Research on DCT shift strategy for various driving style based on “driver-vehicle-cloud” machine learning

Abstract

In order to solve the problem that the DCT static shift strategy cannot adapt to the difference in driving style, the driving style identification model based on multi-dimensional data mining and intelligent algorithm heavily depends on vehicle terminal data storage and calculation, an intelligent shift strategy based on “driver-vehicle-cloud” cooperative control is proposed. Firstly, the dynamic model of the DCT vehicle is analyzed, the primary shift schedule is calculated, and a method to adaptively modify the shifting schedule of DCT according to driving style is proposed. Then, many vehicle driving data are collected, cleaned, and reconstructed by wavelet denoising and other methods, and a driving style database with 80-dimensional features is constructed. Five essential features are selected by the ReliefF method, and the driving style recognition model is constructed by combining random forest, support vector machine, naive Bayesian, and other algorithms. Finally, the support vector machine model with the highest precision is selected, and the “driver-vehicle-cloud” collaborative control system is deployed using cloud computing and vehicle-cloud collaborative technology. The experiment car test shows that the system can identify the driver’s driving style in real time and realize the differential shift schedule and driving experience of DCT.