Operation assistance method, operation assistance system, and operation assistance server

Abstract

The invention provides an operation support method, an operation support system, and an operation support server, which can predict the occurrence risk after a predetermined time for an event according to the biometric data of a driver and the driving state data. The operation assistance method comprises the following steps: generating an accident risk definition model used for calculating a danger occurrence probability as an accident risk; a step of estimating the probability of danger occurrence to generate accident risk estimation data; a step for generating an accident risk prediction model for predicting an accident risk after a predetermined time; a step for calculating second biological index data indicating the state of the driver; predicting an accident risk after the predetermined time by using the accident risk prediction model; a step for calculating second biological index data indicating the state of the driver; a step for calculating the severity of the person using the biological state estimation model; and a step for determining whether or not to issue a warning on the basis of the personal severity.

Description

[0001] This application is a divisional application of the same name as the international application number PCT / JP2021 / 021634 and the application number 202180027297.9 which entered the Chinese national phase on October 9, 2022.

[0002] This application claims the priority of Japanese patent application No. 2020-100060 filed on June 9, 2020 (2020) and Japanese patent application No. 2020-157573 filed on September 18, 2020 (2020), the contents of which are incorporated into this application by reference. Technical Field

[0003] The present invention relates to an operation assistance method, an operation assistance system and an operation assistance server for assisting the operation of a traffic mechanism by predicting the risk of a traffic accident. Background Art

[0004] In recent years, traffic accidents caused by the health and fatigue of drivers in logistics trucks and long-distance buses have become a social problem. In order to prevent traffic accidents, the application of biosensors that monitor the state of the driver while driving and technologies that measure the driving state of the vehicle such as the distance between vehicles and speed in real time are progressing.

[0005] As technologies for assisting the driving operation of a vehicle, there are known patent documents 1 and 2. Patent document 1 discloses a technology for efficiently and accurately assisting driving by predicting driving operations related to sudden braking with a margin based on a table including, in addition to the causal relationship between driving condition data and driving operation data, the causal relationship between physiological state data that is not related to the driving operation and the driving condition.

[0006] Patent document 2 discloses a technology for identifying deviations from the driver's usual internal state by comparing the driving environment risks included in the external environment identified by sensing the external environment of a mobile body with the driver's risk perception state estimated as an internal state based on the driver's driving operations.

[0007] Prior art literature

[0008] Patent Literature

[0009] Patent Document 1: Japanese Patent Application Publication No. 2009-245147

[0010] Patent Document 2: Japanese Patent Application Publication No. 2009-098970 Summary of the invention

[0011] Problems to be solved by the invention

[0012] However, in the above-mentioned prior art, although it is possible to predict the driving operation of sudden braking based on the driver's physiological state data, driving operation data and driving condition data, this is a prediction focusing on specific driving operations and there is a problem that it is difficult to predict events that do not occur frequently, such as traffic accidents.

[0013] Therefore, the present invention has been made in view of the above-mentioned problems, and an object of the present invention is to provide a driving assistance method, a driving assistance system, and a driving assistance server capable of predicting the risk of a traffic accident based on the biological data and driving status data of the driver.

[0014] Technical solutions to solve problems

[0015] The present invention is an operation assistance method for assisting the operation of a vehicle by a computer having a processor and a memory, comprising: a first step of taking first on-vehicle sensor data representing the driving state of the vehicle collected in the past and danger occurrence data in which information on the occurrence of danger is pre-set from the first on-vehicle sensor data as input, and generating an accident risk definition model for estimating the probability of danger occurrence as an accident risk through machine learning; a second step of inputting second on-vehicle sensor data representing the driving state of the vehicle collected in the past into the accident risk definition model, and estimating the probability of danger occurrence to generate accident risk estimation data; a third step of taking first biological indicator data pre-calculated from the biological data of the driver when the second on-vehicle sensor data was collected and the accident risk estimation data as input, and generating an accident risk prediction model for predicting the accident risk after a specified time through machine learning; a fourth step of acquiring the biological data of the driver who is driving the vehicle, and calculating second biological indicator data representing the state of the driver from the biological data; and a fifth step of inputting the second biological indicator data into the accident risk prediction model to predict the accident risk after the specified time.

[0016] Effects of the Invention

[0017] Therefore, according to the present invention, it is possible to predict the risk of an event such as a traffic accident occurring within a predetermined time based on the biological data of the driver and the driving state data.

[0018] Details of at least one embodiment of the subject matter disclosed in this specification will be described in the accompanying drawings and the following description. Other features, methods, and effects of the disclosed subject matter will be described through the following disclosure, drawings, and technical solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a block diagram showing an example of the configuration of the operation support system according to the first embodiment of the present invention.

[0020] Figure 2This is a diagram showing an overview of processing performed in the operation support system according to the first embodiment of the present invention.

[0021] Figure 3 This is a flowchart showing an example of a process for generating an accident risk prediction model performed in an operation support server according to the first embodiment of the present invention.

[0022] Figure 4 This is a flowchart showing an example of a process for generating biometric indicator data performed in an operation supporting server according to the first embodiment of the present invention.

[0023] Figure 5 This is a flowchart showing an example of prediction processing performed in the operation support server according to the first embodiment of the present invention.

[0024] Figure 6 This is a flowchart showing an example of an alarm generation process performed in the operation support server according to the first embodiment of the present invention.

[0025] Fig. 7A This is a flowchart showing an example of a biological indicator calculation process performed in the operation support server according to the first embodiment of the present invention.

[0026] Figure 7B This is a graph showing an example of heartbeat data according to the first embodiment of the present invention.

[0027] Figure 7C This is a graph showing an example of heartbeat fluctuations according to the first embodiment of the present invention.

[0028] Fig.7D This is a graph showing an example of spectral power density of heartbeat fluctuations according to the first embodiment of the present invention.

[0029] Fig. 7E This is a graph showing an example of the autonomic nerve NNXX according to the first embodiment of the present invention.

[0030] Figure 8 This is a diagram showing an example of biological indicator data according to the first embodiment of the present invention.

[0031] Fig. 9 This is a diagram showing an example of business data according to the first embodiment of the present invention.

[0032] Fig.10 This is a diagram showing an example of environmental data according to the first embodiment of the present invention.

[0033] Fig.11 This is a diagram showing an example of vehicle-mounted sensor data according to the first embodiment of the present invention.

[0034] Fig.12 This is a diagram showing an example of risk occurrence data according to the first embodiment of the present invention.

[0035] Fig.13 This is a diagram showing an example of accident risk estimation data according to the first embodiment of the present invention.

[0036] Fig.14 This is a diagram showing an example of alarm definition data according to the first embodiment of the present invention.

[0037] Fig.15 This is a diagram showing an example of accident risk prediction data according to the first embodiment of the present invention.

[0038] Fig.16A This is a diagram showing an example of a prediction result display screen according to the first embodiment of the present invention.

[0039] Fig. 16B This is a diagram showing an example of a prediction result notification screen according to the first embodiment of the present invention.

[0040] Fig.17 This is a block diagram showing an example of the configuration of an operation support system according to the second embodiment of the present invention.

[0041] Fig.18 This is a block diagram showing an example of the configuration of an operation support system according to Embodiment 3 of the present invention.

[0042] Fig.19 This is a diagram showing an overview of processing performed in the operation support system according to the third embodiment of the present invention.

[0043] Fig. 20 This is a flowchart showing an example of a process for generating an accident risk prediction model performed in an operation support server according to a third embodiment of the present invention.

[0044] Fig.21 This is a flowchart showing an example of a process of generating a biological condition estimation model performed in an operation support server according to the fourth embodiment of the present invention.

[0045] Fig. 22 This is a flowchart showing an example of biological data calculation processing performed in the operation support server according to the third embodiment of the present invention.

[0046] Fig.23 This is a flowchart showing an example of a calculation process of biological data using heartbeat data performed in an operation support server according to the third embodiment of the present invention.

[0047] Fig.24This is a flowchart showing an example of risk prediction processing performed in the operation support server according to the third embodiment of the present invention.

[0048] Fig.25 This is a flowchart showing an example of a warning presentation process performed in the operation support server according to the third embodiment of the present invention.

[0049] Fig.26A This is a diagram showing an example of a warning presentation screen according to the third embodiment of the present invention.

[0050] Fig.26B This is a diagram showing another example of the warning presentation screen according to the third embodiment of the present invention.

[0051] Fig.27A This is a diagram showing an example of a data structure representing biological data according to the third embodiment of the present invention.

[0052] Fig.27B This is a diagram showing an example of a data structure representing business / environment data according to Embodiment 3 of the present invention.

[0053] Fig.27C This is a diagram showing an example of a data structure of attribute information data according to Embodiment 3 of the present invention.

[0054] Fig.27D This is a diagram showing an example of a data structure of learning information data according to the third embodiment of the present invention.

[0055] Fig.27E This is a diagram showing an example of a data structure showing risk occurrence data according to the third embodiment of the present invention.

[0056] Fig.27F This is a diagram showing an example of a data structure of accident risk prediction data according to the third embodiment of the present invention.

[0057] Figure 27G This is a diagram showing an example of a data structure showing personal severity level data according to the third embodiment of the present invention.

[0058] Fig.27H This is a diagram showing an example of the data structure of a connection table according to the third embodiment of the present invention.

[0059] Fig.27I This is a diagram showing an example of the data structure of alarm definition data according to the third embodiment of the present invention.

[0060] Fig.27J This is a diagram showing an example of a data structure of countermeasure data according to the third embodiment of the present invention.

[0061] Fig.28This is a diagram showing an overview of processing performed in the operation support system according to the fourth embodiment of the present invention.

[0062] Fig.29 This is a diagram showing an overview of processing performed in the operation support system according to the fifth embodiment of the present invention.

[0063] Fig.30 This is a diagram showing an example of a warning target list screen provided to a driver manager according to a fifth embodiment of the present invention. DETAILED DESCRIPTION

[0064] Hereinafter, embodiments of the present invention will be described based on the drawings.

[0065] Example 1

[0066] Figure 1 This is a block diagram showing an example of the configuration of the operation support system according to the first embodiment of the present invention.

[0067] The operation assistance system of this embodiment includes an operation assistance server 1 that assists the operation of one or more vehicles 7 via a network 13. The vehicle 7 includes an on-board sensor 8 that detects the driving state, a biometric sensor 12 that detects the biological data of the driver, a driver ID reading device 11 that identifies the driver, a driving data collection device 10 that collects the detected sensor data and the driver ID and sends them to the operation assistance server 1, and a prediction result notification device 9 that receives a notification corresponding to the risk of a traffic accident of the driver (hereinafter referred to as accident risk) from the operation assistance server 1 and notifies the driver.

[0068] In the example shown in the figure, the driving data collection device 10 and the prediction result notification device 9 are independent devices, but they can also be configured by a single portable terminal. In this case, it functions as a driving data collection unit and a prediction result notification unit.

[0069] The vehicle-mounted sensor 8 may include a GNSS (Global Navigation Satellite System) 81 for detecting position information of the vehicle 7 , an acceleration sensor 82 for detecting the movement and speed of the vehicle 7 , and a camera 83 for detecting the driving environment using images.

[0070] The vehicle-mounted sensor 8 is not limited to the above, and a distance sensor for detecting an object or distance around the vehicle 7 and a steering angle sensor for detecting a driving operation can be used. In addition, the acceleration sensor 82 is preferably a three-axis acceleration sensor.

[0071] The biosensor 12 includes a heartbeat sensor 121 for detecting heartbeat data and an acceleration sensor 122 for detecting the driver's movement. The biosensor 12 is not limited to the above, and a sensor for detecting sweating, body temperature, blinking, eye movement, or brain waves can be used. As the biosensor 12, a wearable device that can be worn by the driver, a sensor device attached to the inside of the vehicle 7 such as a steering wheel, a seat, and a seat belt, and an image recognition system that captures the driver's expression and movement and analyzes the image can be used.

[0072] The driver ID reading device 11 reads a card recording the driver's identifier. The driving data collecting device 10 collects data from the vehicle-mounted sensor 8 and the biosensor 12 at a predetermined period, and transmits the data to the driving support server 1 via the network 13 .

[0073] The operation 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 functional unit of a biological indicator calculation unit 31, an accident risk definition generation unit 32, an accident risk prediction model generation unit 33, a prediction model selection unit 34, an accident risk prediction unit 35, an accident risk notification unit 36, and a data collection unit 37 as a program. Each program is executed by the processor 2. In addition, the details of each functional unit will be described later.

[0074] The processor 2 operates as a functional unit that provides a predetermined function by executing processing according to the program of each functional unit. For example, the processor 2 performs the function of the biological indicator calculation unit 31 by executing the biological indicator calculation program. The same is true for other programs. Furthermore, the processor 2 also operates as a functional unit that provides the functions of each of the multiple processes executed by each program. Computers and computer systems are devices and systems that include these functional units.

[0075] The storage device 4 stores data used by the above-mentioned functional units. The storage device 4 stores vehicle-mounted sensor data 40, biosensor data 41, business data 42, accident risk prediction model 43, accident risk prediction data 44, hazard occurrence data 45, bio-indicator data 46, environmental data 47, accident risk definition model 48, alarm definition data 49, and accident risk estimation data 50. The details of each data will be described later.

[0076] The input / output device 5 includes an input device such as a mouse, a keyboard, or a touch panel, and an output device such as a display. The communication device 6 communicates with the vehicle 7 via the network 13 .

[0077] <Overview of Operation Support System>

[0078] Figure 21 is a diagram showing an overview of the processing performed in the operation support system. First, the accident risk definition generation unit 32 takes the pre-set danger occurrence data 45 and the vehicle-mounted sensor data 40-m1 collected in the past as input, and generates an accident risk definition model 48 (S1) that estimates the probability of danger occurrence as the accident risk. In addition, the accident risk definition model 48 takes the vehicle-mounted sensor data 40-m2 indicating the driving state of the vehicle 7 as input to estimate the probability of danger occurrence.

[0079] The danger occurrence data 45 is data in which a manager or the like determines an event and a phenomenon leading to the event based on the time-series vehicle sensor data 40 - m1 indicating the driving state of the vehicle 7 , and sets the occurrence date and time and the importance of the event.

[0080] For example, when the vehicle-mounted sensor data 40 - m1 is sensor data of the acceleration sensor 82 , events such as sudden braking and sharp turns or driving conditions leading to the events are detected, and the importance of the danger occurrence is set by the manager, etc., to generate danger occurrence data 45 .

[0081] In addition, an example is shown in which the danger increases as the importance value increases. In addition, the driving state leading to the event indicates a state in which the driver is safe (frightened). In the following description, the event and the driving state leading to the event are regarded as the occurrence of danger.

[0082] In generating the danger occurrence data 45, the operation support server 1 can detect driving conditions where the front and rear accelerations exceed a predetermined threshold and driving conditions where the yaw angular velocity exceeds a predetermined threshold (danger occurrence), and the administrator or the like can set the importance of the detected danger occurrence phenomenon.

[0083] Figure 2 The time series of the vehicle-mounted sensor data 40 - m1 shown is the same time series as the time series of the danger occurrence data 45 .

[0084] The accident risk definition generating unit 32 receives the vehicle-mounted sensor data 40 - m1 and the danger occurrence data 45 as inputs, and generates a machine learning model for estimating the accident risk from the driving state as an accident risk definition model 48 .

[0085] As a machine learning model, well-known or publicly known methods such as neural network and logistic regression can be used, so they are not described in detail in this embodiment. In addition, the accident risk is the probability (percentage) of causing an event (or danger). In addition, in this embodiment, the phenomenon that leads to a traffic accident is regarded as an event.

[0086] Next, the accident risk definition generation unit 32 inputs the vehicle sensor data 40-m2 collected in the past in a different time series from the above-mentioned vehicle sensor data 40-m1 into the generated accident risk definition model 48 to output the probability of accident risk occurrence, thereby generating accident risk estimation data 50 (S2).

[0087] The accident risk estimation data 50 is data in which the probability of an event occurring (or the probability of an accident occurring) is generated for the onboard sensor data 40 - m2 in the same time series as the onboard sensor data 40 - m2 .

[0088] Next, the biometrics calculation unit 31 calculates the driver's biometrics data 46-m based on the past biometrics data 41 (the driver's biometrics data) corresponding to when the vehicle-mounted sensor data 40-m2 was acquired. As the biometrics data 46-m, for example, the power spectrum density calculated from the driver's heartbeat data (described later) and the autonomic nerve index obtained based on the NN interval (the interval between R waves) calculated by time domain analysis (described later) can be used.

[0089] In addition, the vehicle-mounted sensor data 40 - m1 , 40 - m2 and the biometric indicator data 46 - m collected in the past may be stored in a predetermined area of ​​the storage device 4 .

[0090] Next, the accident risk prediction model generation unit 33 takes the accident risk estimation data 50 and the biological indicator data 46-m as input, and generates a machine learning model that outputs the accident risk (probability) after a specified time based on the biological indicator data 46 of the moving vehicle 7 as the accident risk prediction model 43 (S3).

[0091] In addition, regarding the accident risk prediction model 43 , a plurality of models may be generated according to the type of the vehicle-mounted sensor data 40 used and the environment in which the prediction is performed, and the prediction model selection unit 34 may select the model to be used.

[0092] In addition, for the accident risk prediction model 43, multiple types can also be generated according to the types corresponding to the different driving characteristics of the vehicle 7, the driver's driving experience, pre-operation biological data, etc. For example, as the difference in the driving characteristics of the vehicle 7, it is preferable to generate different models for vehicle types such as trucks and trailers that have different steering wheel operations when turning and reversing. In addition, as the driver's driving experience, the model can be generated according to the correctness of the driving action, the sense of danger, the number of years of driving, the vehicles driven (large, special, etc.), the driving route experience, etc. In addition, as the driver's pre-operation biological data, the model can be generated according to the body temperature, blood pressure, blood oxygen concentration, sleep time the previous day, etc.

[0093] Next, the operation support server 1 uses the generated accident risk prediction model 43 to predict the accident risk of the vehicle 7 in actual driving (S4). In the operation support server 1, when there are multiple accident risk prediction models 43, the prediction model selection unit 34 selects an accident risk prediction model 43 corresponding to the vehicle sensor data 40 used and the environmental data 47 of the vehicle 7.

[0094] The accident risk prediction unit 35 inputs the heartbeat data of the biosensor 12 received by the data collection unit 37 via the vehicle 7 to the biometric calculation unit 31 so that it calculates the biometric data 46. Then, the accident risk prediction unit 35 inputs the calculated biometric data 46 to the accident risk prediction model 43 so that it predicts the future accident risk from now until a predetermined time Δt later. In addition, when the accident risk prediction model 43 outputs interpretation data (described later) indicating the state of the driver, the accident risk prediction unit 35 can add interpretation data to the predicted value of the accident risk.

[0095] Then, when the accident risk after the specified time Δt is above the specified threshold Th, the accident risk notification unit 36 ​​generates an alarm and a message according to the value of the accident risk, the interpretation data and the driving status or environmental data, and sends the alarm and the message to the vehicle 7 for which the accident risk is predicted to increase (S5).

[0096] When the vehicle 7 receives an alarm from the operation support server 1, the prediction result notification device 9 outputs an alarm to convey an alarm or message to the driver. In addition, when the accident risk is predicted to increase after a predetermined time Δt, the accident risk notification unit 36 ​​can generate an image including the location information of the vehicle 7, the accident risk of the driver, the traffic volume, etc., for the administrator of the operation support server 1, and display it on the display of the input / output unit device 5.

[0097] The operation support server 1 predicts that the accident risk will reach or exceed the threshold value Th after a predetermined time Δt, and notifies the prediction result notification device 9 of the vehicle 7 and the display of the input / output device 5, thereby alerting the driver and the manager with an alarm and a message before the accident risk increases. Thus, the operation support server 1 can reduce the actual accident risk (51) and promote the safe operation of each vehicle 7.

[0098] In addition, the predetermined time Δt for predicting the future accident risk is preferably set to a value such as 30 minutes or 1 hour, so that the driver can be recommended to take a rest or given advice before the time when the accident risk increases.

[0099] <Data>

[0100] Next, each data used in the operation support system will be described.

[0101] Figure 8 : is a figure which shows an example of the biological indicator data 46. The biological indicator data 46 is a table which stores the data calculated by the biological indicator calculation part 31 for every predetermined period based on the heartbeat data of the biosensor data 41.

[0102] The biometric data 46 includes, in one record, a user ID 460, a date and time 461, an autonomic nerve total power 462, an autonomic nerve LF / HF 463, an autonomic nerve NN50 465, and an average RRI 464. The user ID 460 stores the driver's identifier. The date and time 461 stores the date and time when the biosensor data 41 (heartbeat data) is acquired.

[0103] In the autonomic nerve total power 462, as described later, the total value of the low frequency (LF: Low Frequency) component and the high frequency (HF: High Frequency) component of the power spectrum density of the R wave interval (RRI) of the heartbeat data is saved as a value representing the balance of the autonomic nerves (sympathetic nerves and parasympathetic nerves).

[0104] In the autonomic nerve LF / HF 463, the ratio of the low frequency (LF) component to the high frequency (HF) component of the power spectrum density is stored. The low frequency component represents the activity index of the sympathetic nerve, and the high frequency component represents the activity index of the parasympathetic nerve. In the average RRI 464, the average value of the R wave interval (RRI) within a preset period is stored.

[0105] In the autonomic nerve NN50 465, as described later, the total number of differences of 50 msec or more in the difference sequence ΔRRI(t) of the heartbeat variation time series data RRI is stored as a value indicating the magnitude of the parasympathetic nerve activity.

[0106] The illustrated example shows an example in which the autonomic nerve total power 462, the autonomic nerve LF / HF 463, and the autonomic nerve NN50 465 are calculated at one-minute intervals, but the present invention is not limited thereto, and the bio-indicator data 46 may be calculated at appropriately set time intervals.

[0107] Fig. 9 4 is a diagram showing an example of the business data 42. The business data 42 is a table for storing and managing information on the vehicle 7 driven by the driver and the business, and is set by a person in charge or a manager of the operation support system.

[0108] The service data 42 includes, in one record, a user ID 420, a service start date and time 421, a service end date and time 422, a vehicle ID 423 storing an identifier of the vehicle 7, and a service type 424. The service type 424 stores the content of the driver's service.

[0109] Fig.10 4 is a diagram showing an example of environmental data 47. Environmental data 47 is a table storing information obtained based on vehicle-mounted sensor data 40 of vehicle 7. Environmental data 47 stores a vehicle ID 470 storing an identifier of vehicle 7, a date and time 471 when environmental data 47 is generated, a congestion condition 472 of a road on which vehicle 7 is traveling, weather 473, and temperature 474.

[0110] The congestion condition 472 may be determined based on the image of the camera 83 of the vehicle 7 , or may be obtained from a traffic information service on the network 13 based on the position information calculated by the GNSS 81 .

[0111] Fig.11 2 is a diagram showing an example of the vehicle-mounted sensor data 40. The vehicle-mounted sensor data 40 is a table storing information collected by the data collection unit 37 from the vehicle-mounted sensor 8 of the vehicle 7 at a predetermined period.

[0112] The vehicle sensor data 40 includes, in one record, a vehicle ID 400 storing an identifier of the vehicle 7 , a date and time 401 at which data is acquired from the vehicle-mounted sensor 8 , a position 402 storing position information detected by the GNSS 81 , a speed 403 of the vehicle 7 , and an acceleration 404 detected by the acceleration sensor 82 .

[0113] Fig.12 45 is a diagram showing an example of the danger occurrence data 45. Figure 2 As shown, it is a table generated based on the vehicle sensor data 40-m1.

[0114] The danger occurrence data 45 includes, in one record, a vehicle ID 450 storing an identifier of the vehicle 7, a date and time 451 of danger occurrence, a danger type 452 storing the type of danger that occurred, and an importance level 453 of the danger that occurred. The danger type 452 and the importance level 453 store values ​​determined by a manager or the like.

[0115] Regarding the danger occurrence data 45 , a manager or the like determines an event or a state leading to the event based on the driving state of the vehicle 7 indicated by the vehicle-mounted sensor data 40 - m1 and sets a danger type 452 and an importance level 453 .

[0116] Fig.13 2 is a diagram showing an example of the accident risk estimation data 50. The accident risk estimation data 50 is a table in which the accident risk definition generating unit 32 inputs the vehicle-mounted sensor data 40-m2 to the accident risk definition model 48 to estimate the accident risk.

[0117] The accident risk estimation data 50 includes, in one record, a user ID 500 storing a driver's identifier, a vehicle ID 501 storing an identifier of the vehicle 7 , a date and time 502 when the accident risk occurs, and an accident risk 503 storing a probability (percentage) of reaching an event.

[0118] The accident risk 503 is the result of calculating the probability of an event occurring by giving the speed, acceleration, and position information of the vehicle 7 to the accident risk definition model 48. In the example shown in the figure, the accident risk 503 is calculated at one-minute intervals, but the present invention is not limited to this.

[0119] Fig.14 2 is a diagram showing an example of the alarm definition data 49. The alarm definition data 49 is a table in which alarms or messages notified by the accident risk notification unit 36 ​​are preset.

[0120] The alarm definition data 49 includes an alarm ID 490 , an accident risk condition 491 , a time condition 492 , a congestion condition 493 , a weather condition 494 , a priority 495 , and a comment 496 in one record.

[0121] The alarm ID 490 stores an identifier for identifying an alarm. The accident risk condition 491 stores a condition for selecting the accident risk of the alarm in the record. For example, in the record with the alarm ID 490="1", a condition is set to select when the accident risk value is 70% or less.

[0122] The time condition 492 sets the time period for selecting the alarm of the record. The congestion condition 493 sets the congestion status (presence or absence) for selecting the alarm of the record. The weather condition 494 sets the weather for selecting the alarm of the record.

[0123] The priority 495 sets the priority of the selected record in advance when extracting a plurality of records satisfying the accident risk condition 491. The comment 496 stores the content of the notified alarm (or message).

[0124] Fig.15 : is a diagram showing an example of the accident risk prediction data 44. The accident risk prediction data 44 is a table storing the prediction results of the accident risk calculated by the accident risk prediction model 43 to which the biosensor data 41 of the driver of the traveling vehicle 7 is input.

[0125] The accident risk prediction data 44 includes a user ID 440 , a vehicle ID 441 , a date and time 442 when the vehicle-mounted sensor data 40 is acquired, a predicted date and time 443 , and an accident risk 444 in one record.

[0126] The predicted date and time after the predetermined time Δt is stored in the predicted date and time 443. The predicted value (percentage) of the accident risk after the predetermined time Δt is stored in the accident risk 444.

[0127] <Details of the processing>

[0128] Next, the processing performed in the operation support system will be described.

[0129] Figure 3 2 is a flowchart showing an example of a process for generating the accident risk prediction model 43 performed in the operation support server 1. This process is performed before receiving the vehicle sensor 8 and the biosensor 12 from the vehicle 7, and the accident risk prediction model 43 is generated in advance.

[0130] The accident risk definition generation unit 32 combines the pre-generated danger occurrence data 45 and the vehicle-mounted sensor data 40 ( Figure 2 40-m1) is used as input to generate an accident risk definition model 48 (S11) through machine learning.

[0131] Then, the accident risk definition generation unit 32 inputs the vehicle-mounted sensor data 40 ( Figure 2 40-m2), generate accident risk estimation data 50 (S12).

[0132] Next, the accident risk prediction model generation unit 33 obtains the vehicle-mounted sensor data 40 ( Figure 2 According to the accident risk estimation data 50 and the biological indicator data 46-m, an accident risk prediction model 43 (S13) is generated through machine learning.

[0133] Furthermore, when performing machine learning, the accident risk prediction model generation unit 33 acquires information about the driver from the business data 42 and acquires the driving environment of the vehicle 7 from the environment data 47 to generate the accident risk prediction model 43 .

[0134] Through the above-described processing, after the accident risk definition model 48 is generated and the accident risk estimation data 50 is generated, the accident risk prediction model 43 is generated based on the biological indicator data 46 and the accident risk estimation data 50 collected in the past.

[0135] Figure 4 4 is a flowchart showing an example of a process for generating the biometric indicator data 46 performed in the operation support server 1. This process is executed every time data is received from the vehicle 7. The data collection unit 37 of the operation support server 1 receives data from the vehicle sensor 8 and the biometric sensor 12 from the driving data collection device 10 of the vehicle 7 (S21).

[0136] The data collection unit 37 stores the heartbeat data and acceleration data detected by the heartbeat sensor 121 and the acceleration sensor 122 in the biosensor data 41 ( S22 ). The data of the vehicle-mounted sensor 8 is stored in the vehicle-mounted sensor data 40 by the data collection unit 37 .

[0137] Next, the biological indicator calculation unit 31 reads the biosensor data 41 and calculates the biological indicator as described later (S23). The biological indicator calculation unit 31 stores the calculated biological indicator in the biological indicator data 46 (S24).

[0138] Fig. 7A 1 is a flowchart showing an example of a biological indicator calculation process performed by the biological indicator calculation unit 31. Figure 4 The process is performed in step S23.

[0139] The biometrics calculation unit 31 acquires heartbeat data from the biosensor data 41 (S51). The acquired heartbeat data is data within a predetermined period from the latest heartbeat data. Next, the biometrics calculation unit 31 calculates RRI (RR Interval) from the interval of R waves included in the heartbeat data (S52).

[0140] Figure 7B This is a diagram showing an example of heartbeat data detected by the biosensor 12. The biometric calculation unit 31 detects the time interval between the peak of the R wave in the diagram and the peak of the next R wave as RRI, calculates a specified number (or a specified period) of RRIs as heartbeat variation timing data, and further calculates the average value of the RRIs as average RRI 464.

[0141] Next, the biological indicator calculation unit 31 calculates fluctuations based on the calculated heartbeat variation time series data. Figure 7C This is a graph showing an example of fluctuation (heartbeat variation) of heartbeat data calculated by the biological indicator calculation unit 31. The RRI of heartbeat data is not fixed but varies due to the activity of the autonomic nerves and the like.

[0142] The bio-indicator calculation unit 31 performs spectrum analysis based on the time-series heartbeat variation data (S53), and then calculates the power spectrum density (PSD: Power Spectral density) (S54). The power spectrum density can be calculated by a well-known method.

[0143] Next, the biological indicator calculation unit 31 calculates the intensity LF of the low-frequency component and the intensity HF of the high-frequency component of the power spectrum density. Fig.7D This is a graph showing an example of the power spectrum density of heartbeat fluctuations in the frequency domain.

[0144] The biological indicator calculation unit 31 is as follows Fig.7D As shown, the value obtained by adding up (LF+HF) the intensity (integrated value) LF of the low-frequency component interval (0.05Hz~0.15Hz) of the power spectrum and the intensity (integrated value) HF of the high-frequency component interval (0.15Hz~0.40Hz) is calculated as the total autonomic nerve power 462.

[0145] Furthermore, the bio-index calculation unit 31 calculates the ratio (LF / HF) of the intensity LF of the low-frequency component to the intensity HF of the high-frequency component of the power spectrum as the autonomic nerve LF / HF 463 .

[0146] Through the above processing, the time-series heartbeat variation data is calculated based on the heartbeat data of the biosensor data 41, and the strength of the low-frequency component and the high-frequency component (LF, HF) is calculated to obtain the autonomic nerve total power 462 and the autonomic nerve LF / HF463.

[0147] Here, the high-frequency component appears in the heartbeat fluctuation when the parasympathetic nerves are activated (stressed), and the low-frequency component appears in the heartbeat fluctuation when the sympathetic nerves are activated (stressed) and when the parasympathetic nerves are activated (stressed).

[0148] It is known that the activation of the sympathetic nerves indicates a stress state and the activation of the parasympathetic nerves indicates a relaxation state, so it is possible to determine whether the driver is in a stress state or a relaxation state based on the intensity of the low frequency component LF and the intensity of the high frequency component HF.

[0149] Thereafter, the biological index calculation unit 31 calculates a time domain index obtained by time domain analysis of the heartbeat variation time series data RRI (S55). Fig. 7E This is a graph showing an example of autonomic nerve NNXX, which is one of the time domain indicators, calculated based on the difference series ΔRRI(t) of the heartbeat variation time series data RRI.

[0150] The biological indicator calculation unit 31 is as follows Fig. 7E As shown, the difference sequence of adjacent RRIs is calculated for the heartbeat variation timing data RRI, i.e., ΔRRI(t), and the total number of differences constituting ΔRRI(t) that are greater than XXmsec, i.e., the autonomic nerve NNXX, is calculated based on ΔRRI(t). Typically, Fig. 7E As shown in FIG. 1 , the total number of differences constituting ΔRRI(t) of more than 50 msec (= the total number of differences existing in the range of the grid dots), i.e., the autonomic nerve NN50 465, is calculated when XX=50.

[0151] Here, it is known that the autonomic nerve NN50 465 is similar to the intensity HF of the high-frequency component and is an indicator of the magnitude of the parasympathetic nerve activity, so the degree of the driver's relaxation state can be determined based on the autonomic nerve NN50 465.

[0152] In addition, NNXX is not limited to the above definition. For example, instead of the total number of differences constituting ΔRRI(t) being greater than 50 msec, the total number of differences constituting ΔRRI(t) being greater than 50 msec in absolute value (the total number of differences existing in the grid dot range and the diagonal dot range), i.e., absNN50, may be calculated, or the autonomic nerve NN50 465 may be normalized by dividing the total number of differences included in ΔRRI(t) (= the total number of differences existing in the grid dot range ÷ the total number of differences existing in the dot dot range), i.e., pNN50. In addition, instead of the condition that the differences constituting ΔRRI(t) are greater than 50 msec, the autonomic nerve NN50 465 may be calculated based on the condition that the differences constituting ΔRRI(t) exceed 50 msec. Furthermore, instead of XX=50, NNXX may be calculated with XX=40, 60, 100, etc.

[0153] Then, the bio-indicator calculation unit 31 stores the above-calculated autonomic nerve total power 462, autonomic nerve LF / HF 463, autonomic nerve NN50 465, and average RRI 464 in the bio-indicator data 46 (S56 and Figure 4 S24).

[0154] Thus, the accident risk prediction model 43 can output interpretation data about the driver's state based on bio-indicators such as the autonomic nerve LF / HF 463 obtained based on the intensity LF and HF, and bio-indicators such as the autonomic nerve NN50 465 obtained based on time domain analysis. The interpretation data may include the driver's stress state and relaxation state.

[0155] Figure 5 4 is a flowchart showing an example of prediction processing performed in the operation support server 1. This processing is executed at a predetermined timing, such as when the biological indicator data 46 is updated. The accident risk prediction unit 35 acquires biological indicator data 46 for a predetermined period from the latest data (S31).

[0156] Next, the prediction model selection unit 34 determines the user ID 420 and vehicle ID 423 of the analysis target based on the business data 42, and selects the accident risk prediction model 43 to be used using the environment data 47 (S32). The user ID 420 of the analysis target can be, for example, the user ID 460 of the unprocessed biometric data 46.

[0157] Furthermore, when there are a plurality of accident risk prediction models 43 , the prediction model selection unit 34 selects a model to be used based on the weather 473 , the date and time 471 , the congestion status 472 , and the like of the environmental data 47 .

[0158] Next, the accident risk prediction unit 35 inputs the biological indicator data 46 of the user ID of the analysis target to the selected accident risk prediction model 43 to calculate the accident risk (S33). The calculated accident risk is stored in the accident risk 444 of the accident risk prediction data 44.

[0159] in addition, Fig.15 , an example of saving the accident risk 444 is shown, but when the accident risk prediction model 43 outputs interpretation data about the driver's state (stressed state and relaxed state) based on the autonomic nerve LF / HF 463, etc., the interpretation data can be saved in the bio-indicator data 46.

[0160] In addition, the above shows an example of inputting the biological indicator data 46 to the accident risk prediction model 43, but the present invention is not limited to this, and the environmental data 47 and the vehicle sensor data 40 may be added to the input of the accident risk prediction model 43. For example, when the accident risk prediction model 43 outputs interpretation data, by using the congestion conditions and the like as the input of the accident risk prediction model 43, the prediction accuracy can be improved.

[0161] Through the above-described processing, at a predetermined timing such as after the biological indicator data 46 is updated, the accident risk prediction model 43 is used to predict the accident risk (and interpretation data) after a predetermined time Δt.

[0162] Figure 6 is a flowchart showing an example of an alarm generation process performed in an operation auxiliary server. Figure 5 Executed after the prediction processing is completed.

[0163] The accident risk notification unit 36 ​​acquires the warning definition data 49 (S41). The accident risk notification unit 36 ​​refers to the accident risk prediction data 44 to search for a driver whose accident risk 444 after a predetermined time Δt is greater than a preset threshold value Th (reference value in the figure) (S42).

[0164] If there is a driver whose accident risk 444 value is greater than or equal to the threshold value Th after the predetermined time Δt, the accident risk notification unit 36 ​​proceeds to step S44 and acquires the vehicle ID 423 of the business data 42 as the target of sending the alarm based on the identifier of the driver greater than or equal to the threshold value Th ( S44 ).

[0165] Then, the accident risk notification unit 36 ​​obtains the environmental data 47 according to the vehicle ID 470 of the vehicle 7, selects the alarm ID 490 that satisfies the conditions of the environmental data 47 (time condition 492, congestion condition 493, weather condition 494) and the accident risk condition 491 from the alarm definition data 49, and outputs the accident risk alarm (omitted in the figure) specified by the alarm ID 490 and the annotation 496 of the selected record (S45).

[0166] Furthermore, the accident risk notification unit 36 ​​acquires the position 402 of the vehicle 7 from the vehicle-mounted sensor data 40 , and generates an alarm by adding the position information of the vehicle 7 .

[0167] Through the above processing, an alarm is sent to the vehicle 7 of the driver whose accident risk is predicted to be greater than the threshold value Th after the predetermined time Δt. In the vehicle 7 that has received the alarm, the prediction result notification device 9 notifies the driver of the alarm. In addition, in the operation support server 1, the accident risk notification unit 36 ​​displays the vehicle 7 that has sent the alarm on the display of the input / output device 5.

[0168] in addition, Fig.14 The alarm definition data 49 of FIG. 4 shows an example of determining the annotation and the alarm ID 490 according to the accident risk condition 491, the environmental condition, and the priority 495, but the present invention is not limited thereto. For example, when the accident risk prediction model 43 outputs interpretation data about the driver's state, a condition for determining the alarm ID according to the interpretation data may be set.

[0169] In one example, when the accident risk 444 is the same and the interpretation data is in a pressurized state or a relaxed state, the alarm ID 490 when the accident risk increases in the pressurized state and the alarm ID 490 when the accident risk increases in the relaxed state are set as different alarms. Thus, by adding interpretation data to the value of the accident risk 444, a high-precision prediction can be achieved.

[0170] Fig.16A 1 is a diagram showing an example of a prediction result display screen 600 outputted by the accident risk notification unit 36 ​​on the display of the input / output device 5. The prediction result display screen 600 outputted by the accident risk notification unit 36 ​​displays an icon 602 of the vehicle 7 to be managed on a map, and an alarm icon 601 is added to the icon 602 for issuing an alarm.

[0171] Furthermore, a driving condition 603 of a driver whose accident risk is increasing is displayed near the warning icon 601, so that the workload and the number of breaks of the driver can be notified to a manager or the like.

[0172] In addition, the manager can check the accident risks of a plurality of drivers at a glance on the map, and can select a driver with a high accident risk and give remote instructions.

[0173] In addition, in the example shown in the figure, the position of the vehicle 7 and the driver's condition are displayed on a map, but the present invention is not limited to this, and notification may be performed using text information such as a list.

[0174] In addition, when the accident risk notification unit 36 ​​obtains interpretation data based on the autonomic nerve LF / HF463, etc. for a driver whose accident risk is predicted to increase, it displays the interpretation data in the driving status 603, thereby enabling the manager and the like to know the driver's stress state and relaxation state.

[0175] Fig. 16B 1 is a diagram showing an example of a prediction result display screen 650 output by the prediction result notification device 9. The prediction result notification device 9 includes a display (not shown), and displays the prediction result display screen 650 when receiving an alarm from the operation support server 1.

[0176] The prediction result display screen 650 includes an area 651 for displaying an accident risk warning and an area 652 for displaying a comment. The driver can recognize the accident risk by viewing the prediction result display screen 650 of the prediction result notification device 9.

[0177] As described above, the operation support system of this embodiment inputs the biometric data 46 calculated by the operation support server 1 based on the heartbeat data of the biosensor 12 into the accident risk prediction model 43 to predict the accident risk 444 after a predetermined time Δt. Thus, it is possible to predict in advance the state in which a traffic accident is likely to occur according to the state of the autonomic nerves of the driver of the vehicle 7, and provide feedback such as recommendation for rest to the driver.

[0178] In the first embodiment, the present invention is applied to the operation support system for managing the vehicle 7, but the present invention is not limited thereto. For example, the present invention can be applied to a mobile object requiring a driver or operator, such as a railway vehicle, a ship, or an aircraft, instead of the vehicle 7.

[0179] In addition, in the above-mentioned embodiment 1, an example of generating the accident risk definition model 48 based on the vehicle sensor data 40-m1 and the danger occurrence data 45 collected in the past is shown, but the invention is not limited to this. For example, it is also possible to identify the position determined by the driver as dangerous from the new vehicle sensor data 40 after the business is completed and generate new danger occurrence data 45, add it as learning data, and learn the accident risk definition model 48 again. In this case, the accident risk prediction model 43 is also generated again.

[0180] Example 2

[0181] Fig.17This is a block diagram showing an example of the structure of the operation support system of Example 2. In this embodiment, the biometric calculation unit 31, the accident risk prediction unit 35 and the biometric data 46 are configured in the driving data collection device 10 of the vehicle 7, and the accident risk is predicted using the accident risk prediction model 43A selected by the prediction model selection unit 34 of the operation support server 1.

[0182] The driving data collection device 10 of the vehicle 7 calculates the biometric data and predicts the accident risk, and sends the accident risk to the operation support server 1. In the operation support server 1, the accident risk notification unit 36 ​​outputs an alarm to the vehicle 7 in the same manner as in the first embodiment when the accident risk after a predetermined time Δt is greater than the threshold value Th.

[0183] In this embodiment, by calculating the biological index and predicting the accident risk on the vehicle 7 side, the load of the operation support server 1 can be reduced, and the number of vehicles 7 managed by the operation support server 1 can be increased.

[0184] Example 3

[0185] Next, the third embodiment of the present invention is described. The above-mentioned patent document 2 provides a means for identifying deviations from the normal internal state of the driver by comparing the driving environment risk included in the external environment identified by sensing the external environment of the mobile body with the risk perception state of the driver calculated as the internal state based on the driving operation of the driver (driver).

[0186] However, only by performing state estimation based on the correlation between the driver's driving operation and the external environment and identifying deviations from the driver's usual internal state (narrowly defined dangerous state), it is not possible to detect states that have not yet appeared in the driving operation but require attention (broadly defined dangerous state). Therefore, there is a problem that it is difficult to take measures to avoid broadly defined dangerous states in advance.

[0187] This embodiment is made in view of the above-mentioned problems, and aims to detect a broadly defined dangerous state taking into account individual differences of each driver using biometric data in order to detect and avoid the broadly defined dangerous state in advance, and to determine whether to issue an alarm based on the result.

[0188] <System Structure>

[0189] Fig.18 1 is a block diagram showing an example of a main configuration of an operation support system according to a third embodiment of the present invention. The operation support system of this embodiment includes an operation support server 1 that supports the operation of one or more vehicles 7 via a network 14 .

[0190] The vehicle 7 includes an on-board sensor 8 for detecting the driving status, a biometric sensor 12 for detecting the driver's biometric sensor data 163, a driver ID reading device 11 for identifying the driver, a driving data collection device 10 for collecting the detected on-board and biometric sensor data 163 and the driver ID and sending them to the operation assistance server 1, a prediction result notification device 9 for receiving a warning corresponding to the driver's risk of a traffic accident (hereinafter referred to as accident risk) from the operation assistance server 1 and prompting the driver, and a business status input device 130 for receiving input of the measurement status of the biometric indicator data 164.

[0191] In addition, in the example shown in the figure, the driving data collection device 10, the prediction result notification device 9, the business status input device 130 and the driver ID reading device 11 are shown as independent devices, but they can also be configured by a single portable terminal. In this case, the portable terminal plays the role of a driving data collection unit, a prediction result notification unit, a business status input unit and a driver ID reading unit.

[0192] The vehicle-mounted sensor 8 may include a GNSS (Global Navigation Satellite System) 81 for detecting position information of the vehicle, an acceleration sensor 82 for detecting the movement and speed of the vehicle, and a camera 83 for detecting the driving environment using images.

[0193] The vehicle-mounted sensor 8 is not limited to the above, and may include a distance sensor for detecting objects and / or distances around the vehicle, a steering angle sensor for detecting driving operations, an angular velocity sensor for detecting steering operations of the vehicle, etc. In addition, the acceleration sensor 82 is preferably a three-axis acceleration sensor.

[0194] The biosensor 12 includes a heartbeat sensor 121 for detecting heartbeat data and an acceleration sensor 122 for detecting the driver's motion. The heartbeat sensor 121 can use a sensor that detects heartbeat based on electrocardiogram, pulse, heart sound, or the like.

[0195] The biosensor 12 is not limited to the above, and a sensor that detects sweat, body temperature, blinking, eye movement, myoelectricity, or brain waves, etc. can be used. As the biosensor 12, wearable devices that can be worn by the driver, sensor devices attached to the interior of the vehicle such as the steering wheel, seat, and seat belt, and an image recognition system that captures the driver's expression and movement and analyzes the image, etc. can be used.

[0196] The driver ID reading device 11 reads a card recording the driver's identifier. The driving data collecting device 10 collects data from the vehicle-mounted sensor 8 and the biosensor 12 at a predetermined period, and transmits the data to the driving support server 1 via the network 14 .

[0197] In addition, in the example shown in the figure, the driver ID reading device 11 is configured as a device that reads a card recording the driver's identifier, but a different structure may be adopted. For example, when the driver ID reading device 11 is configured by a portable terminal and the portable terminal functions as a driver ID reading unit, the driver ID can be read by having the driver input the driver's identifier himself, or by recognizing the driver using a known face recognition technology of a camera provided in the portable terminal.

[0198] The work status input device 130 receives the work status of the driver, such as driving, resting, or napping, and transmits it to the operation support server 1 .

[0199] The operation 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 functional unit of a biological indicator calculation unit 131, an accident risk definition generation unit 132, an accident risk estimation unit 133, a biological state estimation model generation unit 134, an accident risk prediction model generation unit 135, a model selection unit 136, a risk prediction unit 137, a warning presentation unit 138, and a data collection unit 139 as a program. Each program is executed by the processor 2. In addition, the details of each functional unit will be described later.

[0200] The processor 2 operates as a functional unit that provides a predetermined function by executing processing according to the program of each functional unit. For example, the processor 2 performs the function of the biological indicator calculation unit 131 by executing the biological indicator calculation program. The same is true for other programs. Furthermore, the processor 2 also operates as a functional unit that provides the functions of each of the multiple processes executed by each program. Computers and computer systems are devices and systems that include these functional units.

[0201] The storage device 4 stores data used by the above-mentioned functional units. The storage device 4 stores vehicle-mounted sensor data 161, biosensor data (first biodata) 163, business / environmental data 174, accident risk prediction model 171, accident risk prediction data 166, hazard occurrence data 162, bio-indicator data 164, accident risk definition model 170, alarm definition data 172, accident risk estimation data 165, response plan data 173, bio-state estimation model 169, personal severity data 167, combination table 168, attribute information data 175, and learning information data 176.

[0202] In addition, as described later, the vehicle sensor data 161 includes the vehicle sensor data 161-1 (first vehicle sensor data) used when generating the accident risk definition model 170 and the vehicle sensor data 161-2 (second vehicle sensor data) used when generating the accident risk prediction model 171. In the following description, the symbol "161" is used to omit the "-" after the vehicle sensor data without distinguishing the vehicle sensor data.

[0203] In addition, as described later, the biometric data 164 includes biometric data 164-1 (first biometric data) used when generating the accident risk prediction model 171, biometric data 164-2 (second biometric data) used when inputting into the accident risk prediction model 171 to generate the accident risk prediction data 166, and biometric data 164-3 (third biometric data) used when generating the biological state estimation model 169. In the following description, the symbol "164" is used to omit the "-" after the biometric data without distinguishing them. The details of each data are described later.

[0204] The input / output device 5 includes input devices such as a mouse, a keyboard, a touch panel, or a microphone, and output devices such as a display and a speaker. The communication device 6 communicates with the vehicle via a network 14 .

[0205] <Processing Overview>

[0206] Fig.19 1 is a diagram showing an overview of processing performed in the operation support system. The example shown in the figure shows an example consisting of a learning phase performed before the start of operation of the vehicle 7 and an operation phase performed during the operation of the vehicle 7.

[0207] First, the operation support system generates, in the operation support server 1 , the accident risk prediction model 171 and the biological condition estimation model 169 necessary for predicting the dangerous state of the driver who is driving the vehicle 7 .

[0208] First, the accident risk definition generation unit 132 inputs the pre-set danger occurrence data 162 and the vehicle sensor data 161-1 collected in the past, and generates the accident risk definition model 170 (S101) for estimating the probability of danger occurrence as the accident risk. In addition, the time series of the vehicle sensor data 161-1 and the time series of the danger occurrence data 162 are the same time series.

[0209] The accident risk definition model 170 takes the vehicle sensor data 161-2 (described later) indicating the past driving state of the vehicle as input and estimates the probability of occurrence of a dangerous phenomenon, that is, the accident risk. In the following, the value of the accident risk is expressed as a percentage.

[0210] The danger occurrence data 162 is data obtained by extracting phenomena judged to be dangerous based on the vehicle-mounted sensor data collected in the past. Specifically, it is data pre-set based on the type of phenomenon and the date and time of occurrence, for events such as sudden braking and sharp turns or phenomena leading to the events, judged by the driver or manager or judged by a commercially available vehicle-mounted alarm or AI. Furthermore, the danger occurrence data 162 can set the importance of each event.

[0211] In addition, in this embodiment, an event is defined as a phenomenon that leads to a traffic accident. In addition, the phenomenon that leads to an event indicates that the driver is in a state of near miss. Furthermore, an example is shown in which the degree of danger increases as the importance increases. In the following, an event or a phenomenon that leads to an event is defined as a dangerous phenomenon.

[0212] In addition, the dangerous state in a narrow sense described in the above-mentioned subject includes the above-mentioned events, and the dangerous state in a broad sense can include the phenomena leading to the above-mentioned events.

[0213] The accident risk definition model 170 is a machine learning model that estimates the accident risk of a dangerous phenomenon by inputting the vehicle-mounted sensor data 161-2. As a machine learning model, well-known or publicly known methods such as support vector machines, neural networks, and logistic regression can be used. In addition, multiple accident risk definition models 170 can be generated for each type of dangerous phenomenon.

[0214] Next, the accident risk estimation unit 133 inputs the vehicle-mounted sensor data 161 - 2 collected in the past into the accident risk definition model 170 , and generates accident risk estimation data 165 that estimates the accident risk of the occurrence of a dangerous phenomenon in a time series manner ( S102 ).

[0215] The accident risk estimation data 165 is data in which the probability of occurrence of a dangerous phenomenon is generated for the onboard sensor data 161 - 2 in the same time series as the onboard sensor data 161 - 2 .

[0216] After that, the accident risk prediction model generation unit 135 inputs the accident risk estimation data 165 and the past biometric data 164-1, predicts and outputs the accident risk after a predetermined time based on the biometric data 164-2 (described later) of the driver of the traveling vehicle 7, and generates the accident risk prediction model 171 (S104). In addition, the past biometric data 164-1 is calculated by the biometric calculation unit 131 based on the past biosensor data 163 corresponding to the vehicle-mounted sensor data 161-2 collected in the past.

[0217] In this embodiment, the biometric data 164-1 calculated based on the biosensor data 163 collected in the past can be calculated using, for example, an average heart rate calculated based on the driver's heartbeat data, a maximum Lyapunov exponent calculated by nonlinear analysis of a heartbeat data sequence, a frequency domain analysis of heartbeat interval data extracted from the heartbeat data, or an autonomic nervous function index (LF / HF, etc.). In addition, the autonomic nervous function index (LF / HF, etc.) can be calculated using methods such as time domain analysis and nonlinear analysis.

[0218] The biometric indicator data 164 - 1 is a biometric indicator calculated by the biometric indicator calculation unit 131 based on the past biometric sensor data 163 corresponding to the time series of the vehicle-mounted sensor data 161 - 2 .

[0219] The accident risk prediction model 171 is composed of a well-known or publicly known machine learning model, and takes the biometric data 164-2 of the driving driver as input and outputs the accident risk after a predetermined time. The biometric data 164-2 of the driving driver is a biometric calculated by the biometric calculation unit 131 based on the driver's biosensor data 163, and is received by the operation support server 1 from the driving vehicle 7.

[0220] In addition, as the accident risk prediction model 171, a plurality of models may be generated in advance according to the type of accident risk, the type of biological indicator data used, the driving environment of the vehicle to which the prediction is applied, and the prescribed time. By generating a plurality of models, it is possible to select and use appropriate models in the risk prediction process.

[0221] In addition, multiple models can also be generated according to the attribute information data 175 that stores the driver's business characteristics (general road driving, highway driving, continuous work day and night, etc.), driving experience (years of driving, driving skills, type of driver's license held), and health characteristics (gender, amount of sleep time, etc.).

[0222] Next, the biological state estimation model generation unit 134 takes the biological indicator data 164-3 measured in the past as input, generates and outputs a biological state estimation model 169 that estimates and outputs the severity of the person's condition, indicating the degree of deviation of the input (current biological indicator data 164-2) from the biological indicator data 164-3 measured in the past (S103). In addition, the degree of deviation of the current biological indicator data 164-2 from the biological indicator data 164-3 measured in the past can also be expressed as an abnormality.

[0223] The biological state estimation model 169 is composed of a well-known or publicly known statistical model or machine learning model. Typically, a plurality of biological state estimation models 169 are generated for each type of biological indicator data 164-3, and are unsupervised models that take a certain type of biological indicator data 164-3 as input and output a statistic representing the degree of abnormality (the severity of the individual) based on the biological indicator data 164-3 used in model generation.

[0224] As a statistical model, for example, a statistical model that can output a z-score that normalizes the input using the mean and standard deviation of the bioindicator data 164-3 used in model generation, a statistical model that normalizes the output to a range of 0 to 1 using the maximum and minimum values ​​of the bioindicator data 164-3 used in model generation, a statistical model that uses the statistic of the bioindicator data 164-3 used in model generation as a reference and outputs the rate of change compared to the value, and a statistical model that outputs a quantile on the data distribution estimated based on the bioindicator data 164-3 used in generating the biological state estimation model 169, etc.

[0225] In addition, as a machine learning model, for example, an abnormality detection model can be used that non-parametrically estimates data distribution from the biological indicator data 164-3 used in generating the biological condition estimation model 169 and outputs the distance to the cluster center of the distribution.

[0226] In addition to the unsupervised biological state estimation model 169, a supervised biological state estimation model 169 can also be used. In this case, the biological state estimation model 169 can be constructed using a known or publicly known statistical model or machine learning model that estimates the driver's sleepiness estimated by a known or publicly known method, the number of blinks (blink count) as an indicator of the driver's wakefulness, and the subjective fatigue obtained by the driver answering the subjective fatigue according to the Visual Analogue Scale (VAS) as a response variable. In this case, for example, a regression model can be used as a statistical model, and a support vector machine and a neural network can be used as a machine learning model.

[0227] In addition, the biological indicator data 164-3 used as the input of the biological state estimation model 169 is typically univariate, but may also be multivariate. When the input is multivariate, it is preferable to use as input a combination of biological indicator data types that are easy to apply in generating a countermeasure for avoiding a dangerous state in the warning presentation process described later. In this case, the biological state estimation model 169 can be constructed using a well-known method such as a multivariate regression model, Mahalanobis-Taguchi method, and multivariate statistical process management.

[0228] In addition, the biological state estimation model 169 may be generated for each driver, or a plurality of drivers may be combined to generate one biological state estimation model 169. When a plurality of drivers are combined, since there are large individual differences in biology, it is preferable to generate the biological state estimation model 169 for a group of drivers that exhibit similar biological responses.

[0229] By generating the biological condition estimation model 169 based on a plurality of driver groups showing similar biological responses, individual differences in biological responses can be absorbed and the time required to measure data for generating the biological condition estimation model 169 can be shortened.

[0230] Furthermore, after measuring the required amount of data, it is preferred to switch from the biological condition estimation model 169 generated based on a plurality of driver groups showing similar biological responses to the biological condition estimation model 169 generated based only on the target driver and use it. This allows outputting the severity level of the individual that takes the individual biological response into more consideration.

[0231] Next, the operation support system uses the operation support server 1 to predict the probability of occurrence of a dangerous phenomenon of the driver who is actually driving the vehicle 7 using the generated accident risk prediction model 171 and the biological condition estimation model 169 .

[0232] First, the risk prediction unit 137 inputs the biosensor data 163 such as heart rate data measured by the biosensor 12 received by the data collection unit 139 from the vehicle 7 to the bio-indicator calculation unit 131 to calculate the bio-indicator data 164-2 (S105).

[0233] In addition, when there are multiple accident risk prediction models 171 and biological state estimation models 169, the risk prediction unit 137 uses the model selection unit 136 to select the accident risk prediction model 171 and biological state estimation model 169 corresponding to the business / environmental data 174 and the attribute information data 175.

[0234] After that, the risk prediction unit 137 inputs the biological indicator data 164-2 to the accident risk prediction model 171 to predict (generate) the accident risk prediction data 166 (S106A). In addition, the risk prediction unit 137 inputs the biological indicator data 164-2 to the biological state estimation model 169 and calculates the personal severity data 167 for each biological indicator data type (S106B).

[0235] Then, the danger prediction unit 137 extracts the type of biological indicator data 164-2 that contributes to the prediction of the accident risk prediction model 166 (contribution index), and saves the accident risk prediction data 166, the accident risk prediction model 171 used to calculate the accident risk prediction data 166, the contribution index, and the personal severity data 167 of the contribution index in the combination table 168 in comparison (S107A).

[0236] In extracting the contribution index, the risk prediction unit 137 extracts the contribution index based on the biological indicator data 164 - 2 , the accident risk prediction data 166 , and the accident risk prediction model 171 used to calculate the accident risk prediction data 166 .

[0237] When a linear model such as logistic regression is used as the accident risk prediction model 171 , the risk prediction unit 137 can simply extract the type of biological indicator data having the largest product of the regression coefficient and the biological indicator data 164 - 2 as a contribution indicator.

[0238] In addition, when using a neural network or the like as the accident risk prediction model 171, for example, the hazard prediction unit 137 can calculate the contribution of each biological indicator data 164-2 using the well-known method SHapley Additive ex Planations (SHAP) used to ensure the interpretability of the machine learning model, and extract the one with the largest contribution as the contribution indicator.

[0239] Then, the warning presentation unit 138 determines whether to issue an alarm (hereinafter referred to as the alarm issuance judgment) based on the accident risk and the severity of the driver stored in the combination table 168 (S107B). When the warning presentation unit 138 determines that the driver is in a broadly dangerous state, it generates a response plan for avoiding the broadly dangerous state (S107C).

[0240] In the judgment of issuing an alarm, when the accident risk or the personal severity exceeds the prescribed threshold value, or when the value related to the accident risk and the personal severity exceeds the prescribed threshold value, the warning presentation unit 138 judges that it is equivalent to a dangerous state (dangerous phenomenon) in a broad sense. In addition, in the former case, when judging to issue an alarm, the alarm presentation unit 138 may not set a threshold value for either the accident risk or the personal severity.

[0241] For example, when the warning presentation unit 138 sets a threshold value only for the accident risk, it can issue an alarm for scenes generally judged to have a high accident risk. In addition, when the warning presentation unit 138 sets a threshold value only for the individual severity, it can issue an alarm for abnormalities in the biometric data 164-2, such as a sudden change in the driver's personal physical condition. In addition, the threshold value can be set not only for the corresponding records of the accident risk and the individual severity, but also for the time series of the accident risk and the individual severity.

[0242] In the generation of the countermeasure implemented by the alarm presentation unit 138, a countermeasure for avoiding a broadly defined dangerous state is generated based on the combination table 168 and the countermeasure data 173. Based on past conclusions, a method of physiologically interpreting a change in a certain type of biological indicator data 164-2 and a method of changing a certain type of biological indicator data 164-2 are known.

[0243] In the countermeasure data 173, for each type of the biometric data 164-2, a physiological explanation when the value increases or decreases and a method for increasing or decreasing the value of the biometric data 164-2 are stored. Since the contribution index that contributes to the prediction of the accident risk is stored in the combination table 168, by acquiring the physiological explanation corresponding to the contribution index (described later) and the method for increasing or decreasing the contribution index, a countermeasure for avoiding a broadly dangerous state can be generated according to the driver's state.

[0244] For example, when the contribution indicator is LF / HF, which is an indicator of autonomic nervous function, and the individual's severity is high when LF / HF is in the positive direction, it can be physiologically interpreted as excessive tension. The warning presentation unit 138 can generate a relaxation-promoting response plan such as "Please breathe steadily" in order to reduce the driver's LF / HF.

[0245] In addition, when the contribution indicator is the average heart rate, and the average heart rate has a downward trend and the severity of the individual is high, it is physiologically interpreted as deepening sleepiness. The warning presentation unit 138 can generate a response plan such as "Please stop at the next convenience store and stretch outside the car" to encourage people to eliminate sleepiness.

[0246] In addition, in the generation of the response plan, the warning presentation unit 138 can add information about the individual's severity based on the contribution index to the response plan. It is difficult to convey the danger to the driver who receives the warning only by using a warning based on the accident risk, but by presenting the warning in a form that the driver can understand based on the individual's severity, it is expected that the driver's acceptance of the warning will be improved. Furthermore, in the generation of the response plan, the warning presentation unit 138 can add information based on the business / environmental data 174 and the attribute information data 175 to the response plan.

[0247] After generating the countermeasure, the warning presentation unit 138 sends a warning (S8) to the vehicle 7 that is determined to be in a broadly dangerous state using an alarm and a message including the countermeasure. In addition, the warning presentation unit 138 may send 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 displayed on the driver's display may include an image including the driver's position information, etc.

[0248] When the vehicle 7 receives the warning from the driving support server 1 , the prediction result notification device 9 outputs a warning to convey a warning or a message to the driver.

[0249] The operation assistance server 1 judges the broad dangerous state based on the accident risk after a specified time predicted using the accident risk prediction model 171 and the individual severity of the biological indicator data 164-2 calculated using the biological state estimation model 169, and is able to detect the broad dangerous state by absorbing the individual differences of each driver included in the biological indicator data 164-2.

[0250] In addition, when the operation support server 1 determines that the vehicle is in a general dangerous state, it generates a countermeasure for avoiding the general dangerous state based on the contribution index and the individual severity of the contribution index, and sends a warning including the countermeasure to the prediction result notification device of the vehicle and the display of the input-output device 5. Thus, the operation support server 1 can warn the driver and the driver manager before the risk of an accident increases, so that they can take countermeasures for avoiding the general dangerous state.

[0251] According to the above, the operation support server 1 can avoid a dangerous state in a broad sense in advance, prevent the driver from falling into a dangerous state in a narrow sense, and support the safe operation of each vehicle 7 .

[0252] <Details of the processing>

[0253] Fig. 20 171 is a flowchart showing an example of the process of generating the accident risk prediction model 171. This process is performed before receiving the biological indicator data 164-2 from the vehicle 7 performing the danger detection while traveling ( Fig.19 The learning phase) is implemented to generate the accident risk prediction model 171 in advance.

[0254] The accident risk definition generation unit 132 generates an accident risk definition model 170 using the risk occurrence data 162 collected in the past and the corresponding vehicle-mounted sensor data 161 - 1 as input ( S101 ).

[0255] Next, the accident risk prediction model generation unit 135 inputs the vehicle-mounted sensor data 161 - 2 collected in the past to the generated accident risk definition model 170 to generate the accident risk estimation data 165 ( S102 ).

[0256] Next, the accident risk prediction model generation unit 135 generates an accident risk prediction model 171 (S104) based on the biometric data 164-1 corresponding to the vehicle-mounted sensor data 161-2 collected in the past and the accident risk estimation data 165. In this case, the accident risk prediction model generation unit 135 may also generate an accident risk prediction model 171 for each similar business characteristic or environmental characteristic based on the business / environmental data 174, and attach the information to the accident risk prediction model 171 as meta-information.

[0257] Through the above-mentioned processing, after the operation support server 1 generates the accident risk definition model 170 and the accident risk estimation data 165 , it generates the accident risk prediction model 171 based on the biological indicator data 164 - 1 which corresponds to the accident risk estimation data 165 in time series.

[0258] The above processing is performed every time the amount or type of hazard occurrence data 162 increases by a certain amount, every time the accident risk estimation data 165 with high accident risk increases to a certain extent, or at a certain frequency, and the accident risk definition model 170 and the accident risk prediction model 171 are updated each time.

[0259] Fig.21 169 is a flowchart showing an example of the generation process of the biological state estimation model 169. This process is performed before the danger prediction of a certain vehicle 7 during driving is performed ( Fig.19 The biological state estimation model 169 is generated for each type of biological indicator data 164-3.

[0260] The biological condition estimation model generation unit 134 first extracts a group of drivers who exhibit similar biological responses as a group of drivers for generating the biological condition estimation model 169 (S111). Typically, the group of drivers who exhibit similar biological responses is extracted by clustering using a known or public method using daily statistics such as the mean, standard deviation, median, and quantile of a certain type of biological indicator data 164-3, or weekly or monthly statistics.

[0261] In this case, the biological condition estimation model generation unit 134 may add information included in the business / environment data 174, gender included in the attribute information data 174, and age group obtained based on age stratification, etc., in addition to the biological indicator data 164-3, to the input for clustering.

[0262] Furthermore, when selecting the biometric data 164-3 to be input for clustering, the biological condition estimation model generation unit 134 may select only the biometric data 164-3 measured under the same measurement conditions based on the learning information data 176, and input only the biometric data 164-3 of the period stored in the learning information data 176. In this way, the measurement conditions are controlled, and it is expected that the error of the biometric data 164-3 will be suppressed.

[0263] Furthermore, when a sufficient amount of biometric data 164-3 has been successfully measured for a specific driver, the biological condition estimation model generation unit 134 may extract only the specific driver instead of the plurality of driver groups, thereby estimating the severity level of the specific individual that better matches the characteristics of the individual.

[0264] Next, the biological condition estimation model generation unit 134 receives the biological indicator data 164 - 3 collected in the past for the extracted driver group as input, and generates a biological condition estimation model 169 for each type of the biological indicator data 164 - 3 ( S103 ).

[0265] In this case, similarly to the above-mentioned driver extraction, the biological condition estimation model generation unit 134 can select only the biological indicator data 164-3 measured under the same measurement conditions based on the learning information data 176 and input only the biological indicator data 164-3 of the period stored in the learning information data 176.

[0266] In addition, when the biological state estimation model generation unit 134 generates a supervised model as the biological state estimation model 169, the response variable of the period stored in the learning information data 176 and the corresponding biological indicator data 164-3 are used as input to generate the biological state estimation model 169 for each type of biological indicator data 164-3.

[0267] In this case, the living body condition estimation model generation unit 134 may add the type of the biological indicator data 164 - 3 and information on similar biological features as input to the living body condition estimation model 169 as meta-information.

[0268] Through the above-described processing, after a group of drivers showing similar biological response characteristics is extracted, a biological state estimation model 169 is generated for each category of the biological indicator data 164 - 3 based on the biological indicator data 164 - 3 regarding the extracted group of drivers.

[0269] In addition, the biological state estimation model 169 is updated every time the number of target drivers increases by a certain amount, every time the biological indicator data 164-3 increases by a certain amount, or every certain period, based on the above-mentioned processing generated by multiple drivers. Furthermore, the biological state estimation model 169 is updated every time the biological indicator data 164-3 increases by a certain amount, or every certain period, based on the above-mentioned processing generated by a specific driver.

[0270] In addition, the biometric data 164-3 used to generate the biological state estimation model 169 is different from the biometric data 164-1 used to generate the accident risk prediction model 171. It does not need to correspond to the vehicle sensor data 161-2, nor does it need to correspond to situations with high accident risks, and there are fewer restrictions on measurement scenarios.

[0271] Therefore, since it is easy to measure the biological indicator data 164-3 for generating the biological condition estimation model 169, data can be measured in a shorter period of time than when generating the accident risk prediction model 171, and it can be expected that the update frequency can be set higher.

[0272] Furthermore, since a broadly defined dangerous state rarely occurs, it takes time to construct the accident risk prediction model 171 based only on a plurality of driver groups with similar biological responses. On the other hand, in this embodiment, the individual differences in biological responses between a plurality of drivers are absorbed in the biological state estimation model 169 rather than in the accident risk prediction model 171, so that the data generation model of drivers with different biological responses can also be applied in the generation of the accident risk prediction model 171. Moreover, in the generation of the biological state estimation model 169, the model can be generated based only on the data of drivers with similar biological characteristics that can be easily collected during driving.

[0273] As described above, compared with a method of taking into account individual differences in biological responses when generating the accident risk prediction model 171, the present embodiment can shorten the time required for model generation required to achieve the risk prediction process.

[0274] Fig. 22 164 is a flowchart showing an example of a calculation process of the biological indicator data 164. This process is performed in the biological indicator calculation unit 131 every time data is received from the vehicle 7.

[0275] The data collection unit 139 of the operation support server 1 receives data from the vehicle-mounted sensor 8 and the biosensor 12 from the driving data collection device 10 of the vehicle 7 ( S121 ).

[0276] The data collection unit 139 stores the heartbeat data detected by the heartbeat sensor 121 and the acceleration data detected by the acceleration sensor 122 in the biosensor data 163 ( S122 ). In addition, the data of the vehicle-mounted sensor 8 is stored in the vehicle-mounted sensor data 161 .

[0277] Next, the biometrics calculation unit 131 uses Fig.23 The method described later in the above-mentioned process calculates the biometric data 164 by taking the biosensor data 163 as input (S123). The biometric calculation unit 131 stores the calculated data in the biometric data 164 (S124). The biometric data 164 is calculated from the biosensor data 163 by the above-mentioned process.

[0278] Fig.23 1 is a flowchart showing an example of a calculation process of the biometric data 164 using the heartbeat data. This process is performed in the biometric calculation unit 131 every time the biosensor data 163 is stored.

[0279] The calculation and processing of the biometric data 164 using the heartbeat data is performed by a well-known or publicly known method, so only its outline is described below. First, the biometric calculation unit 131 reads the heartbeat data in the biosensor data 163 (S131). In addition, the heartbeat data refers to data from which the heartbeat interval can be extracted from the driver who is driving using the heartbeat sensor 121, and specifically, electrocardiogram data, pulse data, or heart sound data can be cited.

[0280] Next, the biometrics calculation unit 131 extracts the beat interval IBI (Inter Beat Interval) from the heartbeat data and calculates its time series (S132). In the case of electrocardiogram data, the IBI can be used to calculate the RRI (RR Interval) time series using the interval of the electrocardiogram R wave. RRI is specifically mentioned as an example below.

[0281] The bio-index calculation unit 131 calculates the autonomic nervous function index by performing a necessary analysis based on the RRI. Examples of the necessary analysis include frequency domain analysis, time domain analysis, and RRI nonlinear analysis.

[0282] The biological indicator calculation unit 131 calculates the frequency domain index according to the power spectral density of the RRI time series in the frequency domain analysis (S133). The RRI time series is non-uniformly spaced time series data, so after resampling to equal intervals using spline interpolation or the like, the power spectral density PSD (Power Spectral Density) is calculated using a known method such as an autoregressive model or a maximum entropy method, or a Lomb-Scargle method that can use non-uniformly spaced data.

[0283] The bio-indicator calculation unit 131 calculates the calculated PSD, such as the integrated value LF of the low frequency band of 0.05 Hz-0.15 Hz, the integrated value HF of the high frequency band of 0.15 Hz-0.40 Hz, the sum TP of LF and HF, LF / HF which is LF divided by HF, LFnu which is LF divided by TP and converted into a percentage, etc., as frequency domain indices of the autonomic nervous function index.

[0284] The biological indicator calculation unit 131 calculates the time domain indicator (S134) by calculating the statistics of the RRI time series and the difference series of the adjacent RRI, that is, the ΔRRI time series. For example, the reciprocal of the average value of the RRI time series, that is, the average heart rate, and the standard deviation of the RRI, that is, SDNN, are calculated. In addition, based on the ΔRRI time series, for example, the total number of data whose absolute value of the difference constituting the ΔRRI time series exceeds 50ms, that is, NN50, pNN50 obtained by dividing NN50 by the total number of data in the ΔRRI time series, and the standard deviation of ΔRRI, that is, SDSD, are calculated as the time domain indicator of the autonomic nervous function indicator.

[0285] The bio-indicator calculation unit 131 calculates the nonlinear feature quantity by various methods in the RRI nonlinear analysis (S135). For example, the bio-indicator calculation unit 131 calculates the ellipse area S of the plotted area by Poincare diagram analysis, which plots the RRI time series RRI(t) on the X-axis and the time series RRI(t+1) obtained by advancing the RRI time series by one time scale on the Y-axis. In addition, the bio-indicator calculation unit 131 calculates α1 and α2 obtained by similarity entropy and Detrended Fluctuation Analysis, and the pitch and entropy obtained by pitch-entropy analysis, and calculates the RRI nonlinear index as an autonomic nervous function index.

[0286] In addition, the biometric calculation unit 131 can also perform nonlinear analysis on the heartbeat data before RRI extraction to calculate the nonlinear index of the heartbeat (S136). For example, the biometric calculation unit 131 calculates the correlation dimension and the maximum Lyapunov exponent of the heartbeat data by applying chaos analysis to the heartbeat data, and calculates the nonlinear index of the heartbeat as an autonomic nervous function index.

[0287] The calculated autonomic nervous function index reflects the biological state of the driver controlled by the autonomic nervous system, so by calculating the index measuring the biological state of the driver necessary for prompting the driver with a countermeasure for avoiding a broadly defined dangerous state, it is possible to realize the presentation of a warning that matches the biological state. For example, LF / HF is an index that measures the balance between sympathetic nerve activity that is mainly active under stress and tension and parasympathetic nerve activity that is mainly active under relaxation, so the biological index calculation unit 131 can present a warning that takes into account the driver's tension-relaxation state by calculating LF / HF.

[0288] Then, the biometric calculation unit 131 stores the calculated autonomic nerve function index group in the biometric data 164 (S124). Through the above-mentioned process, the autonomic nerve function index is calculated based on the heartbeat data in the biosensor data 163 and stored as the biometric data 164.

[0289] In addition, the biological indicator calculation unit 131 may execute the processing of steps S133 to S136 in parallel or in sequence.

[0290] Fig.24 1 is a flowchart showing an example of risk prediction processing performed in the operation support server 1. This processing is executed at a predetermined timing such as when the biological indicator data 164-2 is updated. The risk prediction unit 137 acquires the biological indicator data 164-2 from the latest period to the predetermined period (S112).

[0291] Next, the model selection unit 136 determines the driver ID and vehicle ID of the analysis object based on the business / environmental data 174, and selects the accident risk prediction model 171 and the biological state estimation model 169 (S113 and S114) based on the business / environmental data 174. In selecting the accident risk prediction model 171, the model to be used is selected based on the vehicle size, weather, congestion conditions, etc. of the business / environmental data 174, for example.

[0292] In selecting the biological condition estimation model 169 , for example, the model selection unit 136 selects a model to be used based on information on the biological condition estimation model 169 corresponding to the driver ID stored in the attribute information data 175 .

[0293] Next, the risk prediction unit 137 inputs the biometric data 164-2 of the driver ID of the analysis object to the selected accident risk prediction model 171 to predict the accident risk, and stores it in the accident risk prediction data 166 (S106A). In addition, the above shows an example of inputting the biometric data 164-2 to the accident risk prediction model 171, but it is not limited to this, and the business / environmental data 174 and the attribute information data 175 can also be added to the input of the accident risk prediction model 171.

[0294] For example, by inputting the driving skill information obtained by driving suitability diagnosis, etc., and the personality and temperament information obtained by Temperament and Character Inventory (TCI), Temperament Evaluation of the Memphis, Pisa, Paris, and San Diego Autoquestionnaire (TEMPS-A), Ten Item Personality Inventory for measuring five major personality traits, NEO-FFI, etc. included in the attribute information data 175, it is possible to achieve a more accurate prediction of accident risk while also taking into account the driver's driving skills and personality.

[0295] In addition, the risk prediction unit 137 inputs the biometric data 164-2 of the driver ID of the analysis object to the selected biometric state estimation model 169, and estimates a plurality of personal severity levels for each type of biometric data, and stores them in the personal severity level data 167 (S106B). In addition, the above shows an example of inputting the biometric data 164 and the biometric data 164-2 as a single variable, but in addition to the above, for example, the risk prediction unit 137 may input the biometric data 164-2 of multiple variables to calculate the personal severity level.

[0296] Next, the danger prediction unit 137 uses the predicted accident risk prediction data 166 and the accident risk prediction model 171 for predicting the accident risk, the biometric data 164-2 of the driver ID of the analysis object, and the estimated personal severity data 167 to extract the contribution index that contributes to the prediction of the accident risk, and saves the information of the contribution index, the personal severity of the contribution index, the accident risk and the accident risk prediction model 171 in a comparative manner in the combination table 168 (S107A).

[0297] Through the above processing, at a specified time such as after the biological indicator data 164-2 is updated, the contribution index used to detect a broad dangerous state and issue a warning, the individual severity of the contribution index, the accident risk and the accident risk prediction model information are saved in the combination table 168.

[0298] In addition, the risk prediction unit 137 may execute the processing of the above-mentioned steps S113 and S114 in parallel or sequentially.

[0299] Fig.25 1 is a flowchart showing an example of the warning presentation process performed in the operation support server 1. This process is executed at a predetermined timing, such as every time the connection table 168 is updated.

[0300] The warning presentation unit 138 reads the condition for presenting a warning, that is, the definition for issuing an alarm, from the alarm definition data 172 (S141). A plurality of definitions for issuing an alarm can be set in accordance with the accident risk prediction model 171 and the biological condition estimation model 169. In addition, a plurality of definitions for issuing an alarm can be set according to the object and method of presenting the alarm, such as voice notification using the prediction result notification device 9 for the driver, display on a display, display on a display of the input / output device 5 of the driver manager, and notification by mail via the input / output device 5.

[0301] The warning presentation unit 138 refers to the accident risk and the driver's severity level of each record in the combined table 168 ( S142 ), and determines whether there is a driver whose level exceeds the threshold set in the warning issuance definition ( S107B).

[0302] If there is a driver whose performance exceeds the threshold and meets the judgment of issuing an alarm, the warning presentation unit 138 generates a countermeasure including a warning to be presented ( S107C). ​​The countermeasure is generated by referring to the countermeasure data 173 corresponding to the contribution index of the combination table 168 .

[0303] In addition, regarding the countermeasure, in order to improve the warning acceptance level of the driver receiving the warning, the warning presentation unit 138 may also use the personal severity of the contribution index of the combination table 168 to add information that allows the driver to know the degree of abnormal state. For example, the warning presentation unit 138 may add information such as "this is a state that only occurs once in N years".

[0304] Furthermore, the warning presentation unit 138 can add information obtained based on the business / environmental data 174 and the attribute information data 175 and the combination table 168 to the response plan. For example, the warning presentation unit 138 can update the content of the response plan feedback based on the information of the time N when the accident risk is considered to increase according to the information of the accident risk prediction model 171 stored in the combination table 168. In addition, for example, the warning presentation unit 138 can also update the content of the feedback based on the weather and temperature information stored in the business / environmental data 174.

[0305] Then, the warning presentation unit 138 acquires the vehicle ID of the transmission target from the driver ID based on the business / environmental data 174, and uses the vehicle 7 as the transmission target. In addition, when the driver manager is used as the transmission target instead of the driver, the warning presentation unit 138 can determine the manager ID instead of acquiring the vehicle ID, and use the input / output device 5 of the operation support server 1 operated by the manager as the transmission target.

[0306] Thereafter, a warning is transmitted using an alarm or a message obtained by adding an accident risk alarm based on an alarm issuance definition corresponding to an alarm issuance judgment to the countermeasure generated by the alarm presentation unit 138 (S108).

[0307] Through the above processing, a warning is sent to the vehicle in which the driver corresponding to the broadly dangerous state is detected. In the vehicle 7 that has received the warning, the prediction result notification device 9 notifies the driver of the warning. In addition, in the operation support server 1, the warning presentation unit 138 displays the vehicle that has sent the warning on the display of the input / output device 5.

[0308] Fig.26A , Fig.26B 1 is a diagram 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 receiving a warning from the operation support server 1. The warning presentation screen 1000A includes an area 1010 for displaying an accident risk warning and a comment area 1020 for displaying a countermeasure for avoiding a broadly defined dangerous state.

[0309] In the area 1010 displaying the accident risk alert, for example, warning text 1011 indicating an increased accident risk and information 1012 on the individual's severity based on the biometric data 164-2 of the contribution index are displayed. This is because, rather than merely notifying the driver of a dangerous state, the driver is given a warning in a form that is understandable to the driver based on the individual's severity based on the contribution index, such as "this is a state that only occurs once in half a year for you." This can be expected to increase the driver's acceptance of the warning.

[0310] In addition, in the display of the information 1012 based on the severity of the individual, it is also possible to know whether the display is based on the biological state estimation model 169 generated based only on the biological indicator data 164-2 of the target driver himself, or based on the biological state estimation model 169 generated based on the biological indicator data of multiple drivers who show biological response characteristics similar to those of the target driver.

[0311] For example, in Fig.26B In the warning presentation screen 1000B, the information 1012B based on the individual's severity in the area 1010 displaying the accident risk alert can also be displayed with text such as "For drivers like you, this situation only occurs once in half a year", so that it can be seen that the data is compared with multiple drivers.

[0312] In addition, for example, when comparing with a plurality of drivers, information 1013 on which group of drivers the driver is compared with may be displayed to make it known. In addition, which biological condition estimation model 169 is used may be displayed by an icon or the like. As described above, it is expected that the driver will easily understand the warning as "his own business".

[0313] Furthermore, in the annotation area 1020 of the warning presentation screens 1000A and 1000B, for example, an explanation 1021 of the physiological state based on the contribution index and a specific response plan 1022 for avoiding a broadly dangerous state are presented, whereby the driver who receives the warning presentation does not just receive the warning but is able to understand the next action to be taken to eliminate the dangerous state and take action.

[0314] In addition, although the presenting of the warning on the warning presentation screen is shown in this example, the warning may be presented in other ways. For example, the warning may be presented in the form of a text with the same content as that displayed on the warning presentation screen being read aloud by a machine.

[0315] <Data structure>

[0316] Next, the characteristic structure of each data used in the operation support system is shown.

[0317] Fig.27A This is an example of the data structure of the biometric data 164 stored in the operation support system of the present invention. The biometric data 164 stores the driver ID 1401, the vehicle ID 1402, the date and time 1403, and various autonomic nervous function indices calculated by the biometric calculation unit 131. Here, as the autonomic nervous function indices, for example, LF / HF 1404 as a frequency domain index, average heart rate 1405 as a time domain index, NN50 (1406), or α1 (1407) as an RRI nonlinearity index, etc. can be cited.

[0318] Fig.27B This is an example of the data structure of the task / environment data 174 held in the operation support system of the present invention. The task / environment data 174 stores task information of the driver in the task being performed and driving environment information.

[0319] Typically, the driver ID 411, the start date and time 412 and the end date and time 413 indicating the duration of the information, the vehicle ID 414 used, the vehicle type 415 indicating, for example, the size of the vehicle being driven, the weather 416 and the temperature 417 of a representative location in the driving area, the business type 418 indicating the business status from the start date and time 412 to the end date and time 413, and the business reservation 419 such as the destination are saved.

[0320] Fig.27C This is an example of the data structure of the attribute information data 175 held in the driving support system of the present invention. The attribute information data 175 stores basic information of the driver that rarely changes with time (changes with the passage of time).

[0321] Typically, the driver ID 1421, the start date 1422 and end date 1423 of the duration of the application of the basic information, the vehicle ID 1424 of the vehicle used, the driving skill information measured by the driver suitability diagnostic test (for example, skill A 14215, a pseudonym), the personality temperament information measured by the Temperament and Character Inventrory (TCI) (for example, personality A (1426), a pseudonym), the gender 1427, the age 1428, the driving experience 1429, etc. are stored. In addition to the above, the biological state estimation model ID 1430 and the accident risk prediction model ID 1431 are also stored as various model information used preferentially for the driver.

[0322] Fig.27D This is an example of the data structure of the learning information data 176 maintained in the operation support system of the present invention. The learning information data 176 stores information used when generating the biological state estimation model 169, typically storing the driver ID 1441 and the period information of the biological indicator data 164 used for learning, that is, the start time 1442 and the end time 1443.

[0323] Furthermore, the learning information data 176 may store the measurement conditions 1444 of the data of the period information for the purpose of making the measurement conditions of the data used to generate the biological state estimation model 169 consistent. In this case, the measurement conditions 1444 may be automatically determined and stored by the driving data collection device 10 based on the vehicle-mounted sensor data 161-2, etc. (for example, high-speed driving), or may be manually input by the driver himself from a predetermined condition via the work state input device 130 (for example, lunch break).

[0324] Furthermore, when the biological condition estimation model 169 is generated by a supervised model, in addition to the above, for example, the value 1445 obtained by measuring the subjective fatigue level by VAS may be stored as a response variable required for generating the supervised model.

[0325] Fig.27E This is an example of the data structure of the danger occurrence data 162 stored in the driving support system of the present invention. The danger occurrence data 162 stores data setting the type and occurrence date of the danger phenomenon detected by the driver or driver manager or a commercially available vehicle alarm.

[0326] Typically, the vehicle ID 1451, the date and time of occurrence of the dangerous phenomenon 1452, and the type of danger 1453 are stored. In addition, the importance 1454 of the phenomenon in which the severity of the detected dangerous phenomenon is recorded by visual confirmation or the like may also be stored.

[0327] Fig.27F This is an example of the data structure of the accident risk prediction data 166 held in the operation support system of the present invention.

[0328] The accident risk prediction data 166 stores the accident risk determined based on the biometric data 164-2 using the accident risk prediction model 171. Typically, the data stores the driver ID 1461, date and time 1462, the ID 1463 of the accident risk prediction model 171 used to predict the accident risk, and the accident risk 1464 indicating the predicted probability (percentage) of occurrence of the dangerous phenomenon.

[0329] Figure 27G This is an example of the data structure of the personal severity data 167 held in the operation support system of the present invention.

[0330] The personal severity level data 167 stores the personal severity level determined from the biometric data 164-2 using the biometric state estimation model 169. Typically, the data stores driver ID 1471, date and time 1472, type 1473 of the biometric data 164-2 for estimating the personal severity level, ID 1474 of the biometric state estimation model 169 used for estimation, and personal severity level 1475.

[0331] Furthermore, the biological condition estimation model 169 is generated for each biological indicator data type, and since the method of calculating the individual's severity level is assumed to be different for each biological indicator data type, the output format of the individual's severity level is not necessarily fixed.

[0332] For example, in the record of the first row, for the biological data type 1473 "average heart rate" of the biological indicator data 164-2, a statistic approximated by a normal distribution is output, and the personal severity 1475 is expressed as a value "+4SD" from the mean using the standard deviation (SD).

[0333] In addition, for example, in the record of the second row, the quantiles on the data distribution are output for NN50, and the personal severity 1475 is expressed as a value equivalent to the 99% quantile in the data set used to build the model.

[0334] Fig.27H This is an example of the data structure of the combination table 168 held in the operation support system of the present invention. The combination table 168 is a table generated by comparing the values ​​used in the warning presentation process.

[0335] Typically, the driver ID 481, vehicle ID 482, date and time 483, accident risk prediction model ID 484 used for prediction, predicted accident risk 485, the type of biological indicator data contributing to the accident risk prediction, i.e., contribution index 486, biological state estimation model ID 487 used to estimate the biological state for the contribution index, and the personal severity 488 regarding the contribution index are saved.

[0336] Furthermore, a flag for managing whether the warning presentation unit 138 has presented a warning when determining that a certain record is in a dangerous state in a broad sense, that is, an alarm issued flag 489 may be stored.

[0337] Fig.27I This is an example of the data structure of the alarm definition data 172 held in the operation support system of the present invention. The alarm definition data 172 stores the judgment conditions used when judging whether or not a state corresponds to a dangerous state in a broad sense.

[0338] Typically, the alarm ID 1491, the accident risk threshold 1492, and the personal severity threshold 1493 are stored. In addition, for example, the contribution index condition 1494 for limiting the contribution index when using the personal severity threshold 1493, the weather condition 1495 for limiting the target weather in the execution of the business, and the personal severity type 1496 for limiting the personal severity obtained in a specific output form as the judgment object when executing the alarm judgment can also be stored.

[0339] Fig.27J This is an example of the data structure of the countermeasure data 173 held in the operation support system of the present invention. The countermeasure data stores information for generating a countermeasure in warning presentation.

[0340] Typically, the solution ID 1501 for storing the response solution, the contribution index 1502 as a condition for presenting the response solution, the phenomenon 1503 indicating the state of the contribution index, the explanation 1504 of the driver's physiological state assumed based on the phenomenon of the contribution index, and the response solution 1505 for avoiding the corresponding general dangerous state.

[0341] In addition, for example, a plurality of countermeasures 1505 may be set for the same contribution index 1502, phenomenon 1503, and explanation 1504. In this case, the priority 1506 of the presented countermeasures 1505 may be further stored.

[0342] As described above, the operation assistance system of this embodiment detects whether it corresponds to a broadly dangerous state based on the accident risk prediction data 166 calculated by inputting the biological indicator data 164-2 calculated by the operation assistance server 1 based on the heartbeat data of the biosensor 12 into the accident risk prediction model 171, and the personal severity data 167 calculated by inputting the biological indicator data 164-2 into the biological state estimation model 169.

[0343] Thus, the operation support server 1 of this embodiment can detect a dangerous state before it occurs in the operation of the vehicle 7 based on the individual differences of the biometric data 164-2. In addition, the operation support server 1 presents a warning including a countermeasure for avoiding the generalized dangerous state based on the contribution index that contributes to the prediction of the accident risk when it corresponds to the generalized dangerous state, so that the driver who receives the warning can take the action required to avoid the generalized dangerous state. As described above, the operation support server 1 can enable the driver of the vehicle 7 to avoid the generalized dangerous state in advance.

[0344] In the third embodiment, the present invention is applied to the operation support system for managing the vehicle 7, but the present invention is not limited thereto. For example, the present invention can be applied to a mobile object requiring a driver or operator, such as a railway vehicle, a ship, or an aircraft, instead of the vehicle 7.

[0345] Example 4

[0346] Next, a fourth embodiment of the present invention will be described. Except for the differences described below, the components of the operation support system of the fourth embodiment have the same functions as the components of the third embodiment denoted by the same reference numerals, and thus their descriptions are omitted.

[0347] Fig.28 1 is a diagram showing an outline of processing performed in the operation support system of Embodiment 4. Embodiment 4 of the present invention describes the point that when estimating the accident risk, the calculation is performed in consideration of individual differences in biological indicator data.

[0348] The biological indicator data 164 - 1 is input to the biological condition estimation model 169 to calculate the personal severity level data 167A (first personal severity level) ( S106C), and is input to the accident risk prediction model generation unit 135 .

[0349] The accident risk prediction model generation unit 135 generates the accident risk prediction model 171 by taking the biological indicator data 164 - 1 , the accident risk estimation data 165 , and the personal severity data 167A as inputs ( S104 ).

[0350] Then, the risk prediction unit 137 inputs the biological indicator data 164-2 to the biological state estimation model 169 to calculate the personal severity data 167B (second personal severity) (S106B), and outputs the personal severity data 167B to the accident risk estimation unit 133. The accident risk estimation unit 133 inputs the biological indicator data 164-2 and the personal severity data 167A to the input of the accident risk prediction model 171 (S106A).

[0351] In addition, the process of extracting the contribution index using the personal severity data 167B and generating the combination table 168 by the risk prediction unit 137 in step S107A is the same as that in the third embodiment described above.

[0352] Therefore, in the extraction / combination processing of the contribution index performed by the risk prediction unit 137, the input that contributes to the accident risk estimation is selected from the biological indicator data 164-2 and the personal severity data 167B as the contribution index, and when the biological indicator data 164-2 is selected, the personal severity of the selected type of biological indicator data 164-2 is saved in the combination table 168, and when the personal severity data 167B is selected, the personal severity data 167B of the selected type is saved in the combination table 168.

[0353] As described above, in addition to the judgment of the warning presentation, in the accident risk estimation, the fourth embodiment can also estimate the accident risk by taking into account the individual differences in the biological indicator data 164, thereby achieving the effect of realizing the prediction of the accident risk with higher accuracy.

[0354] Example 5

[0355] Next, a fifth embodiment of the present invention will be described. Except for the differences described below, the components of the operation support system of the fifth embodiment have the same functions as the components of the third embodiment denoted by the same reference numerals, and thus their description will be omitted.

[0356] Fig.29 This is a diagram showing an overview of processing performed in the operation support system of the fifth embodiment. Fig.29 The above is omitted Fig.19 The learning phase in the processing overview of the describes the differences in the processing overview for the operation phase.

[0357] The operation assistance server 1 of Example 5 is an example in which the warning presentation unit 138 does not present a warning every time the danger prediction unit 137 adds a record to the combination table 168, but instead prompts the driver manager with an overview of the warning presentation objects so that the driver manager can manually intervene to assist the driver's operation.

[0358] The warning presentation unit 138 updates the intervention candidates that may be intervened by the driver manager every time the danger prediction unit 137 of the operation support server 1 updates the combination table 168 or reads the combination table 168 at a certain period (S15). For example, only the record group of the last N hours of the operation day in which the warning flag 489 issued in the combination table 168 is not "completed" is extracted.

[0359] Afterwards, the warning presentation unit 138 uses the above method to determine whether to issue an alarm for the extracted record group, and extracts the record group corresponding to the generalized dangerous state (S107B). If there is a record group corresponding to the generalized dangerous state, the warning presentation unit 138 generates a response plan for avoiding the generalized dangerous state for each of the record groups (S107C).

[0360] The warning presenting unit 138 displays a list of warning targets for which countermeasures have been generated (S116) on the display of the input / output device 5. In the list of warning targets, warning targets that are considered to require manual intervention by the driver manager are displayed first so that the driver manager can intervene.

[0361] For example, the warning presentation unit 138 compares the business / environmental data 174 with the record groups corresponding to the broadly defined dangerous states, and can display the record groups for which the presented countermeasures cannot be taken at the driver's discretion due to business reasons in the front.

[0362] The input / output device 5 receives intervention from the driver manager on the driver based on the list of warning targets displayed on the display (S117). For example, for a driver who is shown "Please take a rest at a nearby resting place" as a response plan, but there is no resting place nearby and the delivery time of the goods is urgent, the driver manager can use the interface for voice communication with the driver displayed on the list of warning targets to confirm the status of the driver himself.

[0363] Then, if it is confirmed that the driver does not need to rest immediately, the driver manager can indicate a place where the driver can rest after the delivery is completed, thereby presenting a suboptimal solution to the driver.

[0364] In addition, for example, for a driver who is repeatedly warned that "I am becoming increasingly sleepy and need to rest", the driver manager confirms the driver's condition and, if it is judged that it is difficult to execute the original business schedule, the second half of the day's business schedule and the business portion scheduled for the next morning are omitted, thereby adjusting the driver's business workload and allowing him to rest.

[0365] Fig.302 is a diagram showing an example of a warning target list screen 2000 outputted by the display of the input / output device 5. When the warning presentation unit 138 updates the extracted intervention candidates and generates countermeasures, the input / output device 5 updates the warning target list and displays the warning target list screen 2000.

[0366] At the top of the warning target list screen 2000, the current time 2001 and the manager ID 2002, which is the ID of the driver manager who is browsing the warning target list screen 2000, are displayed. In the warning target list 2003, the record group for which the response plan is generated and extracted and generated by the warning presentation unit 138 is displayed in a table form. For the record group, by pressing the downward triangle button in the figure for each column, it is possible to sort the elements that the driver manager attaches importance to and wants to confirm in order of priority.

[0367] The details of the driver corresponding to the record selected for processing by the driver manager are displayed in the lower area 2005. In the example shown in the figure, information that the driver ID corresponding to the record 2004 in the warning target list 2003 is "ZAX001" is displayed.

[0368] The area 2005 displays the current position 2011 showing the position of the driver on a map, a countermeasure 2012 for the selected record 2004, the driver's work schedule 2013, a note 2014 indicating the driver's work status, and a contact driver button 2015 for directly communicating with the driver. In addition, the time transition of the day for the biometric data 164 of the contribution index may also be displayed.

[0369] In the current position 2011 of the driver, the position 2021 of the target driver and the actual driving route 2022 (solid line) and the scheduled driving route 2022 (dotted line) of the day are shown on the map. In addition, the positions 2023 of other drivers existing around the driver are also shown on the map.

[0370] The driver manager refers to the record 2004, the response plan 2012, the business plan 2013, etc., and determines whether the driver can execute the response plan at his own discretion and whether it is necessary to directly confirm the driver's condition through the warning target list screen 2000. Then, the driver manager presses the contact driver button 215 as needed to directly communicate with the driver, thereby manually intervening to avoid a dangerous state in a broad sense.

[0371] As described above, the driver manager can avoid broadly dangerous conditions and assist the safe operation of the managed driver group by executing a response plan 2012 that is difficult for the driver to execute at his own discretion, or executing a suboptimal response plan different from the response plan 2012.

[0372] As described above, in the fifth embodiment, for a response plan that is difficult to implement at the driver's discretion despite being in a broadly dangerous state, the driver manager allows the driver to implement the response plan according to the driver's own state, thereby enabling the driver to adopt a more fundamental response plan for avoiding the broadly dangerous state. As a result, the operation assistance system can assist the driver to adopt a more flexible response plan to avoid the broadly dangerous state.

[0373] According to the above-mentioned embodiments 3 to 5, it is possible to detect a generalized dangerous state (dangerous phenomenon) based on the accident risk prediction model, the biological state estimation model and the biological data. As a result, it is possible to determine whether an alarm needs to be issued based on the generalized dangerous state, and further, it is possible to suggest a response plan for avoiding the generalized dangerous state.

[0374] Summary

[0375] As described above, the operation assisting methods of the above-mentioned embodiments 1 and 2 can adopt the following structure.

[0376] (1) An operation assisting method for assisting the operation of a vehicle (7) by a computer having a processor (2) and a memory (3), characterized in that it comprises: a first step, the computer uses first on-board sensor data (40-m1) representing the driving state of the vehicle collected in the past and danger occurrence data (45) in which information on the occurrence of danger is pre-set from the first on-board sensor data (40-m1) as input, and generates an accident risk definition model (48) for estimating the probability of danger occurrence as an accident risk through machine learning; a second step, the computer inputs second on-board sensor data (40-m2) representing the driving state of the vehicle (7) collected in the past into the accident risk definition model (48), estimating the probability of danger occurrence to generate accident risk estimation data (50); a third step , the computer uses the first biometric data (46-m) pre-calculated from the first biosensor data (biometric sensor data 41) of the driver when collecting the second vehicle-mounted sensor data (40-m2) and the accident risk estimation data (50) as input, and generates an accident risk prediction model (43) for predicting the accident risk after a specified time through machine learning; a fourth step, the computer obtains the second biometric data (41) of the driver who is driving the vehicle (7), and calculates the second biometric data (46) representing the state of the driver from the second biometric data (41); and a fifth step, the computer inputs the second biometric data (46) into the accident risk prediction model (43) to predict the accident risk (44) after the specified time Δt.

[0377] According to the above configuration, the driving support server 1 can predict the accident risk after the predetermined time Δt from the first biological data (biological sensor data 41 ) of the driver who is driving the vehicle 7 .

[0378] (2) The operation support method described in (1) above is characterized by further comprising a sixth step, in which the computer generates and outputs an alarm (600) when the predicted value of the accident risk after the specified time Δt is greater than a preset threshold value Th.

[0379] According to the above structure, the operation support server 1 predicts that the accident risk will reach or exceed the threshold value Th after a predetermined time Δt, and notifies the prediction result notification device 9 of the vehicle 7 and the display of the input / output device 5, thereby being able to use an alarm and a message to alert the driver and the manager before the accident risk increases. Thus, the operation support server 1 can promote the safe operation of each vehicle 7.

[0380] (3) The operation assist method described in (1) above is characterized in that: the fourth step includes: a step of obtaining the driver's heartbeat data as the second biosensor data (41); a step of calculating RRI from the heartbeat data to generate heartbeat change timing data; a step of performing spectral analysis on the heartbeat change timing data; and a step of calculating the autonomic nervous system total power from the result of the spectral analysis from the sum of the intensity LF of the low-frequency component and the intensity HF of the high-frequency component of the power spectrum density, and calculating the bio-indicator data (autonomic nervous system total power 462).

[0381] According to the above configuration, time-series heartbeat variation data can be calculated from the heartbeat data of the biosensor data 41 to calculate the strengths of the low-frequency component and the high-frequency component (LF, HF), thereby obtaining the autonomic nerve total power 462 and the autonomic nerve LF / HF 463 .

[0382] (4) The operation assistance method described in (3) above is characterized in that: in the fifth step, the intensity of the low-frequency component (LF) and the intensity of the high-frequency component (HF) of the power spectrum density are input into the accident risk prediction model, and the accident risk prediction model (43) outputs interpretation data representing the driver's status.

[0383] According to the above configuration, since it is known that the sympathetic nerve activation indicates a stress state and the parasympathetic nerve activation indicates a relaxation state, it is possible to determine whether the driver is in a stress state or a relaxation state from the intensity LF of the low frequency component and the intensity HF of the high frequency component.

[0384] (5) The operation assistance method described in (2) above is characterized in that: in the sixth step, the position information (position 402) of the vehicle (7) is obtained and a screen (prediction result display screen 600) displaying the position information (402) of the vehicle (7) and information (603) representing the status of the driver is output.

[0385] According to the above configuration, the manager can check the accident risks of a plurality of drivers at a glance on the map, and can select a driver with a high accident risk and remotely instruct him.

[0386] (16) The operation assist method described in (1) above is characterized in that: the fourth step includes: a step of obtaining the driver's heartbeat data as the first biosensor data (41); a step of calculating RRI from the heartbeat data to generate heartbeat change timing data; a step of calculating the difference between adjacent RRIs for the heartbeat change timing data to generate a difference sequence ΔRRI(t); and a step of calculating the total number of differences above a specified value XXmsec from the difference sequence ΔRRI(t) as autonomic nerve NNXX and as bio-indicator data (462).

[0387] According to the above structure, it is possible to calculate the time-series heartbeat change data from the heartbeat data of the biosensor data 41, calculate the difference sequence ΔRRI(t) from the heartbeat change data, calculate the total number of differences above the specified value XXmsec as the autonomic nerve NNXX, and thereby obtain the bioindicator data 46 as an indicator of the magnitude of parasympathetic nerve activity.

[0388] Furthermore, the operation support server 1 of the above-mentioned embodiments 3 to 5 can adopt the following configuration.

[0389] (19) An operation assistance method for assisting the operation of a vehicle (7) by a computer (operation assistance server 1) having a processor (2) and a memory (3), characterized in that it comprises: a first step, wherein the computer (1) uses first on-board sensor data (161-1) representing the driving state of the vehicle (7) collected in the past and pre-set danger occurrence data (162) as input to generate an accident risk definition model (70) for estimating the probability of occurrence of a dangerous phenomenon as an accident risk; a second step, wherein the computer inputs second on-board sensor data (161-2) representing the driving state of the vehicle (7) collected in the past into the accident risk definition model (70) to estimate the probability of occurrence of a dangerous phenomenon to generate accident risk estimation data (165); a third step, wherein the computer uses first biometric indicator data (164-1) pre-calculated from the driver's first biosensor data (biosensor data 163) when the second on-board sensor data (161-2) is collected and the accident risk estimation data (165) as input to generate an accident risk definition model (70) for predicting the accident risk. a fourth step, the computer inputs the pre-calculated third biometric data (164-3) to generate a biometric state estimation model (169) for calculating the individual's severity (individual severity data 167); a seventh step, the computer obtains the second biometric sensor data (163) of the driver who is driving from the biometric sensor (12) of the vehicle (7), and calculates the second biometric data (164-2) representing the state of the driver from the second biometric sensor data (163); an eighth step, the computer inputs the second biometric data (164-2) to the accident risk prediction model (171) to predict the accident risk, and inputs the first biometric data (164-2) to the biometric state estimation model (169) to calculate the individual's severity (167); and a ninth step, the computer determines whether to issue a warning based on the predicted accident risk (accident risk prediction data 166) and the calculated individual's severity (167).

[0390] According to the above configuration, the operation support server 1 can detect a dangerous state (dangerous phenomenon) in a broad sense based on the accident risk prediction model 171 , the biological state estimation model 169 , and the biological indicator data 164 .

[0391] (20) The operation assistance method described in (19) above is characterized in that it also includes: a tenth step, in which the computer, when determining that the warning is to be issued, extracts the accident risk prediction model (71), the accident risk (166), and a contribution index including the type of biological data that contributes to the prediction of the accident risk (166); and an eleventh step, in which the computer generates a response plan for avoiding dangerous phenomena based on the individual's severity (67) regarding the contribution index.

[0392] According to the above structure, the operation support server 1 can detect a broadly dangerous state (dangerous phenomenon) based on the accident risk prediction model 171, the biological state estimation model 169 and the biological indicator data 164, and prompt a response plan for avoiding the broadly dangerous state when issuing a warning.

[0393] (21) The operation support method described in (20) above is characterized in that it also includes a twelfth step in which the computer generates and outputs a warning including a countermeasure (countermeasure data 173) for avoiding the dangerous phenomenon.

[0394] According to the above structure, the operation assistance server 1 can detect a broadly dangerous state (dangerous phenomenon) based on the accident risk prediction model 171, the biological state estimation model 169 and the biological indicator data 164, and generate a response plan for avoiding the broadly dangerous state and issue a warning when issuing a warning.

[0395] (22) The operation assistance method described in (19) above is characterized in that: in the fourth step, the first biological indicator data (164-1) and the learning information data (176) of the object period for generating the biological state estimation model (169) at least including the first biological indicator data (164-1) are used as input to generate the biological state estimation model (169) for calculating the severity level (167) of the individual.

[0396] According to the above configuration, when selecting the biometric data 164-3 to be input for clustering, the biometric data 164-3 measured under the same measurement conditions can be selected based on the learning information data 176, and only the biometric data 164-3 of the period stored in the learning information data 176 can be input. Thus, the measurement conditions are controlled, and it can be expected that the error of the biometric data 164-3 can be suppressed.

[0397] (23) The operation assistance method described in (21) above is characterized in that: in the eighth step, a combined table (168) is generated that stores the accident risk (166), the contribution index and the individual's severity (167) related to the contribution index in an associated manner.

[0398] According to the above structure, at a specified time such as after the biological indicator data 164-2 is updated, the contribution index used to detect a broad dangerous state and issue a warning, the individual severity of the contribution index, the accident risk and the accident risk prediction model information are stored in the combination table 168.

[0399] (24) The operation assistance method described in (21) above is characterized in that: in the twelfth step, information obtained based on the severity level (167) of the individual and a warning containing a response plan (173) for avoiding the dangerous phenomenon are displayed.

[0400] According to the above structure, in the area 1010 displaying the accident risk alarm, for example, in addition to the warning text 1011 indicating the increased accident risk, information 1012 obtained based on the personal severity of the bio-indicator data 164-2 of the contribution index is also displayed. This not only notifies the driver of the dangerous state, but also allows the driver to be prompted with a warning in a form that is understandable to the driver based on the personal severity of the contribution index, such as "this is a state that only occurs once in half a year for you." This can help to increase the driver's acceptance of the warning.

[0401] (25) The operation assist method described in (19) above is characterized in that it also includes a thirteenth step in which the computer inputs the first biometric data (164-1) into the bio-state estimation model (69) to calculate the first personal severity (67A); in the third step, the first biometric data (164-1) pre-calculated from the driver's first biosensor data (163) when collecting the second vehicle-mounted sensor data (161-2), the accident risk estimation data (165) and the first personal severity (167) are used as input to generate an accident risk prediction model (171) for predicting the accident risk (166) after a specified time; in the eighth step, the second biometric data (164-2) is input into the bio-state estimation model (169) to calculate the second personal severity (167B); and the second biometric data (164-2) and the second personal severity (167B) are input into the accident risk prediction model (171) to predict the accident risk (166).

[0402] According to the above structure, the operation assistance server 1 can not only determine the warning presentation, but also estimate the accident risk by considering the individual differences of the biometric data 164 when estimating the accident risk using the biometric data 164, thereby achieving the effect of predicting the accident risk with higher accuracy.

[0403] In addition, the present invention is not limited to the above-mentioned embodiments, and includes various modified examples. For example, the above-mentioned embodiments are described in detail in order to explain the present invention in an easy-to-understand manner, and are not limited to all structures that must be described. In addition, a part of the structure of a certain embodiment can be replaced with the structure of other embodiments, and the structure of other embodiments can be added to the structure of a certain embodiment. In addition, for a part of the structure of each embodiment, the addition, deletion, or replacement of other structures can be applied alone or in combination.

[0404] In addition, for each of the above structures, functions, processing parts, and processing units, for example, a part or all of them can be implemented by hardware by designing in an integrated circuit. In addition, each of the above structures and functions can also be implemented by software by interpreting and executing programs that implement each function by a processor. Information such as programs, tables, and files that implement each function can be stored in a recording device such as a memory, a hard disk, an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

[0405] In addition, the control lines and information lines are shown as necessary for explanation, and not all the control lines and information lines on the product are necessarily shown. In fact, it can be considered that almost all structures are connected to each other.

Claims

1. A method for assisting the operation of a vehicle by a computer having a processor and a memory, characterized in that: include: In a first step, the computer uses first vehicle-mounted sensor data representing the driving state of the vehicle collected in the past and risk occurrence data in which information on the occurrence of a risk is pre-set from the first vehicle-mounted sensor data as input, and generates an accident risk definition model for estimating the probability of the occurrence of the risk as the accident risk through machine learning; In a second step, the computer inputs second vehicle-mounted sensor data representing the driving state of the vehicle collected in the past into the accident risk definition model, and estimates the probability of the danger occurring to generate accident risk estimation data; In a third step, the computer uses the first biometric data pre-calculated from the first biosensor data of the driver when collecting the second vehicle-mounted sensor data and the accident risk estimation data as input, and generates an accident risk prediction model for predicting the accident risk after a specified time through machine learning; In a fourth step, the computer acquires second biometric sensor data of a driver driving a vehicle, and calculates second biometric indicator data representing a state of the driver from the second biometric sensor data; and In a fifth step, the computer inputs the second biological indicator data into the accident risk prediction model to predict the accident risk after the specified time. The third step also includes a personal severity calculation step, in which a biological state estimation model is generated by inputting the pre-calculated third biosensor data to calculate the personal severity; The operation auxiliary method also includes: In a sixth step, the computer acquires the second biometric sensor data of the driver who is driving from the biometric sensor of the vehicle, and calculates second biometric indicator data representing the state of the driver from the second biometric sensor data; In a seventh step, the computer inputs the second biometric data into the accident risk prediction model to predict the accident risk, and inputs the second biometric data into the biological state estimation model to calculate the severity of the individual; and In an eighth step, the computer determines whether to issue a warning based on the predicted accident risk and the calculated severity level of the individual.

2. The operation assisting method according to claim 1, characterized in that: Also includes: In a ninth step, when the computer determines that the warning is to be issued, the computer extracts the accident risk prediction model, the accident risk, and a contribution index including the types of biological data that contribute to the prediction of the accident risk; and In a tenth step, the computer generates a response plan for avoiding dangerous phenomena based on the individual's severity level regarding the contribution index.

3. The operation assisting method according to claim 2, characterized in that: The method further includes an eleventh step in which the computer generates and outputs a warning including the countermeasure.

4. The operation assisting method according to claim 1, characterized in that: In the step of calculating the individual's severity level in the third step, the first bio-indicator data and at least the learning information data of the object period for generating the biological state estimation model in the first bio-indicator data are taken as input to generate the biological state estimation model for calculating the individual's severity level.

5. The operation assisting method according to claim 2, characterized in that: In the seventh step, a combined table is generated that stores the accident risk, the contribution index, and the personal severity level with respect to the contribution index in association with each other.

6. The operation assisting method according to claim 3, characterized in that: In the eleventh step, information based on the severity of the individual's condition and a warning including the coping plan are displayed.

7. The operation assisting method according to claim 1, characterized in that: The method further comprises a twelfth step in which the computer inputs the first biological indicator data into the biological state estimation model to calculate the severity of the first person. In the third step, the first biometric data pre-calculated from the driver's first biosensor data when the second vehicle-mounted sensor data is collected, the accident risk estimation data and the first personal severity level are used as inputs to generate an accident risk prediction model for predicting the accident risk after a specified time; In the seventh step, the second bio-indicator data is input into the biological state estimation model to calculate the second personal severity level, and the second bio-indicator data and the second personal severity level are input into the accident risk prediction model to predict the accident risk.

8. An operation assisting system comprising a server and a vehicle for assisting the operation of the vehicle, wherein the server comprises a processor and a memory, and the vehicle comprises an onboard sensor for detecting a driving state and a biosensor for detecting biosensor data of a driver, The server comprises: an accident risk definition generating unit, which takes as input first vehicle-mounted sensor data representing the driving state of the vehicle collected in the past and danger occurrence data in which information on the occurrence of danger is pre-set from the first vehicle-mounted sensor data, generates an accident risk definition model for estimating the probability of danger occurrence as the accident risk through machine learning, inputs second vehicle-mounted sensor data representing the driving state of the vehicle collected in the past into the accident risk definition model, estimates the probability of danger occurrence and generates accident risk estimation data; an accident risk prediction model generating unit which uses as input first biometric data pre-calculated from first biosensor data obtained as biosensor data of the driver when collecting the second vehicle-mounted sensor data and the accident risk estimation data, and generates an accident risk prediction model for predicting the accident risk after a predetermined time by machine learning; a biometric indicator calculation unit that acquires second biometric sensor data of the driver who is driving from the biometric sensor of the vehicle, and calculates second biometric indicator data indicating the state of the driver from the second biometric sensor data; an accident risk prediction unit, which inputs the second biometric data into the accident risk prediction model to predict the accident risk after the specified time; a biological state estimation unit for generating a biological state estimation model for calculating the severity of the person by inputting the third biological indicator data calculated in advance; a risk prediction unit, which inputs the second biological indicator data into the accident risk prediction model to predict the accident risk, and inputs the second biological indicator data into the biological state estimation model to calculate the severity of the individual; and A warning presentation unit determines whether to issue a warning based on the predicted accident risk and the calculated severity level of the individual.

9. The operation assistance system according to claim 8, characterized in that: When it is determined that the warning is to be issued, the warning presentation unit generates a response plan for avoiding dangerous phenomena based on the accident risk prediction model, the accident risk, the type of biological data that contributes to the prediction of the accident risk, i.e., the contribution index, and the severity of the individual with respect to the contribution index.

10. The operation assistance system according to claim 9, characterized in that: The warning presentation unit generates and outputs a warning including the countermeasure.

11. The operation assistance system according to claim 8, characterized in that: The biological state estimation unit generates the biological state estimation model for calculating the severity level of the person taking as input the first biological indicator data and learning information data of a target period for generating the biological state estimation model at least including the first biological indicator data.

12. The operation assistance system according to claim 9, characterized in that: The risk prediction unit generates a combination table that stores the accident risk, the contribution index, and the personal severity level with respect to the contribution index in association with each other.

13. The operation assistance system according to claim 8, characterized in that: The risk prediction unit displays information based on the severity of the individual and a warning including a countermeasure for avoiding a dangerous phenomenon.

14. The operation assistance system according to claim 8, characterized in that: The risk prediction unit inputs the first biological indicator data into the biological state estimation model to calculate the first person's severity level. The accident risk prediction generating unit uses the first biological indicator data pre-calculated from the first biological data of the driver when collecting the second vehicle-mounted sensor data, the accident risk estimation data and the first personal severity as input to generate an accident risk prediction model for predicting the accident risk after a specified time. The risk prediction unit inputs the second biological indicator data into the biological state estimation model to calculate a second personal severity level, and inputs the second biological indicator data and the second personal severity level into the accident risk prediction model to predict the accident risk.

15. An operation assistance server having a processor and a memory for assisting the operation of a vehicle, characterized in that: include: an accident risk definition model generating unit, which takes as input first vehicle-mounted sensor data representing the driving state of the vehicle collected in the past and danger occurrence data in which information on the occurrence of danger is pre-set from the first vehicle-mounted sensor data, and generates an accident risk definition model for estimating the probability of danger occurrence as the accident risk through machine learning, and inputs second vehicle-mounted sensor data representing the driving state of the vehicle collected in the past into the accident risk definition model, and estimates the probability of danger occurrence to generate accident risk estimation data; an accident risk prediction model generating unit, which uses first biometric data pre-calculated from the driver's first biosensor data when the second vehicle-mounted sensor data is collected and the accident risk estimation data as input, and generates an accident risk prediction model for predicting the accident risk after a specified time through machine learning; a biometric indicator calculation unit that acquires second biometric sensor data of the driver driving the vehicle and calculates second biometric indicator data indicating a state of the driver from the second biometric sensor data; an accident risk prediction unit, which inputs the second biometric data into the accident risk prediction model to predict the accident risk after the specified time; a biological state estimation unit for generating a biological state estimation model for calculating the severity of the person by inputting the third biological indicator data calculated in advance; a risk prediction unit, which inputs the second biological indicator data into the accident risk prediction model to predict the accident risk, and inputs the second biological indicator data into the biological state estimation model to calculate the severity of the individual; and A warning presentation unit determines whether to issue a warning based on the predicted accident risk and the calculated severity level of the individual.

16. The operation auxiliary server according to claim 15, characterized in that: When it is determined that the warning is to be issued, the warning presentation unit generates a response plan for avoiding dangerous phenomena based on the accident risk prediction model, the accident risk, the type of biological data contributing to the prediction of the accident risk, i.e., the contribution index, and the individual's severity with respect to the contribution index.

17. The operation auxiliary server according to claim 16, characterized in that: The warning presentation unit generates and outputs a warning including the countermeasure.

18. The operation auxiliary server according to claim 15, characterized in that: The biological state estimation unit generates the biological state estimation model for calculating the severity level of the person taking as input the first biological indicator data and learning information data of a target period for generating the biological state estimation model at least including the first biological indicator data.

19. The operation auxiliary server according to claim 16, characterized in that: The risk prediction unit generates a combination table that stores the accident risk, the contribution index, and the personal severity level with respect to the contribution index in association with each other.

20. The operation auxiliary server according to claim 15, characterized in that: The risk prediction unit displays information based on the severity of the individual and a warning including a countermeasure for avoiding a dangerous phenomenon.

Images