Information notification control device, information notification control program, driver assistance system, and information collection method

Abstract

This technology provides an improved efficiency in acquiring user feedback information for AI models used in notification control. [Solution] An exemplary information notification control device is an information notification control device that uses an AI model to control the notification of information, and comprises a controller, which, when it determines that it is a suitable time to acquire feedback information for the AI ​​model, performs the notification control according to the inference result with a lower confidence level than the inference result with the highest confidence level among a plurality of inference results with different confidence levels output by the AI ​​model.

Description

【Technical Field】 【0001】 The present invention relates to a technique for performing notification control of information using an AI (Artificial Intelligence) model. 【Background Art】 【0002】 Conventionally, a notification system that gives notifications to a driver (user) for purposes such as driving support has been installed in a vehicle (see, for example, Patent Document 1). As disclosed in Patent Document 1, there are various means of information notification in a vehicle. For example, display, sound, and vibration are used. 【Prior Art Documents】 【Patent Documents】 【0003】 【Patent Document 1】 Japanese Unexamined Patent Application Publication No. 2020-41914 【Summary of the Invention】 【Problems to be Solved by the Invention】 【0004】 In a vehicle, information notification has various notification means (display, sound, vibration, etc.) as described above, and the types of information to be notified are also diverse. For this reason, control of information notification is not easy, and it is even more difficult to perform notification control that suits the preferences of users such as drivers. In view of such circumstances, it is conceivable to apply an AI model to the control of information notification. 【0005】 When applying an AI model to a notification system installed in a vehicle, it is assumed that the AI model is initially a learned model learned using initially prepared initial learning data. Then, when a user (such as a driver) starts using the AI model and obtains feedback data from the user, and relearning is performed based on the feedback data, it is assumed that the AI model approaches an AI model that gives notifications according to the preferences of the user. 【0006】 Methods for acquiring feedback data in vehicles include, for example, panel input using touch panels and acquiring user reactions such as the user's voice or facial expressions after notification is executed. However, if the user does not use the vehicle frequently or if there are few opportunities for information notifications, it may not be possible to collect sufficient feedback data. Also, in the case of panel input, for example, the user may not input due to reasons such as inconvenience, which may result in insufficient feedback data being collected. Furthermore, when using methods to acquire voice or facial expressions, the voice and facial expression data itself may not accurately capture the user's expectations regarding notifications, which may result in insufficient feedback data being collected. 【0007】 In view of the above, the present invention aims to provide a technology that can improve the efficiency of acquiring user feedback information for an AI model used in notification control. [Means for solving the problem] 【0008】 An exemplary information notification control device of the present invention is an information notification control device that uses an AI model to control the notification of information, and comprises a controller, wherein when the controller determines that it is a suitable time to acquire feedback information for the AI ​​model, it performs the notification control according to an inference result with a lower confidence level than the inference result with the highest confidence level among a plurality of inference results with different confidence levels output by the AI ​​model. [Effects of the Invention] 【0009】 According to the exemplary invention, information notifications made at specific timings (timing suitable for acquiring feedback information) will intentionally include information that may not be optimal for the user. Users who receive notifications of potentially inadequate information are more likely to take action to provide feedback on the notification. Furthermore, the user's reaction to notifications of potentially inadequate information is more likely to be useful feedback information for training the AI ​​model. For this reason, according to the exemplary invention, the efficiency of acquiring user feedback information for the AI ​​model used for notification control can be improved. [Brief explanation of the drawing] 【0010】 [Figure 1] Diagram showing the general configuration of the driver assistance system. [Figure 2] Block diagram showing the configuration of the functional unit of the controller of the information notification control device. [Figure 3] Schematic diagram to explain the HMI control model. [Figure 4] This figure shows specific examples of inputs and outputs in an HMI control model. [Figure 5] A flowchart illustrating the flow of notification control processing performed by the information notification control device. [Figure 6] A flowchart showing a modified example of the notification control process performed by the information notification control device. [Figure 7] A flowchart illustrating the processing flow related to feedback information executed by the information notification control device. [Figure 8] Block diagram showing the general configuration of the learning server. [Figure 9A] A diagram showing properly labeled data obtained from positive data. [Figure 9B] A diagram showing properly labeled data obtained from negative data. [Modes for carrying out the invention] 【0011】 Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts will be denoted by the same reference numerals and will not be repeated in the description. 【0012】 <1. Driver assistance systems> Figure 1 shows a schematic configuration of a driver assistance system 100 according to an embodiment of the present invention. The driver assistance system 100 is a system that assists in the driving of a vehicle VE. The driver assistance system 100 can also be described as an information notification system that notifies the driver of the vehicle VE of information related to driving. The information notified may include not only information related to driving, but also entertainment information such as music and movies, and in such a configuration, the system indicated by reference numeral 100 may be understood as an information provision system. When configured as an information provision system, the provision (notification) of information related to driving is not essential, and the information provision system is not limited to vehicles, but can be applied to various uses such as homes, factories, or offices. 【0013】 As shown in Figure 1, the driver assistance system 100 comprises an information notification control device 1, an information source device 2, and an information notification device 3. The driver assistance system 100 also comprises an FB (feedback) information acquisition device 4, an agent system 5, and a learning server 6. In this embodiment, the information notification control device 1, the information source device 2, the information notification device 3, the FB information acquisition device 4, and the agent system 5 are mounted on the vehicle VE. The learning server 6 is located outside the vehicle VE and is configured to communicate with the information notification control device 1 via a communication network such as the Internet (not shown). 【0014】 In this embodiment, the driver assistance system 100 includes an agent system 5 and a learning server 6, but this is merely an example. The driver assistance system 100 may also be configured to not include at least one of the agent system 5 and the learning server 6. 【0015】 The information notification control device 1 performs control related to the notification of information. The information notification control device 1 determines in what notification mode to notify the information to be notified to the notification target person (a detailed example is a driver) received from the information source device 2. Further, the information notification control device 1 controls the information notification device 3 so that the notification is performed in the determined notification mode. The determination of the notification mode includes, for example, the determination of notification means, notification location, and notification parameters such as notification intensity. In the present embodiment, at least the determination of the notification parameters uses an AI model. That is, the information notification control device 1 performs notification control for the information notification device 3 using the AI model by acquiring information from the information source device 2. Details of the information notification control device 1 will be described later. 【0016】 Note that in the present embodiment, the information notification control device 1 is an in-vehicle device mounted on the vehicle VE, but this is an example. The information notification control device 1 may be composed of an in-vehicle device and a server device provided to be communicable via a communication network such as the Internet with the in-vehicle device. Further, the information notification control device 1 mounted on the vehicle VE may be configured as a part of in-vehicle devices such as a navigation device or a display audio. 【0017】 The information source device 2 is a device that serves as an information source of the information to be notified to the notification target person, and provides notification information to the information notification control device 1. Specifically, the information source device 2 is a device that serves as an information source of the information to be notified to the driver. The information source device 2 outputs the information to be notified to the driver toward the information notification control device 1. The information source device 2 may be configured to, for example, read information stored in a storage medium (not shown) and output it toward the information notification control device 1. Further, the information source device 2 may be configured to generate the information to be notified by itself and output it toward the information notification control device 1. Further, the information source device 2 may be configured to receive the information to be notified from another device such as a server device and output it toward the information notification control device 1. 【0018】 In this embodiment, the information (notification information) output from the information source device 2 is information for assisting driving. That is, the information source device 2 is a device that serves as an information source for information for assisting driving. The information for assisting driving may widely include information related to driving, such as information indicating the state of the vehicle VE, information indicating the state of the driver, information informing about the surrounding environment of the vehicle VE, information regarding route guidance, and attention - calling information during driving. The information source device 2 that outputs such notification information may be composed of, for example, a navigation device or a safety monitoring device mounted on the vehicle VE. 【0019】 Note that the information source device 2 may be composed of only one device or a plurality of devices. Also, in this embodiment, the information source device 2 is an in - vehicle device, but at least some of its functions may be composed of a server device that is communicable with the in - vehicle device. Further, when the notification information may also include information other than information related to driving, the notification information may include entertainment information such as music and videos (movies, etc.) in a TV, radio, disc player, etc., and Internet information such as news. 【0020】 The information notification device 3 performs information notification in accordance with a command from the information notification control device 1. Specifically, the information notification device 3 performs information notification using at least one of the driver's vision, hearing, and touch in accordance with a command from the information notification control device 1. In this embodiment, the information notification device 3 performs notification of information for assisting driving. The information notification device 3 is composed of multiple types of devices. The multiple types of devices are a display device 31, an audio output device 32, and a vibration generating device 33. 【0021】 The display device 31 is a means of notifying information using the driver's vision. The display device 31 is positioned in a location where the driver of the vehicle VE to which the driver assistance system 100 is applied can easily see the displayed content. For example, the display device 31 is a liquid crystal display or an organic EL display placed on the dashboard of the vehicle VE. Alternatively, the display device 31 is a head-up display (HUD) that projects information onto the windshield of the vehicle VE. Alternatively, the display device 31 is an indicator light placed on the meter panel, rearview mirror, or side mirror. The display device 31 may also be part of a stationary or portable navigation system or safety monitoring device mounted in the vehicle. The display device 31 may also be a display operation device with an operation function such as a touch panel. There may be one or more display devices 31. If there are multiple, there may be multiple types of display devices 31. 【0022】 The voice output device 32 is a means of notifying information using the driver's hearing. The voice output device 32 is, for example, a speaker positioned in a location where the driver of a vehicle VE to which the driver assistance system 100 is applied can easily hear the voice. The voice output device 32 may also be part of a stationary or portable navigation system or safety monitoring system installed in the vehicle. There may be one or more voice output devices 32. 【0023】 The vibration generator 33 is a means of notifying the driver of information using their sense of touch. The vibration generator 33 includes, for example, a vibrator positioned in a location where the driver of the vehicle can feel the vibration. The vibrator is positioned, for example, in the driver's seat. More specifically, the vibrator is positioned in the seat that supports the driver's buttocks, or in the backrest that supports the driver's back when their buttocks are on the seat. The vibrator may be, for example, a vibrator with an electrical-magnetic circuit configuration in which the diaphragm of a speaker (having a structure suitable for acoustic conversion) is replaced with a diaphragm suitable for vibration transmission, or a vibrator with a configuration utilizing a piezoelectric element. 【0024】 In this embodiment, the system utilizes three senses—sight, hearing, and touch—when notifying information, but this is merely an example. The types of senses that can be used when notifying information may be one or two, for example. For instance, the types of senses that can be used when notifying information may be two, such as sight and hearing. Furthermore, the information notification device 3 may consist of one type of device or multiple devices other than three. 【0025】 The FB information acquisition device 4 is a means for acquiring feedback information from a user (in this embodiment, the driver) who has received a notification from the information notification device 3. The feedback information is the driver's impression, feelings, or opinion regarding the information notification, or information that allows for the inference of such impression, feelings, or opinion. The FB information acquisition device 4 outputs the acquired feedback information to the information notification control device 1. The information notification control device 1 processes the acquired feedback information and reflects it in the control of the notification. The information notification control device 1 also performs processing such as storing the acquired feedback information so that it can be used as data for learning the HMI (Human Machine Interface) control model described later. 【0026】 The FB information acquisition device 4 may be, for example, a camera, a microphone, or an input device such as a touch panel. The camera is, in detail, a camera that photographs the driver, such as a camera provided in a drive recorder. From the driver's facial expressions captured in the camera's images, the driver's feelings regarding the information notification can be inferred, and this information can serve as feedback information. The microphone is, in detail, a microphone placed near the driver's seat of the vehicle VE. From the driver's voice input via the microphone, the driver's feelings regarding the information notification can be inferred, and this information can serve as feedback information. In a configuration using a voice recognition device, the driver can directly input their thoughts and feelings regarding the information notification using the microphone. The input device such as a touch panel is, in detail, a function provided in a navigation system, and the driver can directly input their thoughts and feelings regarding the information notification using the input device. 【0027】 The input device for inputting information may consist of a terminal device such as a smartphone owned by the driver. However, in this case, there may be cases where the terminal device cannot communicate directly with the information notification control device 1, for example, if the vehicle VE's power is turned off. For this reason, the feedback information transmitted from the terminal device may be configured to be sent to a cloud server first, and the information notification control device 1 may be configured to acquire the feedback information from the cloud server. 【0028】 The agent system 5 has agent functions. These agent functions include, for example, a function for interacting with the occupants of the vehicle VE (in a specific example, the driver). Interactions with the occupants conducted by the agent functions may include interactions conducted in response to commands from other devices within the vehicle VE. In this embodiment, interactions with the occupants conducted by the agent functions include interactions conducted in response to commands from the information notification control device 1. 【0029】 In detail, the agent function is implemented by integrating a speech recognition function (a function that converts speech to text) that recognizes the voice of the occupant, a natural language processing function (a function that understands the structure and meaning of text), a dialogue management function, and a search function that searches other devices and predetermined databases owned by the system itself. Some or all of these functions may be implemented using AI technology. Furthermore, some of the configurations for performing these functions (in particular, the speech recognition function and the natural language processing function) may be installed on an agent server (not shown) that can communicate with the communication device equipped in the vehicle VE. Note that the agent server is an external device and, in detail, is a cloud server. 【0030】 As shown in Figure 1, the agent system 5 comprises an agent control device 51, an agent information acquisition device 52, and an agent device 53. These are mounted on the vehicle VE and connected to each other so that they can communicate with one another. The agent control device 51 is a computer device that performs various processes so that the agent system 5 can perform the agent functions described above. This computer device includes a processor and memory, and the agent functions are performed by the processor executing calculations according to the program stored in the memory. The agent control device 51 is connected to the information notification control device 1 (specifically the controller 11) so that they can communicate with each other. 【0031】 The agent information acquisition device 52 is a device that acquires information necessary for the agent control device 51 to perform its functions. The agent information acquisition device 15 includes, for example, an in-vehicle camera, an external camera, various sensors, a communication device for communicating with an external server, and an audio input device (microphone). If the agent information acquisition device 52 contains the same devices as those included in the information notification device 3 or the FB information acquisition device 4, those devices may be configured to be shared. 【0032】 Agent device 53 is an output device that outputs information to the occupants of the vehicle VE. The operation of agent device 53 is controlled by agent control device 51. Agent device 53 is, for example, an audio output device (speaker) or a display device. If agent device 53 contains the same devices as those included in the information notification device 3 or the FB information acquisition device 4, those devices may be configured to be shared. 【0033】 In the following, the service provider entity (service entity) virtually created by Agent System 5 may be referred to as an agent. Furthermore, Agent System 5 is not a mandatory component and may not be provided in all cases. 【0034】 The learning server 6 is, in detail, a cloud server. The learning server 6 is a learning device that trains the AI ​​model used by the information notification control device 1. Note that "training" refers to machine learning. The learning server 6 provides the trained AI model (in detail, the model information of the AI ​​model) to the information notification control device 1 via a communication network such as the internet. The learning server 6 also obtains the above feedback information, or information obtained by processing the feedback information, from the information notification control device 1 and retrains the AI ​​model. 【0035】 In this embodiment, the learning device that performs AI model training (including retraining) is configured as a server, but this is merely an example. The learning device that performs AI model training may be installed in the vehicle's VE (i.e., an in-vehicle device) and does not necessarily have to be configured as a server. 【0036】 <2. Information Notification Control Device> As shown in Figure 1, the information notification control device 1 comprises a controller 11 and a memory 12. The information notification control device 1 is a so-called computer device and, in addition to the controller 11 and memory 12, includes an input / output unit (not shown). In this embodiment, the information notification control device 1 also includes a communication unit (not shown) that enables communication using a communication network such as the Internet. 【0037】 The controller 11 is configured to include an arithmetic circuit that performs calculations. The arithmetic circuit is more specifically composed of a processor. The processor is composed of, for example, a CPU (Central Processing Unit). The controller 11 may consist of one processor or multiple processors. If it consists of multiple processors, those processors should be provided to communicate with each other. 【0038】 Memory 12 consists of volatile memory and non-volatile memory. The volatile memory is specifically RAM (Random Access Memory). The non-volatile memory is specifically ROM (Read Only Memory). The non-volatile memory may also be flash memory or a hard disk drive, etc. The non-volatile memory stores a program 121 and data that can be read by the computer. 【0039】 The functions of the controller 11 are realized by the processor executing arithmetic processing according to the program 121 stored in memory 12. The number of programs 121 that realize the functions of the controller 11 may be one or more. The functions of the controller 11 may be realized by the execution of arithmetic processing according to the program 121 by an arithmetic circuit, i.e., by software, but may also be realized by other methods. At least some of the functions of the controller 11 may be realized using, for example, an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). In other words, at least some of the functions of the controller 11 may be realized by hardware using a dedicated IC or the like. Furthermore, at least some of the functions of the controller 11 may be realized by using a combination of software and hardware. 【0040】 Figure 2 is a block diagram showing the configuration of the functional units of the controller 11 of the information notification control device 1. As shown in Figure 2, the controller 11 comprises, as its functional units, an acquisition unit 110, a preprocessing unit 111, an HMI control model execution unit 112, an output control unit 113, a feedback information processing unit 114, a feedback reflection unit 115, and an agent management unit 116. Note that each functional unit 110 to 116 is a conceptual component. The function performed by one component may be distributed among multiple components. Alternatively, the functions of multiple components may be integrated into a single component. 【0041】 The acquisition unit 110 acquires information from the memory 12 and from devices (in-vehicle or external) and sensors that are configured to communicate with the information notification control device 1. The information acquired by the acquisition unit 110 includes notification information output from the information source device 2 and feedback information output from the FB information acquisition device 4. In addition, the information acquired by the acquisition unit 110 includes various types of information for input into the HMI control model, which will be described in detail later. 【0042】 The preprocessing unit 111 performs various preprocessing steps necessary for providing information notification when it becomes necessary to notify the driver, such as by obtaining notification information from the information source device 2. These preprocessing steps include, for example, determining the notification priority and deciding on the notification format. 【0043】 The priority of a notification is determined based on the urgency of the information to be notified. For example, if the information to be notified is warning of danger, the urgency will be high and the priority will be determined to be high. For example, if the information to be notified is directions such as turning left or right, the urgency will not necessarily be high and the priority will be determined to be low. The priority of a notification may be determined by creating a table in advance that associates the notification content (notification information) with the priority, storing it in memory 12, and using that table. Alternatively, as another example, the priority of a notification may be determined using an AI model that takes the notification content and various information such as the occupant (driver) profile as input and outputs the priority. Such an AI model may be generated by supervised learning performed on a large amount of data with correct labels that associate the notification content and occupant profile with the correct priority value. 【0044】 Determining the notification format includes determining the means and location of information notification. This determination process determines, for example, which of text, graphics, sound, and vibration will be used as the means of information notification. It also determines where the display, sound generation, or vibration generation will occur. Note that if there is only one notification means for each category such as display, the notification location is automatically determined by the determination of the notification means, so in such a configuration, there is no need to perform a notification location determination process. The means of information notification, etc., may be determined by a pre-prepared rule base. This rule base may use, for example, notification priority information or occupant (driver) profiles. The occupant profile is, for example, stored in memory 12 (see Figure 1) in advance. For example, in the case of a driver with poor hearing, it is decided that voice will not be used as a notification means, and other means (display or vibration) will be used instead. Also, if the notification priority is high, it is decided that multiple notification means will be used. 【0045】 The HMI control model execution unit 112 reads the model information 122 (see Figure 1) stored in the memory 12 and executes processing by the HMI control model. The model information 122 is the model information of the HMI control model, and in detail includes the structure and parameters of the HMI control model, as well as code instructions for executing processing by the HMI control model. 【0046】 Figure 3 is a schematic diagram illustrating the HMI control model 7. As described above, the HMI control model 7 is a trained AI model that performs inference processing in response to information input and outputs the inference results. In detail, the HMI control model 7 receives priority information and notification format obtained by the preprocessing unit 111, as well as various other information, and outputs notification parameters for providing information notifications that are estimated to be comfortable for the driver, as a result of inference on said input. Note that the HMI control model 7 with such functions may consist of a single AI model or multiple AI models. For example, the HMI control model 7 may include a visual model for notification control by display, an auditory model for notification control by sound, and a tactile model for notification control by vibration. 【0047】 Other types of information input to the HMI control model 7 may include, for example, notification information, scene information, DMS (Driver Monitoring System) information, occupant profiles, and environmental information. This information is obtained, for example, from devices (in-vehicle or external devices) or sensors that are communicatively connected to the information notification control device 1, or from the memory 12 provided by the information notification control device 1. 【0048】 Notification information includes, for example, "Merge Warning" and "Blind Spot Warning." Scene information includes, for example, ADAS (Advanced Driver-Assistance Systems) activation scenes and surrounding vehicle approach scenes. ADAS activation scenes include automatic braking scenes. DMS information is driver monitoring information from in-vehicle cameras, such as driver drowsiness, concentration level, or posture. Occupant profiles are occupant information such as the driver, such as age, gender, or driving history. Environmental information is environmental information inside and outside the vehicle, such as brightness information inside and outside the vehicle, noise information inside the vehicle, vibration information inside the vehicle, or weather information. 【0049】 The notification parameters output by the HMI control model 7 are, in detail, notification parameters corresponding to the notification format determined by the preprocessor 111. For example, if it is determined that text, graphics, voice, and vibration will be used as notification formats, parameters corresponding to these will be output. For example, for text and graphics, parameter information such as size, placement, and color scheme will be output. For voice, parameter information such as volume and sound quality will be output. For vibration, parameter information such as vibration intensity and vibration frequency will be output. 【0050】 Figure 4 shows specific examples of inputs and outputs in the HMI control model 7. The "input values" shown in Figure 4 are specific examples of values ​​input to the HMI control model 7. In the example shown in Figure 4, the input values ​​input to the HMI control model 7 are "5" for the input item "priority", "1" for the input item "gender", "3" for the input item "age", and "3" for the input item "in-vehicle vibration". 【0051】 Note that the input items shown in Figure 4 represent only a portion of the total input items, and input values ​​are provided similarly for items other than those shown in Figure 4. Furthermore, the numerical format for each input item (priority, gender, age, in-car vibration, etc.) is defined using numerical conversion tables, etc., to ensure that the numerical format is suitable for handling in the AI ​​model. For example, gender is defined as "1" for male, "2" for female, and "3" for others. Similarly, for age, for example, under 20 is defined as "1", 21 to 25 is defined as "2", 26 to 35 is defined as "3", 36 to 59 is defined as "4", 60 to 69 is defined as "5", 70 to 79 is defined as "6", and 80 and over is defined as "7". 【0052】 The "output values" shown in Figure 4 are specific examples of values ​​output from the HMI control model 7. The output values ​​shown in Figure 4 are obtained as a result of inputting all the input values ​​for each input item shown in Figure 4. In the example shown in Figure 4, it is assumed that the vibration generator 33 is used as a notification means, and the parameter values ​​for "vibration intensity" and "vibration frequency" are output as notification parameters. As with the input values ​​mentioned above, the format of the output parameter values ​​is defined to be a numerical format suitable for handling in the AI ​​model. 【0053】 The output also includes a confidence score, which quantifies the level of confidence. In the example shown in Figure 4, the confidence score is expressed as a percentage. The confidence score is an indicator of how confident the HMI control model 7 is in its inference. The closer the confidence score is to 100%, the more the HMI control model 7 has determined that its inference is accurate. As shown in Figure 4, the output obtained as a result of giving a certain input to the AI ​​model includes multiple inference results with different confidence scores. In the example shown in Figure 4, for the input shown in Figure 4, the output is an inference result with a confidence score of 75%, where the parameter value of "vibration intensity" is "3" and the parameter value of "vibration intensity" is "4". Also, for the same input, the output is an inference result with a confidence score of 10%, where the parameter value of "vibration intensity" is "4" and the parameter value of "vibration intensity" is "4". Also, for the same input, the output is an inference result with a confidence score of 5%, where the parameter value of "vibration intensity" is "3" and the parameter value of "vibration intensity" is "3". Furthermore, the number of inference results with different confidence scores that the HMI control model 7 outputs for a given input is not limited to three, but may vary as appropriate. 【0054】 The output control unit 113 (see Figure 2) controls the information notification device 3 according to the notification signal indicating the content of the notification information and the notification parameters output by the HMI control model 7. In other words, the information notification control device 1 uses an AI model to control information notifications. The computer (controller 11) executes a notification control method that uses an AI model to control information notifications. 【0055】 In detail, the output control unit 113 controls the display device 31, the audio output device 32, and the vibration generator 33 (all of which are shown in Figure 1) according to the notification parameters output by the HMI control model 7. As a result of this control, the driver receives information notification from at least one of the display device 31, the audio output device 32, and the vibration generator 33. 【0056】 In this embodiment, the output control unit 113 normally controls the information notification device 3 (notification control) using the most reliable inference result obtained by the HMI control model 7. The most reliable inference result is the notification parameter with the highest reliability score among several notification parameters with different reliability scores obtained when a certain input value is input to the HMI control model 7. In the example shown in Figure 4, the notification parameter with the highest reliability score is vibration intensity "3" and vibration frequency "4", which has a reliability score of 75%, and the output control unit 113 controls the vibration generator 33 using this notification parameter. 【0057】 However, when the output control unit 113 determines that a specific timing has occurred, it controls the information notification device 3 (notification control) according to the inference result with a lower confidence level than the most confident inference result among the multiple inference results with different confidence levels output by the HMI control model 7. Examples of low-confidence inference results include the second and third highest confidence scores mentioned above. For example, in a configuration where the third highest confidence score inference result is used, as shown in the example in Figure 4, the output control unit 113 controls the vibration generator 33 using vibration intensity "3" and vibration frequency "3", which have a confidence score of 5%, as notification parameters. Here, an example of using the third highest confidence score inference result has been given, but this is merely an example. The data with which confidence score is used can be determined using, for example, a fixed method, a random selection method, or a sequential method. A fixed method is, for example, a method in which the data with the third highest confidence score is selected each time. A random selection method is, for example, a method in which one is randomly selected from the data with the second to fifth highest confidence scores. When using a random selection method, it is preferable to exclude data with excessively low confidence scores from selection. The sequential selection method involves sequentially changing the criteria for selecting data. For example, if the data with the second-highest confidence score was selected in the previous selection, the data with the third-highest confidence score would be selected in the next selection, and so on, changing the selection criteria sequentially. More specifically, an example would be selecting data while changing the selection criteria, such as second-highest → third-highest → fourth-highest → fifth-highest → second-highest →... 【0058】 A specific timing is a suitable timing for acquiring feedback information for the HMI control model 7 (AI model). In other words, when the information notification control device 1 determines that it is a suitable timing for acquiring feedback information for the AI ​​model, it performs notification control according to the inference result with a lower confidence level than the inference result with the highest confidence level among multiple inference results with different confidence levels output by the AI ​​model. The information notification control program makes the computer function as a means to do the following: When it determines that it is a suitable timing for acquiring feedback information for the AI ​​model, it performs the notification control according to the inference result with a lower confidence level than the inference result with the highest confidence level among multiple inference results with different confidence levels output by the AI ​​model. 【0059】 With this configuration, information notifications, delivered at a time suitable for acquiring feedback information, will intentionally include information that may not be optimal for the user. Users who receive notifications of potentially inadequate information are more likely to take action to provide feedback on the notification. Furthermore, the user's reaction to notifications of potentially inadequate information is more likely to be useful feedback information for training the AI ​​model (HMI control model 7). For this reason, the efficiency of acquiring user feedback information for the AI ​​model (HMI control model 7) used for notification control can be improved. 【0060】 In other words, in this embodiment, the following (a) and (b) are performed as methods for collecting feedback information by computer for the AI ​​model (HMI control model 7) that performs notification control. (a) Among the multiple inference results with different confidence levels output by the AI ​​model, notification control will be performed according to the inference result with a lower confidence level than the inference result with the highest confidence level. (b) Collecting user responses as feedback information to notification controls that were implemented based on unreliable inference results. Furthermore, feedback information is collected not only in the case of notification control that takes place at specific times, but also in the case of normal notification control as described above. 【0061】 Furthermore, the optimal timing for acquiring feedback information for the HMI control model 7 (AI model) is determined based on the priority of the information notification. Specifically, this timing is performed when a notification with a low priority is issued. Whether or not a notification has a low priority can be determined, for example, by whether or not the notification priority given by the preprocessing unit 111 is less than a predetermined threshold. Here, low priority is synonymous with low urgency. By configuring the system to notify information that may not be optimal for the user (driver) only in the case of low-priority (urgency) notifications, it is possible to suppress adverse effects on driving and distrust of the user towards the driver assistance system 100. 【0062】 Furthermore, the timing suitable for acquiring feedback information for the HMI control model 7 is not limited to when low-priority notifications are issued. Such timing may include, for example, when driving on a road where monotonous driving occurs, such as a highway, when driving situations that may cause the driver to feel bored, such as traffic jams, or when driver drowsiness is detected. When issuing notifications for collecting feedback information at these timings, it is preferable to inform the driver in advance, for example using the agent system 5, that they would like their cooperation in collecting information for the AI ​​model's learning. In such a configuration, since the driver has been informed in advance that a test notification for collecting feedback information will be issued, the content of the notification can be determined regardless of priority. By issuing such test notifications, the learning opportunities for the HMI control model 7 can be increased, and the driver's alertness can be expected. 【0063】 Whether or not the vehicle is traveling on a highway may be determined based on GPS (Global Positioning System) information and map information. Whether or not the road being traveled is congested may be determined by obtaining congestion information using a road traffic information communication system. Driver drowsiness may be determined by photographing the driver with an in-vehicle camera and processing the image data obtained from the photograph. 【0064】 The feedback information processing unit 114 processes the feedback information acquired from the FB information acquisition device 4. Specifically, the feedback information processing unit 114 performs processing to make the acquired feedback information usable for output correction of the HMI control model 7 and as data for relearning the HMI control model 7. Details of the processing by the feedback information processing unit 114 will be described later. 【0065】 The feedback reflection unit 115 uses the data processed by the feedback information processing unit 114 to process the input information to be input to the HMI control model 7, for example, so that notification parameters reflecting the feedback information are output from the HMI control model 7. This processing is just one example of the output correction of the HMI control model 7. For example, if the processing of the feedback information indicates that the volume of the audio output device 32 is low, the input values ​​of input items that affect the volume among the input values ​​input to the HMI control model 7 may be processed (corrected). An example of an input item that affects the volume is the in-car noise, and the HMI control model 7 may be configured to input an input value indicating that the in-car noise is high so that the volume increases. The feedback reflection unit 115 also stores the data processed by the feedback information processing unit 114 as information for updating (retraining) the HMI control model 7. The stored data is used, for example, by being sent to the learning server 6 (see Figure 1). The feedback reflection unit 115 may also store the data processed by the feedback information processing unit 114 in the form of learning data. 【0066】 The agent management unit 116 is configured to communicate with the agent control device 51 via CAN (Controller Area Network) communication or the like, and controls the agent control device 51. A detailed example of how to use the agent system 5, including the agent control device 51, will be described later. 【0067】 <3. Details of processing in the information notification control device> Next, we will explain the details of the processing performed by the information notification control device 1. 【0068】 [3-1. Notification Control Processing] Figure 5 is a flowchart illustrating the flow of notification control processing executed by the information notification control device 1. This flowchart shows the technical content of a computer program (information notification control program) that enables a computer to implement the information notification control method of this embodiment. This computer program can be stored on various non-volatile recording media readable by a computer and provided (sold, distributed, etc.). This computer program may consist of only one program, or it may consist of multiple programs working together. 【0069】 The process shown in Figure 5 is executed as appropriate when the vehicle VE is powered on and the information notification control device 1 becomes capable of executing notification control processing. 【0070】 In step S1, the controller 11 (acquisition unit 110) acquires notification information from the information source device 2. As described above, the notification information is information related to the operation of the vehicle VE, such as information warning of danger or information urging caution. Once the notification information is acquired, the process proceeds to the next step S2. 【0071】 In step S2, the controller 11 (preprocessing unit 111) performs various preprocessing steps necessary for notification control using the HMI control model 7. Specifically, as described above, this includes determining the notification priority and deciding on the notification format. Once the preprocessing is complete, the process proceeds to the next step S3. 【0072】 In step S3, the controller 11 (HMI control model execution unit 112) executes processing by the HMI control model 7. That is, the controller 11 estimates the notification parameters for controlling notifications from the information notification device 3. Once the notification parameters are output from the HMI control model 7, the process proceeds to the next step S4. 【0073】 In step S4, the controller 11 (output control unit 113) determines whether it is a specific timing. The specific timing is, as described above, a timing suitable for obtaining feedback information for the HMI control model 7, and a specific example is the timing for issuing a notification that is not of high priority (urgency). In this case, it is preferable that the notification occurs once every n (for example, 5) notifications of low priority. Examples of notifications of low priority include notifications regarding periodic inspections. If it is not a specific timing (No in step S4), the process proceeds to step S5. If it is a specific timing (Yes in step S4), the process proceeds to step S6. 【0074】 In step S5, the controller 11 (output control unit 113) performs normal notification control. That is, the controller 11 controls the information notification device 3 using the most reliable notification parameters obtained in the HMI control model 7. As a result, information is notified from the information notification device 3 to the driver. 【0075】 In step S6, the controller 11 (output control unit 113) performs feedback (FB) priority notification control aimed at increasing opportunities for driver feedback. Specifically, the controller 11 controls the information notification device 3 according to notification parameters with lower reliability than the most reliable notification parameter among the multiple types of notification parameters with different reliability levels output by the HMI control model 7. For example, notification control is performed using notification parameters with the second or third highest reliability levels. As a result, information is notified from the information notification device 3 to the driver. In this case, the driver is more likely to receive information notifications in an inoptimal form. 【0076】 A driver that receives information in a potentially suboptimal format is more likely to take action to provide feedback on the information notification compared to a driver that receives information in an optimal format. Furthermore, a user's response to information in a potentially suboptimal format is more likely to provide useful feedback for the HMI control model 7's learning process. As a result, an increase in opportunities to acquire feedback information for the HMI control model 7 can be expected. It is also preferable that, even in the case of normal notification control, the user can take action to provide feedback, and that feedback information is collected in this case as well. This configuration allows for the collection of feedback information in a balanced manner, enabling better control. 【0077】 Figure 6 is a flowchart showing a modified version of the notification control process executed by the information notification control device 1. In the modified version shown in Figure 6, if the result in step S4 is "Yes", the notification type notification process (step S7) is executed before the FB priority notification control (step S6) is executed. This is different from the flow shown in Figure 5. 【0078】 The notification process in step S7 is a notification process that notifies that a notification will be made for the purpose of collecting feedback information. In other words, the modified information notification control device performs a notification process that notifies that a notification will be made for the purpose of collecting feedback information (that a notification format for the purpose of collecting feedback information will be used) when notification control is performed according to an unreliable inference result. With this configuration, even if a notification that may not be optimal for the driver is made by FB priority notification control, it is possible to suppress the driver from developing distrust towards the notification control by the information notification control device 1. 【0079】 In detail, the notification method notification process may be configured to inform the system that potentially suboptimal notification methods will be tested with lower-priority (urgency) notifications for model training. Such notifications may be, for example, voice notifications using the voice output device 32 or screen notifications using the display device 31. Alternatively, the notification method may use both voice and display. 【0080】 In this modified example, the notification process is performed before executing the FB priority notification control (processing in step S6), but it is also possible to configure the system so that the notification process is performed after executing the FB priority notification control. 【0081】 [3-2. Processing of Feedback Information] (3-2-1. Overview) As described above, the information notification control device 1 acquires feedback information related to notification control via the FB information acquisition device 4. In other words, the information notification control device 1 acquires feedback information for the HMI control model 7. The controller 11 (feedback information processing unit 114) performs information processing on the acquired feedback information. 【0082】 The feedback information processing unit 114 (i.e., the information notification control device 1) uses the feedback information to determine whether the driver (user) is positive or negative regarding the notification control of the HMI control model 7. By classifying the feedback information as positive or negative in this way, the feedback information can be easily used for output correction of the HMI control model 7 and for relearning the HMI control model 7. 【0083】 In this embodiment, the feedback information processing unit 114 (i.e., the information notification control device 1) determines whether the driver is positive or negative towards the notification control of the HMI control model 7 based on the following two indicators. One of the two indicators is the state of implementation of the notification content after the information is notified. The other of the two indicators is the detection result of a negative response of the driver to the notification control. As will be described in detail later, with this configuration, it is possible to efficiently and appropriately determine whether the driver is positive or negative towards the notification control. 【0084】 (3-2-2. Detailed example) Figure 7 is a flowchart illustrating the processing flow related to feedback information executed by the information notification control device 1. This flowchart shows the technical content of a computer program (FB information processing program) that enables a computer to implement the feedback information processing method of this embodiment. This computer program can be stored on various non-volatile recording media readable by a computer and provided (sold, distributed, etc.). This computer program may consist of only one program, or it may consist of multiple programs working together. 【0085】 The process shown in Figure 7 is executed each time a notification is sent using the information notification device 3. 【0086】 In step S11, the feedback information processing unit 114 acquires the feedback information transmitted from the FB information acquisition device 4. Once the feedback information is acquired, the process proceeds to the next step S12. 【0087】 In step S12, the feedback information processing unit 114 determines whether the information notification that is the subject of the feedback is a first-determinable information notification. The first determination and the second determination, described later, are convenient names to distinguish between two determination processes performed to determine whether the driver is positive or negative towards the information notification. The first determination is a determination regarding the realization state of the notification content after the information notification, and more specifically, it is a determination of whether the state required by the information notification has been achieved. 【0088】 The first determination is that it may or may not be possible to perform the action, depending on the content of the notification. Taking this into consideration, the processing in step S12 is carried out. Examples of information notifications that can be determined in the first determination include information notifications instructing to stop or information notifications instructing to slow down before a pedestrian crossing. Examples of information notifications that cannot be determined in the first determination include information notifications prompting caution at merging points when there are merging lanes ("Merging Caution") or information notifications informing of periodic inspections. These notifications are classified as being on the side of being impossible to determine in the first determination because it is difficult to define or determine what state should be in when the notification is received. 【0089】 Regarding the feasibility of the first determination, a table may be prepared in advance that categorizes each notification content as either "Yes" or "No," and the determination can be made by referring to this table. If the first determination is possible (Yes in step S12), the process proceeds to the next step S13. If the first determination is not possible (No in step S12), the process skips the next step S13 and proceeds to step S14. 【0090】 In step S13, the feedback information processing unit 114 performs a first determination. Specifically, it determines whether the state required by the information notification has been reached. The determination of whether the state required by the information notification has been reached differs depending on the content of the information notification. For example, if the information notification is for a temporary stop, it is determined whether the vehicle VE has stopped at the temporary stop location (whether its speed has become zero). Such a determination can be made based on GPS information obtained from the vehicle VE's position, map information stored in memory 12, and the vehicle VE's speed information. If it is determined that the state required by the information notification has been reached (Yes in step S13), the process proceeds to the next step S14. If it is determined that the state required by the information notification has not been reached (No in step S13), the process proceeds to step S21. 【0091】 In step S14, the feedback information processing unit 114 performs the process of acquiring information for the second determination. The second determination determines whether the driver showed a negative reaction to the information notification, and in detail, the determination is made by detecting negative statements or emotions such as "annoying". In step S14, the information necessary to make such a determination is acquired. This information is included in the information output from the FB information acquisition device 4 described above. 【0092】 For example, information about the driver's facial expressions, biometric information, or voice information may be acquired as information for the second decision. Facial expressions may be information from an in-car camera that photographs the driver. Biometric information may be, for example, brain waves or heart rate, and this information may be from biosensors such as brain wave sensors or heart rate (pulse) sensors worn by the driver. Voice information may be information collected by microphones placed around the driver. Once the information for the second decision is acquired, the process proceeds to the next step S15. 【0093】 In step S15, the feedback information processing unit 114 performs a second determination based on the second determination information acquired earlier, determining whether the driver's response to the information notification is a negative response. Whether or not it is a negative response may be determined, for example, by facial expression analysis of an image of the driver's face, emotion estimation based on biometric information, or voice information analysis. 【0094】 For example, if facial expression analysis using an image estimates that the driver is making an angry or unpleasant expression, the driver's reaction may be judged as negative. Recognition of the driver's facial expression may be performed using known image processing techniques. The image recognition algorithm may be configured to be executed using a known AI model. 【0095】 Furthermore, for example, if emotion estimation based on biometric information estimates the driver's emotion to be "anger" or "displeasure," the driver's response may be judged as negative. Publicly known techniques may be used for emotion estimation based on biometric information. For example, it is publicly known that a person's emotions can be estimated using brain waves and heart rate, and emotion estimation may be performed using such publicly known techniques. An emotion estimation model configured as an AI model may also be used for emotion estimation. 【0096】 Furthermore, for example, if negative statements or sounds are extracted through voice analysis based on voice information, the driver's response may be determined to be negative. Negative sounds include, for example, tongue clicks. Known speech recognition and natural language processing methods may be used to extract the driver's statements through voice analysis. Whether the extracted statements are negative or not can be determined, for example, by creating a list of negative statements in advance, and determining that the extracted statements are negative if they match any of the statements in the list. 【0097】 If the second determination process is performed, the process proceeds to the next step, S16. In step S16, it is confirmed whether or not the second determination was impossible. If the second determination is impossible, it means that it is not possible to determine whether or not the driver's response is a negative response. Examples of such a state include situations where it is not clear whether or not the response can be classified as a negative response, and a more detailed example is when it is estimated to be a negative response, but the accuracy (confidence) of that estimation is low. If the second determination was impossible (Yes in step S16), the process proceeds to step S19. If the second determination was not impossible (No in step S16), the process proceeds to step S17. 【0098】 In step S17, the feedback information processing unit 114 checks whether the driver gave a negative response, as a result of the determination when a second determination was possible. If the driver gave a negative response (Yes in step S17), the process proceeds to step S21. If the driver did not give a negative response (No in step S17), the process proceeds to step S18. 【0099】 In step S18, the feedback information processing unit 114 determines that the acquired feedback information is positive data and collects it as positive data. In addition to the information that the evaluation result of the information notification was positive, the positive data includes the following two pieces of information: the input information that was input to the HMI control model 7 when the information notification received a positive evaluation, and the output information of the HMI control model 7 for that input information. Once the positive data is collected, the process shown in Figure 7 is completed. 【0100】 In step S19, the feedback information processing unit 114 decides to execute agent processing, and under the management of the agent management unit 116, agent processing using the agent system 5 is executed. During agent processing, the agent confirms with the driver whether or not they were negative about the information notification. For example, the agent asks the driver a question such as, "Was it bothersome?" The agent system 5 receives the driver's response (operation of the touch panel, voice operation using voice recognition, etc.), processes it as appropriate, and transmits it to the information notification control device 1. In this way, the execution of agent processing allows the driver's direct feedback on the information notification to be obtained. Upon completion of agent processing, the process proceeds to step S20. 【0101】 In step S20, the feedback information processing unit 114 determines from the driver's response information obtained from the agent system 5 whether the driver's response to the information notification was negative or not. If the result of this determination is that the driver's response to the information notification is negative (Yes in step S20), the process proceeds to step S21. If the driver's response to the information notification is not negative (No in step S20), the process proceeds to step S19, and the feedback information is collected as positive data. 【0102】 In other words, the information notification control device 1 determines whether the driver is positive or negative towards notification control based on the confirmation result from the agent. With this configuration, it is possible to collect accurate information on whether the driver is positive or negative towards notification control. This confirmation process using the agent may be performed every time FB information is acquired. However, as in this embodiment, by configuring the system to perform confirmation processing by the agent only when it is not possible to clearly classify whether the driver's response is negative or not, the frequency of confirmation processing for the driver can be reduced, and the possibility that the driver will find the confirmation process bothersome can be reduced. 【0103】 Furthermore, a state in which it is not possible to determine whether the above driver response is a negative response or not (i.e., a second state of indetermination) also includes a state in which it is not clear that the response can be classified as positive. A detailed example of this state may include a situation where the response is presumed to be positive, but the accuracy (confidence) of that presumption is low. For this reason, confirmation processing by the agent may be performed in such cases as well. However, in order to reduce the frequency of agent confirmation, in such cases, it may be decided not to use the feedback information being processed as feedback data, and thus it may be excluded from the agent's confirmation processing. 【0104】 In step S21, the feedback information processing unit 114 collects the acquired feedback information as negative data. The negative data includes information that the evaluation result of the information notification was negative, as well as the following two pieces of information: the input information that was input to the HMI control model 7 when the information notification received a negative evaluation, and the output information of the HMI control model in response to that input information. Once the negative data is collected, the process proceeds to the next step S22. 【0105】 In step S22, the feedback information processing unit 114 performs a detailed analysis on the negative data to obtain the reasons why it was deemed negative. In other words, the information notification control device 1 performs a process to obtain the problems of the notification control from the negative data. Through this process, the negative data can be made useful for correcting the output of the HMI control model 7 and for relearning the HMI control model 7. 【0106】 The detailed analysis process may be performed automatically by the feedback information processing unit 114, or it may be performed with the help of the agent system 5. In the former configuration, for example, the system may estimate how inappropriate the output result of the HMI control model 7 was by comparing the information notification being processed with the most recent notification with similar inputs. 【0107】 In the latter configuration, the agent may ask "what went wrong" regarding negative data. The system may ask vague questions such as "what went wrong" and repeatedly interact with the driver until it obtains information on which notification method or format was problematic and to what extent, at which point the interaction ends. For example, once information such as "the sound was very quiet" is obtained, the interaction for detailed analysis may end. 【0108】 However, alternatively, a standardized format for questions may be prepared in advance, and questions may be asked according to this format regarding which notification method or format was inadequate and to what extent, and responses regarding what was inadequate and to what extent may be obtained. Alternatively, the inadequacy of the notification method or format may be determined from the timing of the information notification and the timing of the acquisition of the feedback information. For example, if the driver makes a negative statement immediately after the occurrence of vibration, it may be determined that the driver was negative about "vibration." 【0109】 The information obtained through the detailed analysis process is added to the previous negative data, and the data after this information addition becomes the final negative data. Once the process in step S22 is completed, the process shown in Figure 7 is finished. 【0110】 In the process shown in Figure 7, the first determination may be omitted, and the processes from the second determination onward may be performed. However, if the configuration includes the first determination, it may be possible to perform only the first determination, which is a relatively simple process, and omit the second determination. For this reason, it is expected that the efficiency of processing feedback information will be improved by configuring the system to perform the first determination before the second determination. 【0111】 Furthermore, the information notification control device 1 of this embodiment collects notification control information that the user has positively received as positive data, and notification control information that the user has negatively received as negative data. With this configuration, not only positive data but also negative data can be used to retrain the HMI control model 7, and data for retraining can be collected efficiently. 【0112】 <4. Learning (Machine Learning)> Next, the learning (machine learning) of the HMI control model 7 will be described. As described above, in this embodiment, the learning device that performs the learning of the HMI control model 7 is the learning server 6. However, the learning device may be an in-vehicle device mounted on the vehicle VE. 【0113】 Figure 8 is a block diagram illustrating the schematic configuration of the learning server 6. As shown in Figure 8, the learning server 6 comprises a controller 61, a memory 62, and a communication unit 63. The controller 61 is configured to include a processor, such as a CPU. The memory 62 is configured to include volatile memory and non-volatile memory. The functions of the controller 61 are realized by the processor executing arithmetic processing according to the program stored in the memory 62. The communication unit 63 is configured as a communication interface having an interface circuit for connecting to a communication network (not shown), such as the Internet. 【0114】 The controller 61 trains the HMI control model 7 using pre-prepared initial training data. The training of the HMI control model 7 is, in detail, supervised learning. Any known supervised learning method may be used. The initial training data is, in detail, a dataset composed of multiple data points. Each data point that makes up the initial training data is data with a ground truth label. The data with a ground truth label has a ground truth label that is used to compare the input value input to the HMI control model 7 with the output value output from the HMI control model 7. The ground truth label is a notification parameter that the user (driver) perceives as a comfortable form of information notification. 【0115】 In addition to the initial learning described above, the controller 61 uses feedback data collected by the information notification control device 1 to retrain the trained HMI control model 7. By repeatedly retraining the HMI control model 7, the information notification control device 1 can be made to more closely match the user's preferences in providing information notifications. 【0116】 Retraining is performed once a certain amount of feedback data has been accumulated. The positive and negative data mentioned above are used as feedback data. The process of converting the positive and negative data into data for retraining may be performed by the information notification control device 1 or by the learning server 6. In some cases, it may be performed by a device other than either the information notification control device 1 or the learning server 6. 【0117】 The retraining performed by the controller 61 using feedback data may be a known supervised learning method, similar to the initial training described above. The labeled data used for supervised learning includes data obtained from positive data and data obtained from negative data. The labeled data obtained from positive data and the labeled data obtained from negative data will be described below. 【0118】 Figure 9A shows the ground truth labeled data obtained from positive data. More specifically, the table portion shown in Figure 9A represents the ground truth labeled data. As shown in the table, the ground truth labeled data has a ground truth label used to compare the input value input to the HMI control model 7 with the output value output from the HMI control model 7. 【0119】 In Figure 9A, notification A was issued with a vibration intensity of "3" output from the HMI control model 7, and positive feedback (FB) of "no problem" was given in response to the notification. In other words, the above positive data is obtained in Figure 9A. When such positive data is obtained, it can be determined that the control value (notification parameter) "3" used when notification A was issued was correct, and therefore the correct value for vibration intensity is "3". For this reason, "3" is given as the correct label for vibration intensity. The input value in the data with the correct label is the input value that was input to the HMI control model 7 when notification A was issued. 【0120】 Figure 9B shows the ground truth labeled data obtained from negative data. More specifically, the table portion shown in Figure 9B represents the ground truth labeled data. As shown in the table, the ground truth labeled data has a ground truth label used to compare the input value input to the HMI control model 7 with the output value output from the HMI control model 7. This is the same as in the case of positive data described above. 【0121】 In Figure 9B, notification A is performed with a vibration intensity of "3" output from the HMI control model 7, and negative feedback (FB) indicating "vibration intensity is strong" is given in response to the notification. In other words, negative data is obtained in Figure 9B. When such negative data is obtained, it can be determined that the control value (notification parameter) "3" used when notification A was issued was incorrect, and therefore the correct value for vibration intensity is not "3". In the example shown in Figure 9B, the information "intensity is strong" is obtained as a result of the detailed analysis process for the above negative data (see step S22 in Figure 7). Taking this into consideration, an intensity of "2", which is weaker than intensity "3", is given as the correct label for vibration intensity. The input values ​​in the data with the correct label are the input values ​​that were input to the HMI control model 7 when notification A was issued. 【0122】 In Figure 9B, the vibration intensity used for notification A is reduced by one level, taking into account the user's feedback that the vibration was "too strong." However, if the detailed analysis process described above yields user feedback such as "very strong," the vibration intensity used for notification A may be reduced by two levels. In other words, the correct label value should be determined appropriately according to the information obtained from the detailed analysis process described above. 【0123】 <5. Points to note> The various technical features disclosed in the embodiments for carrying out the invention as described herein can be modified in various ways without departing from the spirit of the technical creation. Furthermore, the multiple embodiments and modifications disclosed in the embodiments for carrying out the invention as described herein may be combined to the extent possible. [Explanation of symbols] 【0124】 1. Information notification control device 2... Information source device 3... Information notification device 5. Agent System 7. HMI control model (AI model) 121... Program 100...Driving assistance system VE...vehicle

Claims

[Claim 1] An information notification control device that uses an AI model to control information notification, comprising a controller, The aforementioned controller, An information notification control device that, when it is determined that it is a suitable time to acquire feedback information for the AI ​​model, performs the notification control according to the inference result with a lower confidence level than the inference result with the highest confidence level among a plurality of inference results with different confidence levels output by the AI ​​model. [Claim 2] The information notification control device according to claim 1, wherein the controller determines the timing based on the priority of notification of the information. [Claim 3] The information notification control device according to claim 1, wherein the controller performs a notification process to notify that a notification for the purpose of collecting feedback information will be made when the notification control is performed in accordance with the unreliable inference result. [Claim 4] The aforementioned controller, The aforementioned feedback information is acquired, An information notification control device according to any one of claims 1 to 3, wherein the user determines whether the AI ​​model's response to the notification control is positive or negative, using the aforementioned feedback information. [Claim 5] The information notification control device according to claim 4, wherein the controller collects information on the notification control that the user has given a positive response to as positive data, and information on the notification control that the user has given a negative response to as negative data. [Claim 6] The information notification control device according to claim 5, wherein the controller performs a process to acquire problems of the notification control with respect to the negative data. [Claim 7] The information notification control device according to claim 4, wherein the controller determines whether the user is positive or negative towards the notification control based on the degree to which the notification content is realized after the notification of the information, and the detection result of the user's negative response to the notification control. [Claim 8] The information notification control device according to claim 4, wherein the controller determines whether the user is positive or negative towards the notification control based on the confirmation result by the agent. [Claim 9] The information notification control device according to claim 5, wherein when the controller collects the positive data, it generates data with correct labels, using the output value output by the AI ​​model when notifying the information as the correct label, as training data. [Claim 10] The information notification control device according to claim 5, wherein, when the controller collects the negative data, it generates data with correct labels as training data, using corrected values ​​obtained by correcting the output values ​​output by the AI ​​model when notifying the information as the correct labels. [Claim 11] The aforementioned controller, The negative data is processed to acquire the problems with the notification control. An information notification control device according to claim 10, which determines the corrected value based on the aforementioned problems. [Claim 12] A program that causes a computer to execute an information notification control method that uses an AI model to control information notification, The aforementioned computer, When it is determined that it is a suitable time to obtain feedback information for the AI ​​model, the notification control is performed according to the inference result with a lower confidence level than the inference result with the highest confidence level among the multiple inference results with different confidence levels output by the AI ​​model. An information notification and control program that functions as a means to carry out [the action]. [Claim 13] A driver assistance system that assists in driving a vehicle, An information notification device that provides information to support the aforementioned operation, A source information device that serves as the source of the aforementioned information, An information notification control device that, upon acquiring the information from the information source device, uses an AI model to perform notification control to the information notification device, Equipped with, The aforementioned information notification control device is A driver assistance system that, when it determines that it is a suitable time to acquire feedback information for the AI ​​model, performs the notification control according to the inference result with a lower confidence level than the inference result with the highest confidence level among multiple inference results with different confidence levels output by the AI ​​model. [Claim 14] A computer-based method for collecting feedback information for an AI model that performs notification control, Among the multiple inference results with different confidence levels output by the AI ​​model, the notification control is performed according to the inference result with a lower confidence level than the inference result with the highest confidence level. An information gathering method that collects user responses to notification control performed according to the low-reliability inference results as feedback information.

Images