koopman-learning-and-control

Python simulation and hardware library for Koopman based learning and control.

The simulation framework of this repository is utilizing the Learning and Control Core Library.

Code examples for Koopman bilinear model learning and nonlinear model predictive control

This repository contains code to generate the examples combining learning of control-affine dynamics with Koopman bilinear models and nonlinear model predictive control.

Examples related to the paper "Koopman NMPC: Koopman-based Learning and Nonlinear Model Predictive Control of Control-affine Systems" (ICRA 2021) are contained in two Jupyter notebooks:

1. bilinearizable_sys_mpc.ipynb: Code containing example of a system that has a Koopman bilinear transform that can be computed analytically, and how to design the nonlinear model predictive control law for the exact Koopman bilinear form.
2. planar_quadrotor_learning_mpc.ipynb: Code containing planar quadrotor example with data collection, model learning, and control design.

An example related to the paper "Quadrotor Trajectory Tracking with Learned Dynamics: Joint Koopman-based Learning of System Models and Function Dictionaries" (submitted to ICRA 2022) are contained in the Jupyter notebook:

1. planar_quadrotor_learning_mpc_dnn.ipynb: Code containing planar quadrotor example with data collection, model learning, and control design.

macOS setup

Set up virtual environment

python3 -m venv .venv

Activate virtual environment

source .venv/bin/activate

Upgrade package installer for Python

pip install --upgrade pip

Install requirements

pip3 install -r requirements.txt

Create IPython kernel

python3 -m ipykernel install --user --name .venv --display-name "Virtual Environment"