On the Properties of Optimal-Decay Control Barrier Functions

Abstract

Control barrier functions provide a powerful means for synthesizing safety filters that ensure safety framed as forward set invariance. Key to CBFs' effectiveness is the simple inequality on the system dynamics: $\dot{h} \geq - \alpha(h)$. Yet determining the class $\mathcal{K}^e$ function $\alpha$ is a user defined choice that can have a dramatic effect on the resulting system behavior. This paper formalizes the process of choosing $\alpha$ using optimal-decay control barrier functions (OD-CBFs). These modify the traditional CBF inequality to: $\dot{h} \geq - \omega \alpha(h)$, where $\omega \geq 0$ is automatically determined by the safety filter. A comprehensive characterization of this framework is elaborated, including tractable conditions on OD-CBF validity, control invariance of the underlying sets in the state space, forward invariance conditions for safe sets, and discussion on optimization-based safe controllers in terms of their feasibility, Lipschitz continuity, and closed-form expressions. The framework also extends existing higher-order CBF techniques, addressing safety constraints with vanishing relative degrees. The proposed method is demonstrated on a satellite control problem in simulation.