Learning Control Lyapunov functions to Ensure Stable Motions

This code implements the paper "Learning control Lyapunov function to ensure stability of dynamical system-based robot reaching motions" in python. For details, please refer to:

   S.M. Khansari-Zadeh and A. Billard (2014), "Learning Control Lyapunov Function
   to Ensure Stability of Dynamical System-based Robot Reaching Motions."
   Robotics and Autonomous Systems, vol. 62, num 6, p. 752-765.

The original MATLAB implementation of this paper can be found at:

https://bitbucket.org/khansari/clfdm/.

This code started as a branch of:

https://github.com/robotsorcerer/LyapunovLearner.

Differently from this other repository, here we only focus on the method implementation. Moreover, the LASA handwriting dataset has been included for testing.

Examples

Energy levels Lyapunov function

• Blue: demonstrations
• Red: trajectories from learned dynamical system

Vector field learned dynamical system

• White: demonstrations
• Red: trajectories from learned dynamical system

Setup

Install as follows:

  pip install -r requirements.txt

Usage

In the folder src run:

  python main.py

To change parameters or select a different shape modify the file config.py.

Credits

• Olalekan Ogunmolu
• Rodrigo Pérez Dattari
• Rachel Skye Thompson