Abstract
Autonomous vehicles, also known as self-driving cars, refer to vehicles that can travel on
public roads to destinations with minimal to no interaction with human beings. These types
of cars can respond to traffic-related incidents faster and more precisely than human
beings, thus potentially reducing the number of traffic accidents, subsequent pedestrian
injuries and even fatalities.
Autonomous vehicles have been a major focus of artificial intelligence research over the
past few years, with major developments being contributed to the industry by Google,
Bosch, and leading companies in the automobile industry.
The study presented in the dissertation explores the use of Intelligent Agents as a
computer science abstraction that encapsulates the several components of an
autonomous vehicle, in order to promote component modularity and to allow the inclusion
of newer technologies that could further improve the effectiveness of autonomous
vehicles.
A particular recent advancement in the field of autonomous vehicles is the use of intervehicle
communications, which supplements the array of sensors provided in the vehicles,
in case of failure or inability to produce sufficient data that would be necessary for the
vehicle to make a decision. The agent model proposed in the dissertation places a
paramount importance on the communications mechanism, incorporating it in its agent
architecture, in order to produce an autonomous vehicle model that is safer and more
effective than current solutions.
The autonomous vehicle agent model, given research constraints, was deployed in a
simulated 3D urban traffic environment, where it was tested in a number of scenarios
where a vehicle's sensors failed or provided insufficient data, preventing a safe journey for
the vehicle's passengers, passengers of other vehicles and pedestrians in the simulated
environment. The results of the tests demonstrated that an inter-vehicle communications
mechanism, even with limited transmission range, effectively complements the existing
modules of an autonomous vehicle, and is especially useful in case one of the modules
fails.
M.Sc. (Information Technology)