This repo contains the implementation of the method Lyapunov Neural Ordinary Differential Equation State-Feedback Policies (L-NODEC) found in arXiv. Continuous-time optimal control problems play a role in many decision making tasks and this paper presents a novel Lyapunov-based formulation for neural ODE-based control policies. An exponentially-stabilitizing control Lyapunov function is incorporated into the state feedback policy, which leads to stability guarantees and adversarial robustness to pertubations in the initial state.
Install required dependencies with pip install -r requirements.txt.
This repo uses .ipynb files for visualization and it is assumed the user has a valid environment for them.
To obtain the results for the double integrator case study in the paper, open example_double_integrator.ipynb and click Run all.
If you found this paper helpful, please consider citing our work:
@article{ip2024lyapunov,
title={Lyapunov Neural ODE State-Feedback Control Policies},
author={Ip, Joshua Hang Sai and Makrygiorgos, Georgios and Mesbah, Ali},
journal={arXiv preprint arXiv:2409.00393},
year={2024}
}