Computer Science Department Events
There will be a PhD Defense on Friday July 27 at 9:00 AM by Samy El-TawabE&CS Building RM 3316TIME: 9:00 AMTITLE: FRIEND: A cyber-physical system for traffic Flow Related Information AggrEgatioN and DisseminationSPEAKER: Samy El-TawabAbstract:
FRIDAY July 27th
The major contribution of this thesis is to lay the theoretical foundations of FRIEND - A cyber-physical system for traffic Flow-Related Information aggrEgatioN and Dissemination. By integrating resources and capabilities at the nexus between the cyber and physical worlds, FRIEND will contribute to aggregating traffic flow data collected by the huge fleet of vehicles on our roads into a comprehensive, near real-time synopsis of traffic flow conditions. We anticipate providing the drivers with a meaningful, color-coded, at-aglance view of flow conditions ahead, alerting them to congested traffic. FRIEND can be used to provide accurate information about traffic flow and can be used to propagate this information. The workhorse of FRIEND is the ubiquitous lane delimiters (a.k.a. cat's eyes) on our roadways that, at the moment, are used simply as dumb reflectors. Our main vision is that by endowing cat's eyes with a modest power source, detection and communication capabilities they will play an important role in collecting, aggregating and disseminating traffic flow conditions to the driving public. We envision the cat's eyes system to be supplemented by road-side units (RSU) deployed at regular intervals (e.g. every kilometer or so). The RSUs placed on opposite sides of the roadway constitute a logical unit and are connected by optical fiber under the median. Unlike inductive loop detectors, adjacent RSUs along the roadway are not connected with each other, thus avoiding the huge cost of optical fiber. Each RSU contains a GPS device (for time synchronization), an active Radio Frequency Identification (RFID) tag for communication with passing cars, a radio transceiver for RSU to RSU communication and a laptop-class computing device. The physical components of FRIEND collect traffic flow-related data from passing vehicles. The collected data is used by FRIEND's inference engine to build beliefs about the state of the traffic, to detect traffic trends, and to disseminate relevant traffic flow-related information along the roadway. The second contribution of this thesis is the development of incident classification and detection algorithm that can be used to classify different types of traffic incident. Then, it can notify the necessary target of the incident. We also compare our incident detection technique with other VANET techniques.
Our third contribution is a novel strategy for information dissemination on highways. First, we aim to prevent secondary accidents. Second, we notify drivers far away from the accident of an expected delay that gives them the option to continue or exit before reaching the incident location. A new mechanism tracks the source of the incident while notifying drivers away from the accident. The more time the incident stay, the further the information need to be propagated. Furthermore, the denser the traffic, the faster it will backup. In high density highways, an incident may form a backup of vehicles faster than low density highways. In order to satisfy this point, we need to propagate information as a function of density and time.