Abstract

Continuous network design problem is often formulated as a bilevel program with equilibrium constraints, and only approximation solution can be obtained due to its nonconvexity. Based on geometric programming, this paper develops an equivalent singlelevel model to find an approximated global optimal solution. The principle of solving this problem is to apply a monomial approximation method to transform it to an equivalent nonlinear but convex problem, which is amenable to a global solution.