Optimization reaches minimal step length

[mpt0816]

Hello！ I has a problem when solving an MPC problem. The Hpipm solver returns code 2, similar to issue #56. Strange to say, it would return code Sucess when removing the first general constraint. But, the MPC problem meet the first constraint. In addition, the MPC problem can be solved by OSQP. Switching to the robust mode and reducing the exit tolerances do not work.

[mpt0816]

This is the statistics of Hpipm slover.

[giaf]

From the statistics it looks like the iterate are diverging straight away (see column res_comp), even if the linear KKT system is solved with good accuracy (see last 4 columns lin res ...). If you can share the QP data produced with codegen I can try to reproduce the issue and check what is going on.

[mpt0816]

thans for your reply! it is hard for me, if the QP data you expected is https://github.com/giaf/hpipm/blob/master/examples/c/data/getting_started_data.c. I can provide QP data as csv file, but you have to handle the data with extra work. I use https://github.com/mayataka/hpipm-cpp as a interface of Hpipm, the problem raised before is solved when I use hpipm as https://github.com/giaf/hpipm/blob/master/examples/c/example_d_ocp_qp.c, but the same problem appears when I using in a more complicated MPC. Is it possible to use csv format data? And, Can you check if there is any problem with code wrapper https://github.com/mayataka/hpipm-cpp as a interface of Hpipm? Thanks for your help！

[giaf]

You don't have to write the data files by hands, you just need to add the three function calls here https://github.com/giaf/hpipm/blob/master/test_problems/test_d_ocp.c#L1301 (and next 2 lines) in the source code where the IPM solver is called. This will create a data file with a copy of the QP and the solver arguments such that it can be reproduced easily. Currently there is not a CSV interface so that would be quite a bit of work to implement.

I'm not familiar with C++ so I won't be able to check if there is any problem in the code wrapper just looking at it. If there is any hint at some specific issue with the wrapper it can be investigated more easily.