⚡ electrical engineering

Using Dynamic Safety Margins as Control Barrier Functions

This paper proposes a method to design Control Barrier Functions (CBFs) for arbitrary state and input constraints by leveraging dynamic safety margins from reference governor literature, providing a relative-degree-agnostic approach that ensures safety and feasibility.

Original authors: Victor Freire, Marco M. Nicotra

Published 2026-04-28

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Victor Freire, Marco M. Nicotra

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are teaching a teenager how to drive a car. You want them to get to their destination quickly (performance), but you absolutely cannot allow them to hit a wall or run out of gas (safety).

This paper is about a mathematical way to build a "smart safety net" for robots and automated systems. It solves a classic problem in engineering: How do you keep a machine safe without making it so cautious that it becomes useless?

Here is the breakdown of the paper using everyday analogies.

1. The Problem: The "Over-Cautious" vs. "Reckless" Dilemma

In robotics, there are two common ways to handle safety:

The Reckless Way (Candidate CBFs): This is like telling a driver, "Don't hit the wall." It sounds simple, but if the driver is already going 80 mph toward a wall, the math might suddenly realize, "Wait, there is no way to stop now!" At that moment, the computer "freezes" or crashes because it can't find a solution. This is called infeasibility.
The Over-Cautious Way (Reference Governors): This is like a parent sitting in the passenger seat, constantly tapping the brakes and slowing the car down to a crawl to ensure nothing ever goes wrong. It’s very safe, but the teenager will never actually get anywhere.

2. The Innovation: The "Dynamic Safety Margin" (The Smart Co-Pilot)

The authors propose a middle ground using something called Dynamic Safety Margins (DSMs).

Think of a DSM as a "Buffer Zone" that isn't just about distance, but about momentum and energy.

Instead of just looking at where the robot is now, the DSM looks at the robot's "intentions" (its reference path) and asks: "If we keep following this plan, do we have enough 'braking power' left to stay safe later?"

The Analogy: The Mountain Biker
Imagine a mountain biker approaching a sharp turn.

A standard safety system only looks at the edge of the cliff. If the biker is already leaning too far, it's too late.
The DSM approach looks at the biker's speed, the steepness of the hill, and the grip of the tires. It realizes, "If you go this fast, you won't be able to make the turn in 5 seconds." It then subtly adjusts the "plan" (the virtual reference) to slow the biker down just enough to make the turn perfectly, without ever making them stop pedaling.

3. The Secret Sauce: The "Augmented System"

The clever mathematical trick the authors used is called Augmentation.

Usually, engineers try to control the Robot (the car). The authors decided to control the Robot + The Plan (the car + the GPS route) as if they were one single unit.

By treating the "goal" as part of the system, they can use a mathematical tool called a Control Barrier Function (CBF) to smooth out the goal itself. If the original goal is too dangerous, the math "bends" the goal into a safer version that the robot can actually achieve.

4. Why does this matter? (The Results)

The authors tested this on two tricky systems:

An "Anthill" System: A system that wants to stay in a stable zone but can easily "fall off" into chaos.
An Overhead Crane: A heavy weight swinging on a cable. This is notoriously hard to control because if you move the crane too fast, the weight swings wildly (the "pendulum effect"), which could hit something.

The Verdict:
Their method was a "Goldilocks" solution:

It was safer than the "Reckless" method (it didn't crash or freeze).
It was faster and smoother than the "Over-Cautious" method (it didn't move like a snail).
It was smarter than the "Backup" method (it didn't need to run massive, slow simulations to predict the future; it just used elegant math).

Summary in one sentence:

Instead of just reacting to danger when it appears, this paper provides a way for robots to constantly adjust their "game plan" so they always stay within a safe zone of control, allowing them to move fast without ever risking a crash.

Technical Summary: Using Dynamic Safety Margins as Control Barrier Functions

Authors: Victor Freire and Marco M. Nicotra

1. Problem Statement

The paper addresses a fundamental limitation in the field of Control Barrier Functions (CBFs). While CBFs are widely used to derive optimization-based control laws that guarantee safety (control invariance), there is currently no systematic method to synthesize them for general nonlinear systems with arbitrary state and input constraints.

Existing practical approaches often rely on "candidate CBFs"—functions that are positive in safe sets and negative in unsafe sets. However, candidate CBFs lack formal safety guarantees; specifically, the resulting optimization program (typically a Quadratic Program) can become infeasible if the state reaches a point where the control authority is insufficient to prevent constraint violation. This is a critical failure mode for safety-critical systems.

2. Methodology

The authors bridge two distinct control paradigms: Reference Governors (RG) and Control Barrier Functions (CBF).

Augmented System Approach: Instead of applying a CBF directly to the plant state $x$ , the authors propose applying it to an augmented system consisting of the state $x$ and a virtual reference $v$ . The system dynamics are augmented to include reference dynamics: $\dot{v} = \rho(v, r)$ , where $r$ is the desired target.
Dynamic Safety Margins (DSMs) as CBFs: The core innovation is proving that Dynamic Safety Margins (DSMs)—tools used in the Explicit Reference Governor (ERG) literature to measure the distance to constraint violation—are, in fact, valid CBFs for this augmented system.
Lyapunov-based Construction: The authors provide a systematic way to construct these DSMs using reference-dependent Lyapunov functions ( $V(x, v)$ ). By calculating safety thresholds ( $\Gamma^*$ ) and stability thresholds ( $\Gamma$ ), they define a vector-valued DSM $\Delta(x, v)$ that accounts for multiple state and input constraints simultaneously.
Optimization-based Policy: The control law is formulated as a Quadratic Program (QP) that minimizes the deviation from a nominal controller $\kappa(x, r)$ and a navigation field $\rho(v, r)$ , subject to the DSM-CBF safety constraints.

3. Key Contributions

Theoretical Linkage: Formally proves that DSMs are CBFs for an augmented state-reference system.
Relative Degree Agnosticism: Demonstrates that the constructed DSM-CBFs are independent of the relative degree of the constraints, simplifying the design process for complex systems.
Guaranteed Feasibility: Proves that the proposed DSM-CBF optimization program is persistently feasible, meaning a safe control input is always guaranteed to exist within the safe set.
Vector-Valued Generalization: Extends traditional real-valued CBFs to vector-valued CBFs, allowing for the simultaneous handling of multiple, diverse constraints (e.g., position, velocity, input saturation, and payload limits).
Systematic Synthesis: Provides a recipe for designing DSMs using existing Lyapunov-based tools from the reference governor literature.

4. Results and Numerical Validation

The authors validated their method using two complex nonlinear systems: an Anthill system (representing a system with potential instability/divergence) and an Overhead Crane (representing a multi-constraint mechanical system).

Performance vs. ERG: The DSM-CBF approach significantly outperforms the standard Explicit Reference Governor (ERG) in terms of tracking performance (it is less conservative) while maintaining the same level of safety.
Performance vs. Backup CBFs: While "Backup CBFs" (which use trajectory prediction) can achieve high performance, they are computationally expensive and numerically unstable over long horizons. The DSM-CBF approach provides comparable performance with much lower computational overhead and higher numerical robustness.
Performance vs. Candidate CBFs: Candidate CBFs often fail (become infeasible) when the system approaches the boundaries of its control authority. The DSM-CBF approach successfully avoids these infeasibilities by accounting for input constraints during the synthesis phase.

5. Significance

This work is significant because it provides a systematic, mathematically rigorous bridge between the reference governor and CBF literatures. It transforms the difficult problem of "finding a valid CBF" into the well-studied problem of "finding a Lyapunov function." By doing so, it offers a practical and computationally efficient framework for designing safety filters that are guaranteed to be feasible, even in the presence of strict input saturation and complex state constraints. This has direct implications for the deployment of autonomous systems in safety-critical environments.