giulioturrisi / self-balancing-robot

Control algorithms for a self balancing robot (two wheeled inverted pendulum) using ROS2

Overview

This repo contains the code for controlling both a real and a simulated two wheeled inverted pendulum robot via ROS2.

List of available controllers

1. PID
2. Linear Quadratic Regulator
3. Linear Quadratic Integral Regulator
4. Least Square Linear Quadratic Regulator
5. Adaptive Linear Quadratic Regulator (MIT rule)
6. Sliding Mode
7. RL policy via IsaacGym
8. Nonlinear MPC via Acados
9. Iterative Linear Quadratic Regulator
10. Iterative Linear Quadratic Regulator via Crocoddyl
11. Predictive Sampling

List of available state estimators

1. Extended Kalman Filter

Repository structure

It includes the following folders and subfolders:

1. python_scripts: most of the ROS2 nodes call some classes here

2. coppeliasim_simulation: scenes used for simulating the robot

3. ros2_ws: collection of ROS2 nodes for controlling the robot and some external folders such as simExtROS2 and ros2_bubble_rob(dependencies for CoppeliaSim)

Dependencies

1. ROS2 Humble

2. CoppeliaSim for simulations

Build on Linux

1. clone the repo recursively

git clone --recurse-submodules https://github.com/giulioturrisi/Self-Balancing-Robot.git

2. install miniforge (x86_64)

3. create an environment using the file in the folder installation/conda:

    conda env create -f mamba_environment.yml

4. follow the instruction here to install ros-humble

5. download CoppeliaSim

6. add in your .bashrc

alias twip_env="conda activate twip_env && source your_path_to/Self-Balancing-Robot/ros2_ws/install/setup.bash"
export COPPELIASIM_ROOT_DIR=your_path_to/CoppeliaSim

7. start your environment and go in ros2_ws

twip_env
cd your_path_to/Self-Balancing-Robot/ros2_ws
rosdep install -y -r -q --from-paths src --ignore-src --rosdistro humble
ulimit -s unlimited
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release

8. if you need acados, go inside the acados/acados folder and press

mkdir build
cd build
cmake -DACADOS_WITH_QPOASES=ON  -DACADOS_WITH_OSQP=ON ..
make install -j4
pip install -e ./../interfaces/acados_template

then in your .bashrc, add

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:"/your_path_to/Self-Balancing-Robot/python_scripts/controllers/acados/lib"
export ACADOS_SOURCE_DIR="/your_path_to/Self-Balancing-Robot/python_scripts/controllers/acados"

How to run the simulation

1. Open Coppeliasim and run the scene self_balancing_robot.ttt in the folder coppeliasim_simulation

./coppeliaSim.sh -f your_path_to/Self-Balancing-Robot/coppeliasim_simulation/self_balancing_robot.ttt 

2. on a new terminal

ros2 run controllers <control_node>                     

where in <control_node> you can choose the type of controller you want.

3. to command the robot with a joystick

ros2 launch teleop_twist_joy teleop-launch.py joy_config:='xbox'

Status

Still working in progress, the real robot does not exist yet. Recently added IsaacGym to learn an RL policy (Self-Balancing-Robot-IsaacGym)!

self-balancing-robot's Issues

Can't launch the simulation with CoppeliaSim 4.6

Hi, I'm trying to use the simulation, through docker, following the readme but I get the error :

./coppeliaSim.sh -s self_balancing_robot.ttt 
the argument ('/home/coppeliasim_simulation/self_balancing_robot.ttt') for option '--auto-start' is invalid

when running the command :
./coppeliaSim.sh -s self_balancing_robot.ttt

I don't understand what the coppeliaSim.sh script does, but I belive my path file is ok.

Any hint ?