JP7334057B2 - A system that predicts driving hazards - Google Patents

Description

The present disclosure relates to predicting driving hazard risk.

BACKGROUND ART In recent years, there has been a problem of traffic accidents caused by drivers' health conditions and fatigue in logistics trucks, long-distance buses, and the like. In order to improve safety, biosensors that monitor the state of the driver while driving, and technology that measures vehicle conditions such as inter-vehicle distance and speed in real time are being applied.

For example, as a background technology of the present disclosure, there is Japanese Patent Laying-Open No. 2017-27414 (Patent Document 1). Patent Document 1 states, "The safe driving support device 100 receives lifelog data from the lifelog sensor 300 of the mobile phone terminal 200, receives driving behavior data from the telematics device 400 of the moving object 400M, and receives accident data from the insurance server 500. Data is received, life log data, driving behavior data, and accident data are used to identify the degree of danger when a living organism drives a vehicle, and warning information for avoiding accidents is sent to the mobile phone terminal 200 and the telematics device 400. provide.” (e.g., abstract).

JP 2017-27414 A

In Patent Document 1, an alert is issued based on a remarkable abnormal value in life log data that is seen when an accident occurs, and it was not possible to predict fine daily fluctuations in accident risk. Moreover, Patent Literature 1 is based on general accident data, and cannot reflect the differences between individuals and the characteristics of each individual. Therefore, there is a demand for a technique that can appropriately predict danger risks in driving for each individual.

A system according to one aspect of the present disclosure includes one or more processors that operate according to a program stored in at least one of the one or more storage devices. The one or more processors acquire a prediction model that predicts the number of dangerous driving operations based on biometric data, the prediction model corresponding to past biometric data of the individual, and using the acquired prediction model, Based on biometric data measured prior to the individual driving the vehicle, a predicted level of danger risk in the individual driving the vehicle is determined.

According to one aspect of the present disclosure, it is possible to appropriately predict danger risks in driving for each individual.

A system configuration example including an operation management assistance system is shown. 1 shows an example of the hardware configuration of an operation management assistance system. 4 shows a configuration example of a biometric data DB; It includes a configuration example of the subjective evaluation data DB. 4 shows a configuration example of a dangerous driving operation data DB; 4 shows a configuration example of association data. Fig. 2 shows a flow chart for a biometric data acquisition device to acquire heart rate variability parameters; An example of heart rate variability is shown. 4 shows the time variation of the peak interval RR. An example of a frequency spectrum of time-varying peak intervals RR is shown. 4 shows a flow chart of an example of processing by the subjective evaluation data collection device. Example images of two VAS are shown. FIG. 4 is a diagram for explaining generation of association data, generation of a trained prediction model, and prediction of a danger risk level by the trained prediction model; 4 shows a flowchart of a processing example of an operation management assistance system and data related thereto; 4 shows an example of a flow chart for selection of a biological model by a biological model selection program. 4 shows an image example of a report presented by the prediction result presentation program. 4 shows an image example of a report presented by the prediction result presentation program.

Embodiments of the present invention will be described below with reference to the accompanying drawings. It should be noted that this embodiment is merely an example for realizing the present invention and does not limit the technical scope of the present invention.

In the following, a system for predicting driving hazards is disclosed. An operation management assistance system according to one embodiment will be described below. The operation management assistance system can assist operation management of vehicles in the land transportation industry, such as trucks, buses, and taxis. Note that the prediction of danger risks in driving according to the present disclosure is not limited to operation management of multiple vehicles, and can be applied to prediction of danger risks before driving for each of one or more drivers.

FIG. 1 shows a system configuration example including an operation management assistance system. The operation management assistance system 2, the driving data collection device 15, the prediction result display terminal 31, the biological data collection device 32, and the subjective evaluation data collection device 33 can each communicate via a network.

The driving data collection device 15 is mounted on the vehicle 1 , acquires measurement data (in-vehicle sensor data or driving data) of sensor devices mounted on the vehicle 1 , and transmits the data to the operation management assistance system 2 . In the example of FIG. 1 , the sensor devices mounted on the vehicle 1 include a position sensor (GNSS) 11 based on GNSS (Global Navigation Satellite System), an inter-vehicle distance sensor 12 , a speedometer 13 and an acceleration sensor 14 . The driving data collection device 15 may be installed outside the vehicle 1 and receive data from an in-vehicle sensor device via a network.

The prediction result display terminal 31 displays the prediction result of danger risks during driving by the operation management assistance system 2 . The details of the risk prediction method and the display image of the prediction result will be described later.

The biological data collection device 32 acquires the measured data of the driver from the biological measuring instrument and transmits the data to the operation management assistance system 2 . In the example of FIG. 1, the bioinstruments used include a heart rate monitor 34 , a thermometer 35 and a blood pressure monitor 36 . A driver or an administrator measures the driver's biological data using a biological measuring instrument before and after driving for a day, for example.

The subjective evaluation data collection device 33 acquires the driver's subjective evaluation and transmits it to the operation management assistance system 2 . For example, the driver inputs a subjective evaluation of his/her physical condition to the subjective evaluation data collection device 33 when measuring the driver's biological data. Details of subjective evaluation input will be described later.

The operation management assistance system 2 holds a plurality of processing programs and data processed by the plurality of programs. The programs executed by the operation management assistance system 2 include a risk determination program 221, a biological model selection program 222, a prediction model training program 223, and a prediction result presentation program 224.

The danger determination program 221 analyzes the vehicle-mounted sensor data transmitted from the driving data collection device 15 to detect dangerous driving operations. The biological model selection program 222 selects a biological model suitable for the driver. Details of the biological model will be described later. Predictive model training program 223 trains the predictive models included in the selected biological model to generate trained predictive model 253 . As will be described below, the trained predictive model 253 predicts the driver's driving hazard risk level. The prediction result presentation program 224 presents information about the predicted danger risk level of the driver.

Data held by the operation management assistance system 2 includes history data 24 , biological model DB 251 , association data 252 , trained prediction model 253 , and pre-driving data 26 . Historical data 24 includes historical data history for one or more vehicles and one or more drivers, and in the example described below, historical data history for multiple vehicles and multiple drivers. . The history data 24 specifically includes a biological data database (DB) 241, a subjective evaluation data DB 242, a dangerous driving operation data DB 243, and an in-vehicle sensor data DB 244.

The biometric data DB 241 stores biometric data received from the biometric data collection device 32 . The biometric data DB 241 stores biometric data histories measured for a plurality of drivers. In the example described below, values that can change before and after driving, such as heart rate variability and body temperature, are measured.

The subjective evaluation data DB 242 stores subjective evaluation data received from the subjective evaluation data collection device 33 . The subjective evaluation data indicates the subjective evaluation of the driver's own physical condition. The subjective evaluation data DB 242 stores subjective evaluation data histories of multiple drivers. As will be described later, the personal data in the biometric data DB 241 and the subjective evaluation data DB 242 are referred to for selecting a driving danger risk prediction model.

The in-vehicle sensor data DB 244 stores in-vehicle sensor data (driving data) received from the driving data collection device 15 . The in-vehicle sensor data DB 244 stores a history of in-vehicle sensor data of multiple drivers. The dangerous driving operation data DB 243 stores a history of dangerous driving operations in driving (dangerous driving operation data history). The risk determination program 221 detects dangerous driving operations in each driving unit (for example, one driver's driving in one day) from the data stored in the vehicle-mounted sensor data DB 244, and stores the information in the dangerous driving operation data DB 243. Store.

Any technique can be used for the determination of the dangerous driving operation by the risk determination program 221, and the criteria depend on the design. Examples of driving that is determined to be unsafe include sudden braking, sudden acceleration, vehicle speeds that are too high for the legal speed limit, too narrow inter-vehicle distances, and the like. The danger determination program 221 detects various dangerous driving operations from on-vehicle sensor data. For example, sudden braking and sudden start can be detected from the data of the acceleration sensor 14, and the inter-vehicle distance can be acquired from the data of the inter-vehicle distance sensor 12. The legal speed is known from the position sensor 11 and map information (not shown), and the vehicle speed is known from the speedometer 13 .

The biological model DB 251 stores a plurality of biological models. The biological model includes an explanatory model, semantic interpretation data, and a risk prediction model, as described later. Details of these will be described later. Association data 252 is generated from history data 24 by biological model selection program 222 . Association data 252 is data for a particular driver extracted from historical data 24 .

A trained predictive model 253 is a predictive model included in the biometric model selected for a particular driver and trained with the associated data 252 for that particular driver. The trained prediction model 253 is a prediction model corresponding to the individual's past biometric data.

The pre-driving data 26 is data about the driver acquired before the driver (individual) drives the vehicle, and includes biometric data 261 measured before the driver drives the vehicle and the driver's vehicle driving data. Contains previously evaluated subjective evaluation data 262 . The subjective evaluation data 262 indicates subjective evaluation of the physical condition before the driving by the driver. As will be described later, the subjective evaluation data 262 is referred to for creating an explanation in the pre-driving report.

The biometric data 261 is input to a trained prediction model 253 that predicts the number of dangerous driving maneuvers in the driving. The biometric data 261 includes, for example, biometric data measured before driving on the day of driving, biometric data measured before and after driving on the previous driving day, difference between biometric data measured before and after driving on the previous driving day, and past data. It includes the slope of a regression equation representing the transition of biometric data before and after driving for a predetermined period, or a combination of some or all of these. The data included in pre-drive data 26 may differ between trained predictive models 253 . Inputs to the trained predictive model 253 may include subjective evaluation data.

FIG. 2 shows a hardware configuration example of the operation management assistance system 2. As shown in FIG. The operation management assistance system 2 can have a general computer configuration. The operation management assistance system 2 includes a processor 401 , a memory (main storage device) 402 , an auxiliary storage device 403 , an output device 404 , an input device 405 and a communication interface (I/F) 407 . The above components are connected to each other by buses. Memory 402, secondary storage 403, or a combination thereof are storage devices and store the programs and data shown in FIG.

The memory 402 is composed of, for example, a semiconductor memory, and is mainly used to hold programs and data being executed. Processor 401 executes various processes according to programs stored in memory 402 . Various functional units are implemented by the processor 401 operating according to the program. Auxiliary storage device 403 is composed of a large-capacity storage device such as a hard disk drive or solid state drive, and is used to store programs and data for a long period of time.

Processor 401 can be configured with a single processing unit or multiple processing units, and can include single or multiple arithmetic units or multiple processing cores. Processor 401 is one or more central processing units, microprocessors, microcomputers, microcontrollers, digital signal processors, state machines, logic circuits, graphics processing units, chip-on-systems, and/or manipulates signals based on control instructions. It can be implemented as any device.

The programs and data stored in the auxiliary storage device 403 are loaded into the memory 402 at startup or when necessary, and the programs are executed by the processor 401 to execute various processes of the operation management assistance system 2 . Therefore, the processing executed by the operation management auxiliary system 2 below is processing by the processor 401 or a program.

The input device 405 is a hardware device for the user to input instructions, information, etc. to the operation management assistance system 2 . The output device 404 is a hardware device that presents various input/output images, such as a display device or a printing device. A communication I/F 407 is an interface for connection with a network. The input device 405 and the output device 404 may be omitted, and the user may access the operation management assistance system 2 from a terminal via a network.

The functions of the operation management assistance system 2 can be implemented in a computer system comprising one or more computers including one or more processors and one or more storage devices including non-transitory storage media. Multiple computers communicate via a network. For example, a part of a plurality of functions of the operation management assistance system 2 may be implemented in one computer, and another part may be implemented in another computer.

The driving data collection device 15, the prediction result display terminal 31, the biometric data collection device 32, and the subjective evaluation data collection device 33 can each have a computer configuration like the operation management assistance system 2. The functions of a plurality of devices in the above device may be implemented in one device, for example, the operation management assistance system 2, the prediction result display terminal 31, the biological data collection device 32, and the subjective evaluation data collection device 33 are integrated into one It may be integrated into the device.

In FIG. 2, each element of the software may be stored in any area within the storage device. As described above, the processor 401 functions as a specific functional unit by operating according to specific programs. For example, the processor 401 functions as a risk determination section, a biological model selection section, a prediction model training section, and a prediction result display section according to the above program.

FIG. 3 shows a configuration example of the biometric data DB 241. As shown in FIG. The data stored in the biometric data DB 241 includes data measured before driving on the day of driving. The biometric data DB 241 includes a user ID column, a date and time column, an autonomic nerve Total Power column, an autonomic nerve LF/HF column, and a body temperature column. The user ID column indicates a user ID for identifying each driver who is a user.

The date and time column indicates the measurement date and time of the biometric data. The autonomic nerve total power column indicates the total power (TP), which is one of the electrocardiographic fluctuation parameters. Total Power is the total power of the power spectrum of the specific frequency band of electrocardiographic fluctuations, and is related to fatigue. The autonomic nerve LF/HF indicates the power ratio of the low frequency band (LF) and the high frequency band (HF) in the specific frequency band, which is one of the electrocardiographic fluctuation parameters. LF/HF represents the overall balance of sympathetic and parasympathetic nerves. The Body Temperature column shows the measured body temperature of the driver.

FIG. 4 includes a configuration example of the subjective evaluation data DB 242. As shown in FIG. The data stored in the subjective evaluation data DB 242 includes subjective evaluation data before driving on the day of driving. The subjective evaluation data DB 242 includes a user ID column, a date and time column, and VAS1 to VAS5 columns. The user ID column indicates a user ID for identifying each driver who is a user, and the date and time column indicates the date and time when the subjective evaluation was performed.

Columns VAS1 to VAS5 respectively indicate different VAS (Visual Analog Scale) values. The VAS value represents, for example, self-evaluation of fatigue and sleep. Note that VAS is an example of a self-evaluation method, and a method different from VAS may be used.

FIG. 5 shows a configuration example of the dangerous driving operation data DB 243. As shown in FIG. The dangerous driving operation data DB 243 stores the determination results of the risk determination program 221 . The dangerous driving operation data DB 243 includes a user ID column, a date column, a start time column, an end time column, and a dangerous driving operation 1 column to a dangerous driving operation 4 column.

The user ID column indicates a user ID for identifying each driver who is a user, and the date column indicates the date of a specific driving period for which a dangerous driving operation is to be detected. The start time column and end time column indicate the start time and end time of the operation period. The Dangerous Driving Manipulation 1 column to the Dangerous Driving Manipulation 4 column each indicate the number of predetermined dangerous driving manipulations. Dangerous driving maneuvers 1 to 4 represent different dangerous driving maneuvers.

FIG. 6 shows a configuration example of the association data 252. As shown in FIG. The association data 252 is data of the same user (driver) extracted from the biometric data DB 241 , the subjective evaluation data DB 242 and the dangerous driving operation data DB 243 . Each record of the association data 252 indicates information extracted from the three DBs 241, 242 and 243 for one operating period.

This configuration example includes a user ID column, a date column, a column of autonomic nerve total power before driving on the current day, a column of autonomic nerve total power before driving on the previous day, a column of autonomic nerve total power after driving on the previous day, a column of differential autonomic nerve total power before and after driving on the previous day, and the current day. VAS 1 before driving to VAS 5 before driving on the day, VAS 1 after driving on the day to VAS 5 after driving on the day, and dangerous driving operation 1 column to dangerous driving operation 4 column are included. Some columns are omitted in FIG. As will be described below, association data 252 is used to select from a plurality of hazard risk predictive models the appropriate predictive model for a particular user, and to train the selected predictive model.

Unlike accidents, dangerous driving maneuvers are detected in daily driving, and the number of dangerous driving maneuvers indicates the dangerous risk level of driving. Therefore, by using the history of dangerous driving maneuvers for each individual, it is possible to generate a trained prediction model suitable for each individual.

FIG. 7 shows a flowchart for the biometric data acquisition device 32 to acquire heart rate variability parameters. The biological data collection device 32 measures the biological data of the driver using a biological measuring instrument and transmits the measured biological data to the operation management assistance system 2 . The biological data collection device 32 measures an electrocardiogram (heartbeat) with the heart rate monitor 34 (S101). FIG. 8 shows an example of heart rate variability, where the horizontal axis represents time and the vertical axis represents potential. "RR" indicates the spacing between peaks of a particular type.

Returning to FIG. 7, the biological data collection device 32 detects an interval RR between specific types of peaks in the measurement result of the heart rate monitor 34 (S102). FIG. 9 shows the time variation of the interval RR. Returning to FIG. 7, the biological data collection device 32 analyzes the frequency spectrum of the time variation of the interval RR (S103). FIG. 10 shows an example of a frequency spectrum of time variation of the interval RR. FIG. 10 shows the spectral power density for interval RR.

Returning to FIG. 7, the biological data acquisition device 32 extracts the low frequency band LF power and the high frequency band HF power in the frequency spectrum, and calculates predetermined heart rate variability parameters (S104). Predetermined heart rate variability parameters include Total Power and the ratio of LF and HF power (LF/HF). In this example, Total Power is the sum of LF and HF. The biometric data collection device 32 transmits information about the calculated heart rate variability to the operation management assistance system 2 and records it in the biometric data DB 241 (S105).

FIG. 11 shows a flow chart of an example of processing by the subjective evaluation data collection device 33 . The subjective evaluation data collection device 33 receives the VAS1 input from the driver (S121) and calculates the VAS1 input value (S122). The subjective evaluation data collection device 33 similarly receives VAS2 to VAS4 inputs from the driver and calculates VAS4 input values from the VAS2 input values (not shown). Further, the subjective evaluation data collection device 33 receives the VAS5 input from the driver (S123) and calculates the VAS5 input value (S124). The subjective evaluation data collection device 33 transmits the VAS1 input value to the VAS5 input value to the operation management assistance system 2 and records them in the subjective evaluation data DB 242 .

FIG. 12 shows two VAS example images 501 and 502 . VAS 501 is a VAS for inputting fatigue, and VAS 502 is a VAS for inputting information about last night's sleep.

The left end of the line 511 in the VAS 501 corresponds to a very tired state, and the right end corresponds to a completely non-tired state. The driver inputs his fatigue self-evaluation by moving the black dot 512 on the line 511 to position it. The subjective evaluation data collection device calculates a VAS input value from the input position of the black point 512 .

Similarly, the left end of line 521 in VAS 502 corresponds to a state of poor sleep, and the right end corresponds to a state of good sleep. The driver inputs his sleep self-assessment by moving and positioning black dot 522 on line 521 . The subjective evaluation data collection device calculates the VAS input value from the input position of the black point 522 .

FIG. 13 is a diagram for explaining generation of the association data 252, generation of the trained prediction model 253, and prediction of the number of times of dangerous driving by the trained prediction model 253. FIG. The upper part of FIG. 13 shows the generation of the association data 252, and the lower part shows the training of the prediction model using the association data 252 and the prediction of the number of times of dangerous driving by the trained prediction model 253. FIG.

FIG. 13 shows an example of biological data and dangerous driving operation data included in the association data 252, omitting the subjective evaluation data. In the example of FIG. 13, the association data 252 includes a specific individual's biometric data history from the Kth day to the Nth day. In FIG. 13, the heart marks in the morning and evening of each day indicate biometric data measurements, and the triangles below the vehicle indicate detected dangerous driving maneuvers.

In the example of FIG. 13, the trained prediction model 253 predicts the number of times of dangerous driving in driving on the day from the biometric data before driving on the previous day, the biometric data after driving on the previous day, the biometric data before and after driving on the previous day, and the biometric data on the day before driving. do. The number of times of dangerous driving is one of indices representing the dangerous risk level.

The association data 252 associates the previous day's pre-driving biometric data, the previous day's post-driving biometric data, the previous day's before and after pre-driving biometric data difference, and the current day's pre-driving biometric data with the specific individual's driving from the K+1 day to the Nth day. stored as The association data 252 further stores dangerous driving operation data indicating the number of dangerous driving operations in driving on each of the K+1st to Nth days in association with the biometric data for driving on each day.

In the example of FIG. 13, the biological data 531A is the biological data before driving on the K day and the biological data before driving on the previous day on the K+1 day. The biological data 531B is biological data after driving on the K day and biological data after driving on the previous day on the K+1 day. The biometric data difference 532 is a difference in biometric data before and after driving on the current day on the K day, and a difference in biometric data on the previous day before and after driving on the K+1 day.

The biological data 541A is the biological data before driving on the day (K+1) and the biological data before driving on the previous day (K+2). The number of risk determinations (dangerous driving operation data) 543 on the K+1 day is associated with the biometric data 531A, the biometric data 531B, the biometric data difference 532, and the biometric data 541A.

The biological data and dangerous driving data stored in the association data 252 are used as training data for the predictive model. The predictive model is trained (S 143 ) using machine learning techniques using the association data 252 to generate a trained predictive model 253 . Training updates the parameters of the prediction model based on the error between the prediction model's output and the correct answer. By training the prediction model with the association data 252 of the target user (driver), a prediction model more suitable for the target user can be obtained.

In one example, the trained prediction model 253 predicts the number of dangerous driving operations (number of dangerous determinations) 563 per unit time during driving on the day. In one example, the output of trained predictive model 253 is the total number of all dangerous driving maneuvers per unit time during one driving period (eg, one day). The trained predictive model 253 may predict the number of times per unit time for each type of dangerous driving maneuver.

For example, the trained prediction model 253 is based on the previous day's pre-driving biometric data 551A, the previous day's post-driving biometric data 551B, the previous day's pre- and post-driving biometric data difference 552, and the current day's pre-driving biometric data 561A. Predict 563. The biometric data before and after driving on the previous day indicates changes in the biometric data due to driving, enabling more appropriate prediction. Moreover, the biometric data before driving on the day is the biometric data just before driving, which enables more appropriate prediction. In addition, as described above, the input data may not be common among the prediction models.

FIG. 14 shows a flow chart of a processing example of the operation management assistance system 2 and data related thereto. The biological model selection program 222 reads the association data 252 of the target user (driver) from the history data 24 (S141). The biological model selection program 222 selects an appropriate biological model 280 from a plurality of biological models in the biological model DB 251 based on the association data 252 (S142). The details of the selection method of the biological model 280 will be described later.

Biological model 280 includes explanatory model 281, predictive model 282, and semantic interpretation data 283, which are pre-associated. The predictive model training program 223 uses the association data 252 as training data to train the predictive model 282 (S143). The parameters of predictive model 282 are updated to produce trained predictive model 253 .

The trained prediction model 253 calculates a predicted value of the number of dangerous driving operations in driving by the target user based on the target user's pre-driving data 26 (S144). The predicted value is, for example, the number of times or the total number of each type of dangerous driving operation. As described above, inputs to trained predictive model 253 include biometric data 261 prior to driving (on or before the day before) and may further include subjective evaluation data 262 .

The prediction result presentation program 224 uses the semantic interpretation data 283 of the biological model 280 to generate a semantic interpretation sentence (description), and determines a dangerous risk prediction level based on the number of dangerous driving predictions. The prediction result presentation program 224 presents the semantic interpretation sentence and the danger risk prediction level to the target user and/or administrator (S145).

The prediction result presentation program 224 generates display data 27 and displays it on the output device 404 . The display data 27 includes a danger risk prediction level 271 and a semantic interpretation sentence (description). The prediction result presentation program 224, for example, predicts dangerous risk based on the history of the number of dangerous driving operations of the target user in the past and/or the result of comparison with statistical values (for example, deviation values) of the number of dangerous driving operations of other drivers. Determine level 271.

FIG. 15 shows an example of a flowchart for selection of a biological model (S142) by the biological model selection program 222. FIG. The biological model selection program 222 tests each of the plurality of explanation models (statistical models) in the biological model DB 251 for the association data 252 (S161), and selects an explanation model that matches the association data 252 (S162). .

The explanatory model shows the relationship between subjective evaluation data and biometric data. For example, the relationship between multiple explanatory variables in subjective evaluation data and one objective variable in biological data can be identified by multiple regression analysis. The explanatory model represents, for example, a formula obtained by multiple regression analysis. The biological model selection program 222 may select, for example, an explanatory model whose multiple correlation coefficient between the explanatory model and the association data 252 is the smallest or within a predetermined range.

The explanatory model may represent a relational expression between each of a plurality of objective variables of biometric data and objective variables of one or a plurality of subjective evaluation data. The biological model selection program 222 may select an explanatory model, for example, based on predetermined statistical values of multiple correlation coefficients between the explanatory models and the association data 252 for a plurality of relational expressions.

In one example, the explanatory model shows the relationship between one or more VAS and one or more types of biometric data. An explanatory model as a more specific example shows a correlation coefficient between one VAS (for example, fatigue) and autonomic nerve total power. The biological model selection program 222 selects an explanatory model whose correlation coefficient calculated from the association data 252 is the closest or whose number of correlations is within the assumed range.

FIG. 15 shows an explanatory model A 281A and an explanatory model B 281B as examples. Explanation model A 281A and explanation model B 281B each show a correlation coefficient with autonomic nerve total power. For example, explanatory model A 281A is selected as the explanatory model that best matches association data 252 .

Next, the biological model selection program 222 selects a biological model associated with the biological model including the selected explanatory model from the biological model DB 251 (S163). In the example of FIG. 15, a biological model 280A including explanatory model A 281A is selected. The biological model 280A includes an explanatory model A281A, a prediction model A282A, and semantic interpretation data A283A. Biological model 280B, which includes predictive model B282B and semantic interpretation data B283B in addition to explanatory model B281B, is not selected.

The relationship between subjective evaluation and biometric data may vary from individual to individual. For example, the correlation coefficient between a certain user's autonomic nerve total power and VAS1 may be a negative value, and the correlation coefficient between another user's autonomic nerve total power and VAS1 may be a positive value. As described above, by selecting a biological model based on the relationship between subjective evaluation data and biological data, an appropriate prediction model and semantic interpretation data that match individual characteristics based on medical knowledge are selected. be able to.

In another example, the biological model selection program 222 may select a prediction model without using an explanation model (by omitting the explanation model). For example, the biological model selection program 222 may select a classification model that best matches the target user's historical data. Also, the training of the prediction model may be omitted. The classification model and the untrained prediction model that best match the target user's historical data are prediction models that correspond to the individual's past biometric data.

A trained prediction model obtained by training the prediction model 282 pre-associated with the explanation model 281 and the trained prediction model 282 using the association data 252 is a prediction based on the prediction model associated with the explanation model. is a model.

16 and 17 respectively show image examples of reports presented by the prediction result presentation program 224. FIG. 16 shows report 600 for target user (driver) A, and FIG. 17 shows report 620 for target user (driver) B. FIG.

A report 600 shown in FIG. 16 includes biometric data measurement results 601 . In the example of FIG. 16, the measurement result 601 indicates the biometric data measured after driving on the previous day and the biometric data measured before driving on the current day.

The report 600 also includes information about predicted hazard risks in driving for the day. Specifically, the report 600 shows a hazard risk prediction level 603 and a semantic interpretation (description) 605 related to the hazard risk. The semantic interpretation sentence 605 further includes a semantic interpretation sentence 607 about the biological data and a semantic interpretation sentence 609 about the dangerous risk prediction level 603 .

By presenting the danger risk prediction level, it is possible to alert the target user A, or to prompt the operation manager to take measures such as changing the work content of the target user A. Furthermore, by presenting the semantic interpretation sentence 607 about the biometric data, it is possible to improve the understanding of the target user A's physical condition. Furthermore, by presenting the semantic interpretation sentence 609 about the danger risk prediction level, it is possible to improve the comprehension of the danger risk prediction level. Note that part of the information presented by the report 600 may be omitted.

A report 620 shown in FIG. 17 includes biometric data measurement results 621 . In the example of FIG. 17, the measurement result 621 indicates the biometric data measured after driving on the previous day and the biometric data measured before driving on the current day.

The report 620 also includes information about predicted hazard risks in driving for the day. Specifically, the report 620 shows a hazard risk prediction level 623 and a semantic interpretation (description) 625 related to the hazard risk. Semantic interpretation sentences 625 include semantic interpretation sentences 627 about biometric data. Since the danger risk prediction level is low, the semantic interpretation sentence about the danger risk prediction level is omitted. The report 620 can have the same effect as the report 600 for the target user B. FIG. Also, some of the information presented by report 620 may be omitted.

The prediction result presentation program 224 determines the dangerous risk prediction level based on the number of dangerous driving operations predicted by the trained prediction model 253, as described above. When the trained prediction model 253 indicates the number of times for each dangerous driving maneuver type, the prediction result presentation program 224 may give a predetermined weight to each dangerous driving maneuver type.

The prediction result presentation program 224 determines the representation of the measurement result from the numerical value of the biometric data based on a predetermined criterion. The criteria may be determined based on historical data for each target user.

The prediction result presentation program 224 refers to the semantic interpretation data 283 in the selected biological model 280 to generate a semantic interpretation sentence. Semantic interpretation data 283 indicates a sentence associated with each different level of biometric data, and further indicates a sentence associated with each different hazard risk prediction level.

In the example of the report 600 in FIG. 16, the semantic interpretation data 283 associates the semantic interpretation sentence 607 with the numerical value representing the autonomic nervous system after driving the previous day, which is "slightly high." Furthermore, the semantic interpretation data 283 associates the semantic interpretation sentence 609 with the danger risk prediction level of "slightly high risk". Similarly, in the example of the report 620 in FIG. 17, the semantic interpretation data 283 in the selected biological model 280 associates the semantic interpretation sentence 627 with the numerical value representing the autonomic nerves before driving today is "high."

The semantic interpretation data 283 are selected based on the subjective evaluation data and biometric data of the target user, along with the explanatory model 281 and the predictive model 282 . Therefore, it is possible to present an appropriate semantic interpretation sentence for the target user's biometric data measurement result and danger risk prediction level.

In the example above, semantic interpretation data 283 is associated with explanatory model 281 and predictive model 282 . Semantic interpretation data 283 is associated with predictive model 282 even when explanatory model 281 is omitted. The prediction result presentation program 224 refers to the semantic interpretation data 283 associated with the selected prediction model 282 to generate a semantic interpretation sentence.

In addition, the present invention is not limited to the above-described embodiments, and includes various modifications. For example, the above embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the described configurations. Also, part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. Moreover, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

Further, each of the configurations, functions, processing units, etc. described above may be realized by hardware, for example, by designing a part or all of them using an integrated circuit. Moreover, each of the above configurations, functions, etc. may be realized by software by a processor interpreting and executing a program for realizing each function. Information such as programs, tables, and files that implement each function can be stored in a recording device such as a memory, a hard disk, an SSD (Solid State Drive), or a recording medium such as an IC card or SD card.

In addition, the control lines and information lines indicate those considered necessary for explanation, and not all control lines and information lines are necessarily indicated on the product. In fact, it may be considered that almost all configurations are interconnected.

1 vehicle 2 operation management auxiliary system 11 position sensor 12 inter-vehicle distance sensor 13 speedometer 14 acceleration sensor 15 driving data collection device 24 history data 26 data before driving 27 display data 31 prediction result display terminal 32 biological data collection device 33 subjective evaluation data Collection device 34 Heart rate monitor 35 Thermometer 36 Sphygmomanometer 221 Risk determination program 222 Biological model selection program 223 Prediction model training program 224 Prediction result presentation program 242 Subjective evaluation data DB
243 Dangerous driving operation data DB
244 In-vehicle sensor data DB
251 Biological Model DB
252 Association data 253 Prediction model 261 Biological data 262 Subjective evaluation data 271 Danger risk prediction level 280 Biological model 281 Explanation model 282 Prediction model 283 Semantic interpretation data 401 Processor 402 Memory 403 Auxiliary storage device 404 Output device 405 Input device 407 Communication I/F
501, 502 VAS images 511, 521 Lines 512, 522 Black dots 531A, 531B, 541A, 551A, 551B, 561A Biological data 532, 552 Biological data difference 563 Number of dangerous driving operations 600, 620 Reports 601, 621 Measurement results 603, 623 Danger risk prediction level 605, 607, 609, 625, 627 Semantic interpretation sentence

Claims (5)

one or more storage devices;
one or more processors that operate according to programs stored in the one or more storage devices;
The one or more storage devices are
a plurality of prediction models that predict the number of dangerous driving operations based on biometric data;
Subjective evaluation data history of the individual's physical condition associated with the individual's biometric data history;
storing association data including the individual's biometric data history and the individual's dangerous driving operation frequency history that are associated with each other;
each of the plurality of prediction models is associated with an explanation model indicating the relationship between subjective evaluation data and biometric data;
The one or more processors
selecting an explanatory model corresponding to the relationship between the subjective evaluation data history and the biometric data history from the explanatory models associated with each of the plurality of prediction models;
selecting a predictive model associated with the selected explanatory model from the plurality of predictive models;
training the selected predictive model using the association data;
A system that uses the trained predictive model to determine a predicted level of hazard risk in driving the vehicle for the individual based on biometric data and subjective assessment data measured prior to the driving of the vehicle for the individual. 2. The system of claim 1, wherein
The one or more storage devices are associated with each of the explanation models and store semantic interpretation data indicating semantic interpretation of biometric data;
The one or more processors measure biometric data measured prior to driving the individual's vehicle and semantic interpretation data associated with the selected explanatory model. A system for creating a semantic interpretation sentence of biometric data and presenting it together with information on the determined danger risk prediction level. 2. The system of claim 1, wherein
The one or more storage devices are associated with each explanation model of the explanation model and store semantic interpretation data indicating a semantic interpretation of a danger risk prediction level in driving;
The one or more processors determine the determined hazard risk prediction level based on the determined hazard risk prediction level and semantic interpretation data associated with the obtained prediction model via the explanation model. A system for creating a semantic statement and presenting it with the information of the determined hazard risk prediction level. 2. The system of claim 1, wherein
The one or more processors compare the obtained prediction result of the prediction model with the statistical value of the history of dangerous driving operations of the individual and/or others, based on the dangerous risk in driving the vehicle by the individual. A system that determines the level of prediction. A method performed by a system comprising:
The system includes:
a plurality of prediction models that predict the number of dangerous driving operations based on biometric data;
Subjective evaluation data history of the individual's physical condition associated with the individual's biometric data history;
storing association data including the individual's biometric data history and the individual's dangerous driving operation frequency history that are associated with each other;
each of the plurality of prediction models is associated with an explanation model indicating the relationship between subjective evaluation data and biometric data;
The method includes:
wherein the system selects an explanatory model corresponding to the relationship between the subjective evaluation data history and the biometric data history from the explanatory models associated with each of the plurality of prediction models;
the system selecting from the plurality of predictive models a predictive model associated with the selected explanatory model;
the system training the selected predictive model using the association data;
wherein the system uses the trained predictive model to determine a predicted level of hazard risk in driving the individual's vehicle based on biometric data and subjective assessment data measured prior to driving the individual's vehicle; method involving

Images