(VANET IR-CAS): Utilizing IR Techniques in Building Context Aware Systems for VANET

Abstract

Most of the available context aware dissemination systems for the Vehicular Ad hoc Network (VANET) are centralized systems with low level of user privacy and preciseness. In addition, the absence of common assessment models deprives researchers from having fair evaluation of their proposed systems and unbiased comparison with other systems. Due to the importance of the commercial, safety and convenience services, three IR-CAS systems are developed to improve three applications of these services: the safety Automatic Crash Notification (ACN), the convenience Congested Road Notification (CRN) and the commercial Service Announcement (SA). The proposed systems are context aware systems that utilize the information retrieval (IR) techniques in the context aware information dissemination. The dispatched information is improved by deploying the vector space model for estimating the relevance or severity by calculating the Manhattan distance between the current situation context and the severest context vectors. The IR-CAS systems outperform current systems that use machine learning, fuzzy logic and binary models in decentralization, effectiveness by binary and non-binary measures, exploitation of vehicle processing power, dissemination of informative notifications with certainty degrees and partial rather than binary or graded notifications that are insensitive to differences in severity within grades, and protection of privacy which achieves user satisfaction. In addition, the visual-manual and speech-visual dual-mode user interface is designed to improve user safety by minimizing distraction. An evaluation model containing ACN and CRN test collections, with around 500,000 North American test cases each, is created to enable fair effectiveness comparisons among VANET context aware systems. Hence, the novelty of VANET IR-CAS systems is: First, providing scalable abstract context model with IR based processing that raises the notification relevance and precision. Second, increasing decentralization, user privacy, and safety with the least distracting user interface. Third, designing unbiased performance evaluation as a ground for distinguishing significantly effective VANET context aware systems.