Intelligent Vehicle Driving Behavior Decision Algorithm In Complex Traffic Scenes

Whether in academia or industry,intelligent vehicles are the focus of research and transformation,as well as the research focus of overcoming technical problems and occupying the technological high ground.It has broad market application prospects and important research significance.The decision-making system is the "driving brain" in an intelligent vehicle.Whether the driving behavior it decides is reasonable or not directly affects whether the intelligent vehicle is safe and comfortable,and it also determines the level of autonomous driving of the intelligent vehicle.However,the intelligent vehicle decision-making system is not perfect.Traffic accidents caused by errors in the decisionmaking system occur from time to time.The complex urban traffic environment faced by intelligent vehicles is full of the constraints of traffic laws,the complexity of the interaction between vehicles,and the uncertainty of other vehicles that are difficult to observe.The above problems are very big challenges for intelligent vehicle decision-making systems,and they are also problems that need to be solved urgently in the current driving behavior decision-making algorithms.This paper aims at the challenging problem of vehicle driving behavior decisionmaking in the context of smart cities and intelligent transportation,and conducts theoretical research on it from two aspects: "construction of candidate driving behavior set" and "driving behavior decision-making adapted to complex traffic scenarios".A decisionmaking method for hierarchical processing of static traffic law information and dynamic traffic participant information is designed,and a simulation verification platform is built through the Matlab Automated Driving Toolbox,and the proposed driving behavior decision-making method is tested in a continuous traffic scenario.Firstly,in view of the problem that the existing methods usually combine traffic laws and participants to make decisions,which results in poor adaptability to traffic laws,this paper summarizes and extracts the existing traffic laws and regulations at the upper level of the driving behavior decision-making algorithm,and summarizes their factors into road markings,indicator lights,restriction signs,and the legality of behavior migration,so as to realize the systematic modeling of traffic laws and regulations,improve the adaptability to traffic laws.According to the current position and behavior of the intelligent vehicle,combined with the above four aspects of information,the construction of the candidate driving behavior set is completed to ensure the legitimacy of the candidate behavior.Secondly,according to the information of dynamic traffic participants,four indicators are formulated at the lower level of the decision-making algorithm: risk estimation,human-like lane selection,mandatory lane change evaluation,and impact on following vehicles: in order to fully consider the uncertainty of vehicle driving and driving habits of drivers,an accident risk assessment method based on intention recognition and motion prediction is proposed;when calculating the human-like lane selection index,the weighted vehicle speed of the target lane,the feasible approach space ahead,the density of vehicles in the target lane and the type of following vehicles are comprehensively considered for weighted calculation;obtain the mandatory lane change evaluation index by comparing the lateral distance between the target lane and the exit lane that meets the requirements of the macro path;quantify the impact of lane change behavior on the following vehicles based on the collision time.Taking into account that different indicators have different priorities,the first-level mandatory lane changing threshold and the second-level accident risk threshold are set,and the corresponding driving behavior is selected as the optimal driving behavior according to the optimal utility.The method proposed in this paper can process the information of dynamic traffic participants relatively completely,has strong adaptability to the scene,and conforms to the characteristics of human driving behavior.Finally,the Matlab Automated Driving Toolbox is used to build a simulation verification platform,and the autonomous driving ability of the smart car controlled by the proposed method is simulated and tested in a time-continuous scenario.Through 6 different typical scenarios including the process of self-car lane change,closed road section and vehicle cut-in,and intersection,the effectiveness and rationality of the decision-making method are verified,and the decision result is highly consistent with the human decision result of the driver while driving.