Research On Intelligent Decision-Making,Trajectory Planning And Tracking Of Driverless Vehicle In High Speed Driving Conditions

With the increasing of vehicle ownership,energy crisis,traffic accidents,traffic jams,and parking problems continuously come up.Therefore,the development of efficient and environmentally friendly clean energy electric vehicles,and the realization of safe and reliable intelligent driving have attracted much attention.As a high-tech complex integrating environmental perception,trajectory planning,decision-making,control execution,and information interaction,intelligent driving vehicle has become a research hotspot.How to accurately extract effective information to make a safe and reliable control instruction is the technical basis for intelligent driving in the complex driving environment with multi-source and heterogeneous structure;how to plan a smooth and feasible trajectory in the dynamic environment in response to the various driving conditions is the key link to achieve intelligent driving.And the ultimate goal of intelligent driving is to establish a unified and efficient multi-actuator integrated coordinated controller to ensure the accuracy of path tracking control.In this paper,several key technologies of intelligent driving under the high speed driving conditions are studied.A complete driverless technical realization scheme is designed based on the intelligent driving decision-making,trajectory planning,and path tracking control.The main work of this paper are as follows:(1)Research on intelligent driving decision-making according to the characteristics of multi-source and heterogeneous information in complex driving environment.Firstly,a database of driving environment decision results is established based on the driving rules and vehicle driving states.Then a fuzzy classification rule is established to preprocess the information database.Finally,a decision-making maker of driving behavior is established in combination with the neural network.In order to further improve the decision-making accuracy of driving behavior,the gradient boosting decision tree(GBDT)algorithm is used in this study.The simulation results show that the decision accuracy of the GBDT algorithm in combination with fuzzy classification preprocessing of the driving environment information database can reach 99%,which indicates that the GBDT algorithm has great potential in the study of intelligent driving behavior decision.(2)Research on lane change trajectory planning in static driving environment.Firstly,the multi-objective function with weight coefficient is established based on the trapezoid lateral acceleration expression,and a fuzzy logic controller is used to complete the adaptive control of weight coefficient under different driving conditions,then the lane change trajectory adapted to different vehicle speeds is planned out.Secondly,in order to deal with the emergency collision avoidance condition when obstacles suddenly appear in front of the driving,another lane change trajectory planning method based on the fifth-order Bezier curve is proposed.By seeking the left and right boundaries of the feasible region,the physical problem is transformed into an optimization problem with respect to the coordinate points of Bezier curve expression.Then,the optimal collision avoidance trajectory curve satisfying the boundary conditions is obtained.In addition,the optimal trajectory is transformed into the steering wheel angle signal based on the preview following driver model,which can be output directly.The proposed method breaks the limitation of traditional planning method needs to know the position of the end point in advance,and realizes the dynamic follow under different vehicle speeds in the lane change process considering both driving comfort and efficiency.(3)Research on lane changing trajectory planning in dynamic driving environment.A dynamic lane changing trajectory planning method in highway scene is proposed based on the model predictive control theory.Firstly,the kinematic models of the ego-vehicle and surrounding vehicles are established,which are transformed into a state equation and then used as the prediction model.Then,the cost function and constraint conditions are constructed to complete the rolling optimization of the model predictive control.In order to facilitate the solution,the control vector is transformed from a two-dimensional coordinate system to a polar coordinate system,and the dispersion point of the optimized control vector is fitted by using quintic polynomials to obtain a smooth and feasible lane change trajectory.(4)Research on the tracking control method of the planned ideal trajectory.To track the planned trajectory,the vehicle state equation including the dynamic characteristics of the slip rate of each wheel is established based on the model predictive control theory.Then,the multi index function including tracking accuracy,actuator efficiency,and actuator performance is established.The optimal control sequence of steering wheel angle and driving torque signal of each wheel is obtained by optimal solution.An integrated coordinated control method with multiple actuators is proposed to realize the tracking of the ideal trajectory,and a speed follower is also established to maintain the target speed during lane change process.(5)Since it is difficult to realize the intelligent driving in real vehicle experiment at high speed,this paper firstly uses Prescan to build a driving scene that reflects the real driving situation,and combines the trajectory generator built in MATLAB/Simulink to simulate and analyze the proposed lane change trajectory planning method under the dynamic scene.The simulation results show that the proposed method can well respond to the dynamic state of surrounding vehicles,realizing dynamic trajectory planning.Furthermore,the high-precision vehicle model built in Car Sim instead of the real vehicle is combined with the coordinated controller built in MATLAB/Simulink to carry out the co-simulation analysis of the proposed integrated coordination control method.The results show that the proposed method not only improves the vehicle running stability but also significatly increases the trajectory tracking accuracy,which is beneficial to fully explore the potential of distributed driverless vehicle.