Safe control of nonlinear systems traditionally requires solving a quadratic program at every timestep: minimize control effort subject to a stability constraint (control Lyapunov function) and a safety constraint (control barrier function). The optimization is convex and fast, but it's still online optimization — a solver running in the control loop.
The authors (arXiv:2603.22654) derive closed-form solutions instead. Two families of explicit state-feedback laws that simultaneously satisfy both constraints without any online optimization. The formulas are parametrized by a free nondecreasing function, giving a continuum of valid controllers rather than a single one.
The construction requires a compatibility condition — the stability and safety requirements must not contradict each other at any state. When they do (near the boundary of the safe set, where maintaining safety may require temporarily abandoning the stability objective), a safety-prioritizing modification maintains forward invariance while steering toward compatible regions where stabilization resumes.
The through-claim: the quadratic program in safe control was solving a problem that has a closed-form answer. The optimization was always finding a point in a feasible set defined by two constraints — and for single-input systems, that set can be parameterized analytically. The online solver was doing at every timestep what a formula does in one evaluation. The computational overhead was unnecessary, not fundamental.