An Iterative Method to Learn a Linear Control Barrier Function

TL;DR

This paper introduces an iterative, optimization-based method to learn linear control barrier functions for control affine systems, ensuring safety and respecting input constraints, demonstrated on a jet engine model.

Contribution

It presents a novel framework for constructing control barrier functions from user-defined sets using linear parameterization and sampling, applicable to general control affine systems.

Findings

Successfully learned a CBF for a nonlinear jet engine model

Ensured the system's safety by preventing unsafe set entry

Explicitly incorporated control input constraints in the CBF construction

Abstract

Control barrier function (CBF) has recently started to serve as a basis to develop approaches for enforcing safety requirements in control systems. However, constructing such function for a general system is a non-trivial task. This paper proposes an iterative, optimization-based framework to obtain a CBF from a given user-specified set for a general control affine system. Without losing generality, we parameterize the CBF as a set of linear functions of states. By taking samples from the given user-specified set, we reformulate the problem of learning a CBF into an optimization problem that solves for linear function coefficients. The resulting linear functions construct the CBF and yield a safe set which has forward invariance property. In addition, the proposed framework explicitly addresses control input constraints during the construction of CBFs. Effectiveness of the proposed method is demonstrated by learning a CBF for an nonlinear Moore Greitzer jet engine, where the system trajectory is prevented from entering unsafe set.