Scribd
Upload
130 views4 pages

Hw01sol PDF

This document contains the solutions to three homework problems for the optimal control course ECSE 6440. Problem 1 involves finding the closest point of approach and distance between a meteor's trajectory and a space station. The solution uses the Hamiltonian to find the optimal point as (1.71, 2.05) with cost 0.087. Problem 2 derives the closed-loop state and costate equations for an optimal control problem with a bilinear system and quadratic cost. Problem 3 considers a bilinear system with a final state constraint and derives: (a) the optimal control and equations, (b) the optimal trajectories for a example with N=2, and (c) reformulates it as a static optimization problem

Uploaded by

Dao Yan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
130 views4 pages

Hw01sol PDF

This document contains the solutions to three homework problems for the optimal control course ECSE 6440. Problem 1 involves finding the closest point of approach and distance between a meteor's trajectory and a space station. The solution uses the Hamiltonian to find the optimal point as (1.71, 2.05) with cost 0.087. Problem 2 derives the closed-loop state and costate equations for an optimal control problem with a bilinear system and quadratic cost. Problem 3 considers a bilinear system with a final state constraint and derives: (a) the optimal control and equations, (b) the optimal trajectories for a example with N=2, and (c) reformulates it as a static optimization problem

Uploaded by

Dao Yan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

ECSE 6440

Optimal Control

Jan 15, 2004

Homework Solution # 1
Due: Jan 26, 2004 (Monday)
1. (30%)
A meteor is moving along the path (Earth is at the origin)
y = x2 + 3x 6.
A space station is at (x, y) = (2, 2). Find the closest point of approach, (x, y), and the
distance from this point to the space station.
Solution:
This problem may be stated as:
min J := (x 2)2 + (y 2)2
x,y

subject to y = x2 + 3x 6, x 0, y 0.

Form the Hamiltonian:


H = J + f = (x 2)2 + (y 2)2 + 1 (x2 + 3x 6 y) + lambdax + 3 y.
First order condition:
H
x
H
y

= 0 2(x 2) + 21 x + 31 + 2 = 0
= 0 2(y 2) 1 + 3 = 0.

First assume the inequality constraints are not active, i.e., 2 = 3 = 0. Then
x=

4 31
, 1 = 2(y 2)
2(1 + 1

which implies
x =

3y + 8
4 6(y 2)
=
.
2(1 + 2(y 2))
2y 3

Substitute into the constraint equation y =2 +3x 6, we obtain


y=

3y 8
2y 3

!2

3y 8
3
6.,
2y 3

which may be solved to give three candidate solutions: {8.22, 2.05, 0.92}, and the
corresponding x: {1.68, 1.71, 4.53}. Take the positive solutions only, we therefore
obtain
(x , y ) = (1.71, 2.05).
Since the inequality constraints are indeed inactive, this is the optimal solution, and
the optimal cost is J = 0.087.
(Run hw01 1.m for the numerical solution and plot.)
1

2. (30%) Consider the bilinear system


x(i + 1) = Ax(i) + Dx(i)u(i) + bu(i)
where x(i) Rn , u(i) R, with quadratic performance index
J = x(N )T Q(N )x(N ) +

N
1 
X

x(i)T Qx(i) + ru(i)2

i=0

where Q(N ) 0, Q 0, r > 0. Find the optimal control, and the closed loop state
and co-state equations. Dont worry about the solution of these equations.
Solution:
First form the Hamiltonian:
Hi = xTi Qxi + ru2i + Ti+1 (Axi + Dxi ui + bui ).
Stationarity condition:
Hi
= 2rui b + (Ti+1 Dxi + b) = 0,
ui
which may be solved to give
1
ui = Ti+1 (Dxi + b).
2r
The costate equation is given by
i =

Hi
xi

!T

= 2Qxi + (AT + ui DT )i+1

with the boundary condition

= 2QN xN .
xN
The TPBVP corresponding to the optimal trajectory is given by
N =

Dxi T
(Dxi + b)
2r i+1
DT i+1 T
= 2Qxi + AT i+1
i+1 (Dxi + b)
2r
= 2QN xN , x0 given.

xi+1 = Axi
i
N

3. (40%)
Consider the following bilinear system:
x(i + 1) = x(i)u(i) + 1
with the performance index
J=
and final state constraint x(N ) = 0.
2

1
1 NX
u2
2 i=0 i

(a) (15%) Find the optimal control and the closed loop state and co-state equations.
(b) (15%) Suppose that x(0) = 1 and N = 2. Use the result from part 3a to find the
optimal state and co-state sequences, the optimal control, and the optimal value
of the performance index.
(c) (10%) Convert the optimal control problem in part 3b into a static optimization
and find the optimal control sequence and the optimal value of the performance
index.
Solution:
xi+1 = xi ui + 1, x0 = 1, x2 = 0.
(a) Form the Hamiltonian
u2i
+ i+1 (xi ui + 1).
2
i
Solve for the optimal control from H
= ui + i+1 xi = 0:
ui
Hi =

ui = i+1 xi .
Costate equation:
Hi
= i = i+1 ui .
xi
TPBVP:
xi+1 = i+1 x2i + 1, x0 , xN given
i = xi 2i+1 .
(b) For N = 2, x0 = 1, x2 = 0, we have
x1 = 1 + 1
x2 = 2 x21 + 1
= l2 (1 1 )2 + 1
1 = x1 22 .
Use 1 and x1 equations to solve for x1 , we get
x1 = 1 + 1 = x1 22 + 1 =
And then solve for 1 :
1 =

1
.
1 22

22
.
1 22

Substitute into x2 and simplify:


x2 =

2
+ 1 = 0.
(1 22 )2
3

Solve for 2 , we get two real solutions: {1.49, 0.52}. Compute J for each candidate
2 , we get J = {4.8, 0.67}. Therefore, the optimal 2 is 0.52, and 1 = 0.38,
u0 = 0.38, u1 = 0.72. the optimal state trajectory is
x0 = 1, x1 = 1.38, x2 = 0.
(c) The equivalent static optimization problem is
1
min (u20 + u21 )
2

u0 ,u1

subject to
0 = (u0 + 1)u1 + 1.
For the Hamiltonian:
1
H = (u20 + u21 ) + ((u0 + 1)u1 + 1).
2
First order condition:
u0 + u1 = 0
u1 + (u0 + 1) = 0.
Solve u0 , u1 in terms of :
2
1 2

=
.
1 2

u0 =
u1

Substitute into the constraint equation and solve for , we get the same two real
roots as 2 in the previous part: {1.49, 0.52}. Now check the second order condition:
"
#
2H
1
=
1
u2
which is positive definite if and only if 1 2 > 0. Therefore, the optimal is
0.52. Substituting into u0 and u1 , we obtain the same solution as before.

You might also like