Driver driving behavior correction training scheme generation method and device based on artificial intelligence, computer equipment and storage medium

Abstract

The invention relates to a driver driving behavior correction training scheme generation method and device based on artificial intelligence, computer equipment and a storage medium. The method comprises the following steps: acquiring historical data of a to-be-corrected driver; the historical data at least comprises historical operation behavior data, historical vehicle data and historical in-vehicle behavior video data of the driver to be corrected and historical environment data of the surrounding environment of the vehicle; according to the historical data, determining current driving style information of the driver to be corrected, and inputting the historical data into a preset multi-mode driving behavior recognition model to obtain abnormal driving behavior information of the driver to be corrected; determining a driving behavior risk value of the driver to be corrected according to the driving style information and the abnormal driving behavior information; and generating a behavior correction training scheme of the driver to be corrected according to the driving style information, the abnormal driving behavior information and the driving behavior risk value. By adopting the method, the accuracy of behavior correction training of the driver can be improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for generating driver behavior correction training programs based on artificial intelligence. Background Technology

[0002] Currently, in order to improve driving safety and reduce the risk of accidents, it is crucial to accurately conduct driving behavior correction training for drivers.

[0003] Traditional techniques for correcting driver behavior typically rely on human experience, standardized training content, or post-incident intervention. However, these methods lack a deep understanding of individual driver differences, making it difficult to provide personalized and targeted behavioral guidance, resulting in low accuracy in driver behavior correction training. Summary of the Invention

[0004] Therefore, it is necessary to provide an AI-based method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of driver behavior correction training by addressing the aforementioned technical problems.

[0005] Firstly, this application provides a method for generating a driver driving behavior correction training program based on artificial intelligence, including:

[0006] Obtain historical data of the driver to be corrected; the historical data includes at least the driver's historical operational behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment;

[0007] Based on the historical data, the current driving style information of the driver to be corrected is determined, and the historical data is input into a preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected.

[0008] Based on the driving style information and the abnormal driving behavior information, the driving behavior risk value of the driver to be corrected is determined;

[0009] Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, a behavior correction training plan is generated for the driver to be corrected.

[0010] In one embodiment, determining the current driving style information of the driver to be corrected based on the historical data includes:

[0011] Obtain the sample driving style vector corresponding to the driving style information of the sample drivers;

[0012] Clustering is performed on the sample driving style vectors to obtain the cluster centers corresponding to each driving style information;

[0013] Based on the historical data, a driving style vector of the driver to be corrected is constructed, and the driving style information with the smallest distance between the corresponding cluster center and the driving style vector is selected from the driving style information and used as the current driving style information of the driver to be corrected.

[0014] In one embodiment, constructing the driving style vector of the driver to be corrected based on the historical data includes:

[0015] Based on the historical operational behavior data, the first data feature of the driver to be corrected is determined; based on the historical vehicle data, the second data feature of the driver to be corrected is determined; based on the historical in-vehicle behavior video data, the third data feature of the driver to be corrected is determined; and based on the historical environmental data, the fourth data feature of the driver to be corrected is determined.

[0016] Based on the first data feature, the operational behavior index value of the driver to be corrected is determined; based on the second data feature, the state behavior index value of the driver to be corrected is determined; based on the third data feature, the in-vehicle behavior index value of the driver to be corrected is determined; and based on the fourth data feature, the environmental behavior index value of the driver to be corrected is determined.

[0017] Based on the operational behavior index value, the state behavior index value, the in-vehicle behavior index value, and the environmental behavior index value, a driving style vector of the driver to be corrected is constructed.

[0018] In one embodiment, determining the driving behavior risk value of the driver to be corrected based on the driving style information and the abnormal driving behavior information includes:

[0019] A preset driving style risk mapping table is queried to obtain the first driving behavior risk value corresponding to the driving style information; the preset driving style risk mapping table is used to represent the mapping relationship between driving style information and driving behavior risk value.

[0020] Identify the abnormal behavior type corresponding to the abnormal driving behavior information, determine the driving behavior risk value corresponding to each abnormal behavior type, and sum the driving behavior risk values ​​corresponding to each abnormal behavior type to obtain a second driving behavior risk value.

[0021] The first driving behavior risk value and the second driving behavior risk value are fused together to obtain the driving behavior risk value of the driver to be corrected.

[0022] In one embodiment, generating a behavior correction training plan for the driver to be corrected based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value includes:

[0023] Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, the target driver with the greatest similarity to the driver to be corrected is determined from the sample drivers;

[0024] Obtain the sample behavior correction training scheme of the target driver, and determine the initial behavior correction training scheme of the driver to be corrected based on the sample behavior correction training scheme of the target driver.

[0025] Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, the initial behavior correction training program is updated to obtain the behavior correction training program for the driver to be corrected.

[0026] In one embodiment, determining the target driver with the highest similarity to the driver to be corrected from the sample drivers based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value includes:

[0027] Obtain sample driving style information, sample abnormal driving behavior information, and sample driving behavior risk value of sample drivers;

[0028] Based on the difference between the driving style information and the sample driving style information, a first similarity between the driver to be corrected and the sample driver is determined; based on the difference between the abnormal driving behavior information and the sample abnormal driving behavior information, a second similarity between the driver to be corrected and the sample driver is determined; and based on the difference between the driving behavior risk value and the sample driving behavior risk value, a third similarity between the driver to be corrected and the sample driver is determined.

[0029] The first similarity, the second similarity, and the third similarity are fused to obtain the similarity between the driver to be corrected and the sample driver;

[0030] From the sample drivers, the sample driver with the highest similarity is selected as the target driver corresponding to the driver to be corrected.

[0031] Secondly, this application also provides a driver driving behavior correction training program generation device based on artificial intelligence, comprising:

[0032] The data acquisition module is used to acquire historical data of the driver to be corrected; the historical data includes at least the driver's historical operating behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment;

[0033] The data processing module is used to determine the current driving style information of the driver to be corrected based on the historical data, and to input the historical data into a preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected.

[0034] The risk determination module is used to determine the driving behavior risk value of the driver to be corrected based on the driving style information and the abnormal driving behavior information.

[0035] The scheme generation module is used to generate a behavior correction training scheme for the driver to be corrected based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value.

[0036] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0037] Obtain historical data of the driver to be corrected; the historical data includes at least the driver's historical operational behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment;

[0038] Based on the historical data, the current driving style information of the driver to be corrected is determined, and the historical data is input into a preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected.

[0039] Based on the driving style information and the abnormal driving behavior information, the driving behavior risk value of the driver to be corrected is determined;

[0040] Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, a behavior correction training plan is generated for the driver to be corrected.

[0041] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0042] Obtain historical data of the driver to be corrected; the historical data includes at least the driver's historical operational behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment;

[0043] Based on the historical data, the current driving style information of the driver to be corrected is determined, and the historical data is input into a preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected.

[0044] Based on the driving style information and the abnormal driving behavior information, the driving behavior risk value of the driver to be corrected is determined;

[0045] Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, a behavior correction training plan is generated for the driver to be corrected.

[0046] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0047] Obtain historical data of the driver to be corrected; the historical data includes at least the driver's historical operational behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment;

[0048] Based on the historical data, the current driving style information of the driver to be corrected is determined, and the historical data is input into a preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected.

[0049] Based on the driving style information and the abnormal driving behavior information, the driving behavior risk value of the driver to be corrected is determined;

[0050] Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, a behavior correction training plan is generated for the driver to be corrected.

[0051] The aforementioned method, device, computer equipment, storage medium, and computer program product for generating driver behavior correction training schemes based on artificial intelligence first acquire historical operational behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment of the driver to be corrected, as historical data of the driver to be corrected. Then, based on the historical data, the current driving style information of the driver to be corrected is determined, and the historical data is input into a preset multimodal driving behavior recognition model to obtain abnormal driving behavior information of the driver to be corrected. Then, based on the driving style information and abnormal driving behavior information, the driving behavior risk value of the driver to be corrected is determined. Finally, based on the driving style information, abnormal driving behavior information, and driving behavior risk value, a behavior correction training scheme for the driver to be corrected is generated. In this way, when conducting driving behavior correction training for drivers, the individual driving style of the driver to be corrected is first characterized based on historical data. At the same time, a multimodal model is used to accurately identify specific abnormal driving behavior information. Combining the two methods, the driving behavior risk value of the driver to be corrected can be determined comprehensively and accurately. Based on the above information, a behavior correction training plan that is more suitable for the driver to be corrected can be generated, which helps to improve the accuracy of the determination of the behavior correction training plan, thereby improving the accuracy of the driver's behavior correction training. Moreover, the entire process does not require human intervention, avoiding the shortcomings of relying on human experience judgment, uniform training content, or post-intervention methods that lack a deep understanding of individual driving differences and are difficult to achieve personalized and targeted behavioral guidance, resulting in low accuracy of driver behavior correction training. This further improves the accuracy of driver behavior correction training. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart illustrating a method for generating a driver's driving behavior correction training scheme based on artificial intelligence in one embodiment.

[0054] Figure 2 This is a flowchart illustrating the steps for determining the current driving style information of the driver to be corrected in one embodiment.

[0055] Figure 3 This is a flowchart illustrating a method for generating a driver's driving behavior correction training scheme based on artificial intelligence, as described in another embodiment.

[0056] Figure 4This is a structural block diagram of an AI-based driver behavior correction training scheme generation device in one embodiment.

[0057] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0059] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0060] In one exemplary embodiment, such as Figure 1 As shown, a method for generating driver behavior correction training schemes based on artificial intelligence is provided. This embodiment illustrates the application of this method to a server; it is understood that this method can also be applied to a terminal, or to a system including a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, and tablets; the server can be a standalone server or a server cluster composed of multiple servers. In this embodiment, the method includes the following steps:

[0061] Step S101: Obtain historical data of the driver to be corrected; the historical data includes at least the driver's historical operation behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment.

[0062] Among them, drivers awaiting correction refer to drivers who need to undergo driving behavior correction training.

[0063] The historical operational behavior data includes the accelerator pedal depth, brake pedal depth, steering wheel angle, gear status, and number of braking actions of the driver to be corrected at historical times.

[0064] The historical vehicle data includes the speed, acceleration, and position of the vehicle driven by the driver to be corrected at a historical time.

[0065] Among them, historical in-vehicle behavior video data includes videos of the driver's in-vehicle behavior at historical times, such as the driver's gaze direction (e.g., looking straight ahead, looking away from the phone or out the window).

[0066] Among them, the vehicle's surrounding environment is used to represent the external environment in which the driver to be corrected was driving the vehicle at a historical time.

[0067] The historical environmental data includes the distribution of driving time of the drivers to be corrected in the past (such as the proportion of night driving time) and the road condition complexity index (a quantitative score of the road condition complexity through preset rules, such as 0-10 points, with higher values ​​indicating more complex road conditions).

[0068] For example, the server obtains the driver identifier (such as driver's name, driver number, etc.) of the driver to be corrected, and retrieves the historical operation behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment corresponding to the driver identifier from the database, as the historical operation behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment of the driver to be corrected; then, the server performs time alignment of the historical operation behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment according to the timestamp, and performs structured representation to obtain the historical data of the driver to be corrected.

[0069] Step S102: Based on historical data, determine the current driving style information of the driver to be corrected, and input the historical data into the preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected.

[0070] Among them, the current driving style information is used to represent the driving style information of the driver to be corrected at the current time.

[0071] Among them, driving style information is used to indicate the classification of the driving characteristics of the driver to be corrected (such as aggressive, steady, distracted, etc.).

[0072] Among them, the multimodal driving behavior recognition model refers to a network model that can use the historical data of the driver to be corrected to obtain information on the abnormal driving behavior of the driver to be corrected, such as a fusion model based on Transformer.

[0073] Among them, abnormal driving behavior information is used to represent the specific abnormal driving behavior of the driver to be corrected, including quantitative information such as the type of abnormal behavior of the driver to be corrected (such as rapid acceleration, long-term gaze deviation, speeding at night, etc.), the number of occurrences (such as 12 rapid accelerations per month), and the duration (such as a cumulative 8 hours of gaze deviation per month).

[0074] For example, the server queries the correspondence between historical data and driving style information to obtain the driving style information corresponding to the historical data, which serves as the current driving style information of the driver to be corrected. Specifically, the server queries the correspondence between operational behavior data and driving style information to obtain the driving style information corresponding to historical operational behavior data, which serves as the first driving style information of the driver to be corrected; it queries the correspondence between vehicle data and driving style information to obtain the driving style information corresponding to historical vehicle data, which serves as the second driving style information of the driver to be corrected; and it queries the correspondence between in-vehicle behavior video data and driving style information to obtain the driving style information corresponding to historical in-vehicle behavior video data, which serves as the third driving style information of the driver to be corrected. The server firstly retrieves three types of driving style information. It then queries the correspondence between environmental data and driving style information to obtain the driving style information corresponding to historical environmental data, which serves as the fourth driving style information for the driver to be corrected. Next, the server fuses the first, second, third, and fourth driving style information to determine the current driving style information of the driver to be corrected. Finally, the server inputs historical data into a preset multimodal driving behavior recognition model to obtain the predicted probability of the driver to be corrected under each preset abnormal driving behavior information. From these preset abnormal driving behavior information, the server selects the preset abnormal driving behavior information with the highest predicted probability, which is then used as the abnormal driving behavior information for the driver to be corrected.

[0075] It should be noted that the preset multimodal driving behavior recognition model can be trained in the following way: obtain sample data of sample drivers; input the sample data into the multimodal driving behavior recognition model to be trained to obtain the predicted abnormal driving behavior information of the sample drivers; obtain the actual abnormal driving behavior information of the sample drivers, and iteratively train the multimodal driving behavior recognition model to be trained based on the difference between the predicted abnormal driving behavior information and the actual abnormal driving behavior information of the sample drivers to obtain the preset multimodal driving behavior recognition model.

[0076] Step S103: Based on driving style information and abnormal driving behavior information, determine the driving behavior risk value of the driver to be corrected.

[0077] Among them, the driving behavior risk value of the driver to be corrected is used to represent the overall degree of danger corresponding to the driving behavior of the driver to be corrected.

[0078] For example, the server uses driving style information as primary data and abnormal driving behavior information as secondary data, inputting them into a feature extraction model for feature extraction processing to obtain a first feature vector of driving style information. Next, the server uses abnormal driving behavior information as primary data and driving style information as secondary data, inputting them into the feature extraction model for feature extraction processing to obtain a second feature vector of abnormal driving behavior information. Then, the server concatenates the first and second feature vectors to obtain a concatenated feature vector. Next, the server inputs the concatenated feature vector into multiple driving behavior risk value prediction models to obtain multiple predicted driving behavior risk values ​​for the driver to be corrected. Finally, the server sums the multiple predicted driving behavior risk values ​​for the driver to be corrected according to the model weights of each driving behavior risk value prediction model to obtain the driving behavior risk value for the driver to be corrected.

[0079] Step S104: Generate a behavior correction training plan for the driver to be corrected based on driving style information, abnormal driving behavior information, and driving behavior risk value.

[0080] Among them, the behavior modification training program refers to a personalized driving behavior improvement plan for drivers undergoing correction.

[0081] For example, the server preprocesses driving style information, abnormal driving behavior information, and driving behavior risk value to obtain preprocessed driving style information, preprocessed abnormal driving behavior information, and preprocessed driving behavior risk value. Then, the server inputs the preprocessed driving style information into a behavior correction training scheme prediction model to obtain a first behavior correction training scheme for the driver to be corrected; inputs the preprocessed abnormal driving behavior information into the behavior correction training scheme prediction model to obtain a second behavior correction training scheme for the driver to be corrected; and inputs the preprocessed driving behavior risk value into the behavior correction training scheme prediction model to obtain a third behavior correction training scheme for the driver to be corrected. Finally, the server fuses the first, second, and third behavior correction training schemes to obtain the behavior correction training scheme for the driver to be corrected.

[0082] In the aforementioned method for generating a driver behavior correction training scheme based on artificial intelligence, the following steps are taken: First, historical operational behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment are acquired from the driver to be corrected. These are used as the driver's historical data. Then, based on the historical data, the driver's current driving style information is determined. The historical data is then input into a preset multimodal driving behavior recognition model to obtain the driver's abnormal driving behavior information. Next, based on the driving style information and the abnormal driving behavior information, the driver's driving behavior risk value is determined. Finally, based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, a behavior correction training scheme for the driver to be corrected is generated. In this way, when conducting driving behavior correction training for drivers, the individual driving style of the driver to be corrected is first characterized based on historical data. At the same time, a multimodal model is used to accurately identify specific abnormal driving behavior information. Combining the two methods, the driving behavior risk value of the driver to be corrected can be determined comprehensively and accurately. Based on the above information, a behavior correction training plan that is more suitable for the driver to be corrected can be generated, which helps to improve the accuracy of the determination of the behavior correction training plan, thereby improving the accuracy of the driver's behavior correction training. Moreover, the entire process does not require human intervention, avoiding the shortcomings of relying on human experience judgment, uniform training content, or post-intervention methods that lack a deep understanding of individual driving differences and are difficult to achieve personalized and targeted behavioral guidance, resulting in low accuracy of driver behavior correction training. This further improves the accuracy of driver behavior correction training.

[0083] In one exemplary embodiment, such as Figure 2 As shown, step S102 above determines the current driving style information of the driver to be corrected based on historical data, specifically including the following steps:

[0084] Step S201: Obtain the sample driving style vector corresponding to the driving style information of the sample driver.

[0085] Step S202: Cluster the sample driving style vectors to obtain the cluster centers corresponding to each driving style information.

[0086] Step S203: Based on historical data, construct the driving style vector of the driver to be corrected, and select the driving style information with the smallest distance between the corresponding cluster center and the driving style vector from each driving style information, as the current driving style information of the driver to be corrected.

[0087] Among them, the sample driver refers to the historical driver corresponding to the driver to be corrected.

[0088] Among them, the sample driving style vector refers to the multi-dimensional vector formed after quantifying the driving behavior characteristics of the sample drivers.

[0089] Here, cluster center refers to the feature vector of each driving style information obtained after clustering the sample driving style vectors.

[0090] Among them, the driving style vector refers to the multi-dimensional vector formed after quantifying the driving behavior characteristics of the driver to be corrected.

[0091] For example, the server identifies historical drivers corresponding to the driver to be corrected as sample drivers and obtains their driving style information. Next, the server inputs the sample driver's driving style information into a feature extraction model for feature extraction processing, obtaining a sample driving style vector corresponding to the sample driver's driving style information. Then, the server inputs the sample driving style vector into a clustering model, performing clustering processing on the sample driving style vector to obtain cluster centers corresponding to each driving style information. Next, the server constructs the driving style vector of the driver to be corrected based on historical data. Finally, the server selects the driving style information with the smallest distance between the corresponding cluster center and the driving style vector from all driving style information, and uses this driving style information as the current driving style information of the driver to be corrected.

[0092] In this embodiment, by using the cluster centers obtained from clustering sample driver data, a standardized and quantifiable reference system is provided for driving style, avoiding the subjectivity of traditional manual judgment. This allows for an objective match between the driving style and the driver's actual driving characteristics, providing a precise basis for designing personalized correction schemes for different styles.

[0093] In an exemplary embodiment, step S203 above, which constructs a driving style vector for the driver to be corrected based on historical data, specifically includes the following: determining a first data feature of the driver to be corrected based on historical operational behavior data; determining a second data feature of the driver to be corrected based on historical vehicle data; determining a third data feature of the driver to be corrected based on historical in-vehicle behavior video data; and determining a fourth data feature of the driver to be corrected based on historical environmental data; determining operational behavior index values ​​of the driver to be corrected based on the first data feature; determining state behavior index values ​​of the driver to be corrected based on the second data feature; determining in-vehicle behavior index values ​​of the driver to be corrected based on the third data feature; and determining environmental behavior index values ​​of the driver to be corrected based on the fourth data feature; and constructing a driving style vector for the driver to be corrected based on the operational behavior index values, state behavior index values, in-vehicle behavior index values, and environmental behavior index values.

[0094] The first data feature includes the normalized frequency of rapid acceleration (number of rapid acceleration events per unit mileage), the normalized frequency of emergency braking (number of emergency braking events per unit mileage), and the normalized number of rapid lane changes (number of rapid lane change events per unit mileage) of the driver to be corrected.

[0095] The second data feature includes the normalized average vehicle speed (the average speed during driving), the normalized standard deviation of vehicle speed (a quantitative indicator of the degree of vehicle speed fluctuation), and the normalized acceleration volatility (the degree of drastic change in acceleration) of the driver to be corrected.

[0096] The third data feature includes the normalized percentage of time the driver looks straight ahead (the proportion of time spent looking at the road ahead during driving) and the normalized frequency of gaze deviation (the ratio of the number of times the gaze deviates from straight ahead to the number of times per unit time).

[0097] The fourth data feature includes the normalized driving time distribution information and the normalized road condition complexity index of the drivers to be corrected.

[0098] Among them, the operational behavior index value is used to indicate the degree of aggression in the operational behavior of drivers to be corrected.

[0099] Among them, the state behavior index value is used to represent the stability of the vehicle operation state of the driver to be corrected.

[0100] Among them, the in-vehicle behavior index is used to represent the level of attention concentration of drivers who need correction while driving.

[0101] Among them, the environmental behavior index value is used to represent the adaptability of drivers to be corrected in different environments (such as nighttime, complex road conditions).

[0102] For example, based on historical operational behavior data, the server determines the frequency of sudden acceleration (e.g., the number of events within a unit mileage where the accelerator pedal depth is ≥80% or the increase within 1 second is ≥50% and lasts for more than 1 second), the frequency of sudden braking (e.g., the number of events within a unit mileage where the brake pedal depth is ≥70% or the increase within 1 second is ≥60% and lasts for more than 0.5 seconds), and the number of rapid lane changes (e.g., within a unit mileage where the steering wheel angle is ≥30° and the steering speed is ≥45° / s). The server first analyzes the number of events, and then normalizes the frequency of rapid acceleration, rapid braking, and rapid lane changes to obtain normalized rapid acceleration, rapid braking, and rapid lane change frequencies, which serve as the first data feature of the driver to be corrected. Next, based on historical vehicle data, the server determines the driver's average speed (e.g., total distance traveled on a road segment divided by total travel time), speed standard deviation (e.g., standard deviation of speed at all times within a road segment), and acceleration volatility (e.g., average absolute value of acceleration within a road segment). The server then analyzes the average speed, speed... The standard deviation and acceleration variability are normalized to obtain the normalized average vehicle speed, normalized vehicle speed standard deviation, and normalized acceleration variability, which serve as the second data feature of the driver to be corrected. Then, based on historical in-vehicle behavior video data, the server determines the proportion of time the driver gazes directly forward (e.g., duration of gazes directly forward divided by total driving time) and the frequency of gaze deviation (e.g., number of gaze deviations divided by total driving time), and normalizes the proportion of time the driver gazes directly forward and the frequency of gaze deviation to obtain the normalized proportion of time the driver gazes directly forward and the normalized acceleration variability. The frequency of gaze deviation after the initial gaze shift is used as the third data feature of the driver to be corrected. Next, based on historical environmental data, the server determines the driver's driving time distribution information (e.g., the proportion of nighttime driving time to total driving time) and road condition complexity index (e.g., average road condition level calculated by road segment). The server then normalizes the driving time distribution information and road condition complexity index to obtain normalized driving time distribution information and normalized road condition complexity index, which serve as the fourth data feature of the driver to be corrected. Finally, the server inputs the first data feature into the operational behavior index prediction model. The system obtains the operational behavior index values ​​of the driver to be corrected, inputs the second data feature into the state behavior index prediction model to obtain the state behavior index values ​​of the driver to be corrected, inputs the third data feature into the in-vehicle behavior index prediction model to obtain the in-vehicle behavior index values ​​of the driver to be corrected, and inputs the fourth data feature into the environmental behavior index prediction model to determine the environmental behavior index values ​​of the driver to be corrected. Then, the server constructs the driving style vector of the driver to be corrected according to the corresponding arrangement order of the operational behavior index values, state behavior index values, in-vehicle behavior index values, and environmental behavior index values.

[0103] The operational behavior index value is calculated using the following formula:

[0104] Equation (1)

[0105] In the formula, Indicates the value of operational behavior indicators. , and All represent weight parameters. This represents the normalized acceleration frequency. This represents the standardized emergency braking frequency. This represents the number of fast lane changes after standardization.

[0106] The state behavior index value is calculated using the following formula:

[0107] Equation (2)

[0108] In the formula, Indicates the state behavior index value, , and All represent weight parameters. This represents the standardized average vehicle speed. This represents the standard deviation of vehicle speed after standardization. This represents the standardized acceleration fluctuation.

[0109] The in-vehicle behavior index value is calculated using the following formula:

[0110] Equation (3)

[0111] In the formula, Indicates in-vehicle behavior index values, and All represent weight parameters. This represents the standardized gaze shift frequency. This represents the percentage of time spent looking directly ahead after standardization.

[0112] The environmental behavior index values ​​are calculated using the following formula:

[0113] Equation (4)

[0114] In the formula, Indicates the environmental behavior index value, and All represent weight parameters. This represents the standardized proportion of nighttime driving time. This represents the standardized road condition complexity index.

[0115] In this embodiment, features are extracted from four core dimensions: operation, vehicle status, attention, and environmental adaptation, and transformed into standardized index values. These values ​​are then combined into a driving style vector according to fixed dimensions. This approach preserves the detailed differences in driving features across each dimension while eliminating dimensional differences between different features through a unified quantification standard. This lays a standardized and quantifiable foundation for generating targeted correction schemes.

[0116] In an exemplary embodiment, step S103, which determines the driving behavior risk value of the driver to be corrected based on driving style information and abnormal driving behavior information, specifically includes the following: querying a preset driving style risk mapping table to obtain a first driving behavior risk value corresponding to the driving style information; the preset driving style risk mapping table is used to represent the mapping relationship between driving style information and driving behavior risk value; identifying the abnormal behavior type corresponding to the abnormal driving behavior information, determining the driving behavior risk value corresponding to each abnormal behavior type, and summing the driving behavior risk values ​​corresponding to each abnormal behavior type to obtain a second driving behavior risk value; and fusing the first driving behavior risk value and the second driving behavior risk value to obtain the driving behavior risk value of the driver to be corrected.

[0117] Among them, the first driving behavior risk value refers to the degree of danger corresponding to the driving behavior of the driver to be corrected, which is obtained based on driving style information.

[0118] Among them, the mapping relationship between driving style information and driving behavior risk value, for example, the driving behavior risk value corresponding to the aggressive type is 80, the driving behavior risk value corresponding to the distracted type is 70, and the driving behavior risk value corresponding to the steady type is 30.

[0119] Among them, abnormal behavior type refers to the specific category of illegal or dangerous behavior identified from abnormal driving behavior information, such as rapid acceleration, prolonged gaze deviation, nighttime speeding, and frequent sudden braking.

[0120] The second driving behavior risk value refers to the degree of danger corresponding to the driving behavior of the driver to be corrected, obtained based on abnormal driving behavior information.

[0121] For example, the server queries a preset driving style risk mapping table to obtain the driving behavior risk value corresponding to the driving style information, which is used as the first driving behavior risk value. Next, the server inputs the abnormal driving behavior information into the trained abnormal behavior type prediction model to obtain the predicted abnormal behavior type corresponding to the abnormal driving behavior information, as well as the predicted probability of each predicted abnormal behavior type. From each predicted abnormal behavior type, the server selects the predicted abnormal behavior type with a predicted probability greater than a preset probability, which is used as the abnormal behavior type corresponding to the abnormal driving behavior information. Then, the server queries the correspondence between the abnormal behavior type and the driving behavior risk value to obtain the driving behavior risk value corresponding to each abnormal behavior type. Next, the server sums the driving behavior risk values ​​corresponding to each abnormal behavior type according to the preset weights corresponding to each abnormal behavior type to obtain the second driving behavior risk value. Finally, the server sums the first driving behavior risk value and the second driving behavior risk value according to the first weight of the first driving behavior risk value and the second weight of the second driving behavior risk value to obtain the driving behavior risk value of the driver to be corrected.

[0122] In this embodiment, by fusing the first driving behavior risk value obtained based on driving style information and the second driving behavior risk value obtained based on abnormal driving behavior information, it avoids risk misjudgment caused by relying solely on style, and also prevents the one-sidedness of focusing only on a single abnormality while ignoring style-related hidden dangers. This allows the obtained driving behavior risk value to accurately and comprehensively measure the driver's overall driving hazard level.

[0123] In an exemplary embodiment, step S104 above, which generates a behavior correction training plan for the driver to be corrected based on driving style information, abnormal driving behavior information, and driving behavior risk value, specifically includes the following: determining the target driver with the highest similarity to the driver to be corrected from the sample drivers based on the driving style information, abnormal driving behavior information, and driving behavior risk value; obtaining the sample behavior correction training plan of the target driver, and determining the initial behavior correction training plan for the driver to be corrected based on the sample behavior correction training plan of the target driver; updating the initial behavior correction training plan based on the driving style information, abnormal driving behavior information, and driving behavior risk value to obtain the behavior correction training plan for the driver to be corrected.

[0124] The similarity score is used to represent the degree of matching between the driver to be corrected and the sample drivers.

[0125] The target driver refers to the sample driver selected from the sample drivers who has the highest similarity to the driver to be corrected.

[0126] Among them, the sample behavior correction training program refers to the behavior correction training program used by the target driver in the historical correction process.

[0127] The initial behavior modification training program refers to a preliminary behavior modification training program adapted to the drivers to be corrected.

[0128] For example, the server determines the similarity between the driver to be corrected and the sample drivers based on driving style information, abnormal driving behavior information, and driving behavior risk value, and identifies the target driver with the highest similarity to the driver to be corrected from the sample drivers. Next, the server retrieves the target driver's sample behavior correction training scheme from the database and uses this scheme as the initial behavior correction training scheme for the driver to be corrected. Then, the server updates the initial behavior correction training scheme based on the driving style information, abnormal driving behavior information, and driving behavior risk value to obtain the behavior of the driver to be corrected. Corrective training schemes; for example, if the driving style information is not the preset driving style information, or the abnormal driving behavior information is not the preset abnormal driving behavior information, or the driving behavior risk value is greater than the preset driving behavior risk value, the server updates the initial behavior correction training scheme to obtain the behavior correction training scheme for the driver to be corrected; if the driving style information is the preset driving style information, the abnormal driving behavior information is the preset abnormal driving behavior information, and the driving behavior risk value is less than or equal to the preset driving behavior risk value, the server uses the initial behavior correction training scheme as the behavior correction training scheme for the driver to be corrected.

[0129] In this embodiment, the target driver with the highest similarity is first matched from the sample drivers, and the validated sample scheme is reused as the initial scheme to ensure the basic effectiveness and generation efficiency of the scheme. Then, the specific driving style, abnormal behavior and risk value of the driver to be corrected are combined for targeted updates, which makes up for the adaptation gap between the sample scheme and individual differences, and significantly improves the effect and efficiency of driving behavior correction.

[0130] In an exemplary embodiment, based on driving style information, abnormal driving behavior information, and driving behavior risk value, a target driver with the highest similarity to the driver to be corrected is determined from the sample drivers. Specifically, this includes: obtaining sample driving style information, sample abnormal driving behavior information, and sample driving behavior risk value of the sample drivers; determining a first similarity between the driver to be corrected and the sample drivers based on the difference between the driving style information and the sample driving style information; determining a second similarity between the driver to be corrected and the sample drivers based on the difference between the abnormal driving behavior information and the sample abnormal driving behavior information; determining a third similarity between the driver to be corrected and the sample drivers based on the difference between the driving behavior risk value and the sample driving behavior risk value; fusing the first, second, and third similarities to obtain the total similarity between the driver to be corrected and the sample drivers; and selecting the sample driver with the highest similarity from all the sample drivers as the target driver corresponding to the driver to be corrected.

[0131] Among them, the sample driving style information refers to the driving style information of the sample drivers.

[0132] Among them, abnormal driving behavior information refers to abnormal driving behavior information of sample drivers.

[0133] Among them, the sample driving behavior risk value refers to the driving behavior risk value of the sample drivers.

[0134] The first similarity refers to the similarity obtained based on the difference between driving style information and sample driving style information.

[0135] The second similarity refers to the similarity obtained based on the difference between abnormal driving behavior information and sample abnormal driving behavior information.

[0136] The third similarity refers to the similarity obtained based on the difference between the driving behavior risk value and the sample driving behavior risk value.

[0137] For example, the server retrieves sample driving style information, sample abnormal driving behavior information, and sample driving behavior risk value of sample drivers from the database. Then, based on the difference between the driving style information and the sample driving style information, the server determines the similarity between the driver to be corrected and the sample driver as a first similarity. Based on the difference between the abnormal driving behavior information and the sample abnormal driving behavior information, the server determines the similarity between the driver to be corrected and the sample driver as a second similarity. Based on the difference between the driving behavior risk value and the sample driving behavior risk value, the server determines the similarity between the driver to be corrected and the sample driver as a third similarity. Then, the server sums the first similarity, second similarity, and third similarity according to a first preset weight for the first similarity, a second preset weight for the second similarity, and a third preset weight for the third similarity to obtain the similarity between the driver to be corrected and the sample driver. Finally, the server selects the sample driver with the highest similarity from all sample drivers and designates this sample driver as the target driver corresponding to the driver to be corrected.

[0138] In this embodiment, a comprehensive matching result is obtained by fusing three types of similarity, which avoids the bias caused by single-dimensional comparison and ensures that the selected target drivers are highly similar to the drivers to be corrected in multiple dimensions. As a result, the sample scheme is more valuable for reference and provides a reliable reference basis for generating a highly adaptable initial correction scheme, thereby improving the efficiency and accuracy of scheme generation.

[0139] In one exemplary embodiment, such as Figure 3 As shown, another method for generating driver behavior correction training schemes based on artificial intelligence is provided. Taking the application of this method to a server as an example, the specific steps include:

[0140] Step S301: Obtain historical data of the driver to be corrected; the historical data includes at least the driver's historical operation behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment.

[0141] Step S302: Obtain the sample driving style vector corresponding to the driving style information of the sample driver; perform clustering processing on the sample driving style vector to obtain the cluster center corresponding to each driving style information.

[0142] Step S303: Based on historical operational behavior data, determine the first data feature of the driver to be corrected; based on historical vehicle data, determine the second data feature of the driver to be corrected; based on historical in-vehicle behavior video data, determine the third data feature of the driver to be corrected; and based on historical environmental data, determine the fourth data feature of the driver to be corrected.

[0143] Step S304: Based on the first data feature, determine the operational behavior index value of the driver to be corrected; based on the second data feature, determine the state behavior index value of the driver to be corrected; based on the third data feature, determine the in-vehicle behavior index value of the driver to be corrected; and based on the fourth data feature, determine the environmental behavior index value of the driver to be corrected.

[0144] Step S305: Construct a driving style vector for the driver to be corrected based on the operational behavior index value, state behavior index value, in-vehicle behavior index value, and environmental behavior index value.

[0145] Step S306: From each driving style information, select the driving style information with the smallest distance between the corresponding cluster center and the driving style vector, and use it as the current driving style information of the driver to be corrected.

[0146] Step S307: Input historical data into a preset multimodal driving behavior recognition model to obtain abnormal driving behavior information of the driver to be corrected.

[0147] Step S308: Query the preset driving style risk mapping table to obtain the first driving behavior risk value corresponding to the driving style information; the preset driving style risk mapping table is used to represent the mapping relationship between driving style information and driving behavior risk value.

[0148] Step S309: Identify the abnormal behavior type corresponding to the abnormal driving behavior information, determine the driving behavior risk value corresponding to each abnormal behavior type, and sum the driving behavior risk values ​​corresponding to each abnormal behavior type to obtain the second driving behavior risk value.

[0149] Step S310: The first driving behavior risk value and the second driving behavior risk value are fused to obtain the driving behavior risk value of the driver to be corrected.

[0150] Step S311: Obtain sample driving style information, sample abnormal driving behavior information, and sample driving behavior risk value of the sample drivers.

[0151] Step S312: Based on the difference between driving style information and sample driving style information, determine the first similarity between the driver to be corrected and the sample driver; based on the difference between abnormal driving behavior information and sample abnormal driving behavior information, determine the second similarity between the driver to be corrected and the sample driver; and based on the difference between driving behavior risk value and sample driving behavior risk value, determine the third similarity between the driver to be corrected and the sample driver.

[0152] Step S313: The first similarity, the second similarity and the third similarity are fused to obtain the similarity between the driver to be corrected and the sample drivers; from each sample driver, the sample driver with the highest similarity is selected as the target driver corresponding to the driver to be corrected.

[0153] Step S314: Obtain the sample behavior correction training plan of the target driver, and determine the initial behavior correction training plan of the driver to be corrected based on the sample behavior correction training plan of the target driver; update the initial behavior correction training plan according to driving style information, abnormal driving behavior information and driving behavior risk value to obtain the behavior correction training plan of the driver to be corrected.

[0154] In the aforementioned AI-based driver behavior correction training program generation method, when conducting driver behavior correction training, the method first characterizes the driver's personalized driving style based on historical data, and then uses a multimodal model to accurately identify specific abnormal driving behavior information. Combining these two methods allows for a comprehensive and accurate determination of the driver's driving behavior risk value. By integrating this information, a behavior correction training program that is more suitable for the driver can be generated, which helps improve the accuracy of determining the behavior correction training program and thus improves the accuracy of driver behavior correction training. Moreover, the entire process does not require human intervention, avoiding the shortcomings of relying on human experience judgment, uniform training content, or post-intervention methods that lack a deep understanding of individual driver differences and are difficult to achieve personalized and targeted behavioral guidance, resulting in low accuracy of driver behavior correction training. This further improves the accuracy of driver behavior correction training.

[0155] In an exemplary embodiment, to more clearly illustrate the method for generating an AI-based driver behavior correction training scheme provided in this application, the following specific embodiment will be used to describe the method. In one embodiment, this application also provides an AI-based driver behavior correction training method. Specifically, it includes the following:

[0156] Step 1: Obtain the historical driving behavior data of the driver to be corrected.

[0157] Historical driving behavior data includes historical operational behavior data, historical vehicle data, historical video data, and historical environmental data. Historical operational behavior data includes accelerator pedal depth, brake pedal depth, steering wheel angle, gear status, and number of braking actions. Historical vehicle data includes vehicle speed, acceleration, and position. Historical video behavior data includes forward road video and driver in-vehicle behavior video.

[0158] Step 2: Analyze the behavioral characteristics based on historical driving behavior data to obtain the driving style of the driver to be corrected.

[0159] Calculate the frequency of rapid acceleration, rapid braking, and rapid lane changes based on historical operational behavior data; calculate the average vehicle speed, vehicle speed standard deviation, and acceleration volatility based on historical vehicle data; perform behavior detection on in-vehicle behavior videos to obtain the proportion of time spent looking straight ahead and the frequency of gaze deviation; and calculate the driving time distribution and road condition complexity index based on historical environmental data.

[0160] Based on the above features, operational behavior indicators, state behavior indicators, in-vehicle behavior indicators, and environmental behavior indicators are calculated. Then, a driving style vector of the driver to be corrected is constructed using these indicators.

[0161] Then calculate the distance between the driving style vector and each cluster center, and select the driving style corresponding to the cluster center with the smallest distance to the driving style vector as the driving style of the driver to be corrected.

[0162] The calculation formula for the operational behavior indicators is as follows:

[0163] Equation (1)

[0164] In the formula, Indicators representing operational behavior. , and All represent weight parameters. This represents the normalized acceleration frequency. This represents the standardized emergency braking frequency. This represents the number of standardized rapid lane changes;

[0165] The formula for calculating the state behavior index is:

[0166] Equation (2)

[0167] In the formula, Indicators representing state behavior , and All represent weight parameters. This represents the standardized average vehicle speed. This represents the standard deviation of vehicle speed after standardization. This represents the standardized acceleration fluctuations;

[0168] The formula for calculating in-vehicle behavior indicators is:

[0169] Equation (3)

[0170] In the formula, Indicators of in-vehicle behavior and All represent weight parameters. This represents the standardized gaze shift frequency. This represents the standardized percentage of time spent looking directly ahead.

[0171] The formula for calculating environmental behavior indicators is:

[0172] Equation (4)

[0173] In the formula, Indicators of environmental behavior and All represent weight parameters. This represents the standardized proportion of nighttime driving time. This represents the standardized road condition complexity index.

[0174] Step 3: Input historical driving behavior data into a preset multimodal behavior recognition model for behavior recognition to obtain the abnormal behavior of the driver to be corrected. The abnormal behavior includes the type, frequency and duration of the abnormal behavior.

[0175] Step 4: Calculate the hazard risk score of the driver to be corrected based on driving style and abnormal behavior.

[0176] Based on driving style and a preset driving style risk mapping table, the driving style risk score of the driver to be corrected is obtained. Then, the abnormal risk area of ​​each type of abnormal behavior is calculated by considering the frequency, duration, and weight of the abnormal behavior.

[0177] The abnormal risk areas of all abnormal behaviors are summed to obtain the abnormal behavior risk value. Then, the abnormal behavior risk value is standardized to obtain the abnormal behavior risk score.

[0178] The driving style risk score and abnormal behavior risk score are weighted and fused to obtain the danger risk score of the driver to be corrected.

[0179] Step 5: Generate a personalized correction training plan based on driving style, abnormal behavior, and hazard risk scores.

[0180] The similarity between the driver to be corrected and each corrected driver is calculated based on the driving style, abnormal behavior, and risk scores of the driver to be corrected and the driving style, abnormal behavior, and risk scores of the trained drivers.

[0181] The personalized correction training plan of the trained driver with the highest similarity was selected as the initial plan for the driver to be corrected. Then, the driver status was constructed based on the driver's driving style, abnormal behavior, and hazard risk score. Finally, the initial plan was optimized based on the driver status to obtain the personalized correction training plan for the driver to be corrected.

[0182] The process involves selecting an action based on a chosen strategy and the driver's state, updating the initial plan based on the selected action, and obtaining an optimized plan. The optimized plan is then used to simulate training the driver to be corrected, updating the driver's state. A reward value is calculated based on the updated driver state, and the chosen strategy is updated using the reward value. Finally, it is determined whether the iteration conditions have been met. If so, the optimized plan is used as the personalized correction training plan for the driver to be corrected; otherwise, the optimized plan is used as the initial plan for the next iteration, and the next iteration is conducted based on the updated chosen strategy and the driver's state.

[0183] The formula for calculating the reward value is:

[0184] Equation (5)

[0185] Equation (6)

[0186] In the formula, Indicates the reward value. Indicates the rate of decrease in the number of abnormal behaviors. This indicates the percentage reduction in the duration of abnormal behavior. Indicates the magnitude of the decrease in risk score. Indicates driving style rewards. Represents Euclidean distance. This represents the driving style vector before correction. This represents the corrected driving style vector. Represents the ideal driving style vector. , , and All of these represent reward weight parameters.

[0187] Step 6: Construct a virtual driving simulation scenario through a personalized correction training program, and conduct behavioral correction training for the driver to be corrected through the virtual driving simulation scenario.

[0188] Based on the personalized correction training plan, the corresponding environment is configured in the virtual driving simulation platform to construct a virtual driving training scenario that suits the driver to be corrected.

[0189] In the above embodiments, when conducting driving behavior correction training for drivers, the individual driving style of the driver to be corrected is first characterized based on historical data. Simultaneously, a multimodal model is used to accurately identify specific abnormal driving behavior information. Combining these two methods allows for a comprehensive and accurate determination of the driver's driving behavior risk value. This information, combined with the driver's own experience, generates a more suitable behavior correction training plan, improving the accuracy of the training plan and consequently the accuracy of the driver's behavior correction training. Furthermore, the entire process requires no human intervention, avoiding the shortcomings of relying on human experience, standardized training content, or post-intervention methods that lack a deep understanding of individual driving differences, making it difficult to achieve personalized and targeted behavioral guidance and resulting in low accuracy in driver behavior correction training. This further improves the accuracy of driver behavior correction training. Simultaneously, by extracting driving behavior features from multimodal data, combining driving style and abnormal behavior, a hazard risk score is calculated, and a personalized training plan is generated. Differentiated corrections are implemented for different driving styles such as aggressive, distracted, and cautious, effectively enhancing the targeting and effectiveness of the intervention. Furthermore, by introducing a reinforcement learning mechanism, training actions are selected based on the driver's state, and rewards are calculated according to the degree of improvement in abnormal behaviors and changes in driving style, dynamic optimization of the training program is achieved. The training content covers multiple dimensions such as attention concentration, smooth operation, and lane change judgment, and the training methods are flexible. Therefore, it can not only reduce the frequency of risky behaviors by drivers, but also guide drivers' driving behavior towards an ideal style, and is applicable to the correction of drivers' non-standard operations and the improvement of their abilities.

[0190] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0191] Based on the same inventive concept, this application also provides an AI-based driver behavior correction training scheme generation device for implementing the aforementioned AI-based driver behavior correction training scheme generation method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more AI-based driver behavior correction training scheme generation device embodiments provided below can be found in the limitations of the AI-based driver behavior correction training scheme generation method described above, and will not be repeated here.

[0192] In one exemplary embodiment, such as Figure 4 As shown, an artificial intelligence-based driver behavior correction training scheme generation device is provided, including: a data acquisition module 401, a data processing module 402, a risk determination module 403, and a scheme generation module 404, wherein:

[0193] The data acquisition module 401 is used to acquire the historical data of the driver to be corrected; the historical data includes at least the historical operation behavior data of the driver to be corrected, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment.

[0194] The data processing module 402 is used to determine the current driving style information of the driver to be corrected based on historical data, and to input the historical data into a preset multimodal driving behavior recognition model to obtain abnormal driving behavior information of the driver to be corrected.

[0195] The risk assessment module 403 is used to determine the risk value of the driver's driving behavior to be corrected based on driving style information and abnormal driving behavior information.

[0196] The scheme generation module 404 is used to generate a behavior correction training scheme for the driver to be corrected based on driving style information, abnormal driving behavior information and driving behavior risk value.

[0197] In an exemplary embodiment, the data processing module 402 is further configured to obtain the sample driving style vector corresponding to the driving style information of the sample driver; perform clustering processing on the sample driving style vector to obtain the cluster center corresponding to each driving style information; construct the driving style vector of the driver to be corrected based on historical data, and select the driving style information with the smallest distance between the corresponding cluster center and the driving style vector from each driving style information as the current driving style information of the driver to be corrected.

[0198] In an exemplary embodiment, the data processing module 402 is further configured to: determine a first data feature of the driver to be corrected based on historical operational behavior data; determine a second data feature of the driver to be corrected based on historical vehicle data; determine a third data feature of the driver to be corrected based on historical in-vehicle behavior video data; and determine a fourth data feature of the driver to be corrected based on historical environmental data; determine operational behavior index values ​​of the driver to be corrected based on the first data feature; determine state behavior index values ​​of the driver to be corrected based on the second data feature; determine in-vehicle behavior index values ​​of the driver to be corrected based on the third data feature; and determine environmental behavior index values ​​of the driver to be corrected based on the fourth data feature; and construct a driving style vector of the driver to be corrected based on the operational behavior index values, state behavior index values, in-vehicle behavior index values, and environmental behavior index values.

[0199] In an exemplary embodiment, the risk determination module 403 is further configured to query a preset driving style risk mapping table to obtain a first driving behavior risk value corresponding to the driving style information; the preset driving style risk mapping table is used to represent the mapping relationship between driving style information and driving behavior risk value; identify the abnormal behavior type corresponding to the abnormal driving behavior information, determine the driving behavior risk value corresponding to each abnormal behavior type, and sum the driving behavior risk values ​​corresponding to each abnormal behavior type to obtain a second driving behavior risk value; and perform fusion processing on the first driving behavior risk value and the second driving behavior risk value to obtain the driving behavior risk value of the driver to be corrected.

[0200] In an exemplary embodiment, the scheme generation module 404 is further configured to: identify the target driver with the highest similarity to the driver to be corrected from the sample drivers based on driving style information, abnormal driving behavior information, and driving behavior risk value; obtain the sample behavior correction training scheme of the target driver; and determine the initial behavior correction training scheme of the driver to be corrected based on the sample behavior correction training scheme of the target driver; and update the initial behavior correction training scheme based on the driving style information, abnormal driving behavior information, and driving behavior risk value to obtain the behavior correction training scheme of the driver to be corrected.

[0201] In an exemplary embodiment, the scheme generation module 404 is further configured to acquire sample driving style information, sample abnormal driving behavior information, and sample driving behavior risk value of the sample driver; determine a first similarity between the driver to be corrected and the sample driver based on the difference between the driving style information and the sample driving style information; determine a second similarity between the driver to be corrected and the sample driver based on the difference between the abnormal driving behavior information and the sample abnormal driving behavior information; and determine a third similarity between the driver to be corrected and the sample driver based on the difference between the driving behavior risk value and the sample driving behavior risk value; fuse the first similarity, the second similarity, and the third similarity to obtain the similarity between the driver to be corrected and the sample driver; and select the sample driver with the highest similarity from among the sample drivers as the target driver corresponding to the driver to be corrected.

[0202] The modules in the aforementioned AI-based driver behavior correction training program generation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0203] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores historical operational behavior data, historical vehicle data, and historical in-vehicle behavior video data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an artificial intelligence-based driver behavior correction training scheme generation method.

[0204] Those skilled in the art will understand that Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0205] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0206] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above-described method embodiments.

[0207] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0208] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0209] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0210] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for generating a driver driving behavior correction training program based on artificial intelligence, characterized in that, The method includes: Obtain historical data of the driver to be corrected; the historical data includes at least the driver's historical operational behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment; Based on the historical data, the current driving style information of the driver to be corrected is determined, and the historical data is input into a preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected. Based on the driving style information and the abnormal driving behavior information, the driving behavior risk value of the driver to be corrected is determined; Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, a behavior correction training plan is generated for the driver to be corrected.

2. The method according to claim 1, characterized in that, The step of determining the current driving style information of the driver to be corrected based on the historical data includes: Obtain the sample driving style vector corresponding to the driving style information of the sample drivers; Clustering is performed on the sample driving style vectors to obtain the cluster centers corresponding to each driving style information; Based on the historical data, a driving style vector of the driver to be corrected is constructed, and the driving style information with the smallest distance between the corresponding cluster center and the driving style vector is selected from the driving style information and used as the current driving style information of the driver to be corrected.

3. The method according to claim 1, characterized in that, The step of constructing the driving style vector of the driver to be corrected based on the historical data includes: Based on the historical operational behavior data, the first data feature of the driver to be corrected is determined; based on the historical vehicle data, the second data feature of the driver to be corrected is determined; based on the historical in-vehicle behavior video data, the third data feature of the driver to be corrected is determined; and based on the historical environmental data, the fourth data feature of the driver to be corrected is determined. Based on the first data feature, the operational behavior index value of the driver to be corrected is determined; based on the second data feature, the state behavior index value of the driver to be corrected is determined; based on the third data feature, the in-vehicle behavior index value of the driver to be corrected is determined; and based on the fourth data feature, the environmental behavior index value of the driver to be corrected is determined. Based on the operational behavior index value, the state behavior index value, the in-vehicle behavior index value, and the environmental behavior index value, a driving style vector of the driver to be corrected is constructed.

4. The method according to claim 1, characterized in that, The step of determining the driving behavior risk value of the driver to be corrected based on the driving style information and the abnormal driving behavior information includes: A preset driving style risk mapping table is queried to obtain the first driving behavior risk value corresponding to the driving style information; the preset driving style risk mapping table is used to represent the mapping relationship between driving style information and driving behavior risk value. Identify the abnormal behavior type corresponding to the abnormal driving behavior information, determine the driving behavior risk value corresponding to each abnormal behavior type, and sum the driving behavior risk values ​​corresponding to each abnormal behavior type to obtain a second driving behavior risk value. The first driving behavior risk value and the second driving behavior risk value are fused together to obtain the driving behavior risk value of the driver to be corrected.

5. The method according to any one of claims 1 to 4, characterized in that, The step of generating a behavior correction training plan for the driver to be corrected based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value includes: Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, the target driver with the greatest similarity to the driver to be corrected is determined from the sample drivers; Obtain the sample behavior correction training scheme of the target driver, and determine the initial behavior correction training scheme of the driver to be corrected based on the sample behavior correction training scheme of the target driver. Based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value, the initial behavior correction training program is updated to obtain the behavior correction training program for the driver to be corrected.

6. The method according to claim 5, characterized in that, The step of identifying the target driver with the highest similarity to the driver to be corrected from the sample drivers based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value includes: Obtain sample driving style information, sample abnormal driving behavior information, and sample driving behavior risk value of sample drivers; Based on the difference between the driving style information and the sample driving style information, a first similarity between the driver to be corrected and the sample driver is determined; based on the difference between the abnormal driving behavior information and the sample abnormal driving behavior information, a second similarity between the driver to be corrected and the sample driver is determined; and based on the difference between the driving behavior risk value and the sample driving behavior risk value, a third similarity between the driver to be corrected and the sample driver is determined. The first similarity, the second similarity, and the third similarity are fused to obtain the similarity between the driver to be corrected and the sample driver; From the sample drivers, the sample driver with the highest similarity is selected as the target driver corresponding to the driver to be corrected.

7. A driver driving behavior correction training program generation device based on artificial intelligence, characterized in that, The device includes: The data acquisition module is used to acquire historical data of the driver to be corrected; the historical data includes at least the driver's historical operating behavior data, historical vehicle data, historical in-vehicle behavior video data, and historical environmental data of the vehicle's surrounding environment; The data processing module is used to determine the current driving style information of the driver to be corrected based on the historical data, and to input the historical data into a preset multimodal driving behavior recognition model to obtain the abnormal driving behavior information of the driver to be corrected. The risk determination module is used to determine the driving behavior risk value of the driver to be corrected based on the driving style information and the abnormal driving behavior information. The scheme generation module is used to generate a behavior correction training scheme for the driver to be corrected based on the driving style information, the abnormal driving behavior information, and the driving behavior risk value.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images