schlagenhauf/ocp-modules

ocp-modules

This library provides a set of modules to assemble optimal control problems (OCPs) by generating symbolic expressions based on casadi.

Installation

For Users

ocp-modules is (SOON) available via pip:

pip3 install --user ocp-modules

For Developers

1. Install setuptools and virtualenv:

   pip3 install --user setuptools virtualenv

2. Create a new virtual environment in the folder venv and source it:

   git clone <this_repo>
   cd <this_repo>
   python3 -m virtualenv venc
   source ./venv/bin/activate

3. Install ocp-modules via symlink into the virtual environment

   cd <this_repo>
   pip3 install -e .

A note regarding CasADi

Casadi can make use of proprietary solvers such as HSL_MA57 Follow the instructions obtaining_HSL to install them and modify the solver options to use the built-in default.

Common issues with HSL solvers:

• Cannot find libhsl.so. In this case, manually create a symlink libcoinhsl.so -> libhsl.so in /usr/lib
• Failed to load libmetis.so: It is hard to get a working instance of metis (the version packaged in Ubuntu 18.04 is not compatible). It seems to work without, but spams into stdout.

Running the examples

Run examples via:

./examples/<folder>/<script>.py

ocp-modules's Issues

The @ operator does not work anymore with new casadi versions

the @ operator does not work anymore with new casadi versions.

Implement Affine Constraints

Required to implement constraints resulting from linearization, should take the form

or

and

or (and this is the most flexible and general version

where and are constant

Implement better solution for auxilliary data handling

Currently, collecting internal data (dubbed "metadata") will create huge arrays until the memory is full.

Implement linear cost terms

For slack variables and linear programs a linear cost terms would be nice.