High-Order Control Barrier Functions for Safety-Critical Control

This direction is split into other directions: high relative degree constraints, high-order control barrier functions. Its last version stays readable and citable here.

Version 1 of 1134 papers152 conceptsfrozen 2026-10-04

In autonomous systems, "safety" is often defined as staying within a specific region of the state space—for example, keeping a drone a minimum distance from a wall. Mathematically, we define a safe set $\mathcal{C}$ where a scalar function $h(x) \geq 0$. To guarantee safety, a controller must ensure that if the robot starts safe, it never leaves this set.

The standard tool for this is the Control Barrier Function (CBF). However, a fundamental physical mismatch often makes standard CBFs fail: the "relative degree" problem.

The Problem of High Relative Degree

In simple systems, the control input $u$ (like velocity) directly affects the safety metric $h(x)$ (like position). This is a relative degree one system. If you are getting too close to a wall, you simply command a velocity away from it.

But real robots have inertia. As noted by Aaron D. Ames and colleagues in Control Barrier Function Based Quadratic Programs with Application to Automotive Safety Systems, most mechanical systems are high relative degree. You don't control position or even velocity directly; you control torque or acceleration.

If you only constrain the system when $h(x)$ reaches zero, it might already be moving so fast that no amount of braking (maximum negative acceleration) can prevent a collision. As Guang Yang explains in Continuous-time Signal Temporal Logic Planning with Control Barrier Function, planners must account for the system's inertia to prevent "unavoidable constraint violations."

Bridging the Gap with HOCBFs

To solve this, researchers use High-Order Control Barrier Functions (HOCBFs). Instead of a single constraint on the boundary, an HOCBF creates a chain of increasingly restrictive constraints on the derivatives of the safety function.

If $h(x)$ is your distance to a wall, and you control acceleration, an HOCBF doesn't just look at distance; it looks at a combination of distance and velocity. It forces the robot to start slowing down long before it hits the boundary. Wei Xiao and Calin Belta demonstrate this in Bridging the Gap between Optimal Trajectory Planning and Safety-Critical Control, showing how these functions allow complex systems like autonomous vehicles to maintain safety even when the control influence is delayed through multiple layers of integration.

Safety Under Uncertainty and Learning

A major challenge in high-order systems is that the mathematical relationship between input and safety is often unknown or noisy. Mohammad Javad Khojasteh and Vikas Dhiman address this by combining Probabilistic CBFs with Gaussian Processes in Probabilistic Safety Constraints for Learned High Relative Degree System Dynamics. This allows a robot to learn its own complex physics on-the-fly while maintaining a "safety buffer" that accounts for model uncertainty.

Similarly, Armin Lederer and Sandra Hirche explore this in the context of elastic joint robots, where the "relative degree" is even higher due to the flexible coupling between motors and links.

Real-Time Execution via Optimization

Because safety must be guaranteed every millisecond, HOCBFs are typically implemented as Quadratic Programs (QP). The QP acts as a "safety filter": it takes a desired performance command (like a path from a neural network or a human operator) and minimally modifies it to satisfy the HOCBF constraints.

Recent work has focused on making these filters faster and more robust:

  • Discrete-Time HOCBFs (DHOCBFs): Shuo Liu and Jun Zeng developed iMPC-DHOCBF, which slashes computational costs by 100x, allowing high-order safety to be checked in real-time for complex vehicles.
  • Robustness: J. W. H. Chun and Peter J. Seiler introduced Robust Time-Varying CBFs to handle actuator nonlinearities like tire slip or motor saturation, which are common in high-order mechanical systems.
  • Sampled-Data Safety: Ming Gao and colleagues proposed SdHOCBFs to ensure quadrotors remain safe between the discrete time steps of a digital controller.

By mathematically formalizing the "delay" between action and outcome, High-Order Control Barrier Functions provide a rigorous way to ensure that robots with complex, high-inertia dynamics never reach a point of no return.

Papers in this version

Showing 60 of 134 papers.

History & citationVersion 1 · 2026-10-04

Split into other directions · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_high_order_control_barrier_functions_for_5e749ca6,
  title = {High-Order Control Barrier Functions for Safety-Critical Control},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/high-order-control-barrier-functions-for-safety-critical-control/rdr_5e749ca64c79467c9069c471cd66be62},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_high_order_control_barrier_functions_for_5e749ca6_v1,
  title = {High-Order Control Barrier Functions for Safety-Critical Control},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/high-order-control-barrier-functions-for-safety-critical-control/rdr_5e749ca64c79467c9069c471cd66be62/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}