Skip to content

Installation

  1. Clone or download ParNMPC from Github: ParNMPC.

  2. Extract the downloaded file.

    Requirements

    ParNMPC supports MATLAB R2016a or later, and requires the following tools:

    • MATLAB Coder
    • MATLAB Optimization Toolbox
    • MATLAB Parallel Computing Toolbox
    • MATLAB Symbolic Math Toolbox
    • C/C++ compiler supporting parallel code generation

    For compilers that do not support OpenMP, code generation will treat parallel computing as serial computing. Check Supported and Compatible Compilers for details.

    Warning

    ParNMPC may not work properly on MATLAB R2017b and R2018a because of the bug that exists in certain Windows 10 builds. Check sym bug report for details.

File Layout

The file layout of ParNMPC looks like this:

ParNMPC/                          % The toolkit folder.
docs/                             % The documents folder.
SemiActiveDamper/       % Example of controlling a semi-active damper.
    NMPC_Problem_Formulation.m    % NMPC formulation and initialization.
    Simu_Matlab.m                 % Closed-loop simulation in MATLAB.
    Simu_Simulink.slx             % Closed-loop simulation in Simulink.
    Simu_Simulink_Setup.m         % Setup for the Simulink simulation.
    Timer.m                       % Your own timer function.
    GEN_initData.mat              % Initial guess
DoubleInvertedPendulum/ % Example of swinging up a double inverted pendulum on a cart.
    ...                           % Files
Quadrotor/              % Example of controlling a quadrotor.
    ...                           % Files

Getting Started

This section shows how to do the closed-loop simulation in Simulink using MATLAB R2016a and Microsoft Visual C++ 2015 Professional as an example.

  1. Run the following MATLAB command and select the Microsoft Visual C++ 2015 Professional (C) compiler: 

    >> mex -setup

  2. Navigate to the Quadrotor/ folder. 

    >> cd  Quadrotor/

  3. Open NMPC_Problem_Formulation.m and run. By running this file, the following things are done:

    • The NMPC controller is defined and configured, and necessary files are automatically generated to the ./funcgen/ and ./codegen/ folders.
    • The very first OCP is solved and its solution is saved to GEN_initData.mat.
    • The controlled plant for simulation is defined and auto-generated.

  4. Open Simu_Simulink_Setup.m and run. By doing this, the NMPC controller is generated into C codes and compilied into a DLL file.

  5. Open Simu_Simulink.slx and run. Simu_Simulink.slx calls the generated NMPC controller function from the DLL file.

    Warning

    Running Simu_Matlab.m directly is not in parallel.