Road surface condition prediction device and system

Abstract

This technology provides a method for predicting road surface conditions, including the state of the water film and slipperiness. [Solution] The road surface condition prediction device comprises a weather information acquisition unit and a prediction unit. The weather information acquisition unit acquires current weather information relating to at least one of precipitation and river flood information in a specific area. Based on the acquired weather information, the prediction unit predicts at least one of the water film state and slipperiness on the road surface in the specific area at a point in time from the present. The prediction unit uses a model that represents the correspondence between the estimated result of at least one of the water film state and slipperiness on the road surface, which was previously estimated based on vehicle data indicating the driving conditions of vehicles that have traveled on the road surface, and past weather information in the specific area at the time the vehicle data was acquired, to predict at least one of the water film state and slipperiness on the road surface in the specific area.

Description

Technical Field

[0001] The present invention relates to a road surface condition prediction device that predicts road surface conditions and a system including the same.

Background Art

[0002] Conventionally, a technique has been proposed for estimating the ponding condition of a road surface on which a vehicle is traveling based on vehicle data indicating the running state of the vehicle (for example, Patent Document 1). In addition, a technique has also been proposed for estimating the friction coefficient μ of a road surface on which a vehicle is traveling based on vehicle data indicating the running state of the vehicle (for example, Patent Document 2). The friction coefficient μ of the road surface indicates the slipperiness of the road surface, and has a lower value on a snow-packed road or a frozen road surface.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] As disclosed in Patent Documents 1 and 2, the road surface conditions estimated for each vehicle are used to issue a warning to the driver of the vehicle in real time or to guide a detour around a ponding point. However, these estimations are made based on the vehicle data actually acquired, and a technique for predicting road surface conditions including the state of the water film and the slipperiness has not yet been proposed.

[0005] An object of the present invention is to provide a technique for predicting road surface conditions including the state of the water film and the slipperiness.

Means for Solving the Problems

[0006] A road surface condition prediction device according to a first aspect of the present invention comprises a weather information acquisition unit and a prediction unit. The weather information acquisition unit acquires current weather information relating to at least one of precipitation and river flood information in a specific area. Based on the acquired weather information, the prediction unit predicts at least one of the water film state and slipperiness on the road surface in the specific area at a point in time from the present. The prediction unit is characterized by using a model that represents the correspondence between the estimation result of at least one of the water film state and slipperiness on the road surface, which was previously estimated based on vehicle data indicating the driving conditions of vehicles that have traveled on the road surface, and past weather information in the specific area at the time the vehicle data was acquired, to predict at least one of the water film state and slipperiness on the road surface in the specific area.

[0007] A road surface condition prediction device according to a second aspect of the present invention is a road surface condition prediction device according to a first aspect, wherein the model is a machine learning model that has learned the correspondence between a risk label indicating the susceptibility of the road surface to flooding or freezing, past weather information, and the estimation result; a regression model that uses the risk label and the past weather information as explanatory variables and the estimation result as the dependent variable; or a model that takes the risk label and the acquired weather information as input and outputs the estimation result, wherein the parameters of the current model are successively updated based on the newly acquired estimation result, and the prediction unit predicts at least one of the state of the water film on the road surface and the slipperiness based on the risk label and the current weather information.

[0008] A road surface condition prediction device according to a third aspect of the present invention is a road surface condition prediction device according to the first or second aspect, further comprising a route information creation unit that creates route information for guiding users on a route that bypasses road surfaces where water film formation is predicted or road surfaces where freezing is predicted, based on the prediction results from the prediction unit.

[0009] A road surface condition prediction device according to the fourth aspect of the present invention is a road surface condition prediction device according to any of the first or third aspects, characterized in that when the vehicle travels in accordance with the route information guidance created by the route information creation unit, the insurance premium of the vehicle user is reduced or points are awarded.

[0010] A road surface condition prediction device according to the fifth aspect of the present invention is a road surface condition prediction device according to any of the first or fourth aspects, wherein the vehicle is a rental car or a car-sharing vehicle, and when the vehicle travels in accordance with the route information guidance created by the route information creation unit, points are awarded to the user of the vehicle.

[0011] A road surface condition prediction system according to the sixth aspect of the present invention is characterized by comprising a road surface condition prediction device according to any of the first or fifth aspects, and a distribution device that distributes data of at least one of the prediction results and the route information to an external computer. [Effects of the Invention]

[0012] According to the present invention, a technology is provided for predicting road surface conditions, including the state of the water film and slipperiness. [Brief explanation of the drawing]

[0013] [Figure 1] A block diagram showing the overall configuration of the road surface condition prediction system. [Figure 2] A flowchart illustrating the flow of system operation. [Modes for carrying out the invention]

[0014] Hereinafter, with reference to the drawings, a road surface condition prediction device and a road surface condition prediction system equipped therewith according to an embodiment of the present invention will be described.

[0015] <1. Overall System Configuration> Figure 1 is a block diagram showing the overall configuration of the road surface condition prediction system 1 (hereinafter also simply referred to as "System 1") according to this embodiment. System 1 comprises an on-board CPU 20 mounted on a vehicle 2, a road surface condition prediction device 3 (hereinafter simply referred to as "prediction device 3"), at least one server device 4 constituting a network, and at least one computer 5. The network is a general-purpose wide-area communication network such as the Internet, and the on-board CPU 20, prediction device 3, and computer 5 can communicate data with each other via the server device 4. The server device 4 is further connected to a server device (not shown) that distributes weather information. Here, the weather information includes at least one of observed precipitation (including converted amounts due to snowfall) and river flood information for one or more areas. In addition to the above, the weather information may also include predicted precipitation, observed temperature, and predicted temperature for the same area.

[0016] <2. Vehicles> Vehicle 2 includes an on-board CPU 20, a communication unit 21, a display unit 22, a location information acquisition unit 23, a sensor 24, and a storage unit 25 mounted on the vehicle body. Vehicle 2 may be equipped with an engine (internal combustion engine) as a power source, or with an electric motor, or with both an engine and an electric motor.

[0017] The in-vehicle CPU 20 is a known CPU (Central Processing Unit) in terms of hardware, and has RAM (Random Access Memory) 202 and ROM (Read Only Memory) 203. The ROM 203 of some in-vehicle CPUs 20 stores a program 204 for performing estimation processing to estimate at least one of the water film state and slipperiness of the road surface on which the vehicle 2 is traveling. By reading and executing the program 204, the in-vehicle CPU 20 virtually operates as a vehicle data acquisition unit 200 and an estimation unit 201. The vehicle data acquisition unit 200 acquires sensing data output from the sensor 24 (described later) at predetermined intervals as vehicle data indicating the driving state of the vehicle 2. The estimation unit 201 estimates at least one of the water film state and slipperiness of the road surface on which the vehicle 2 is traveling based at least on the above vehicle data, and creates an estimation result. The created estimation result is used for road surface condition prediction by the prediction device 3 (described later).

[0018] Here, the "state of the water film" on the road surface includes the "presence or absence of a water film (flooding)" on the road surface or the "thickness of the water film" on the road surface. Furthermore, the "slipperiness" of the road surface includes the coefficient of friction of the road surface. A specific method for estimating the "presence or absence of a water film (flooding)" in the in-vehicle CPU 20 is, for example, the method disclosed in Patent Document 1. A specific method for estimating the "thickness of the water film" in the in-vehicle CPU 20 is, for example, the method described in Japanese Patent Application Publication No. 2024-077834, which is an application filed by the present applicant. A specific method for estimating the "slipperiness" in the in-vehicle CPU 20 is, for example, the method disclosed in Patent Document 2.

[0019] The RAM 202 is used as appropriate for calculations by the in-vehicle CPU 20 together with the memory unit 25. The memory unit 25 is a non-volatile, rewritable memory device, and can be configured as flash memory or the like. Map information may be stored in the memory unit 25.

[0020] The communication unit 21 is a known communication module that performs data communication with an external computer such as the server device 4.

[0021] The display unit 22 is a display device for displaying various types of information to the passengers (mainly the driver) of the vehicle 2, and can be realized in any form such as a liquid crystal display element, a liquid crystal monitor, a plasma display, an organic EL display, etc. The mounting position of the display unit 22 is not particularly limited.

[0022] The position information acquisition unit 23 receives a positioning signal from a satellite positioning system (Global Navigation Satellite System: GNSS) such as GPS, and acquires position information indicating the position where the vehicle 2 is currently traveling. The acquired position information is associated with the vehicle data acquired by the vehicle data acquisition unit 200, the estimation result created by the estimation unit 201, and the time information by the in-vehicle CPU 20. The time information can be information indicating the acquisition time of the vehicle data. The estimation result associated with the position information and the time information may be transmitted by the communication unit 31 to the server device 4 described later in a timely manner and stored in the storage unit 42 as estimation result data. The stored estimation result data is further associated with the weather information corresponding to the position information and the time information included therein, and the risk label described later, and is utilized as a learning data set for learning or re-learning the model 320 described later.

[0023] Sensor 24 outputs sensing data indicating the driving state of vehicle 2. Examples of sensors 24 include a wheel speed sensor for detecting the rotational speed of each wheel of vehicle 2, a wheel torque sensor for detecting wheel torque, an acceleration sensor for detecting the longitudinal acceleration of vehicle 2, an accelerator sensor for detecting the amount the accelerator pedal is pressed, and a brake sensor for detecting the amount the brake pedal is pressed or the brakes are ON / OFF. For example, if the estimation unit 201 is configured to estimate the presence or absence of flooding (presence or absence of a water film) on the road surface on which vehicle 2 is driving using the method disclosed in Patent Document 1, then sensor 24 includes a wheel speed sensor, a wheel torque sensor, an accelerator sensor, and a brake sensor for detecting the amount the brake pedal is pressed. Alternatively, if the estimation unit 201 is configured to estimate the thickness of the water film using the method disclosed in Japanese Patent Application Publication No. 2024-077834, then sensor 24 includes a wheel speed sensor, a wheel torque sensor, an acceleration sensor, and a brake sensor for detecting the brakes being ON / OFF. For example, if the estimation unit 201 is configured to estimate the slipperiness (coefficient of friction of the road surface) of the road surface on which the vehicle 2 travels in the manner disclosed in Patent Document 2, then the sensor 24 includes a wheel speed sensor.

[0024] The various sensors exemplified above are not limited in their detection method or mounting location on the vehicle 2, as long as they output sensing data indicating the driving state of the vehicle 2, such as the rotational speed of each wheel of the vehicle 2, the longitudinal acceleration of the vehicle 2, wheel torque, accelerator pedal depression, brake pedal depression, or brake ON / OFF status. Furthermore, the longitudinal acceleration of the vehicle 2 may be derived, for example, from the rotational speed of each wheel of the vehicle 2 or from the change in the vehicle 2's position information over time, and the acceleration sensor may be omitted. Similarly, the wheel torque may be derived from a torque signal representing the torque of the drive source, and the wheel torque sensor may be omitted.

[0025] <3. Prediction device> The prediction device 3 can be configured as a general-purpose computer in terms of hardware. The prediction device 3 has a CPU 30, a communication unit 31, and a storage unit 32. The CPU 30 has RAM 302 and ROM 303. The ROM 303 stores a program 304 for executing the road surface condition prediction processing described later. The program 304 is installed in the ROM 303 from a non-volatile storage medium readable by a computer, such as a CD-ROM, or via a network. By reading and executing the program 304, the CPU 30 virtually operates as a weather information acquisition unit 305, a prediction unit 300, and a route information creation unit 301.

[0026] The weather information acquisition unit 305 acquires weather information for the area to which the road surface to be estimated belongs from a server device that distributes weather information, via the server device 4. The prediction unit 300 uses the model 320, described later, to predict at least one of the water film condition and slipperiness (coefficient of friction of the road surface) of the road surface within a specific area. The route information creation unit 301 creates route information that guides a route that bypasses the road surface where water film formation or freezing is predicted, based on the prediction results created by the prediction unit 300 and the geographic information contained in the geographic data 321, described later. The route information may include information on alternative means of transportation (railway, etc.) that are available at the time when water film formation or freezing is predicted.

[0027] The communication unit 31 is a well-known communication module and performs data communication with an external computer such as the server device 4.

[0028] The memory unit 32 is a non-volatile, rewritable memory device, and can be configured as flash memory or the like. The memory unit 32 stores parameters that define the road surface condition prediction model 320 (hereinafter simply referred to as "model 320"). In this embodiment, model 320 is constructed as a single, unified model independent of the area, but it may also be constructed as multiple models defined by different parameters for each area. The memory unit 32 also stores geographic risk data 321 (hereinafter simply referred to as "geographic data 321"). Geographic data 321 is, for example, data that associates geographic information with the likelihood of road surface flooding or freezing. Geographic information can include general road map information and elevation information. In this geographic information, roads are divided into predetermined grids (for example, 30m squares), and each divided grid is assigned identification information to identify it. The susceptibility of road surfaces to flooding or slipperiness can be represented by risk labels (numerical indicators within a predetermined range corresponding to the susceptibility of road surfaces to flooding or slipperiness) assigned to each grid, for example, based on information from flood hazard maps issued by administrative agencies. Alternatively, the susceptibility of road surfaces to flooding or slipperiness can be represented by risk labels (numerical values) that distinguish roads prone to flooding, such as underpasses.

[0029] Model 320 is a model that has learned the correspondence between the risk label of the road surface to be predicted, past weather information for a specific area including the road surface, and the estimated road surface conditions created by the onboard CPU 20 of a vehicle 2 that has previously traveled on the road surface. When data corresponding to the risk label of the road surface to be predicted and the current weather information is input, it outputs data corresponding to at least one of the water film state and slipperiness of the road surface at a point in time from the present onward. In this embodiment, the input risk label is a label indicating the susceptibility of the road surface to flooding. In this embodiment, the input weather information is the latest observed precipitation and the latest river flood information for a specific area. The observed precipitation is the amount of precipitation observed at one or more observation points within the specific area. The river flood information is a numerical value corresponding to the warning level announced by the Japan Meteorological Agency for rivers within the specific area. This value can range from 0 to 5, for example: 0 if there are no rivers in a particular area, 1 if an early warning is issued by the Japan Meteorological Agency, 2 if a flood warning is issued, 3 if a flood alert is issued, 4 if a flood danger is issued, and 5 if a flood has occurred.

[0030] In this embodiment, the data output by Model 320 corresponds to the predicted thickness of the water film on the road surface in a specific area, and will be a value greater than or equal to 0. If the output value is 0, it is predicted that there is no water film on the road surface (no flooding). If the current (latest) observed precipitation is input to Model 320, the data output by Model 320 can be the predicted thickness of the water film on the road surface for a period of approximately 15 minutes from the present. If, for example, the predicted precipitation for one hour from now is input to Model 320, the data output by Model 320 can be the predicted thickness of the water film on the road surface for a period of approximately one hour from the present.

[0031] Model 320 can be, for example, a pre-trained machine learning model that, when inputting data corresponding to the risk label of the road surface to be predicted and weather information for a specific area to which the road surface belongs, using the above-mentioned training dataset, outputs data corresponding to the thickness of the water film on the road surface in that specific area. The structure of the machine learning model is not particularly limited, and known model structures such as neural networks and convolutional neural networks can be adopted. Furthermore, the training method for Model 320 (the method for tuning the parameters that define Model 320) is also not particularly limited, and known methods can be adopted.

[0032] <4. Server Equipment> The server device 4 can be configured as a general-purpose server computer in terms of hardware. The server device 4 has a CPU 40, a communication unit 41, and a storage unit 42. The CPU 40 has RAM 401 and ROM 402. A program for controlling the operation of the server device 4 is installed in ROM 402. The CPU 40 virtually operates as a distribution unit 400 by reading and executing this program. The operation of the distribution unit 400 will be described later. The storage unit 42 is a non-volatile, rewritable memory device, and can be configured as flash memory or the like in terms of hardware.

[0033] The communication unit 41 is a known communication module that communicates data with an external computer. The server device 4 may collect estimated result data transmitted from the vehicle 2's communication unit 21 via the communication unit 41. The server device 4 may also collect weather information corresponding to the location and time from an external server device based on the location and time information included in the estimated result data. The CPU 40 may create a dataset by associating the thus collected estimated result data and weather information with each other and store it in the storage unit 42. The numerous datasets thus stored can be used to retrain the model 320. Risk labels corresponding to geographic data 321 may be further associated with these numerous datasets based on location information. From this perspective, the server device 4 can also be described as a training dataset creation device for training a model that, when input, takes data corresponding to current weather information regarding at least one of precipitation and river flood information in a specific area, and outputs data corresponding to at least one of the water film state and slipperiness on the road surface in the specific area at a point in time from the present.

[0034] <5. Computers> Computer 5 is not particularly limited as long as it is an information processing terminal that is communicatively connected to the server device 4, and can be configured as a desktop PC, laptop PC, smartphone, tablet, or (excluding smartphones) mobile phone. Computer 5 may also include an in-vehicle computer such as an in-vehicle CPU 20. In the processing described later, computer 5 receives at least one of the prediction results and route information created by the prediction device 3.

[0035] Computer 5 includes a display unit 50 for displaying at least one of the received prediction results and route information. The display unit 50 can be implemented in any form, such as a liquid crystal display element, liquid crystal monitor, plasma display, organic EL display, etc. When computer 5 is a general-purpose desktop PC, laptop PC, smartphone, tablet, or (excluding smartphones) mobile phone, computer 5 may acquire the above route information in a timely manner based on the user's schedule (scheduled time and destination, etc.) that has been registered in computer 5, and display a detour route that bypasses the road surface where water film formation or freezing is predicted, or display information on alternative means of transportation to the vehicle 2 on the display unit 50. When computer 5 is an in-vehicle computer, computer 5 may, for example, receive at least one of the prediction results and route information for the area in which vehicle 2 is currently traveling and the surrounding area, and display a warning on the display unit 22 that water film formation or freezing is predicted in the vicinity, or display a detour route for the above road surface. Furthermore, the in-vehicle computer, which is computer 5, does not necessarily have to have a function (estimation unit 201) for estimating the state and slipperiness of the water film.

[0036] <6. System Operation> Figure 2 is a flowchart illustrating the flow of road surface condition prediction and prediction result distribution performed by System 1. The process shown in Figure 2 begins when the prediction device 3 receives a request command requesting prediction results for a specific road surface corresponding to one or more grids. The request command may be sent to the prediction device 3 from, for example, a computer 5 via a server device 4.

[0037] In step S1, the weather information acquisition unit 305 of the prediction device 3 identifies the road surface to be predicted. The road surface to be predicted is the specific road surface specified in the request command. Based on the request command, the weather information acquisition unit 305 identifies one or more grids of the corresponding road surface in the geographic data 321.

[0038] In step S2, the weather information acquisition unit 305 reads one or more risk labels assigned to the road surface of the grid identified in step S1 from the geographic data 321.

[0039] In step S3, the weather information acquisition unit 305 acquires weather information for the area to which the road surface identified in step S1 belongs from a server device that distributes weather information. As described above, in this embodiment, the weather information is the latest observed precipitation and river flood information for the area.

[0040] In step S4, the prediction unit 300 inputs the risk labels read in step S2 and the meteorological information acquired in step S3 into the model 320 and derives the output from the model 230. As described above, the output from the model 230 is the predicted water film thickness for the road surface corresponding to one grid. If there are multiple grids identified in step S1, the prediction unit 300 can sequentially input the risk labels and meteorological information corresponding to each grid into the model 320 and sequentially derive the output from the model 230.

[0041] In step S5, the prediction unit 300 determines, based on the output derived in step S4, whether or not flooding requiring a detour is predicted to occur on the road surface. This determination can be made, for example, based on a predetermined threshold stored in the memory unit 32 or ROM 303. If at least one of the outputs exceeds the threshold, it is predicted that flooding requiring a detour will occur (YES). If the answer in step S5 is YES, step S6 is executed. On the other hand, if all of the outputs are below the threshold and it is predicted that flooding requiring a detour will not occur (NO), step S6 is omitted and step S7 is executed.

[0042] In step S6, the route information creation unit 301 creates route information. The route information creation unit 301 identifies, for example, available alternative means of transportation in the area closest to the road surface identified in step S1. Alternatively, for example, the route information creation unit 301 creates a detour route that bypasses the road surface corresponding to one or more grids where flooding is predicted to occur and which should be bypassed in step S6. At this time, the route information creation unit 301 may perform steps S2 to S5 for grids adjacent to the above one or more grids and confirm that flooding is not predicted to occur on the road surfaces corresponding to these grids.

[0043] In step S7, the prediction unit 300 transmits at least one of the predicted road surface conditions and the route information created in step S6 to the server device 4 via the communication unit 31. The predicted road surface conditions may include at least one of the predicted values ​​of the water film thickness on each road surface derived in step S4 and the presence or absence of flooding that requires detour, as determined in step S5. The prediction processing in the prediction device 3 then ends.

[0044] In step S41, the server device 4 receives at least one of the predicted road surface conditions and route information. Next, in step S42, it distributes this to the computer 5. The server device 4 may transmit at least one of the predicted road surface conditions and route information only to the computer 5 that sent the request command. Alternatively, the server device 4 may create a webpage on the Internet containing at least one of the predicted road surface conditions and route information, making it accessible to an unspecified computer 5. Alternatively, the server device 4 may transmit at least one of the predicted road surface conditions and route information to a computer 5 that has previously registered its desire to receive at least one of the predicted road surface conditions and route information for a specific area. The server device 4 is an example of a distribution device of the present invention.

[0045] If the server device 4 is configured to transmit the above route information to a specific computer 5, the storage unit 32 of the prediction device 3 may be configured to further store incentives for the user of vehicle 2 (typically the driver or owner of vehicle 2). In addition, the CPU 30 may be configured to further operate as an incentive identification unit. Incentives are, for example, parameters that work to reduce the insurance premium when calculating the insurance premium for vehicle 2's automobile insurance, or points that can be exchanged for cash, goods, or gift certificates. For example, if vehicle 2 travels a detour route in accordance with the guidance of the above route information, the computer 5 creates a detour route travel record indicating that vehicle 2 actually traveled a detour route based on positioning signals from the satellite positioning system, and transmits this to the prediction device 3. Upon receiving the detour route travel record, the CPU 30 of the prediction device 3 identifies the incentive for the user of vehicle 2, associates it with that user, and stores the latest incentive in the storage unit 32. The method for identifying incentives from detour route driving records is not particularly limited, but possible methods include updating parameters so that the insurance premium is reduced by a predetermined percentage for each transmission of a detour route driving record, or awarding points of a predetermined value. If vehicle 2 is a rental car or car-sharing vehicle, and the owner of vehicle 2 is a rental car company or car-sharing company, the points can be awarded to the driver of vehicle 2, who is a customer of that company. As can be seen from the above, users of the prediction device 3 may include automobile insurance companies, rental car companies, car-sharing companies, or information providers who provide information to these companies for determining the above incentives.

[0046] <7. Features> (1) According to the system 1 or prediction device 3 of the above embodiment, the road surface conditions of the road surface to be predicted can be predicted even if the vehicle 2 is not currently traveling on that road surface. This makes it possible to predict road surface conditions over a wide area and over a relatively long period of time.

[0047] (2) According to the system 1 or prediction device 3 of the above embodiment, it becomes easier for public agencies that manage roads to determine the priority of locations for on-site inspections to decide whether to close roads. Therefore, more appropriate measures can be taken when flooding occurs, and the adverse effects of flooding can be suppressed.

[0048] (3) According to the system 1 or prediction device 3 of the above embodiment, the driver of vehicle 2 will be able to avoid routes where flooding is likely to occur or routes where freezing is likely to occur. This will reduce the risk of the driver of vehicle 2 being caught in flooding or slipping on icy roads. Furthermore, when vehicle 2 approaches a road surface where flooding or freezing is predicted to occur, and the sensing data of vehicle 2 shows some kind of abnormality, or when the estimation unit 201 of vehicle 2 that has approached the road surface estimates the occurrence of flooding or freezing on that road surface, the computer 5 or the onboard CPU 20 can be configured to automatically send a request signal from vehicle 2 to a rescue team. An example of when the sensing data of vehicle 2 shows some kind of abnormality is when the value of the wheel speed sensor remains at 0 or nearly 0 for a certain period of time in an area where there are no traffic lights and there is no need to stop temporarily. Alternatively, if vehicle 2 approaches a road surface where flooding or freezing is predicted, an external computer, such as a server device 4, can inquire with the occupants of vehicle 2 about the need for rescue. The inquiry may be sent, for example, from an external computer to computer 5 or the onboard CPU 20 and displayed, for example, on display unit 50 or display unit 22, or output from computer 5 or the onboard CPU 20 in the form of a buzzer or voice guidance. The occupants of vehicle 2 can respond to the inquiry by, for example, selecting a button displayed on display unit 50 or display unit 22, or by answering verbally, to indicate whether rescue is needed.

[0049] (4) According to the system 1 of the above embodiment, if vehicle 2 drives while avoiding danger, the user of vehicle 2 is given an incentive. This makes it possible to avoid situations where public resources necessary for rescue and road repair are strained due to people carelessly approaching dangerous places.

[0050] <8. Variation> Although one embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications are possible without departing from the spirit of the invention. For example, the following modifications are possible. Furthermore, the gist of the following modifications can be combined as appropriate.

[0051] (1) Model 320, as a pre-trained machine learning model, may be configured to take meteorological information for a specific area as input and output an estimated result of at least one of the water film state and slipperiness (coefficient of friction) of the target road surface. Model 320 does not have to be a pre-trained machine learning model. Model 320 may be a regression model in which meteorological information for a specific area is used as an explanatory variable and the estimated result of at least one of the water film state and slipperiness (coefficient of friction) of the target road surface is used as an dependent variable, and may be expressed, for example, by a regression equation. In this case, the parameters that define Model 320 are the regression coefficients of the above regression equation. The parameters that define Model 320 may be identified based on a large dataset of past meteorological information and past estimation results. Model 320 may also be a sequential model in which the parameters of the current model are sequentially updated based on newly acquired estimation results, and which takes meteorological information for a specific area as input and outputs an estimated result of at least one of the water film state and slipperiness (coefficient of friction) of the target road surface as input. For example, a Kalman filter or a particle filter can be applied to update the parameters. Such a sequential model allows for predictions that reflect the trends of the most recently acquired data at the time of prediction, while suppressing the computational load on the prediction device 3. However, from the viewpoint of prediction accuracy for at least one of the water film state and slipperiness, it is preferable that the vehicle 2 equipped with an on-board CPU 20 having an estimation unit 201 travels on the road surface to be predicted at a time as close as possible to the time of prediction (for example, 1 hour to 0 minutes before the prediction). The above also applies when predicting the estimation result of at least one of the water film state and slipperiness of the road surface using the regression model or sequential model described above, based on both weather information and risk labels.

[0052] (2) The order in which steps S2 to S3 above are performed may be reversed.

[0053] (3) In addition to or in lieu of the above-mentioned risk labels, the geographic data 321 may include statistical data of water film thickness or friction coefficient estimated by the in-vehicle CPU 20 in the past. The statistical data may include, for example, the maximum value of the estimated water film thickness over a predetermined period (e.g., one year), the maximum difference between the water film thickness estimated at the same estimation period and the road surface of another adjacent grid, the number of times the estimated water film thickness exceeded a certain value over a predetermined period, the minimum value of the estimated friction coefficient over a predetermined period, the maximum difference between the friction coefficient estimated at the same estimation period and the road surface of another adjacent grid, the number of times the estimated friction coefficient fell below a certain value over a predetermined period, and so on. Model 320 may use these statistical data as input (or explanatory variables) in addition to or in lieu of the above-mentioned risk labels.

[0054] (4) When the output from model 320 corresponds to the predicted friction coefficient of the road surface, the prediction unit 300 compares the threshold predetermined in step S5 with the output from model 320 (or the target variable), and can predict road surface freezing if the output from model 320 (or the target variable) is less than or equal to the threshold. The output from model 320 (or the target variable) may include both the predicted thickness of the water film on the road surface and the predicted friction coefficient of the road surface. When the output from model 320 (or the target variable) corresponds to the friction coefficient of the road surface, it is preferable that the weather information includes the temperature of a specific area (observed temperature or predicted temperature).

[0055] (5) Geographic information may include soil classification labels related to drainage, such as sandy soil, clayey soil, and asphalt soil.

[0056] (6) In System 1, the on-board CPU 20 and the computer 5 may be omitted. Also, System 1 may include multiple prediction devices 3 or multiple server devices 4, or the functions of the prediction device 3 and the server device 4 may be mounted on a single device. Furthermore, at least one of the model 320 and geographic data 321 may be stored in the server device 4, and the prediction device 3 may read this information from the server device 4 as appropriate to perform predictions. In addition, the creation of route information may be performed by the server device 4 or each computer 5 instead of the route information creation unit 301. In this case, a dedicated program for creating route information based on the prediction results of the prediction device 3 and geographic information may be installed in the server device 4 or each computer 5.

[0057] (7) The geographic information in geographic data 321 does not have to be divided into a grid. For example, geographic data 321 may be data in which a section ID table in accordance with the road section ID method proposed by the National Institute for Land and Infrastructure Management of the Ministry of Land, Infrastructure, Transport and Tourism and the Japan Digital Road Map Association is associated with risk labels.

[0058] (8) Model 320 may be multiple models with optimized parameters for a predetermined area including multiple grids, or a predetermined area including multiple section IDs. The prediction device 3 may also be configured to automatically acquire the latest weather information without requiring a request command and to automatically output prediction results and route information for a wide area of ​​road surface. [Explanation of Symbols]

[0059] 1 System 2 vehicles 3 Prediction device 4 Server Devices 5 Computers

Claims

1. A weather information acquisition unit that acquires current weather information relating to at least one of precipitation and river flood information for a specific area, Based on the acquired weather information, a prediction unit predicts at least one of the water film condition and slipperiness on the road surface within the specific area at a point in time from the present. Equipped with, The prediction unit is characterized by predicting at least one of the water film state and slipperiness of the road surface within the specific area by using a model that represents the correspondence between the estimation result of at least one of the water film state and slipperiness of the road surface, which was previously estimated based on vehicle data indicating the driving conditions of vehicles that have traveled on the road surface, and past weather information for the specific area at the time the vehicle data was acquired. Road surface condition prediction device.

2. The model is a machine learning model that has learned the correspondence between risk labels indicating the susceptibility of the road surface to flooding or freezing, past weather information, and the estimation results; a regression model that uses the risk labels and past weather information as explanatory variables and the estimation results as the dependent variable; or a model that takes the risk labels and acquired weather information as input and outputs the estimation results, wherein the parameters of the current model are updated sequentially based on the newly acquired estimation results. The prediction unit is characterized by predicting at least one of the water film condition and slipperiness of the road surface based on the risk label and the current weather information. The road surface condition prediction device according to claim 1.

3. A route information creation unit creates route information that guides users on a route that bypasses road surfaces where water film formation or freezing is predicted, based on the prediction results from the aforementioned prediction unit. Features further comprising, A road surface condition prediction device according to claim 1 or 2.

4. The road surface condition prediction device according to claim 3, characterized in that if the vehicle travels in accordance with the route information guidance created by the route information creation unit, the insurance premium of the vehicle user is reduced or points are awarded.

5. The road surface condition prediction device according to claim 3, wherein the vehicle is a rental car or a car-sharing vehicle, and points are awarded to the user of the vehicle when the vehicle travels in accordance with the route information provided by the route information creation unit.

6. A road surface condition prediction device according to claim 3, A distribution device that distributes at least one of the prediction results and the route information to an external computer. A feature comprising: Road surface condition prediction system.

Images