JP7633013B2 - Safe driving support method, safe driving support system, and safe driving support server - Google Patents

Description

The present invention relates to a safe operation support method and a safe operation support system that predict the risk of traffic accidents and support the safe operation of transportation.

In recent years, traffic accidents caused by driver health and fatigue in logistics trucks, long-distance buses, etc. have become a social issue. In order to prevent traffic accidents, technology that monitors the driver's condition while driving using biosensors and technology that monitors the behavior of the vehicle while driving are being increasingly applied.

For example, Patent Document 1 discloses a "safe driving support system that can continually learn the relationship between the driver's driving operations and the external environment, recognize deviations from the driver's usual internal state, and support safe driving."

JP 2009-098970 A

Patent Document 1 provides a means for recognizing deviations from the driver's usual internal state by comparing the driving environment risks contained in the external environment recognized by sensing the external environment of the vehicle with the driver's risk recognition state estimated as an internal state from the driver's driving operation.

However, simply performing state estimation on the relationship between the driver's driving operation and the external environment and recognizing deviations from the driver's usual internal state (dangerous states in the narrow sense) is not enough to detect conditions that are not apparent in driving operation but require caution (dangerous states in the broad sense). This has led to the problem that it is difficult to take measures to avoid dangerous states in the broad sense in advance.

The present invention was made in consideration of the above problems, and aims to detect dangerous conditions in a broad sense, taking into account the individual differences between drivers, using biometric data, and to determine whether or not to issue an alert based on the results, in order to detect dangerous conditions in a broad sense and avoid them in advance.

Solution

The present invention is a safe driving support method in which a computer having a processor and a memory supports driving of a vehicle, the method including a first step of generating an accident risk definition model in which the computer receives first on-board sensor data indicating a driving state of the vehicle collected in the past and preset danger occurrence data, and estimates the probability of a dangerous event occurring as an accident risk; a second step of the computer inputting second on-board sensor data indicating the driving state of the vehicle collected in the past to the accident risk definition model, and estimating the probability of a dangerous event occurring to generate accident risk estimation data; a third step of the computer receiving first biodata calculated in advance from first biosensor data of the driver when the second on-board sensor data was collected and the accident risk estimation data, and generating an accident risk prediction model in which the computer predicts an accident risk after a predetermined time; The method includes a fourth step of generating a biological state estimation model that uses third biological data collected in the past using either a seat, a sensing device mounted on the vehicle, or an image recognition system that analyzes an image of the driver , and calculates a person seriousness indicating to what extent second biological data calculated from second biological sensor data of the driver while driving deviates from the third biological data; a fifth step of the computer acquiring the second biological sensor data and calculating the second biological data from the second biological sensor data; a sixth step of the computer inputting the second biological data into the accident risk prediction model to predict the accident risk and inputting the second biological data into the biological state estimation model to calculate the person seriousness; and a seventh step of the computer determining whether or not to issue a warning based on the predicted accident risk and the calculated person seriousness.

Therefore, according to one embodiment of the present invention, it becomes possible to detect a dangerous state (hazardous event) in a broad sense based on the accident risk prediction model, the biological state estimation model, and the biological data. As a result, it becomes possible to determine whether or not an alarm should be issued based on the dangerous state in a broad sense, and further to present countermeasures for avoiding the dangerous state in a broad sense.

Details of at least one implementation of the subject matter disclosed herein are set forth in the accompanying drawings and the description below. Other features, aspects, and advantages of the disclosed subject matter will become apparent from the following disclosure, drawings, and claims.

FIG. 1 is a block diagram illustrating an example of a configuration of a safe driving support system according to a first embodiment of the present invention. FIG. 1 is a diagram illustrating an overview of a process performed in a safe driving support system according to a first embodiment of the present invention. 4 is a flowchart illustrating an example of a generation process of an accident risk prediction model performed by the safe driving support server according to the first embodiment of the present invention. 10 is a flowchart illustrating an example of a biological state estimation model generation process performed by the safe driving support server according to the first embodiment of the present invention. 4 is a flowchart illustrating an example of a calculation process of biological data performed by the safe driving support server according to the first embodiment of the present invention. 5 is a flowchart illustrating an example of a calculation process of biological data using heart rate data, which is performed in the safe driving support server according to the first embodiment of the present invention. 4 is a flowchart illustrating an example of a risk prediction process performed by the safe driving support server according to the first embodiment of the present invention. 4 is a flowchart illustrating an example of a warning presentation process performed by the safe driving support server according to the first embodiment of the present invention. FIG. 11 illustrates an example of a warning display screen according to the first embodiment of the present invention. FIG. 11 illustrates another example of the warning display screen according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of biometric data according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of task and environment data according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of attribute information data according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of learning information data according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of risk occurrence data according to the first embodiment of the present invention. FIG. 2 is a diagram illustrating an example of a data structure of accident risk prediction data according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of person seriousness data according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of a join table according to the first embodiment of the present invention. FIG. 1 illustrates an example of a data structure of alert definition data according to the first embodiment of the present invention. FIG. 2 illustrates an example of a data structure of countermeasure proposal data according to the first embodiment of the present invention. FIG. 11 is a diagram illustrating an outline of a process performed in the safe driving support system according to the second embodiment of the present invention. FIG. 11 is a diagram illustrating an outline of a process performed in the safe driving support system according to the third embodiment of the present invention. FIG. 13 is a diagram illustrating an example of a warning destination list screen provided to a driver manager according to the third embodiment of the present invention.

The following describes an embodiment of the present invention with reference to the drawings.

First, we will explain Example 1 of the present invention.

<System Configuration>
1 is a block diagram showing an example of a main configuration of a safe driving support system according to a first embodiment of the present invention. The safe driving support system according to the present embodiment includes a safe driving support server 1 that supports the driving of one or more vehicles 7 via a network 14.

The vehicle 7 includes an on-board sensor 8 that detects the driving condition, a biosensor 12 that detects the driver's biometric data 64, a driver ID reader 11 that identifies the driver, a driving data collection device 10 that collects the detected on-board and biometric sensor data and the driver ID and transmits them to the safe driving support server 1, a prediction result notification device 9 that receives a warning from the safe driving support server 1 according to the driver's risk of a traffic accident (hereinafter, accident risk) and presents it to the driver, and a business status input device 13 that receives input of the measurement status of the biometric data 64.

In the illustrated example, the driving data collection device 10, the prediction result notification device 9, the business status input device 13, and the driver ID reader 11 are shown as separate devices, but they can be configured as a single mobile terminal. In this case, the mobile terminal functions as a driving data collection unit, a prediction result notification unit, a business status input unit, and a driver ID reader.

The on-board sensors 8 can include a Global Navigation Satellite System (GNSS) 81 that detects the vehicle's position information, an acceleration sensor 82 that detects the vehicle's behavior and speed, and a camera 83 that detects the driving environment as an image.

The on-board sensor 8 is not limited to the above, and may be a distance sensor that detects objects and/or distances around the vehicle, a steering angle sensor that detects driving operations, an angular velocity sensor that detects turning operations of the vehicle, etc. Also, it is preferable that the acceleration sensor 82 is a three-axis acceleration sensor.

The biosensor 12 includes a heart rate sensor 121 that detects heart rate data, and an acceleration sensor 122 that detects the driver's movement. The heart rate sensor 121 can be a sensor that detects the heart rate based on an electrocardiogram, a pulse wave, or heart sounds.

The biosensor 12 is not limited to the above, and can be a sensor that detects the amount of sweat, body temperature, blinking, eye movement, electromyography, or brain waves. As the biosensor 12, a wearable device that can be worn by the driver, a sensing device attached to the inside of the vehicle such as the steering wheel, seat, or seat belt, or an image recognition system that captures the driver's facial expressions and behavior and analyzes the images can be used.

The driver ID reader 11 reads the card on which the driver's identifier is recorded. The driving data collector 10 collects data from the vehicle sensor 8 and biosensor 12 at a predetermined interval and transmits it to the safe driving support server 1 via the network 14.

In the illustrated example, the driver ID reader 11 is configured as a device that reads a card on which the driver's identifier is recorded, but a different configuration may be used. For example, if the driver ID reader 11 is configured as a mobile terminal and the mobile terminal is made to function as a driver ID reader, the driver ID may be read by having the driver input the driver's identifier, or the driver ID may be read by identifying the driver using a known face recognition technology using the camera of the mobile terminal.

The work status input device 13 accepts the driver's work status, such as driving, resting, or napping, and transmits it to the safe driving support server 1.

The safe driving support server 1 is a computer including a processor 2, a memory 3, a storage device 4, an input/output device 5, and a communication device 6. The memory 3 loads each of the functional units, namely, a biometric data calculation unit 31, an accident risk definition generation unit 32, an accident risk estimation unit 33, a biometric state estimation model generation unit 34, an accident risk prediction model generation unit 35, a model selection unit 36, a danger prediction unit 37, a warning presentation unit 38, and a data collection unit 39, as programs. Each program is executed by the processor 2. The details of each functional unit will be described later.

The processor 2 operates as a functional unit that provides a specific function by executing processing according to the program of each functional unit. For example, the processor 2 functions as the biometric data calculation unit 31 by executing a biometric data calculation program. The same applies to other programs. Furthermore, the processor 2 also operates as a functional unit that provides each function of the multiple processes executed by each program. Computers and computer systems are devices and systems that include these functional units.

The storage device 4 stores data used by each of the above functional units. The storage device 4 stores vehicle sensor data 61, biosensor data 63, task and environment data 74, accident risk prediction model 71, accident risk prediction data 66, hazard occurrence data 62, biometric data 64, accident risk definition model 70, alert definition data 72, accident risk estimation data 65, countermeasure proposal data 73, biometric state estimation model 69, individual severity data 67, combination table 68, attribute information data 75, and learning information data 76.

As described below, the on-board sensor data 61 includes on-board sensor data 61-1 (first on-board sensor data) used when generating the accident risk definition model 70, and on-board sensor data 61-2 (second on-board sensor data) used when generating the accident risk prediction model 71. In the following description, when there is no need to distinguish between the on-board sensor data, the symbol "61" is used with the parts after "-" omitted.

As described below, the biometric data 64 includes biometric data 64-1 (first biometric data) used when generating the accident risk prediction model 71, biometric data 64-2 (second biometric data) used when inputting the accident risk prediction model 71 to generate the accident risk prediction data 66, and biometric data 64-3 (third biometric data) used when generating the biological state estimation model 69. In the following description, when there is no need to distinguish between the biometric data, the symbol "64" is used with the parts after "-" omitted. Details of each data will be described later.

The input/output device 5 includes input devices such as a mouse, keyboard, touch panel, or microphone, and output devices such as a display and speaker. The communication device 6 communicates with the vehicle via the network 14.

<Processing Overview>
2 is a diagram showing an overview of the processing performed by the safe driving support system. In the example shown in the figure, an example is shown that is configured with a learning phase that is performed before the vehicle 7 starts driving, and an operation phase that is performed while the vehicle 7 is driving.

First, the safe driving support system generates an accident risk prediction model 71 and a biological condition estimation model 69 in the safe driving support server 1, which are necessary to predict the dangerous state of the driver while driving the vehicle 7.

First, the accident risk definition generation unit 32 inputs preset risk occurrence data 62 and previously collected vehicle sensor data 61-1 to generate an accident risk definition model 70 that estimates the probability of risk occurrence as an accident risk (S1). Note that the time series of the vehicle sensor data 61-1 and the time series of the risk occurrence data 62 are the same time series.

The accident risk definition model 70 estimates the accident risk, which is the probability that a dangerous event will occur, using on-board sensor data 61-2 (described later) that indicates the vehicle's past driving conditions as input. In the following, the accident risk value is expressed as a percentage.

The danger occurrence data 62 is data obtained by extracting events that are determined to be dangerous based on previously collected in-vehicle sensor data. Specifically, incidents such as sudden braking or sharp turns, or events that are linked to incidents, are determined by the driver or manager, or by a commercially available in-vehicle alarm or AI, and the data is preset based on the type of event and the date and time of occurrence. Furthermore, the danger occurrence data 62 may be set to a level of importance for each incident.

In this embodiment, an incident is defined as an event that can lead to a traffic accident. An event that leads to an incident is a state in which the driver feels a close call or is startled. In addition, an example is shown in which the danger increases as the importance level increases. In the following, an incident or an event that leads to an incident is defined as a dangerous event.

The dangerous situation in the narrow sense described in the above problem includes the above incident, and the dangerous situation in the broad sense can include events that are linked to the above incident.

The accident risk definition model 70 is a machine learning model that estimates the risk of an accident resulting from a hazardous event by inputting the in-vehicle sensor data 61-2. For the machine learning model, a well-known or publicly known method such as a support vector machine, a neural network, or logistic regression can be used. In addition, multiple accident risk definition models 70 may be generated for each type of hazardous event.

Next, the accident risk estimation unit 33 inputs the previously collected vehicle sensor data 61-2 into the accident risk definition model 70 to generate accident risk estimation data 65 that estimates the accident risk of a hazardous event occurring in a time series (S2).

The accident risk estimation data 65 is data that represents the probability of a dangerous event occurring for the in-vehicle sensor data 61-2, and is generated as the same time series as the in-vehicle sensor data 61-2.

Then, the accident risk prediction model generation unit 35 inputs the accident risk estimation data 65 and the past biometric data 64-1, predicts and outputs the accident risk after a predetermined time from the biometric data 64-2 (described later) of the driver of the vehicle 7 while it is moving, and generates an accident risk prediction model 71 (S4). Note that the past biometric data 64-1 is calculated by the biometric data calculation unit 31 from the past biometric sensor data 63 corresponding to the on-board sensor data 61-2 collected in the past.

In this embodiment, the biometric data 64-1 calculated from the biometric sensor data 63 collected in the past can be calculated using, for example, the average heart rate calculated from the driver's heart rate data, the maximum Lyapunov exponent calculated by nonlinear analysis of the heart rate data series, frequency domain analysis of the heart rate interval data extracted from the heart rate data, or an autonomic nerve function index (LF/HF, etc.). Note that the autonomic nerve function index (LF/HF, etc.) can be calculated using methods such as time domain analysis and nonlinear analysis.

The biometric data 64-1 is a biometric indicator calculated by the biometric data calculation unit 31 from past biometric sensor data 63 corresponding to the time series of the in-vehicle sensor data 61-2.

The accident risk prediction model 71 is composed of a known or publicly known machine learning model, and takes as input the biometric data 64-2 of the driver while driving, and outputs the accident risk after a predetermined time. The biometric data 64-2 of the driver while driving is a biometric indicator calculated by the biometric data calculation unit 31 from the driver's biosensor data 63, and is received by the safe driving support server 1 from the vehicle 7 while driving.

As the accident risk prediction model 71, it is advisable to generate multiple models in advance according to the type of accident risk, the type of biometric data to be used, the driving environment of the vehicle to which the prediction is applied, and the specified time. By generating multiple models in advance, it becomes possible to select and use the most appropriate one in the risk prediction process.

In addition, multiple models may be generated based on attribute information data 75 that stores the driver's work characteristics (driving on ordinary roads, driving on highways, working continuously day and night, etc.), driving experience (years of driving, driving skills, type of license held), and health characteristics (gender, amount of sleep, etc.).

Next, the biological state estimation model generation unit 34 uses previously measured biological data 64-3 as input and generates a biological state estimation model 69 that estimates and outputs a degree of severity of the person indicating the degree to which the input (current biological data 64-2) deviates from the previously measured biological data 64-3 (S3). The degree of deviation of the current biological data 64-2 from the previously measured biological data 64-3 may be expressed as an abnormality degree.

The biological state estimation model 69 is composed of a known or publicly known statistical model or machine learning model. Typically, multiple biological state estimation models 69 are generated for each type of biological data 64-3, and are unsupervised models that input a certain type of biological data 64-3 and output statistics indicating the degree of abnormality (the individual's severity) based on the biological data 64-3 used to generate the model.

Examples of statistical models that can be used include a statistical model that outputs a z-score obtained by normalizing the input using the average value and standard deviation of the biological data 64-3 used to generate the model, a statistical model that normalizes the output to a range of 0 to 1 using the maximum and minimum values of the biological data 64-3 used to generate the model, a statistical model that outputs the rate of change from a value based on the statistics of the biological data 64-3 used to generate the model, and a statistical model that outputs a quantile on the data distribution estimated from the biological data 64-3 used to generate the biological state estimation model 69.

As a machine learning model, for example, an anomaly detection model can be used that estimates a data distribution nonparametrically from the biological data 64-3 used to generate the biological state estimation model 69, and outputs the distance of that distribution from the cluster center.

In addition to the unsupervised biological state estimation model 69, a supervised biological state estimation model 69 can also be adopted. In this case, the biological state estimation model 69 can be configured using a known or known statistical model or machine learning model that estimates, as a target variable, learning information data 76 that collects the driver's drowsiness estimated by a known or known method, the number of blinks (number of blinks) that is an index of the driver's wakefulness, and the subjective fatigue level answered by the driver on a visual analogue scale (VAS). In this case, for example, a regression model or the like can be adopted as the statistical model, and a support vector machine or a neural network can be adopted as the machine learning model.

The biometric data 64-3 input to the biometric state estimation model 69 is typically univariate, but may be multivariate. When the input is multivariate, it is desirable to input a combination of biometric data types that can be easily used to generate countermeasures to avoid dangerous situations in the warning presentation process described below. In this case, the biometric state estimation model 69 can be configured using known methods such as a multiple regression model, the Mahalanobis-Taguchi method, or multivariate statistical process management.

The biological state estimation model 69 may be generated for each driver, or a single biological state estimation model 69 may be generated for a group of multiple drivers. When multiple driver groups are grouped together, it is desirable to generate a biological state estimation model 69 for a group of drivers who show similar biological responses, since there is a large degree of individual variation in the body.

By generating a biological state estimation model 69 from a group of multiple drivers who show similar biological responses, it is possible to absorb individual differences in biological responses and shorten the time required to measure the data required to generate the biological state estimation model 69.

Furthermore, after the required amount of data has been measured, it is desirable to switch from using the biological state estimation model 69 generated from a group of multiple drivers who show similar biological responses to using the biological state estimation model 69 generated from only the target driver. This makes it possible to output the individual severity level that takes into greater consideration the individual's biological responses.

Then, the safe driving support system uses the safe driving support server 1 to predict the probability of a dangerous event occurring to the driver while actually driving the vehicle 7, using the generated accident risk prediction model 71 and biological state estimation model 69.

First, the risk prediction unit 37 inputs biosensor data 63, such as heart rate data measured by the biosensor 12, received by the data collection unit 39 from the vehicle 7, to the biodata calculation unit 31, and calculates the biodata 64-2 (S5).

In addition, when there are multiple accident risk prediction models 71 or biological state estimation models 69, the risk prediction unit 37 selects the accident risk prediction model 71 or biological state estimation model 69 according to the task/environment data 74 and attribute information data 75 in the model selection unit 36.

Then, the risk prediction unit 37 inputs the biometric data 64-2 to the accident risk prediction model 71 to predict (generate) the accident risk prediction data 66 (S6A). The risk prediction unit 37 also inputs the biometric data 64-2 to the biological state estimation model 69 to calculate the person's severity data 67 for each biometric data type (S6B).

Then, the risk prediction unit 37 extracts the type of biometric data 64-2 that contributed to the prediction of the accident risk prediction data 66 (contribution index), and matches the accident risk prediction data 66 with the accident risk prediction model 71 used to calculate the accident risk prediction data 66, the contribution index, and the individual severity data 67 of the contribution index, and stores them in the combination table 68 (S7A).

When extracting the contribution index, the risk prediction unit 37 extracts the contribution index from the biometric data 64-2, the accident risk prediction data 66, and the accident risk prediction model 71 used to calculate the accident risk prediction data 66.

For example, when a linear model such as logistic regression is used as the accident risk prediction model 71, the danger prediction unit 37 can simply extract the biometric data type that maximizes the product of the regression coefficient and the biometric data 64-2 as the contributing index.

In addition, when a neural network or the like is used as the accident risk prediction model 71, the risk prediction unit 37 can calculate the contribution of each piece of biometric data 64-2 using, for example, SHAP, a known method used to ensure the explainability of machine learning models, and extract the maximum contribution as the contribution index.

The warning unit 38 then makes a decision to issue a warning (hereinafter, "alert decision") based on the accident risk and the person's severity stored in the combination table 68 (S7B). If the warning unit 38 determines that the driver is in a broadly dangerous state, it generates a countermeasure plan for avoiding the broadly dangerous state (S7C).

In determining whether to issue an alert, if the accident risk or the severity of the person concerned exceeds a predetermined threshold, or if the values related to the accident risk and the severity of the person concerned exceed a predetermined threshold, the alert presentation unit 38 determines that the situation corresponds to a dangerous state (dangerous event) in the broad sense. Note that in the former case, the alert presentation unit 38 does not have to set a threshold for either the accident risk or the severity of the person concerned when determining whether to issue an alert.

For example, if the warning presentation unit 38 sets a threshold only for the accident risk, it can issue a warning for situations that are generally determined to have a high accident risk. If the warning presentation unit 38 sets a threshold only for the person's seriousness, it can issue a warning for abnormalities in the biometric data 64-2, such as a sudden change in the driver's personal physical condition. Furthermore, thresholds may be set not only for the records of the corresponding accident risk and person's seriousness, but also for the time series of the accident risk and person's seriousness.

When generating countermeasures implemented by the warning presentation unit 38, countermeasures to avoid dangerous situations in a broad sense are generated based on the combination table 68 and countermeasure data 73. From past knowledge, there are known methods for physiologically interpreting changes in a certain type of biometric data 64-2 and methods for changing a certain type of biometric data 64-2.

Measures proposal data 73 stores physiological interpretations for when the value of each type of biometric data 64-2 increases or decreases, and methods of biometric data 64-2 for increasing or decreasing the value. Since the combined table 68 stores contribution indices that contributed to the prediction of accident risk, it is possible to generate measures to avoid dangerous situations in a broad sense according to the driver's condition by obtaining physiological interpretations corresponding to the contribution indices (described below) and methods for increasing or decreasing the contribution indices.

For example, if the contributing index is the autonomic nerve function index LF/HF, and the LF/HF is in the positive direction and the individual's severity is high, this can be physiologically interpreted as being under hypertension, and the warning presentation unit 38 may generate countermeasures to encourage relaxation, such as "try breathing slowly" in order to reduce the driver's LF/HF.

In addition, if the contributing index is the average heart rate, the average heart rate is on a downward trend, and the individual's seriousness is high, this may be physiologically interpreted as an increase in drowsiness, and the warning presentation unit 38 may generate a countermeasure to encourage the driver to relieve drowsiness, such as "Stop at the next convenience store, get out of the car, and stretch."

In addition, when generating a countermeasure proposal, the warning presentation unit 38 can add information based on the personal severity of the contribution index to the countermeasure proposal. A warning based on accident risk alone is unlikely to convey the danger to the driver receiving the warning, but by presenting a warning in a form that is understandable to the driver based on the personal severity, it is expected that the driver's acceptance of the warning will be improved.In addition, when generating a countermeasure proposal, the warning presentation unit 38 can add information based on the task/environment data 74 and attribute information data 75 to the countermeasure proposal.

After generating the proposed countermeasures, the warning presentation unit 38 transmits a warning as an alert or message including the proposed countermeasures to the vehicle 7 that has been determined to be in a broadly dangerous state (S8). The warning presentation unit 38 may also transmit a warning not only to the vehicle 7 but also to the driver manager, and display it on the display of the input/output device 5. In this case, the content to be displayed on the driver's display may include an image including the driver's position information, etc.

When the vehicle 7 receives a warning from the safe driving support server 1, the vehicle 7 outputs an alert from the prediction result notification device 9 and transmits the alert or message to the driver.

The safe driving support server 1 determines a dangerous state in a broad sense based on the accident risk after a specified time predicted using the accident risk prediction model 71 and the severity of the individual in the biometric data 64-2 estimated using the biometric state estimation model 69, thereby absorbing the individual differences between drivers contained in the biometric data 64-2 and detecting a dangerous state in a broad sense.

When the safe driving support server 1 determines that a dangerous condition exists in a broad sense, it generates a countermeasure plan for avoiding the dangerous condition in a broad sense based on the contribution index and the individual's seriousness of the contribution index, and transmits a warning including the countermeasure plan to the prediction result notification device of the vehicle in question or the display of the input/output device 5. This allows the safe driving support server 1 to present a warning to the driver or driver manager before the risk of an accident increases, and to have them take measures to avoid the dangerous condition in a broad sense.

As a result, the safe driving support server 1 can prevent dangerous situations in the broad sense in advance and prevent the driver from falling into dangerous situations in the narrow sense, thereby supporting the safe driving of each vehicle 7.

<Processing details>
3 is a flowchart showing an example of a process for generating the accident risk prediction model 71. This process is performed before receiving the biometric data 64-2 from the vehicle 7 that performs hazard detection while traveling (the learning stage in FIG. 2), to generate the accident risk prediction model 71 in advance.

The accident risk definition generation unit 32 generates an accident risk definition model 70 by inputting previously collected risk occurrence data 62 and corresponding in-vehicle sensor data 61-1 (S1).

Then, the accident risk prediction model generation unit 35 inputs the previously collected vehicle sensor data 61-2 into the generated accident risk definition model 70 to generate accident risk estimation data 65 (S2).

Next, the accident risk prediction model generation unit 35 generates an accident risk prediction model 71 from the biometric data 64-1 corresponding to the previously collected in-vehicle sensor data 61-2 and the accident risk estimation data 65 (S4). In this case, the accident risk prediction model generation unit 35 may generate an accident risk prediction model 71 for each similar task characteristic or environmental characteristic based on the task and environment data 74, and may add this information to the accident risk prediction model 71 as meta information.

Through the above process, the safe driving support server 1 generates an accident risk definition model 70 and generates accident risk estimation data 65, and then generates an accident risk prediction model 71 from biometric data 64-1 that corresponds in time series to the accident risk estimation data 65.

The above process is also performed each time the amount or type of hazard occurrence data 62 increases by a certain amount, each time the amount of high accident risk estimation data 65 increases by a certain amount, or at a certain frequency, and the accident risk definition model 70 and accident risk prediction model 71 are updated each time.

Figure 4 is a flowchart showing an example of a process for generating a biological state estimation model 69. This process is performed before performing risk prediction for a certain vehicle 7 while it is traveling (the learning stage in Figure 2), and a biological state estimation model 69 is generated for each type of biological data 64-3.

The biological state estimation model generation unit 34 first extracts a group of drivers who show similar biological responses as a group of drivers to be used in generating the biological state estimation model 69 (S11). Typically, the unit performs clustering or the like using known or publicly known methods to extract a group of drivers who show similar biological responses using daily statistics such as the average, standard deviation, median, and quantiles of a certain type of biological data 64-3, or weekly or monthly statistics.

In this case, in addition to the biometric data 64-3, the biological state estimation model generation unit 34 may also add information contained in the business/environment data 74, and information such as gender and age stratified based on age contained in the attribute information data 75, to the input for clustering.

In addition, when selecting the biological data 64-3 to be input for clustering, the biological state estimation model generation unit 34 may select only the biological data 64-3 measured under the same measurement conditions based on the learning information data 76, and input only the biological data 64-3 for the period stored in the learning information data 76. This is expected to control the measurement conditions and suppress the variation in the biological data 64-3.

The biological condition estimation model generation unit 34 may further extract only the specific driver instead of a group of multiple drivers if a sufficient amount of biological data 64-3 has been measured for the specific driver. This allows the estimation of the severity of the specific individual that is more in line with the characteristics of the specific individual.

Next, the biological state estimation model generation unit 34 uses the biological data 64-3 previously collected for the extracted driver group as input and generates a biological state estimation model 69 for each type of biological data 64-3 (S3).

In this case, the biological state estimation model generation unit 34 may select only the biological data 64-3 measured under the same measurement conditions based on the learning information data 76, as in the above driver extraction, and input only the biological data 64-3 for the period stored in the learning information data 76.

When the biological state estimation model generation unit 34 generates a supervised model as the biological state estimation model 69, the unit 34 inputs the objective variable for the period stored in the learning information data 76 and the corresponding biological data 64-3, and generates the biological state estimation model 69 for each type of biological data 64-3.

In this case, the biological state estimation model generation unit 34 may add information regarding the type of input biological data 64-3 and similar biological characteristics to the biological state estimation model 69 as meta-information.

After extracting groups of drivers who exhibit similar biological response characteristics through the above process, a biological state estimation model 69 is generated for each type of biological data 64-3 from the biological data 64-3 for the extracted groups of drivers.

The above process of generating from multiple drivers is performed each time the number of target drivers increases by a certain amount, each time the biometric data 64-3 increases by a certain amount, or at regular intervals, and the biometric state estimation model 69 is updated. The above process of generating from a specific driver is performed each time the biometric data 64-3 increases by a certain amount, or at regular intervals, and the biometric state estimation model 69 is updated.

The biometric data 64-3 used to generate the biometric state estimation model 69 differs from the biometric data 64-1 used to generate the accident risk prediction model 71 in that it does not need to correspond to the in-vehicle sensor data 61-2 or to cases where the risk of an accident is high, and there are fewer limitations on the measurement scenes.

Therefore, since the biological data 64-3 used to generate the biological state estimation model 69 is easy to measure, the data can be measured in a shorter period of time than the generation of the accident risk prediction model 71, and it is therefore expected that the update frequency can be set higher.

Furthermore, because the occurrence of dangerous conditions in a broad sense is rare, it takes time to construct an accident risk prediction model 71 only from a group of multiple drivers with similar biological responses. On the other hand, in this embodiment, the individual differences in biological responses between multiple drivers are absorbed in the biological state estimation model 69, not the accident risk prediction model 71, so it is possible to generate a model that utilizes data on drivers with dissimilar biological responses when generating the accident risk prediction model 71. And, it is possible to generate a model only from data on drivers with similar biological characteristics that can be easily collected while driving.

As a result, this embodiment can reduce the time required to generate a model required to realize hazard prediction processing, compared to a method that absorbs individual differences in biological responses when generating an accident risk prediction model 71.

Figure 5 is a flowchart showing an example of the calculation process of the biometric data 64. This process is performed by the biometric data calculation unit 31 each time data is received from the vehicle 7.

The data collection unit 39 of the safe driving support server 1 receives data from the on-board sensor 8 and biosensor 12 from the driving data collection device 10 of the vehicle 7 (S21).

The data collection unit 39 stores the heart rate data detected by the heart rate sensor 121 and the acceleration data detected by the acceleration sensor 122 in the biosensor data 63 (S22). The data from the in-vehicle sensor 8 is stored in the in-vehicle sensor data 61.

Next, the biometric data calculation unit 31 calculates the biometric data 64 using the biometric sensor data 63 as input (S23) using a method described later in FIG. 6. The biometric data calculation unit 31 stores the calculated data in the biometric data 64 (S24). Through the above process, the biometric data 64 is calculated from the biometric sensor data 63.

Figure 6 is a flowchart showing an example of a calculation process of the biometric data 64 using heart rate data. This process is performed by the biometric data calculation unit 31 each time the biometric sensor data 63 is stored in the biometric data 64.

Since the calculation process of the biometric data 64 using the heart rate data is performed by a known or publicly known method, only an outline of the process will be described below. First, the biometric data calculation unit 31 reads the heart rate data from the biometric sensor data 63 (S31). Note that the heart rate data refers to data from which the interval between heart rates obtained from the driver while driving by the heart rate sensor 121 can be extracted, and specific examples of the data include electrocardiogram data, pulse wave data, and heart sound data.

Then, the biological data calculation unit 31 extracts the beat interval IBI (Inter Beat Interval) from the heartbeat data and calculates its time series (S32). In the case of electrocardiogram data, the IBI may be calculated as an RRI (R-R Interval) time series that uses the interval between electrocardiogram R waves. In the following, the RRI will be mentioned in particular as an example.

The biological data calculation unit 31 calculates the autonomic nerve function index by performing analysis as needed from the RRI. Examples of analysis as needed include frequency domain analysis, time domain analysis, and RRI nonlinear domain analysis.

In the frequency domain analysis, the biological data calculation unit 31 calculates frequency domain indices from the RRI time series via power spectral density (S33). Since the RRI time series is irregularly spaced time series data, it is resampled at regular intervals using spline interpolation or the like, and then the power spectral density PSD (Power Spectral Density) is calculated using an autoregressive model or maximum entropy method, or a known method such as the Lomb-Scargle method that can use irregularly spaced data.

The biometric data calculation unit 31 calculates, from the calculated PSD, for example, the integral value LF of the low frequency region of 0.05 Hz-0.15 Hz, the integral value HF of the high frequency region of 0.15 Hz-0.40 Hz, TP which is the sum of LF and HF, LF/HF which is LF divided by HF, and LFnu which is LF divided by TP as a percentage, as frequency domain indices of autonomic nerve function indexes.

In the time domain analysis, the biological data calculation unit 31 calculates a time domain index by calculating statistics of the RRI time series and the ΔRRI time series, which is a difference series between adjacent RRIs (S34). For example, the average heart rate, which is the reciprocal of the average value of the RRI time series, and SDNN, which is the standard deviation of the RRI, are calculated. In addition, from the ΔRRI time series, for example, NN50, which is the total number of data whose absolute value of the difference value constituting the ΔRRI time series exceeds 50 ms, pNN50, which is NN50 divided by the total number of data in the ΔRRI time series, and SDSD, which is the standard deviation of the ΔRRI, are calculated and calculated as a time domain index of the autonomic nerve function index.

In the RRI nonlinear region analysis, the biometric data calculation unit 31 calculates nonlinear features using various methods (S35). For example, the biometric data calculation unit 31 calculates an ellipse area S that approximates the plotted region through Poincaré plot analysis in which the RRI time series RRI(t) is plotted on the X-axis and the time series RRI(t+1) obtained by advancing the RRI time series by one time is plotted on the Y-axis. In addition, the biometric data calculation unit 31 calculates similarity entropy, α1 and α2 by Detrended Fluctuation Analysis, and tone and entropy based on tone-entropy analysis, and calculates them as the RRI nonlinear region index of the autonomic nerve function index.

The biological data calculation unit 31 can also perform a heart rate nonlinear region analysis on the heart rate data before extracting the RRI to calculate a heart rate nonlinear region index (S36). For example, the biological data calculation unit 31 applies chaos analysis to the heart rate data to calculate the correlation dimension and maximum Lyapunov exponent of the heart rate data, and calculates them as the heart rate nonlinear region index of the autonomic nerve function index.

The calculated autonomic nervous function index reflects the driver's biological state controlled by the autonomic nervous system, so by calculating an index that measures the driver's biological state and presenting it to the driver as a countermeasure to avoid dangerous situations in a broad sense, it is possible to present a warning that matches the biological state. For example, since LF/HF is an index that measures the balance between sympathetic nervous activity, which is mainly increased in stress or tension states, and parasympathetic nervous activity, which is mainly increased in relaxation states, the biological data calculation unit 31 can calculate LF/HF to present a warning that takes into account the driver's tension-relaxation state.

Then, the biometric data calculation unit 31 collectively stores the calculated autonomic nerve function indexes in the biometric data 64 (S24). Through the above process, the autonomic nerve function index is calculated from the heart rate data in the biometric sensor data 63, and stored as the biometric data 64.

The biometric data calculation unit 31 may execute the processes in steps S33 to S36 in parallel or sequentially.

Figure 7 is a flowchart showing an example of a risk prediction process performed by the safe driving support server 1. This process is executed at a predetermined timing, such as each time the biometric data 64-2 is updated. The risk prediction unit 37 acquires the biometric data 64-2 from the latest period to a predetermined period (S12).

Next, the model selection unit 36 identifies the driver ID and vehicle ID to be analyzed from the task and environment data 74, and selects an accident risk prediction model 71 and a biological state estimation model 69 based on the task and environment data 74 (S13 and S14). When selecting the accident risk prediction model 71, for example, the model to be used is selected based on the vehicle size, weather, traffic congestion, etc. of the task and environment data 74.

When selecting the biological state estimation model 69, the model selection unit 36 selects the model to be used based on, for example, information on the biological state estimation model 69 corresponding to the driver ID stored in the attribute information data 75.

Next, the risk prediction unit 37 inputs the biometric data 64-2 of the driver ID to be analyzed into the selected accident risk prediction model 71 to predict the accident risk, and stores the predicted accident risk in the accident risk prediction data 66 (S6A). Note that, although an example of inputting the biometric data 64-2 into the accident risk prediction model 71 has been shown above, this is not limiting, and task/environment data 74 and attribute information data 75 may also be added to the input of the accident risk prediction model 71.

For example, by inputting driving skill information obtained from a driving aptitude test included in the attribute information data 75, or personality and temperament information obtained from the Temperature and Character Inventory (TCI), the Temperature Evaluation of the Memphis, Pisa, Paris, and San Diego Autoquestionnaire (TEMPS-A), the Ten Item Personality Inventory for measuring the Big Five personality, or the NEO-FFI, it is possible to take into account the driver's driving skill and personality and realize more accurate prediction of accident risk.

The risk prediction unit 37 also inputs the biometric data 64-2 of the driver ID to be analyzed into the selected biometric state estimation model 69, estimates multiple person seriousness levels for each biometric data type, and stores them in the person seriousness data 67 (S6B). Note that, although an example of inputting the biometric data 64-2 as a single variable has been shown above, the risk prediction unit 37 may also input multivariate biometric data 64-2 to calculate the person seriousness level.

Next, the danger prediction unit 37 extracts contribution indices that contributed to the prediction of the accident risk using the predicted accident risk prediction data 66, the accident risk prediction model 71 used to predict the accident risk, the biometric data 64-2 of the driver ID to be analyzed, and the estimated person seriousness data 67, and matches the contribution indices, the person seriousness of the contribution indices, the accident risk, and the accident risk prediction model 71 information, and stores them in the combination table 68 (S7A).

By the above process, at a specified timing, such as after the biometric data 64-2 is updated, the contribution index, the individual's seriousness of the contribution index, the accident risk, and the accident risk prediction model information used to detect a broadly defined dangerous condition and issue a warning are stored in the combination table 68.

The risk prediction unit 37 may execute the processes of steps S13 and S14 in parallel or sequentially.

Figure 8 is a flowchart showing an example of a warning display process performed by the safe driving support server 1. This process is executed at a predetermined timing, such as each time the connection table 68 is updated.

The warning presentation unit 38 reads the notification definition, which is the condition for presenting a warning, from the alert definition data 72 (S41). A plurality of notification definitions may be set according to the accident risk prediction model 71 and the biological state estimation model 69. In addition, a plurality of notification definitions may be set according to the target and method of presenting a warning, such as using a voice notification from the prediction result notification device 9 to the driver, displaying on a display, displaying on the display of the driver manager's input/output device 5, or sending an email notification via the input/output device 5.

The warning presentation unit 38 refers to the accident risk and personal severity of each record in the combined table 68 (S42) and determines whether a driver has exceeded the threshold set in the alert definition (S7B).

When a driver exceeds the threshold and is judged to be an alert driver, the warning presentation unit 38 generates a countermeasure plan to be included in the warning to be presented (S7C). The countermeasure plan is generated by referring to the countermeasure plan data 73 corresponding to the contribution index in the combination table 68.

In addition, in order to increase the acceptability of the warning to the driver who receives the warning, the warning presentation unit 38 may add information to the proposed countermeasures that indicates how abnormal the condition is for the driver by using the severity of the contribution index in the combination table 68. For example, the warning presentation unit 38 may add information such as "This condition occurs only once every N years."

Furthermore, the warning presentation unit 38 can add information based on the task/environment data 74, attribute information data 75, and the combination table 68 to the proposed countermeasures. For example, the warning presentation unit 38 may update the content to be fed back as a proposed countermeasure based on information on the time N at which the risk of an accident is deemed to be high from the information on the accident risk prediction model 71 stored in the combination table 68. Also, for example, the warning presentation unit 38 may update the content of the feedback based on weather and temperature information stored in the task/environment data 74.

Then, the warning presentation unit 38 acquires the vehicle ID of the transmission target from the driver ID based on the business and environment data 74, and sets the vehicle 7 as the transmission target. Note that, when the transmission target is the driver manager rather than the driver, the warning presentation unit 38 may specify the manager ID instead of acquiring the vehicle ID, and set the input/output device 5 of the safe driving support server 1 operated by the manager as the transmission target.

Then, the warning presentation unit 38 sends a warning as an alert or message to the generated countermeasure plan with an accident risk alert based on the alarm definition that was determined to be triggered (S8).

By the above process, a warning is sent to the vehicle of a driver who is detected as being in a broadly defined dangerous state. In the vehicle 7 that receives the warning, the prediction result notification device 9 notifies the driver of the warning. In addition, in the safe driving support server 1, the warning presentation unit 38 displays the vehicle that sent the warning on the display of the input/output device 5.

Figures 9A and 9B are diagrams showing an example of a warning presentation screen output by the prediction result notification device 9. The prediction result notification device 9 has a display (not shown), and displays a warning presentation screen 1000A when it receives a warning from the safe driving support server 1. The warning presentation screen 1000A includes an area 1010 that displays an accident risk alert, and a comment area 1020 that displays countermeasures for avoiding dangerous situations in a broad sense.

In the area 1010 that displays the accident risk alert, for example, in addition to a warning message 1011 about an increased accident risk, information 1012 based on the personal severity of the contributing indicator biometric data 64-2 can be displayed, so that rather than simply notifying the driver that a dangerous situation is occurring, the warning can be presented in a form that the driver can understand based on the personal severity of the contributing indicator, such as "For you, this is a situation that occurs only once every six months," and this is expected to improve the driver's acceptance of the warning.

In addition, when displaying the information 1012 based on the individual's severity, the display may be made in such a way that it is clear whether it is based on a biological state estimation model 69 generated only from the target driver's own biological data 64-2, or based on a biological state estimation model 69 generated from the biological data of multiple drivers who show biological response characteristics similar to that of the target driver.

For example, in the warning display screen 1000B of FIG. 9B, the information 1012B based on the individual's severity in the area 1010 displaying the accident risk alert may be displayed in a form that makes it clear that the information is being compared with data on multiple drivers, such as "This is a condition that occurs only once every six months for drivers similar to you."

For example, when a comparison is being made with multiple drivers, information 1013 regarding the group of drivers being compared may be displayed so that the driver can understand the comparison. Also, which biological state estimation model 69 is being used may be displayed using an icon or the like. This is expected to have the effect of making it easier for the driver to perceive the warning as something that concerns them personally.

Furthermore, by presenting, for example, an interpretation 1021 of the physiological state based on the contribution index and specific countermeasures 1022 for avoiding a dangerous situation in a broad sense in the comment area 1020 of the warning presentation screens 1000A and 1000B, the driver who receives the warning can understand and take the next action to resolve the dangerous situation, rather than simply receiving the warning and leaving it at that.

Note that, although this example shows an example of presenting a warning using a warning display screen, the warning may be presented by other methods. For example, the warning may be presented in the form of a mechanically read-out text that is equivalent to the content displayed on the warning display screen.

<Data structure>
Next, we will show the characteristic structure of each piece of data used in the safe driving support system.

Figure 10A shows an example of the data structure of biometric data 64 held in the safe driving support system of the present invention. In addition to the driver ID 401, vehicle ID 402, and date and time 403, the biometric data 64 stores various autonomic nerve function indices calculated by the biometric data calculation unit 31. Here, examples of the autonomic nerve function indices include LF/HF 404, which is a frequency domain index, average heart rate 405, which is a time domain index, NN50 406, or α1 (407), which is an RRI nonlinear domain index, etc.

Figure 10B shows an example of the data structure of task and environment data 74 held in the safe driving support system of the present invention. Task and environment data 74 stores task information and driving environment information during the driver's work.

Typically, in addition to the driver ID 411, the start date and time 412 and end date and time 413 indicating the duration of the information, and the vehicle ID 414 being used, the vehicle type 415 indicating, for example, the size of the vehicle being driven, the weather 416 and temperature 417 at a representative point in the driving area, the task type 418 indicating the task status during the period from the start date and time 412 to the end date and time 413, and the task schedule 419 such as the destination are stored.

Figure 10C shows an example of the data structure of attribute information data 75 held in the safe driving support system of the present invention. The attribute information data 75 stores basic information about the driver that does not change over time.

Typically, in addition to the driver ID 421, the start date 422 and end date 423, which are the duration for which the basic information applies, and the vehicle ID 424 of the vehicle being used, driving skill information measured in a driver aptitude diagnostic test or the like (e.g., skill A 425, pseudonym), personality and temperament information measured in the Temperature and Character Inventory (TCI) or the like (e.g., personality A (426), pseudonym), gender 427, age 428, driving experience 429, etc. are stored. In addition to the above, various model information that is preferentially used for the driver, such as a biological state estimation model ID 430 and an accident risk prediction model ID 431, are stored.

Figure 10D shows an example of the data structure of the learning information data 76 held in the safe driving support system of the present invention. The learning information data 76 stores information used when generating the biological state estimation model 69, and typically stores a driver ID 441 and a start time 442 and an end time 443, which are period information of the biological data 64 used for learning.

Furthermore, the learning information data 76 may store the measurement conditions 444 of the data of the period information in order to align the measurement conditions of the data used to generate the biological state estimation model 69. In this case, the measurement conditions 444 may be stored by the driving data collection device 10 automatically determining the state from the in-vehicle sensor data 61-2, etc. (e.g., high-speed driving), or the driver may manually input from among the defined conditions via the work state input device 13 (e.g., lunch break).

When the biological state estimation model 69 is generated as a supervised model, in addition to the above, a value 445 of subjective fatigue level measured by VAS, for example, may be stored as a target variable required for generating the supervised model.

Figure 10E shows an example of the data structure of danger occurrence data 62 held in the safe driving support system of the present invention. The danger occurrence data 62 stores data that sets the type of dangerous event detected by the driver, the driver manager, a commercially available vehicle alarm, etc., and the date and time of occurrence.

Typically, the vehicle ID 451, the date and time of occurrence of the dangerous event 452, and the type of danger 453 are stored. In addition, the importance of the event 454, which records the seriousness of the situation by visual confirmation or the like, may also be stored for the detected dangerous event.

Figure 10F shows an example of the data structure of accident risk prediction data 66 held in the safe driving support system of the present invention.

Accident risk prediction data 66 stores the accident risk determined from biometric data 64-2 using accident risk prediction model 71. Typically, driver ID 461, date and time 462, ID 463 of accident risk prediction model 71 used to predict the accident risk, and accident risk 464 indicating the occurrence probability (percentage) of the predicted hazardous event are stored.

Figure 10G shows an example of the data structure of the person's seriousness data 67 stored in the safe driving support system of the present invention.

The person seriousness data 67 stores the person seriousness determined from the biometric data 64-2 using the biometric state estimation model 69. Typically, the driver ID 471, date and time 472, type 473 of the biometric data 64-2 from which the person seriousness was estimated, ID 474 of the biometric state estimation model 69 used for the estimation, and the person seriousness 475 are stored.

Note that the biological condition estimation model 69 is generated for each type of biological data, and the method of calculating the severity of the individual is expected to differ for each type of biological data, so the output format of the severity of the individual is not necessarily constant.

For example, in the first record, the type 473 of the biometric data 64-2 outputs a statistic that approximates the normal distribution for the "average heart rate," and the individual's severity level 475 is shown to be a value of "+4 SD" from the average, using standard deviation (SD).

For example, the record on the second line outputs the quantile of the data distribution for NN50, indicating that the individual severity level of 475 is the 99% quantile of the data set used to build the model.

Figure 10H shows an example of the data structure of a combination table 68 stored in the safe driving support system of the present invention. The combination table 68 is a table generated by matching values used in the warning presentation process.

Typically, the driver ID 481, vehicle ID 482, date and time 483, accident risk prediction model ID 484 used for the prediction, predicted accident risk 485, contribution index 486 which is the type of biometric data that contributed to the accident risk prediction, biometric condition estimation model ID 487 used to estimate the biometric condition in relation to the contribution index, and the person's severity in relation to the contribution index 488 are stored.

Furthermore, when a record is determined to be in a broadly defined dangerous state, a warning issued flag 489 may be stored, which is a flag that manages whether or not the warning issuing unit 38 has already issued a warning.

Figure 10I shows an example of the data structure of alert definition data 72 held in the safe driving support system of the present invention. The alert definition data 72 stores judgment conditions used to judge whether or not a situation corresponds to a dangerous state in the broad sense.

Typically, an alert ID 491, an accident risk threshold 492, and a person seriousness threshold 493 are stored. In addition, for example, a contribution index condition 494 for limiting the contribution index when using the person seriousness threshold 493, a weather condition 495 for limiting the target weather during work, and a person seriousness type 496 for limiting the judgment target to only person seriousness obtained in a specific output format when making an alert judgment may be stored.

Figure 10J shows an example of the data structure of countermeasure proposal data 73 held in the safe driving support system of the present invention. The countermeasure proposal data stores information used to generate countermeasure proposals for issuing warnings.

Typically, the countermeasure proposal ID 501, the contribution index 502 that is the condition for presenting the countermeasure proposal, the phenomenon 503 that indicates the state of the contribution index, the interpretation 504 of the driver's physiological state assumed from the phenomenon of the contribution index, and the countermeasure proposal 505 for avoiding the corresponding broadly defined dangerous state are stored.

For example, multiple countermeasure proposals 505 may be set for the same contribution index 502, phenomenon 503, and interpretation 504. In this case, the priority 506 for presenting the countermeasure proposal 505 may be further stored.

As described above, the safe driving support system of this embodiment detects whether a broadly defined dangerous state exists based on accident risk prediction data 66 calculated by inputting biometric data 64-2 calculated by the safe driving support server 1 from the heart rate data of the biometric sensor 12 into an accident risk prediction model 71, and the person's seriousness data 67 calculated by inputting the biometric data 64-2 into a biometric state estimation model 69.

As a result, the safe driving support server 1 of this embodiment is able to absorb individual differences in the biometric data 64-2 and detect dangerous conditions before they appear in the behavior of the vehicle 7. Furthermore, when a dangerous condition is detected in a broad sense, the safe driving support server 1 presents a warning including proposed measures to avoid the dangerous condition in a broad sense based on the contribution index that contributed to the prediction of the accident risk, thereby allowing the driver who receives the warning to take the action required to avoid the dangerous condition in a broad sense. As a result, the safe driving support server 1 is able to allow the driver of the vehicle 7 to avoid dangerous conditions in a broad sense before they occur.

In the above embodiment 1, an example of applying the present invention to a traffic support system that manages vehicles 7 is shown, but the present invention is not limited to this. For example, instead of vehicles 7, the present invention can be applied to moving bodies that require a driver or operator, such as railway vehicles, ships, or aircraft.

Next, a second embodiment of the present invention will be described. Except for the differences described below, each part of the safe driving support system of the second embodiment has the same function as each part of the first embodiment with the same reference numerals, and therefore the description of those parts will be omitted.

Figure 11 is a diagram showing an overview of the processing performed by the safe driving support system of the second embodiment. In the second embodiment of the present invention, we will explain how the calculation takes into account individual differences in biometric data when estimating accident risk.

The biometric data 64-1 is input to the biometric state estimation model 69 to calculate the person's seriousness data 67A (first person's seriousness) (S6C), which is then input to the accident risk prediction model generation unit 35.

The accident risk prediction model generation unit 35 generates an accident risk prediction model 71 by inputting the biometric data 64-1, the accident risk estimation data 65, and the person's severity data 67A (S4).

Then, the risk prediction unit 37 inputs the biometric data 64-2 to the biometric state estimation model 69 to calculate the person's seriousness data 67B (second person's seriousness) (S6B), and outputs the person's seriousness data 67B to the accident risk estimation unit 33. The accident risk estimation unit 33 inputs the person's seriousness data 67A in addition to the biometric data 64-2 to the input of the accident risk prediction model 71 (S6A).

The process in step S7A in which the risk prediction unit 37 uses the individual seriousness data 67B to extract contribution indices and generate the combined table 68 is the same as in Example 1.

Therefore, in the process of extracting and combining contribution indices performed by the risk prediction unit 37, the input that contributed to the accident risk estimation is selected from the biometric data 64-2 and the person seriousness data 67B as a contribution index, and when the biometric data 64-2 is selected, the person seriousness of the selected type of biometric data 64-2 is stored in the combination table 68, and when the person seriousness data 67B is selected, the person seriousness data 67B of the selected type is also stored in the combination table 68.

As a result, in Example 2, in addition to determining whether to present a warning as described above, it is also possible to estimate the risk of an accident by taking into account individual differences in biometric data 64, thereby achieving the effect of making it possible to predict the risk of an accident with greater accuracy.

Next, a third embodiment of the present invention will be described. Except for the differences described below, each part of the safe driving support system of the third embodiment has the same function as each part of the first embodiment with the same reference numerals, and therefore the description of those parts will be omitted.

Figure 12 is a diagram showing an overview of the processing performed by the safe driving support system of Example 3. In Figure 12, the learning phase of the processing overview in Figure 2 is omitted, and an overview of the processing is described with respect to the differences in the operation phase.

The safe driving support server 1 of Example 3 is an example in which the warning presentation unit 38 does not present a warning each time the danger prediction unit 37 adds a record to the combined table 68, but presents a list of targets for which a warning should be presented to the driver manager, allowing the driver manager to manually intervene to support the driver's safe driving.

Each time the danger prediction unit 37 of the safe driving support server 1 updates the connection table 68, or at regular intervals, the warning presentation unit 38 reads the connection table 68 and updates intervention candidates for which the driver manager may intervene (S15). For example, only records in the connection table 68 for which the report issued flag 489 is not "issued" and which are from the most recent N hours on the day of operation are extracted.

Then, the warning presentation unit 38 performs an alarm determination for the extracted record groups using the method described above, and extracts record groups that correspond to a broadly defined dangerous state (S7B). If there are record groups that correspond to a broadly defined dangerous state, the warning presentation unit 38 generates countermeasures for each of the record groups to avoid the broadly defined dangerous state (S7C).

The warning presentation unit 38 displays a list of warning destinations, which are records for which countermeasure proposals have been generated, on the display of the input/output device 5 (S16). In the list of warning destinations, warning destinations that are considered to require manual intervention by the driver manager are displayed at the top with priority, allowing intervention by the driver manager.

For example, the warning presentation unit 38 can compare the business and environment data 74 with records that correspond to dangerous situations in a broad sense, and display at the top a list of records that are presented for business reasons and that the driver cannot take at his own discretion.

The input/output device 5 accepts intervention by the driver manager with respect to the driver based on the list of warning recipients displayed on the display (S17). For example, if a driver is shown a suggested countermeasure such as "Please take a rest at a nearby rest point," but there is no rest point nearby and the delivery time for the package is approaching, the driver manager can use an interface for voice communication with the driver, which is displayed in the list of warning recipients, to check the driver's own condition.

If it is determined that the driver does not need to take a break immediately, the driver manager can provide the driver with a second-best solution by showing him/her a place where he/she can take a break after completing a delivery.

For example, if a driver is repeatedly warned that he or she is "feeling sleepy and needs to rest," and the driver manager checks the driver's condition and determines that it will be difficult for the driver to complete the original work schedule, the manager can adjust the driver's workload and allow the driver to rest by skipping some of the work scheduled for the latter half of the day's work schedule or first thing the next morning.

Figure 13 is a diagram showing an example of a warning destination list screen 2000 output by the display of the input/output device 5. When the warning presentation unit 38 updates the extracted intervention candidates and generates a countermeasure proposal, the input/output device 5 updates the warning destination list and displays the warning destination list screen 2000.

At the top of the warning destination list screen 2000, the current time 2001 and an administrator ID 2002, which is the ID of the driver manager viewing the warning destination list screen 2000, are displayed. In the warning destination list 2003, a group of records for which countermeasures have been extracted and generated by the warning presentation unit 38 are displayed in a table format. By pressing the downward triangular button in the figure for each column, the group of records can be sorted by priority order according to the elements that the driver manager wants to check.

The lower area 2005 displays details of the driver corresponding to the record selected by the driver manager for action. In the illustrated example, information is displayed for the driver ID "ZAX001" corresponding to record 2004 in the warning destination list 2003.

In area 2005, a current location 2011 showing the driver's location on a map, a countermeasure proposal 2012 relating to the selected record 2004, the driver's work schedule 2013, notes 2014 showing the driver's work status, and a driver contact button 2015 for directly communicating with the driver are displayed. In addition, the time progression of biometric data 64 relating to the contribution index for the day may also be displayed.

The driver's current location 2011 shows the target driver's location 2021, the actual driving route for that day 2022 (solid line), and the planned driving route 2022 (dotted line) on a map. In addition, the locations 2023 of other drivers around the driver are also displayed on the map.

The driver manager refers to the records 2004, countermeasure proposals 2012, work schedule 2013, etc., and judges through the warning destination list screen 2000 whether the driver can take countermeasures at his/her own discretion or whether it is necessary to directly check the driver's condition. Then, the driver manager can manually intervene to avoid a dangerous situation in a broad sense by pressing the driver contact button 2015 as necessary and directly communicating with the driver.

As a result, the driver manager can avoid dangerous situations in a broad sense and support the safe driving of the group of drivers under his/her management by having the driver execute countermeasure plan 2012 that would be difficult for the driver to execute at his/her own discretion, or by having the driver execute a second-best countermeasure plan that is different from countermeasure plan 2012.

As described above, in the third embodiment, when a driver is in a dangerous state in a broad sense but has a countermeasure that is difficult for the driver to implement at his/her discretion, the driver manager gives permission to implement the countermeasure based on the driver's own condition, and the driver is able to take a more fundamental countermeasure to avoid the dangerous state in a broad sense. As a result, the safe driving support system has the effect of supporting the driver to take more flexible countermeasures to avoid the dangerous state in a broad sense.

<Conclusion>
As described above, the safe driving support server 1 of the first to third embodiments can be configured as follows.

(1) A safe driving support method in which a computer (safe driving support server 1) having a processor (2) and a memory (3) supports the driving of a vehicle (7), comprising the steps of: (1) generating an accident risk definition model (70) that estimates the probability of a dangerous event occurring as an accident risk using first on-board sensor data (61-1) indicating the driving state of the vehicle (7) collected in the past and preset danger occurrence data (62) as input; a second step of inputting second on-board sensor data (61-2) indicating a state into the accident risk definition model (70) to estimate the probability of a hazardous event occurring and generate accident risk estimation data (65); and an accident risk prediction model in which the computer inputs first biometric data (64-1) calculated in advance from first biometric sensor data (63) of the driver when the second on-board sensor data (61-2) was collected and the accident risk estimation data (65) to predict an accident risk after a predetermined time. a third step of generating a biological state estimation model (71) by the computer; a fourth step of generating a biological state estimation model (69) by inputting a third biological data (64-3) calculated in advance and calculating a person's seriousness (person's seriousness data 67) by the computer; a fifth step of acquiring second biological sensor data (63) of the driver while driving from a biological sensor (12) of the vehicle (7) and calculating a biological state estimation model (64-2) indicating the state of the driver from the second biological sensor data (63); A safe driving support method characterized by including a sixth step in which a computer inputs the biological state estimation model (64-2) into the accident risk prediction model (71) to predict the accident risk, and inputs the biological state estimation model (64-2) into the biological state estimation model (69) to calculate the person's seriousness (67), and a seventh step in which the computer determines whether or not to issue a warning based on the predicted accident risk (accident risk prediction data 66) and the calculated person's seriousness (67).

With the above configuration, the safe driving support server 1 is able to detect dangerous conditions (dangerous events) in a broad sense based on the accident risk prediction model 71, the biological state estimation model 69, and the biological data 641.

(2) The safe driving support method described in (1) above, further comprising: an eighth step of extracting the accident risk prediction model (71), the accident risk (66), and a contribution index, which is the type of biometric data that contributed to the prediction of the accident risk (66), when the computer determines that the warning should be issued; and a ninth step of generating a countermeasure plan for avoiding a dangerous event based on the personal seriousness (67) of the contribution index.

With the above configuration, the safe driving support server 1 can detect dangerous conditions (dangerous events) in a broad sense based on the accident risk prediction model 71, the biological state estimation model 69, and the biological data 641, and when issuing a warning, can present proposed measures to avoid dangerous conditions in a broad sense.

(3) The safe driving support method described in (2) above, further comprising a tenth step in which the computer generates and outputs a warning including a countermeasure plan (countermeasure plan data 73) for avoiding the dangerous event.

With the above configuration, the safe driving support server 1 can detect dangerous conditions (dangerous events) in a broad sense based on the accident risk prediction model 71, the biological state estimation model 69, and the biological data 641, and when issuing a warning, generate countermeasures to avoid the dangerous conditions in a broad sense and issue the warning.

(4) A safe driving support method as described in (1) above, characterized in that in the fourth step, the first biometric data (64-1) and learning information data (76) including at least a target period of the first biometric data (64-1) to be used for generating the biometric state estimation model (69) are input to generate the biometric state estimation model (69) for calculating the person's severity (67).

With the above configuration, when selecting the biological data 64-3 to be input for clustering, the biological state estimation model generation unit 34 may select only the biological data 64-3 measured under the same measurement conditions based on the learning information data 76, and input only the biological data 64-3 for the period stored in the learning information data 76. This is expected to control the measurement conditions and suppress the variation in the biological data 64-3.

(5) The safe driving support method described in (1) above, wherein in the sixth step, a combination table (68) is generated that associates and stores the accident risk (66), the contribution index, and the person's seriousness (67) related to the contribution index.

With the above configuration, at a specified timing, such as after the biometric data 64-2 is updated, the contribution index used to detect a broadly defined dangerous condition and issue a warning, the individual's severity of the contribution index, the accident risk, and accident risk prediction model information are stored in the combination table 68.

(6) The safe driving support method described in (1) above, characterized in that in the tenth step, a warning including information based on the personal seriousness (67) and a proposed measure to avoid the dangerous event is displayed.

With the above configuration, in the area 1010 that displays the accident risk alert, in addition to a warning message 1011 about an increased accident risk, information 1012 based on the personal severity of the contributing index biometric data 64-2 can be displayed. This not only notifies the driver that a dangerous situation is occurring, but also presents a warning in a form that the driver can understand based on the personal severity of the contributing index, such as "For you, this is a situation that occurs only once every six months." This is expected to improve the driver's acceptance of warnings.

(7) The safe driving support method described in (1) above, further comprising a twelfth step in which the computer inputs the first biometric data (64-1) into the biometric state estimation model (69) to calculate a first person severity (67A), and in the third step, the first biometric data (64-1) calculated in advance from the driver's first biometric sensor data (63) when the second in-vehicle sensor data (61-2) was collected, the accident risk estimation data (65), and the first person severity (67A) are input into the biometric state estimation model (69) to calculate a first person severity (67A). A safe driving support method, characterized in that an accident risk prediction model (71) is generated that predicts an accident risk (66) after a predetermined time using the person seriousness (67) as input, and in the sixth step, the biological state estimation model (64-2) is input into the biological state estimation model (69) to calculate a second person seriousness (67B), and the biological state estimation model (64-2) and the second person seriousness (67B) are input into the accident risk prediction model (71) to predict the accident risk (66).

With the above configuration, the safe driving support server 1 can estimate the accident risk by taking into account individual differences in the biometric data 64, in addition to determining whether to issue a warning, and can achieve the effect of making it possible to predict the accident risk with greater accuracy.

The present invention is not limited to the above-described embodiments, but includes various modified examples. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and are not necessarily limited to those having all of the configurations described. It is also possible to replace part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

The above configurations, functions, processing units, processing means, etc. may be realized in part or in whole in hardware, for example by designing them as integrated circuits. Also, the above configurations, functions, etc. may be realized in software by the processor 2 interpreting and executing a program that realizes each function. Information on the programs, tables, files, etc. that realize each function can be stored in a storage device such as the memory 3, a hard disk drive, or an SSD (Solid State Drive), or in a computer-readable non-transitory data storage medium such as an IC card, an SD card, or a DVD.

The drawings also show control lines and information lines that are considered necessary to explain the embodiments, and do not necessarily show all of the control lines and information lines contained in an actual product to which the present invention is applied. In reality, it can be assumed that almost all components are interconnected.

LIST OF SYMBOLS 1 Operation support server 2 Processor 3 Memory 4 Storage device 7 Vehicle 8 On-vehicle sensor 12 Biometric sensor 31 Biometric data calculation unit 32 Accident risk definition generation unit 33 Accident risk estimation unit 34 Biometric state estimation model generation unit 35 Accident risk prediction model generation unit 36 Model selection unit 37 Risk prediction unit 38 Warning presentation unit 61 On-vehicle sensor data 62 Risk occurrence data 63 Biometric sensor data 64 Biometric data 65 Accident risk estimation data 66 Accident risk prediction data 67 Person seriousness data 68 Joint table 69 Biometric state estimation model 70 Accident risk definition model 71 Accident risk prediction model 72 Alert definition data 73 Countermeasure proposal data 74 Business/environment data 75 Attribute information data 76 Learning information data

Claims (20)

A safe driving support method for supporting driving of a vehicle by a computer having a processor and a memory, comprising:
A first step in which the computer generates an accident risk definition model that estimates the probability of a hazardous event occurring as an accident risk using first on-board sensor data indicating a vehicle's traveling state collected in the past and preset hazard occurrence data as inputs;
a second step in which the computer inputs second on-board sensor data indicating a vehicle driving state collected in the past into the accident risk definition model, and estimates a probability of a hazardous event occurring to generate accident risk estimation data;
a third step in which the computer generates an accident risk prediction model that predicts an accident risk after a predetermined time period using as input first biometric data calculated in advance from first biometric sensor data of the driver when the second in-vehicle sensor data is collected and the accident risk estimation data;
a fourth step in which the computer inputs third biometric data previously collected using any one of a wearable device worn by the driver, a sensing device mounted on the vehicle, and an image recognition system that analyzes an image of the driver, and generates a biological state estimation model that calculates a personal severity indicating how much second biometric data calculated from second biometric sensor data of the driver while driving deviates from the third biometric data;
a fifth step of the computer acquiring the second biosensor data and calculating the second biodata from the second biosensor data;
A sixth step in which the computer inputs the second biological data into the accident risk prediction model to predict the accident risk, and inputs the second biological data into the biological state estimation model to calculate the person's severity;
A seventh step in which the computer determines whether or not to issue a warning based on the predicted accident risk and the calculated personal seriousness;
A safe driving support method comprising: The safe driving support method according to claim 1,
an eighth step of extracting, when it is determined that the warning should be issued, the accident risk prediction model, the accident risk, and a contribution index which is a type of biometric data that contributed to the prediction of the accident risk;
A ninth step in which the computer generates a countermeasure plan for avoiding the hazardous event based on the personal seriousness regarding the contribution index;
The safe driving support method further comprises: The safe driving support method according to claim 2,
The safe driving support method further comprises a tenth step in which the computer generates and outputs a warning including the countermeasure proposal. The safe driving support method according to claim 1,
In the fourth step,
A safe driving support method characterized by generating the biological condition estimation model that calculates the person's severity by inputting the first biological data and learning information data that includes a target period of at least the first biological data to be used for generating the biological condition estimation model. The safe driving support method according to claim 2,
In the sixth step,
A safe driving support method, characterized by generating a combination table that associates and stores the accident risk, the contribution index, and the person's seriousness regarding the contribution index. The safe driving support method according to claim 3,
In the tenth step,
A safe driving support method, characterized by displaying information based on the individual's seriousness and a warning including the proposed countermeasures. The safe driving support method according to claim 1,
The method further includes a twelfth step of inputting the first biological data into the biological state estimation model to calculate a first person severity level,
In the third step,
Generate the accident risk prediction model using the first biometric data, the accident risk estimation data, and the first person severity as inputs;
In the sixth step,
A safe driving support method characterized by inputting the second biometric data into the biometric condition estimation model to calculate a second person's seriousness, and inputting the second biometric data and the second person's seriousness into the accident risk prediction model to predict the accident risk. a server having a processor and a memory;
A driving assistance system that supports driving of a vehicle including a vehicle having an on-board sensor that detects a driving state and a biosensor that detects biometric data of a driver,
The server,
an accident risk definition generation unit that generates an accident risk definition model that estimates the probability of a hazardous event occurring as an accident risk using first on-board sensor data indicating a traveling state of the vehicle collected in the past and preset hazard occurrence data as input;
an accident risk estimation unit that inputs second on-board sensor data that indicates a traveling state of the vehicle that has been collected in the past into the accident risk definition model, estimates a probability of a hazardous event occurring, and generates accident risk estimation data;
an accident risk prediction model generation unit that generates an accident risk prediction model that predicts an accident risk after a predetermined time using as input first biological data calculated in advance from first biological sensor data of the driver when the second in-vehicle sensor data was collected and the accident risk estimation data;
a biological state estimation unit that generates a biological state estimation model that receives third biological data previously collected using a wearable device worn by the driver, a sensing device mounted on the vehicle, or an image recognition system that analyzes an image of the driver, and calculates a personal severity level that indicates how much second biological data calculated from second biological sensor data of the driver while driving deviates from the third biological data;
a biometric data calculation unit that acquires the second biometric sensor data and calculates the second biometric data from the second biometric sensor data;
a danger prediction unit that inputs the second biological data into the accident risk prediction model to predict the accident risk, and also inputs the second biological data into the biological state estimation model to calculate the person's severity;
a warning issuing unit that determines whether or not to issue a warning based on the predicted accident risk and the calculated personal seriousness;
A safe driving support system comprising: The safe driving support system according to claim 8,
The warning presentation unit is
A safe driving support system characterized by generating countermeasures to avoid dangerous events based on the accident risk prediction model, the accident risk, a contribution index which is a type of biometric data that contributed to the prediction of the accident risk, and the individual's seriousness regarding the contribution index when it is determined that the warning should be issued. The safe driving support system according to claim 9,
The warning presentation unit is
A safe driving support system characterized by generating and outputting a warning including the countermeasure proposal. The safe driving support system according to claim 8,
The biological state estimation unit is
A safe driving support system characterized by generating the biological condition estimation model that calculates the person's severity by inputting the first biological data and learning information data that includes a target period of at least the first biological data to be used for generating the biological condition estimation model. The safe driving support system according to claim 9,
The risk prediction unit is
A safe driving support system, characterized by generating a combination table that associates and stores the accident risk, the contribution index, and the person's seriousness regarding the contribution index. The safe driving support system according to claim 8,
The risk prediction unit is
A safe driving support system characterized by displaying information based on the individual's seriousness and a warning including proposed measures to avoid dangerous events. The safe driving support system according to claim 8,
The risk prediction unit inputs the first biological data into the biological state estimation model to calculate a first person seriousness,
The accident risk prediction model generation unit
Generate the accident risk prediction model using the first biometric data, the accident risk estimation data, and the first person severity as inputs;
The risk prediction unit is
A safe driving support system characterized by inputting the second biometric data into the biometric condition estimation model to calculate a second person's seriousness, and inputting the second biometric data and the second person's seriousness into the accident risk prediction model to predict the accident risk. A vehicle operation support server having a processor and a memory, the server supporting a vehicle operation,
an accident risk definition generation unit that generates an accident risk definition model that estimates the probability of a hazardous event occurring as an accident risk using first on-board sensor data indicating a traveling state of the vehicle collected in the past and preset hazard occurrence data as input;
an accident risk estimation unit that inputs second on-board sensor data that indicates a traveling state of the vehicle that has been collected in the past into the accident risk definition model, estimates a probability of a hazardous event occurring, and generates accident risk estimation data;
an accident risk prediction model generation unit that generates an accident risk prediction model that predicts an accident risk after a predetermined time period using as input first biological data calculated in advance from first biological sensor data of the driver when the second in-vehicle sensor data was collected and the accident risk estimation data;
a biological state estimation unit that generates a biological state estimation model that receives third biological data previously collected using a wearable device worn by the driver, a sensing device mounted on the vehicle, or an image recognition system that analyzes an image of the driver, and calculates a personal severity level that indicates how much second biological data calculated from second biological sensor data of the driver while driving deviates from the third biological data;
a biometric data calculation unit that acquires the second biometric sensor data and calculates the second biometric data from the second biometric sensor data;
a danger prediction unit that inputs the second biological data into the accident risk prediction model to predict the accident risk, and also inputs the second biological data into the biological state estimation model to calculate the person's severity;
a warning issuing unit that determines whether or not to issue a warning based on the predicted accident risk and the calculated personal seriousness;
A safe driving support server comprising: The safe driving support server according to claim 15,
The warning presentation unit is
A safe driving support server characterized in that, when it is determined that the warning should be issued, a countermeasure plan for avoiding a dangerous event is generated based on the accident risk prediction model, the accident risk, a contribution index which is a type of biometric data that contributed to the prediction of the accident risk, and the personal seriousness regarding the contribution index. The safe driving support server according to claim 16,
The warning presentation unit is
A safe driving support server that generates and outputs a warning including the countermeasure proposal. The safe driving support server according to claim 15,
The biological state estimation unit is
A safe driving support server characterized by generating the biological condition estimation model that calculates the person's severity by inputting the first biological data and learning information data that includes a target period of at least the first biological data to be used for generating the biological condition estimation model. The safe driving support server according to claim 16,
The risk prediction unit is
A safe driving support server that generates a combination table that associates and stores the accident risk, the contribution index, and the person's seriousness regarding the contribution index. The safe driving support server according to claim 15,
The risk prediction unit is
A safe driving support server that displays information based on the individual's seriousness and a warning including a proposed countermeasure for avoiding a dangerous event.

Images