Road Network Design Problems (RNDP) involve two levels: the main design decision and the subsequent network performance that derives from that network design. In real-world cities, solving these two levels together in a single-level framework through mathematical methods (optimization) increases the overall numerical complexity of the problem, turning the problem very computationally expensive and the solving process becomes cumbersome (most of the times).
For larger case studies, single-level optimization is not an advisable method to solve these problems, given the numerical complexity and the time-consuming solving process. Simulation methods appear as a method to reduce the complexity of the mathematical problem by solving one part of the issue. In RNDP, simulation solves the lower-level problem by estimating the performance of the network for a given network design solution in a much faster way than the mathematical methods (at larger road networks). Yet, joining optimization together with simulation requires an interface and a framework that can be very computationally expensive; plus, the methodology for solving the higher-level problem through optimization is usually through (meta)heuristics that generate design solutions not through mathematical programming. Joining simulation and optimization is challenging in terms of programming, software interfaces and time resources; yet a feasible solution (the global optimal solution is not guaranteed) can be obtained even when the solution process does not reach the end.
The main objective of this presentation is to discuss simulation-optimization (SO) frameworks, in particular, the one recently developed for solving the problem of reversible lanes in a smart traffic control system. The limitations of each subroutine (optimization and simulation) and its effect on the whole design problem will be debated, as well as future research lines to solve this problem in a more efficient way.
