Can you help me find the bug in my optimal control code?

  • Thread starter Thread starter enroger0
  • Start date Start date
  • Tags Tags
    Control Nonlinear
AI Thread Summary
The discussion revolves around using local linear control for trajectory control in nonlinear systems, emphasizing that linear controllers can be effective within certain limits. It highlights the importance of ensuring the trajectory is defined by continuous position, speed, and acceleration functions to facilitate easier tracking. The user is seeking assistance with an optimal control problem, specifically identifying bugs in their Fortran code that implements a fourth-order Runge-Kutta method for state and adjoint equations. They note discrepancies between their results and those from a reference article, prompting a request for algorithm improvement suggestions. The conversation underscores the challenges of nonlinear control and the need for precise trajectory definitions in control systems.
enroger0
Messages
21
Reaction score
0
I've been looking at nonlinear control system and fuzzy control in particular.

If I simply want to do trajectory control (have the system to follow a certain trajectory in phase space), can I use much simpler local linear control?

I mean, for each point in phase space, do a local linearization and find the local linear control transfer function, as long as the system stay close to the intended trajectory this would work right?
 
Engineering news on Phys.org
What to you refer to by "phase space"?

And yes linear controllers usually can be used successfully within certain limits on nonlinear systems.
Most real systems are nonlinear, some much more so than others.
 
enroger0 said:
I've been looking at nonlinear control system and fuzzy control in particular.

If I simply want to do trajectory control (have the system to follow a certain trajectory in phase space), can I use much simpler local linear control?

I mean, for each point in phase space, do a local linearization and find the local linear control transfer function, as long as the system stay close to the intended trajectory this would work right?

Is this trajectory fixed, or will it change in time?

First the trajectory setting.

It is easier to follow a trajectory if the path has the following conditions:

1.- position follows a continiuos function through the path. let's say x=f(x).
2.- speed function is continuous in the segment, then df(x) continous
3.- aceleration ddf(x) is continous.

this will prevent from trying to follow a difficult or impossible trajectory.

Choose the input variable to use as control input. Usually we use time in order to decide the position. This would mean that time will be use in order to find the error, and you will have some errors in position. perhaps you want to input position and try to accomplish the positioning in the time required, giving you errors in time and no in position. I usually use position rather than time because I prefer to have erros in time and no in position.


If you can create a plant or function on what "should" be the inputs in order to follow the trajectory (Output), use it as the main entrance, then add a PI controll and you will have a good result.
 
Hello ,Guys!
I have been trying to solve an optimal control problem that needs a code written for it.I have used the forward RK 4th order method (for the state variable) and backward RK 4th order method (for the adjoint equation) but the results I obtained are not right.I am saying this because the article I found the problem from includes the graph of the solution which is nothing like mine .Now I want you to help me find the bug if any or suggest a better algorithm.
Here is the Fortran code:
!===========================
PROGRAM opct
IMPLICIT NONE
!Declaration
real,DIMENSION(1010)::t,z,eta,w,u,x_1,x_2,A,B,C,D
INTEGER::k,i,j
real :: h,n,k1,k2,k3,k4,l1,l2,l3,l4,dz,deta,tol
real:: gama,g_1,g_2,phi_1,phi_2,z_0,eta_T,z_1,z_2,u_1,u_2
!Initial conditions
x_1(1)=427.22;x_2(1)=157.10
z(1)= x_2(1)/x_1(1)
!z(1)=0.37
phi_1=0.14;phi_2=0.1;gama=0.5;g_1=0.67;g_2=1.0;u_1=0.01;u_2=1.0
u(1)=0.05;n=1000.00;t(1)= 0.0;z_1=0.1033; z_2=31.5502
t(1001) = 5.0;eta(1001)=0.0;h=(t(1001)-t(1))/n;tol=0.001
w(1)= (phi_1 + phi_2 * (z(1)**gama))*z(1)/(g_1 * z(1) + g_2)
!print*,w(1)
!The header of the output
write(19,100)
100 format(5X,"t",10X,"z",9X,"eta",8X,"u")
!201 format (2f10.6,2f10.6,2f16.6,2f10.6)
!do i = 1,1001
!Do loop of Runge-Kutta fourth order forward method to get state variable
do i = 1,1001

do j = 1,1001
k1 = h*dz(t(j),z(j))
k2 = h*dz(t(j) + h/2.0,z(j) + k1/2.0)
k3 = h*dz(t(j) + h/2.0,z(j) + k2/2.0)
k4 = h*dz(t(j) + h,z(j) + k3)
!The values of z will be calculated
z(j+1) = z(j) + (k1 + 2.0*k2 + 2.0*k3 + k4)/6.0
t(j+1) = t(j)+ h !Getting set the time for the next iteration
w(j+1) = (phi_1 + phi_2 * (z(j+1)**gama))*z(j+1)/(g_1 * z(j+1) + g_2)
!print*, k1,k2,k3,k4
A(j)=k1;B(j)=k2;C(j)=k3;D(j)=k4
end do ! for z
!print*,w
!end do
!write(19,101)
!101 format(5X,"A",10X,"B",9X,"C",8X,"D")
!202 format (2f10.6,2f10.6,2f10.6,2f10.6)

!Do loop of Runge-Kutta fourth order backward method to get adjoint variable
!do i=1001,2,-1
do k = 1001,1,-1
l1 = h*deta(t(k),z(k),eta(k))
l2 = h*deta(t(k) - h/2.0,z(k)+A(k)/2.0,eta(k) +l1/2.0)
l3 = h*deta(t(k) - h/2.0,z(k)+B(k)/2.0,eta(k) + l2/2.0)
l4 = h*deta(t(k) - h,z(k)+ C(k),eta(k) + l3)
!===========================
!The values of lambda will be calculated
eta(k-1) = eta(k) - (l1 + 2.0*l2 + 2.0*l3 + l4)/6.0
t(k-1)=t(k)-h !Getting set the time for the next iteration
!write(8,*) l1,l2,l3,l4
!write(9,*) eta(k)
end do !for eta
!end do
!print*,eta
!Do i=1,1001
if (eta(i) .lt. (g_1/g_2)) then
u(i) = u_1
else if (eta(i).eq.(g_1/g_2)) then
if (z(i) .le. z_1) then
u(i)= u_1
else if (z_1 .lt. z(i) .and. z(i) .lt. z_2) then
u(i)= w(i)

else
u(i)= u_2
end if
else
u(i)=u_2
end if
end do !for u
!======================================= saved slopes
do j=1,100
write(*,*)j,A(j),B(j),C(j),D(j)
end do
!=======================================

!if (abs(x(i)-x(i+1)) + abs(u(i)-u(i+1))+ abs(lamd(i)-lamd(i+1)) .lt. tol )then
! stop
!else
201 format (2f10.6,2f10.6,2f16.6,2f10.6)
do i=1,1001
write(19,201) t(i),z(i),eta(i),u(i)
end do
!end if
!print*,u
end program opct

!=======================================================================================
function dz(t,z)
integer::i
real, intent(in)::t,z
real,DIMENSION(1010)::u
dz = -(phi_1 + phi_2*(z**gama))*z + (g_1*z + g_2)*u(1)
end function dz
!================================
function deta(t,z,eta)
integer::i
real, intent(in)::t,z,eta
real,DIMENSION(1010)::u
deta = ( phi_1 + (1-gama)*phi_2*(z**gama))*eta + g_2*u(1)*(eta-g_1/g_2)*eta-(gama*phi_2/(z**(1-gama)))
end function deta

!=================================================

Thanks a lot!
 
Hi all, I have a question. So from the derivation of the Isentropic process relationship PV^gamma = constant, there is a step dW = PdV, which can only be said for quasi-equilibrium (or reversible) processes. As such I believe PV^gamma = constant (and the family of equations) should not be applicable to just adiabatic processes? Ie, it should be applicable only for adiabatic + reversible = isentropic processes? However, I've seen couple of online notes/books, and...
I have an engine that uses a dry sump oiling system. The oil collection pan has three AN fittings to use for scavenging. Two of the fittings are approximately on the same level, the third is about 1/2 to 3/4 inch higher than the other two. The system ran for years with no problem using a three stage pump (one pressure and two scavenge stages). The two scavenge stages were connected at times to any two of the three AN fittings on the tank. Recently I tried an upgrade to a four stage pump...
Back
Top