zhangbaozhe / controllibrarydemo

This repo contains demo ROS code based on Control-Toolbox and ACADO Toolkit

Code Repo for Tutorial: Introducing Two Control Libraries -- Control Toolbox and ACADO Toolkit

This repo contains the following packages:

• control-toolbox (modified from upstream): from https://github.com/ethz-adrl/control-toolbox
  • ct: meta package
  • ct_core: core library
  • ct_optcon: optimal control library
• kindr: needed by control-toolbox
• invpend_experiment: containing the necessary description of a inverted pendulum, I forked from https://github.com/linZHank/invpend_experiment
• acado: containing ACADO source code. This folder will be ignored during catkin compilation
• control_demo: containing demo control programs using the two libraries written in C++
  • ct_quad_circle_mpc.cpp -> executable ct_quad_circle_mpc_node: (non-linear) model predictive control for quadrotor tracking a circle using control-toolbox
  • ct_invpend_pid.cpp -> executable ct_invpend_pid_node: using one PID feedback controller to control the inverted pendulum stay stable, but not working
  • ct_invpend_lqr.cpp -> executable ct_invpend_lqr_node: using linear quadratic regulator feedback controller to control the inverted pendulum staying stable, working, but less robust
  • ct_invpend_mpc.cpp -> executable ct_invpend_mpc_node: (non-linear) model predictive control for the inverted pendulum staying stable, working
  • acado_invpend_mpc.cpp -> executable acado_invpend_mpc_node: Linear MPC for the inverted pendulum staying stable, working

Installation Guide

First, make a workspace folder, say demo_ws

$ mkdir demo_ws

Next clone this repo as src folder

$ git clone https://github.com/zhangbaozhe/ControlLibraryDemo.git src

Then install the following dependencies:

• gflags:

  $ sudo apt install libgflags-dev

• catkin tools:

  $ sudo apt install python-catkin-tools
  or
  $ sudo apt install python3-catkin-tools

• Control Toolbox dependencies:
  • HPIPM & blasfeo: Go to src/control-toolbox/ct

    $ cd src/control-toolbox/ct
    $ bash install_hpipm.sh

• invpend_experiment dependencies:

  $ sudo apt install ros-noetic-ros-control ros-noetic-ros-controllers ros-noetic-gazebo-ros ros-noetic-gazebo-ros-control

• mavros:

  $ sudo apt-get install ros-noetic-mavros ros-noetic-mavros-extras
  $ wget https://raw.githubusercontent.com/mavlink/mavros/master/mavros/scripts/install_geographiclib_datasets.sh
  $ sudo bash ./install_geographiclib_datasets.sh 

In the workspace folder:

$ catkin build -DCMAKE_BUILD_TYPE=Release -j2

If you have more than 8GB RAM or even 16GB, you can maybe drop -j2 option. If you want compile the examples and tests in control toolbox you can type

$ catkin build -DCMAKE_BUILD_TYPE=Release -DBUILD_EXAMPLES=true -DBUILD_TESTS=true -j2

The compilation may take a while, since C++ templates compilation consumes huge resources.

(Optional) If you want to use the quadrotor demo, you need to use PX4-Autopilot simulation.

Executables

Run the inverted pendulum demo: In one terminal (in the workspace folder):

$ source devel/setup.bash
$ roslaunch invpend_control load_invpend.launch

In another terminal

$ source devel/setup.bash
$ rosrun control_demo ct_invpend_mpc_node