Research On Lane-changing Game And Decision-making Control Considering Driving Style And Interaction Between Vehicles

Autonomous driving technology is considered as a promising means to reduce road traffic accidents and improve traffic safety and efficiency,and thus has been extensively studied in both industry and academia.Therefore,many universities and technology companies are committed to developing higher levels of autonomous driving technology.The lane-changing decision of intelligent vehicles is one of the most basic decision-making behaviors of autonomous vehicles,which has a crucial impact on the safety,comfort and traffic efficiency of vehicles.Research on the interaction between intelligent vehicles and human-driven vehicles will help promote the personalized development of high-level automatic driving assistance systems and achieve human-like decision-making.The main research work of this paper is as follows:First,a driving style classification study is conducted based on naturalistic driving data.The NGSIM public data set is used as the data source for driving style recognition under lane-changing conditions in this paper.The original data is filtered and outliers are processed using the symmetrical exponential moving average filter algorithm.The driving style multivariate feature matrix is composed of the average speed of horizontal and vertical lane changes,the average acceleration of horizontal and vertical directions,the maximum value of speed,and the standard deviation of speed,and the interactive parameters are introduced as new feature quantities of driving style clustering for lane changing.Accuracy,analyzing the basic characteristics and changing laws of lane changing.Then design an improved K-means driving style recognition algorithm that can dynamically eliminate redundant features.According to the clustering results,the lane-changing behavior characteristics of people with different driving styles under dynamic lane-changing conditions and the differences in motion parameters such as acceleration and speed are analyzed.In order to introduce driving style characterization parameters in the subsequent development of smart car final calculation algorithms to achieve personalized decision-making.Secondly,research on the lane-changing game and decision-making considering workshop interaction and driving style is carried out.Analyze the dynamic interaction and game mechanism between the main vehicle and the surrounding vehicles,and use the Stackelberg game theory to describe the micro-interaction and game behavior between themain vehicle and the conflicting vehicles,and determine the game object,strategy space and profit function of the lane change.Aiming at solving the two-level optimization problem in equilibrium,a two-level hybrid particle swarm optimization algorithm is designed to solve the game equilibrium,and then the optimal strategy of the vehicles participating in the game and the optimal acceleration of longitudinal planning are obtained.Then,a coupled DRF-MPC controller is designed based on the Driving Risk Field(DRF)theory and the Model Predictive Control(MPC)theory.The risk field models of obstacle vehicles and structured roads are respectively established,and then the unified risk modeling of the driving environment of the main vehicle is realized;combined with the MPC control theory,the movement prediction of the main vehicle changing lanes is performed,and the field strength distribution of the predicted trajectory position As a key component of the composite cost function of MPC.The controller is coupled with the lateral planning and tracking control functions of the smart car,which can adjust the DRF-MPC control amount to track the planned lane-changing trajectory while obtaining the lowest-risk lane-changing trajectory,and feed back the motion state at the next moment to the decision-making module,to solve iteratively at the next moment.Finally,the validity and scalability of the model are verified through simulation tests.Based on Matlab/Simulink,a typical two-lane lane-changing scene,a two-lane overtaking scene and a three-lane multi-vehicle game scene are established.The simulation results show that smart cars can play games in different scenarios and under different combinations of driving styles,which can achieve safe and reasonable Lane-changing decision-making and planning control effectively simulate the dynamic game process of workshops in real traffic scenarios,and verify the effectiveness and scalability of the algorithm.