Automobile dynamic property and economy expectation quantification method

Abstract

The invention discloses an automobile dynamic property and economy expectation quantification method, which belongs to the technical field of driving intention identification. The method is divided into two stages of offline modeling and real-time quantification. In the off-line modeling stage, the automobile road test speed, the accelerator opening degree and the dynamic property expectation subjective evaluation of different drivers are collected and processed; a fuzzy neural network is adopted to establish a driver dynamic property expectation quantification model; and a driving manipulation characteristic parameter and vehicle operation state parameter prediction model are established based on the exponential smoothing-Markov model. In the real-time quantification stage, the vehicle speed and the accelerator opening degree are collected in real time; parameter prediction is carried out through a driving manipulation characteristic parameter and vehicle running state parameter prediction model established offline; the predicted value is input into a driver dynamic property expectation quantification model to obtain a dynamic property expectation value; and the expected economicvalue is further calculated. The method can continuously quantify the expectation of the driver for the dynamic property and economy of the automobile, and has the characteristic of high quantizationprecision.

Description

technical field [0001] The invention relates to an automobile driving intention recognition technology, specifically a method for quantifying the driver's power and economic expectations, and selecting different shift schedules based on the quantification results. Background technique [0002] In "A Driving Intention Recognition Method Based on Improved HMM and SVM Double-layer Algorithm" (CN106971194A), the invention uses the improved HMM and SVM double-layer algorithm for off-line training, and the driver's sudden left lane change and normal left change It recognizes driving intentions such as lane keeping, lane keeping, normal right lane change and sudden right lane change. [0003] In "A Driver's Intention Recognition Method" (CN103318181A), the invention uses a multi-dimensional discrete hidden Markov model to propose a two-layer recognition structure to recognize the driver's acceleration / braking behavior and steering / lane-changing behavior. [0004] In "A Shift Corre...

AI Technical Summary

Problems solved by technology

[0007] Existing recognition technologies are mostly used to classify driving behaviors or intentions, while a small number of technologies that quantify power and / or economic expectations generally use fuzzy reasoning methods. However, due to the inability to optimize the membership function, the accuracy of the quantified results needs improvement

Images