Research On Collision Warning And Decision Algorithm For Traffic Safety-Related Applications

Despite the rapid development and improvement of Intelligent Transportation Systems(ITS),increasing car ownership still causes serious traffic accidents.How to accurately detect accident risks and adopt effective early warning measures to establish a safe traffic environment has always been an important research hotspot,which is essential to protect life and property.At the same time,the use of on-board sensors,communication and computing technologies has stimulated research on new types of traffic safety applications.However,the complex and variable traffic environment leads to unstable sensor performance,inaccurate positioning data,and unreliable communication.In addition,the driver’s driving characteristics and status can also cause distraction,misjudgment,and misoperation while driving,as well as varies the degree of response to the warning information.The above situations bring great challenges to the corresponding accident detection,accident warning,and decision control algorithm design.The Autonomous Vehicle(AV),as a major technological breakthrough in the future automotive and transportation industry,has more advanced sensing,communication,decisionmaking,and control capabilities,and can greatly liberate drivers while reducing traffic accident rates.However,the development of AVs will bring a mixed traffic scenario where AVs and driver-controlled(or common)vehicles coexist.In addition,massive sensing data will be generated.In the face of increasingly complex traffic environments and large amounts of traffic data,Artificial Intelligence(AI)algorithms can quickly and accurately process large amounts of data and adapt to the driving environment,which can play an important role in improving traffic safety.Based on this,this thesis mainly studies the traffic safety issues in the current mixed traffic scenarios.Specifically,this thesis combines the Internet of Vehicles(IoV)technology with AI algorithms for traffic safety applications,and to design effective warning,accurate positioning,and precise control algorithms in consideration of the characteristics of mixed traffic scenarios,aiming at improving traffic safety and passenger’s driving experience.This thesis investigates the application of communication technology and AI technology in traffic safety issues.Firstly,this thesis analyzes the scenario of vehicle rear-end collision and intersection collision accidents and designs effective collision avoidance algorithms.In addition,considering the importance of vehicle positioning information in safety applications,a vehicle positioning error correction method in mixed traffic scenarios is studied.Finally,based on the effective early warning algorithm and accurate positioning information,this thesis explores the vehicle autonomous braking control strategy in emergency scenarios,thereby further improving the success rate of collision avoidance.According to the chapter setting,the main research work and contributions of this thesis can be summarized as follows.(1)Research on neural network(NN)-aided graded warning algorithm of rear-end collision.Vehicle rear-end collisions account for a very high proportion of traffic accidents.However,achieving ultra-low delay and ultra-high reliability collision avoidance in this scenario still faces challenges such as positioning errors,inaccurate risk assessment,and difficulty in ensuring passenger comfort.Aiming at the above problems,an NN-aided graded warning algorithm is proposed.First,the vehicular communication technology is used to judge whether the leading vehicle and the following vehicle is in the same lane,which is used to compensate for the influence of the positioning error.In addition,an online NN model is proposed for risk assessment.Finally,considering the urgency of the accident and the comfort of the passengers,a graded warning strategy to achieve reasonable avoidance of the rear-end collision is proposed.The effectiveness and accuracy of the proposed algorithm outperform existing algorithms in terms of relative lane positioning,risk assessment,and collision avoidance.(2)Design of infrastructure-cooperated intersection collision avoidance algorithms.Traffic accidents at intersections are particularly important for the performance of traffic networks.However,the characteristics of intersection accidents limit the development of collision avoidance methods.Based on this,an infrastructure-cooperated intersection collision avoidance algorithm is proposed.Firstly,the Dynamic Bayesian Network(DBN)is used to model the vehicle state evolution.The model can deal with some uncertain factors such as driver’s behavior,and overcome the defects of the previous model that need to know the full information of the vehicle state.At the same time,the risk assessment is adopted to distinguish the driver’s intentions and safe behavior at intersections,which avoids the complicated vehicle trajectory prediction process.Finally,according to different dangerous situations,the corresponding collision avoidance warning is made.(3)Research on vehicle positioning error correction method in mixed traffic scenarios.Aiming at the mixed traffic scenario where Connected and Autonomous Vehicles(CAVs)and driver-controlled(or common)vehicles coexist,a vehicular blockchain-based secure and efficient GPS positioning error evolution sharing framework is proposed for improving vehicle positioning accuracy,which fully exploits the characteristics of blockchain and edge computing.First,by analyzing the GPS error,a bridge can be established between the sensor-rich vehicles and the common vehicles to achieve cooperation by sharing the positioning error evolution at a specific time and location.Moreover,the positioning error evolution is obtained by a deep neural network(DNN)-based prediction algorithm running on the edge server.Based on this,a blockchain-based vehicle error storage and sharing mechanism are proposed to ensure the security of cooperative vehicles and mobile edge computing nodes(MECNs).In addition,the corresponding smart contracts are designed to automate and efficiently perform storage and sharing tasks as well as solve inconsistencies in time scales.(4)Research on autonomous braking control strategy combined with reinforcement learning.The autonomous braking control is an effective way to reduce accidents caused by driver operation,and the autonomous braking control strategy in emergency situations still deserves further study in terms of effectiveness and accuracy.This thesis proposes an automatic braking control strategy combined with deep reinforcement learning(DRL).First,the vehicle lane-changing process and the braking process are analyzed in detail,and a multi-objective reward function is designed,which can compromise the rewards achieved of different brake moments,the degree of the accident,and the comfort of the passenger.Next,inspired by intelligence algorithms for vehicle autonomous control,DRL is used to develop optimal braking strategies.Finally,a typical actor-critic(AC)algorithm named deep deterministic policy gradient(DDPG)is adopted for solving the autonomous braking problem,which can improve the efficiency of the optimal strategy and be stable in continuous control tasks.