Towards efficient and secured intelligent transportation system (ITS)

Author/Creator

Author/Creator ORCID

Date

2015-08-28

Department

Towson University. Department of Computer and Information Sciences

Program

Citation of Original Publication

Rights

Copyright protected, all rights reserved.
There are no restrictions on access to this document. An internet release form signed by the author to display this document online is on file with Towson University Special Collections and Archives.

Subjects

Abstract

According to the National Highway Traffic Safety Administration (NHTSA), in the U.S., road traffic congestions, and accidents are responsible for over $80 billion economic loss, and over 32,800 deaths per year. Intelligent transportation system (ITS)/vehicular ad hoc networks (VANETs), however, promises improved mobility/traffic efficiency, safety, security, and greener transportation, etc. using vehicle-to-vehicle (V2V), and/or vehicle-to-infrastructure (V2I) communication. However, in light of the aforementioned challenges, these proclaimed levels of improvements have not fully/comprehensively been critically evaluated/examined especially in a realistic setting i.e. using real-world data, and road networks as corroborated by several authors/authorities in the ITS/VANET domain [1-14]; as a result, one of the major goals of this dissertation is to fill this pertinent gap. Consequently, in this dissertation research, using both real-world road traffic data consisting of a total of 6 months traffic data of the Maryland (MD)/Washington DC and Virginia (VA) areas from July 1st, to December 31st, 2012 - of which 6 weeks of this was used as a representative sample after a comprehensive/exhaustive data analysis - and real-world road networks, we first evaluate the performance of two popular vehicular routing algorithms namely: A* (Astar), and Dijkstra's routing algorithms respecting travel time performance in our developed generic real-world ITS test-bed using both small, and large road networks. Next, using the two major VANET architectures - V2V, and V2I communication architectures - we evaluate their performance respecting safety and traffic efficiency. In order to do this, we developed a mobile application we called Incident Warning Application (IWA) of which IWA-equipped vehicles utilize this application to evade a compound road accident consisting of a blocking of the entire roadway lanes, presence of slippery/frozen ice, and reduced speed limit as a result of fog. Vehicles (classic vehicles) unequipped with this mobile application are unaware of this congested condition - they, therefore, drive heedlessly unto the congested road and eventually suffer the consequences in the form of delayed arrival time/increased travel time. In addition, we analyze the performance of V2V and V2I communication in the presence of a type of denial of service (DoS) attack - jamming attack - with the view of ascertaining which is most resilient/effective when part of the system is under attack or is being compromised also respecting the evaluation metrics of traffic efficiency, and safety. Also, using our real-world data, and road network, we evaluated the performance of over 24 supervised machine learning classification, and regression algorithms with respect to the evaluation metrics of predictive accuracy, and prediction speed with the view of having a comprehensive, and comparative reference manual i.e. a taxonomy. Finally, we examine the influence of driver distractions/attentiveness on traffic efficiency, and safety performances with our developed Driver Notification Application (DNA) using two popular driver models/age groups - young drivers (ages 16 - 25 years), and middle-age drivers (ages 30 - 45 years) respectively employing ad hoc/decentralized communication. Overall, our results show that no significant difference respecting travel time performance was observed between Dijkstra and A* (Astar) algorithms in both small, and large road networks. Next, V2I communication outperformed V2V communication respecting traffic efficiency, and safety performances before, and during the execution of the jamming (availability) attack. Also, classification tree (Ctree), and regression tree (Rtree) gave the best performances respecting predictive accuracy and prediction speed amongst all the algorithms examined/evaluated. In general, with respect to all other evaluated supervised machine learning algorithms, a tradeoff between speed, and accuracy is imperative and will be largely dependent on the scenario in question i.e. this tradeoff must be determined on an individual/case-by-case basis. Lastly, our results lucidly shows that middle-age drivers outperformed younger drivers respecting their ability to maintain their attention/concentration levels for longer time periods while in transit; thereby resulting in better safety, and traffic efficiency performances.