Python software for Strategies and Heuristics Regarding Lifting of Optimal Control problems (PySHeRLOC)

This software implements the benchmarks and algorithms described in [1].

Using blockSQP2

To use blockSQP2, you have to follow the installation instructions from [2]. Afterwards, change the path inside the file blocksqp_path.txt to your local installation:

# path to your local installation of blockSQP2:
{PATH_TO_LOCAL_INSTALLATION}/blockSQP2/Python

Using the CasADi version of blockSQP

To use the CasADi version of blockSQP, you have to configure the ma27 solver.

Getting started

First, create a new virtual python environment, e.g.,

python -m venv .venv

Then activate the virtual environment and install the necessary packages via:

source .venv/bin/activate
python -m pip install -r requirements.txt

Now you can start the solver with user interface by running

python GUI.py   # for the normal optimal control problems
python OED_GUI.py   # for OED problems

from the same folder as the file itself.

To benchmark all problems listed in benchmark_problems.txt, run:

bash run_all_alg_benchmarks.sh   # to benchmark all algorithms
bash run_all_def_benchmarks.sh   # to benchmark all shooting discretizations

Alternatively, you can run the corresponding benchmark with the desire arguments, e.g.,

# benchmarks the Bioreactor problem using the exact Hessian variant of blockSQP2
python benchmark_def.py -n "Bioreactor" -hess y -solver blockSQP2

# benchmarks the automatic condensing algorithm for Lotka OED using the Quasi-Newton variant of blockSQP2 
python benchmark_algs.py -n "Lotka OED" -hess n -fs n -cond y

The problem titles refer to the description on mintOC. They can be solved using IPOPT, blockSQP, blockSQP2, or fatrop. Benchmark problems can be found inside the Apps folder. Those include, among many others:

Bioreactor	Bioreactor Mayer	Bryson Denham	Bryson Denham Mayer
Cart Pendulum	Cart Pendulum Mayer	Catalyst Mixing	Cushioned Oscillation Mayer
Dielectrophoretic Particle Mayer	Double Oscillator	Double Oscillator Mayer	Ducted Fan
Egerstedt	Egerstedt Mayer	Electric Car	Electric Car Mayer
Fuller	Fuller Mayer	Hang Glider	Hanging Chain
Lotka Competitive	Lotka Competitive Mayer	Lotka Shared	Lotka Shared Mayer
Lotka Volterra	Lotka Volterra Mayer	LQR	Moon Landing
Mountain Car Mayer	Ocean	Quadrotor	Rao Mease
Three Tank	Three Tank Mayer	Van der Pol	Van der Pol Mayer
Lotka OED	Dielectr Particle OED	Jackson OED	Van der Pol OED

References

[1]: Lampel, R., Sager, S.: "On lifting strategies for optimal control problems"

[2]: Wittmann, R.: "blockSQP2"