ECE 594D

Robot Locomotion

Winter 2010

Cart-pole system: Equations of motion Nonlinear Dynamics

This document provides a derivation of the equations of motion (EOM) for the cart-pole system. The (true) nonlinear dynamic equations are derived rst, using a Lagrangian approach; then the system is linearized about the upright equilibrium (inverted pendulum) position. Figure 1 shows the system. There are two degrees of freedom: the position of the cart, x, and the angle of the pendulum, . The system is underactuated, since there is only one actuation: a force, Fx , applied on the cart.

mp g

Figure 1: Cart-pole system. To begin, we will derive the simple relationships giving the position and velocity of the pendulum: xp = L sin xp = L cos + x yp = L cos yp = L sin The Lagrangian for a dynamic system is dened as: L = T V where T is the kinetic energy and V is the potential energy. For the cart-pole system: 1 1 p 2 T = M x2 + m p x2 + y p 2 2 1 1 2 = M x + mp L2 cos2 2 2L cos x + x2 + L2 sin2 2 2 2 1 1 = (M + mp ) x2 + mp L2 cos2 2 + L2 sin2 2 mp L cos x 2 2 1 1 = (M + mp ) x2 + mp L2 2 mp L cos x 2 2
(Last revised January 10, 2010)

(1) (2) (3) (4) (5)

(6)

Homework 1

ECE 594D

Robot Locomotion

Winter 2010

And the potential energy is: V = mp gyp = mp gL cos So the Lagrangian is: L= 1 1 (M + mp ) x2 + mp L2 2 mp L cos x mgL cos p 2 2 (8) (7)

Next, for each generalized coordinate, qn , and its associated actuating force (if any), n , we can derive an equation of motion: d L dL = n (9) dt qn qn For the x coordinate of the cart: n L x d L dt x L x = Fx = (M + mp ) x mp L cos = (M + mp ) xp + mp L sin 2 mp L cos =0 (10) (11) (12) (13)

Combining these expressions as required by Equation 8, we obtain one equation of motion: Fx = (M + mp ) x + mp L sin 2 mp Lcos For the (unactuated) angle of the pendulum, p , L d L dt L =0 = mp L2 mp L cos xp x = mp L2 mp L cos + mp L sin xp = mp L sin x + mp gL sin (15) (16) (17) (18) (14)

Again using Equation 8, we get the second equation of motion: 0 = mp L2 mp L cos + mp L sin xp mp L sin x mp gL sin x x = mp L2 mp L cos mp gL sin (19)

(Last revised January 10, 2010)

Homework 1

ECE 594D

Robot Locomotion

Winter 2010

Linearization
Assuming the pendulum remains near = 0, we can linearize the equations of motion given in 14 and 19. For the linearization, we assume: sin cos 1 2 0 Equation 14 can then be approximated as: Fx = (M + mp ) x mp L and equation 19 becomes: 0 = mp L + mp L2 mp gL x (20) (21)

Decoupling the equations of motion

It is trivial to rewrite the dynamics represented by 20 and 21 to give the following two (explicit) expressions for the accelerations of the state variables of the linearized system: M + mp g 1 + M L ML 1 mp x= g + Fx M M = Fx (22) (23)

After a bit of mathematical gymnastics, we can also rewrite the two nonlinear equations of motion (14 and 19) to express each acceleration explicitly: 2 sinmp L cos + sin gM + sin mp g + cos Fx = L (mp cos2 + M + mp ) mp L sin cos 2 + (M + mp ) g sin + cos Fx = (M + mp (1 cos2 )) L x= = sinmp L2 + sin mp cos g + Fx mp cos2 + M + mp mp L sin 2 + mp g sin cos + Fx M + mp (1 cos2 )

(24)

(25)

(Last revised January 10, 2010)

Homework 1