 )] (26)

Page 10 of 20

H l1 (m1 + m2 )p2 l2 m2 p1 cos(1 2 ) 2 = = 2 p2 l1 l2 m2 [m1 + m2 sin2 (1 2 )] p1 = H = (m1 + m2 )gl1 sin 1 C1 + C2 1 H = m2 gl2 sin 2 + C1 C2 2

(27)

Go Back

(28)

Full Screen

p2 =
Close

(29)

Quit

C1
Home Page

p1 p2 sin(1 2 ) l1 l2 [m1 + m2 sin2 (1 2 )]

(30)

C2
Title Page

2 2 l2 m2 p2 + l1 (m1 + m2 )p2 l1 l2 m2 p1 p2 cos(1 2 ) 1 2 sin[2(1 2 )] 2 2 2l1 l2 [m1 + m2 sin2 (1 2 )]2

(31)

Contents

Page 11 of 20

Go Back

Full Screen

Close

Quit

3. Finding a Numerical Solution

3.1. Creating a Function M-File
Home Page

Solutions for 1 and 2 can be found by solving the system of rst-order dierential equations numerically using Matlabs ODE45 routine. The rst step is to create a function m-le in the Matlab editor. function xprime=doublependulum(t,x,flag,g,l1,l2,m1,m2)

Title Page

Contents

C1=(x(3).*x(4).*sin(x(1)-x(2)))./... l1*l2*(m1+m2*(sin(x(1)-x(2))).^2); C2=((l2^2*m2*(x(3)).^2+l1^2*(m1+m2)*... (x(4)).^2-l1*l2*m2*x(3).*x(4).*cos(x(1)-x(2)))... ./2*l1^2*l2^2*(m1+m2*(sin(x(1)-x(2))).^2).^2)... .*sin(2*(x(1)-x(2))); xprime=zeros(4,1); xprime(1)=(l2*x(3)-l1*x(4).*cos(x(1)-x(2)))... ./l1^2*l2*(m1+m2*(sin(x(1)-x(2))).^2); xprime(2)=(l1*(m1+m2)*x(4)-l2*m2*x(3).*cos(x(1)-x(2)))... ./l1*l2^2*m2*(m1+m2*(sin(x(1)-x(2))).^2); xprime(3)=-(m1+m2)*g*l1*sin(x(1))-C1+C2; xprime(4)=-m2*g*l2*sin(x(2))+C1-C2; This m-le denes the system of four rst-order dierential equations, where flag sets the parameters of the system.

Page 12 of 20

Go Back

Full Screen

Close

Quit

3.2. Using ODE45

Below is an example of how to execute Matlabs ODE45 routine. First the function is initiated, then a time interval is given followed by the initial conditions of the system, and nally the parameters of the system are given values. Four numerical solutions are produced by this example, where x is a vector that contains the solutions, x = [1 (t), 2 (t), p1 (t), p2 (t)]. [t,x]=ode45(doublependulum,[0,25],[pi;pi;0;0],[],9.8,1,1,1,1);
Contents

Home Page

Title Page

To call up the numerical equation 1 , the rst column of the vector x is initiated x(:,1). Solutions for 2 , p1 , and p2 are produced by calling the appropriate column of the vector x. These solutions can then be plugged into the position Equation 1, Equation 2, Equation 3, and Equation 4 to model the behavior of the masses in the system. x1=sin(x(:,1)); y1=-cos(x(:,1)); x2=sin(x(:,1))+sin(x(:,2)); y2=-cos(x(:,1))-cos(x(:,2));

Page 13 of 20

Go Back

Full Screen

Close

Quit

4. Behaviors of the System

Depending on the given initial conditions, the double pendulum exhibits three types of behavior:
Home Page

Periodic motion
Title Page

Quasiperiodic motion Chaotic motion

Contents

4.1. Periodic Motion

In periodic behavior, the systems motion is cyclic. This occurs when the energy of the system is either small or extremely large. Figure 2 shows the trace of m2 when the system is given a substantial amount of energy. The mass stays on the same path indenitely, or the motion is periodic. The following is the code that produced Figure 2.
Page 14 of 20

Go Back

Full Screen

[t,x]=ode45(doublep,[0,0.7],[pi;pi;10;-8.898788999]); x1=sin(x(:,1)); y1=-cos(x(:,1)); x2=sin(x(:,1))+sin(x(:,2)); y2=-cos(x(:,1))-cos(x(:,2)); plot(x2,y2) axis([min(-2) max(2)... min(-2) max(2)])

Close

Quit

2
Home Page

1.5
Title Page

1 0.5 yaxis 0 0.5 1

Contents

Page 15 of 20

1.5 2 2

Go Back

0 xaxis

Full Screen

Figure 2: Periodic Behavior

Close

Quit

4.2. Quasiperiodic Motion

In quasiperiodic behavior, the systems motion is almost periodic. This behavior occurs at all energy levels, yet is rare at middle range energy values. Figure 3 shows the trace of m2 when the system is given appropriate values to produce quasiperiodic behavior. Here is the code that produced Figure 3. [t,x]=ode45(doublep,[0,1],[pi;pi;20;-50]); x1=sin(x(:,1)); y1=-cos(x(:,1)); x2=sin(x(:,1))+sin(x(:,2)); y2=-cos(x(:,1))-cos(x(:,2)); plot(x2,y2) axis([min(-2) max(2) ... min(-2) max(2)])

Home Page

Title Page

Contents

Page 16 of 20

Go Back

Full Screen

Close

Quit

4.3. Chaotic Motion

In chaotic behavior, the systems motion is unpredictable and non-repetitive. Two systems with almost identical initial conditions show very dierent long term behavior. The system produces chaotic behavior at all energy levels unless the system is given an innite amount of energy where only periodic and quasiperiodic behavior exist. Figure 4 is an example of chaotic behavior exhibited by the system. The code for Figure 4 follows. [t,x]=ode45(doublep,[0,25],[pi;pi;0;0]); x1=sin(x(:,1)); y1=-cos(x(:,1)); x2=sin(x(:,1))+sin(x(:,2)); y2=-cos(x(:,1))-cos(x(:,2)); plot(x2,y2) axis([min(-2) max(2) ... min(-2) max(2)])

Home Page

Title Page

Contents

Page 17 of 20

Go Back

Full Screen

Close

Quit

2
Home Page

1.5
Title Page

1 0.5 yaxis 0 0.5 1

Contents

Page 18 of 20

1.5 2 2

Go Back

0 xaxis

Full Screen

Figure 3: Quasiperiodic Behavior

Close

Quit

2
Home Page

1.5
Title Page

1 0.5 yaxis 0 0.5 1

Contents

Page 19 of 20

1.5 2 2

Go Back

0 xaxis

Full Screen

Figure 4: Chaotic Behavior

Close

Quit

References
[1] Arya, Atam P. (1998). Introduction to Classical Mechanics (2nd ed.). Upper Saddle River, NJ: Prentice Hall.
Home Page

Title Page

[2] Marion, Jerry B., & Thornton, Stephen T. (1988). Classical Dynamics of Particles & Systems(3rd ed.). San Diego, CA: Harcourt Brace Jovanovich. [3] Instsitute fur den Wissenschaftlichen (Producer). (1985). Planar Double Pendulum [Videotape]. Chicago, IL: International Film Bureau.

Contents

Page 20 of 20

Go Back

Full Screen

Close

Quit