| With the development of the country,the progress of the society and the rise of the tide of intelligent technology,autonomous driving technology began to develop rapidly.ADAS and autonomous driving systems are increasingly used.However,most of the current L2,L2.5level ADAS and L3 level automatic driving systems do not take into account the usual driving habits and characteristics of human drivers in the curve sections,which reduces the acceptance and satisfaction of drivers and passengers on intelligent vehicles.This paper uses typical curve driving tests and data mining theories to study the characteristics of skilled drivers’ curve driving behavior,and establishes data samples.By studying the skilled drivers driving decision-making information sources of the bend,design curve dynamic trajectory planning and trajectory tracking algorithm,thus improve the ADAS and automatic driving system in the bend of planning control ability,optimize the environment of human-vehicle-environment loop performance of the system,enhance for intelligent vehicle ride comfort and satisfaction.The main research contents of this paper include the following four aspects:(1)Data acquisition and analysis of typical curves for skilled driversFirstly,a real vehicle driving data acquisition platform is built based on d SPACE realtime simulation platform,Carsim RT real-time vehicle dynamics simulation software,Senso Wheel torque steering wheel and its supporting sensors,and industrial computer.Secondly,32 skilled drivers were recruited through the Internet,and typical driving conditions with different characteristics were designed according to the research requirements,and the data of the drivers were collected.Then the collected data are quantified and analyzed,and the trajectory characteristics of skilled drivers under different steering conditions are summarized.Finally,the similarity analysis and cleaning of the obtained data are carried out based on dynamic time warping to ensure the validity of the samples.(2)Intelligent Vehicle Curve Dynamic Trajectory Planning StrategyFirstly,the mechanism of visual attention of human drivers when driving on curves was explored,and its mechanism was studied and a reasonable model was established.Then the cyclic neural network is used to simulate the driving process of the driver in the curve environment.The results show that the model has a good prediction effect.Finally,considering the internal weak interpretation of the deep neural network,the trajectory correction strategy is further designed,and replanning is carried out when necessary to ensure the safety of the vehicle running.Finally,the simulation results show the effectiveness of the warning and correction strategies.(3)Intelligent vehicle trajectory tracking strategyA tracking control method based on preview feedforward and robust feedback is designed.Firstly,the feedforward control is solved by previewing the optimal curvature model,and then the feedback gain is solved based on the theory of linear matrix inequality.Finally,the verification test of the tracking control method is carried out under two working conditions.The experimental results show that the preview feedforward and robust feedback tracking control method can improve the vehicle stability and reduce the path following error.(4)Verification and analysisFirstly,the dynamic humanoid programming method was integrated with the vehicle tracking control method,and the joint software simulation platform was built.Secondly,combined with the research objectives,the left turn and right turn condition tests are designed,and the similarity analysis between the test results and the real driver data is made.The results show that the human-imitated driving model based on RNN has strong generalization ability,can plan the trajectory in line with the driver’s actual driving characteristics and turning mode under the curve section,and can control the vehicle to pass the curve smoothly,and has good applicability. |
PDF Full Text Request