Consolidated Control Barrier Functions

• Consolidated control barrier functions are smooth, unified formulations that algorithmically combine multiple candidate CBFs using state‐ and time-dependent weights.
• They ensure forward invariance and feasible QP synthesis under simultaneous state and input constraints, addressing challenges like discontinuities and infeasibility.
• Their versatile framework supports Boolean compositions and real-time adaptation, making them highly applicable to multi-agent navigation, robotic systems, and temporal logic tasks.

A consolidated control barrier function (C-CBF) is a formal, algorithmic synthesis of multiple candidate control barrier functions (CBFs) into a single, smooth barrier that provably enforces safety for control-affine systems under multiple simultaneous state and input constraints. The consolidated CBF paradigm addresses feasibility, regularity, and unification issues that can arise when enforcing several individual CBF constraints in parallel, such as infeasibility of quadratic programs (QPs), discontinuities or loss of invariance at set intersections, and tuning complexity. This approach tightly couples safety and control authority by adaptively combining the constituent CBFs using state- and time-dependent weighting, and ensures the resulting barrier admits forward invariance of an (under-approximating) safe set, even under input constraints. The development and online adaptation of C-CBFs have enabled robust, real-time, and high-dimensional safety filtering for complex systems and formal task specifications.

1. Fundamentals of Control Barrier Functions and the Need for Consolidation

A control barrier function (CBF) for a control-affine system

$$\dot x = f(x) + g(x)u, \quad x \in \mathbb{R}^n,\,u\in \mathcal{U} \subset\mathbb{R}^m$$

is a scalar function $h(x)$ whose nonnegative superlevel set encodes the "safe" set $\mathcal{C} = \{x \mid h(x) \ge 0\}$. The invariance of $\mathcal{C}$ is established by enforcing, pointwise in $x$, the condition

$L_f h(x) + L_g h(x)\,u + \alpha(h(x)) \ge 0$

for all $x \in \mathcal{D} \supset \mathcal{C}$, where $L_f h = \nabla h \cdot f$, $L_g h = \nabla h \cdot g$, and $\alpha$ is a class-$h(x)$0 function. This constraint can be incorporated into a QP-based controller synthesis, often in combination with a (soft) control Lyapunov function (CLF) constraint for performance objectives, yielding controllers that are provably safe and well-posed under regularity and relative degree conditions (Ames et al., 2016, Xu et al., 2016).

However, in realistic systems, multiple safety constraints are present:

• Physical or regulatory limits (e.g., collision avoidance, actuator bounds);
• Logical or task-based constructs (e.g., signal temporal logic (STL) objectives, task sequencing);
• Composite or networked system constraints (e.g., multi-agent or distributed networks).

Simply enforcing $h(x)$1 individual CBF constraints can lead to infeasibility in the presence of input constraints, chattering and nonsmoothness when using min/max or switching constructions, loss of safety at safe set intersections, and difficulties with tuning class-$h(x)$2 functions for compatibility (Breeden et al., 2022, Black et al., 2023, Black et al., 2022).

2. Synthesis of Consolidated CBFs

The core idea of consolidation is to algorithmically combine $h(x)$3 candidate CBFs, each enforcing forward invariance of a set $h(x)$4, into one smooth function $h(x)$5, parameterized by state- or time-dependent weights $h(x)$6, whose nonnegative superlevel set $h(x)$7 under-approximates $h(x)$8 and can be certified as a valid CBF.

A canonical form is

$h(x)$9

where $\mathcal{C} = \{x \mid h(x) \ge 0\}$0 is a monotone function (e.g., $\mathcal{C} = \{x \mid h(x) \ge 0\}$1) and each $\mathcal{C} = \{x \mid h(x) \ge 0\}$2 is a positive adaptation weight. As $\mathcal{C} = \{x \mid h(x) \ge 0\}$3, the zero-superlevel set $\mathcal{C} = \{x \mid h(x) \ge 0\}$4 closes up to $\mathcal{C} = \{x \mid h(x) \ge 0\}$5; for finite $\mathcal{C} = \{x \mid h(x) \ge 0\}$6, $\mathcal{C} = \{x \mid h(x) \ge 0\}$7 (Black et al., 2022, Black et al., 2023).

The C-CBF admits a forward-invariance guarantee if the following holds for all $\mathcal{C} = \{x \mid h(x) \ge 0\}$8: $\mathcal{C} = \{x \mid h(x) \ge 0\}$9 where $\mathcal{C}$0 and $\mathcal{C}$1 decompose into linear combinations of the constituent $\mathcal{C}$2 and $\mathcal{C}$3, weighted by terms $\mathcal{C}$4 (the partial derivatives).

The online adaptation of weights $\mathcal{C}$5 is critical: it ensures that the consolidated barrier remains well-defined, nonsingular (enabling synthesis of safe controls for any state on the boundary), and keeps the C-CBF constraint feasible even as the system evolves.

• (Black et al., 2022) proposes an explicit adaptation law for $\mathcal{C}$6, enforcing that the controllable dynamics of $\mathcal{C}$7 never vanish (i.e., the control effectiveness stays nonzero on the boundary).
• (Black et al., 2023) develops a predictor-corrector (interior-point) flow for $\mathcal{C}$8 as the optimizer of a constrained problem that balances regularity against feasibility.

3. Feasibility and Theoretical Guarantees

A principal concern for QP-based synthesis under multiple constraints is feasibility, i.e., the existence of a control $\mathcal{C}$9 satisfying all (possibly consolidated) CBF inequalities and input constraints. The consolidated approach guarantees:

• Under mild regularity conditions on the system (local feedback linearizability, smoothness, relative degree one), and provided the state remains within $x$0, the consolidated CBF condition is always feasible for a range of input constraints (Black et al., 2023).
• Any measurable selection $x$1 from the feasible set $x$2 renders $x$3 forward-invariant.
• For time-varying CBFs (e.g., for STL or mission specifications), the adaptation law ensures that the forward-invariance property holds throughout the task execution, provided the adaptation remains within the feasible region.

If each constraint is of mixed or higher relative degree, the consolidated CBF can still be constructed; no assumption is made that all constituent CBFs have the same relative degree (Black et al., 2023).

The QP-based controller synthesizes the safe action: $x$4 which is always feasible by construction on the C-CBF domain.

4. Consolidated CBFs for Boolean and Complex Compositions

Beyond simple set intersections, consolidated CBF constructions permit the encoding of arbitrary Boolean compositions—conjunctions (AND), disjunctions (OR), and even NOT—of primitive constraints. The methodology is as follows (Molnar et al., 2023):

• Conjunctions are encoded via smooth approximations of min(·) (e.g., log-sum-exp or smooth harmonic mean).
• Disjunctions via smooth versions of max(·) (e.g., log-sum-exp of negated arguments).
• Algorithms recursively build a single, smooth composite barrier function capturing the Boolean formula over all primitive constraints.

Such a composite CBF $x$5 enjoys:

• $x$6 regularity: differentiable everywhere in $x$7.
• Soundness: as the smoothing parameter increases, the zero-superlevel set of $x$8 approaches the exact safe set defined by the Boolean formula.
• Forward invariance: provided each constituent CBF is valid and the compatibility condition is met, enforcing the consolidated constraint ensures safety for the composite set (Molnar et al., 2023).

5. Adaptation and Online Regularization

Adaptive and time-varying consolidation is essential when constituent CBFs become singular, conflict, or lead to vanishing effectiveness due to input constraints or the geometry of the combined set. Both (Black et al., 2022) and (Black et al., 2023) present adaptation-based methods for online regularization:

• State- and time-dependent adaptation laws update a vector of weights $x$9, always keeping the consolidated constraint regular.
• The adaptation may be expressed as the solution of a real-time convex QP or an interior-point ODE, combined with log-barrier penalties to guarantee strict feasibility and maintain minimum gain on controllable directions.
• These adaptation laws guarantee that for all time, the consolidated barrier retains full rank (that is, the relative degree-1 derivative along the control vector field never vanishes on the boundary).

This ensures not only the well-posedness and feasibility of the CBF-QP at each instant but also global safety over mission horizons.

6. Applications and Performance

Consolidated CBFs have been demonstrated on a range of safety-critical systems:

• Multi-agent navigation in crowded, cluttered environments: the C-CBF framework scales efficiently to dozens of simultaneously imposed constraints (each encoding, e.g., inter-agent collision avoidance, workspace limits, and obstacle exclusion) (Black et al., 2022, Black et al., 2023).
• Temporal logic task planning with nested and conflicting objectives: for STL specifications, consolidated CBFs encode both reachability and sequential requirements, ensuring that all subtasks are met within their prescribed temporal windows under actuation saturation (Buyukkocak et al., 2022).
• Robotic systems with geometric and functional constraints: consolidated Boolean CBFs enable complex, non-convex, and non-monotonic safe set definitions with guarantees of smoothness and forward invariance (Molnar et al., 2023).

Empirical results confirm that the consolidated approach retains real-time capability (e.g., solve times at 1 ms per QP on typical platforms), eliminates chattering, and avoids infeasibility or loss of safety at constraint intersections previously problematic for min/max or switching approaches.

7. Comparison to Alternative Multi-CBF Approaches

Traditional schemes for handling multiple CBFs include joint QPs with one constraint per CBF, which generally require independence or non-interference conditions to guarantee nonempty safe input sets (Breeden et al., 2022, Reis et al., 20 Mar 2025). Other methods based on pointwise min/max, switching logic, or composition through small-gain theorems (Jagtap et al., 2020) can suffer from regularity issues, QP infeasibility, or the creation of spurious or stable boundary equilibria (cf. the boundary equilibrium analysis in (Reis et al., 20 Mar 2025)). Consolidated CBFs, by contrast:

• Collapse the enforcement to a single constraint, enabling easier feasibility analysis and streamlined adaptation.
• Allow for weighted regularization to mitigate the loss of control authority and guarantee persistent forward-invariance.
• Generalize efficiently to high-order and high-dimensional settings, compositional logic, and complex or time-varying safety tasks.

Recent advances have integrated consolidated CBFs with reference governor and motion corridor approaches, enabling persistent safe motion planning under dynamic environments (Arslan et al., 6 Mar 2026).

---

Principal References:

• (Ames et al., 2016): control barrier function-based QP synthesis for safety-critical systems
• (Black et al., 2022): adaptation law for validating consolidated CBFs and preserving controllability
• (Black et al., 2023): synthesis and online verification of time-varying consolidated CBFs under input constraints
• (Molnar et al., 2023): smooth algorithmic construction of consolidated CBFs to encode Boolean logic
• (Breeden et al., 2022, Reis et al., 20 Mar 2025, Buyukkocak et al., 2022): feasibility, compositionality, and temporal logic integration

This unified theory extends the practicality and scope of safety-critical control in nonlinear, input-constrained, and high-dimensional dynamical systems.