Neural Control Barrier Functions from Physics Informed Neural Networks

Shreenabh Moujesh Agrawal
BS (Research) Thesis · Indian Institute of Science · 2025
Supervised by Dr. Shishir N. Y. Kolathaya and Dr. G. R. Jayanth

Download PDF

Abstract

As autonomous systems become increasingly prevalent in daily life, ensuring their safety is paramount. Recent work has shown that Control Barrier Functions (CBFs) provide verifiable guarantees of safety and have consequently emerged as effective tools for this task. Nevertheless, manually designing them for specific applications remains a significant challenge. With the advent of deep learning techniques, recent research has explored synthesizing CBFs using neural networks—commonly referred to as neural CBFs. This paper introduces a novel class of neural CBFs leveraging a physics-inspired neural network framework by incorporating Zubov’s Partial Differential Equation (PDE) within the context of safety. The proposed approach is scalable and suitable for synthesizing neural CBFs even for systems with high-dimensional state spaces, as demonstrated by an aerial navigation example. By utilizing neural reciprocal CBFs as described in this work, users can modify the “conservativeness” of the safe regions by appropriately selecting corresponding sublevel sets. To validate the effectiveness of the approach, case studies are presented on three different systems: an inverted pendulum, autonomous ground navigation, and aerial navigation in obstacle-laden environments.

Cite

@misc{agrawal2025neural,
  title  = {Neural Control Barrier Functions from Physics Informed Neural Networks},
  author = {Shreenabh Moujesh Agrawal},
  school = {Indian Institute of Science},
  year   = {2025},
  note   = {BS (Research) Thesis}
}

← All publications