The integral of an increasing delay function is convex
ProvedKellyStochasticNetworks.wardrop_objective_convexLet be continuous and increasing. Then
is a convex function on .
This is the analytic fact behind the proof of Theorem 4.3. The objective of the optimization problem whose solutions are the Wardrop equilibria is ; since each link's delay function is increasing, each summand is convex in the link flow , and since is linear in the route flows the whole objective is convex in . Convexity on a compact feasible region is what delivers an optimum and lets the Lagrangian characterization be read off.
Note what is not assumed: need not be convex, and the book explicitly declines to assume it. Monotonicity alone gives convexity of the integral, because the integral's derivative is itself.
Formalization Note "Increasing" is taken in the weak sense, so a delay function that is flat on an interval is allowed; this is the weaker hypothesis and the one the argument needs. Convexity is asserted on the whole real line rather than on the non-negative half, which is where flows live; the stronger statement is true and avoids carrying a domain restriction through the composition with .
import Mathlib import Definitions.Def_KellyStochasticNetworks_Wardrop
namespace KellyStochasticNetworks
theorem wardrop_objective_convex (D : ℝ → ℝ) (hD : Continuous D) (hmono : Monotone D) :
ConvexOn ℝ Set.univ (fun y : ℝ => ∫ u in (0:ℝ)..y, D u) := by sorry
end KellyStochasticNetworksConfirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.