|M.Sc Student||Reznikov Elena|
|Subject||Algorithms for Cross Nested Logit Stochastic User|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Tomer Toledo|
|Professor Shlomo Bekhor|
|Full Thesis text - in Hebrew|
Most stochastic user equilibrium (SUE) model applications reported in the literature are based on the multinomial logit model (MNL). The analytical simplicity of the MNL-SUE model led to development of many efficient algorithms. However, the structure of the MNL model does not take into account the similarity between different alternatives. In typical transportation networks there may be significant overlaps among different paths. This property affects the fidelity of the assignment results. In recent years new route choice models were developed and applied to the SUE traffic assignment problem, that try to overcome the problems of the model MNL.
This study discusses the potential application of the CNL model for procedure of stochastic loading as solution of the SUE problem. The significant advantage of the CNL-SUE is that the solution considers the similarity between the different paths.
The research develops path-based algorithms for the solution of the CNL-SUE problem based on adaptation of the disaggregate simplicial decomposition (DSD) method. The algorithms differ with respect to the step size determination, in which three different methods are considered. The algorithms are tested in two well-known networks. Two main tests are conducted: first, the impact of the CNL model parameters on the assignment results is analyzed, and second the difference between the CNL-SUE and MNL-SUE solutions are investigated. The results indicate that the path-based algorithm with Armijo's step size rule outperforms other step size determinations. The research shows that depending on the model parameters, particularly the nesting coefficient, the CNL-SUE path flows may be quite different from the MNL-SUE path flows.