Context Aware Drivers' Behaviour Detection System for VANET
Wireless communications and mobile computing have led to the enhancement of, and improvement in, intelligent transportation systems (ITS) that focus on road safety applications. As a promising technology and a core component of ITS, Vehicle Ad hoc Networks (VANET) have emerged as an application of Mobile Ad hoc Networks (MANET), which use Dedicated Short Range Communication (DSRC) to allow vehicles in close proximity to communicate with one another, or to communicate with roadside equipment. These types of communication open up a wide range of potential safety and non-safety applications, with the aim of providing an intelligent driving environment that will offer road users more pleasant journeys. VANET safety applications are considered to represent a vital step towards improving road safety and enhancing traffic efficiency, as a consequence of their capacity to share information about the road between moving vehicles. This results in decreasing numbers of accidents and increasing the opportunity to save people's lives. Many researchers from different disciplines have focused their research on the development of vehicle safety applications. Designing an accurate and efficient driver behaviour detection system that can detect the abnormal behaviours exhibited by drivers (i.e. drunkenness and fatigue) and alert them may have an impact on the prevention of road accidents. Moreover, using Context-aware systems in vehicles can improve the driving by collecting and analysing contextual information about the driving environment, hence, increasing the awareness of the driver while driving his/her car. In this thesis, we propose a novel driver behaviour detection system in VANET by utilising a context-aware system approach. The system is comprehensive, non-intrusive and is able to detect four styles of driving behaviour: drunkenness, fatigue, reckless and normal behaviour. The behaviour of the driver in this study is considered to be uncertain context and is defined as a dynamic interaction between the driver, the vehicle and the environment; meaning it is affected by many factors and develops over the time. Therefore, we have introduced a novel Dynamic Bayesian Network (DBN) framework to perform reasoning about uncertainty and to deduce the behaviour of drivers by combining information regarding the above mentioned factors. A novel On Board Unit (OBU) architecture for detecting the behaviour of the driver has been introduced. The architecture has been built based on the concept of context-awareness; it is divided into three phases that represent the three main subsystems of context-aware system; sensing, reasoning and acting subsystems. The proposed architecture explains how the system components interact in order to detect abnormal behaviour that is being exhibited by driver; this is done to alert the driver and prevent accidents from occurring. The implementation of the proposed system has been carried out using GeNIe version 2.0 software to construct the DBN model. The DBN model has been evaluated using synthetic data in order to demonstrate the detection accuracy of the proposed model under uncertainty, and the importance of including a large amount of contextual information within the detection process.
- PhD