Model Predictive Contouring Control

This repo contains an inprogress implementation of model predictive contouring control (MPCC) based on the ACADOS (kinematic model only) and forces PRO solver (both models). A registered licence for forces PRO is required to run this part of the repo.

Problem formulation

The mathematical formulation can be found here. The resulting controller is shown below.

ACADOS Setup

1. Clone this repo

2. Head to https://github.com/acados/acados and install acados following their instructions. Below is a brief summary.

clone acados and unpack

git clone [email protected]:acados/acados.git
cd acados
git submodule update --recursive --init

then build it

mkdir -p build
cd build
cmake ..  
make install

install dependencies

pip3 install casadi

make python bindings:

 cd <acados_root>/build
 cmake -DACADOS_WITH_QPOASES=ON ..
 make install -j4
 pip3 install <acados_root>/interfaces/acados_template
 export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:"<acados_root>/lib"
 export ACADOS_SOURCE_DIR=<acados_root>

Note: I ended up putting the last two commands into my .bashrc

It should look something like this:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:"/home/pw/acados/lib/"
export ACADOS_SOURCE_DIR=/home/pw/acados

Test installation:

cd <acados_root>/examples/acados_python/getting_started
python3 minimal_example_closed_loop.py

3. All set up! Currently, I am working in mpcc/scripts/acados try running "$python3 python_sim.py kin" or "$python3 python_sim.py dyn"

Debug

Currently there are issues with the dynamic model. The solver is reporting error code 3, which corresponds to the minimal step size in the QP solver being reached.

The system equations are described using CasADi in models.py (in the function dynamic_model()). The solver is set up in acados_settings.py (in the function acados_settings_dyn())