Research On Key Technologies Of Traffic Information System Based On VANETs And Intellisense

With the continuous growth of the economy and the rapid development of the automobile manufacturing industry,the number of cars continues to increase.This has led to a series of social problems such as traffic accidents,vehicle theft,road congestion,and environmental pollution.In order to solve the above problems,the intelligent transportation system has emerged.It integrates modern communication technology,sensor technology,automatic control technology and data processing technology into the entire transportation system,making the transportation network more efficient,safe,convenient and comfortable,and establishing a transportation system coordinated with modern social civilization.It aims to maximize the efficient use of public resources by working closely with roads,vehicles and people to reduce traffic accidents,ease traffic congestion,improve transport efficiency,save resource consumption,and mitigate environmental pollution.As two key technologies in intelligent transportation systems,Vehicular Ad Hoc Networks(VANETs)and intelligent sensing have attracted widespread attention from industry and academia.Currently,smartphones equipped with rich sensing technology have powerful perception,computing,storage and communication functions,making smartphone sensing a good solution for IntelliSense in vehicle-mounted.Faced with the outstanding problems of the current transportation system,this dissertation focuses on the key technologies in the VANETs and IntelliSense traffic information system.Based on the comprehensive analysis and summary of VANETs and IntelliSense,the application feasibility of combining VANETs with smartphone sensing and deep learning technology in traffic information system is discussed in detail.The research and exploration on vehicle safety,transportation efficiency and safe driving of traffic information system were conducted respectively.To solve these problems,several achievements are gained in this dissertation.The major contributions of this dissertation are as follows:1.In view of the fact that vehicle theft has become an increasingly serious public safety issue,a low-cost vehicle anti-theft tracking system based on an obsolete smartphone called PhoneInside is proposed.The function of vehicle anti-theft and tracking can be achieved without the use of additional special device.Specifically,in the identification of illegal driver intrusion,the driver’s identity information is verified based on the ad hoc network authentication.In order to eliminate false alarms,based on historical driving habits and travel trajectories,the system can establish a individual mobility model of the vehicle and distinguish illegal driving behaviors by the LSTM(Long Short-Term Memory)network.In order to achieve accurate vehicle tracking,A novel VelocityAware Dead Reckoning(VA-DR)method is presented,which utilizes map knowledge and vehicle’s turns at road curves and intersections to estimate velocity for trajectory computation.Compared with traditional dead reckoning,the cumulative error is effectively reduced.The realistic experiments based on the driving data prove that PhoneInside can detect theft of vehicles effectively,track and locate a stolen vehicle accurately.At the same time,the system provides a new solution for the reuse of obsolete smartphones.2.Aiming at the problem of low efficiency in current traffic transportation,a ParkingLot-Assisted Carpool(PLAC)method based on Vehicular ad hoc networks is proposed.The historical travel data of parked vehicles is collected and exchanged in common or adjacent parking lots by VANETs,and the future trajectories of parked vehicles are analyzed and predicted based on the data.The shared carpooling service is provided for drivers with the same or similar journeys.Specifically,the parking lot is regarded as a distributed cluster,and the parked vehicles are managed according in a cluster structure.The cluster head node maintains the information of the member vehicles.The on-board equipment records the driving track of the vehicle,and synchronizes the travel data to the cluster head node to save through VANETs after entering the parking lot cluster.Then,in order to cover the drivers with carpooling requirements in the nearby parking lot,an effective routing strategy for the cluster head to submit the trajectory information of parked vehicle to the surrounding parking lot clusters is established.The appropriate coincidence degree scheme of vehicle trajectory is designed to select the best carpool driver.The simulation results based on real city map and parking lot data show that compared with the existing two carpooling methods,the proposed PLAC effectively improves the vehicle transportation capacity.3.It has been proven to be an effective method to judge the driver’s fatigue state by monitoring changes in the driver’s respiratory rate in safe driving.We use the smartphone’s microphone to sense and collect the driver’s breathing information,and regard respiration as a special speech recognition to calculate breathing frequency.A finegrained breathing frequency monitoring model called DeepFilter based on deep learning is proposed for smart phones.DeepFilter is a bi-directional Recurrent Neural Network started with input of multiple convolutional layers,and trained in an end-to-end manner using the cross entropy loss function.In addition,batch normalization is adapted to speed up the training.Specifically,a frame-aligned training label is created in breathing rate monitoring,and it is translated to a classification problem at the frame level to determine whether the frames belong to inhaling/exhaling or not.On the input sequence of convolution layer,the time domain sequence of the fixed length frame is converted into time-frequency spectrum by Fast Fourier Transform,so the diversity of respiratory speech signal is overcome by using the translation invariance of convolution.An inhalation or exhalation event may involve several frames,so bidirectional recurrent layers can effectively capture sequence history information to improve performance.After collecting a large amount of respiratory data and post-processing for noise reduction to train the mode,the comparison experimental results data show that DeepFilter has a superior accuracy in breathing frequency monitoring.Based on the realistic experiment of vehicle and sleep environments,the results show that DeepFilter has good practicability.