טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentRozen Neta
SubjectDynamic Assignment Based on Mixed Strategies in Congestion
Networks
DepartmentDepartment of Civil and Environmental Engineering
Supervisors Professor David Mahalel
Professor Shlomo Bekhor
Full Thesis textFull thesis text - English Version


Abstract

Traffic assignment models have been well studied over the years, and new methods of data sharing, specifically real-time traffic data, have evolved. As a result, there has been a great interest in developing stochastic and mixed strategy models. Still, the equilibrium conditions in most traffic assignment models are based on pure strategies. In the mixed strategy approach, each user chooses a probability distribution over his or her pure strategies (where a pure strategy defines a path). The mixed strategy approach relaxes the assumption of homogeneity among network users.

Accordingly, mixed strategy models such as the Mixed-Strategy User-Equilibrium model (MSUE) for general networks were developed, based on realistic assumptions about the users: acting selfishly, aware of the network topology and links performance functions, but do not have a perfect knowledge of other users' path-choices.

However, the mixed strategy path-choices models proposed so far assumed static conditions. This research introduces a new realistic model in which general network users employ mixed strategies under dynamic conditions. The existence and uniqueness conditions for dynamic mixed strategy user equilibrium (DMSUE) are investigated.

Furthermore, the theoretical distribution of network's flow in the DMSUE is examined, depends on the user' path-choices. For single O-D networks, three path-choice patterns were observed depend on the demand level and the network topology. Two demand levels defining the bounds between the flow patterns are analytically formulated for disjoint paths' networks and algorithmic approach is presented for networks with joint paths. With this powerful tool, the demand bounds can be computed for every single O-D network, in order to predict the users' path-choices over time, given the demand level.