Heavy truck driving behavior data processing method and device, equipment and medium

Abstract

The invention relates to the technical field of driving behavior data processing, and discloses a heavy truck driving behavior data processing method and device, equipment and a medium, and the method comprises the steps: firstly obtaining a driving behavior vector of a to-be-evaluated object; then the driving behavior vector is input into a driving style recognition model, the driving style recognition model comprises a multi-level cascade model, and the cascade model comprises a plurality of sub-models; performing primary feature mapping on the driving behavior vector by using a plurality of sub-models of a first level to obtain a first probability set; and then inputting the driving behavior vector and the first probability set into a second hierarchy, performing advanced feature fusion by using a plurality of sub-models on the second hierarchy to obtain a second probability set, finally determining a target driving style of the to-be-evaluated object based on the second probability set, and generating performance assessment data of the to-be-evaluated object based on the target driving style. Through a multi-level cascade structure, the driving behavior characteristics are gradually optimized and enhanced, so that the accuracy of the performance assessment data of the to-be-assessed object is improved.

Description

Technical Field

[0001] The present application relates to the technical field of driving behavior data processing, and in particular to a method, device, equipment and medium for processing driving behavior data of a heavy truck. Background Art

[0002] Emissions from heavy trucks are receiving increasing attention. In addition to the configuration of the vehicle itself, the driver's driving behavior is also an important cause of high fuel consumption. Effective supervision of the driver's driving behavior can help improve the economy of the vehicle. However, in the relevant technology, the accuracy of heavy truck driver driving behavior evaluation still has room for improvement.

[0003] Therefore, it is urgent to propose a new heavy truck driving behavior data processing method. Summary of the invention

[0004] The present application provides a method, device, equipment and medium for processing heavy truck driving behavior data, which solves the technical problem that the accuracy of heavy truck driver driving behavior evaluation needs to be improved, and achieves the technical effect of more accurate heavy truck driver driving behavior evaluation.

[0005] In order to achieve the above objectives, the main technical solutions adopted in this application include:

[0006] In a first aspect, an embodiment of the present application provides a method for processing heavy truck driving behavior data, the method comprising:

[0007] Obtaining a driving behavior vector corresponding to the process of the subject to be evaluated driving a heavy truck;

[0008] Inputting the driving behavior vector into a driving style recognition model, wherein the driving style recognition model includes a multi-level cascade model; the cascade model includes a plurality of sub-models;

[0009] Using multiple sub-models on the first level to perform primary feature mapping on the driving behavior vector respectively, to obtain a first probability set; wherein the first probability set includes the predicted probability of each sub-model on the first level under multiple preset driving styles;

[0010] Inputting the driving behavior vector and the first probability set together into a second layer connected to the first layer, and performing high-level feature fusion using multiple sub-models on the second layer to obtain a second probability set; wherein the second probability set includes the predicted probability of each sub-model on the second layer under the multiple preset driving styles;

[0011] The target driving style of the object to be evaluated is determined based on the second probability set, and the performance evaluation data of the object to be evaluated is generated according to the safety performance data and the economic performance data corresponding to the target driving style.

[0012] Optionally, the step of obtaining a driving behavior vector corresponding to a process in which the subject to be evaluated drives a heavy truck includes:

[0013] Acquire current driving behavior data generated by the subject to be evaluated while driving a heavy truck;

[0014] A driving behavior vector is constructed based on the current driving behavior data.

[0015] Optionally, the driving style recognition model includes a multi-level cascade forest, and the first-level cascade forest includes a plurality of first forest models and a plurality of second forest models;

[0016] The using of the plurality of sub-models on the first level to respectively perform primary feature mapping on the driving behavior vector to obtain a first probability set includes:

[0017] The driving behavior vector is respectively subjected to primary feature mapping using multiple first forest models and multiple second forest models on the first level to obtain the first probability set; wherein the first probability set includes a first prediction probability of each first forest model under the multiple preset driving styles, and a second prediction probability of each second forest model under the multiple preset driving styles.

[0018] Optionally, the driving style recognition model comprises a multi-level cascade forest, wherein the second level has a plurality of third forest models and a plurality of fourth forest models;

[0019] The step of using the multiple sub-models on the second level to perform high-level feature fusion to obtain a second probability set includes:

[0020] The second probability set is obtained by performing high-level feature fusion using multiple third forest models and multiple fourth forest models on the second level; wherein the second probability set includes the third prediction probability of each third forest model under the multiple preset driving styles, and the fourth prediction probability of each fourth forest model under the multiple preset driving styles.

[0021] Optionally, the last level of the driving style recognition model is recorded as the nth level; the nth level has a plurality of fifth forest models and a plurality of sixth forest models;

[0022] The determining the target driving style of the object to be evaluated based on the second probability set includes:

[0023] At the nth level, driving style prediction is performed according to the second probability set and the driving behavior vector to obtain a fifth prediction probability of each fifth forest model under the plurality of preset driving styles and a sixth prediction probability of each sixth forest model under the plurality of preset driving styles;

[0024] For each preset driving style, average calculation is performed based on the plurality of fifth prediction probabilities and the plurality of sixth prediction probabilities to obtain a prediction probability corresponding to each preset driving style;

[0025] The maximum probability is determined among the predicted probabilities corresponding to each preset driving style, and the preset driving style corresponding to the maximum probability is used as the target driving style.

[0026] Optionally, the plurality of preset driving styles are determined in the following manner:

[0027] Obtaining historical driving behavior vectors corresponding to each of the plurality of historical driving behavior data;

[0028] Determining a plurality of vector clusters corresponding to the historical driving behavior vectors;

[0029] In response to the labeling operation on the plurality of vector clusters, the plurality of preset driving styles are obtained.

[0030] Optionally, the determining a plurality of vector clusters corresponding to the historical driving behavior vectors includes:

[0031] Perform correlation division based on the historical driving behavior vector to obtain a target cluster number;

[0032] The historical driving behavior vectors are clustered according to the target number of clusters to obtain a plurality of vector clusters.

[0033] In a second aspect, an embodiment of the present application provides a device for processing driving behavior data of a heavy truck, the device comprising:

[0034] A vector acquisition module, used to acquire a driving behavior vector corresponding to the process of the subject to be evaluated driving a heavy truck;

[0035] A vector input module, used for inputting the driving behavior vector into a driving style recognition model, wherein the driving style recognition model comprises a multi-level cascade model; the cascade model comprises a plurality of sub-models;

[0036] A feature mapping module, configured to perform primary feature mapping on the driving behavior vector using a plurality of sub-models on the first level to obtain a first probability set; wherein the first probability set includes a predicted probability of each sub-model on the first level under a plurality of preset driving styles;

[0037] a feature fusion module, configured to input the driving behavior vector and the first probability set into a second layer connected to the first layer, and perform high-level feature fusion using a plurality of sub-models on the second layer to obtain a second probability set; wherein the second probability set includes a predicted probability of each sub-model on the second layer under the plurality of preset driving styles;

[0038] A driving style determination module is used to determine a target driving style of the object to be evaluated based on the second probability set, and generate performance evaluation data of the object to be evaluated according to the safety performance data and economic performance data corresponding to the target driving style.

[0039] In a third aspect, an embodiment of the present application provides a computer device, including:

[0040] A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method described in any of the above embodiments by executing the computer instructions.

[0041] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to enable a computer to execute the method described in any of the above embodiments.

[0042] In the embodiment of the present application, the driving behavior vector corresponding to the process of the subject to be evaluated driving a heavy truck is first obtained; then the driving behavior vector is input into the driving style recognition model, and the driving style recognition model includes a multi-level cascade model; the cascade model includes multiple sub-models; then the multiple sub-models on the first level are used to perform primary feature mapping on the driving behavior vector respectively to obtain a first probability set; then the driving behavior vector and the first probability set are input into the second level connected to the first level together, and the multiple sub-models on the second level are used to perform high-level feature fusion to obtain a second probability set, and finally the target driving style of the subject to be evaluated is determined based on the second probability set, and the performance evaluation data of the subject to be evaluated is generated according to the safety performance data and economic performance data corresponding to the target driving style. Through the cascade structure of multiple levels, the driving behavior characteristics are gradually optimized and enhanced, thereby improving the accuracy of the performance evaluation data of the subject to be evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1a A flow chart of a method for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0045] Figure 1b A flow chart of a method for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0046] Figure 1c A schematic diagram of a method for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0047] Figure 2 A flow chart of a method for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0048] Figure 3 A flow chart of a method for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0049] Figure 4 A flow chart of a method for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0050] Figure 5 A flow chart of a method for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0051] Figure 6 A schematic diagram of a device for processing heavy truck driving behavior data provided in an embodiment of this specification;

[0052] Figure 7 A schematic diagram of the structure of a computer device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0054] The issue of emissions from heavy trucks is receiving increasing attention. In addition to factors such as the vehicle's configuration, engine technology, and fuel type, the driver's driving behavior is also a key factor leading to high fuel consumption. Different driving styles, especially unsafe operations such as sudden acceleration, frequent braking, and speeding, often significantly increase fuel consumption and aggravate pollution emissions. Therefore, effective monitoring and assessment of the driver's driving behavior can not only help improve the vehicle's fuel economy, reduce carbon emissions, and improve operating cost-effectiveness, but also enhance driving safety. However, the accuracy of the heavy truck driver driving behavior assessment method in related technologies needs to be improved.

[0055] Based on this, this application provides a method for processing heavy truck driving behavior data, please refer to Figure 1a , the method comprising:

[0056] S101. Acquire current driving behavior data generated by the subject to be evaluated while driving a heavy truck.

[0057] S102: adaptively adjust the trained recognition model according to the driving mode corresponding to the current driving behavior data to obtain a driving style recognition model.

[0058] Among them, the driving mode can be divided into a simple mode and a complex mode. The simple mode can be a mode in which the vehicle speed does not change much, and the complex mode can be a mode in which the vehicle speed changes greatly.

[0059] The trained recognition model is adaptively adjusted. Specifically, if the driving mode belongs to a simple mode, a part of the recognition model is used as the driving style recognition model; if the driving mode belongs to a complex mode, the entire recognition model is used as the driving style recognition model.

[0060] S103: Inputting the driving behavior vector corresponding to the current driving behavior data into a driving style recognition model, wherein the driving style recognition model includes a multi-level cascade model; the cascade model includes a plurality of sub-models.

[0061] S104 , using the multiple sub-models on the first level to perform primary feature mapping on the driving behavior vector respectively to obtain a first probability set.

[0062] The first probability set includes the predicted probability of each sub-model on the first level under multiple preset driving styles.

[0063] S105: Input the driving behavior vector and the first probability set together into a second level connected to the first level, and use multiple sub-models on the second level to perform high-level feature fusion to obtain a second probability set.

[0064] The second probability set includes the predicted probability of each sub-model on the second level under the multiple preset driving styles.

[0065] S106. Determine a target driving style of the object to be evaluated based on the second probability set, and generate performance evaluation data of the object to be evaluated according to the safety performance data and economic performance data corresponding to the target driving style.

[0066] In some embodiments, during a certain journey or period of time, it is detected that the speed of the heavy truck is maintained in the medium speed range (i.e., the speed is greater than 30 km / h and less than 60 km / h) or in the high speed range (the speed is greater than 60 km / h and less than 80 km / h), and the acceleration change range is small, such as -1 m / s 2 ~1m / s 2 , and judge that the driver's driving behavior belongs to the simple mode. During a certain journey or a certain period of time, it is detected that the speed of a heavy truck is in the low speed range, medium speed range, and high speed range, and the acceleration change range is also large, such as -5m / s 2 ~5m / s 2 , determining that the driver’s driving behavior belongs to a complex mode.

[0067] In some embodiments, the recognition model includes a five-level cascade model. If the current driving mode belongs to a simple mode, the first three levels of the cascade model are used as the driving style recognition model; if it is detected that the current driving mode belongs to a complex mode, the five-level cascade model is used to predict the driving style.

[0068] In some embodiments, the cascade model of each level of the recognition model includes four sub-models. If the current driving mode belongs to a simple mode, the cascade model of each level uses two sub-models; if it is detected that the current driving mode belongs to a complex mode, the cascade model of each level uses four sub-models.

[0069] It is understandable that driving modes can be divided into more detailed categories, corresponding to a more refined driving style recognition model.

[0070] In the above embodiment, by adaptively adjusting the trained recognition model according to the driving mode corresponding to the current driving behavior data, the performance and accuracy of driving style recognition can be taken into account, thereby improving the performance and accuracy of generating performance evaluation data of the object to be evaluated. In simple mode, using a simpler model for recognition can reduce computational complexity and resource consumption, thereby improving the performance of the model. In complex mode, using a more complex model for recognition can capture more detailed features and improve the accuracy of recognition. Through this adaptive adjustment, the appropriate model can be dynamically selected according to different driving modes, which can not only ensure higher performance, but also provide more accurate recognition results when needed, thereby achieving a balance between performance and accuracy.

[0071] According to an embodiment of the present application, an embodiment of a method for processing heavy truck driving behavior data is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0072] See also Figure 1b and Figure 1c In this embodiment, a method for processing heavy truck driving behavior data is provided, the method comprising:

[0073] S110, obtaining a driving behavior vector corresponding to a process in which the subject to be evaluated drives a heavy truck.

[0074] The driving behavior vector may be a vector obtained based on driving behavior data of a heavy truck driven by the subject to be evaluated in one trip, and the elements of the vector may be characteristic values ​​of the driving behavior index.

[0075] In some embodiments, the terminal (T-BOX) collects various driving behavior data (such as speed, brake pedal opening, brake pedal opening, etc.) of heavy trucks in real time during driving through the CAN bus at a frequency of 1 Hz, and uploads these data to the Internet of Vehicles platform through a communication link. Secondly, the Internet of Vehicles platform processes these data (such as eliminating outliers, filling missing values, etc.), and finally calculates and obtains the characteristic values ​​of some required driving behavior indicators (such as speed mean, accelerator pedal standard deviation, etc.), and uses the characteristic values ​​to construct a driving behavior vector.

[0076] S120: Input the driving behavior vector into a driving style recognition model, where the driving style recognition model includes a multi-level cascade model; the cascade model includes a plurality of sub-models.

[0077] The driving style recognition model can be used to recognize the driving style of the object to be evaluated according to the driving behavior vector, such as "very intense", "intense", "normal", "smooth" and "very smooth". The sub-model can be a random forest model and / or a completely random forest model.

[0078] Specifically, the driving style recognition model includes a multi-level cascade model, which can process the original input driving behavior vector layer by layer. The cascade model of each level takes the output of the previous level as input and outputs the processed information to the next level. The output of the cascade model of each level can be a driving style recognition result. Exemplarily, there are five types of preset driving styles, and the corresponding driving style recognition results can be the probability that the driving style belongs to these five types respectively. The driving style recognition model gradually optimizes and enhances the driving behavior characteristics through the cascade model of each level, so as to have a stronger classification ability and finally output a more accurate prediction result.

[0079] It should be noted that the driving style recognition model can be a trained model or a model that has been adaptively adjusted.

[0080] S130 , using the multiple sub-models on the first level to perform primary feature mapping on the driving behavior vector respectively, to obtain a first probability set.

[0081] The primary feature map may be each sub-model on the first level, and according to the input driving behavior vector, after inference analysis, the predicted probability of the driving behavior under multiple preset driving styles is obtained. The first probability set includes the predicted probability of each sub-model on the first level under multiple preset driving styles.

[0082] In some embodiments, the cascade model includes four sub-models, and there are three types of preset driving styles. After the driving behavior vector is input into the driving style recognition model, the four sub-models on the first level perform primary feature mapping according to the driving behavior vector, and obtain the predicted probability of each preset driving style, which can be output in the form of a vector, such as P11 = (0.6, 0.3, 0.1), P12 = (0.7, 0.1, 0.2), P13 = (0.65, 0.25, 0.1) and P14 = (0.55, 0.35, 0.1), P1, P2, P3 and P4 constitute the first probability set.

[0083] S140, inputting the driving behavior vector and the first probability set together into a second layer connected to the first layer, and performing high-level feature fusion using multiple sub-models on the second layer to obtain a second probability set.

[0084] Among them, the advanced feature fusion can be that each sub-model of the second level combines the prediction probability output by the corresponding sub-model in the first level with the driving behavior vector, and further obtains the prediction probability of the driving behavior under multiple preset driving styles through reasoning analysis. The second probability set includes the prediction probability of each sub-model on the second level under multiple preset driving styles.

[0085] In some embodiments, the cascade model includes four sub-models. After the four sub-models on the first level output prediction probabilities respectively, they are input into the corresponding four sub-models on the second level. At the same time, the driving behavior vectors are also input into the four sub-models on the second level respectively. The four sub-models perform high-level feature fusion on the input prediction probabilities and driving behavior vectors respectively, and after reasoning, the prediction probabilities of driving behaviors under multiple preset driving styles are obtained, which can be P21, P22, P23 and P24, which constitute the second probability set.

[0086] Combining the driving behavior vector and the predicted probability output of the first level as the input information of the second level can provide richer information for the classification prediction of the second level, which helps to improve the prediction probability accuracy of the sub-model in the second level.

[0087] S150: Determine a target driving style of the object to be evaluated based on the second probability set, and generate performance evaluation data of the object to be evaluated according to the safety performance data and economic performance data corresponding to the target driving style.

[0088] The target driving style may be the driving style of the object to be evaluated. The safety performance data may be driving safety related data. The economic performance data may be fuel economy related data.

[0089] Specifically, the driving style recognition model includes multiple levels, the second probability set output by the second level is input to the third level, and the sub-models of the third level respectively input the predicted probability output by the corresponding sub-model of the second level, as well as the driving behavior vector, and obtain the predicted probability on each preset driving style by performing high-level feature fusion, and so on, until each sub-model of the nth level outputs the nth predicted probability vector on each preset driving style, and all nth predicted probability vectors constitute the nth probability set. The nth probability set can be averaged to obtain the nth predicted probability mean vector, and finally the maximum value is taken from the nth predicted probability mean vector, and the driving style corresponding to the maximum value is the target driving style of the object to be evaluated.

[0090] In some embodiments, the cascade model includes four sub-models, and there are three types of preset driving styles: fierce, normal, and stable. The nth predicted probability vector can be Pn1=(p 11 , p 12 , p 13)、Pn2=(p 21 , p 22 , p 23 )、Pn3=(p 31 , p 32 , p 33 ) and Pn4=(p 41 , p 42 , p 43 ), the elements in each nth prediction vector correspond to the prediction probabilities of fierce, normal, and stable driving styles, respectively. The prediction probability mean vector can be P = ((p 11 +p 21 +p 31 +p 41 ) / 4,(p 12 +p 22 +p 32 +p 42 ) / 4,(p 13 +p 23 +p 33 +p 43 ) / 4), the maximum value can be taken from the three elements of P. Assuming that the value of the first element is the largest, the driving style corresponding to the value is fierce, and the target driving style of the object to be evaluated is "fierce". It can be understood that the preset driving style can also be more types.

[0091] In some embodiments, the corresponding relationship between driving style and safety performance data and economic performance data is shown in Table 1. After obtaining the driving style of the object to be evaluated, its performance evaluation data is generated according to the corresponding relationship. For example, when the driving style is "intense", the performance evaluation data may be "relatively unsafe driving and relatively low fuel economy".

[0092] Table 1 Example of the correspondence between driving style, safety performance data, and economic performance data

[0093] Driving style Safety performance data Economic performance data Very intense Very unsafe Very poor fuel economy fierce Relatively unsafe Relatively low fuel economy generally General safety Average fuel economy smooth Safer Higher fuel economy Very stable Very safe Very high fuel economy

[0094] In the above embodiment, firstly, the driving behavior vector corresponding to the process of the object to be evaluated driving a heavy truck is obtained; then the driving behavior vector is input into the driving style recognition model, and the driving style recognition model includes a multi-level cascade model; the cascade model includes multiple sub-models; then the multiple sub-models on the first level are used to perform primary feature mapping on the driving behavior vector respectively to obtain a first probability set; then the driving behavior vector and the first probability set are input into the second level connected to the first level, and the multiple sub-models on the second level are used to perform high-level feature fusion to obtain a second probability set, and finally the target driving style of the object to be evaluated is determined based on the second probability set, and the performance evaluation data of the object to be evaluated is generated according to the safety performance data and economic performance data corresponding to the target driving style. By including multiple sub-models in the cascade model, the overfitting problem of a single sub-model can be effectively avoided. At the same time, through the cascade structure of multiple levels, the driving behavior characteristics are gradually optimized and enhanced, thereby improving the accuracy of the performance evaluation data of the object to be evaluated.

[0095] See also Figure 2 In some embodiments, obtaining a driving behavior vector corresponding to the process of the subject to be evaluated driving a heavy truck includes:

[0096] S210, obtaining current driving behavior data generated by the subject to be evaluated while driving a heavy truck.

[0097] S220: Construct a driving behavior vector based on current driving behavior data.

[0098] When the driver is driving, the terminal (T-BOX) collects dozens of driving behavior data including speed in real time at a frequency of 1Hz through the CAN bus, and uploads these data to the Internet of Vehicles platform through the 4G network. Abnormal values ​​can be data outside the range that does not conform to objective laws and need to be eliminated. For example, for driving behavior data, the speed range is 0km / h~150km / h, the throttle opening range is 0~100%, and the brake pedal opening range is 0~100%; the vehicle data obtained through calculation has an acceleration range of -5m / s 2 ~5m / s 2 .

[0099] After the outliers are removed, there will be missing data in the original position. The missing values ​​can be filled by interpolation. For example, linear interpolation can be used. x0 and y0 are the data before the missing value, and x1 and y1 are the data after the missing value. The interpolation formula is:

[0100]

[0101] Due to the vibration and random noise of heavy trucks, the driving behavior data is not very smooth. The filtering method can be used for smoothing. For example, the smoothing filter is used to process the data. The formula can be:

[0102]

[0103] Among them, M is the sliding filter window size, x k-i Driving behavior data.

[0104] Please refer to Table 2 for driving behavior indicators.

[0105] Table 2 Example of driving behavior index system

[0106]

[0107] The calculation method of each driving behavior indicator characteristic value is as follows:

[0108] 1) Idle speed ratio

[0109] Idle speed can be a state where the vehicle speed is 0 but the engine speed is not 0. The calculation formula is:

[0110]

[0111] Among them, F1 is the idle ratio, T1 is the idle time, and T is the total actual operation time of a single trip, in seconds. The calculation result of F1 is rounded to four decimal places.

[0112] 2) Low speed ratio

[0113] Low speed can be a state where the speed is greater than 10km / h and less than 30km / h. The calculation formula is:

[0114]

[0115] Among them, F2 is the low-speed ratio, T2 is the low-speed time, and T is the total actual operation time of a single trip, in seconds. The calculation result of F2 is rounded to four decimal places.

[0116] 3) Medium speed ratio

[0117] Medium speed can be a state where the speed is greater than 30km / h and less than 60km / h. The calculation formula is:

[0118]

[0119] Among them, F3 is the medium-speed ratio, T3 is the medium-speed time, and T is the total actual operation time of a single trip, in seconds. The calculation result of F3 is rounded to four decimal places.

[0120] 4) High-speed ratio

[0121] High speed can be a state where the speed is greater than 60km / h and less than 80km / h. The calculation formula is:

[0122]

[0123] Among them, F4 is the high-speed ratio, T4 is the high-speed time, and T is the total actual operation time of a single trip, in seconds. The calculation result of F4 is rounded to four decimal places.

[0124] 5) Ultra-high-speed ratio

[0125] Super high speed can be a state where the speed is greater than 80km / h, and the calculation formula is:

[0126]

[0127] Among them, F5 is the high-speed ratio, T5 is the high-speed time, and T is the total actual operation time of a single trip, in seconds. The calculation result of F5 is rounded to four decimal places.

[0128] 6) Speed ​​average

[0129] It is necessary to remove the data with speed V < 10km / h, and then calculate the speed mean. The calculation formula is:

[0130]

[0131] Among them, V m is the mean velocity in m / s, v i is any speed value, and n is the number of samples. m Keep two decimal places.

[0132] 7) Speed ​​standard deviation

[0133] It is necessary to eliminate the data with speed V < 10km / h, and then calculate the standard deviation of vehicle speed. The calculation formula is:

[0134]

[0135] Among them, V std is the standard deviation of vehicle speed, v i is any speed value, n is the number of samples, V m is the mean velocity. std The calculation result should be rounded to two decimal places.

[0136] 8) Accelerator pedal average

[0137] The calculation formula of the average accelerator pedal opening is:

[0138]

[0139] in, is the average accelerator pedal opening, is a single opening, and n is the total number of samples. Keep four decimal places.

[0140] 9) Accelerator pedal standard deviation

[0141] Eliminate the data with accelerator pedal opening of 0, and then calculate the standard deviation of accelerator pedal opening. The calculation formula is:

[0142]

[0143] Among them, F(lp) is the standard deviation of the accelerator pedal opening, For a single opening, is the average value of the accelerator pedal opening, n is the total number of samples. F(lp) is rounded to four decimal places.

[0144] 10) Low accelerator pedal opening ratio

[0145] The low throttle pedal opening can be a state where the throttle opening is greater than 0 and less than 30%, and the calculation formula is:

[0146]

[0147] Among them, p 0-30 is the proportion of low accelerator pedal opening, t 0-30 is the low accelerator pedal opening time, and T is the total actual operation time of a single trip. 0-30 The calculation result is rounded to four decimal places.

[0148] 11) Medium accelerator pedal opening ratio

[0149] The medium accelerator pedal opening can be a state where the accelerator opening is greater than or equal to 30% and less than 60%, and the calculation formula is:

[0150]

[0151] Among them, p 30-60 is the accelerator pedal opening ratio, t 30-60 is the accelerator pedal opening time, and T is the total actual operation time of a single trip. 30-60 The calculation result is rounded to four decimal places.

[0152] 12) High accelerator pedal opening ratio

[0153] The high throttle pedal opening can be a state where the throttle opening is greater than or equal to 60% and less than 80%, and the calculation formula is:

[0154]

[0155] Among them, p 60-80is the proportion of high accelerator pedal opening, t 60-80 is the high accelerator pedal opening time, and T is the total actual operation time of a single trip. 60-80 The calculation result is rounded to four decimal places.

[0156] 13) Ultra-high accelerator pedal ratio

[0157] The ultra-high throttle pedal opening can be a state where the throttle opening is greater than or equal to 80%, and the calculation formula is:

[0158]

[0159] Among them, p 80 is the proportion of ultra-high accelerator pedal opening, t 80 is the ultra-high accelerator pedal opening time, and T is the total actual operation time of a single trip. 80 The calculation result is rounded to four decimal places.

[0160] 14) Brake pedal average

[0161] It is necessary to remove the data with a brake pedal opening of 0, and then calculate the average brake pedal opening. The formula for the brake pedal average is:

[0162]

[0163] in, is the average opening of a single stroke, is the single stroke opening, and n is the total number of samples. Keep two decimal places.

[0164] 15) Brake pedal standard deviation

[0165] It is necessary to remove the data with a brake pedal opening of 0, and then calculate the standard deviation of the brake pedal opening. The calculation formula for the standard deviation of the brake pedal opening is:

[0166]

[0167] Among them, F(l q ) is the standard deviation of brake pedal opening, is the single stroke opening, is the average opening of a single stroke, and n is the total number of samples. q ) Keep two decimal places.

[0168] 16) Braking time percentage

[0169] It is necessary to eliminate the data with a brake pedal opening of 0 and then calculate the braking time. The calculation formula is:

[0170]

[0171] Among them, F6 is the braking time percentage, T6 is the braking time, and T is the total actual operating time of a single trip, in seconds. The calculation result of F6 is rounded to four decimal places.

[0172] 17) Average impact strength

[0173] The impact degree can be expressed as the second-order derivative of the velocity, and the calculation formula is:

[0174]

[0175] Among them, jerk is the impact degree, a2 is the acceleration at time t2, and a1 is the acceleration at time t1.

[0176]

[0177] Among them, a is the acceleration, v2 is the velocity at time t2, and v1 is the velocity at time t1.

[0178] The calculation formula of the mean impact degree is:

[0179]

[0180] in, is the mean value of shock, n is the number of shock samples, jerk i is the impact degree at any moment.

[0181] 18) Impact strength standard deviation

[0182] The calculation formula of impact standard deviation is:

[0183]

[0184] Among them, σ(jerk) is the standard deviation of the shock degree, is the mean value of shock, n is the number of shock samples, jerk i For any impact degree.

[0185] It should be noted that among the above driving behavior indicators, in order to eliminate the dimensional impact between different indicators, the mean and standard deviation indicators need to be normalized.

[0186]

[0187] Among them, x′ i is the normalized index, x max is the maximum value of this indicator in the historical driving behavior indicators, x min is the minimum value of this indicator in the historical driving behavior indicators, x i This is the original data of this indicator.

[0188] Finally, based on the eigenvalues ​​of the above indicators, the driving behavior vector of the driver can be constructed, for example, it can be a row vector composed of the eigenvalues ​​of the above 18 indicators, or it can be a column vector.

[0189] In the above embodiment, the driving behavior index system for constructing the driving behavior vector includes multiple dimensions such as speed, accelerator pedal, brake pedal and impact degree, and each dimension includes multiple specific indicators. Compared with identifying the driving style of the object to be evaluated by only a few indicators, the driving behavior vector constructed by using this index system helps to more accurately identify the driving style and conduct performance evaluation.

[0190] In some embodiments, the driving style recognition model includes a multi-level cascade forest, and a first-level cascade forest includes a plurality of first forest models and a plurality of second forest models;

[0191] The driving behavior vectors are respectively mapped to primary features using multiple sub-models on the first level to obtain a first probability set, including:

[0192] A plurality of first forest models and a plurality of second forest models on the first level are used to perform primary feature mapping on the driving behavior vector respectively to obtain a first probability set.

[0193] The first probability set includes a first prediction probability of each first forest model under a plurality of preset driving styles, and a second prediction probability of each second forest model under a plurality of preset driving styles.

[0194] In some embodiments, the first forest model may be a random forest model, and the second forest model may be a completely random forest model. The first level includes two random forest models and two completely random forest models, and there are three types of preset driving styles. After the driving behavior vector is input into the driving style recognition model, the two random forest models on the first level perform primary feature mapping according to the driving behavior vector respectively, and obtain the first prediction probability on each preset driving style, which can be output in the form of a vector, such as P1 = (0.6, 0.3, 0.1), P2 = (0.7, 0.1, 0.2); the two completely random forest models on the first level perform primary feature mapping according to the driving behavior vector respectively, and obtain the second prediction probability on each preset driving style, which is output in the form of a vector, such as P3 = (0.65, 0.25, 0.1), P4 = (0.55, 0.35, 0.1), P1, P2, P3 and P4 constitute the first probability set.

[0195] In the above embodiment, two different types of forest models are combined into a cascade model, which can give full play to the advantages of each model, improve the diversity and robustness of the model, and reduce overfitting in the process of gradually optimizing and enhancing features, thereby improving the final prediction accuracy.

[0196] In some embodiments, the driving style recognition model includes a multi-level cascade forest, with a plurality of third forest models and a plurality of fourth forest models on the second level; and a plurality of sub-models on the second level are used to perform high-level feature fusion to obtain a second probability set, including:

[0197] A second probability set is obtained by performing high-level feature fusion using multiple third forest models and multiple fourth forest models on the second level.

[0198] The second probability set includes the third prediction probability of each third forest model under multiple preset driving styles, and the fourth prediction probability of each fourth forest model under multiple preset driving styles.

[0199] In some embodiments, the third forest model may be a random forest model, and the fourth forest model may be a completely random forest model. The second level includes two random forest models and two completely random forest models. After the two random forest models on the first level output the first prediction probability respectively, they are input into the two random forest models corresponding to the second level; after the two completely random forest models on the first level output the second prediction probability respectively, they are input into the two completely random forest models corresponding to the second level. At the same time, the driving behavior vectors are also input into the four models of the second level respectively. The two random forest models perform high-level feature fusion on the first prediction probability and the driving behavior vector respectively to obtain the third prediction probability, which can be output in the form of a vector; the two completely random forest models perform high-level feature fusion on the second prediction probability and the driving behavior vector respectively to obtain the fourth prediction probability, which can be output in the form of a vector. The third prediction probability and the fourth prediction probability constitute the second probability set.

[0200] In the above embodiment, two different types of forest models are combined into a cascade model, which can give full play to the advantages of each model, improve the diversity and robustness of the model, and reduce overfitting in the process of gradually optimizing and enhancing features, thereby improving the final prediction accuracy.

[0201] See also Figure 3 In some embodiments, the last level of the driving style recognition model is recorded as the nth level; the nth level has a plurality of fifth forest models and a plurality of sixth forest models;

[0202] Determining a target driving style of the object to be evaluated based on the second probability set includes:

[0203] S510. At the nth level, driving style prediction is performed according to the second probability set and the driving behavior vector to obtain a fifth prediction probability of each fifth forest model under multiple preset driving styles and a sixth prediction probability of each sixth forest model under multiple preset driving styles.

[0204] Specifically, the driving style recognition model includes multiple levels. The sub-model of the third level inputs the second probability set output by the corresponding sub-model of the second level, as well as the driving behavior vector. By performing high-level feature fusion, the predicted probability of each preset driving style is obtained, and so on, until each sub-model of the nth level outputs the predicted probability of each preset driving style.

[0205] In some embodiments, the fifth forest model may be a random forest model, and the sixth forest model may be a completely random forest model. There are two random forest models and two completely random forest models on the nth level. There are three types of preset driving styles: fierce, normal, and stable. The two random forest models output the fifth prediction probability Pn1=(p 11 , p 12 , p 13 ) and Pn2=(p 21 , p 22 , p 23 ), the two completely random forest models output the sixth prediction probability Pn3 = (p 31 , p 32 , p 33 ) and Pn4=(p 41 , p 42 , p 43 ).

[0206] S520: For each preset driving style, average calculation is performed based on the multiple fifth prediction probabilities and the multiple sixth prediction probabilities to obtain a prediction probability corresponding to each preset driving style.

[0207] In some embodiments, for each preset driving style, the average calculation is performed based on the plurality of fifth prediction probabilities and the plurality of sixth prediction probabilities, and the prediction probability mean vector can be P=((p 11 +p 21 +p 31 +p 41 ) / 4,(p 12 +p 22 +p 32 +p 42 ) / 4,(p 13 +p 23 +p 33 +p 43 ) / 4).

[0208] S530: Determine a maximum probability among the predicted probabilities corresponding to each preset driving style, and use the preset driving style corresponding to the maximum probability as the target driving style.

[0209] In some embodiments, the maximum value is taken from all elements of the predicted probability mean vector P. For example, the value of the first element is the largest, and the driving style corresponding to the value is aggressive, so the target driving style of the object to be evaluated is aggressive.

[0210] In some embodiments, the accuracy of driving style recognition is evaluated by comparing some evaluation indicators of the driving style recognition model and other algorithm models (such as random forest, support vector machine, naive Bayes, XGBoost, etc.). These evaluation indicators include: ACC (Accuracy), RMSE (Root Mean Squared Error), MAE (Mean Absolute Error), MAPE (Mean Absolute Percentage Error and AUC (Area Under the Curve), see Table 3, where the higher the values ​​of ACC and AUC, the higher the accuracy of the model in driving style recognition; and the lower the values ​​of RMSE, MAE and MAPE, the higher the accuracy of the model in driving style recognition.

[0211] Table 3 Example table of accuracy evaluation of driving style recognition

[0212] Model ACC RMSE MAE MAPE AUC Driving style recognition model 0.973046 0.286723 0.044474 2.223720 0.999252 Random Forest 0.909704 0.526904 0.148248 5.480683 0.991577 Support Vector Machine 0.948787 0.370764 0.078167 4.177898 0.997157 Naive Bayes 0.892183 0.567541 0.168464 6.019766 0.985249 XGboost 0.901617 0.455573 0.132075 5.884996 0.990060

[0213] In the above embodiment, on the one hand, combining two different types of forest models into a cascade model can give full play to the advantages of each model, improve the diversity and robustness of the model, and reduce overfitting in the process of gradually optimizing and enhancing features, thereby improving the final prediction accuracy. On the other hand, by averaging the prediction probabilities of multiple models, the deviation that may be caused by a single model is reduced, making the final target driving style prediction more reliable.

[0214] See also Figure 4 In some embodiments, the plurality of preset driving styles are determined by:

[0215] S610: Obtain historical driving behavior vectors corresponding to each of the plurality of historical driving behavior data.

[0216] S620: Determine multiple vector clusters corresponding to the historical driving behavior vectors.

[0217] S630: In response to the labeling operation on the multiple vector clusters, multiple preset driving styles are obtained.

[0218] The historical driving behavior data may be a large amount of driving behavior data of heavy trucks obtained from the Internet of Vehicles platform. The historical driving behavior vector may be obtained by using the historical driving behavior data to calculate the required historical driving behavior indicators, and then construct the historical driving behavior vector corresponding to each trip through the historical driving behavior indicators. The preset driving style may be a classification name related to the driving behavior mode assigned to each vector cluster by analyzing the vectors in the vector cluster, such as fierce, smooth, etc.

[0219] Specifically, first, a large amount of driving behavior data of heavy trucks can be obtained from the Internet of Vehicles platform, and then the data can be used to obtain historical driving behavior indicators of a large number of trips according to the calculation method of the driving behavior indicators in Table 2. Through the historical driving behavior indicators, a historical driving behavior vector corresponding to each trip is constructed, for example, 10,000.

[0220] Furthermore, by clustering the historical driving behavior vectors into multiple vector clusters, a clustering algorithm without specifying the number of clusters can be used to discover clusters in the historical driving behavior vectors, such as the DBSCAN clustering algorithm. Compared with clustering with manually specified cluster numbers, clustering without specifying the number of clusters can more dynamically adjust the clustering structure and avoid problems such as unstable clustering caused by manually setting the wrong number of clusters.

[0221] Furthermore, after determining multiple vector clusters, the driving behavior characteristics of the vectors of each cluster can be manually analyzed. For example, the corresponding driving style can be obtained based on the analysis of the indicator range in the vector, and then the preset driving style can be used to label the vectors of each cluster, such as "very intense", "intense", "normal", "smooth" and "very smooth".

[0222] In the above embodiment, firstly, the historical driving behavior vectors corresponding to the plurality of historical driving behavior data are obtained; secondly, the plurality of vector clusters corresponding to the historical driving behavior vectors are determined. Finally, in response to the labeling operation of the plurality of vector clusters, a plurality of preset driving styles are obtained, laying a foundation for the subsequent driving style recognition.

[0223] See also Figure 5 In some embodiments, determining a plurality of vector clusters corresponding to the historical driving behavior vectors includes:

[0224] S710 , performing correlation division based on the historical driving behavior vectors to obtain a target number of clusters.

[0225] S720 . Cluster the historical driving behavior vectors according to the target number of clusters to obtain multiple vector clusters.

[0226] The correlation division may be a process of preliminarily grouping the historical driving behavior vectors.

[0227] Specifically, the SOM neural network model can be used to divide the historical driving behavior vectors into related parts. The main steps are as follows:

[0228] 1) Data preprocessing, mainly normalizing the data and eliminating the dimensions between different data;

[0229] 2) Initialization: assign smaller weights to the neuron connections from the i input layer to the output layer;

[0230] 3) Calculate the similarity. Input the historical driving behavior vector into the input layer of the SOM neural network and calculate the similarity between the weight vector of the mapping layer and the historical driving behavior vector. The Euclidean distance can be used for calculation. The formula is:

[0231]

[0232] Among them, ω ij is the weight between the i-th neuron in the input layer and the j-th neuron in the mapping layer. After calculating the similarity of each neuron, the most similar one is selected, i.e., d j The neuron with the smallest value is regarded as the winning neuron.

[0233] 4) Update weights and correct the winning neuron j * and the weights of the neighboring neurons

[0234]

[0235] Among them, η(t) is the learning rate, which gradually decreases to 0 over time. is a neighborhood function.

[0236] 5) Iterative training, repeat the above steps until the stopping condition is met, such as reaching a predetermined number of training rounds, or the learning rate drops below a preset threshold, that is, the SOM algorithm converges.

[0237] Furthermore, after the training is completed, the number of feature regions output by the SOM neural network is used as the target cluster number.

[0238] Furthermore, a clustering method with a specified number of clusters can be used to accurately cluster the historical driving behavior vectors. For example, the target number of clusters is used as the K value of the K-means algorithm for clustering. The clustering result evaluation formula can be the sum of squared errors (SSE), which is calculated as:

[0239]

[0240] Among them, d(x, C i ) is the distance from the cluster center to other particles in space,

[0241]

[0242] C i is the i-th cluster center, and m is the dimension of the historical driving behavior vector. When the change of the cluster center is lower than the preset threshold or the decrease of S is lower than the preset threshold, the K-means algorithm converges, the clustering ends, and the vector clusters of the target number of clusters can be obtained.

[0243] In the above embodiment, correlation division is first performed based on the historical driving behavior vectors to obtain the target number of clusters; then the historical driving behavior vectors are clustered according to the target number of clusters to obtain multiple vector clusters. A more accurate historical driving behavior vector clustering result is obtained through the two stages of preliminary clustering and precise clustering, providing more reliable data support for the determination of the preset driving style.

[0244] See also Figure 6 In this embodiment, a heavy truck driving behavior data processing device 800 is also provided. The heavy truck driving behavior data processing device 800 includes:

[0245] The vector acquisition module 810 is used to acquire the driving behavior vector corresponding to the process of the subject to be evaluated driving a heavy truck;

[0246] A vector input module 820, used to input the driving behavior vector into a driving style recognition model, wherein the driving style recognition model includes a multi-level cascade model; the cascade model includes a plurality of sub-models;

[0247] The feature mapping module 830 is used to perform primary feature mapping on the driving behavior vector respectively using the multiple sub-models on the first level to obtain a first probability set; wherein the first probability set includes the predicted probability of each sub-model on the first level under multiple preset driving styles;

[0248] A feature fusion module 840 is used to input the driving behavior vector and the first probability set into a second layer connected to the first layer, and perform high-level feature fusion using multiple sub-models on the second layer to obtain a second probability set; wherein the second probability set includes the predicted probability of each sub-model on the second layer under multiple preset driving styles;

[0249] The driving style determination module 850 is used to determine the target driving style of the object to be evaluated based on the second probability set, and generate performance evaluation data of the object to be evaluated according to the safety performance data and economic performance data corresponding to the target driving style.

[0250] In some implementations, the vector acquisition module 810 includes:

[0251] A data acquisition unit, used to acquire current driving behavior data generated by the subject to be evaluated during driving the heavy truck;

[0252] The vector construction unit is used to construct a driving behavior vector based on current driving behavior data.

[0253] In some embodiments, the driving style recognition model includes a multi-level cascade forest, and the first level of the cascade forest includes a plurality of first forest models and a plurality of second forest models; the feature mapping module 830 includes:

[0254] A feature mapping unit is used to use multiple first forest models and multiple second forest models on the first level to perform primary feature mapping on the driving behavior vector respectively to obtain a first probability set; wherein the first probability set includes a first prediction probability of each first forest model under multiple preset driving styles, and a second prediction probability of each second forest model under multiple preset driving styles.

[0255] In some embodiments, the driving style recognition model includes a multi-level cascade forest, with a second level having a plurality of third forest models and a plurality of fourth forest models; the feature fusion module 840 includes:

[0256] A feature fusion unit is used to use multiple third forest models and multiple fourth forest models on the second level to perform high-level feature fusion to obtain a second probability set; wherein the second probability set includes the third prediction probability of each third forest model under multiple preset driving styles, and the fourth prediction probability of each fourth forest model under multiple preset driving styles.

[0257] In some implementations, the last level of the driving style recognition model is recorded as the nth level; the nth level has a plurality of fifth forest models and a plurality of sixth forest models; the driving style determination module 850 includes:

[0258] a prediction probability acquisition unit, configured to perform driving style prediction at the nth level according to the second probability set and the driving behavior vector, to obtain a fifth prediction probability of each fifth forest model under a plurality of preset driving styles, and a sixth prediction probability of each sixth forest model under a plurality of preset driving styles;

[0259] an average calculation unit, configured to perform an average calculation for each preset driving style according to the plurality of fifth prediction probabilities and the plurality of sixth prediction probabilities, to obtain a prediction probability corresponding to each preset driving style;

[0260] The maximum probability acquisition unit is used to determine the maximum probability among the predicted probabilities corresponding to each preset driving style, and use the preset driving style corresponding to the maximum probability as the target driving style.

[0261] In some embodiments, the heavy truck driving behavior data processing device 800 further includes:

[0262] A historical vector acquisition module, used to acquire historical driving behavior vectors corresponding to each of a plurality of historical driving behavior data;

[0263] A vector cluster determination module, used to determine multiple vector clusters corresponding to historical driving behavior vectors;

[0264] The preset driving style acquisition module is used to obtain multiple preset driving styles in response to the labeling operation on multiple vector clusters.

[0265] In some implementations, the vector cluster determination module further includes:

[0266] A target cluster number determination unit, used for performing correlation division based on the historical driving behavior vector to obtain a target cluster number;

[0267] The clustering unit is used to cluster the historical driving behavior vectors according to the target cluster number to obtain multiple vector clusters.

[0268] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0269] The processing device for heavy truck driving behavior data in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above-mentioned functions.

[0270] See also Figure 7 , Figure 7 is a schematic diagram of the structure of a computer device provided in an embodiment of the present application, such as Figure 7 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 7 A processor 10 is taken as an example.

[0271] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0272] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0273] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0274] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0275] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0276] The embodiment of the present application also provides a computer-readable storage medium. The above method according to the embodiment of the present application can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0277] The embodiment of the present application provides a computer program product, which includes computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method of any embodiment of the present application.

[0278] Although the embodiments of the present application are described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

[0279] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0280] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0281] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0282] The present application is described with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, and the combination of the process and / or box in the flowchart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for realizing the function specified in one process or multiple processes in the flowchart and / or one box or multiple boxes in the block diagram.

[0283] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0284] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0285] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0286] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0287] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

[0288] Although the embodiments of the present application have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for processing heavy truck driving behavior data, characterized in that: The method comprises: Obtaining a driving behavior vector corresponding to the process of the subject to be evaluated driving a heavy truck; Inputting the driving behavior vector into a driving style recognition model, wherein the driving style recognition model includes a multi-level cascade model; the cascade model includes a plurality of sub-models; Using multiple sub-models on the first level to perform primary feature mapping on the driving behavior vector respectively, to obtain a first probability set; wherein the first probability set includes the predicted probability of each sub-model on the first level under multiple preset driving styles; Inputting the driving behavior vector and the first probability set together into a second layer connected to the first layer, and performing high-level feature fusion using multiple sub-models on the second layer to obtain a second probability set; wherein the second probability set includes the predicted probability of each sub-model on the second layer under the multiple preset driving styles; The target driving style of the object to be evaluated is determined based on the second probability set, and the performance evaluation data of the object to be evaluated is generated according to the safety performance data and the economic performance data corresponding to the target driving style.

2. The method according to claim 1, characterized in that The step of obtaining a driving behavior vector corresponding to a process in which the subject to be evaluated drives a heavy truck includes: Acquire current driving behavior data generated by the subject to be evaluated while driving a heavy truck; A driving behavior vector is constructed based on the current driving behavior data.

3. The method according to claim 1, characterized in that The driving style recognition model includes a multi-level cascade forest, and the first-level cascade forest includes a plurality of first forest models and a plurality of second forest models; The using of the plurality of sub-models on the first level to respectively perform primary feature mapping on the driving behavior vector to obtain a first probability set includes: The driving behavior vector is respectively subjected to primary feature mapping using multiple first forest models and multiple second forest models on the first level to obtain the first probability set; wherein the first probability set includes a first prediction probability of each first forest model under the multiple preset driving styles, and a second prediction probability of each second forest model under the multiple preset driving styles.

4. The method according to claim 1, characterized in that: The driving style recognition model includes a multi-level cascade forest, wherein the second level has a plurality of third forest models and a plurality of fourth forest models; The step of using the multiple sub-models on the second level to perform high-level feature fusion to obtain a second probability set includes: The second probability set is obtained by performing high-level feature fusion using multiple third forest models and multiple fourth forest models on the second level; wherein the second probability set includes the third prediction probability of each third forest model under the multiple preset driving styles, and the fourth prediction probability of each fourth forest model under the multiple preset driving styles.

5. The method according to claim 1, characterized in that The last level of the driving style recognition model is recorded as the nth level; The nth level has a plurality of fifth forest models and a plurality of sixth forest models; The determining the target driving style of the object to be evaluated based on the second probability set includes: At the nth level, driving style prediction is performed according to the second probability set and the driving behavior vector to obtain a fifth prediction probability of each fifth forest model under the plurality of preset driving styles and a sixth prediction probability of each sixth forest model under the plurality of preset driving styles; For each preset driving style, average calculation is performed based on the plurality of fifth prediction probabilities and the plurality of sixth prediction probabilities to obtain a prediction probability corresponding to each preset driving style; The maximum probability is determined among the predicted probabilities corresponding to each preset driving style, and the preset driving style corresponding to the maximum probability is used as the target driving style.

6. The method according to claim 1, characterized in that The multiple preset driving styles are determined in the following way: Obtaining historical driving behavior vectors corresponding to each of the plurality of historical driving behavior data; Determining a plurality of vector clusters corresponding to the historical driving behavior vectors; In response to the labeling operation on the plurality of vector clusters, the plurality of preset driving styles are obtained.

7. The method according to claim 6, characterized in that The determining of a plurality of vector clusters corresponding to the historical driving behavior vectors includes: Perform correlation division based on the historical driving behavior vector to obtain a target cluster number; The historical driving behavior vectors are clustered according to the target number of clusters to obtain a plurality of vector clusters.

8. A device for processing heavy truck driving behavior data, characterized in that: The device comprises: A vector acquisition module, used to obtain a driving behavior vector corresponding to the process of the subject to be evaluated driving a heavy truck; A vector input module, used for inputting the driving behavior vector into a driving style recognition model, wherein the driving style recognition model comprises a multi-level cascade model; the cascade model comprises a plurality of sub-models; A feature mapping module, configured to perform primary feature mapping on the driving behavior vector using a plurality of sub-models on the first level to obtain a first probability set; wherein the first probability set includes a predicted probability of each sub-model on the first level under a plurality of preset driving styles; a feature fusion module, configured to input the driving behavior vector and the first probability set into a second layer connected to the first layer, and perform high-level feature fusion using a plurality of sub-models on the second layer to obtain a second probability set; wherein the second probability set includes a predicted probability of each sub-model on the second layer under the plurality of preset driving styles; A driving style determination module is used to determine a target driving style of the object to be evaluated based on the second probability set, and generate performance evaluation data of the object to be evaluated according to the safety performance data and economic performance data corresponding to the target driving style.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images