Control barrier functions for safe robotics (2017)
Researchers led by Aaron Ames at Georgia Tech and Caltech formalized control barrier functions (CBFs), a way to prove a control system will never leave a safe region. The functions translate safety requirements into constraints that can be enforced in real time on complex nonlinear systems.
CBFs became a standard tool in safety-critical robotics and autonomous driving research within a few years.
The framework showed that hard safety guarantees and strong performance can coexist when constraints are handled explicitly.
HardFlow applies the same philosophy, enforcing hard constraints without crippling performance, but shifts it from control inputs to generative AI sampling.
