Convergence Properties of Fast quasi-LPV Model Predictive Control

General

This repository contains the simulation code to reproduce the tables and figures presented in

C. Hespe and H. Werner, "Convergence Properties of Fast quasi-LPV Model Predictive Control," 60th Conference on Decision and Control. Austin, TX, USA, 2021, doi: 10.1109/CDC45484.2021.9683612.

The code has two main entry points, the scripts benchmark_adip.m and benchmark_unicycle.m, both in the root directory of this repository. They will produce among others the material presented in the paper.

Prerequisites

To run the simulation files, some additional packages need to be installed first.

1. Install acados, e.g. according to this guide
2. Install CasADi into the external directory of acados, as described on the acados interface page
3. Create the environment variable ACADOS_INSTALL_DIR pointing to the main acados directory

The simulation code in this repository was tested in the following environment:

• Windows 10 Version 21H1
• Matlab 2021a
• acados 0.1.6
• CasADi 3.5.5

In the default configuration, you will also need the Matlab Coder. It is used to generate C code for the qLMPC solvers. If you do not have access to Matlab Coder and want to run the simulation nonetheless, you can disable code generation in the main simulation scripts by setting generate_mex = false.