You are on page 1of 10

1

EML 6934, Spring 2010 Instructor: Dr. W. Dixon EXAM #2

RAMIN SHAMSHIRI UFID#: 9021-3353

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

2
I. Consider the following dynamic system = + + = where (), (), (), () , and () is a continuous unknown function that satisfies () < where is a known positive constant. Assuming only that () and () are measurable, what is the best result that you can obtain where the objective is 0? You can implement any controller - continuous or discontinuous. Show a complete analysis, including all signal chasing arguments to prove the control can be implemented. Problem 1

Solution: This can be considered as an integrator back-stepping problem since the control input is not mentioned explicitly in the dynamic. Adding and subtracting virtual control input into the dynamic, = + + + Let the back-stepping error be: = Substituting the back-stepping error, can be expressed as: = + + + The virtual control input can be designed as: = 1 Substituting (1.4) in (1.3) yield the closed loop dynamic = 1 + + Consider a positive definite, radially unbounded, decrescent candidate Lyapunov function = where = 1 2 2 (1.6) (1.5) (1.4) (1.3) (1.2) (1.1)

. Differentiating with respect to time yields = + = 1 + + + = 1 2 + . + +

where =

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

3
Designing the control input u(t) as: = 2 (1.7)

To show that this control input is implementable, since () and () are assumed to be measurable, therefore will be measurable as well (we can write using the statement for () and () as = + 1 + ). Also can be calculated as: () = 1 1 ( + + ()) = 1 1 + + = (1.8) 2

It can be seen that all the terms in are known and measurable, except the function , however we can use the upper bound for this function as given by the problem. Since () < , we can write as: = 1 1 + + 2 Substituting () as expressed in (1.7) into the expression of yields: = 1 2 2 2 + . () Since () < , we can write as: min 1 , 2 } 2 + { Letting 1 = 2 = , 2 2 +

(1.9)

(1.10)

(1.11)

(1.12)

0 => 2 + 0, which implies . In the other words, at some point will be such small that 2 will be dominated by . Therefore we can only conclude that the system is Globally Uniformly Stable, and we can show GUUB. Trying to complete the square in (1.10), = 2 2 1 2 = 2 2 1 We can upper bound
2 4 1

2 2 + + 2 1 41 41 2 2 + 21 41

(1.13)

by

2 4 1

, hence: 2 2 1 21
2

2 41

(1.14)

In order to get N.S.D result for , the constant (


2 4 1

2 4 1

can be made arbitrary small by choosing large values of 1 .

0 as 1 ). So it is Globally Uniformly Stable, and GUUB.

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

4
II. Consider the following dynamic system = + = + where (), (), (), () , and () is a continuous unknown function that satisfies () < where is a known positive constant. Assuming only that () and () are measurable, what is the best result that you can obtain where the objective is 0? You can implement any controller - continuous or discontinuous. Show a complete analysis, including all signal chasing arguments to prove the control can be implemented. Problem 2

Solution: Using integrator back-stepping approach, adding and subtracting virtual control input into the dynamic, = + + Let the back-stepping error be: = Substituting the back-stepping error, can be expressed as: = + + The virtual control input can be designed as: = 1 Substituting (2.4) in (2.3) yield the closed loop dynamic = 1 + Consider a positive definite, radially unbouded, decrescent candidate Lyapunov function = where = 1 2 2 (2.6) (2.5) (2.4) (2.3) (2.2) (2.1)

. Differentiating with respect to time yields = + = 1 + + + = 1 2 + + + ()

where = +

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

5
Designing the control input () as: = 2 (2.7)

This controller was shown to be implementable in the previous problem. Substituting () into the expression of yields: = 1 2 2 2 + . () Since () < , we can write as: min 1 , 2 } 2 + { Letting 1 = 2 = , 2 2 + (2.10) (2.9) (2.8)

Since we cannot clearly conclude anything regarding N.D or N.S.D from the expression for given in (2.10), we consider another positive definite, radially unbouded, decrescent candidate Lyapunov function as: 1 = 2 2 Differentiating with respect to time yields = = ( + ) Designing the control input () as: = 2 (2.13) (2.12) (2.11)

To show that this control input is implementable, since () and () are assumed to be measurable, therefore will be measurable as well (we can write using the statement for () and () as = + 1 + ). Also can be calculated as: () = 1 1 ( + ) = 1 1 + 2 = It can be seen that all the terms in are known and measurable. Substituting () as expressed in (2.13) into the expression for in (2.12) yields: = 2 2 + . () Since () < , we upper-bound . () by and write the expression for as: 2 2 + (2.15) (2.14)

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

6
Let 2 = 1 + 2 to complete the square in (2.14), = 1 2 2 2 = 1 2 2 We can upper bound
2 42

2 2 + + 2 42 4 2 2 2 2 + 22 42

(2.16)

by =

2 42

, hence: 1 2 + 21 + () 0 21 + (2.17)

And therefore, () can be written as: () ( 0 ) + (2.18)

The constant Cb can be made arbitrary small. It can be seen that x(t) is upper bounded. So it is Globally Uniformly Stable, and we can show GUUB.

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

7
III. Consider the following dynamic system = 2 + + + where (), () , and , , are positive unknown constants. Assume only that () is measurable, and given the objective to show 0? a. b. What is the best result that you can obtain with a continuous controller? What is the best result that you can obtain with a discontinuous controller? Problem 3

Show a complete analysis, including all signal chasing arguments to prove the control can be implemented. Solution: a. Implementing the following continuous control input into the dynamic yields: = 2 . sin = 2 + 2 . sin Where > 0, and =

(3.1) 2 . . sin

and . 2 = . sin . sin

= 1 2 + ( ) 2

Considering a positive definite, radially unbouded, decrescent candidate Lyapunov function as: = Differentiating with respect to time yields . sin = 1 3 + 2 2 sin Since sin 1, we can upper bound (3.3) by: 1 3 + 2 3 In order to have 0, we should have: 1 3 3 < 0 1 2 < 3 3 2 <
1

1 2 2

(3.2)

(3.3)

(3.4)

<

3 1

Also 2 < 0, 1 + < 0 > 1

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

8
b. Implementing the following discontinuous control input into the dynamic yields: = 2 + 2 + () where 1 , 2 > 0. Using the fact that dynamic yields, = and = 1, implementing the control input given by (3.5) into the 1 + 1 + 2 (2 )

(3.5)

= 2 + 2 + 2 + 1 + 1 + 2 2 2 + . 2 + () . 1 + . 1 + . 2 . (2 ) Since 1 , 2 > 0, we have 1 = 2 = . Letting = 1 2 and Simplifying (3.6): = 1 2 + + . 2 + Considering a positive definite, radially unbouded, decrescent candidate Lyapunov function as: = Differentiating with respect to time yields = = 1 3 + 2 + 2 Since sin 1, we can upper bound (3.9) by: 1 3 + ( + ) 2 In order to have 0, we should have: 1 3 < 0 > 1 Also ( + ) 2 < 0 + 1 < 0 > + 1

(3.6)

(3.7)

1 2 2

(3.8)

(3.9)

(3.10)

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

9
IV. Consider the following dynamic system = 2 + . sin + . tanh + = . tanh + = + . sin where (), , () , and , , are positive unknown constants. What is the best result that you can obtain with a continuous controller for the tracking objective = 0? for some given desired trajectory ()? Assume that only (), (), (), () are measurable, and that () is sufficiently smooth and bounded with bounded higher order derivatives. Show a complete analysis, including all signal chasing arguments to prove the control can be implemented. Solution: Define filter tracking error = + Open loop error system is: = + = 2 . sin . tanh() + Adding and subtracting virtual control input = 2 . sin . tanh() + 1 + where 1 = . Design : = 2 . sin + Then = . tanh() + 1 + = 2 . sin . cos + Differentiating the virtual control input 1 and adding and subtracting : 1 = + 1 = 2 . sin . cos + . tanh +2 Where 2 = . Plugging the expression for into 1 and rearranging yields: 1 = 2 2 + . sin + . cos + 2 . sin . tanh + + . tanh Problem 4

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

10
We can write 1 = 2 , Where = 2 + sin() + + 2 + . sin + Design : 1 = 2 = + 2 1 So, the closed loop error system is 1 = 2 2 1 Where = Considering a positive definite, radially unbouded, decrescent candidate Lyapunov function as: = Differentiating with respect to time yields 1 1 1 = + 1 1 + 2 2 1 2 2 2 Designing such that we get 0 1 2 1 2 1 2 1 1 + 1 + 2 + 2 2 2 2 . tanh() tanh () 2

Ramin Shamshiri

EML 6934, Exam #2

Due March.29.2010

You might also like