Biological systems — predator-prey dynamics, population models, chemical reactions — live in the positive orthant. Populations can't go negative. Concentrations can't go negative. Controls (harvesting rates, drug dosages) can't go negative. Everything is positive.
Note (2026-11-10): I checked the abstract (first and current versions) of Krstic's paper. It builds inverse optimal stabilizers for control-affine systems in the positive orthant with positive controls using a strict control Lyapunov function, through two frameworks (A: a CLF, a stabilizing feedback and an expander; B: a CLF and a contractor function) on orthants of arbitrary dimension, notes that a stronger CLF condition gives a stronger construction, and presents two explicit Sontag-like universal formulae, one with inverse optimality. The essay says the method works for any positive-orthant system; the abstract requires a strict CLF, and its conclusion that the constraint must be in the foundation is the essay's own.
Standard inverse optimal control methods fail here. They're designed for systems that range freely over all of state space. Restrict to the positive orthant, and the cost functions become asymmetric in a way the classical Sontag formula can't handle. The mathematics of optimal control was developed for systems that don't have to stay positive.
The fix is a pair of new functions: contractors (which pull the system toward the origin) and expanders (which push it away from the boundary). Combined with a control Lyapunov function, these enable inverse optimal stabilization for any positive-orthant system in any dimension.
Two explicit universal formulae emerge — analogues of Sontag's formula that respect positivity. One of them achieves inverse optimality, meaning the stabilizing control is not just effective but provably optimal for some cost function.
The structural lesson: the constraint of positivity isn't just a restriction on where the system can go. It's a fundamentally different mathematical structure. Systems that must stay positive don't just need positive control — they need a control theory that was built for positivity from the start. Patching the general theory doesn't work. The constraint has to be in the foundation.