FlowBarrier: New P-CBF method achieves zero safety violations in robot navigation

Control barrier functions (CBFs) have been a cornerstone for real-time safety guarantees in autonomous systems, but they suffer from myopic behavior—only considering the current state. A new paper by Amirsaeid Safari and Jesse B. Hoagg from the University of Kentucky introduces Predicted-Flow Control Barrier Functions (P-CBFs), which extend CBFs to a functional of a predicted flow under a parametrized control plan over a finite prediction horizon. This allows the safety certificate to cover the entire prediction horizon, not just the instantaneous state. To address the challenge of ensuring validity under control constraints, the authors introduce a terminal candidate P-CBF that requires the predicted flow to end in a backup safe set at the terminal time, plus a planning-time shift that modulates the prediction horizon for added feasibility. The real-time control and evolution of the control-plan parameter are jointly determined by a single convex optimization—a quadratic program (QP) when control constraints form a convex polytope. They call this QP implementation FlowBarrier.

FlowBarrier was validated on a nonholonomic ground robot navigating a dense obstacle environment. In 100 randomized trials, it was compared against nonlinear model predictive control (NMPC) and two CBF-based safety filter methods. FlowBarrier achieved the highest goal-reaching rate, zero safety violations (all other methods had some violations), and the lowest computation time. This is a significant step for real-time safe optimal control, unifying finite-horizon integral-cost optimization with rigorous safety certification. The key innovation is that P-CBFs eliminate the need to handcraft a valid CBF, which is notoriously difficult, and instead provide a systematic, provably feasible optimization framework. For robotics and autonomous systems practitioners, this means safer, faster, and more reliable navigation in cluttered environments.