CN-116788261-B - Method for discriminating behavior habit of driver of self-adaptive commercial vehicle load

CN116788261BCN 116788261 BCN116788261 BCN 116788261BCN-116788261-B

Abstract

The invention relates to a method for discriminating behavior habits of drivers of self-adaptive commercial vehicles, which mainly comprises offline data acquisition, data preprocessing and feature definition, clustering features, storing clustering discrimination results of online running vehicles into a result matrix, converting the result matrix into a probability matrix, calculating information entropy and entropy weight of the behavior habits, and finally representing classification of the behavior habits of the drivers in different regions on a time domain.

Inventors

YAN FANGCHAO
TANG ZHICHENG

Assignees

天津布尔科技有限公司

Dates

Publication Date: 20260508
Application Date: 20230628

Claims (4)

1. The method is characterized by comprising a first module, a second module and a third module, wherein the first module performs offline classification model training through the commercial vehicle driving data stored by a cloud platform, the classification model training comprises the steps of collecting a sign signal, preprocessing data, calculating a characteristic value and clustering, and a cluster center is obtained after the clustering and comprises an aggressive driver, a general driver and a conservative driver; the second module records online vehicle data to be classified through a cloud platform, and the frequency of 1 time/min is based on characteristic calculation and driver behavior habit classification of the cluster center in the first module, so that a single judgment result is obtained, and a driver behavior habit category label corresponding to the result is recorded; The third module performs a scrolling calculation based on the category labels of the travel records in the second module, The rolling calculation specifically comprises the following steps: step S1, classifying and storing the behavior habit of the driver obtained by each data transmission into a class result matrix, wherein the expression is as follows: Wherein x 11 ,x 12 ,x 13 is a result vector obtained by the first data transmission, three element positions respectively represent aggressive type, general type and conservative type, the corresponding element position is assigned as 1 according to the calculation result, the other two values are 0, x n1 ,x n2 ,x n3 is a result vector of the nth time, and the latest returned data is represented; Step S2, elements in the result matrix are standardized according to the following formula to obtain a standardized matrix Z; Wherein x ij is a single element in the result matrix; step S3, calculating a probability matrix P based on the standardized result matrix Z; wherein z ij is a single element in the normalized result matrix; And S4, taking p ij as a probability and bringing the probability into an information entropy formula: E j is the information entropy of each label, and the larger the information entropy is, the smaller the information quantity contained in the label is, and the lower the credibility is; Step S5, obtaining entropy weight of each driver behavior habit classification result: wherein w 1 ,w 2 ,w 3 represents the entropy weight of the aggressive type, general type and conservative type driver classification result, and the classification with the largest entropy weight is selected as the result of the driver behavior habit discrimination to be output.
2. The method for judging the behavior habit of the driver of the self-adaptive commercial vehicle according to claim 1, wherein the characteristic signal acquisition content comprises a vehicle speed sensor signal, an accelerator position sensor signal, a steering wheel angle sensor signal and a brake pedal position sensor signal.
3. The method for judging the behavior habit of the driver of the self-adaptive commercial vehicle according to claim 2, wherein the vehicle speed characteristics commonly used in traditional researches are removed in the data processing, and the data processing takes acceleration change rate variance, acceleration maximum value, accelerator pedal change rate variance, brake pedal change rate variance and steering wheel angle change rate variance as characteristic values to participate in a clustering algorithm.
4. The method for judging the behavior habits of the drivers of the self-adaptive commercial vehicle load according to claim 3, wherein the method for recording the on-line vehicle data to be classified by the cloud platform comprises the steps of collecting sensor signals in real time based on a CAN bus of the vehicle through a vehicle-mounted terminal T-BOX, carrying out on-line calculation through a vehicle networking cloud platform, solving Euclidean distance with an off-line clustering algorithm cluster center, and finally transmitting the calculated driver behavior habit classification result to the vehicle according to time intervals.

Description

Method for discriminating behavior habit of driver of self-adaptive commercial vehicle load Technical Field The invention relates to the field of identification of behavior habits of drivers of commercial vehicles, in particular to a method for identifying the behavior habits of drivers of self-adaptive commercial vehicles. Background In a fleet or individual user of commercial vehicles, fuel cost is usually the highest priority cost management item, and under the background of cost pressure, energy conservation and emission reduction of the commercial vehicle are imperative as one of main tools of logistics industry. In the energy-saving field of commercial vehicles, besides optimizing on hardware such as a flow baffle plate, tires and the like to obtain lower running resistance, the purpose of saving oil can be achieved by controlling and improving the control of a driver, and the behavior habit of the driver is researched to effectively identify the action of the driver for controlling the vehicle such as stepping on an accelerator, braking and the like, so that the oil consumption and the emission of the whole vehicle are improved by optimizing control signals. In the existing related researches, the research content of the passenger car is usually carried on a characteristic selection and learning algorithm by a driver behavior identification model of the commercial car, but in objective fact, the difference of the running speeds among individual passenger cars under the same working conditions is not large, the parameters of the vehicle are relatively close to the actual conditions of the commercial car, and the problems of unsuitable characteristic selection and low algorithm matching degree are caused when the driver behavior identification research of the commercial car is transplanted to the commercial car. Disclosure of Invention The invention provides a method for identifying behavior characteristics of a commercial vehicle driver, which is suitable for all types and loads, and can avoid the situation that the behavior characteristics of the traditional driver deviate due to the difference of the commercial vehicle types and the difference of the loads, thereby influencing the judgment accuracy of the driving behavior. The aim of the invention can be achieved by adopting the following technical scheme: The method comprises the steps that the first module carries out offline classification model training through commercial vehicle driving data stored by a cloud platform, the classification model training comprises characteristic signal acquisition, data preprocessing, characteristic value calculation and clustering, a cluster center is obtained after the clustering, and the cluster center comprises an aggressive driver, a general driver and a conservative driver; The second module records online vehicle data to be classified through a cloud platform, and the frequency of 1 time/min is based on characteristic calculation and driver behavior habit classification of the cluster center in the first module, so that a single judgment result is obtained, and a driver behavior habit category label corresponding to the result is recorded; the third module performs a scrolling calculation based on the category labels of the travel records in the second module, The rolling calculation method specifically comprises the following steps: step S1, classifying and storing the behavior habit of the driver obtained by each data transmission into a class result matrix, wherein the expression is as follows: Wherein x 11,x12,x13 is a result vector obtained by the first data transmission, three element positions respectively represent aggressive type, general type and conservative type, the corresponding element position is assigned as 1 according to the calculation result, the other two values are 0, x n1,xn2,xn3 is a result vector of the nth time, and the latest returned data is represented; Step S2, elements in the result matrix are standardized according to the following formula to obtain a standardized matrix Z; Wherein x ij is a single element in the result matrix; step S3, calculating a probability matrix P based on the standardized result matrix Z; wherein z ij is a single element in the normalized result matrix; And S4, taking p ij as a probability and bringing the probability into an information entropy formula: E j is the information entropy of each label, and the larger the information entropy is, the smaller the information quantity contained in the label is, and the lower the credibility is; Step S5, obtaining entropy weight of each driver behavior habit classification result: wherein w 1,w2,w3 represents the entropy weight of the aggressive type, general type and conservative type driver classification result, and the classification with the largest entropy weight is selected as the result of the driver behavior habit discrimination to be output. Preferably, the acquisition content of the characteristic