[Ipopt] Dual Feasibility is always high (feasible problem)

[Nittin Arora]

Hello all,

I am new to IPOPT and am experiencing issues getting any convergence. I
have checked that my derivatives are correct. I have not yet computed the
hessian code, so am using the limited-memory option.

I have an very very good guess of the solutions. The problem I am having is
that I get to stage where my constraint violations are zero and the
objective is fairly small but the "inf_du" is still large and hence the
optimizer keeps on going and ends with a "restoration failed" comment. Does
anyone know whats going on ? See below for first few and last few
iterations:
---

iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr
ls

   0  5.3875588e-10 2.05e+00 1.73e+01   0.0 0.00e+00    -  0.00e+00
0.00e+00   0

   1r 5.3875588e-10 2.05e+00 9.99e+02  -0.1 0.00e+00    -  0.00e+00
1.46e-13R 35

   2r 1.2827230e-01 0.00e+00 7.81e-02  -2.3 8.20e+02    -  1.00e+00
9.90e-04f  1

   3  2.4369526e-01 0.00e+00 3.55e+01  -2.2 7.86e-02    -  1.00e+00
1.00e+00h  1

   4  1.3480672e-01 0.00e+00 3.10e+01  -3.0 8.89e-02    -  1.00e+00
1.00e+00f  1

   5  1.2105967e-01 0.00e+00 4.22e+01  -2.7 1.76e+00    -  1.00e+00
2.25e-02f  6

   6  8.2795336e-02 0.00e+00 2.35e+01  -4.1 4.91e-02    -  1.00e+00
8.79e-01f  1

   7  9.0081218e-02 0.00e+00 4.22e+01  -5.5 3.37e-03    -  1.00e+00
1.00e+00h  1

   8  7.7203884e-02 1.92e-05 3.24e+01  -5.6 4.63e-03    -  1.00e+00
1.00e+00f  1

   9  7.4325412e-02 1.54e-05 3.49e+01  -7.3 2.56e-02    -  1.00e+00
2.50e-01f  3

.

.

.

iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du alpha_pr
ls

1100  8.9721257e-04 0.00e+00 1.69e+01 -11.0 3.06e-06    -  1.00e+00
2.50e-01f  3

1101  8.9718809e-04 0.00e+00 1.08e+01 -11.0 1.89e-06    -  1.00e+00
5.00e-01f  2

1102  8.9714023e-04 0.00e+00 1.63e+01 -11.0 2.36e-06    -  1.00e+00
2.50e-01h  3

1103  8.9701825e-04 0.00e+00 4.94e+00 -11.0 1.72e-06    -  1.00e+00
2.50e-01h  3

1104  8.9699042e-04 0.00e+00 6.59e+00 -11.0 2.47e-06    -  1.00e+00
2.50e-01h  3

1105  8.9696387e-04 0.00e+00 7.74e+00 -11.0 6.71e-06    -  1.00e+00
1.25e-01f  4

1106  8.9790578e-04 0.00e+00 6.14e+00 -11.0 8.75e-06    -  1.00e+00
1.00e+00H  1


Restoration phase is called at point that is almost feasible,

  with constraint violation 3.784646e-14. Abort.

---
I am using the following settings:
hessian_approximation="limited-memory"
max_iter=10000
tol=0.002
obj_scaling_factor=2.0
bound_mult_init_method="mu-based"
mu_init=1e-3

Thanks for all the help,
Nitin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20170804/e220c19a/attachment.html>