Application of High Order Traffic Flow Models for Incident Detection and Modeling Multiclass Flow

This work is focused on the application of high order traffic flow theory to the problem of traffic incident detection, and for modeling multiclass traffic flow composed of different vehicle types. For incident detection applications, a class of generic second order traffic flow models (GOSM) is applied to detect traffic incidents in real time by posing the problem as a hybrid state estimation problem. To incorporate the incident dynamics in the model, a regime variable is introduced to describe where and how many lanes are blocked during an incident, resulting in a multiple model framework. This work develops a multiple model extension to the GOSM on a road network. Then, a discrete version of the GOSM known as the second order cell transmission model (2CTM) is presented under the framework of the cell transmission model. Next, this multiple model predictor is integrated with a particle filter to obtain an estimate of the traffic state and the incident location if it exists. The proposed algorithm is tested on a road segment in numerical simulation using the CORSIM traffic microsimulation software as the true state. In the second application, a new family of high order traffic flow models is considered as an extension to the scalar Lighthill Whitham Richards (LWR) model. Under this framework, a heterogeneous traffic model with two vehicle classes is developed to capture an important phenomenon in highly heterogeneous traffic flows called creeping. Creeping occurs when small vehicles such as motorcycles continue to advance in congestion even though larger vehicles have completely stopped, for example via lane sharing. The new model is a phase transition model which applies a system of conservation laws in the noncreeping phase, and a scalar model in the creeping phase. The solution to the Riemann problem is obtained by investigating the elementary waves, in particular for the cases when one vehicle class is absent, as well as in the presence of a phase transition. Based on the proposed Riemann solver, the solution to the Cauchy problem is constructed using wavefront tracking. Numerical tests are carried out using a Godunov scheme to illustrate the creeping phenomenon.