Integrated Modeling Framework for Dynamic Information Flow and Traffic Flow Under Vehicle-to-Vehicle Communications: Theoretical Analysis and Application

Yong Hoon Kim, Purdue University

Abstract

Advances in information and communication technologies enable new paradigms for connectivity involving vehicles, infrastructure, and the broader road transportation system environment. Vehicle-to-vehicle (V2V) communications under the aegis of the connected vehicle are being leveraged for novel applications related to traffic safety, management, and control, which lead to a V2V-based traffic system. Within the framework of a V2V-based traffic system, this study proposes an integrated modeling framework to model the dynamics of a V2V-based traffic system that entails spatiotemporal interdependencies among the traffic flow dynamics, V2V communication constraints, the dynamics of information flow propagation, and V2V-based application. The proposed framework systematically exploits their spatiotemporal interdependencies by theoretical and computational approaches. First, a graph-based multi-layer framework is proposed to model the V2V-based advanced traveler information system (ATIS) as a complex system which is comprised of coupled network layers. This framework addresses the dynamics of each physical vehicular traffic flow, intervehicle communication, and information flow propagation components within a layer, while capturing their interactions among layers. This enables the capabilities to transparently understand the spatiotemporal evolution of information flow propagation through a graph structure. A novel contribution is the systematic modeling of an evolving information flow network that is characterized as the manifestation of spatiotemporal events in the other two networks to enhance the understanding of the information flow evolution by capturing the dynamics of the interactions involving the traffic flow and the inter-vehicle communication layers. The graph-based approach enables the computationally efficient tracking of information propagation using a simple graphbased search algorithm and the computationally efficient storage of information through a single graph database. Second, this dissertation proposes analytical approaches that enable theoretical investigation into the qualitative properties of information flow propagation speed. The proposed analytical models, motivated from spatiotemporal epidemiology, introduce the concept of an information flow propagation wave (IFPW) to facilitate the analysis of the information propagation characteristics and impacts of traffic dynamics at a macroscopic level. The first model consists of a system of difference equations in the discrete-space and discrete-time domains where an information dissemination is described in the upper layer and a vehicular traffic flow is modeled in the lower layer. This study further proposes a continuous-space and continuous-time analytical model that can provide a closed-form solution for the IFPW speed to establish an analytical relationship between the IFPW speed and the underlying traffic flow dynamics. It can corporate the effects of congested traffic, such as the backward traffic propagation wave, on information flow propagation. Thereby, it illustrates the linkage between information flow propagation and the underlying traffic dynamics. Further, it captures V2V communication constraints in a realistic manner using a probabilistic communication kernel (which captures the probability). Third, within the integrated modeling framework, this dissertation captures the impact of information flow propagation on traffic safety and control applications. The proposed multi-anticipative forward collision warning system predicts the driver’s maneuver intention using a coupled hidden Markov model, which is one of statistical machine learning techniques. It significantly reduces the false alarm rates by addressing the uncertainty associate improves the performance of the future motion prediction, while currently available sensor-based kinematic models for addressing the uncertainty associated with the future motion prediction.

Degree

Ph.D.

Advisors

Peeta, Purdue University.

Subject Area

Communication|Epidemiology|Artificial intelligence|Economics|Information Technology|Recreation|Transportation

COinS