Research On High Speed Lane Change Game Decision And Planning Considering Driving Style

In the foreseeable future traffic scenarios,human-driven vehicles,human-robot collaborative vehicles and autonomous vehicles will share the right of way and interact with each other during driving.The decision-making model adopted by the existing lane-changing assisted driving system does not fully consider the differences in driving styles among traffic participants and the individual needs of its own occupants in decision-making.It is difficult to cope with the changes in driving behavior of vehicles with different driving styles during the lane-changing process.And it focuses more on the joint development of intention recognition and decision-making,ignoring the matching with the planning control part,which makes it difficult to improve the tracking accuracy.In view of the above problems,this paper takes autonomous vehicles as the research object,and studies the autonomous lane change problem considering driving style in detail.Taking the NGSIM driving data set as the data source,the data is cleaned and S-G filtered,and 16 driving style characteristic parameters are preliminarily selected,and the data is reduced by factor analysis method.The optimal number of clustering clusters is determined to be 3 by AIC information criterion,BIC information criterion,elbow method and average contour method.Then,the data after dimensionality reduction are input into K-Means clustering algorithm to obtain three types of driving style data : conservative,normal and aggressive,and the corresponding distribution of driving characteristics,which provides a basis for the design of decision-making model and the simulation of different driving styles for lane change decision.In order to improve the matching between decision-making and planning control parts,a closed-loop lane-changing decision-making model architecture based on Stackelberg game is constructed by using complete information non-cooperative dynamic game theory to realize the integration of longitudinal speed planning and decision-making.In addition,the Stackelberg game decision-making model is proposed,and the driving style elements of each vehicle are introduced into the model.The decision-making cost considering safety,comfort and traffic efficiency is established for the self-vehicle and the obstacle vehicle.Different driving styles are simulated by assigning different weight coefficients to the three.Based on this,the particle swarm optimization algorithm is introduced to solve the Stackelberg game decision-making model,and the optimal lane-changing decision considering driving style is obtained.In order to ensure the safe and collision-free operation of autonomous vehicles,a DPFNMPC planning controller is designed.The risk potential field of road markings and the conical dynamic risk potential field of obstacle vehicles are established to characterize the risk of autonomous vehicles during driving,and the relative velocity element is introduced into the conical dynamic risk potential field to improve the adaptability of the risk potential field.In order to consider the kinematic constraints of the vehicle prediction model in motion planning,the total risk potential field is used as the penalty term of the NMPC part,and the optimal control quantity is obtained by solving the constrained nonlinear programming problem.The model proposed in this paper is simulated and verified under different driving styles.The results show that the closed-loop lane-changing decision-making model architecture based on Stackelberg game and the proposed Stackelberg game decision-making model can effectively combine the interaction behavior of self-vehicle and obstacle vehicle and driving style characteristics to make correct lane-changing decisions under different test conditions,which verifies the feasibility and effectiveness of the proposed model.