Viscosity CBFs: Bridging the Control Barrier Function and Hamilton-Jacobi Reachability Frameworks in Safe Control Theory

Basic Information

• Paper ID: 2510.09929
• Title: Viscosity CBFs: Bridging the Control Barrier Function and Hamilton-Jacobi Reachability Frameworks in Safe Control Theory
• Authors: Dylan Hirsch, Jaime Fernández Fisac, Sylvia Herbert
• Classification: eess.SY (Systems and Control), cs.SY (Systems and Control)
• Publication Date: October 11, 2025 (arXiv preprint)
• Paper Link: https://arxiv.org/abs/2510.09929

Abstract

Control Barrier Functions (CBFs) and Hamilton-Jacobi Reachability (HJR) are two core frameworks in safe control theory. Traditionally, these two frameworks have been viewed as distinct approaches, with the former focusing on optimal safe controller design and the latter providing sufficient conditions for safety. Prior work introduced the concept of Control Barrier Value Functions (CB-VFs), defined similarly to other value functions studied in HJR but possessing certain CBF properties. This paper approaches the problem from another direction, generalizing CBFs to non-differentiable "viscosity" CBFs. The authors demonstrate deep connections between viscosity CBFs and CB-VFs, bridging the CBF and HJR frameworks. Through this bridge, the authors precisely characterize viscosity CBFs as functions providing CBF-class safety guarantees (control invariance and boundary smooth approximation). Furthermore, favorable theoretical properties of viscosity CBFs are demonstrated, including ideal closure under maximization and limit operations.

Research Background and Motivation

Problem Background

1. Separation of Two Major Safe Control Frameworks: CBFs and HJR are two main frameworks in safe control theory, but traditionally viewed as independent approaches
2. Limitations of CBFs: Traditional CBFs require continuous differentiability, which is overly restrictive in certain applications and limits their applicability
3. Missing Theoretical Connections: Lack of a unified theoretical framework to understand the intrinsic connections between these two methods

Research Motivation

1. Theoretical Unification: Establish deep theoretical connections between CBFs and HJR, providing a more complete theoretical foundation for safe control
2. Method Extension: Generalize CBFs to non-differentiable functions, expanding their applicable scope
3. Property Characterization: Precisely characterize which functions can provide CBF-type safety guarantees

Core Contributions

1. Introduction of Viscosity CBFs Concept: Generalize traditional CBFs to non-differentiable functions, requiring only satisfaction of the CBF inequality in the viscosity sense
2. Establishment of Theoretical Bridge: Prove complete equivalence between viscosity CBFs and time-invariant CB-VFs, bridging CBF and HJR frameworks
3. Complete Characterization of Safety Guarantees: Prove that viscosity CBFs are precisely the continuous functions providing "Barrier Guarantees"
4. Extension of CB-VF Theory: Extend CB-VF from exponential discounting to nonlinear discounting determined by class K functions
5. Theoretical Property Analysis: Prove favorable properties of viscosity CBFs, including closure under maximization and limit operations

Detailed Methodology

Basic Definitions and Setup

Consider the dynamical system: $\dot{x} = f(x,u)$

where $f : \mathbb{R}^n \times U \to \mathbb{R}^n$ is a Lipschitz function, and $U \subset \mathbb{R}^m$ is a compact set of admissible control inputs.

Traditional CBF Definition

A function $h : \mathbb{R}^n \to \mathbb{R}$ is a Control Barrier Function with respect to a class K function $\alpha$ if it is continuously differentiable and satisfies: $\max_{u \in U} \nabla h(x) \cdot f(x,u) \geq -\alpha(h(x))$ for all $x \in \mathbb{R}^n$ satisfying $h(x) > 0$.

Core Innovation: Viscosity CBFs

Viscosity CBF Definition

A continuous function $h : \mathbb{R}^n \to \mathbb{R}$ is a viscosity Control Barrier Function with respect to a class K function $\alpha$ if for each $x \in \mathbb{R}^n$ satisfying $h(x) > 0$, the inequality $\max_{u \in U} \nabla h(x) \cdot f(x,u) \geq -\alpha(h(x))$ holds in the viscosity sense.

Viscosity Sense: For any continuously differentiable function $\phi : \mathbb{R}^n \to \mathbb{R}$, if $h - \phi$ has a local maximum at $x$, then $\max_{u \in U} \nabla \phi(x) \cdot f(x,u) \geq -\alpha(h(x))$

Barrier Guarantee

A continuous function $B : \mathbb{R}^n \to \mathbb{R}$ provides a Barrier Guarantee with respect to a locally Lipschitz class K function $\alpha$ if for each $\theta \in [0,1)$ and each $x \in \mathbb{R}^n$ satisfying $B(x) > 0$, there exists $u \in U$ such that for all $t \geq 0$: $B(x_x^u(t)) \geq \beta_\alpha(\theta B(x), t)$

where $\beta_\alpha$ is a class KL function associated with $\alpha$.

Extension of Control Barrier Value Functions (CB-VF)

Given a continuous function $g : \mathbb{R}^n \to \mathbb{R}_{\geq 0}$ and a locally Lipschitz class K function $\alpha$, the CB-VF $v : \mathbb{R}^n \times \mathbb{R}_{\geq 0} \to \mathbb{R}_{\geq 0}$ is implicitly defined by: $\beta_\alpha(v(x,T), T) = \sup_{u \in U} \min_{t \in [0,T]} \beta_\alpha(g(x_x^u(t)), T-t)$

Main Theoretical Results

Theorem 1: HJ-PDE Characterization of CB-VF

The CB-VF $v$ is the unique viscosity solution to the HJ-PDE: $\max\left\{\frac{\partial}{\partial T}v - H_\alpha(x,v,\nabla v), v - g(x)\right\} = 0$

where the Hamiltonian is $H_\alpha(x,r,\lambda) = \max_{u \in U} \lambda \cdot f(x,u) + \alpha(r)$.

Theorem 2: Core Equivalence Result

For a locally Lipschitz class K function $\alpha$ and a continuous function $h : \mathbb{R}^n \to \mathbb{R}_{\geq 0}$, the following propositions are equivalent:

1. $h$ is a viscosity CBF with respect to $\alpha$
2. $h$ is a viscosity solution to the HJ-PDE $-\min\{H_\alpha(x,h,\nabla h), 0\} = 0$
3. The corresponding CB-VF is time-invariant, i.e., $v(\cdot, T) \equiv h(\cdot)$ for all $T \geq 0$
4. $h$ provides a Barrier Guarantee with respect to $\alpha$

Composition Properties of Viscosity CBFs

Theorem 3: Closure Under Maximization

If $h_1, h_2$ are both viscosity CBFs with respect to a class K function $\alpha$, then $\max\{h_1, h_2\}$ is also a viscosity CBF with respect to $\alpha$.

Theorem 4: Closure Under Limits

If $h_1, h_2, \ldots$ is a sequence of viscosity CBFs with respect to a class K function $\alpha$, and $h_i \to h$ locally uniformly, then $h$ is also a viscosity CBF with respect to $\alpha$.

Technical Innovations

1. Application of Viscosity Solution Theory

• Introduce the concept of viscosity solutions from PDE theory into CBF theory
• Eliminate the strict requirement for continuous differentiability in traditional CBFs
• Preserve the essential characteristics of safety guarantees

2. Framework Bridging

• Establish bidirectional connections between CBF and HJR frameworks
• Achieve theoretical unification through CB-VF as an intermediary
• Provide alternative perspectives for understanding safe control from two different angles

3. Nonlinear Discounting Extension

• Generalize from exponential discounting to arbitrary discounting determined by class K functions
• Enhance the expressiveness and applicability of CB-VF

Application Examples

Example 1: Non-Differentiable CBF

Consider the scalar system $\dot{x} = x + x + x^3/(1+|x|) \cdot u$ with control set $U = [-1,1]$.

Define $h(x) = 1 - |x|$ (signed distance function to the unsafe set $(-\infty,-1] \cup [1,+\infty)$).

Although $h$ is non-differentiable at $x = 0$, it can be verified that it is a viscosity CBF with respect to $\alpha(r) = r$.

Theoretical Significance and Practical Value

Theoretical Significance

1. Completeness: Completely characterize the class of functions that can provide CBF-type safety guarantees
2. Unification: Unify two important safe control frameworks under the same theoretical foundation
3. Extensibility: Provide broader development space for CBF theory

Practical Value

1. Expanded Applicability: Allow the use of non-differentiable functions as CBFs, applicable to more practical problems
2. Composition Methods: Provide systematic methods for constructing new CBFs from known ones
3. Computational Advantages: Enable the use of HJR computational methods for CBF construction

Limitations and Future Directions

Limitations

1. Computational Complexity: Verification and construction of viscosity CBFs may be computationally more complex
2. Implementation Details: The paper focuses primarily on theoretical results; practical computational methods require further research
3. Application Verification: Lack of application verification on large-scale practical systems

Future Directions

1. Computational Methods: Develop efficient algorithms for viscosity CBF computation and verification
2. Application Extension: Specific applications in robotics, autonomous driving, and other fields
3. Multi-Agent Systems: Extension to safe control in multi-agent systems
4. Stochastic Systems: Consider the case of stochastic dynamical systems

In-Depth Evaluation

Strengths

1. Theoretical Depth: Establish deep theoretical connections between two important frameworks
2. Mathematical Rigor: Complete proofs with theoretically significant results
3. Conceptual Innovation: The concept of viscosity CBF is original and practically valuable
4. Systematicity: Provide a complete theoretical system with comprehensive property analysis

Weaknesses

1. Lack of Experimental Verification: The paper is primarily theoretical work lacking numerical experiments and practical application validation
2. Computational Methods: No specific computational algorithms or implementation methods provided
3. Complexity Analysis: Missing analysis of computational complexity of the methods

Impact Assessment

1. Academic Value: Provides important theoretical contributions to safe control theory
2. Practical Potential: Expands the applicable scope of CBFs with significant practical value
3. Future Research: Establishes theoretical foundation for subsequent research in related fields

Applicable Scenarios

1. Complex Geometric Constraints: Control problems with non-differentiable safety constraints
2. Multi-Modal Systems: Hybrid systems with switching dynamics
3. Robust Control: Safety-critical systems requiring uncertainty handling
4. Path Planning: Robot navigation and obstacle avoidance problems

References

The paper cites important literature in safe control theory, including:

• Foundational work on CBF theory (Ames et al.)
• Hamilton-Jacobi reachability theory (Mitchell, Bayen, Tomlin)
• Viscosity solution theory (Bardi, Capuzzo-Dolcetta)
• Original work on Control Barrier Value Functions (Choi et al.)

Summary: This is a high-quality theoretical paper making important contributions to safe control theory. By introducing the concept of viscosity CBFs and establishing connections with the HJR framework, it provides a more complete and unified theoretical foundation for safe control. Although lacking experimental verification, its theoretical value and potential application prospects make it an important work in the field.