MPC implementation with Python, solved by PANOC/IPOPT

This repository provides samples of Model Predictive Controller(MPC) implementation with python.

The options of optimization solver is PANOC and IPOPT.

Just change "solver_type" in simulation_setting.yaml as you like.

What is IPOPT?

It is a famous solver for a nonlinear optimization problem. See IPOPT official page for detail.

What is OpEn?

It is a numerical optimization solver written in Rust. The algorythm is called PANOC(Proximal Averaged Newton-type method for Optimal Control).

Following features are highlighted.

• Embeddable
• Accurate
• Fast
• User Friendly
• Community
• Documented

See Official Page for detail.

Simulation Results

Simple Pathtrack with MPC, predicting vehicle behavior by Kinematic Bicycle Model.

SVG Animation

MPC log

Simulator log

Realtime visualizer

realtime_visualizer.mp4

Environment construction

Here are procedures for ubuntu users.

• Install Python
  • pyenv is helpful to switch version of python interpreters.
  • Python version 3.8.x is recommended.
• Install Poetry
  • $ curl -sSL https://install.python-poetry.org | python3 -
  • $ echo export PATH="/home/mizuho/.local/bin:$PATH" >> ~/.bashrc
• Install cargo
  • $ sudo apt-get -y install cargo
• Clone this repository
  • git clone https://github.com/MizuhoAOKI/pympc_panoc_ipopt.git
• Install python libraries
  • $ cd pympc_panoc_ipopt
  • $ poetry install

How to run

$ cd pympc_panoc_ipopt
$ poetry shell
$ cd pathtrack
$ python main.py

References

• OpEn official HP
• About installation of OpEn
• Python Examples of OpEn