Control Barrier Function (CBF)

The primary function of a CBF is to ensure forward invariance of a safe set. If a controller satisfies the CBF condition derived from a function h(x), then any system state starting within the safe set {x | h(x) ≥ 0} is guaranteed to remain within that set for all future time. This provides a formal proof of safety, moving beyond heuristic or probabilistic assurances. For example, in robot navigation, a CBF can guarantee a vehicle never enters a region designated as an obstacle.

Safety constraints (e.g., "maintain a minimum distance from obstacles," "keep joint angles within limits") are encoded into a scalar-valued function h(x), where x is the system state. The safe set is defined as the superlevel set where h(x) ≥ 0. The CBF framework then transforms this geometric constraint into a differential constraint on the system's control inputs. This mathematical formulation allows complex spatial and operational limits to be integrated directly into the control synthesis process.

CBFs are typically used to synthesize safe controllers by solving an online Quadratic Program (QP). The core process is:

• A nominal, performance-driven controller (e.g., tracking a desired path) generates a desired input u_des.
• The CBF condition, ∇h(x)·f(x) + ∇h(x)·g(x)u ≥ -α(h(x)), forms a linear constraint on the admissible control u.
• The QP finds the control input u* that is minimally invasive relative to u_des while satisfying the CBF safety constraint. This ensures safety without unnecessarily sacrificing performance.

A key strength of CBFs is their use as a safety filter. They do not replace performance-oriented controllers but act as a corrective layer. The QP-based synthesis finds the closest safe control to the desired one. This allows systems to pursue primary objectives (e.g., speed, efficiency) until a safety boundary is approached, at which point the CBF overrides just enough to maintain safety. This is superior to overly conservative controllers that permanently degrade performance.

For systems with relative degree greater than one (where the safety constraint h(x) does not directly depend on the control input u), the basic CBF condition cannot be applied directly. High-Order Control Barrier Functions (HOCBFs) solve this by recursively defining a sequence of functions based on the Lie derivatives of h(x) along the system dynamics. This allows safety guarantees for complex dynamics, such as ensuring a car maintains a safe following distance when control is through acceleration (second-order relative degree).

CBFs are often paired with Control Lyapunov Functions (CLFs) in a unified CLF-CBF Quadratic Program. The CLF encodes stability or convergence objectives (e.g., "reach the goal"), while the CBF encodes safety. The QP then solves for a control input that satisfies both sets of constraints, formally guaranteeing both stability and safety. This framework is central to applications like safe navigation, where an agent must stabilize to a target while avoiding obstacles.

A Control Lyapunov Function is a mathematical tool used to formally guarantee the asymptotic stability of a dynamical system's equilibrium point. While a CBF ensures a system stays within a safe set, a CLF ensures it converges to a desired state. In quadratic program-based controllers, CLFs and CBFs are often combined to achieve both stability and safety objectives.

• Primary Role: Stability guarantee.
• Typical Form: A scalar function V(x) where the derivative along system trajectories is negative definite.
• Common Use: Paired with CBFs in a single optimization to satisfy both safety and convergence constraints.

A Barrier Certificate is a scalar function of the system state whose zero-level set defines the boundary of a safe region, providing a formal proof of set invariance without requiring the explicit synthesis of a controller. It is a verification tool, whereas a CBF is a synthesis tool.

• Key Difference: Proves an existing controller is safe vs. synthesizing a new safe controller.
• Verification Focus: Used in formal methods to certify that all possible trajectories of a closed-loop system remain within the safe set.
• Application: Often employed in offline analysis of hybrid and cyber-physical systems.

Hamilton-Jacobi (HJ) Reachability is a formal method for computing the Backward Reachable Tube (BRT)—the set of all states from which a system can be driven into an unsafe set despite optimal control. It provides the ultimate safety guarantee but is computationally intensive for high-dimensional systems.

• Computational Challenge: Suffers from the curse of dimensionality.
• Output: A value function whose zero sub-level set is the BRT.
• Relation to CBF: The value function from HJ analysis can sometimes be used to construct a CBF, providing a less conservative safety filter.

A Quadratic Program Safety Filter is a real-time optimization-based controller that minimally modifies a nominal, potentially unsafe control input to ensure safety constraints derived from a CBF are satisfied. This is the standard implementation architecture for CBFs.

• Standard Form: min ||u - u_nom||² subject to CBF constraint ẋ ≥ -α(h(x)).
• Real-Time Operation: Solved at high frequency (e.g., 100Hz) to filter every control command.
• Advantage: Decouples performance (handled by the nominal controller) from safety (handled by the QP filter).

A High-Order Control Barrier Function extends the CBF framework to handle safety constraints with relative degree greater than one, meaning the control input does not appear in the first derivative of the constraint function. This is critical for many robotic systems.

• Relative Degree: The number of times the constraint function must be differentiated before the control input appears.
• Method: Defines a sequence of functions based on Lie derivatives to eventually expose the control input.
• Application: Essential for enforcing constraints on acceleration, jerk, or other higher-order states in vehicle and manipulator control.

An Adaptive Control Barrier Function is a CBF designed for systems with parametric uncertainties (e.g., unknown mass, friction). It incorporates an adaptive law to estimate unknown parameters online while simultaneously guaranteeing safety.

• Core Challenge: Maintaining safety guarantees despite model uncertainty.
• Mechanism: Uses Lyapunov-like arguments to adapt parameters while ensuring the CBF condition holds.
• Use Case: Safe control of robots interacting with unknown payloads or operating on uncertain terrain.