Data communication system, center device, on-vehicle device, data processing method, and data processing program

Abstract

A center device (2) is configured to: create trigger conditions for detecting a predetermined scene on a vehicle side; create a machine learning model; create a collection data list corresponding to the trigger conditions; distribute the trigger conditions, the machine learning model, and the collection data list to an on-vehicle device; and acquire collection data indicated by the collection data list from the on-vehicle device. The on-vehicle device (4) performs trigger condition determination on the basis of an execution result of the machine learning model based on detection data, creates collection data on the basis of the determination result of the trigger condition determination, and transmits the collection data to the center device.

Description

Data communication system, center device, vehicle device, data processing method, and data processing program CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application is based on Japanese Application No. 2024-013006, filed on January 31, 2024, the contents of which are incorporated herein by reference.

[0002] The present disclosure relates to a data communication system, a center device, a vehicle device, a data processing method, and a data processing program.

[0003] A data communication system is provided in which a center device receives vehicle data transmitted from vehicle devices mounted in an unspecified number of vehicles, thereby collecting vehicle data from the vehicle devices. For example, Patent Document 1 discloses a technology for reducing the frequency of transmission of access requests from the vehicle devices to a center device and reducing the processing load of authentication for collecting vehicle data.

[0004] JP 2023-084379 A

[0005] Users may experience dissatisfaction in certain situations when using a vehicle. Techniques for identifying user dissatisfaction can be envisioned to directly or indirectly identify the user's dissatisfaction. For example, when a user performs a steering operation while a lane tracing assist (hereinafter referred to as LTA) function is running while driving on a highway, the user's dissatisfaction can be directly identified by identifying the cancellation of the LTA function. For example, when the use of an autoparking function is recommended when starting parking, the user's dissatisfaction can be directly identified by identifying the user's refusal to use the autoparking function. Furthermore, after the LTA function is improved at the request of another user, the user's dissatisfaction can be indirectly identified by identifying a difference between the behavior of the logic during operation of the LTA function and the behavior of the logic after the improvement. For these reasons, vehicle manufacturers and app developers, for example, desire to analyze situations in which users experience dissatisfaction and improve vehicle functions.

[0006] In this case, the center device creates a trigger condition for detecting a scene in which the user is dissatisfied, and collects and analyzes the sensor values ​​of each sensor and the control status of each system as collected data from the vehicle device before and after the trigger condition is met. When detecting a scene in which the user is dissatisfied, it is considered relatively easy to detect dissatisfaction triggered by a direct or one-off user operation, such as a switch operation. However, there is a demand for detecting latent user dissatisfaction that is not manifested in a direct or one-off user operation. However, when attempting to detect dissatisfaction from a source other than a direct or one-off user operation, it becomes complicated to set a trigger condition that can accurately detect user dissatisfaction. Therefore, setting a trigger condition that can accurately detect user dissatisfaction is a challenge.

[0007] The present disclosure aims to appropriately set trigger conditions that can accurately detect user dissatisfaction, and accurately detect potential user dissatisfaction.

[0008] According to one aspect of the present disclosure, a data communication system is capable of data communication between a center device and a vehicle device mounted on a vehicle. The center device includes: a trigger condition creation unit that creates trigger conditions for detecting a predetermined scene in the vehicle; a machine learning model creation unit that creates a machine learning model; a collected data list creation unit that creates a collected data list corresponding to the trigger conditions; a distribution unit that distributes the trigger conditions, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition unit that acquires collected data indicated in the collected data list from the vehicle device. The vehicle device includes: a detection data acquisition unit that acquires detection data necessary for trigger condition determination; a trigger condition determination unit that performs trigger condition determination based on a result of execution of the machine learning model based on the detection data; a collected data creation unit that creates the collected data based on a result of the trigger condition determination; and a collected data transmission unit that transmits the collected data to the center device.

[0009] According to one aspect of the present disclosure, a center device is mounted on a vehicle, performs a trigger condition determination based on the execution result of a machine learning model based on detection data required for trigger condition determination, and is capable of data communication with a vehicle device that creates collected data based on the determination result of the trigger condition determination and transmits the collected data to the center device. The center device includes: a trigger condition creation unit that creates trigger conditions for detecting a predetermined scene on the vehicle side; a machine learning model creation unit that creates a machine learning model; a collected data list creation unit that creates a collected data list corresponding to the trigger conditions; a distribution unit that distributes the trigger conditions, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition unit that acquires the collected data indicated in the collected data list from the vehicle device.

[0010] According to one aspect of the present disclosure, a vehicle device is equipped in a vehicle, and the vehicle device creates trigger conditions for detecting a predetermined scene in the vehicle, creates a machine learning model, creates a list of collected data corresponding to the trigger conditions, distributes the trigger conditions, the machine learning model, and the list of collected data to the vehicle device, and is capable of data communication with a center device that acquires collected data from the vehicle device. The vehicle device includes: a detection data acquisition unit that acquires detection data necessary for trigger condition determination; a trigger condition determination unit that performs trigger condition determination based on a result of execution of the machine learning model based on the detection data; a collection data creation unit that creates the collected data based on a determination result of the trigger condition determination; and a collection data transmission unit that transmits the collected data to the center device.

[0011] According to a data processing method for a data communication system of one aspect of the present disclosure, the center device executes a trigger condition creation procedure for creating a trigger condition for detecting a predetermined scene in a vehicle, a machine learning model creation procedure for creating a machine learning model, a collected data list creation procedure for creating a collected data list corresponding to the trigger condition, a distribution procedure for distributing the trigger condition, the machine learning model, and the collected data list to the vehicle device, and a collected data acquisition procedure for acquiring collected data indicated in the collected data list from the vehicle device.The vehicle device executes a detection data acquisition procedure for acquiring detection data necessary for trigger condition determination, a trigger condition determination procedure for determining a trigger condition based on a result of execution of the machine learning model based on the detection data, a collected data creation procedure for creating the collected data based on a result of the trigger condition determination, and a collected data transmission procedure for transmitting the collected data to the center device.

[0012] According to one aspect of the data processing method for a center device of the present disclosure, the center device is capable of data communication with a vehicle device that is mounted on a vehicle and performs a trigger condition determination based on the execution result of a machine learning model based on detection data required for trigger condition determination, and creates collected data based on the determination result of the trigger condition determination and transmits it to the center device.The center device executes a trigger condition creation procedure that creates trigger conditions for detecting a specified scene on the vehicle side, a machine learning model creation procedure that creates a machine learning model, a collected data list creation procedure that creates a collected data list corresponding to the trigger conditions, a distribution procedure that distributes the trigger conditions, the machine learning model, and the collected data list to the vehicle device, and a collected data acquisition procedure that acquires the collected data indicated by the collected data list from the vehicle device.

[0013] According to one aspect of the data processing method for a vehicle device of the present disclosure, a trigger condition for detecting a specified scene in the vehicle is created, a machine learning model is created, a list of collected data corresponding to the trigger condition is created, the trigger condition, the machine learning model, and the list of collected data are distributed to the vehicle device, and data communication with a center device that acquires collected data from the vehicle device is possible, and the vehicle device mounted on the vehicle executes a detection data acquisition procedure for acquiring detection data necessary for trigger condition determination, a trigger condition determination procedure for making a trigger condition determination based on the execution result of the machine learning model based on the detection data, a collected data creation procedure for creating the collected data based on the determination result of the trigger condition determination, and a collected data transmission procedure for transmitting the collected data to the center device.

[0014] According to one aspect of the data processing program of the center device of the present disclosure, the control unit of the center device, which is capable of data communication with a vehicle device that is mounted on a vehicle and performs a trigger condition determination based on the execution result of a machine learning model based on detection data required for trigger condition determination, and creates collected data based on the determination result of the trigger condition determination and transmits it to the center device, executes a trigger condition creation procedure for creating trigger conditions for detecting a specified scene on the vehicle side, a machine learning model creation procedure for creating a machine learning model, a collected data list creation procedure for creating a collected data list corresponding to the trigger conditions, a distribution procedure for distributing the trigger conditions, the machine learning model, and the collected data list to the vehicle device, and a collected data acquisition procedure for acquiring the collected data indicated in the collected data list from the vehicle device.

[0015] According to a data processing program for a vehicle device of one aspect of the present disclosure, a trigger condition for detecting a specified scene in the vehicle is created, a machine learning model is created, a list of collected data corresponding to the trigger condition is created, the trigger condition, the machine learning model, and the list of collected data are distributed to the vehicle device, and data communication with a center device that acquires collected data from the vehicle device is possible, and the control unit of the vehicle device mounted on the vehicle executes a detection data acquisition procedure for acquiring detection data necessary for trigger condition determination, a trigger condition determination procedure for making a trigger condition determination based on the execution result of the machine learning model based on the detection data, a collection data creation procedure for creating the collection data based on the determination result of the trigger condition determination, and a collection data transmission procedure for transmitting the collection data to the center device.

[0016] According to the above disclosure, the center device distributes trigger conditions, machine learning models, and a list of collected data to the vehicle device, and the vehicle device determines the trigger conditions based on the execution results of the machine learning model based on the detection data, and creates collected data based on the determination results of the trigger condition determination and transmits the collected data to the center device. It is possible to appropriately set trigger conditions that can accurately detect user dissatisfaction, and to accurately detect the user's latent dissatisfaction. Furthermore, by inputting data into the machine learning model indicating whether the trigger conditions captured the user's feelings of dissatisfaction, i.e., whether the trigger conditions were correct or incorrect, it is possible to train and evaluate the accuracy of the machine learning model of the trigger conditions, and the accuracy of the trigger conditions can be improved by evolving the machine learning model.

[0017] The above and other objects, features, and advantages of the present disclosure will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which Fig. 1 is a functional block diagram showing the overall configuration of an embodiment, Fig. 2 is a functional block diagram of a center device, Fig. 3 is a functional block diagram of a vehicle device, Fig. 4 is a functional block diagram of a control unit of the center device, Fig. 5 is a diagram explaining a trigger condition file, Fig. 6 is a diagram explaining a collected data list file, Fig. 7 is a diagram explaining the trigger condition file, Fig. 8 is a diagram explaining a collected data list file, Fig. 9 is a functional block diagram of a control unit of the vehicle device, Fig. 10 is a flowchart showing processing of the vehicle device, Fig. 11 is a flowchart showing processing of the center device, Fig. 12 is a diagram explaining the processing flow, Fig. 13 is a diagram explaining the processing flow,

[0018] The present invention relates to a data communication system for a vehicle, a data communication method for a vehicle, and a data communication system for a vehicle.

[0019] As shown in FIG. 2 , the center device 2 includes a control unit 6, a communication unit 7, and a storage unit 8. The control unit 6 is mainly configured with a microcomputer (hereinafter referred to as "microcomputer") having a CPU, ROM, RAM, I / O, etc., and controls the operation of the center device 2 by executing software processing by the CPU running a computer program stored in a non-transitory tangible storage medium, and hardware processing by a dedicated electronic circuit. By executing the computer program, a method corresponding to the computer program is performed. The control unit 6 may include one or more microcomputers. The communication unit 7 controls data communication between the multiple vehicle devices 4 via the wide-area wireless communication network 5. The storage unit 8 stores various data.

[0020] As shown in FIG. 3 , the vehicle device 4 includes a control unit 9, a communication unit 10, a storage unit 11, and a CAN (Controller Area Network) communication unit 12. The control unit 9 is mainly configured with a microcomputer having a CPU, ROM, RAM, I / O, etc., and controls the operation of the vehicle device 4 by executing software processing by the CPU running a computer program stored in a non-transient tangible storage medium and hardware processing by a dedicated electronic circuit. By executing the computer program, a method corresponding to the computer program is performed. The control unit 9 may include one or more microcomputers. The communication unit 10 controls data communication with the center device 2 via the wide-area wireless communication network 5. The storage unit 11 stores various data.

[0021] The CAN communication unit 12 is connected to CAN buses 13 to 15. The CAN communication unit 12 is connected to a plurality of ECUs 161 to 16n via the CAN bus 13 so as to be able to communicate data with them. The plurality of ECUs 161 to 16n include, for example, an engine ECU that controls the engine, an accelerator ECU that controls the accelerator, a brake ECU that controls the brakes, a steering ECU that controls the steering, a meter ECU that controls the meters, and a navigation ECU that controls navigation. For example, the accelerator ECU transmits data indicating the accelerator opening to the CAN communication unit 12 as detected data. For example, the brake ECU transmits data indicating the brake pedal operation amount to the CAN communication unit 12 as detected data. For example, the steering ECU transmits data indicating the steering angular velocity of the steering wheel to the CAN communication unit 12 as detected data. For example, the navigation ECU transmits data indicating latitude and longitude to the CAN communication unit 12 as detected data. The plurality of ECUs 161 to 16n are connected to the same bus for each system. For example, an ADAS (Advanced Driving Assistant System) ECU, a body ECU, a chassis ECU, etc. are all connected to the same bus.

[0022] The CAN communication unit 12 is connected to a plurality of cameras 171-17n and a plurality of sensors 181-18n via the CAN bus 14 so as to be able to communicate data with them. The plurality of cameras 171-17n may be, for example, a forward-facing camera that captures images of the area ahead of the vehicle, an interior-facing camera that captures images of the interior of the vehicle, a driver-facing camera that captures images of the driver, etc. For example, a forward-facing camera that captures images of the area ahead of the vehicle transmits data indicating an image of the area ahead of the vehicle as detection data to the CAN communication unit 12. For example, an interior-facing camera transmits data indicating an image of the interior of the vehicle as detection data to the CAN communication unit 12. The plurality of sensors 181-18n may be, for example, a light detection and ranging (LIDAR), a millimeter-wave radar, an ultrasonic sensor, an illuminance sensor, a rain sensor, etc. For example, a sensor that detects the area ahead of the vehicle using LIDAR, millimeter-wave radar, etc. transmits data indicating the distance from a leading vehicle to the CAN communication unit 12 as detection data. For example, an illuminance sensor transmits data indicating illuminance as detected data to the CAN communication unit 12. For example, a rain sensor transmits data indicating the amount of rainfall as detected data to the CAN communication unit 12.

[0023] The CAN communication unit 12 is connected to a DSM (Driver Status Monitor) 19, an ACC (Adaptive Cruise Control) system 20, an auto light system 21, an auto air conditioning system 22, etc. via a CAN bus 15 so as to be able to communicate data with them. For example, the DSM 19 transmits data indicating the driver's state to the CAN communication unit 12 as detected data. For example, the ACC system 20 transmits data indicating the control state of the ACC system 20 to the CAN communication unit 12 as detected data. For example, the auto light system 21 transmits data indicating the control state of the auto light system 21 to the CAN communication unit 12 as detected data. For example, the auto air conditioning system 22 transmits data indicating the control state of the auto air conditioning system 22 to the CAN communication unit 12 as detected data.

[0024] The number of CAN buses 13-15 connected to the CAN communication unit 12 is not limited to three, and may be two or less, or four or more. The various ECUs 161-16n, the various cameras 171-17n, the various sensors 181-18n, the DSM 19, and the various systems 20-22 may be connected to the CAN buses 13-15 in any manner. The various devices connected to the CAN buses 13-15 are not limited to the ECUs, cameras, sensors, and systems described above. Furthermore, the in-vehicle communication network is not limited to a configuration in which the CAN buses 13-15 are used, and may also be a configuration in which Ethernet or the like is used. Ethernet is a registered trademark.

[0025] 4, in the center device 2, the control unit 6 includes, for each function, a trigger condition creation unit 6a, a machine learning model creation unit 6b, a collected data list creation unit 6c, a distribution unit 6d, and a collected data acquisition unit 6e. These units 6a to 6e execute the data processing method of the center device 2 and the data processing program of the center device 2.

[0026] The trigger condition creation unit 6a creates a trigger condition for detecting a scene in which the user is dissatisfied (corresponding to a predetermined scene on the vehicle side). The machine learning model creation unit 6b creates a machine learning model. The collected data list creation unit 6c creates a collected data list corresponding to the trigger condition. The distribution unit 6d distributes the trigger condition created by the trigger condition creation unit 6a, the machine learning model created by the machine learning model creation unit 6b, and the collected data list created by the collected data list creation unit 6c from the communication unit 7 to the vehicle device 4. The collected data acquisition unit 6e acquires collected data from the vehicle device 4 by receiving collected data transmitted from the vehicle device 4 via the communication unit 7. When distributing a machine learning model, both a machine learning model program, such as a neural network program, and parameters used in the machine learning model program, such as trained weighting coefficients in the neural network, may be distributed. Furthermore, when switching the parameters of a machine learning model installed on the vehicle side from those for young people to those for elderly people, for example, only parameters used in the machine learning model program, such as trained weighting coefficients in the neural network, may be distributed.

[0027] We will now explain the machine learning model. There are two cases where the trigger condition is created using only the machine learning model, and two where the trigger condition is created using the machine learning model and other items. Each case will be explained below.

[0028] (1) When trigger conditions are created only by a machine learning model (see Figures 5 and 6) The driver's excitement and anger are detected by inputting the time series data of accelerator operation, image data, and voice data from the last 60 seconds into the machine learning model. "Trigger condition" Machine learning model: model_angry_no.1 (accelerator opening (60 seconds), driver image (60 seconds), driver voice (60 seconds)) "Collected data" Latitude and longitude, vehicle speed, accelerator opening, and brake pedal operation amount from 60 seconds before the trigger condition is ignited

[0029] The trigger condition creation unit 6a creates a trigger condition file and creates trigger conditions, as shown in Fig. 5. The trigger condition file illustrated in Fig. 5 has the following items: trigger condition, machine learning model, accelerator opening, driver image, and driver voice. The detection data indicating the accelerator opening, the detection data indicating the driver image, and the detection data indicating the driver voice are time-series data that serve as input to the machine learning model.

[0030] The collected data list creating unit 6c creates a collected data list file and creates collected data, as shown in Fig. 6. The collected data list file shown in Fig. 6 has the following items: acquisition period (before trigger firing), acquisition period (after trigger firing), control state of the ACC system, latitude and longitude, vehicle speed, accelerator opening, and brake pedal operation amount.

[0031] (2) When trigger conditions are created using a machine learning model and other items (see Figures 7 and 8) When a vehicle is traveling on a highway, the driver's excitement or anger is detected by inputting the time-series data of accelerator operation, image data, and voice data from the last 60 seconds into the machine learning model. "Trigger conditions" Machine learning model: model_angry_no.1 (accelerator opening (60 seconds), driver image (60 seconds), driver voice (60 seconds)) Road type = 2 (highway) "Collected data" Latitude and longitude, vehicle speed, accelerator opening, and brake pedal operation amount from 60 seconds before the trigger condition is ignited

[0032] The trigger condition creation unit 6a creates a trigger condition file and creates trigger conditions, as shown in Fig. 7. The trigger condition file shown in Fig. 7 has the following items: trigger condition, machine learning model, accelerator opening, driver image, driver voice, and road type. The detection data indicating the accelerator opening, the detection data indicating the driver image, and the detection data indicating the driver voice are time-series data that serve as input to the machine learning model.

[0033] The collected data list creating unit 6c creates a collected data list file and creates collected data, as shown in Fig. 8. The collected data list file shown in Fig. 8 has the following items: acquisition period (before trigger firing), acquisition period (after trigger firing), control state of the ACC system, latitude and longitude, vehicle speed, accelerator opening, and brake pedal operation amount.

[0034] 9, in the vehicle device 4, the control unit 9 includes, for each function, a detection data acquisition unit 9a, a trigger condition determination unit 9b, a collected data creation unit 9c, and a collected data transmission unit 9d. These units 9a to 9d execute the data processing method of the vehicle device 4 and the data processing program for the vehicle device 4.

[0035] The detection data acquisition unit 9a acquires detection data necessary for determining the trigger condition. For example, when determining the accelerator opening, the detection data acquisition unit 9a acquires data indicating the accelerator opening from the accelerator ECU as detection data. For example, when determining the driver image, the detection data acquisition unit 9a acquires data indicating the driver image from the DSM as detection data. For example, when determining the driver voice, the detection data acquisition unit 9a acquires data indicating the driver voice from the DSM as detection data. As described above, these detection data are time-series data that serve as input to the machine learning model.

[0036] When the detection data acquisition unit 9a acquires the detection data, the trigger condition determination unit 9b performs a trigger condition determination based on the execution result of the machine learning model based on the acquired detection data. The trigger condition determination unit 9b determines whether the machine learning model can be executed in the vehicle device 4 based on the vehicle resources. The vehicle resources include, for example, the number of activations of the trigger condition, the CPU usage rate, the memory usage rate, etc.

[0037] When the trigger condition determination unit 9b determines that vehicle resources are sufficient and the machine learning model can be executed in the vehicle device 4, it inputs the detection data as time-series data to the machine learning model of the vehicle device 4, and performs trigger condition determination based on the execution result of the function learning model executed in the vehicle device 4. When the trigger condition determination unit 9b determines that vehicle resources are insufficient and the machine learning model cannot be executed in the vehicle device 4, it causes the detection data to be transmitted as time-series data from the communication unit 10 to the center device 2, inputs the time-series data to the machine learning model of the center device 2 to execute the machine learning model, and obtains the execution result from the center device 2, and performs trigger condition determination based on the execution result of the function learning model executed in the center device 2.

[0038] The collected data creating unit 9 c creates collected data based on the result of the trigger condition determination. When the collected data creating unit 9 c creates collected data, the collected data transmitting unit 9 d transmits the created collected data from the communication unit 10 to the center device 2.

[0039] Next, the operation of the above-described configuration will be described with reference to Figures 10 to 13. Here, the processing performed by the control unit 9 of the vehicle device 4 and the processing performed by the control unit 6 of the center device 2 will be described.

[0040] (1) Processing performed by the control unit 9 of the vehicle device 4 (see FIG. 10) In the vehicle device 4, the control unit 9 starts vehicle-side processing. For example, when detecting that a user has boarded the vehicle, the control unit 9 acquires a driver ID unique to the boarded user (A1). The control unit 9 causes the communication unit 10 to transmit the acquired driver ID to the center device 2 (A2, S1 shown in FIGS. 12 and 13), and waits for reception of a trigger condition, a machine learning model, and a list of collected data from the center device 2.

[0041] When the control unit 9 acquires the trigger conditions, machine learning model, and collected data list distributed from the center device 2 via the communication unit 10 (A3), it determines whether the machine learning model can be executed in the vehicle device 4 based on the vehicle resources (A4, S3 shown in Figures 12 and 13).

[0042] The control unit 9 acquires the detection data required for determining the trigger condition as time-series data (A5, which corresponds to S4 in FIGS. 12 and 13, the detection data acquisition procedure). When the control unit 9 determines that the machine learning model is executable in the vehicle device 4 (A6: YES), the control unit 9 executes the machine learning model using the acquired time-series data as input, and determines whether the trigger condition is satisfied based on the execution result of the machine learning model (A7, which corresponds to S5 in FIG. 12, the trigger condition determination procedure).

[0043] On the other hand, if the control unit 9 determines that the machine learning model is not executable in the vehicle device 4 (A6: NO), it causes the communication unit 10 to transmit the time-series data to the center device 2 (A8) and waits to receive the execution result of the machine learning model from the center device 2 (A9). When the execution result of the machine learning model distributed from the center device 2 is received by the communication unit 10 and the control unit 9 acquires the execution result of the machine learning model from the center device 2 (A9: YES), the control unit 9 determines whether the trigger condition determination is satisfied based on the acquired execution result of the machine learning model (A7, which corresponds to S6 in FIG. 13 , the trigger condition determination procedure).

[0044] When the control unit 6 determines that the trigger condition is met (A7: YES), it creates collected data (A10, corresponding to the collected data creation procedure), transmits the created collected data from the communication unit 10 to the center device 2 (A11, corresponding to S7 shown in Figures 12 and 13, the collected data transmission procedure), and terminates the vehicle-side processing.

[0045] When the above-mentioned trigger condition is created using only a machine learning model, the control unit 9 determines that the trigger condition is determined to be true when the following machine learning model is true: model_angry_no.1 (accelerator opening (60 seconds), driver image (60 seconds), driver voice (60 seconds)). In other words, the control unit 9 determines that the trigger condition is determined to be true when it identifies that the user is dissatisfied through machine learning of the time-series data of the accelerator opening, driver image, and driver voice. The control unit 9 creates collected data including latitude and longitude, vehicle speed, accelerator opening, and brake pedal operation amount from 60 seconds before the trigger condition is ignited, and causes the communication unit 10 to transmit the created collected data to the center device 2.

[0046] When the above-mentioned trigger condition is created using a machine learning model and other items, the control unit 9 determines that the trigger condition is determined to be true if the following is true: Machine learning model: model_angry_no.1 (accelerator opening (60 seconds), driver image (60 seconds), driver voice (60 seconds)) Road type = 2 (expressway) . In other words, when the control unit 9 determines that the user is dissatisfied through machine learning of time-series data of accelerator opening, driver image, and driver voice while the vehicle is traveling on an expressway, it determines that the trigger condition is determined to be true. The control unit 9 creates collected data including latitude and longitude, vehicle speed, accelerator opening, and brake pedal operation amount from 60 seconds before the trigger condition is ignited, and causes the communication unit 10 to transmit the created collected data to the center device 2.

[0047] (2) Processing performed by the control unit 6 of the center device 2 (see FIG. 11 ). In the center device 2, when the center-side processing is started, the control unit 6 acquires the driver ID from the center device 2 by receiving the driver ID transmitted from the center device 2 via the communication unit 7 (B1). The control unit 6 then creates a trigger condition, a machine learning model, and a collected data list (B2, corresponding to the trigger condition creation procedure, machine learning model creation procedure, and collected data list creation procedure). In this case, the control unit 6 creates the trigger condition including adjustment parameters indicating the preferences and characteristics of the driver associated with the acquired driver ID. The control unit 6 then distributes the created trigger condition, machine learning model, and collected data list from the communication unit 7 to the vehicle device 4 (B3, corresponding to S2, the distribution procedure, shown in FIGS. 12 and 13 ), and determines whether time-series data has been received from the vehicle device 4 (B4).

[0048] When the control unit 6 acquires the time series data from the vehicle device 4 (B4: YES) as a result of the time series data being received by the communication unit 7 from the vehicle device 4, the control unit 6 executes a machine learning model using the acquired time series data as input, distributes the execution results of the machine learning model from the communication unit 7 to the vehicle device 4 (B5), and waits to receive collected data from the vehicle device 4.

[0049] When the collected data transmitted from the vehicle device 4 is received by the communication unit 10, the control unit 6 acquires the collected data from the vehicle device 4 (B6, corresponding to the collected data acquisition procedure) and ends the center-side processing. After this, by analyzing the collected data acquired from the vehicle device 4, it becomes possible to analyze the scene that caused the user dissatisfaction.

[0050] The above example illustrates a configuration in which the detection data indicating the accelerator opening, the detection data indicating the driver image, and the detection data indicating the driver voice are input to the machine learning model, but data indicating the driver's preferences and characteristics, which are adjustment parameters linked to the driver ID, may also be input to the machine learning model.Furthermore, data indicating driving preferences, such as whether the driver prefers brisk driving or relaxed driving, may also be input to the machine learning model.

[0051] A plurality of machine learning models may be created for each driver's preferences or characteristics. For example, machine learning models that use detection data indicating the accelerator opening, detection data indicating the driver's image, and detection data indicating the driver's voice as input may be created for young people, elderly people, drivers who prefer brisk driving, and drivers who prefer relaxed driving.

[0052] In the center device 2, an example configuration has been given in which a machine learning model is created after a driver ID is acquired, but it is also possible to create multiple models in advance and select a model linked to the driver ID after acquiring the driver ID.

[0053] The machine learning model may be configured to be able to detect multiple predetermined scenes and may be configured to output whether or not each of the multiple predetermined scenes has occurred. For example, one machine learning model may be configured to receive detection data indicating the accelerator pedal position, detection data indicating a driver image, and detection data indicating the driver's voice as input, and output whether or not a first predetermined scene in which the driver is angrily depressing the accelerator pedal and a second predetermined scene in which the driver is fearfully depressing the accelerator pedal have occurred. If the driver is angrily and fearfully depressing the accelerator pedal, the first predetermined scene and the second predetermined scene will occur simultaneously.

[0054] The machine learning model may be composed of multiple machine learning models. For example, the configuration is not limited to one in which one machine learning model determines both the first predetermined scene and the second predetermined scene. Alternatively, the first machine learning model may determine the first predetermined scene, and the second machine learning model may determine the second predetermined scene. When the machine learning model is configured to be able to detect multiple predetermined scenes, the trigger condition file distributed from the center device 2 may specify how the occurrence of a predetermined scene among the multiple predetermined scenes detectable by the machine learning model is to be included in the trigger condition, for example, by logical sum or logical product with other items.

[0055] A machine learning model may be created, for example, by the following method. For example, correct answers to inputs to the machine learning model, such as accelerator position, driver image, and driver voice, as well as data indicating the driver's preferences and characteristics, such as output of whether a first predetermined scene occurred or whether a second predetermined scene occurred, are prepared as a training dataset. For example, data on whether a predetermined scene occurred in various situations represented by inputs in a test vehicle or the like is collected for each driver with different preferences and characteristics, and prepared as a training dataset. This training data is used for training to obtain a trained model. For example, trained weighting coefficients in a neural network are obtained.

[0056] As described above, according to the embodiment, the following advantageous effects can be achieved. The center device 2 distributes the trigger conditions, the machine learning model, and the list of collected data to the vehicle device 4. The vehicle device 4 determines the trigger conditions based on the execution results of the machine learning model based on the detection data, and creates collected data based on the determination results of the trigger condition determination and transmits the data to the center device. It is possible to appropriately set trigger conditions that can accurately detect user dissatisfaction, and to accurately detect the user's latent dissatisfaction. Furthermore, by inputting data into the machine learning model indicating whether the trigger conditions captured the user's feelings of dissatisfaction, i.e., whether the trigger conditions were correct or incorrect, it is possible to train and evaluate the accuracy of the machine learning model of the trigger conditions, and the accuracy of the trigger conditions can be improved by evolving the machine learning model.

[0057] When the vehicle resources are sufficient and the machine learning model can be executed in the vehicle device 4, the machine learning model is executed in the vehicle device 4, and a trigger condition determination is made based on the execution result of the machine learning model. When the vehicle resources are sufficient, the machine learning model can be executed by utilizing the vehicle resources.

[0058] When the machine learning model cannot be executed in the vehicle device 4 due to a lack of vehicle resources, the machine learning model is executed in the center device 2, and a trigger condition determination is made based on the execution result of the machine learning model. Even when vehicle resources are insufficient, the center device 2 can be made to execute the machine learning model.

[0059] Trigger conditions are created based on the driver ID. By reflecting the driver's preferences and characteristics in the trigger conditions, it is possible to easily manage the reuse of trigger conditions, including the driver's preferences and characteristics.

[0060] Although the present disclosure has been described with reference to the embodiments, it is understood that the present disclosure is not limited to the embodiments or structures. The present disclosure also encompasses various modifications and modifications within the scope of equivalents. In addition, various combinations and forms, as well as other combinations and forms including only one element, more than one element, or less than one element, are also within the scope and spirit of the present disclosure.

[0061] The time-series data input to the machine learning model is not limited to detection data indicating the accelerator opening, detection data indicating the driver's image, and detection data indicating the driver's voice, but may also include data indicating biometric information associated with the user's dissatisfaction, etc. Furthermore, the data indicating biometric information may be data acquired from either an in-vehicle device installed in the vehicle or a user terminal such as a smartphone or tablet terminal owned by the user.

[0062] The control unit and the method described herein may be implemented by a special-purpose computer configured by configuring a processor and memory programmed to perform one or more functions embodied in a computer program. Alternatively, the control unit and the method described herein may be implemented by a special-purpose computer configured by configuring a processor with one or more dedicated hardware logic circuits. Alternatively, the control unit and the method described herein may be implemented by one or more special-purpose computers configured by combining a processor and memory programmed to perform one or more functions with a processor configured with one or more hardware logic circuits. Furthermore, the computer program may be stored as instructions executed by a computer on a computer-readable non-transitory tangible storage medium.

[0063] In addition to the claims, the present disclosure also includes the following disclosure. a machine learning model creation unit (6b) that creates a machine learning model; a collected data list creation unit (6c) that creates a collected data list corresponding to the trigger condition; a distribution unit (6d) that distributes the trigger condition, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition unit (6e) that acquires collected data indicated in the collected data list from the vehicle device; and the vehicle device includes: a detection data acquisition unit (9a) that acquires detection data necessary for trigger condition determination; a trigger condition determination unit (9b) that performs trigger condition determination based on a result of execution of the machine learning model based on the detection data; a collected data creation unit (9c) that creates the collected data based on a determination result of the trigger condition determination; and a collected data transmission unit (9d) that transmits the collected data to the center device.

[0064] [2] The data communication system according to [1], wherein, if the machine learning model can be executed in the vehicle device, the trigger condition determination unit determines the trigger condition based on the execution result of the machine learning model executed in the vehicle device.

[0065] [3] A data communication system according to [1] or [2], wherein, if the machine learning model cannot be executed in the vehicle device, the trigger condition determination unit determines the trigger condition based on the execution result of the machine learning model executed in the center device.

[0066] [4] The data communication system according to claim 1 in any one of [1] to [3], wherein the trigger condition creating unit creates the trigger condition based on driver identification information that identifies a driver.

[0067] [5] The data communication system according to claim 1, wherein the trigger condition creating unit creates the trigger condition using only the machine learning model.

[0068] [6] The data communication system according to claim 1 in any one of [1] to [4], wherein the trigger condition creation unit creates the trigger condition based on the machine learning model and other items.

Claims

1. A data communication system (1) in which a center device (2) and a vehicle device (4) mounted on a vehicle can communicate data, wherein the center device comprises: a trigger condition creation unit (6a) that creates trigger conditions for detecting a predetermined scene on the vehicle side; a machine learning model creation unit (6b) that creates a machine learning model; a collected data list creation unit (6c) that creates a collected data list corresponding to the trigger conditions; a distribution unit (6d) that distributes the trigger conditions, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition unit (6e) that acquires collected data indicated in the collected data list from the vehicle device, and the vehicle device comprises: a detection data acquisition unit (9a) that acquires detection data necessary for trigger condition determination; a trigger condition determination unit (9b) that performs trigger condition determination based on the execution result of the machine learning model based on the detection data; a collected data creation unit (9c) that creates the collected data based on the determination result of the trigger condition determination; and a collected data transmission unit (9d) that transmits the collected data to the center device.

2. A data communication system as described in claim 1, wherein the trigger condition determination unit, if the machine learning model can be executed in the vehicle device, determines the trigger condition based on the execution results of the machine learning model executed in the vehicle device.

3. A data communication system as described in claim 1, wherein the trigger condition determination unit determines the trigger condition based on the execution result of the machine learning model executed in the center device when the machine learning model cannot be executed in the vehicle device.

4. A data communication system according to claim 1, wherein the trigger condition creating unit creates the trigger condition based on driver identification information that identifies a driver.

5. A data communication system according to claim 1, wherein the trigger condition creation unit creates the trigger condition solely using the machine learning model.

6. A data communication system according to claim 1, wherein the trigger condition creation unit creates the trigger condition based on the machine learning model and other items.

7. A center device (2) capable of data communication with a vehicle device (4) that is mounted on a vehicle and performs trigger condition determination based on the execution result of the machine learning model based on detection data required for trigger condition determination, creates collected data based on the determination result of the trigger condition determination, and transmits the data to the center device, the center device comprising: a trigger condition creation unit (6a) that creates trigger conditions for detecting a specified scene on the vehicle side; a machine learning model creation unit (6b) that creates a machine learning model; a collected data list creation unit (6c) that creates a collected data list corresponding to the trigger conditions; a distribution unit (6d) that distributes the trigger conditions, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition unit (6e) that acquires collected data indicated by the collected data list from the vehicle device.

8. A vehicle device (4) mounted on a vehicle, capable of data communication with a center device (2) that creates trigger conditions for detecting a specified scene on the vehicle side, creates a machine learning model, creates a list of collected data corresponding to the trigger conditions, distributes the trigger conditions, the machine learning model, and the list of collected data to the vehicle device, and acquires collected data from the vehicle device, the vehicle device comprising: a detection data acquisition unit (9a) that acquires detection data necessary for trigger condition determination; a trigger condition determination unit (9b) that performs trigger condition determination based on the execution result of the machine learning model based on the detection data; a collection data creation unit (9c) that creates the collection data based on the determination result of the trigger condition determination; and a collection data transmission unit (9d) that transmits the collection data to the center device.

9. A data processing method executed in a data communication system (1) capable of data communication between a center device (2) and a vehicle device (4) mounted on a vehicle, wherein the center device executes: a trigger condition creation procedure for creating trigger conditions for detecting a specified scene on the vehicle side; a machine learning model creation procedure for creating a machine learning model; a collected data list creation procedure for creating a collected data list corresponding to the trigger conditions; a distribution procedure for distributing the trigger conditions, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition procedure for acquiring collected data indicated in the collected data list from the vehicle device; and the vehicle device executes: a detection data acquisition procedure for acquiring detection data necessary for trigger condition determination; a trigger condition determination procedure for determining a trigger condition based on the execution result of the machine learning model based on the detection data; a collected data creation procedure for creating the collected data based on the determination result of the trigger condition determination; and a collected data transmission procedure for transmitting the collected data to the center device.

10. A data processing method in a center device (2) capable of data communication with a vehicle device (4) that is mounted on a vehicle and performs trigger condition determination based on the execution result of a machine learning model based on detection data required for trigger condition determination, creates collected data based on the determination result of the trigger condition determination, and transmits it to the center device, the data processing method comprising: a trigger condition creation procedure for creating trigger conditions for detecting a specified scene on the vehicle side; a machine learning model creation procedure for creating a machine learning model; a collected data list creation procedure for creating a collected data list corresponding to the trigger conditions; a distribution procedure for distributing the trigger conditions, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition procedure for acquiring collected data indicated by the collected data list from the vehicle device.

11. A data processing method in a vehicle device (4) mounted on a vehicle, capable of data communication with a center device (2) that creates trigger conditions for detecting a specified scene on the vehicle side, creates a machine learning model, creates a list of collected data corresponding to the trigger conditions, distributes the trigger conditions, the machine learning model, and the list of collected data to a vehicle device, and acquires collected data from the vehicle device, the data processing method comprising: a detection data acquisition procedure for acquiring detection data necessary for trigger condition determination; a trigger condition determination procedure for making a trigger condition determination based on the execution result of the machine learning model based on the detection data; a collection data creation procedure for creating the collection data based on the determination result of the trigger condition determination; and a collection data transmission procedure for transmitting the collection data to the center device.

12. A data processing program that causes a control unit (6) of a center device (2) capable of data communication with a vehicle device (4) that is mounted on a vehicle and performs trigger condition determination based on the execution result of a machine learning model based on detection data required for trigger condition determination, creates collected data based on the determination result of the trigger condition determination, and transmits it to a center device, to execute: a trigger condition creation procedure for creating trigger conditions for detecting a specified scene on the vehicle side; a machine learning model creation procedure for creating a machine learning model; a collected data list creation procedure for creating a collected data list corresponding to the trigger conditions; a distribution procedure for distributing the trigger conditions, the machine learning model, and the collected data list to the vehicle device; and a collected data acquisition procedure for acquiring collected data indicated by the collected data list from the vehicle device.

13. A data processing program that creates trigger conditions for detecting specified scenes on the vehicle side, creates a machine learning model, creates a list of collected data corresponding to the trigger conditions, distributes the trigger conditions, the machine learning model, and the list of collected data to a vehicle device, and is capable of data communication with a center device (2) that acquires collected data from the vehicle device, and causes a control unit (9) of the vehicle device (4) mounted on the vehicle to execute: a detection data acquisition procedure for acquiring detection data necessary for trigger condition determination; a trigger condition determination procedure for making a trigger condition determination based on the execution result of the machine learning model based on the detection data; a collection data creation procedure for creating the collection data based on the determination result of the trigger condition determination; and a collection data transmission procedure for transmitting the collection data to the center device.

Images