Investigation of the Applicability of Surrogate Models in Transportation Network Design Problems
Bildiriler > Tam Metin Bildiri
Yazarlar: Mervegül Uysal, Yalçın ALVER
International Conference on Civil, Structural and Transportation Engineering, , , Uluslararası, Tam Metin Bildiri, Scopus
Özet
The rapidly increasing rate of vehicle ownership causes significant transportation problems, especially in developing countries.The increase in travel costs and safety are among these issues.To solve this problem, transportation-related institutions generally plan to change the existing geometry of the roads or open alternative roads.These solutions are quite costly and timeconsuming.In addition, existing conditions are not always sufficient for applying these solutions.At the same time, the results of these improvements are not known with certainty.Both travel costs and road and driving safety should be increased with these improvements.This study developed a bi-level optimization model considering these two objectives.According to the network design decisions taken in the lower-level problem of the developed bi-level model, the route choice of the transportation network users is determined by the deterministic user equilibrium assignment model.The assumptions made by Wardrop (1952) were considered in the assignment model.It has been observed that the user equilibrium conditions are ensured in the developed assignment model.The objective function of the upper-level optimization model is to minimize the expected number of accidents to increase road safety.The solution of optimization models developed to work on large-scale road networks takes quite a long time.Since the assignment problem solved in the lower-level optimization model contains continuous loops, the time spent on the solution of the model increases.Therefore, surrogate models were used to shorten this period in the study.Surrogate models can solve largescale problems in less time by producing their data with actual data.As a result of the study, it has been determined that using surrogate models in a medium-sized network shortens the long processing time and gives the most accurate result in a short time.In this respect, surrogate models hope to obtain accurate and applicable results by conducting field studies in an entire transportation network with a short processing time.