Computer Science Department News

There will be a Ph.D. Defense on March 25, 2011 at 10:00 AM by Mahmoud Abuelela.

Title: A Framework for Incident Detection and Notification in Vehicular Ad-hoc Networks.
Location: E&CS Building first floor auditorium
Student: Mahmoud Abuelela
Advisor: Dr. Olariu

Abstract:
The US Department of Transportation (US-DOT) estimates that over half of all con- gestion events are caused by highway incidents rather than by rush-hour traffic in big cities. The US-DOT also notes that in a single year, congested highways due to traffic incidents cost over $75 billion in lost worker productivity and over 8.4 billion gallons of fuel. Further, the National Highway Traffic Safety Administration (NHTSA) in- dicates that congested roads are one of the leading causes of traffic accidents, and in 2005 an average of 119 persons died each day in motor vehicle accidents.
Recently, Vehicular Ad-hoc Networks (VANET) employing a combination of Vehicle- to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) wireless communication have been proposed to alert drivers to traffic events including accidents, lane closures, slowdowns, and other traffic-safety issues.
In this thesis, we propose a novel framework for incident detection and notification dissemination in VANETs. This framework consists of three main components: a system architecture, traffic incident detection engine and a notification dissemina- tion mechanism. The basic idea of our framework is to collect and aggregate traffic- related data from passing cars and to use the aggregated information to detect traffic anomalies. Finally, the suitably filtered aggregated information is disseminated to alert drivers about traffic delays and incidents.
The first component of our framework is an architecture for the notification of traffic incidents, NOTICE for short. In NOTICE, sensor belts are embedded in the road at regular intervals, every mile or so. Each belt consists of a collection of pressure sensors, a simple aggregation and fusion engine, and a few small transceivers. The pressure sensors in each belt allow every message to be associated with a physical vehicle passing over that belt. Thus, no one vehicle can pretend to be multiple vehi- cles and there is no need for an ID to be assigned to vehicles.
Vehicles in NOTICE are fitted with a tamper-resistant Event Data Recorder (EDR), very much like the well-known black-boxes onboard commercial aircraft. EDRs are responsible for storing vehicles behavior between belts such as acceleration, deceler- ation and lane changes. Importantly, drivers can provide input to the EDR, using a simple menu, either through a dashboard console or through verbal input.
The second main contribution of this thesis is to develop incident detection tech- niques that use the information provided by cars in detecting possible incidents and traffic anomalies using intelligent inference techniques. For this purpose, We devel- oped deterministic and probabilistic techniques to detect both blocking incidents, accidents for examples, as well as non-blocking ones such as potholes. To the best of our knowledge, we are the first in attempting to have a complete automatic inci- dent detection techniques using VANETs specially to detect non blocking incidents like potholes. Our third contribution is to develop data dissemination techniques specifically adapted to VANETs. For this purpose, we start by performing analyt- ical analysis, confirming empirical results found by other researchers, proving that under many highway scenarios VANETs tend to be disconnected, consisting of a collection of disjoint clusters. With this in mind, we have developed data dissemi- nation approaches that efficiently propagate messages between cars and belts on the road. To the best of our knowledge, we are the first in VANETs community to prove analytically that disconnection is the norm rather than the exceptions in VANETs. The proposed framework has been enhanced with security and privacy techniques to avoid possible attacks from malicious drivers as well as preserving driver's privacy. Extensive simulation has revealed that the proposed techniques and algorithms out- perform the best-known techniques in use today.