Modeling Trust in Multiagent Mobile Vehicular Ad-Hoc Networks through Enhanced Knowledge Exchange for Effective Travel Decision Making

Abstract

This thesis explores how to effectively model trust in the environment of mobile vehicular ad-hoc networks. We consider each vehicle’s travel path planning to be guided by an intelligent agent that receives traffic reports from other agents in the environment. Determining the trustworthiness of these reports is thus a critical task. We take as a starting point the multi-dimensional trust model of Minhas et al. That work had a two-phased approach: i) model trust and ii) execute an algorithm for using that trust modeling, when deciding what route to take. The framework presented in this thesis aims to clarify i) the messaging that should be supported, ii) the internal representation of the messaging and the trust information and iii) the algorithms for sending and receiving information (thus updating knowledge) in order to perform decision making during route planning. A significant contribution is therefore offered through clarification and extension of the original trust modeling approach. In addition we design a comprehensive, extensive simulation testbed that is used to validate the effectiveness and robustness of the model. This testbed supports a variety of metrics and is able to perform testing in environments with a large number of cars. This constitutes the second significant contribution of the thesis.

Overall, we present a valuable model for knowledge management in mobile vehicular ad-hoc networks through a combination of trust modeling, ontological representation of concepts and facts, and a methodology for discovering and updating user models. Included is a representation and implementation of both a push-based and pull-based messaging protocol. We also demonstrate the effectiveness of this model through validation conducted using our simulation testbed, focusing first on a subset of the multi-faceted trust model in order to highlight the value of the underlying representation, decision making algorithm and simulation metrics. One very valuable result is a demonstration of the importance of the combined use of the different dimensions employed in the trust modeling.