A new arXiv preprint tackles a subtle failure mode in safety-critical control. When a system must satisfy multiple second-order control barrier function (CBF) constraints, the available control authority may be insufficient to meet all of them simultaneously—even if the current state is technically feasible. This can happen under bounded acceleration, where the limits on input magnitude make it impossible to satisfy every constraint at once.
The authors propose a signed local joint-feasibility margin to quantify how close the system is to losing joint feasibility. The margin is designed to capture whether a valid control input exists that satisfies all constraints, and its sign indicates whether the constraints are currently jointly feasible. The abstract does not provide further details on the margin's computation or application, but the concept directly addresses the gap between individual feasibility and joint feasibility under input bounds.
This work is relevant for autonomous systems that rely on CBFs for safety guarantees, such as robots and vehicles, where actuator limits are unavoidable. By detecting potential conflicts before they occur, the margin could enable more robust safety filters or online constraint management. However, the abstract alone does not describe experimental validation or algorithmic implementation, so the practical impact remains to be seen.