A method for analyzing the correlation between driver's driving behavior

Abstract

The present invention discloses a method for analyzing the correlation of a driver's driving behavior, including: S1: collecting driver travel data; S2: dividing the travel route and parsing the signal source code in S1; S3: selecting and analyzing driving behavior evaluation indicators; S4: sequentially labeling the driving behavior evaluation indicator parameters in S3; S5: constructing an FP-tree model according to the travel indicator data set in S4; S6: performing ecological driving behavior correlation analysis according to the FP-tree model result constructed in S5; S7: driving behavior judgment. Combined with urban driving data and positioning information, the travel route is divided to reflect the influence of different road types and traffic conditions on driving behavior, accurately analyze the driving behavior evaluation indicators, propose a coding and labeling method for driving behavior evaluation indicators and convert it into a digital set, which is convenient for big data sample analysis and calculation, construct FP-tree for driving behavior correlation analysis, combine big data efficiently and quickly, divide the driving behavior evaluation indicator interval, and provide a basis for driving behavior judgment.

Description

Technical Field

[0001] The invention relates to a method for analyzing the correlation of a driver's driving behavior, and belongs to the technical field of automobile energy conservation and emission reduction. Background Art

[0002] Vehicle energy consumption is the result of the combined effects of people, vehicles, roads and weather. To reduce vehicle energy consumption, we can start from many aspects. However, due to the limitations of traditional fuel-saving technologies that have almost reached the limit of fuel-saving potential and the difficulty in improving urban congestion in the short term, reducing vehicle energy consumption from the driver's perspective has become a more feasible method. How to evaluate and improve the driver's driving behavior to further improve the performance and range of pure electric vehicles has become a focus of current exploration. In the prior art, traditional on-board data recorders are usually used to collect driving behavior data, such as vehicle speed, acceleration, etc. Then, based on the sample data set, characteristic factors are extracted, and the correlation between complex item sets is mined and searched through the association method. In the research direction of heavy-duty engineering vehicle driving behavior, Yu Linghua et al. extracted factors such as sleep patterns, road conditions, load status and time characteristics from the data set, and explored the influencing factors of heavy-duty engineering vehicle drivers' driving performance with the help of association rule mining (Apriori) method, and put forward suggestions for driver management and road safety assurance; in the research direction of vehicle lane changing, Long Yan et al. used the Apriori method to describe the characteristics of perception-operation and their correlation in the lane changing process, providing support for vehicle lane changing operations; when analyzing bad driving behavior, Liang Chenlei et al. used the Apriori algorithm based on OBD data to analyze the intrinsic correlation and trip probability of bad driving behavior, providing theoretical expansion for personalized driving and driving style research. In addition, scholar L et al. proposed a method of multi-target search for frequent item sets for multi-target.

[0003] However, traditional vehicle data cannot provide detailed information about driver behavior and habits, and cannot locate vehicle trajectories in real time to reflect the impact of the external environment on driving behavior. The processing methods for data sets are very different. In the case of big data, the indicator set is redundant, and the Apriori method generates a large number of candidate sets. Each calculation requires rescanning the data set to calculate the support of each item set. Summary of the invention

[0004] Purpose of the invention: In view of the deficiencies in the prior art, the present invention provides a method for analyzing the correlation between driver driving behaviors. The present invention combines the advantages of big data, studies the intrinsic correlation between driver driving behaviors from the perspective of individual characteristics, and improves computational efficiency by constructing a frequent pattern tree (FP-tree) using the frequent pattern growth (FP-Growth) method, thereby solving the problems in the prior art.

[0005] Technical solution: A method for analyzing the correlation between a driver's driving behavior, comprising the following steps:

[0006] S1: Collecting driver travel data: Collecting data through the collection device on the test vehicle, recording the driver's complete trip as a travel event, that is, the starting point and the end point overlap to form a closed-loop route, collecting a total of m travel events of the same driver, the same model of electric car, and the same route, collecting the signal source code obtained by the collection device in each travel event, the signal includes: time, vehicle speed, acceleration pedal travel percentage, brake pedal travel percentage, steering wheel steering, steering wheel angle, and steering wheel angle velocity, while locating the driving position of the test vehicle through the GPS inertial navigation system, collecting longitude and latitude parameters and time parameters;

[0007] S2: Divide the travel distance and analyze the signal source code in S1, use sliding window mean filtering to filter the vehicle driving data, and record every 1 minute as a time period;

[0008] S3: Selection and analysis of driving behavior evaluation indicators: The average speed of the test vehicle during driving, pedal travel percentage, vehicle speed fluctuation difference, acceleration and deceleration change value, speed change frequency and parking time are selected as driving behavior evaluation indicator parameters and the average vehicle speed threshold is set. Speed ​​fluctuation difference threshold E w , Acceleration / deceleration fluctuation difference threshold E a , speed change frequency threshold E f , Pedal travel percentage threshold E p and parking time threshold E t As a threshold value for driving behavior evaluation indicators;

[0009] S4: sequentially label the driving behavior evaluation index parameters in S3, and convert the driving behavior evaluation index parameters exceeding the threshold into a digital set, and define the trip index data set through the digital set;

[0010] S5: FP-tree model construction based on the trip indicator dataset in S4;

[0011] S6: Conduct correlation analysis on eco-driving behavior indicators based on the operating results obtained by the FP-tree model constructed in S5;

[0012] S7: Driving behavior judgment: judging the driver's driving behavior by setting judgment criteria based on support and confidence.

[0013] Preferably, the specific steps of S2 are as follows:

[0014] The signal source code obtained by the acquisition device is analyzed offline, and the speed and acceleration are processed by sliding window mean filtering. The filtered travel data is sorted according to each trip, and the maximum value of the time column is obtained to determine the upper limit of the time period. The time period is created and the number of times is counted, with 1 minute as a time period. The traffic light intersection entering the journey boundary section is selected as the dividing point, and the corresponding longitude and latitude at this moment are found according to the data collected by the GPS inertial navigation system equipment, and the corresponding time from the departure at this moment is obtained, and the departure time obtained by the GPS inertial navigation system is synchronized to the acquisition device.

[0015] Preferably, the specific steps of S3 are as follows:

[0016] Calculate the average speed threshold Analyze the average speed data collected from m trips, analyze the average speed range of electric vehicles, find the concentrated range, and calculate the average speeding threshold of pure electric vehicles according to the following formula

[0017]

[0018] Where: E M is the 75% quantile of the total distribution of all samples of average vehicle speed; E N is the quartile difference, that is, the difference between the 75% quantile and the 25% quantile of the total distribution of all samples of average vehicle speed; is the average speed exceeding threshold of pure electric vehicles;

[0019] Calculate the speed fluctuation threshold E of pure electric vehicles w :If the speed fluctuation difference exceeds the speed fluctuation difference threshold E w The trip is considered as non-ecological driving, and the speed fluctuation threshold E w is the difference between the maximum speed and the average speed during driving, that is

[0020]

[0021] Where: E w is the speed fluctuation value of the tram (unit: km / h); E Vmax is the maximum speed of the trip (unit: km / h); is the speeding threshold of the average trip speed (unit: km / h);

[0022] Obtain the acceleration change threshold E a :Calculate the range of acceleration change values ​​of group m:

[0023] df["acceleration"] = df.iloc[:,1].diff() / df.iloc[:,0].diff()

[0024] accel_min = df["acceleration"].min()

[0025] accel_max = df["acceleration"].max()

[0026] Then the acceleration changes from accel_min to accel_max m / s 2 , according to the acceleration change and test conditions, select the change threshold E a ; If the acceleration is greater than the maximum acceleration or the deceleration is less than the maximum deceleration, it is considered to be over the threshold;

[0027] Calculate the speed change frequency threshold E f :According to the travel divided by S1, set the acceleration change value caused by speed change during normal driving in am / s 2 Inside, calculate E f :

[0028] E f =Q / T

[0029] Where: E f is the speed change frequency; Q is the number of speed changes; T is the unit time (unit: min); 1 time / minute is set as the speed change frequency threshold, and a speed change behavior greater than the speed change frequency threshold is considered frequent speed change behavior, which is non-ecological driving;

[0030] Calculate the pedal travel change threshold E p :Calculate the pedal stroke percentage change value range of group m, the pedal stroke percentage change is pedal_min~pedal_max%;

[0031] Calculate the parking time threshold E t : Detect and record the duration of each time when the speed is 0km / h in the vehicle speed data, and calculate the average parking threshold E of the electric vehicle t :

[0032] E t =E t1 +1.5E t2

[0033] Where: E t1 is the 75% quantile of the total distribution of all samples of average vehicle speed; E t2 It is the interquartile difference, that is, the difference between the 75% quantile and the 25% quantile of the total distribution of all samples of average vehicle speed.

[0034] Preferably, the specific steps of S4 are as follows:

[0035] A threshold is judged for the driving behavior evaluation index parameters in each trip. If the threshold is exceeded, it is recorded as an event, which is represented by a digital set {1}, {2}, {3}, {4}, {5}, {6}. All events occurring in the trip are defined as the trip index data set {1, 2, 3...}.

[0036] Preferably, the specific steps of S5 are as follows:

[0037] Support is p, and the calculation formula is:

[0038]

[0039] Among them, α is the index item set of the predecessor of the associated item, β is the index item set of the successor of the associated item, and P is the probability;

[0040] Confidence is q, and the calculation formula is:

[0041]

[0042] S501: Encode the driving behavior evaluation index parameter database E, i.e., the travel index data set, and set the minimum support p min and the minimum confidence q min ;

[0043] S502: Select frequent itemsets F from database E;

[0044] S503: Verify whether the frequent item set F satisfies the support p; if the frequent item set F does not satisfy the support p, delete the data; if the frequent item set satisfies the support p, create an item header table, construct an FP-Tree, and construct a conditional pattern base, that is, a set of itemsets remaining after removing an item from a frequent item set;

[0045] S504: Perform recursive iteration until the tree contains an element item, and construct a conditional FP tree;

[0046] S505: Search for frequent items in the conditional FP tree, obtain the frequent item set L and calculate its support p;

[0047] S506: Verify whether the frequent itemset L satisfies the minimum support p min and the minimum confidence q min ; If the frequent itemset L does not satisfy the minimum support p min and the minimum confidence q min , then delete this frequent item set; if the frequent item set L satisfies the minimum support p min and the minimum confidence q min , then the ecological driving behavior association method is generated;

[0048] S507: Output FP-tree operation time T1.

[0049] Preferably, the specific steps of S6 are as follows:

[0050] The associated items formed by the obtained driving behavior evaluation index and the subsequent evaluation index are arranged from large to small according to the support percentage. The maximum support between the associated items indicates that the correlation between the driving behavior evaluation indicators of the item is the strongest, and the maximum confidence between the associated items indicates that the probability that the preceding driving behavior evaluation index of the item has the highest confidence will affect the subsequent driving behavior evaluation index.

[0051] Preferably, the specific steps of S7 are as follows:

[0052] The support judgment interval is divided into three levels: strong (50%~100%), medium (37.5%~50%), and weak (12.5%~37.5%). The confidence judgment interval is divided into three levels: high (95%~100%), medium (80%~95%), and low (50%~80%). In the analysis of driving behavior, we mainly focus on the related items of driving behavior evaluation indicators with strong support and high confidence. Combined with the actual collected data results, the related items with support lower than 12.5% ​​and confidence lower than 50% are excluded.

[0053] Beneficial effects: The present invention can combine urban driving data and positioning information to more accurately divide the itinerary, reflect the impact of different road types and different traffic conditions, i.e., urban congested sections and suburban unobstructed sections, on driving behavior, and make the analysis of driving behavior evaluation indicators more accurate. A coding and numbering method for driving behavior evaluation indicators is proposed, which is converted into a digital set to facilitate the analysis and calculation of big data samples. The FP-Growth method is used to construct an FP-tree for driving behavior correlation analysis, which is more efficient and faster than the existing Apriori method in combination with big data. At the same time, the driving behavior evaluation index interval is divided to provide a basis for driving behavior judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0055] Figure 1 A flow chart of a method for analyzing the correlation between a driver's driving behavior provided by the present invention;

[0056] Figure 2 This is an example diagram of the analysis of the proportion of speeding during travel;

[0057] Figure 3 This is an example diagram of the average speed of the trip;

[0058] Figure 4 This is an example diagram of acceleration fluctuation difference;

[0059] Figure 5 This is an example diagram of the threshold ratio of some driving behavior evaluation indicators;

[0060] Figure 6 This is a flow chart of the driving behavior correlation index modeling using FP-tree provided by the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] In the description of the present invention, it is necessary to understand that the terms "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0063] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0064] like Figure 1As shown, S1: collecting driver travel data: collecting data through the collection device on the test vehicle, recording the driver's complete trip as a travel event, that is, the starting point and the end point overlap to form a closed-loop route, collecting a total of m travel events of the same driver, the same model of electric car, and the same route, collecting the signal source code obtained by the collection device in each travel event, the signal includes: time, vehicle speed, acceleration pedal travel percentage, brake pedal travel percentage, steering wheel steering, steering wheel angle, and steering wheel angle velocity, while the driving position of the test vehicle is located through the GPS inertial navigation system, collecting longitude and latitude parameters and time parameters;

[0065] S2: Divide the travel route into trips and analyze the signal source code obtained by the acquisition device offline. Use sliding window mean filtering to filter the analyzed vehicle form data, and record every 1 minute as a time period;

[0066] The trip data obtained by data filtering preprocessing S1 is used to filter abnormal data, and the sliding window size is defined. The data collection in this embodiment is set to 5s. The pandas library in the Python programming software is used to read Excel data, and the speed and acceleration data are subjected to sliding window mean filtering. The processed trip data is arranged according to the travel date, and the maximum value of the time column is obtained to determine the upper limit of the time period. The time period is created and the number of times is counted, and 1 minute is recorded as a time period. The traffic light intersection entering the trip boundary section is selected as the dividing point, and the corresponding longitude and latitude at this moment are found according to the data collected by the GPS inertial navigation system equipment, and the corresponding time from the departure at this moment is obtained. The departure time obtained by the GPS inertial navigation system is synchronized to the collection device.

[0067] In this embodiment, considering the practical significance of driving behavior research for the ecological driving of pure electric vehicles, taking the pure electric vehicle driver in the city driving condition as an example, the sampling time interval of the high-frequency data, i.e., the vehicle speed, acceleration pedal travel percentage, and brake pedal travel percentage in this embodiment is 0.01s, and the sampling time interval of the steering wheel angle, steering wheel steering, and steering wheel angular velocity is 0.02s, is collected and processed. Simulate the city commuting, select any location in the suburbs as the starting point of the trip, select the travel route, including urban roads passing through downtown areas, hospitals and schools, the data format obtained after the signal source code collected by the collection device is parsed as shown in Table 1, and the data format collected by the GPS inertial navigation system is shown in Table 2:

[0068] Table 1 Collection device signals

[0069]

[0070]

[0071] Table 2 GPS inertial navigation signal

[0072] GNSS Week time latitude longitude high 2270 128387.090 32.196170898 119.509352203 11.0307 2270 128387.100 32.196171094 119.509350944 11.0299 2270 128387.110 32.196171290 119.509349685 11.0292 2270 128387.120 32.196171486 119.509348426 11.0285 2270 128387.130 32.196171682 119.509347168 11.0278

[0073] Since the GPS inertial navigation system and the corresponding time of the acquisition device do not match, the traffic light intersection of the junction section, that is, the traffic light intersection where the latitude and longitude values ​​have a trend of increasing first and then decreasing or decreasing first and then increasing, is selected as the demarcation point for entering the trip. The corresponding longitude and latitude at this moment are found according to the data collected by the GPS inertial navigation system equipment, and the corresponding time from departure at this moment is obtained. The departure time obtained by the inertial navigation is synchronized to the acquisition device, and the raw data obtained by the acquisition device is pre-processed. Among them, the speed and accelerator pedal sampling intervals are 0.01s, and the steering wheel angle and steering sampling intervals are 0.02s. The reaction time of a normal person is 0.15-0.4s, and the fastest reaction time is also greater than 0.1s. Therefore, this embodiment adopts 0.1s as the sampling interval, and uses MatLab to process the parsed data to obtain the trip data of the same driver and the same tram m times with a sampling interval of 0.1s.

[0074] S3: Selection and analysis of driving behavior evaluation indicators: The average speed of the test vehicle during driving, pedal travel percentage, vehicle speed fluctuation difference, acceleration and deceleration change value, speed change frequency and idling time are selected as driving behavior evaluation indicator parameters and the average vehicle speed threshold is set. Speed ​​fluctuation difference threshold E w , Acceleration / deceleration fluctuation difference threshold E a , speed change frequency threshold E f , Pedal travel percentage threshold E p and idling time threshold E t As a threshold value for driving behavior evaluation indicators;

[0075] In this embodiment, the speed limit on urban roads is 50km / h. Based on the actual road conditions in the comprehensive experiment, the speeding threshold for electric vehicles in urban areas is set at 50km / h, and the speeding threshold for suburban roads is set at 80km / h. The speeding time window is set to 4s. The speeding percentage example is as follows: Figure 2 As shown, the horizontal axis is the speeding percentage, that is, the range of 10%, 20%, 30%, 40%, and 50% exceeding 50, and the vertical axis is the ratio of the speeding times corresponding to each interval to the total speeding times. It can be seen that the vast majority of speeding in urban areas occurs within 40%, and is mainly concentrated in the range of 0% to 10%.

[0076] Calculate the average speed threshold The average speed data of m trips is analyzed. Taking 60 trips in one month as an example, the average speed of the trip is as follows: Figure 3 As shown. Figure 3The pie chart is the average speed distribution interval, which is divided into average speed intervals of 5km / h. The average speed between 15-20km / h accounts for 20%, and the average speed between 20-25km / h accounts for as high as 56.67%. It can be seen that the concentrated intervals are mainly concentrated between 15-20km / h and 20-25km / h. Compared with the average speed of 29km / h in urban road conditions in the CATC standard, the average speeding threshold of pure electric vehicles is calculated according to the following formula

[0077]

[0078] Where: E M is the 75% quantile of the total distribution of all samples of average vehicle speed; E N is the quartile difference, that is, the difference between the 75% quantile and the 25% quantile of the total distribution of all samples of average vehicle speed; is the average speed exceeding threshold of pure electric vehicles; in this example, E M =24.36km / h, E N =5.88km / h, then

[0079] Calculate the speed fluctuation threshold E of pure electric vehicles w :According to the maximum speed limit standard of urban roads, the speed fluctuation difference threshold is set to 35km / h. If it exceeds this threshold, it is considered as non-ecological driving. Speed ​​fluctuation threshold E w is the difference between the maximum speed and the average speed during driving, that is

[0080]

[0081] Where: E w is the speed fluctuation value of the tram (unit: km / h); E Vmax is the maximum speed of the trip (unit: km / h); is the speeding threshold of the average trip speed (unit: km / h);

[0082] Obtain the acceleration change threshold E a : Calculate the range of acceleration changes for the m groups of travels, and convert the speed (km / h) in the second column of the data table into speed (m / s):

[0083] df["acceleration"] = df.iloc[:,1].diff() / df.iloc[:,0].diff()

[0084] accel_min = df["acceleration"].min()

[0085] accel_max = df["acceleration"].max()

[0086] Then the acceleration changes from accel_min to accel_max m / s 2 , taking the acceleration fluctuation difference of 60 groups of trips in one month as an example, Figure 4 As shown, the horizontal axis is each stroke, and the vertical axis is the acceleration and deceleration range of the stroke. It can be seen that the acceleration is mainly concentrated at 2m / s 2 Around, the deceleration is mainly concentrated at 2.75m / s 2 According to the acceleration change and test conditions, select the change threshold E a This embodiment is measured under full load of air conditioner in summer, so according to the full load condition of air conditioner in summer, the maximum acceleration is selected as 2.28m / s 2 , maximum deceleration -2.73m / s 2 , if the acceleration is greater than the maximum acceleration or the deceleration is less than the maximum deceleration, it is considered to be an over-threshold. According to the actual road conditions, affected by the urban environment, congested sections are all caused by traffic light intersections. The speed of congested vehicles in urban areas is set to be below 20km / . It is defined that the electric vehicle decelerates from 20km / h to 0km / h as entering a congested section, and the electric vehicle accelerates from 0km / h to 20km / h as leaving a congested section. The number of times the speed is 0km / h in the original data is detected and recorded, and the speed range sections from 0 to 20km / h are extracted from the data to obtain the acceleration of congested sections and unobstructed sections, so as to reflect the impact of traffic flow and signal lights on the driving behavior of electric vehicle drivers.

[0087] Calculate the speed change frequency threshold E f :According to the travel divided by S1, set the acceleration change value caused by speed change during normal driving in am / s 2 Inside, calculate E f :

[0088] E f =Q / T

[0089] Where: E f is the speed change frequency; Q is the number of speed changes; T is the unit time (unit: min); 1 time / minute is set as the speed change frequency threshold, and a speed change behavior greater than the speed change frequency threshold is considered frequent speed change behavior, which is non-ecological driving;

[0090] Calculate the pedal travel change threshold E p :Calculate the pedal stroke percentage change value range of group m, the pedal stroke percentage change is pedal_min~pedal_max%;

[0091] Calculate the idle time threshold E t : Detect and record the duration of the moment when the speed is 0km / h in each data, and calculate the average idle threshold E of the electric vehiclet :

[0092] E t =E t1 +1.5E t2

[0093] Where: E t1 is the 75% quantile of the total distribution of all samples of average vehicle speed; E t2 It is the interquartile difference, that is, the difference between the 75% quantile and the 25% quantile of the total distribution of all samples of average vehicle speed.

[0094] Analyze the proportion of driving behavior evaluation indicators exceeding the threshold. Take 100 groups of trips as an example, select the probability of speeding, average speed, speed fluctuation difference, acceleration fluctuation difference, and speed change frequency exceeding the threshold as follows Figure 5 As shown in the figure, the probability of the average speed and acceleration being normal is high, while the probability of the acceleration exceeding the threshold is high. This is because the traffic volume in the urban area is large and the driving speed is slow; there are many traffic lights in the urban area, which will frequently cause stop-and-go phenomena. Drivers are easily affected by the sudden situation of non-motor vehicles and will frequently change lanes at intersections, resulting in a high probability of acceleration exceeding the threshold.

[0095] S4: The driving behavior evaluation index parameters in S3 are numbered in sequence, and the driving behavior evaluation index parameters exceeding the threshold are converted into digital sets, and the trip index data set is defined by the digital set: the driving behavior evaluation index parameters in each trip are judged by the threshold. If the threshold is exceeded, it is recorded as an event, which is represented by the digital set {1}, {2}, {3}, {4}, {5}, {6}, and all events occurring in the trip are defined as the trip index data set {1, 2, 3...}.

[0096] The six evaluation indicators of pure electric vehicle driving behavior are numbered in sequence: average speed, speed fluctuation difference, acceleration fluctuation difference, pedal travel, idling time, and speed change frequency. The driving behavior indicator values ​​of all travels are converted into corresponding digital sets, as shown in Table 3:

[0097] Table 3 Digital set label table

[0098]

[0099] If a driving behavior evaluation index exceeds the threshold of this driving behavior evaluation index, it is recorded as an event, which is represented by a digital set: For example, when the pedal travel percentage exceeds the pedal travel threshold, it is recorded as event 5, which is represented by the digital set {5}; (greater than ), vehicle speed fluctuation difference E w1 (greater than E w ), acceleration / deceleration change value E a1 (greater than Ea ), speed change frequency E f1 (greater than E f ), then the trip indicator data set is {1,2,3,4}.

[0100] S5: Build the FP-tree model based on the trip indicator dataset in S4:

[0101] Each trip is called a transaction, and m trips contain m transactions. Each driving behavior evaluation index of a trip is called an item, and an item is a basic analysis object, such as the average speed item {1}. A set containing zero or more driving behavior index items is called a driving behavior index item set, such as H = {1, 2, 3, 4, 5, 6} is a set of driving behavior index items for a trip. The index data set in this embodiment is a set containing 6 items, which is called a 6-item set.

[0102] The driving behavior evaluation indexes are to find out the relationship between each item, that is, each driving behavior evaluation index, from the driving behavior evaluation index item set.

[0103] The correlation of driving behavior evaluation indicators can be expressed as satisfy: and α is the index item set of the preceding items of the associated items, and β is the index item set of the following items of the associated items, that is, a large change in acceleration will lead to an increase in the frequency of speed changes.

[0104] Correlation of driving behavior evaluation indicators With support p, for example It refers to the proportion of item set 3∪4 in the entire driving behavior evaluation index item set. The confidence level is q, for example It means that the probability that the evaluation index item set contains 3 and 4 at the same time is q, which is actually the conditional probability P(4|3).

[0105] Support is p, expressed as:

[0106]

[0107] Confidence is q, expressed as:

[0108]

[0109] The driving behavior evaluation index lift represents the degree to which the probability of item β increasing due to item α. The lift is expressed as:

[0110]

[0111] When the lift is greater than 1, there is an improvement; when the lift is less than 1, there is a decrease; and when the lift is equal to 1, there is no change.

[0112] The frequent item sets of driving behavior evaluation indicators are represented as the evaluation indicator item sets whose support is greater than or equal to the minimum support threshold.

[0113] like Figure 6 As shown in the FP-tree operation flow chart, take 8 groups of travel data as an example:

[0114] S501: Encode and label the driving behavior evaluation index parameter database E, i.e., the travel index data set, and set the minimum support p min = 10% and minimum stability q min =50%. According to the S4 label table, the 8 groups of driving behavior evaluation indicators are coded as [2,3,6],[2,3,6],[2,3,6],[2,6],[2,6],[2,6],[2],[2].

[0115] S502: Select a frequent itemset F from database E: {2}{6}{2,6}.

[0116] S503: Verify whether the frequent item set F satisfies the support p. If the frequent item set F does not satisfy the support p, delete the data; if the frequent item set F satisfies the support p, create item table 4, repeat the above operation for each evaluation index item set, sort the driving behavior evaluation index frequent items from large to small according to the support p, and construct the FP-Tree to build the conditional pattern base, that is, the set of itemsets remaining after removing an item from a frequent item set.

[0117] Table 4 Project table

[0118] project Support Linked List 2 1 / 6 0.75 /

[0119] S504: Perform recursive iteration until the tree contains an element item, and construct a conditional FP tree;

[0120] Construct a conditional FP tree and perform recursive iteration until the tree contains an element item. Create a root node. If the item already exists in the child node of the FP tree, increase the frequency of the item; if the item does not exist in the tree, create a new node, and repeat the above operations for the frequent item set F of the driving behavior evaluation index according to the ranking obtained in S503, and expand the FP tree through the frequent item set F. For example: root node root→2-8→6-6, the previous item of 2-8 is item 2, and the subsequent item is frequency 8.

[0121] S505: Search for frequent items in the conditional FP tree, obtain the frequent item set L and calculate its support p;

[0122] S506: Verify whether the frequent itemset L satisfies the minimum support p min and the minimum confidence q min If the frequent itemset L does not satisfy the minimum support p min and the minimum confidence q min , then delete this frequent item set; if the frequent item set L satisfies the minimum support p min and the minimum confidence q min , then the ecological driving behavior association method is generated:

[0123] frozenset({6})-->frozenset({2})conf:1.0

[0124] frozenset({2})-->frozenset({6})conf:0.75

[0125] S507: Output FP-tree operation time T1. Use Apriori method to perform driving behavior correlation analysis on the same data set and output operation time T2. T2 is the operation time of Apriori method, and T1 can be obtained. <T2。

[0126] S6: Conduct correlation analysis on eco-driving behavior indicators based on the operating results obtained by the FP-tree model constructed in S5;

[0127] The results of the FP-tree method are shown in Table 5:

[0128] Table 5 FP-tree operation results

[0129] Related items Support Confidence Lift 5-4 60％ 87％ 2 5-1 40％ 90％ 0.6 1-5-4 35％ 54％ 1 2-3-1 25％ 98％ 1.7

[0130] For the same driver, the same model of tram, and the same route, m trip events, the associated items formed by the preceding evaluation index and the following evaluation index of the obtained driving behavior are arranged from large to small according to the support percentage. As shown in Table 5, the maximum support of the associated items between the two driving behavior evaluation indicators is 60%, which indicates that the pedal stroke percentage has the strongest correlation with the driving behavior of the speed change frequency. If the maximum confidence of the associated items between the two driving behavior evaluation indicators is 90%, this indicates that the preceding indicator, the pedal stroke percentage, has a maximum confidence of 90% probability of affecting the following evaluation indicator, the average vehicle speed. Find out the maximum support of the three driving behavior evaluation indicators, which is 35%, which indicates that the preceding two driving behavior indicators, namely the average vehicle speed and the pedal stroke percentage, have the strongest correlation with the following driving behavior, the speed change frequency. If the maximum confidence of the associated items of the three driving behavior evaluation indicators is 98%, this indicates that the preceding two driving behavior evaluation indicators, namely the speed fluctuation difference and the acceleration change, have a maximum confidence of 98% probability of affecting the following evaluation indicator, the average vehicle speed. The changes in the associated items and the number of associated items correspond to the corresponding conclusions.

[0131] S7: Driving behavior judgment: judging the driver's driving behavior by setting judgment criteria based on support and confidence.

[0132] This method can analyze the data of different drivers, obtain the evaluation results of each driver, and judge the driver's driving behavior according to the evaluation results. As shown in Table 6, the driving behavior evaluation criteria are set to divide the support judgment interval into three levels: strong (50% to 100%), medium (37.5% to 50%), and weak (12.5% ​​to 37.5%), and divide the confidence judgment interval into three levels: high (95% to 100%), medium (80% to 95%), and low (50% to 80%).

[0133] Table 6 Driving behavior evaluation criteria

[0134]

[0135] In the analysis of driving behavior, we mainly focus on the related items of driving behavior evaluation indicators with strong support and high confidence. Combined with the actual collected data results, we exclude the related items with support lower than 12.5% ​​and confidence lower than 50%.

[0136] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0137] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for analyzing the correlation of a driver's driving behavior, characterized in that: The following steps are involved: S1: Collecting driver travel data: Collecting data through the collection device on the test vehicle, recording the driver's complete trip as a travel event, that is, the starting point and the end point overlap to form a closed-loop route, collecting a total of m travel events of the same driver, the same model of electric car, and the same route, collecting the signal source code obtained by the collection device in each travel event, the signal includes: time, vehicle speed, acceleration pedal travel percentage, brake pedal travel percentage, steering wheel steering, steering wheel angle, and steering wheel angle velocity, while locating the driving position of the test vehicle through the GPS inertial navigation system, collecting longitude and latitude parameters and time parameters; S2: Divide the travel distance and analyze the signal source code in S1, use sliding window mean filtering to filter the vehicle driving data, and record every 1 minute as a time period; S3: Selection and analysis of driving behavior evaluation indicators: The average speed of the test vehicle during driving, pedal travel percentage, vehicle speed fluctuation difference, acceleration and deceleration change value, speed change frequency and parking time are selected as driving behavior evaluation indicator parameters and the average vehicle speed threshold E is set. V , vehicle speed fluctuation difference threshold E w , Acceleration / deceleration fluctuation difference threshold E a , speed change frequency threshold E f , Pedal travel percentage threshold E p and parking time threshold E t As a threshold value for driving behavior evaluation indicators; S4: sequentially label the driving behavior evaluation index parameters in S3, and convert the driving behavior evaluation index parameters exceeding the threshold into a digital set, and define the trip index data set through the digital set; S5: FP-tree model construction based on the trip indicator dataset in S4; The specific steps of S5 are as follows: Support is p, and the calculation formula is: Among them, α is the index item set of the predecessor of the associated item, β is the index item set of the successor of the associated item, and P is the probability; Confidence is q, and the calculation formula is: S501: Encode the driving behavior evaluation index parameter database E, i.e., the travel index data set, and set the minimum support p min and the minimum confidence q min ; S502: Select frequent itemsets F from database E; S503: Verify whether the frequent item set F satisfies the support p; if the frequent item set F does not satisfy the support p, delete the data; if the frequent item set satisfies the support p, create an item header table, construct an FP-Tree, and construct a conditional pattern base, that is, a set of itemsets remaining after removing an item from a frequent item set; S504: Perform recursive iteration until the tree contains an element item, and construct a conditional FP tree; S505: Search for frequent items in the conditional FP tree, obtain the frequent item set L and calculate its support p; S506: Verify whether the frequent itemset L satisfies the minimum support p min and the minimum confidence q min ; If the frequent itemset L does not satisfy the minimum support p min and the minimum confidence q min , then delete this frequent item set; if the frequent item set L satisfies the minimum support p min and the minimum confidence q min , then the ecological driving behavior association method is generated; S507: Output FP-tree operation time T1; S6: Conduct correlation analysis on eco-driving behavior indicators based on the operating results obtained by the FP-tree model constructed in S5; S7: Driving behavior judgment: judging the driver's driving behavior by setting judgment criteria based on support and confidence.

2. The method for analyzing the correlation between driver's driving behavior according to claim 1, characterized in that: The specific steps of S2 are as follows: The signal source code obtained by the acquisition device is analyzed offline, and the speed and acceleration are processed by sliding window mean filtering. The filtered travel data is sorted according to each trip, and the maximum value of the time column is obtained to determine the upper limit of the time period. The time period is created and the number of times is counted, with 1 minute as a time period. The traffic light intersection entering the journey boundary section is selected as the dividing point, and the corresponding longitude and latitude at this moment are found according to the data collected by the GPS inertial navigation system equipment, and the corresponding time from the departure at this moment is obtained, and the departure time obtained by the GPS inertial navigation system is synchronized to the acquisition device.

3. The method for analyzing the correlation between driver's driving behavior according to claim 1, characterized in that: The specific steps of S3 are as follows: Get the average speed threshold Analyze the average speed data collected from m trips, analyze the average speed range of electric vehicles, find the concentrated range, and calculate the average speeding threshold of pure electric vehicles according to the following formula Where: E M is the 75% quantile of the total distribution of all samples of average vehicle speed; E N is the quartile difference, that is, the difference between the 75% quantile and the 25% quantile of the total distribution of all samples of average vehicle speed; is the average speed exceeding threshold of pure electric vehicles; Calculate the speed fluctuation threshold E of pure electric vehicles w :If the speed fluctuation difference exceeds the speed fluctuation difference threshold E w The trip is considered as non-ecological driving, and the speed fluctuation threshold E w is the difference between the maximum speed and the average speed during driving, that is Where: E w is the speed fluctuation value of the tram (unit: km / h); E Vmax is the maximum speed of the trip (unit: km / h); is the speeding threshold of the average trip speed (unit: km / h); Obtain the acceleration change threshold E a :Calculate the range of acceleration change values ​​of group m: df["acceleration"] = df.iloc[:,1].diff() / df.iloc[:,0].diff() accel_min = df["acceleration"].min() accel_max = df["acceleration"].max() Then the acceleration changes from accel_min to accel_max m / s 2 , according to the acceleration change and test conditions, select the change threshold E a ; If the acceleration is greater than the maximum acceleration or the deceleration is less than the maximum deceleration, it is considered to be over the threshold; Calculate the speed change frequency threshold E f :According to the travel divided by S1, set the acceleration change value caused by speed change during normal driving in am / s 2 Inside, calculate E f : E f =Q / T Where: E f is the frequency of speed change; Q is the number of speed changes; T is the unit time (unit: min); 1 time / minute is set as the speed change frequency threshold. If the speed change frequency is greater than the threshold, it is considered as frequent speed change behavior, which is non-ecological driving. Calculate the pedal travel change threshold E p :Calculate the pedal stroke percentage change value range of group m, the pedal stroke percentage change is pedal_min~pedal_max%; Calculate the parking time threshold E t : Detect and record the duration of each time when the speed is 0km / h in the vehicle speed data, and calculate the average parking threshold E of the electric vehicle t : AND t =And t1 +1.5E t2 Where: E t1 is the 75% quantile of the total distribution of all samples of average vehicle speed; E t2 It is the interquartile difference, that is, the difference between the 75% quantile and the 25% quantile of the total distribution of all samples of average vehicle speed.

4. The method for analyzing the correlation between driver's driving behavior according to claim 1, characterized in that: The specific steps of S4 are as follows: A threshold is judged for the driving behavior evaluation index parameters in each trip. If the threshold is exceeded, it is recorded as an event, which is represented by a digital set {1}, {2}, {3}, {4}, {5}, {6}. All events occurring in the trip are defined as the trip index data set {1, 2, 3...}.

5. The method for analyzing the correlation between driver's driving behavior according to claim 1, characterized in that: The specific steps of S6 are as follows: The associated items formed by the obtained driving behavior evaluation index and the subsequent evaluation index are arranged from large to small according to the support percentage. The maximum support between the associated items indicates that the correlation between the driving behavior evaluation indicators of the item is the strongest, and the maximum confidence between the associated items indicates that the probability that the preceding driving behavior evaluation index of the item has the highest confidence will affect the subsequent driving behavior evaluation index.

6. The method for analyzing the correlation between driver's driving behavior according to claim 1, characterized in that: The specific steps of S7 are as follows: The support judgment interval is divided into three levels: strong 50%~100%, medium 37.5%~50%, and weak 12.5%~37.5%. The confidence judgment interval is divided into three levels: high 95%~100%, average 80%~95%, and low 50%~80%. In the analysis of driving behavior, we mainly focus on the related items of driving behavior evaluation indicators with strong support and high confidence. Combined with the actual collected data results, we exclude the related items with support lower than 12.5% ​​and confidence lower than 50%.

Images