A pervasive prediction model for vehicular ad-hoc network (VANET)

The growth of city traffic has contributed to severe traffic congestion and traffic accidents in the most of the cities in the world. Since people’s travel demand rise at a rate usually greater than the addition of road capacity to lead many other issues, such as environmental problems and the quality of life. Intelligent Transportation System (ITS) is committed to solving the worsening traffic problems. Wide deployment of such ITS can eventually provide more dynamic, real-time and efficient solutions to transportation problems. ITS uses a variety of high technologies, especially electronic information technology and data communications technology to improve road traffic efficiency, road traffic safety and environmental protection. A number of researchers have depended on the wireless mobile communication to improve data collection and utilisation. The data could be used for early warning and forecasting traffic conditions in real-time. The benefit of wireless mobile communication research, especially Car to Car (C2C) communication is to abandon the expensive wireline-deployed and central processing units. Through the interconnection of many personal mobile devices, a low- cost freely extended, high-performance and parallel system can be formed. Car to Car communication can make possible efficient and reliable data transmission by wireless links in a traffic area. It is based on principles of mobile ad-hoc network (MANET) and applies to the domain of vehicles, being Vehicular ad-hoc network (VANET) which is a key component of ITS. The C2C communication system has become essential for driving safety and comfort and also for improving road condition. Also, the traffic prediction system is also an important part of ITS, traffic condition prediction can be regarded as an extension application of VANET. It provides traffic condition in advance via a variety of prediction models and helps the people make better driving safety, travel decisions and route selections regarding departure or driving time. The challenge of wireless traffic prediction technology is the uncertainty of traffic and real-time traffic data collection. It is widely known that urban transport system is a participatory, time-varying and complex nonlinear system. This uncertainty comes not only from the natural causes, such as seasonal and weather factors, but also from human factors, such as traffic accidents, emergencies and driver’s behaviour. In particular, the short-term traffic prediction is more affected by random interference factors. Current wireless traffic prediction research is usually based on a combination of wireless technology and traditional prediction model. The predictable traffic conditions include travel speed, travel time, traffic density, traffic accident, congestion level. However, in a large network environment, as the number of nodes increases, the transmission performance degrades and the prediction accuracy decreases because the prediction model does not obtain enough data. In this thesis, a novel traffic prediction framework (PPM-C2C) is proposed – Pervasive Prediction Model (PPM) based on the C2C communication. The framework utilises ad-hoc data via C2C communications for a short time traffic prediction in each car. This project builds and investigates the behaviour of a pervasive traffic simulation model in Ad-hoc network, with a particular part of it embedded into each vehicle’s equipment. It includes the data collection, aggregation and application aim to be running in all individual cars so that cars have up to date information on the traffic at all times. Moreover, those cars could predict the traffic conditions of a road section in a short time through the proposed prediction framework, especially travel speed prediction. When the car receives the current traffic information about other vehicles, the prediction system will incorporate the information, analyse the data and predict the traffic conditions of this road section for a future time. The design does not depend on any roadside communications infrastructure. It is a simple and flexible car communication and processing technology to collect real-time traffic information. This process will be aided by car to car wireless communication technology available nowadays. To achieve this goal, a mobility model adapted to VANET needs to be generated that a realistic city scenario based on the actual traffic traces is carried out through simulation. Based on this, we investigate the necessary influencing factors for predicted results. The simulation results illustrate that the prediction model can be applied to wireless network environment for a short time prediction, and our results demonstrate the viability and effectiveness of the proposed prediction framework over Car to Car communications. Furthermore, the wireless environment and derived factors can result in decreased application performance