Driving Assistance Mechanism Design Based On Context Recognition In VANET

The quick development of automobile industry benefits people a lot. However, the accompanying traffic accidents and traffic jams have been one of the most enormous problems which need to be solved urgently in modern society. The three major causes of traffic accidents, namely, drunk driving, fatigue driving and over-speeding are almost all resulted by drivers’ improper operations. Hence, it is necessary to study and analyze people’s driving behavior so as to design effective measures to reduce the incidence rate of traffic accidents. The dissertation proposes a driving assistance mechanism based on context recognition in VANET through collecting and judging the driver’s environment and behaviors.First, this dissertation presents definitions and classification of the concepts which are often mentioned in driving behavior. Secondly, we set up a three-layer driving behavior model based on BP neural network. The three-layer model can be divided into perception and processing layer, driving behavioral decision layer, and application layer. Data of human, vehicles and road conditions can be collected though sensors and then transferred to BP neural network. Then, a driver’s status can be judged though studying and analyzing his or her driving behavior model. Once abnormal driving behavior is observed, relevant reminding measures can be taken and messages can be sent out through VANET according to specific situation. Especially, we propose a pre-judgment method towards the most hazardous drunk driving in this dissertation. Data and distribution of alcohol concentration in the air can be obtained through installing alcohol detectors in three different locations in a car. Status of driver can be judged according to Fuzzy Logic Control Theory, then corresponding treatment can be processed in accordance with the results.On the other hand, speeding behavior of each driver stands for his urgent need to reach the destination. Over-speeding, which belongs to reckless driving, is one of the most extreme driving behavior. The time setting of traffic lights doesn’t match the distribution of traffic flow, which will result in low efficiency of traffic capacity and dangerous over-speeding driving behavior at crossing roads. Therefore, the dissertation proposes a concept of driver’s fast marching aspiration value (Will-Value). It combines driver’s will-value with the virtual traffic lights (VTL). The phase and time of traffic lights can be dynamically changed according to the actual traffic distribution at the crossing roads and the average will-value of vehicles on different lanes. The experiment data shows that the method proposed in this dissertation can efficiently improve the traffic capacity and decrease time waste due to unnecessary wait, especially for those crossing roads with uneven distribution of vehicles.