A Simple Python Implementation of Information-Theoretic Model Predictive Path Integral Controller

A simple python implementation of Information-Theoretic Model Predictive Path-Integral (MPPI) controller introduced by G. Williams et al. in their 2018 paper. This project is based on the work by Mizuho Aoki but contains the following modifications

• Notebook to generate and visualize custom trajectories
• JIT acceleration of cost function evaluation using Numba.
• Ideal-Differential Drive Kinematic model
• Both stage and terminal costs implemented as simple quadratic const function.

This work was part of my final project for EE 7500: Model Predictive Control class taught by Dr. Xiangyu Meng at Louisiana State University.

Supported Dynamics models

• Ideal Differential Drive: See my project report in pdfs_notes for more details
• Ideal Bicycle model --> Head over to to Mizuho's repository.

Video demonstrations

• Oval path with circular obstacles

oval_with_obs.mp4

• Figure-8 path.

figure_eight.mp4

Setup

• Clone this repo, create a mamba (or conda enviornment) and install the required dependencies

mamba create --name ee7500_proj --clone base
mamba activate ee7500_proj
mamba install ipykernel numpy pandas matplotlib seaborn numba

Optional

• ffmpeg

  • mp4 movie writer

  • installation details

    • For Ubuntu Users
      • sudo apt-get update
      • sudo apt-get -y install ffmpeg
    • For Windows Users
      • Install scoop
      • scoop install ffmpeg
    • For macOS Users
      • Install homebrew
      • brew install ffmpeg
    • Check the official website if necessary
      • https://ffmpeg.org/

Usage

• To see the differential drive model in action run either mppi_ee7500_proj_oval.ipynb or mppi_ee7500_proj_figure_eight.ipynb notebooks.
• To visualize or create custom paths, use the traj_gen.ipynb.
• To add new dynamics models see the ./scripts/diffdrive.py.
• To add/modify cost functions/distribution sampling see the ./scripts/mppi.py.
• Various plotting and computational utility functions are in ./scripts/utils.py.

References

1. G. Williams et al. "Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving"
   • URL : https://ieeexplore.ieee.org/document/8558663
   • PDF : https://arxiv.org/pdf/1707.02342.pdf

Other notable projects

• mpc_python: https://github.com/mcarfagno/mpc_python
• nav2's MPPI impelemtation: https://vimeo.com/879001391