Access Violation when running locally

[mastwood]

I have been trying to solve an optimal control problem, so my settings are:

g.options.IMODE = 6  # optimal control
g.options.NODES = 3  # collocation nodes
g.options.SOLVER = 3 # solver (IPOPT)

The example code given in the documentation works, but my code does not run. This error doesn't seem to have to do with how my problem my set up, but I am not sure.

Error: Exception: Access Violation
Traceback: not available, compile with -ftrace=frame or -ftrace=full

Error: 'results.json' not found. Check above for additional error details
Traceback (most recent call last):
  File "gekko_solver.py", line 71, in <module>
    g.solve(disp=True,debug=0) # Solve
  File "C:\Users\Michael\AppData\Local\Programs\Python\Python38-32\lib\site-packages\gekko\gekko.py", line 2145, in solve
    self.load_JSON()
  File "C:\Users\Michael\AppData\Local\Programs\Python\Python38-32\lib\site-packages\gekko\gk_post_solve.py", line 13, in load_JSON
    f = open(os.path.join(self._path,'options.json'))
FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\Michael\\AppData\\Local\\Temp\\tmp5bbvzwdwgk_model0\\options.json'

[mastwood]

My full code is attached here:

from gekko import GEKKO
import numpy as np
import matplotlib.pyplot as pl

g=GEKKO(remote=False)

nt=101
g.time=np.linspace(0,10,nt)

u1=g.MV(value=0)
u2=g.MV(value=0)

y1=g.CV(value=5)
y2=g.CV(value=0)

y1f=g.FV()
y2f=g.FV()

g.Equation(y1f==50)
g.Equation(y2f==-50)
g.Connection(y1f,y1,pos2='end')
g.Connection(y2f,y2,pos2='end')

theta1=g.Var(value=0)
theta2=g.Var(value=0)

g.Equation(u1==theta1.dt())
g.Equation(u2==theta2.dt())

#ydot=6*pi*(v1+dot(y-x1,v1)*(y-x1)/dot(y-x1,y-x1))/sqrt(dot(y-x1,y-x1)) \
#    + 6*pi*(v2+dot(y-x2,v2)*(y-x2)/dot(y-x2,y-x2))/sqrt(dot(y-x2,y-x2)) 

def stokeslet(y1,y2,theta1,theta2,u1,u2):
    k=6*np.pi 
    v11=u1*g.sin(theta1)
    v12=-u1*g.cos(theta1)
    v21=u2*g.sin(theta2)
    v22=-u1*g.cos(theta2)
    x11=g.cos(theta1)
    x12=g.sin(theta1)
    x21=g.cos(theta2)
    x22=g.sin(theta2)

    d1=(y1-x11)*v11+(y2-x12)*v12
    d2=(y1-x11)**2+(y2-x12)**2
    d3=g.sqrt(d2)

    d4=(y1-x21)*v21+(y2-x12)*v22
    d5=(y1-x21)**2+(y2-x22)**2
    d6=g.sqrt(d5)

    ydot1=(k*(v11+d1*(y1-x11))/d2)/d3+(k*(v21+d4*(y1-x21))/d5)/d6
    ydot2=(k*(v12+d1*(y2-x12))/d2)/d3+(k*(v22+d4*(y2-x22))/d5)/d6
    return [ydot1,ydot2]

g.Equation(y1.dt()==stokeslet(y1,y2,theta1,theta2,u1,u2)[0])
g.Equation(y2.dt()==stokeslet(y1,y2,theta1,theta2,u1,u2)[1])

J=g.Var(value=0)
Jf=g.FV()

g.Equation(J.dt()==u1**2+u2**2)

g.Connection(Jf,J,pos2='end')
g.Obj(Jf)

g.options.IMODE = 6  # optimal control
g.options.NODES = 3  # collocation nodes
g.options.SOLVER = 3 # solver (IPOPT)

g.solve(disp=False,debug=0) # Solve

[APMonitor]

Here is a version of your code that works but with soft terminal constraints:

from gekko import GEKKO
import numpy as np
import matplotlib.pyplot as pl

g=GEKKO(remote=False)

nt=101
g.time=np.linspace(0,10,nt)

u1=g.MV(value=0); u1.STATUS=1; u1.DCOST=0.1  # add bounds? lb=-1,ub=1
u2=g.MV(value=0); u2.STATUS=1; u2.DCOST=0.1

y1=g.CV(value=5)
y2=g.CV(value=0)

theta1=g.Var(value=0)
theta2=g.Var(value=0)

g.Equation(u1==theta1.dt())
g.Equation(u2==theta2.dt())

#ydot=6*pi*(v1+dot(y-x1,v1)*(y-x1)/dot(y-x1,y-x1))/sqrt(dot(y-x1,y-x1)) \
#    + 6*pi*(v2+dot(y-x2,v2)*(y-x2)/dot(y-x2,y-x2))/sqrt(dot(y-x2,y-x2)) 

def stokeslet(y1,y2,theta1,theta2,u1,u2):
    k=6*np.pi 
    v11=u1*g.sin(theta1)
    v12=-u1*g.cos(theta1)
    v21=u2*g.sin(theta2)
    v22=-u1*g.cos(theta2)
    x11=g.cos(theta1)
    x12=g.sin(theta1)
    x21=g.cos(theta2)
    x22=g.sin(theta2)

    d1=(y1-x11)*v11+(y2-x12)*v12
    d2=(y1-x11)**2+(y2-x12)**2
    d3=g.sqrt(d2)

    d4=(y1-x21)*v21+(y2-x12)*v22
    d5=(y1-x21)**2+(y2-x22)**2
    d6=g.sqrt(d5)

    ydot1=(k*(v11+d1*(y1-x11))/d2)/d3+(k*(v21+d4*(y1-x21))/d5)/d6
    ydot2=(k*(v12+d1*(y2-x12))/d2)/d3+(k*(v22+d4*(y2-x22))/d5)/d6
    return [ydot1,ydot2]

g.Equation(y1.dt()==stokeslet(y1,y2,theta1,theta2,u1,u2)[0])
g.Equation(y2.dt()==stokeslet(y1,y2,theta1,theta2,u1,u2)[1])

J=g.Var(value=0)

g.Equation(J.dt()==u1**2+u2**2)

final = g.Param(np.zeros(nt)); final[-1]=1
g.Minimize(J*final)

g.Minimize(final*1e5*(y1-50)**2)
g.Minimize(final*1e5*(y2+50)**2)

g.options.IMODE = 6  # optimal control
g.options.NODES = 3  # collocation nodes
g.options.SOLVER = 3 # solver (IPOPT)
g.options.MAX_ITER = 500

g.options.COLDSTART = 1
g.solve(disp=True) # Solve

g.options.COLDSTART = 0
g.options.TIME_SHIFT = 0
g.solve(disp=True) # Solve

pl.plot(g.time,u1.value,label='u1')
pl.plot(g.time,u2.value,label='u2')
pl.plot(g.time,y1.value,label='y1')
pl.plot(g.time,y2.value,label='y2')
pl.legend()
pl.show()

I tried adding some move suppression (DCOST=0.1) but it appears that the MV movement is quite dramatic near the end. It appears that it is still somewhat under-determined.

I've marked it as a bug and I'll add it to the list of things to look at. Even though hard terminal constraints typically don't solve as well, it shouldn't raise and exception.

[APMonitor]

The new function fix_final resolves this error.

#y1f=g.FV(value=50)
#y2f=g.FV(value=-50)

#g.Equation(y1f==50)
#g.Equation(y2f==-50)
#g.Connection(y1f,y1,pos2='end')
#g.Connection(y2f,y2,pos2='end')
g.fix_final(y1,50)
g.fix_final(y2,-50)

Here is the complete script that runs successfully. The solver returns the message that the problem is infeasible and no longer crashes.

from gekko import GEKKO
import numpy as np
import matplotlib.pyplot as pl

g=GEKKO(remote=False)

nt=101
g.time=np.linspace(0,10,nt)
z = 0.0
u1=g.MV(value=z)
u2=g.MV(value=z)

y1=g.CV(value=5)
y2=g.CV(value=z)

g.fix_final(y1,50)
g.fix_final(y2,-50)

theta1=g.Var(value=0)
theta2=g.Var(value=0)

g.Equation(u1==theta1.dt())
g.Equation(u2==theta2.dt())

#ydot=6*pi*(v1+dot(y-x1,v1)*(y-x1)/dot(y-x1,y-x1))/sqrt(dot(y-x1,y-x1)) \
#    + 6*pi*(v2+dot(y-x2,v2)*(y-x2)/dot(y-x2,y-x2))/sqrt(dot(y-x2,y-x2)) 

def stokeslet(y1,y2,theta1,theta2,u1,u2):
    k=6*np.pi 
    v11=u1*g.sin(theta1)
    v12=-u1*g.cos(theta1)
    v21=u2*g.sin(theta2)
    v22=-u1*g.cos(theta2)
    x11=g.cos(theta1)
    x12=g.sin(theta1)
    x21=g.cos(theta2)
    x22=g.sin(theta2)

    d1=(y1-x11)*v11+(y2-x12)*v12
    d2=(y1-x11)**2+(y2-x12)**2
    d3=g.sqrt(d2)

    d4=(y1-x21)*v21+(y2-x12)*v22
    d5=(y1-x21)**2+(y2-x22)**2
    d6=g.sqrt(d5)

    ydot1=(k*(v11+d1*(y1-x11))/d2)/d3+(k*(v21+d4*(y1-x21))/d5)/d6
    ydot2=(k*(v12+d1*(y2-x12))/d2)/d3+(k*(v22+d4*(y2-x22))/d5)/d6
    return [ydot1,ydot2]

g.Equation(y1.dt()==stokeslet(y1,y2,theta1,theta2,u1,u2)[0])
g.Equation(y2.dt()==stokeslet(y1,y2,theta1,theta2,u1,u2)[1])

J=g.Var(value=0)
Jf=g.FV()

g.Equation(J.dt()==u1**2+u2**2)

g.Connection(Jf,J,pos2='end')
g.Obj(Jf)

g.options.IMODE = 6  # optimal control
g.options.NODES = 3  # collocation nodes
g.options.SOLVER = 3 # solver (IPOPT)

g.solve(disp=True,debug=0) # Solve

EXIT: Converged to a point of local infeasibility. Problem may be infeasible.

 An error occured.
 The error code is  2

 ---------------------------------------------------
 Solver         :  IPOPT (v3.12)
 Solution time  :  15.677300000000002 sec
 Objective      :  0.0018925553482170296
 Unsuccessful with error code  0
 ---------------------------------------------------

 Creating file: infeasibilities.txt
 Use command apm_get(server,app,'infeasibilities.txt') to retrieve file
 @error: Solution Not Found