| With the vigorous development of automotive electronic technology,automotive intelligence is gradually becoming a new direction of development.Intelligent vehicle is the carrier of human,vehicle,infrastructure and Internet interconnection.It can grasp the surrounding traffic conditions more fully and provide rich and efficient information to drivers and decision-making modules in time.The intelligent algorithm is used for decision-making control to further improve driving safety.So it can also reduce the difficulty of driving,improve the operational efficiency of the overall traffic system,and alleviate the current situation of urban road traffic congestion.Among them,the lateral tracking control of intelligent vehicle is an important core technology to realize the function of intelligent vehicle.At present,domestic and foreign scholars have carried out a lot of research work on the lateral tracking control of intelligent vehicles,but there are still many shortcomings.Aiming at the problems of low control accuracy and poor robustness of traditional methods in actual driving conditions,this paper studies the control method,control mode switching strategy and vehicle experiment.The main contents include:(1)A transverse tracking controller with low speed and small curvature is designed and its simulation function is verified.Under the condition of low speed and small target curvature,a two-degree-of-freedom dynamic model is established for the lateral control characteristics of path tracking under this condition,and the main parameters affecting the lateral control are analyzed.Combining with the characteristic that BP neural network can fit any non-linear function,the state of the system can be predicted by using BP neural network.Then the error caused by the training of neural network is corrected by using PID algorithm to improve the control accuracy,ensure real-time performance and improve the accuracy of lateral control.Vehicle data under the control of excellent drivers are acquired by driver's in-the-loop simulation platform based on Prescan/Simulink to train CNNPID controller,and the lateral control effect of path tracking is simulated.The tracking error of CNNPID is controlled within 0.2m,which is better than that of PID controller.(2)A transverse tracking controller with high speed and large curvature is designed and its simulation function is verified.Firstly,a three-degree-of-freedom dynamic model is established under the condition of high speed and large target curvature.Then,the lateral controller is designed according to the advantage that the model predictive controller can deal with multiple constraints in the control process.Finally,RBF neural network is used to compensate the control errors caused by simplified vehicle dynamics model in the non-linear region(such as complex curvature variation conditions)to ensure the accuracy of the control.Through Carsim/Simulink joint simulation platform,the control effect is verified under complex curvature conditions including straight line section,serpentine section and double-shift section.The control error is guaranteed in the range of 0.285 m,which proves that RBFMPC has high control accuracy.(3)Based on the analysis of the coupling relationship between vehicle speed and target curvature,the optimal control method under current working conditions is selected by using support vector machine(SVM)with control accuracy and calculation amount as criteria.The evaluation method of control mode switching strategy is proposed from three aspects: accuracy,stability and comfort.(4)Considering the complexity of the real road environment,the performance of the controller needs to be further verified.In this paper,two lateral control methods designed in this paper are experimented with the help of Harvard H8 intelligent vehicle platform.The applicability of the controller and the control handover strategy designed in this paper in real vehicle environment is verified.The actual tracking error is guaranteed in the range of 0.3m. |
PDF Full Text Request