Evaluation Of Intelligent Vehicle Lane Change Model Based On Hybrid Game In Urban Environment

Traffic accidents caused by lane changes and traffic delays have been on the rise in recent years.By taking advantage of the environmental perception advantages of intelligent cars and introducing the lane changing decision-making model to guide the vehicles to change lanes reasonably,safely and comfortably,the driving safety can be effectively guaranteed,the driving delay can be reduced,and the road resources can be fully and reasonably utilized.And lane changing decision is vital part of the intelligent vehicle behavior decision,in the field of microscopic traffic,combined with intelligent made car environmental awareness and the advantage of the cloud platform communication,set up lane-changing model,for smart car made to deal with complex traffic environment to provide decision support,intelligent decision and automatic drive of the vehicle.In order to solve the problem of choosing the lane change model for intelligent vehicle overtaking,a typical intelligent vehicle lane change model needs to be evaluated.Traditional evaluation of intelligent vehicle lane-changing model is based on a single lane changing model validation under simple scenario,the lane changing behavior under the environment of urban road is a more complex strategy game behavior,which influenced by itself in the strategy,also affected by the adjacent traffic lane changing strategy,is a mixed strategy game,to its evaluation needs to macroscopic traffic scene as the backing,construct mixed traffic scene,can reflect the variety of lane changing model in urban road traffic behavior characteristics of the scene.The construction of urban road traffic scene for hybrid strategy game not only needs to be able to simulate a large number of traffic objects,but also needs to be able to describe the microscopic driving characteristics and lane changing behavior of vehicles.Therefore,we need to build simulation systems and scenarios that can realize real-time interaction between micro traffic simulation model and micro vehicle lane changing model.In view of this,this paper proposes a data-driven construction of urban road traffic scenarios,and at the same time selects a typical intelligent vehicle lane change model for algorithm construction.On the basis of game theory,the analysis method of macroscopic and microscopic evaluation analysis of lane change model takes into account the two experimental means of pure strategy game and mixed strategy game.The main work and achievements of this paper are summarized as follows:1.Analysis of lane changing model and algorithm construction of intelligent vehicle overtaking.The typical lane changing models at home and abroad are investigated and analyzed and divided into four categories.They are: rule-based lane change model,utility-based lane change model,fuzzy logic-based lane change model and prediction-based lane change model.In this paper,the classical model of four types of lane changing models is selected for detailed analysis,and the algorithm is implemented by Matlab.2.Based on data-driven,the urban road traffic scene was constructed with the micro traffic flow simulation software SUMO as the platform,which was used as the scenario support for lane change strategy evaluation.Firstly,based on the floating vehicle data and bayonet flow data in some regions of shenzhen,data processing was carried out,OD pairs were divided,and map matching was performed.Finally,the OD matrix was solved.Then take SUMO as the platform,import OD matrix,automatically generate the executable route,complete the construction of urban traffic scene,and based on bayonet data,complete the verification of urban traffic scene;Finally,based on TraCI4 Matlab interface,the real-time interaction between urban road traffic scene and Matlab lane change algorithm is realized,and the construction of hybrid game simulation platform is completed.3.Using the hybrid game simulation platform and based on the game experiment,the macroscopic evaluation of the lane changing strategy of the intelligent car is completed.The mixed strategy game experiments of pure strategy and four different probability distributions of lane changing models were carried out under free flow,flat peak flow and peak flow,respectively.The results show that there are significant differences in the performance of lane changing models under different traffic flows.At the same time,this paper also gives the simulation evaluation results of the mixed strategy game of various combinations under different traffic flow densities,which provides a theoretical basis for the lane changing model selection of intelligent vehicles according to the current traffic flow density.4.Completed the micro evaluation of intelligent vehicle lane changing strategy based on the game experiment.Firstly,road subdivision is carried out to complete the micro-process modeling of lane changing behavior.Then,the pure strategy and mixed strategy game experiments of lane changing model were carried out under free flow,flat peak flow and peak flow,respectively.The micro-evaluation of the intelligent vehicle lane changing model was completed by taking the acceleration distribution at lane changing time and the spacing distribution of the vehicle in front as the evaluation indexes of comfort and safety.The evaluation results reveal the distribution characteristics of the acceleration and the distance between the vehicles in front of the vehicle under different traffic flow densities.At the same time,for the comfort and safety of intelligent vehicle lane changing process,under different traffic environments and traffic flow density,all lane changing strategies cannot guarantee the superiority in any traffic environment and traffic density,and no lane changing model has the lowest profit in all traffic environments and traffic density.5.Through the comparative analysis of mixed strategy game and pure strategy game,the importance of mixed strategy game to the evaluation of lane change model is revealed,which provides a more objective evaluation method for the evaluation of lane change model.By pure strategy and mixed strategy in to the model of radar contrast analysis diagram can be obtained in a pure strategy or fixed under mixed strategy can be found in line with the evaluation index of the optimal lane changing model,but because of the complexity of the road traffic scene,a variety of real widespread of lane changing strategy combinations,only use mixed strategy game on specific strategy distribution of concrete analysis is useful,but also further proves the mixed strategy game exchange model evaluation of objective correctness.