k_aug (kay-ogg)

Basic feature list:

• Sensitivity matrices with K*dsdp = [d2Ldxdp^T dcdp^T]
• Reduced Hessian matrices
• Hessian and Jacobian computation via ASL.
• dot_sens computations (for sensitivity)

k_aug is compatible with Pyomo and AMPL via ASL. The main functionality can be used through suffixes.

Requirements

• Ipopt Version 3.13.X, with its thirdparty libraries.
• In some systems zlib might be required. For example, in ubuntu one can get it by using sudo apt install zlib1g-dev. On a mac brew install zlib reportedly works.
• gcc, g++, gfortran compilers
• cmake

Requirements [CYGWIN 2.900]

• cmake
• gcc-core
• gcc-gfortran (gfortran)
• gcc-g++
• git
• make
• wget
• zlib-devel
• patch
• openblas
• lapack-devel
• pkg-config

Dependencies (NEW!)

It is now assumed that Ipopt from the coin-or project has been previously compiled with HSL. This will enable MA57 as the default linear solver. However, it is crucial to verify that the coinor libraries have been generated in standard locations. Namely, /usr/local/lib, and the libraries include:

• coinasl
• coinmetis
• coinhsl

Depending on your platform, these are typically named libcoinX.so or libcoinX.dll.a, where X is equal to asl, metis or hsl.

Installation

1. Visit the Ipopt repository, follow the instructions and make sure it is compiled, with all the thirdparty libraries.
2. At the root directory use cmake to generate the makefile e.g. cmake CMakeLists.txt
3. Run make
4. Check the bin directory to find the k_aug executable
5. (Windows) add the LAPACK library directory to the PATH (Cygwin typically /usr/lib/lapack, as well the libcoinX.dll.a files (typically located at /usr/local/lib/)

Known issues

• AMPL can not recognize command line options
• Mac os is currently supported (I think!).

So far, dozens of times tested .

k_aug is an essential part of the NMPC-MHE framework(caprese). Written by David Thierry 2020, under BSD 3-Clause license.