Intelligent Vehicle Driving Situation Assessment And Interactive Decisionmaking Considering Uncertainties

As the‘brain’of the intelligent vehicle,the driving decision-making module depends on the quality of information obtained by the environmental perception system,and needs to comprehensively recognize the perceived information,in order to make reasonable decisions to control the safe and stable driving of the vehicle.However,the environmental perception system has uncertainty,which may lead to incomplete or unreliable perceived information.At the same time,when using the perceived information for comprehensive cognition,the uncertainty of traffic participants’ driving behavior may be ignored,resulting in inappropriate decisions.In addition,in the dynamic traffic environment,the behavior decisions between vehicles and their surrounding traffic participants have interactive effects.Therefore,aiming at the uncertainty of perception,traffic participants’ driving behavior and the interactivity of traffic environment,this thesis studies the decisionmaking and control of the intelligent vehicle in dynamic traffic environment from four aspects: driving situation assessment,fault-tolerant perception,vehicle driving intention recognition and interactive behavior decision-making.The detailed works are summarized as follows:1.The situation assessment method based on the driving safety field(DSF)theory does not consider the uncertainty of dynamic traffic environment.To address this issue,a generalized driving situation assessment system considering uncertainty is proposed.In the situation assessment layer of the system,firstly,the driving space around the ego vehicle is divided into six areas.In view of the uncertainty of perception and the uncertainty of driving behavior of traffic participants,the driving safety field forces in different areas at the current time and the future prediction time are calculated,combined with the sensor fault-tolerant perception module and the vehicle driving intention recognition module.Secondly,considering the concept uncertainty of the situation assessment,based on the forward Gaussian cloud(FGC)model,the calculated field forces are converted into the degree of certainty under the three conceptual indicators of driving dangerous situation,desired situation and free situation.Then,the weight coefficient of the certainty degree of three sub-situation in each area when calculating the comprehensive situation is determined based on analytic hierarchy process(AHP),and then the comprehensive situation under different driving modes is calculated.The simulation results show that the generalized driving situation assessment system considering uncertainty can make an accurate and comprehensive assessment of the potential driving situation,which lays a good foundation for the subsequent behavior decision-making of intelligent vehicles.2.Aiming at the perception uncertainty during the process of signal acquisition and transmission of vehicle-borne millimeter wave radar sensor,an active fault-tolerant perception and control method of ACC system is proposed.Firstly,in the fault-tolerant perception module,the motion states of the target vehicles are estimated based on the fourth-order Sage Husa adaptive Kalman filter algorithm,and the sensor faults are identified based on the rules.Secondly,sensor faults are taken as discrete events,and the mixed logical dynamical(MLD)model of ACC upper control system is built.Then,the active fault tolerant control model of ACC system is proposed combining the MLD model with model predictive control(MPC)framework.Finally,simulation experiments show that the proposed method can realize the active fault-tolerant perception and control of intelligent vehicle ACC system without adding other redundant environment perception devices,and has important practical value.In addition,the proposed method can be easily transplanted to other ADAS,and the faulttolerant perception module is applied to the driving situation assessment system in Chapter 2.3.Aiming at the driving behavior uncertainty of traffic participants,a dynamic driving intention recognition method for surrounding vehicles with different driving styles based on adaptive multi-dimension continuous Gaussian mixture-HMM(AMCGM-HMM)is proposed without considering Internet of Vehicles(Io V)technology.Firstly,considering that driving styles of surrounding vehicles are different,and the extracted lane changing data samples are series of time series with different lengths and multiple dimensions,a feature-based multi-dimensional time series clustering algorithm is adopted to cluster the lane changing data samples and the results are used as the input of AMCGM-HMM.Parameters that may affect the recognition accuracy are optimized by multi-objective particle swarm optimization(MOPSO)algorithm to realize the adaptive of the method.Secondly,the proposed driving intention recognition method is trained and tested on NGSIM dataset.Finally,simulation experiments show that the proposed method can identify the driving intention of surrounding vehicles in advance and has high accuracy.Combined with the vehicle decision and control algorithm,it can ensure the safe operation of intelligent vehicles in complex traffic scenes.In addition,the proposed driving intention recognition method is applied to the driving situation assessment system in Chapter 2.4.Aiming at the interactive behavior decision-making problem between vehicles in dynamic traffic environment,an interactive driving behavior decision-making model based on non-cooperative incomplete information dynamic game is proposed.Based on the driving style,the sequential action sequence of the game is specified.Combined with the situation assessment results of the driving situation assessment system considering uncertainty proposed in this paper,the multi-objective decision-making income function of non-cooperative dynamic game is designed,and the sequential equilibrium of the game is solved.Secondly,in order to carry out closed-loop simulation experiments on the proposed decision-making model,the intelligent vehicle trajectory planning method under the Frenet coordinate system is used to plan the local trajectory of the vehicle,and then the planned trajectories are tracked and controled based on MPC algorithm.Finally,the simulation experiment results show that the interaction effects between vehicles and the influence of driving style on the game are fully considered by the driving behavior decision-making game,which is of great significance to improve the flexibility of intelligent vehicle decision-making.