Study On Intelligent Control Of DCT Vehicle Considering Driving Behavior And Driving Environment

Dual clutch transmission(DCT)has the advantages of high transmission efficiency,fast shift response and good inheritance with manual transmission.Since it was listed globally in 2002,its market share has developed rapidly.Due to its technical characteristics,DCT transmission is particularly suitable for the current domestic technical basis and manufacturing conditions.Domestic independent brand automobile enterprises have successively developed DCT products and achieved the listing of complete vehicles.However,there are still many problems in the practical application of DCT.Due to the low intelligence of the gear-shift decision-making system and its poor adaptability to the driver’s driving behavior and driving environment,DCT vehicles are prone to frequent gear changes,which reduces the ride comfort and fuel economy,and easily leads to excessive clutch temperature rise.Because it is difficult to plan the clutch engagement trajectory that balances the engagement time,jerk and sliding friction work according to the driver’s driving intention,the start / shift response of DCT vehicles is generally slow.In addition,due to the characteristics of uncertain parameters and high nonlinearity of DCT system,it is difficult to establish an accurate dynamic model,resulting in low control accuracy of the model-based control method,resulting in problems such as large shake and jerk of the transmission system in the starting / shifting process,which reduces the starting / shifting quality of DCT vehicles and restricts the development of domestic DCT industry.Based on the key project "DCT intelligent control and evaluation method considering service performance,driving behavior and driving environment" of the National Natural Science Foundation China automotive industry innovation and development joint fund,this paper takes wet DCT as the research object,and carries out research work from four aspects: driver’s driving behavior modeling and recognition,real-time estimation of road slope,intelligent decision-making of vehicle gear-shift considering driving behavior,ramp and traffic environment and DCT vehicle start / shift intelligent control based on data-driven,which provides theoretical basis and technical support for improving the intelligent level of DCT vehicle.The main contents of this paper are as follows:(1)In order to accurately obtain the driver’s driving behavior information,the naturalistic driving data acquisition experiment and data preprocessing are carried out.On this basis,a driving style and driving intention modeling and recognition method based on machine learning / deep learning is proposed.The driving data of each driver is modeled as a five-dimensional Gaussian mixture model.Markov Monte Carlo sampling and Kullback Leibler divergence algorithm are used to measure the difference between Gaussian mixture models of different drivers,and the accurate classification of driver driving style is realized.Simultaneously,a driving intention progressive classification method based on K-means algorithm is proposed,and an accurate driving intention classification data set is obtained.On this basis,a driving intention recognition method based on long short-term memory network is proposed,and a long short-term memory network model which can be used to accurately recognize driving intention is constructed and trained.(2)In order to obtain accurate road slope information and improve the adaptability of DCT vehicles to the ramp environment,a road slope estimation method based on multiple model and multiple data fusion is proposed.This method establishes two kinds of slope and two kinds of slope change rate estimation models based on kinematics and dynamics respectively.The interacting multiple model algorithm is used to estimate the slope and slope change rate hierarchically,and a two-stage interacting multiple model slope estimation algorithm is constructed to improve the estimation accuracy.On this basis,for the two slope data estimated based on kinematics and dynamics,the two local estimates are fused at the decision level through filter tracking gate,probabilistic data association algorithm,optimal convex combination fusion algorithm and evidence theory,so as to reduce the impact of vehicle states such as braking,gear shifting,rapid acceleration and vehicle stationary on slope estimation and improve the overall accuracy of slope estimation.(3)In order to improve the adaptability of the gear-shift decision-making system to the driver’s driving behavior,ramp and traffic environment while maintaining low fuel consumption,dynamic programming and multi-objective genetic algorithm are used to solve the optimal gear decision points under different driving styles,driving intentions,ramps and traffic environment based on naturalistic driving data.On this basis,a datadriven gear decision model is established by using random forest method.In addition,a gear-shift decision strategy update method based on over the air technology is proposed.Gaussian mixture model and Kullback Leibler divergence algorithm are used to continuously update the gear-shift decision database,so that the gear decision strategy can continuously adapt to the changes of driving behavior and driving environment,and significantly improve the self-learning ability of gear decision strategy.(4)In order to reduce the shake and jerk in the process of vehicle starting / shifting and adapt to the change of driver’s driving intention,taking the jerk,starting time and sliding friction work as optimization indexes,the adaptive pseudo spectral method is used to accurately solve the optimal target engagement trajectory of the clutch in the process of starting / shifting.On this basis,a real-time planning method of clutch target trajectory in starting / shifting process based on deep gated recurrent unit network and gradient boosting decision tree model is proposed.In addition,based on the naturalistic driving data,the data-driven model of DCT system is constructed through autoregressive exogenous model and partial least squares identification,which realizes the data-driven predictive control of starting / shifting process.Based on the input and output data,the model parameters are updated online by Kalman filter algorithm,which adapts to the time-varying and nonlinear characteristics of the system parameters in real time,and significantly improves the starting / shifting quality of DCT vehicles.(5)Based on the real vehicle control platform,the experimental research on road slope estimation method,intelligent gear-shift decision method and start / shift intelligent control method is carried out,which verifies the effectiveness of the method proposed in this paper from the overall law,and verifies the practicality of the control method proposed in this paper in real-time vehicle control.