Driving Skills Evaluation System

Abstract

To appropriately evaluate a driving technique at each point.SOLUTION: A driving technique evaluation system executes: processing for acquiring a value of driving operation data of a vehicle of an evaluation object person when the vehicle travels at an evaluation point as a value of evaluation object data R2 (step S20); ideal data calculation processing for calculating a value of ideal driving operation data in traveling at the evaluation point as a value of ideal operation data R3; and processing for evaluating a driving technique based on a comparison result between the value of the evaluation object data R2 and the value of the ideal operation data R3 (step S80). In the ideal data calculation processing, an average value R4A of values of a plurality of pieces of other vehicle operation data R4 is defined as the value of the ideal operation data R3 (step S70) when the number of pieces of the other vehicle operation data R4 is equal to or more than a specified value K1 (step S40: YES), and a value of driving operation data which is output by an automatic operation model DU is defined as the value of the ideal operation data R3 (step S170) when the number of pieces of the other vehicle operation data is less than the specified value K1 (step S40: NO).SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a driving skill evaluation system. [Background technology]

[0002] Patent Document 1 discloses an information management system that manages information on multiple vehicles. The information management system has an information management server. The information management server receives various data from the multiple vehicles while they are traveling. The various data are vehicle position information associated with information on the driving operation of the vehicle at that position. The information on the driving operation is, for example, the timing of accelerator pedal operation. The information management server extracts information on the driving operation at a specific point from the various collected data. The information management server then performs statistical processing on the driving operation for each point. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2011-096061 A Summary of the Invention [Problem to be solved by the invention]

[0004] It is conceivable that data statistically processed for each location as in Patent Document 1 can be used to evaluate the driving skills of a certain subject. In this case, the quality of the driving skills of the subject when the subject's vehicle travels through a certain location is judged based on the statistical values ​​of data from a large number of vehicles. In evaluating driving skills using this type of method, it is necessary to prepare data from a large number of vehicles that can be said to be statistically significant at each location where the driving skills are to be evaluated. However, depending on the location, it may not be possible to collect a sufficient amount of data. [Means for solving the problem]

[0005] A driving skill evaluation system for solving the above problem includes an analysis device and a storage device, and the analysis device executes a data acquisition process for acquiring, as an evaluation target data value, a driving operation data value of a vehicle of an evaluation target person when the vehicle travels through an evaluation point, an ideal data calculation process for calculating, as an ideal operation data value, an ideal driving operation data value when the vehicle travels through the evaluation point, and an evaluation process for evaluating the driving skill of the evaluation target person based on a comparison result between the evaluation target data value and the ideal operation data value, and the storage device executes a data acquisition process for acquiring, as an evaluation target data value, a driving operation data value of a vehicle of an evaluation target person when the vehicle travels through the evaluation point, an ideal data calculation process for calculating, as an ideal operation data value, an ideal driving operation data value when the vehicle travels through the evaluation point, and an evaluation process for evaluating the driving skill of the evaluation target person based on a comparison result between the evaluation target data value and the ideal operation data value. The system stores the values ​​of other vehicle operation data, which are driving operation data, and an autonomous driving model that, by inputting the evaluation point, outputs the value of the driving operation data that should be taken when the vehicle travels through the evaluation point.In the ideal data calculation process, if the number of the other vehicle operation data is equal to or greater than a predetermined value set in advance as an integer greater than or equal to 2, the value of the driving operation data obtained by statistically processing the values ​​of the multiple other vehicle operation data is taken as the ideal operation data, and if the number of the other vehicle operation data is less than the specified value, the value of the driving operation data output by the autonomous driving model is taken as the ideal operation data.

[0006] In the above configuration, when the number of driving operation data of other vehicles that have traveled through the evaluation point is small, the value of the ideal operation data is calculated using the automatic driving model. Therefore, even at a point where the number of driving operation data of other vehicles is small, the driving skill can be appropriately evaluated. On the other hand, in the above configuration, when the number of driving operation data of other vehicles that have traveled through the evaluation point is large, the value of the ideal operation data is calculated based on the driving operation data of multiple other vehicles. Therefore, at a point where the number of driving operation data of other vehicles is large, the driving skill can be evaluated based on the value of the ideal operation data that reflects the traffic environment specific to each point, which cannot be incorporated into the automatic driving model.

[0007] The autonomous driving model may be a trained model based on machine learning. When such a model is used, the processing load on the analysis device can be reduced when the autonomous driving model outputs the ideal operation data. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a schematic diagram of a driving skill evaluation system. [Diagram 2] FIG. 2 is a flowchart showing the procedure of the diagnosis process. [Diagram 3] FIG. 3 is a diagram showing an example of a scene to be evaluated. [Figure 4] FIG. 4 is a flowchart showing a part of the processing contents when the autonomous driving model of the second embodiment is used. [Diagram 5] FIG. 5 is a schematic diagram showing a modified example of the driving skill evaluation system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] First Embodiment Hereinafter, a first embodiment of a driving skill evaluation system will be described with reference to the drawings. <Overall composition> As shown in FIG. 1, the driving skill evaluation system 100 is owned by, for example, a car subscription service provider 300. The subscription service provider 300 manages a plurality of vehicles 20 to provide to users. The driving skill evaluation system 100 manages information on the plurality of vehicles 20. The details of the driving skill evaluation system 100 will be described later.

[0010] Each vehicle 20 has a plurality of information acquisition devices 30 as a common configuration. The information acquisition devices 30 are devices for acquiring various information when the vehicle 20 is traveling. The information acquisition devices 30 are, for example, a GPS receiver 31, a periphery monitoring device 32, an accelerator sensor 33, a brake sensor 34, a vehicle speed sensor 35, and a steering sensor 36. The periphery monitoring device 32 includes an imaging device such as a camera, a radar, and a lidar.

[0011] The GPS receiver 31 receives a signal related to the current position coordinate PN of the vehicle 20 from a GPS satellite. The position coordinate PN is a coordinate value of latitude and longitude. The surroundings monitoring device 32 acquires surroundings monitoring information JM of the vehicle 20, such as a captured image of the surroundings of the vehicle 20 and the presence or absence of obstacles around the vehicle 20. The accelerator sensor 33 detects the operation amount of the accelerator pedal of the vehicle 20 as the accelerator operation amount AP. The brake sensor 34 detects the operation amount of the brake pedal of the vehicle 20 as the brake operation amount BK. The vehicle speed sensor 35 detects the traveling speed of the vehicle 20 as the vehicle speed SP. The steering sensor 36 detects the steering angle of the steering wheel as the steering angle ST. The information acquisition device 30 repeatedly transmits a signal corresponding to the information detected or acquired by itself to the control device 22 described later.

[0012] Each vehicle 20 has a control device 22 that controls various parts of the vehicle 20. The control device 22 has a communication device for connecting to the outside via the external communication line network 12. The control device 22 stores in advance an identification number assigned to the vehicle. The control device 22 generates vehicle transmission data W at a predetermined period. The vehicle transmission data W includes a group of the latest information received from each information acquisition device 30, the identification number of the vehicle, and information on the date and time at the time of generating the vehicle transmission data W. When the control device 22 generates the vehicle transmission data W, it transmits the vehicle transmission data W to the driving skill evaluation system 100. The control device 22 repeatedly generates and transmits the vehicle transmission data W.

[0013] <Details of the driving skill evaluation system> The driving skill evaluation system 100 includes a server 101 , a display device 150 , and an input device 160 .

[0014] The server 101 may be configured as one or more processors that execute various processes according to a computer program (software). The server 101 may be configured as one or more dedicated hardware circuits, such as an application specific integrated circuit (ASIC), that execute at least some of the various processes, or a circuitry including a combination thereof. The processor includes a CPU 102 and memories such as a RAM and a ROM 103. The memory stores program codes or instructions configured to cause the CPU 102 to execute processes. The memory, i.e., computer readable media, includes any available media accessible by a general-purpose or dedicated computer. The server 101 also has a storage device 104, which is an electrically rewritable non-volatile memory. The server 101 also has a communication device 105 for connecting to the outside via the external communication line network 12. The CPU 102, the RAM, the ROM 103, the storage device 104, and the communication device 105 can communicate with each other through an internal bus 106. The CPU 102 and the ROM 103 constitute an analysis device 107. The server 101 repeatedly receives the vehicle transmission data W transmitted by each vehicle 20 via the external communication line network 12.

[0015] The storage device 104 stores the vehicle transmission data W received from each vehicle 20. The manner of storage will be described using one vehicle 20 as an example. The storage device 104 stores the vehicle transmission data W received sequentially from the vehicle 20 over time. The storage device 104 handles this group of data over time as vehicle data DC. In storing the vehicle transmission data W over time, the storage device 104 stores a plurality of elements received collectively as the vehicle transmission data W at a certain timing in a linked state. Therefore, for example, the position coordinate PN when the vehicle 20 travels through a certain point is linked to the vehicle speed SP at that point. The storage device 104 stores in advance an identification number assigned to each vehicle 20. The storage device 104 stores the vehicle data DC of each vehicle 20 for each vehicle 20 in association with the identification number.

[0016] The storage device 104 stores a user number for each user to identify the user who is registered with the subscription service. The storage device 104 stores the user number in association with user information. The user information includes the name, date of birth, and contact information. The user information also includes information about the vehicle 20 that the user is currently using.

[0017] The storage device 104 stores three-dimensional map data DF in advance. The map data DF has a plurality of nodes and a plurality of links. Each node indicates a position coordinate. Each link is defined as a line segment connecting adjacent nodes. Each link indicates a road. The map data DF includes detailed information about the road, such as the width of the road and the sidewalks on the side of the road, as well as information about three-dimensional objects around the road, such as traffic lights, road signs, and buildings.

[0018] The storage device 104 stores an automatic driving model DU in advance. The automatic driving model DU is a group of programs for performing a driving simulation assuming that the vehicle 20 is driven by automatic driving. Automatic driving refers to the autonomous driving of the vehicle 20 by the control device 22 of the vehicle 20 without the driver's operation. The automatic driving model DU outputs the values ​​of various driving parameters to be taken when the vehicle 20 drives the point by inputting the position coordinate PN of a certain point and information on the surrounding environment of the point. Here, a situation is considered in which the driver actually drives the vehicle 20 instead of automatic driving. The above driving parameters are parameters whose values ​​change depending on the driving operation of the driver. Examples of the driving parameters include the accelerator operation amount AP, the brake operation amount BK, the vehicle speed SP, and the steering angle ST. In this specification, a model is a program in which an algorithm outputs a result after going through various processes with a predetermined algorithm for a certain input. In the case of the automatic driving model DU of the first embodiment, the above various processes include a process for performing a driving simulation and a process for calculating premise information required for the driving simulation, which will be described later.

[0019] The display device 150 is connected to the server 101 by wire or wirelessly. The display device 150 is capable of communicating with the server 101. The display device 150 displays various information based on a video signal output by the CPU 102 of the server 101. The display device 150 is, for example, a display.

[0020] The input device 160 is connected to the server 101 by wire or wirelessly. The input device 160 is capable of communicating with the server 101. The input device 160 is used to input information from outside to the server 101. The input device 160 is, for example, a keyboard and a mouse.

[0021] <Diagnosis process overview> The CPU 102 is capable of executing a diagnostic process. The diagnostic process is a process for evaluating the driving skill of an evaluation subject. The CPU 102 executes a program stored in the ROM 103 to realize each step of the diagnostic process.

[0022] In the diagnosis process, the CPU 102 evaluates the driving skill for a predetermined scene. As shown in FIG. 3, in this embodiment, the scene for evaluating the driving skill is a scene where the vehicle 20 passes beside a pedestrian 302 on a road Y where a sidewalk Y2 exists beside a roadway Y1. The CPU 102 evaluates the driving skill in the above scene for a predetermined evaluation point. In this embodiment, a plurality of evaluation points are set. At one of the plurality of evaluation points, a guardrail exists between the roadway Y1 and the sidewalk Y2. At another point, a curbstone exists between the roadway Y1 and the sidewalk Y2 instead of a guardrail. At another point, there is neither a guardrail nor a curbstone between the roadway Y1 and the sidewalk Y2, and a white line exists. In this way, the plurality of evaluation points have different road characteristics. The storage device 104 stores the position coordinates PNQ of the plurality of evaluation points in advance. The CPU 102 evaluates the driving skill when the vehicle 20T of the evaluation subject travels through any of the plurality of evaluation points.

[0023] The CPU 102 can execute a data acquisition process as part of the diagnosis process. In the data acquisition process, the CPU 102 acquires the value of the driving operation data R1 of the vehicle 20T of the evaluation subject when the vehicle 20T of the evaluation subject travels through the evaluation point as the value of the evaluation subject data R2. Hereinafter, the driving operation data R1 of the vehicle 20 is simply referred to as the driving operation data R1. The parameters constituting the driving operation data R1 are the above-mentioned driving parameters whose values ​​change depending on the driving operation of the driver. In this embodiment, the vehicle speed SP is used as the driving operation data R1 that is an index for evaluating the driving skill.

[0024] The CPU 102 can execute an ideal data calculation process as part of the diagnosis process. In the ideal data calculation process, the CPU 102 calculates the value of the ideal driving operation data R1 when the vehicle 20 travels through the evaluation point, that is, the ideal vehicle speed SP, as the value of the ideal operation data R3. The CPU 102 uses the following first method or second method to calculate the value of the ideal operation data R3. The vehicle speed SP when the other vehicle 20H other than the vehicle 20T of the evaluation subject travels through the evaluation point is called the other vehicle operation data R4. In the first method, the CPU 102 sets the vehicle speed SP obtained by statistically processing the values ​​of multiple other vehicle operation data R4 as the value of the ideal operation data R3. Specifically, the CPU 102 sets the average value R4A of the values ​​of multiple other vehicle operation data R4 as the value of the ideal operation data R3. In the second method, the CPU 102 sets the vehicle speed SP output by the automatic driving model DU as the value of the ideal operation data R3.

[0025] CPU 102 can execute a method determination process as part of the diagnosis process. In the method determination process, CPU 102 determines whether the first method or the second method is to be used in the ideal data calculation process. CPU 102 uses the first method when a determination condition is satisfied, and uses the second method when the determination condition is not satisfied. The determination condition is that both of the following two items (a) and (b) are satisfied. (a) The number of the multiple other vehicle operation data R4 is equal to or greater than a specified value K1. (b) The standard deviation R4B of the values ​​of the multiple other vehicle operation data R4 is equal to or smaller than the judgment value K2.

[0026] The specified value K1 is determined in advance, for example, from a statistical point of view, as the minimum value that satisfies the following reliability condition with respect to the number of multiple other vehicle operation data R4. The reliability condition is that a highly significant value is obtained when the average value R4A of the values ​​of the multiple other vehicle operation data R4 is calculated. The specified value K1 is an integer equal to or greater than "2". The judgment value K2 is determined in advance, for example, from a statistical point of view, as the maximum value that satisfies the above reliability condition with respect to the variance of the multiple other vehicle operation data R4. The storage device 104 stores the specified value K1 and the judgment value K2 in advance.

[0027] The CPU 102 can execute an evaluation process as part of the diagnosis process. In the evaluation process, the CPU 102 evaluates the driving skill of the person to be evaluated based on a comparison result between the value of the evaluation target data R2 and the value of the ideal operation data R3. In this embodiment, the CPU 102 compares an evaluation value, which is a value obtained by subtracting the value of the ideal operation data R3 from the value of the evaluation target data R2, with a reference value K3. Then, the CPU 102 evaluates the driving skill of the person to be evaluated according to the magnitude relationship between the evaluation value and the reference value K3. The reference value K3 is previously determined, for example, by an experiment or a simulation, as a maximum value at which the driving manner is considered to be good with respect to the difference between the value of the evaluation target data R2 and the value of the ideal operation data R3. The storage device 104 previously stores the reference value K3.

[0028] <Specific steps of diagnostic processing> A user who is registered with the subscription service can receive an evaluation of driving skills as a privilege. When a user requests that the user's driving skills be evaluated, the business operator 300 evaluates the user's driving skills using the above-mentioned diagnostic process as an evaluation subject. When evaluating driving skills using the diagnostic process, the business operator 300 inputs information required for executing the diagnostic process to the server 101 via the input device 160. As described above, the storage device 104 of the server 101 stores a plurality of evaluation points in advance. In one diagnostic process, the driving skills are evaluated at one of the plurality of evaluation points. The necessary information includes an instruction as to which evaluation point is to be the target. The necessary information also includes the user's name. After inputting the necessary information, the business operator 300 inputs an instruction to execute the diagnostic process to the server 101. In the following explanation of the diagnostic process, the evaluation point refers to a point among the plurality of evaluation points that the business operator 300 inputs to the server 101.

[0029] The CPU 102 of the server 101 starts the diagnostic process in response to an execution instruction from the business operator 300. Although a detailed description will be omitted, when executing each step of the diagnostic process, the CPU 102 appropriately refers to information stored in the storage device 104 or the ROM 103, such as the specified value K1, the judgment value K2, and the map data DF.

[0030] As shown in FIG. 2, when the CPU 102 starts the diagnosis process, it first executes the process of step S10. In step S10, the CPU 102 judges whether or not the vehicle 20T of the person to be evaluated encountered the scene to be evaluated during a certain period from the present time to the past. The certain period is a period during which the driving skill of the person to be evaluated can be evaluated relatively recently, such as one month. As a specific process of step S10, the CPU 102 identifies the user number of the person to be evaluated based on the information input by the business entity 300. Then, the CPU 102 identifies the identification number of the vehicle 20T currently used by the person to be evaluated from the information of the user linked to the user number. After that, the CPU 102 refers to the vehicle data DC corresponding to the identification number of the vehicle 20T of the person to be evaluated among the data stored in the storage device 104. The CPU 102 handles the vehicle data DC as the person to be evaluated vehicle data DCT. After that, the CPU 102 judges whether or not the scene to be evaluated occurred with respect to the person to be evaluated vehicle data DCT from the data from the present time to the past certain period. The CPU 102 performs the following process to determine whether or not there is a scene to be evaluated. That is, the CPU 102 determines whether or not there is a scene that matches the position coordinate PNQ of the evaluation point among the transitions of the position coordinate PN of the vehicle 20. If there is a position coordinate PN that matches the position coordinate PNQ of the evaluation point, the CPU 102 identifies the surroundings monitoring information JM linked to that position coordinate PN. Then, the CPU 102 determines whether or not there is a pedestrian based on the identified surroundings monitoring information JM. If there is no scene to be evaluated (step S10: NO) as a result of analyzing the target person vehicle data DCT, the CPU 102 advances the process to step S200.

[0031] In step S200, CPU 102 displays a message indicating that there is no scene to be evaluated on display device 150. After this, CPU 102 ends the series of steps of the diagnosis process. On the other hand, in step S10, if there is a scene to be evaluated (step S10: YES), the CPU 102 advances the process to step S20. Note that, regarding the process of step S10, if there are multiple scenes to be evaluated when analyzing the subject vehicle data DCT, the CPU 102 subjects the latest case to the process of steps S20 and S160 described below.

[0032] In step S20, the CPU 102 identifies the value of the evaluation target data R2. Specifically, the CPU 102 refers to the subject vehicle data DCT. Then, the CPU 102 identifies the vehicle speed SP associated with the position coordinate PN determined in step S10 to match the position coordinate PNQ of the evaluation point as the value of the evaluation target data R2. After identifying the value of the evaluation target data R2, the CPU 102 advances the process to step S30. Note that the CPU 102 identifying the value of the evaluation target data R2 corresponds to the CPU 102 acquiring the value of the evaluation target data R2. In other words, the process of step S20 is a data acquisition process.

[0033] In step S30, the CPU 102 identifies the value of the other vehicle operation data R4 from the vehicle data DC of the multiple other vehicles 20H. Specifically, the CPU 102 refers to the vehicle data DC of all other vehicles 20H other than the vehicle 20T of the person to be evaluated among the data stored in the storage device 104. The CPU 102 handles each of the vehicle data DC of the multiple other vehicles 20H as the other vehicle data DCH. The CPU 102 then sequentially analyzes the multiple other vehicle data DCH. The CPU 102 then determines whether or not there is a scene to be evaluated for each of the multiple other vehicle data DCH for the same fixed period as in step S10. The method of determining whether or not there is a scene to be evaluated is the same as that described in step S10. If there is a scene to be evaluated, the CPU 102 identifies the vehicle speed SP linked to the position coordinate PN that coincides with the position coordinate PNQ of the evaluation point as the value of the other vehicle operation data R4. As in step S10, when a single other vehicle data DCH contains a plurality of scenes to be evaluated, the value of the other vehicle operation data R4 is identified for the latest case. The CPU 102 sequentially analyzes the plurality of other vehicle data DCH to identify the values ​​of the other vehicle operation data R4 for the number of other vehicles 20H that have traveled through the evaluation point. After executing the process of step S30, the CPU 102 advances the process to step S40.

[0034] In step S40, the CPU 102 determines whether the number of the other vehicle operation data R4 identified in step S30 is equal to or greater than the specified value K1. If the number of the other vehicle operation data R4 is equal to or greater than the specified value K1 (step S40: YES), the CPU 102 proceeds to step S50.

[0035] In step S50, the CPU 102 performs statistical processing on the values ​​of the multiple other vehicle operation data R4 identified in step S30 to calculate an average value R4A and a standard deviation R4B of the values ​​of the multiple other vehicle operation data R4. After that, the CPU 102 advances the process to step S60.

[0036] In step S60, the CPU 102 determines whether the standard deviation R4B calculated in step S50 is equal to or less than the judgment value K2. If the standard deviation R4B is equal to or less than the judgment value K2 (step S60: YES), the CPU 102 advances the process to step S70. The processes in steps S40 and S60 are a method determination process.

[0037] In step S70, the CPU 102 calculates the value of the ideal operation data R3. Specifically, the CPU 102 sets the average value R4A calculated in step S50 as the value of the ideal operation data R3. After this, the CPU 102 advances the process to step S80. The process after step S80 will be described later. The process of steps S50 and S70 is an ideal data calculation process using the first method.

[0038] If the number of other vehicle operation data R4 is less than the specified value K1 in step S40 (step S40: NO), the CPU 102 advances the process to step S160. Also, if the standard deviation R4B is greater than the judgment value K2 in step S60 (step S60: NO), the CPU 102 advances the process to step S160.

[0039] In step S160, the CPU 102 performs a calculation using the automatic driving model DU. This calculation process includes a process of grasping premise information necessary for the driving simulation. The premise information indicates the traffic environment that is the premise of the driving simulation, and is an external factor that the driver himself cannot change. Specifically, the CPU 102 refers to the position coordinate PNQ of the evaluation point and the map data DF. Then, the CPU 102 specifies the position coordinate PNQ of the evaluation point on the map data DF. Then, the CPU 102 grasps information on the road before and after the evaluation point including the evaluation point. The road information includes the width of the roadway, the width of the sidewalk, and the curvature of the road. In addition, the CPU 102 grasps object information around the evaluation point from the surroundings monitoring information JM specified in step S10. The object information is the presence or absence of a guardrail or a curb between the roadway and the sidewalk, the position of a pedestrian, etc. As described above, the CPU 102 uses the position coordinate PNQ of the evaluation point and the surroundings monitoring information JM as initial information for obtaining the premise information. In this way, using the position coordinate PNQ of the evaluation point and the surroundings monitoring information JM as initial information is equivalent to inputting the position coordinate PNQ of the evaluation point and the surroundings monitoring information JM to the automatic driving model DU. After grasping the premise information, the CPU 102 performs a driving simulation assuming that the vehicle 20 is driven around the evaluation point without driver operation. Then, the CPU 102 calculates the transition of the values ​​of various driving parameters as the output of the automatic driving model DU. At this time, the CPU 102 also calculates the transition of the position coordinate PN of the vehicle 20 by the driving simulation. When the CPU 102 completes the calculation by the automatic driving model DU, the CPU 102 advances the process to step S170.

[0040] In step S170, the CPU 102 calculates the value of ideal operation data R3. Specifically, the CPU 102 sets the vehicle speed SP at the position coordinate PNQ of the evaluation point, among the transitions of the values ​​of the various driving parameters calculated in step S160, as the value of the ideal operation data R3. After this, the CPU 102 advances the process to step S80. The processes in steps S160 and S170 are ideal data calculation processes that utilize the second method.

[0041] In step S80, the CPU 102 evaluates the driving skill of the evaluation target. Specifically, the CPU 102 subtracts the value of the ideal operation data R3 calculated in step S70 or step S170 from the value of the evaluation target data R2 specified in step S20. The CPU 102 sets the obtained value as an evaluation value. Then, the CPU 102 judges whether the evaluation value is equal to or less than a reference value K3. If the evaluation value is equal to or less than the reference value K3 (step S80: YES), the CPU 102 advances the process to step S90.

[0042] In step S90, the CPU 102 determines that the vehicle speed SP when passing by the pedestrian is appropriate. In this case, the CPU 102 displays on the display device 150 that the driving style of the person to be evaluated is good. After this, the CPU 102 ends the series of processes of the diagnosis process.

[0043] On the other hand, if the evaluation value is greater than the reference value K3 in step S80 (step S80: NO), CPU 102 advances the process to step S100. In step S100, the CPU 102 determines that the vehicle speed SP when passing by the pedestrian is too high. In this case, the CPU 102 displays a message to that effect and a message to the effect that the driver's driving style needs to be improved on the display device 150. After this, the CPU 102 ends the series of processes in the diagnosis process. Note that steps S80, S90, and S100 are evaluation processes.

[0044] <Operation of the First Embodiment> The subscription service provider 300 uses diagnostic processing to evaluate the user's driving skills at multiple evaluation points. The provider 300 then notifies the user of the results. For example, the provider 300 notifies the user of the general tendency of the user's driving operation, as well as any points that require special attention.

[0045] <Advantages of the First Embodiment> In judging the quality of driving skills at a certain evaluation point, it is preferable to take into account the traffic environment specific to that point. From this perspective, it is preferable to calculate the value of the ideal operation data R3 based on the values ​​of the other vehicle operation data R4 of multiple other vehicles 20H that actually traveled through that evaluation point as a criterion for evaluating driving skills. On the other hand, depending on the point, there may not be a large number of other vehicle operation data R4 that are statistically significant.

[0046] Therefore, in this embodiment, when the number of other vehicle operation data R4 at the evaluation point is small, the value of the ideal operation data R3 is calculated using the automatic driving model DU. Therefore, even at a point where the number of other vehicle operation data R4 is small, the driving skill can be appropriately evaluated. On the other hand, in this embodiment, when the number of other vehicle operation data R4 at the evaluation point is large, the value of the ideal operation data R3 is calculated based on the values ​​of multiple other vehicle operation data R4. Therefore, at a point where there is a lot of other vehicle operation data R4, the driving skill can be evaluated based on the value of the ideal operation data R3 that reflects the traffic environment specific to each point and cannot be incorporated into the automatic driving model DU.

[0047] <Second embodiment> A second embodiment of the driving skill evaluation system will be described. In the second embodiment, only the configuration of the automatic driving model differs from the first embodiment. In the following, the parts different from the first embodiment will be mainly described, and the explanation of the contents overlapping with the first embodiment will be simplified or omitted.

[0048] As in the first embodiment, the storage device 104 stores the automatic driving model DU2 in advance. This automatic driving model DU2 has a read model, an image recognition model, and a running system model. The running system model is a learned model that has been machine-learned before being implemented in the server 101. Details of the mapping that defines the running system model will be described later. The running system model outputs the value of an output variable by inputting the values ​​of the following six input variables. Specifically, the input variables are the width A1 of the roadway, the width A2 of the sidewalk, the curvature A3 of the road, the first object identification value A4, the second object identification value A5, and the third object identification value A6. The first object identification value A4 is a variable that indicates the presence or absence of a guardrail between the roadway and the sidewalk. The first object identification value A4 takes a value of "0" when there is no guardrail, and takes a value of "1" when there is a guardrail. The second object identification value A5 is a variable that indicates the presence or absence of a curb between the roadway and the sidewalk. The second object identification value A5 takes a value of "0" when there is no curb, and takes a value of "1" when there is a curb. The third object identification value A6 is a variable that indicates the presence or absence of a pedestrian. The third object identification value A6 takes a value of "0" when a pedestrian is not present, and takes a value of "1" when a pedestrian is present. The output variable is the vehicle speed SP.

[0049] The image recognition model is a trained model that has been machine-learned before being implemented in the server 101. The image recognition model takes image information included in the perimeter monitoring information JM as input, and outputs a first object identification value A4, a second object identification value A5, and a third object identification value A6. For example, the output of the image recognition model, for the first object identification value A4, is "0" when no guardrail exists, and "1" when a guardrail exists. For the second object identification value A5 and the third object identification value A6, the values ​​defined above are output.

[0050] When a certain position coordinate PN is input, the readout model outputs the road width A1, the sidewalk width A2, and the road curvature A3 at the input position coordinate PN based on the map data DF.

[0051] When the above-described automatic driving model DU2 is adopted, in the diagnosis process, the CPU 102 performs the processes of the following steps S610, S620, and S630 instead of the process of step S160 in the first embodiment.

[0052] As shown in FIG. 4, in step S610, the CPU 102 acquires values ​​of input variables to the traveling system model. Specifically, the CPU 102 inputs the position coordinates PNQ of the evaluation point to the readout model. Then, the CPU 102 calculates the width A1 of the roadway, the width A2 of the sidewalk, and the curvature A3 of the road at the evaluation point as the output of the readout model. The CPU 102 also inputs the surroundings monitoring information JM identified in step S10 to the image recognition model. The CPU 102 calculates the first object identification value A4, the second object identification value A5, and the third object identification value A6 as the output of the image recognition model. After this, the CPU 102 advances the process to step S620. Note that the calculation of the values ​​of the above variables by the CPU 102 corresponds to the acquisition of the values ​​of the above variables by the CPU 102.

[0053] In step S620, as pre-processing for using the traveling system model, CPU 102 assigns the values ​​of each variable acquired in the processing of step S610 to input variables x(1) to x(6) for input to the traveling system model. Specifically, CPU 102 assigns roadway width A1 to input variable x(1). CPU 102 assigns sidewalk width A2 to input variable x(2). CPU 102 assigns road curvature A3 to input variable x(3). CPU 102 assigns first object identification value A4 to input variable x(4). CPU 102 assigns second object identification value A5 to input variable x(5). CPU 102 assigns third object identification value A6 to input variable x(6). After this, CPU 102 advances the processing to step S630.

[0054] In step S630, CPU 102 inputs input variables x(1) to x(6) into the traveling system model to calculate the value of output variable y. As described above, output variable y is vehicle speed SP. After that, in step S170, CPU 102 sets vehicle speed SP calculated in step S630 as the value of ideal operation data R3.

[0055] A mapping that defines the traveling system model will be described. The mapping is a mapping configured as a fully connected forward propagation type neural network with a single layer of intermediate layers. The neural network includes input side coefficients wFjk (j=0 to n, k=0 to 6) and an activation function h(x) as an input side nonlinear mapping that nonlinearly transforms each of the outputs of the input side linear mapping, which is a linear mapping defined by the input side coefficients wFjk. In this embodiment, the activation function h(x) is exemplified as a hyperbolic tangent "tanh(x)". In addition, the neural network includes an activation function f(x) as an output side nonlinear mapping that nonlinearly transforms each of the outputs of the output side linear mapping, which is a linear mapping defined by the output side coefficients wSj (j=0 to n). In this embodiment, the activation function f(x) is exemplified as a hyperbolic tangent "tanh(x)". Note that the value n indicates the dimension of the intermediate layer. The input coefficient wFj0 ​​is a bias parameter and is the coefficient of the input variable x(0). The input variable x(0) is set to "1." The output coefficient wS0 is also a bias parameter.

[0056] A learning method of the traveling system model will be described. As described above, the traveling system model is subjected to machine learning before being implemented in the server 101. In learning the traveling system model, a plurality of learning data sets required for learning are created in advance. One learning data set is composed of training data and teacher data. The teacher data is the vehicle speed SP. The training data is a set of six variables that are input to the traveling system model. That is, the training data is composed of the width A1 of the roadway, the width A2 of the sidewalk, the curvature A3 of the road, the first object identification value A4, the second object identification value A5, and the third object identification value A6. In creating the learning data set, a traveling simulation of the automatic driving model DU of the first embodiment is used. That is, various patterns are set for the width of the roadway, the width of the sidewalk, the curvature of the road, the presence or absence of a guardrail, the presence or absence of a curb, and the presence or absence of a pedestrian, and a traveling simulation is performed. Then, the vehicle speed SP in each pattern is calculated. A learning data set can be obtained from the combination of the roadway width, sidewalk width, road curvature, the presence or absence of guardrails, the presence or absence of curbs, and the presence or absence of pedestrians set in each pattern, and the vehicle speed SP at that time. For example, assume that a driving simulation is performed as a first pattern with a first value for the roadway width, a second value for the sidewalk width, a third value for the road curvature, the presence of guardrails, the absence of curbs, and the presence of pedestrians. Assume that the vehicle speed SP obtained from this driving simulation is a fourth value. The training data for this first pattern are a first value for the roadway width A1, a second value for the sidewalk width A2, a third value for the road curvature A3, a first object identification value A4 of "1", a second object identification value A5 of "0", and a third object identification value A6 of "1". The teacher data for the first pattern is the fourth value. In this way, multiple learning data sets are created. When the number of learning data sets necessary for learning the mapping is accumulated, the mapping is learned using the multiple learning data sets. That is, for each learning data set, the input and output coefficients of the mapping are adjusted so that the difference between the vehicle speed SP value output by the mapping using the training data as input and the teacher data value is a predetermined value or less. When the difference is equal to or less than the predetermined value, learning is considered to be complete.

[0057] Although we will not go into details here, the image recognition model is also trained using multiple training data prepared in advance. For example, in the case of a guardrail, the training data would be images of a pattern in which the guardrail is present and images of a pattern in which the guardrail is not present. Using such training data, the model is trained to output a value of "0" or "1" depending on whether or not a guardrail is present.

[0058] <Effects of the second embodiment> As described above, the driving system model and the image recognition model of the second embodiment are trained models that have been machine-learned in advance. When using such trained models, the accuracy of the model output can be guaranteed by preparing a sufficient amount of training data and training the model appropriately. When such an autonomous driving model DU2 is used in diagnostic processing, the processing load on the server 101 when performing calculations using the autonomous driving model DU2 can be reduced.

[0059] <Example of change> The above-described embodiments may be modified as follows: The above-described embodiments and the following modifications may be combined with each other to the extent that no technical contradiction occurs.

[0060] The determination conditions used in the method determination process are not limited to the examples in the above embodiment. Other contents may be adopted as the determination condition items instead of or in addition to those in the above embodiment. The number of items in the determination condition may be one. The determination condition may include an item that the number of multiple other vehicle operation data R4 is equal to or greater than the specified value K1. The processing contents of the diagnosis conditions may be appropriately changed so that the adopted determination condition can be realized.

[0061] The specified value K1 is not limited to the example of the above embodiment. The specified value K1 may be an integer of 2 or more. For example, in a case where a point where it is known in advance that there are a considerable number of factors that cannot be incorporated into the automatic driving model is set as the evaluation point, the following may happen. That is, even if the significance is somewhat low, calculating the value of the ideal operation data R3 from a plurality of other vehicle operation data R4 may be more effective in terms of evaluating the driving skill than calculating the value of the ideal operation data R3 by the automatic driving model. The specified value K1 may be set to an appropriate value taking such points into consideration. The same applies to the judgment value K2. That is, the judgment value K2 is not limited to the example of the above embodiment.

[0062] The method of evaluating the driving skills in the evaluation process is not limited to the example of the above embodiment. For example, the driving skills may be evaluated using a value obtained by dividing the value of the evaluation target data R2 by the value of the ideal operation data R3 as an index. When such an index is adopted, the reference value K3 may be changed to an appropriate value according to the index. As long as the evaluation is performed using the comparison result between the evaluation target data R2 and the ideal operation data R3, any method may be used to evaluate the driving skills.

[0063] The method of determining whether or not a target scene exists is not limited to the example of the above embodiment. For example, the presence or absence of a target scene may be determined using data in which image information included in the surrounding monitoring information JM is arranged over time, i.e., video data. Any method may be used as long as it can appropriately determine whether or not a target scene exists.

[0064] Acquiring the evaluation target data R2 is not limited to identifying a value of the subject vehicle data DCT as the value of the evaluation target data R2 as in the above embodiment. For example, as one aspect of acquiring the evaluation target data R2, the value of the evaluation target data R2 may be calculated through various arithmetic processing. If information on the value of the evaluation target data R2 can be obtained, this may be treated as acquiring the evaluation target data R2.

[0065] The contents of the statistical processing when the values ​​of the multiple other vehicle operation data R4 are statistically processed are not limited to the examples in the above embodiment. For example, instead of calculating the standard deviation R4B, a variance may be calculated. Also, for example, instead of calculating the average value R4A, a median or a mode may be calculated. Such a value may then be adopted as the value of the ideal operation data R3. The contents of the statistical processing may be any that can calculate a value indicating a tendency or characteristic of the values ​​of the multiple other vehicle operation data R4.

[0066] As described in the above modification, when calculating the value of the ideal operation data R3 from the values ​​of the multiple other vehicle operation data R4, the value of the ideal operation data R3 may be a value other than the average value R4A of the multiple other vehicle operation data R4. The value of the ideal operation data R3 may be an appropriate combination of statistical values ​​such as the average value, median, mode, and standard deviation. It is sufficient to be able to calculate an appropriate value as the value of the ideal operation data R3 for comparison with the value of the evaluation target data R2.

[0067] The driving parameters are not limited to those exemplified in the above embodiment. Any parameter whose value changes depending on the driving operation of the driver may be treated as a driving parameter. For example, an element grasped from the surroundings monitoring information JM may be treated as a driving parameter. Examples of such driving parameters include the distance between the pedestrian and the vehicle 20 and the distance between the vehicle and the preceding vehicle.

[0068] Regarding the driving operation data R1 used as an index for evaluating driving skills, the parameters adopted as the driving operation data R1 are not limited to those in the above embodiment. For example, the accelerator operation amount AP may be adopted as the driving operation data R1. Also, as described in the above modified example, the distance between the pedestrian and the vehicle 20 may be adopted as the driving operation data R1. Also, the distance from the vehicle in front may be adopted. As described in the modified example below, when the scene for evaluating driving skills is changed from that of the above embodiment, parameters suitable for the target scene may be adopted as the driving operation data R1.

[0069] The target scenes for evaluating driving skills are not limited to the examples in the above embodiment. For example, a scene of passing an intersection, a scene of overtaking a vehicle ahead, or a scene of another vehicle 20 entering a merging point may be adopted as the above scene. Any scene may be targeted as long as it is suitable for evaluating driving skills. Evaluation points may be set according to the target scene.

[0070] The diagnostic process may be performed in real time while the vehicle 20 is traveling. The server 101 successively receives the vehicle transmission data W from the vehicle 20. Therefore, it is possible to perform the diagnostic process while the vehicle 20 is traveling.

[0071] Regarding the autonomous driving model DU of the first embodiment, the information used to grasp the premise information is not limited to the example of the above embodiment. For example, data in which image information included in the surrounding monitoring information JM is arranged over time, i.e., video data, may be used. Also, for example, weather information may be used. Then, information on the road surface condition grasped from this weather information may be included in the premise information. It is sufficient to use information required as appropriate so that the driving simulation can be performed appropriately. It is sufficient that one of the pieces of information used to grasp the premise information includes information on the evaluation point.

[0072] Regarding the autonomous driving model DU2 of the second embodiment, the input variables adopted in the driving system model are not limited to those in the above embodiment. For example, a variable indicating the weather or a variable indicating the time may be adopted instead of or in addition to the input variables adopted in the above embodiment. The input variables may include the position coordinate PN of the evaluation point. The input variables may be determined so that appropriate values ​​of the output variables can be obtained. The number of input variables may be changed from the example of the above embodiment. Even if the type or number of input variables is changed from the example of the above embodiment, the driving system model may be appropriately trained so that appropriate values ​​of the output variables can be obtained.

[0073] The output variables used in the driving system model are not limited to those in the above embodiment. For example, when the parameters of the driving operation data R1 used as an index for evaluating driving skills are changed from those in the above embodiment, the values ​​of the output variables can be changed accordingly. The number of output variables may be two or more. Appropriate output variables can be used according to the situation in which the driving skills are to be evaluated. Then, appropriate input variables can be determined according to the output variables.

[0074] Driving skills may be evaluated using multiple types of driving parameters as indexes at one evaluation point. For example, in the case of the scene of the above embodiment, the values ​​of the ideal operation data R3 may be calculated for both the vehicle speed SP and the distance between the pedestrian and the vehicle 20. Then, driving skills may be evaluated using both of these two parameters as indexes. In this way, when evaluating driving skills using multiple types of driving parameters as targets, it is effective to have multiple output variables of the driving system model as in the above modified example.

[0075] The mapping that defines the driving system model is not limited to the example of the above embodiment. The mapping may be any mapping that can output appropriate output variable values ​​for the input variables used to output the driving operation data R1 that is used as an index for evaluating the driving skill.

[0076] The configuration of the autonomous driving model is not limited to that of the first and second embodiments. The autonomous driving model may be configured to input information about an evaluation point and output the value of the driving operation data R1 that the vehicle 20 should take when traveling through the evaluation point. Furthermore, the autonomous driving model may be configured to input information about the traffic environment that the driver cannot change and output driving parameters. The information about the traffic environment is, for example, the structure of the road, such as the road width, the curvature of the road, and the presence or absence of an intersection. The information about the traffic environment is, for example, traffic rules defined by signs, road surface conditions according to the weather, etc.

[0077] The autonomous driving model may be one that utilizes a table or a graph, instead of the trained model shown in the second embodiment. For example, when creating a table for the scene exemplified in the above embodiment, the following aspects are possible. That is, a table is created that shows the relationship between several patterns of the roadway width, sidewalk width, and road curvature, and the vehicle speed SP that the vehicle 20 should take in a situation where a pedestrian is present on the sidewalk in each pattern. Then, based on this table, the autonomous driving model is configured to output the vehicle speed SP corresponding to the roadway width, sidewalk width, and road curvature at the evaluation point. If the number of evaluation points is somewhat limited and the range of values ​​that the roadway width, sidewalk width, and road curvature can take is determined to some extent, the above table can be easily created. Note that the travel simulation of the first embodiment may be used to create the above table.

[0078] In calculating the value of the ideal operation data R3 using an automatic driving model, multiple types of automatic driving models may be combined. For example, both the automatic driving model DU of the first embodiment and the automatic driving model DU2 of the second embodiment are stored in the storage device 104. Then, depending on the situation, one of the automatic driving models is selected to calculate the value of the ideal operation data R3. For example, when the road structure is relatively complex, such as when there are a series of winding roads, the automatic driving model DU of the first embodiment is used. On the other hand, when the road structure is relatively simple, such as when there are a series of straight roads, the automatic driving model DU2 of the second embodiment is used. In this manner, two automatic driving models may be appropriately selected.

[0079] When inputting information about evaluation points to the autonomous driving model, the information about the evaluation points does not have to be location coordinates. For example, index values ​​such as "1" and "2" may be defined for each evaluation point. The index values ​​may then be used as input to the autonomous driving model.

[0080] The owner of the driving skill evaluation system 100 and the use of evaluating driving skills through diagnostic processing are not limited to the examples of the above embodiment. For example, the owner of the driving skill evaluation system 100 may be an insurance service provider. The insurance service provider may evaluate driving skills for the purpose of changing the contents of insurance depending on driving skills. Furthermore, the person to be evaluated may evaluate his / her own driving skills for the purpose of improving his / her driving skills, etc.

[0081] The storage device 104 and the analysis device 107 may be configured as parts of separate processing devices. For example, as shown in FIG. 5, the driving skill evaluation system 500 may have a first processing device 501 and a second processing device 502. The first processing device 501 has a communication device for connecting to the outside via the external communication line network 12, and has a storage device 104 that functions in the same manner as the above embodiment. That is, the first processing device 501 repeatedly receives vehicle transmission data W from a plurality of vehicles 20. Then, the storage device 104 of the first processing device 501 sequentially stores the received vehicle transmission data W. The second processing device 502 has a communication device for connecting to the outside via the external communication line network 12, and has an analysis device 107 that functions in the same manner as the above embodiment. That is, the analysis device 107 of the second processing device 502 can execute evaluation processing. The display device 150 and the input device 160 are connected to the second processing device 502. The analysis device 107 of the second processing device 502 acquires various pieces of information stored in the storage device 104 of the first processing device 501 when performing the evaluation process.

[0082] In the above configuration, each of the processing devices 501, 502 may be configured as one or more processors that execute various processes according to a computer program (software). Each of the processing devices 501, 502 may be configured as one or more dedicated hardware circuits, such as application specific integrated circuits (ASICs), that execute at least a part of the various processes, or a circuitry including a combination thereof. The processor includes a CPU and memory, such as a RAM and a ROM. The memory stores program codes or instructions that are configured to cause the CPU to execute the processes. The memory, i.e., computer-readable medium, includes any available medium that can be accessed by a general-purpose or special-purpose computer. [Explanation of symbols]

[0083] 100...Driving Skills Evaluation System 101...Server 104...Storage device 107…Analysis device

Claims

[Claim 1] An analysis device and a storage device, The analysis device includes: A data acquisition process for acquiring, as an evaluation target data value, a driving operation data value of the vehicle when the vehicle of the evaluation target travels through the evaluation point; an ideal data calculation process for calculating an ideal driving operation data value when the vehicle travels through the evaluation point as an ideal operation data value; An evaluation process for evaluating the driving skill of the person to be evaluated based on a comparison result between the value of the evaluation target data and the value of the ideal operation data; The storage device includes: A value of other vehicle operation data, which is the driving operation data when a vehicle other than the vehicle of the evaluation subject travels through the evaluation point; and an automatic driving model that outputs, by inputting the evaluation point, the value of the driving operation data that should be taken when the vehicle travels through the evaluation point; In the ideal data calculation process, When the number of the other vehicle operation data is equal to or greater than a predetermined value that is an integer equal to or greater than 2, the value of the driving operation data obtained by statistically processing the values ​​of the multiple other vehicle operation data is set as the ideal operation data value; When the number of the other vehicle operation data is less than the specified value, the value of the driving operation data output by the automatic driving model is set as the value of the ideal operation data. Driving skills evaluation system.

Images