Driving Skills Evaluation System

Abstract

To appropriately and simply evaluate driving skill in a situation where an object that should be recognized by a driver exists.SOLUTION: A driving skill estimation system executes: processing (step S30) for identifying a situation where a target object which need to be recognized by an evaluation subject exists within an objective range including a vehicle on the basis of traveling data obtained when the evaluation subject drove the vehicle; processing (step S40) for identifying a timing when the evaluation subject recognizes the target object in the identified situation; processing (step S50) for calculating a recognition period A1P indicating a length of time from when the target object appears within the objective range to when the evaluation subject recognizes the target object; and processing (step S100) for calculating an evaluation variable by inputting a plurality of input variables including the recognition period A1P into a mapping.SELECTED DRAWING: Figure 11

Description

[Technical field]

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

[0002] The safe driving support system of Patent Document 1 includes a server and a plurality of devices connected to the server via a communication network. Each device includes a device for measuring the driving ability of the vehicle driver and a device for measuring the health information of the driver. The server evaluates the driving risk of the driver by applying an evaluation score related to the driver's driving ability and an evaluation score related to the driver's health condition to a predetermined evaluation criterion. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2015-141536 A Summary of the Invention [Problem to be solved by the invention]

[0004] In an actual driving environment on the road, there are objects that the driver should recognize, such as road signs. When evaluating driving skills in a situation where such objects exist, it is necessary to make a comprehensive evaluation by taking into consideration various factors regarding the driver's behavior and the vehicle's driving state in the situation. However, the factors that should be considered important in calculating the final evaluation result may change depending on the situation. Therefore, it is difficult to express each factor and the final evaluation result in a unique relational equation. Due to these circumstances, when evaluating driving skills in the above-mentioned situation, it is necessary for an evaluator with extensive experience and specialized knowledge to compare the evaluation results for each factor and make a comprehensive judgment, which is time-consuming. [Means for solving the problem]

[0005] A driving skill evaluation system for solving the above problem is a driving skill evaluation system for evaluating the driving skills of a person to be evaluated, the system having a storage device and an analysis device, the storage device pre-stores mapping data that specifies a mapping that outputs an evaluation variable, which is a variable that indicates an evaluation of the driving skill of the person to be evaluated, when a plurality of input variables are input, the storage device stores driving data of the vehicle including a change in the position of the vehicle when the person to be evaluated drives the vehicle, and physical data of the person to be evaluated over time, and the analysis device executes a scene identification process that identifies, based on the driving data, a scene in which a recognition object that the person to be evaluated needs to recognize is present within a predetermined target range including the vehicle during a period in which the person to be evaluated drives the vehicle; a recognition timing identification process that identifies, based on the physical data, the timing at which the person to be evaluated recognizes the recognition object in the identified scene; a pre-processing process that calculates a recognition variable, which is a variable that indicates the length of time from when the recognition object appears within the target range to when the person to be evaluated recognizes the recognition object; and a calculation process that calculates the evaluation variable by inputting a plurality of the input variables including the recognition variable into the mapping.

[0006] In the above configuration, the evaluation variable is calculated using a mapping. The cognitive variable is included as one of the multiple input variables input to the mapping. The time required to recognize a recognition object is a very important driving skill in a situation where the recognition object is present. Therefore, by including the cognitive variable as one of the multiple input variables, the reliability of the evaluation result can be guaranteed. Furthermore, when using the mapping, if appropriate teacher data and training data can be prepared, appropriate evaluation results can be easily obtained. Therefore, even if the relationship between the cognitive variable and the evaluation variable changes significantly due to input variables other than the cognitive variable, appropriate evaluation results can be obtained without relying on human experience and skill. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a schematic diagram of a driving skill evaluation system. [Diagram 2] FIG. 2 is a diagram showing an example of an image of the internal imaging information. [Diagram 3] FIG. 3 is a diagram showing an example of a virtual world constructed using simulation data. [Figure 4] FIG. 4 is a diagram for explaining how to identify a meeting point. [Diagram 5] FIG. 5 is a diagram for explaining how to identify a meeting point. [Figure 6] FIG. 6 is a diagram for explaining a method for identifying a scene in which a vehicle ahead is present. [Figure 7] FIG. 7 is a diagram for explaining how to specify a scene in which a sign is present. [Figure 8] FIG. 8 is a diagram for explaining a method for identifying a scene in which a pedestrian is present. [Figure 9] FIG. 9 is a diagram for explaining a method for calculating the acceleration difference integrated value and the deceleration difference integrated value. [Figure 10] FIG. 10 is a diagram for explaining a method for calculating the steering angle difference integrated value. [Figure 11] FIG. 11 is a flowchart showing the procedure of the evaluation process. [Figure 12] FIG. 12 is a diagram showing an example of an image displayed in the display process. [Figure 13] FIG. 13 is a diagram for explaining the cognitive distance in a merging scene. [Figure 14] FIG. 14 is a diagram for explaining the recognition distance in a scene where a vehicle ahead is present and a scene where a sign is present. [Figure 15] FIG. 15 is a diagram for explaining the recognition distance in a scene where a pedestrian is present. [Figure 16] FIG. 16 is a diagram for explaining pedestrian variables in a scene where pedestrians are present. [Figure 17] FIG. 17 is a schematic diagram showing a modified example of the driving skill evaluation system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, an 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 capable of communicating with a plurality of vehicles 20, a plurality of road cameras 80, and the like via an external communication network 12. The driving skill evaluation system 100 includes a data server that integrates data from the plurality of vehicles 20 and the plurality of road cameras 80, and the like. The plurality of vehicles 20 and the plurality of road cameras 80 are registered in advance in the driving skill evaluation system 100 as targets for data integration in the system 100. In this embodiment, substantially all vehicles owned in a predetermined target area are subject to registration. In addition, in this embodiment, substantially all road cameras present in the target area are subject to registration.

[0009] The vehicles 20 are, for example, four-wheeled automobiles such as hybrid vehicles, electric vehicles, and fuel cell vehicles. 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 related to the vehicle 20 while the vehicle 20 is traveling. Each information acquisition device 30 is any one of a GPS receiver 31, a vehicle speed sensor 32, an accelerator sensor 33, a brake sensor 34, an acceleration sensor 35, a steering sensor 36, an exterior camera 37, and an interior camera 38.

[0010] 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 vehicle speed sensor 32 detects the traveling speed of the vehicle 20 as the vehicle speed SP. 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 acceleration sensor 35 detects the acceleration in the front-rear direction of the vehicle 20 as the front-rear acceleration AC. The steering sensor 36 detects the steering angle of the steering wheel as the steering angle ST. The exterior camera 37 captures an image of the outside of the vehicle 20, and obtains image information obtained thereby as external imaging information JM. The interior camera 38 captures an image of the inside of the vehicle 20, particularly the driver's seat and its surroundings, and obtains image information obtained thereby as internal imaging information JN.

[0011] 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 prestores an identification number assigned to the vehicle. The control device 22 simultaneously receives the latest information acquired by each information acquisition device 30 at a predetermined cycle. When the control device 22 receives information from each information acquisition device 30, it generates vehicle transmission data WP that consolidates the information. At that time, the control device 22 includes the date and time when the information was received from each information acquisition device 30 and the identification number of the vehicle in the vehicle transmission data WP. When the control device 22 generates the vehicle transmission data WP, it transmits the vehicle transmission data WP to the driving skill evaluation system 100. The control device 22 repeats the generation and transmission of the vehicle transmission data WP at a predetermined control cycle.

[0012] The road cameras 80 are cameras installed on roads. The road cameras 80 are installed at various locations on expressways and general roads, such as at entrances and exits of junctions, intersections, and near crosswalks. The road cameras 80 capture images of the surroundings of the locations where the road cameras 80 are installed, and obtain the image information obtained thereby as road image information JR. When the road cameras 80 obtain the road image information JR, they generate road transmission data WR including the road image information JR, the date and time when the road image information JR was obtained, and the identification number of the camera itself. The road cameras 80 then transmit the road transmission data WR to the driving skill evaluation system 100 via the external communication network 12. The road cameras 80 repeat the generation and transmission of the road transmission data WR at a predetermined control period.

[0013] <Driving Skill Evaluation System> The driving skill evaluation system 100 is owned, for example, by a car sharing service provider 300. The car sharing service provider 300 manages a plurality of vehicles 20 for renting to users. The plurality of vehicles 20 managed by the provider 300 are a part of the plurality of vehicles 20 registered in the driving skill evaluation system 100 as targets for data integration.

[0014] The driving skill evaluation system 100 includes a server 101 , a display device 150 , and an input device 160 . 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 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 receives the vehicle transmission data WP transmitted by each vehicle 20 via the external communication network 12. The server 101 also receives the road transmission data WR transmitted by each road camera 80 via the external communication network 12.

[0015] 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 the center line of 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.

[0016] The storage device 104 stores the above-mentioned identification number for each vehicle 20 for distinguishing between a plurality of preregistered vehicles 20. The storage device 104 stores the identification number in association with basic data D1 of each vehicle 20. The basic data D1 includes hardware design and software design information of each vehicle 20. The hardware design and software design information is the latest information according to the vehicle model, type, grade, and customization from standard equipment. An example of the hardware design information is computer-aided design (CAD) data representing the frame structure of the vehicle 20. An example of the software design information is the source code of a program of an advanced driver assistance system (ADAS) for assisting the occupant in driving, such as cruise control.

[0017] The storage device 104 sequentially stores the vehicle transmission data WP received from each vehicle 20 via the communication device 105. The storage device 104 handles the position coordinate PN of the vehicle 20, the vehicle speed SP, the accelerator operation amount AP, the brake operation amount BK, the longitudinal acceleration AC, the steering angle ST, and the external image information JM, among the information included in the vehicle transmission data WP, as driving-related information. The storage device 104 stores each element of the driving-related information for each vehicle 20 in association with the identification number of each vehicle 20. At that time, the storage device 104 stores each element chronologically based on the date and time information included in the vehicle transmission data WP. Hereinafter, a time-series group of each element of the driving-related information is referred to as driving data D2. The driving data D2 is data indicating the driving state of the vehicle 20 when the vehicle 20 is driven in a real driving environment in the real world. As described above, the driving data D2 includes the transition of the position coordinate PN of the vehicle 20.

[0018] Among the information included in the vehicle transmission data WP, the storage device 104 handles the internal image information JN as driver information. The storage device 104 stores the internal image information JN for each vehicle 20 in association with the identification number of each vehicle 20. The storage device 104 stores the internal image information JN over time based on the date and time information included in the vehicle transmission data WP, similar to the traveling data D2. That is, the time-series data of the internal image information JN is video data. Note that the internal image information JN is image data of the driver's seat and its surroundings in the vehicle 20, and therefore includes image data of the driver of the vehicle 20. Therefore, the time-series data of the internal image information JN is physical data D3 indicating the physical state of the driver when the vehicle 20 is traveling in a real driving environment in the real world.

[0019] The storage device 104 stores the above-mentioned identification number for distinguishing between the multiple road cameras 80 that have been registered in advance, for each road camera 80. The storage device 104 stores the identification number in association with the position coordinates where each road camera 80 is installed.

[0020] The storage device 104 sequentially stores the road transmission data WR from each road camera 80 received via the communication device 105. The storage device 104 stores the road image information JR included in the road transmission data WR for each road camera 80 in association with the identification number of each road camera 80. The storage device 104 stores the road image information JR over time based on the date and time information included in the road transmission data WR. That is, the time series data of the road image information JR is video data. Hereinafter, the time series data of the road image information JR is referred to as road camera data D4. By referring to the road camera data D4, the change in the position of the moving object in the image captured by the road camera 80 can be seen, and the change in the position of the moving object between the image capture ranges of the multiple road cameras 80 can also be seen by comparing the road camera data D4 from multiple road cameras 80 over time. That is, the road camera data D4 has information regarding the change in the position of the moving object. The moving object may be, for example, a vehicle 20 or a pedestrian 20M.

[0021] When storing the running data D2, the body data D3, and the road camera data D4, the storage device 104 synchronizes all of these data over time based on date and time information.

[0022] The storage device 104 stores an identification number for each user to identify the user who is registered with the car sharing service. The storage device 104 stores the identification number in association with information about each user. 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.

[0023] The storage device 104 prestores mapping data DS as information necessary for evaluating driving skills. The mapping data DS includes four mappings. A mapping is prepared for each target situation for evaluating driving skills. That is, in this embodiment, there are four types of situations that are the target of evaluation of driving skills. A mapping is a program that outputs an output variable when multiple input variables are input. The input variables include the following:

[0024] Cognitive variable A1 Accelerated Differential Variable A2 Deceleration differential variable A3 Steering angle differential variable A4 The specific contents of each input variable will be described later. The output variable is an evaluation variable B, which is a variable indicating an evaluation of the driving skill of the evaluation subject. The evaluation subject is a user who is registered with the car sharing service.

[0025] The storage device 104 prestores an automatic driving model DU. The automatic driving model DU is a group of programs required to realize automatic driving of the vehicle 20. Automatic driving refers to the control device 22 of the vehicle 20 autonomously driving the vehicle 20 without the operation of the driver.

[0026] 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 has a display screen 150A. That is, the display device 150 displays various types of information on the display screen 150A by light. The display device 150 displays various types of information on the display screen 150A based on a video signal output by the CPU 102 of the server 101.

[0027] 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.

[0028] <Outline of driving skill evaluation process> The CPU 102 can execute a driving skill evaluation process (hereinafter, simply referred to as the evaluation process). The evaluation process is a process for evaluating the driving skill of the person to be evaluated. The CPU 102 realizes each step of the evaluation process by executing a program stored in the ROM 103. Below, an overview of each step of the evaluation process will be explained in order.

[0029] <Extraction process> The CPU 102 can execute an extraction process as part of the evaluation process. Here, in the evaluation process, the CPU 102 analyzes various data including the driving data D2 stored in the storage device 104. In the extraction process, the CPU 102 extracts data to be analyzed from the data stored in the storage device 104. Specifically, the CPU 102 extracts the driving data D2 of the vehicle 20T of the person to be evaluated as the subject driving data DA for a certain analysis target period. In addition, the CPU 102 extracts the driving data D2 of a vehicle (hereinafter referred to as another vehicle) 20H other than the vehicle 20T of the person to be evaluated as the first other object data DB1 for the analysis target period. In addition, the CPU 102 extracts the road camera data D4 of each road camera 80 as the second other object data DB2 for the analysis target period.

[0030] Here, when the subject of evaluation drives the vehicle 20T, there is a recognition object around the vehicle 20T of the subject of evaluation that the subject of evaluation needs to recognize. In the evaluation process, the driving skill is evaluated taking into consideration the timing of recognition of the recognition object. As information for this purpose, in this embodiment, the direction of the subject of evaluation's line of sight is used. As described above, the storage device 104 stores the driver's physical data D3 for the vehicle 20T of the subject of evaluation. The internal image information JN on which this physical data D3 is based is obtained by capturing an image of the inside of the vehicle 20. That is, as shown in FIG. 2, the internal image information JN includes image information of the upper body including the face of the driver. By using this internal image information JN, the direction of the driver's line of sight can be specified as follows. For example, when the iris is located at the center of gravity 240 of the outline of the driver's eyes, it can be specified that the line of sight is directly ahead. Also, for example, as shown in FIG. 2, when the iris is located on the left side of the driver with respect to the center of gravity 240 of the outline of the eyes, the direction of the line of sight can be specified as the left. Furthermore, for example, when the pupil is located above the center of gravity 240 of the eye outline, the direction of the gaze can be identified as upward. Therefore, in the extraction process, CPU 102 extracts physical data D3 of the subject for evaluation for the analysis period as subject physical data DC. After extracting subject physical data DC, CPU 102 generates gaze data DY representing the transition of the subject's gaze using a method based on the center of gravity 240 of the eye. In generating the gaze data DY, the gaze direction may be appropriately quantified.

[0031] <Simulation data generation process> The CPU 102 can execute a simulated data generation process as a part of the evaluation process. In the simulated data generation process, the CPU 102 generates simulated data DM based on the subject travel data DA extracted in the extraction process, the first other object data DB1, the second other object data DB2, the basic data D1 of each vehicle 20, and the map data DF. As shown in FIG. 3, the simulated data DM is a simulation of the travel state of the vehicle 20T of the subject of evaluation in the actual driving environment over the analysis target period in the virtual world. That is, the simulated data DM is data of a four-dimensional space-time consisting of a three-dimensional space and one-dimensional time. In addition, the simulated data DM also simulates the transition of the position of moving objects other than the vehicle 20T of the subject of evaluation, such as other vehicles 20H and pedestrians 20M. In addition, the simulated data DM also simulates non-moving objects, such as road signs RS. The CPU 102 constructs the simulated data DM by executing a simulation program capable of simulating the real world. This technology, which recreates an environment identical to the real world in a digital space, is sometimes called a digital twin.

[0032] <Scene-specific processing> The CPU 102 can execute a scene identification process as part of the evaluation process. In the scene identification process, the CPU 102 identifies a scene (hereinafter, referred to as an evaluation target scene) that is a target for evaluating the driving skill during the period in which the evaluation target person drives the vehicle 20T. The evaluation target scene is a scene in which a recognition object that the evaluation target person needs to recognize exists within a target range including the evaluation target person's vehicle 20T. The CPU 102 identifies the evaluation target scene based on the simulated data DM. As described above, the simulated data DM is generated using the target person driving data DA. Therefore, identifying the evaluation target scene based on the simulated data DM is equivalent to identifying the evaluation target scene based on the target person driving data DA.

[0033] In this embodiment, the CPU 102 can specify four evaluation target scenes. The four scenes are a merging scene, a scene with a preceding vehicle, a scene with a sign, and a scene with a pedestrian. A method for specifying each scene will be described below.

[0034] The merging scene is a scene where another vehicle 20H merges with the road on which the vehicle 20T of the evaluation subject is traveling. Specifically, as shown in FIG. 4, consider a situation in which the vehicle 20T of the evaluation subject is traveling. The road on which the vehicle 20T of the evaluation subject is traveling is called the traveling road R1. Assume that a cross road R2 intersects with the traveling road R1 ahead of the vehicle 20T of the evaluation subject. The above merging scene is a scene where another vehicle 20H merges from the cross road R2 to the traveling road R1 ahead of the vehicle 20T of the evaluation subject. The recognized object in this merging scene is the other vehicle 20H merging from the cross road R2.

[0035] In identifying the merging scene, the CPU 102 sets, for example, a virtual circle CR centered on the vehicle 20T of the evaluation subject as a predetermined target range including the vehicle 20T of the evaluation subject in the virtual world constructed by the simulation data DM. Then, the CPU 102 identifies a scene in which a merging point of the travel road R1 and the cross road R2 exists within the range of this virtual circle CR, and another vehicle 20H exists near the merging point. More specifically, as shown in FIG. 5, the CPU 102 identifies the timing when the other vehicle 20H traveling on the cross road R2 enters the inside of the virtual circle CR from the outside while the vehicle 20T of the evaluation subject travels on the travel road R1 toward the merging point as the reference timing TD1. Then, the CPU 102 identifies the period from the reference timing TD1 until the countermeasure time has elapsed as the merging scene.

[0036] In addition, even if the other vehicle 20H enters inside the virtual circle CR, if the other vehicle 20H is not located near the merging point, the CPU 102 does not identify the scene as a merging scene. In addition, if the other vehicle 20H leaves the virtual circle CR during the period from the reference timing TD1 to the elapse of the response time, the CPU 102 does not identify the scene as a merging scene. This is to eliminate irregular situations, such as the other vehicle 20H making a U-turn on the cross road R2 or the vehicle 20T of the evaluation subject stopping along the way.

[0037] The radius of the virtual circle CR is determined in advance by experiment or simulation as the following value. That is, the radius of the virtual circle CR is a distance slightly longer than the distance at which the subject of evaluation should notice the presence of the other vehicle 20H when a merging point exists in front of the subject's vehicle 20T and the other vehicle 20H exists near the merging point. The radius of the virtual circle CR is, for example, 100 m. The response time is determined in advance by experiment or simulation as the length of time required for the subject of evaluation to complete various responses that should be performed in response to the presence of the other vehicle 20H. One example of various responses is when the subject of evaluation's vehicle 20T needs to decelerate and the subject of evaluation releases the accelerator pedal and steps on the brake pedal. Another example of various responses is when the subject of evaluation's vehicle 20T needs to change lanes and the subject of evaluation operates the steering wheel.

[0038] As shown in Fig. 6, the vehicle-in-front scene is a scene in which another vehicle 20H is present in front of the vehicle 20T of the subject on the road R1. That is, the object to be recognized in the vehicle-in-front scene is the other vehicle 20H located in front of the vehicle 20T.

[0039] As shown in Fig. 7, the sign-present scene is a scene in which a road sign RS is present ahead of the vehicle 20T of the subject on the travel road R1. That is, the recognition target in the sign-present scene is the road sign RS. The road sign RS is, for example, a sign indicating a speed limit.

[0040] As shown in Fig. 8, the pedestrian-present scene is a scene in which a sidewalk R1A exists beside a travel road R1, and a pedestrian 20M exists on the sidewalk R1A in front of the vehicle 20T of the evaluation subject. That is, the recognition target in the pedestrian-present scene is the pedestrian 20M.

[0041] When identifying any of the above-mentioned scenes of a preceding vehicle presence, a sign presence, and a pedestrian presence, the CPU 102 sets a virtual circle CR centered on the vehicle 20T of the evaluation subject, as in the case of the merging scene. Then, the CPU 102 identifies the timing when the recognized object enters the inside of the virtual circle CR from the outside while the vehicle 20T of the evaluation subject travels on the travel road R1 as a reference timing TD1. Then, the CPU 102 identifies the period from the reference timing TD1 until the action time has elapsed as the scene in which each recognized object exists. As in the case of the merging scene, if the evaluation object deviates from the virtual circle CR during the period from the reference timing TD1 until the action time has elapsed, the CPU 102 does not identify the scene as the evaluation subject scene. Note that the radius of the virtual circle CR and the action time are determined in advance for each scene, for example, by experiment or simulation, from the same viewpoint as the merging scene, as values ​​appropriate for each scene. The radius of the virtual circle CR and the action time for each evaluation subject scene are stored in the storage device 104.

[0042] <Cognitive timing specific processing> The CPU 102 can execute a recognition timing specification process as part of the evaluation process. In the recognition timing specification process, the CPU 102 specifies the timing (hereinafter, referred to as recognition timing) TD2 at which the person to be evaluated recognizes the recognition object in the evaluation target scene specified in the scene specification process. The CPU 102 specifies the recognition timing TD2 based on the line of sight data DY. Specifically, the CPU 102 specifies the timing at which the line of sight of the person to be evaluated turns toward the direction of the recognition object as the recognition timing TD2. For example, in a merging scene, the CPU 102 specifies the timing at which the line of sight of the person to be evaluated turns toward the side of the road as the recognition timing TD2. Also, for example, in a scene where a vehicle ahead is present, the CPU 102 specifies the timing at which the line of sight of the person to be evaluated shifts upward as the recognition timing TD2. Also, for example, in a scene where a sign is present, the CPU 102 specifies the timing at which the line of sight of the person to be evaluated turns toward the road sign RS as the recognition timing TD2. Also, for example, in a scene where a pedestrian is present, the CPU 102 identifies the timing at which the line of sight of the evaluation subject turns toward the sidewalk R1A and toward the pedestrian 20M as the recognition timing TD2.

[0043] <First pre-treatment> The CPU 102 can execute a first pre-processing as a part of the evaluation process. In the first pre-processing, the CPU 102 calculates a recognition variable A1, which is one of the above four variables that are input variables to the mapping. The recognition variable A1 is a variable that indicates the length of time from when the recognition object appears in the virtual circle CR until the person being evaluated recognizes the evaluation object. In this embodiment, the CPU 102 calculates a recognition period A1P as the recognition variable A1. The recognition period A1P is the length of time from the above reference timing TD1 to the above recognition timing TD2.

[0044] <Ideal driving data generation process> As part of the evaluation process, the CPU 102 can execute an ideal driving data creation process. In the ideal driving data creation process, the CPU 102 generates data indicating an ideal driving state in the evaluation target scene identified in the scene identification process. In this embodiment, the ideal driving state is a driving state when the vehicle 20T of the evaluation target person is assumed to be driven by automatic driving under the same circumstances as the evaluation target scene identified in the scene identification process. The CPU 102 performs a driving simulation of the vehicle 20 using the automatic driving model DU for the evaluation target scene identified in the scene identification process. As a result, the CPU 102 generates data of the ideal driving state (hereinafter, referred to as ideal driving data) DI. The ideal driving data DI is a time series group of various driving parameters such as the vehicle speed SP, the longitudinal acceleration AC, and the steering angle ST.

[0045] <Second pre-treatment> The CPU 102 can execute a second pre-processing as a part of the evaluation process. In the second pre-processing, the CPU 102 calculates three of the four variables that are input variables to the mapping, other than the recognition variable A1. That is, the CPU 102 calculates an acceleration difference variable A2, a deceleration difference variable A3, and a steering angle difference variable A4.

[0046] The acceleration difference variable A2 is a variable indicating the amount of deviation between the time change of the longitudinal acceleration AC in an ideal driving state and the time change of the actual longitudinal acceleration AC in the evaluation target scene specified in the scene specification process. In detail, the acceleration difference variable A2 is a variable indicating the amount of deviation under a situation where the longitudinal acceleration AC in the ideal driving state is equal to or greater than zero, that is, under a situation during acceleration. In this embodiment, the CPU 102 calculates an acceleration difference integrated value A2P as the acceleration difference variable A2. As shown in the hatched area in FIG. 9, the acceleration difference integrated value A2P is a value obtained by integrating the absolute value of the difference between the ideal longitudinal acceleration AC and the actual longitudinal acceleration AC at each timing, for a series of periods TDH of the evaluation target scene when the ideal longitudinal acceleration AC is equal to or greater than zero.

[0047] The deceleration difference variable A3, like the acceleration difference variable A2, is a variable indicating the amount of deviation between the time change of the longitudinal acceleration AC in an ideal driving state and the time change of the actual longitudinal acceleration AC in the evaluation target scene identified by the scene identification process. In detail, the deceleration difference variable A3 is a variable indicating the amount of deviation under a situation in which the longitudinal acceleration AC in the ideal driving state is less than zero, that is, under a situation during deceleration. In this embodiment, the CPU 102 calculates the deceleration difference integrated value A3P as the deceleration difference variable A3. As shown in the dotted area in FIG. 9, the deceleration difference integrated value A3P is a value obtained by accumulating the absolute value of the difference between the ideal longitudinal acceleration AC and the actual longitudinal acceleration AC at each timing, for the period TDH of the evaluation target scene when the ideal longitudinal acceleration AC is less than zero.

[0048] The steering angle difference variable A4 is a variable indicating the amount of deviation between the time change of the steering angle ST in an ideal driving state and the time change of the actual steering angle ST in the evaluation target scene specified by the scene specification process. In this embodiment, the CPU 102 calculates the steering angle difference integrated value A4P as the steering angle difference variable A4. As shown in the hatched area in FIG. 10, the steering angle difference integrated value A4P is a value obtained by integrating the absolute values ​​of the differences between the ideal steering angle ST and the actual steering angle ST at each timing for a series of periods TDH in the evaluation target scene. Note that the steering angle ST may be set to zero when the vehicle 20 is running straight, a positive value when the vehicle 20 is turned to the left from the driver's perspective, and a negative value when the vehicle 20 is turned to the right.

[0049] <Selection process> As part of the evaluation process, the CPU 102 can execute a selection process. In the selection process, the CPU 102 selects a map corresponding to the scene to be evaluated identified in the scene identification process from among the four maps stored in the storage device 104.

[0050] <Calculation process> The CPU 102 can execute a calculation process as part of the evaluation process. In the calculation process, the CPU 102 calculates the value of an evaluation variable B, which is an output variable, by inputting the values ​​of the input variables calculated in the first pre-processing and the second pre-processing to the mapping selected in the selection process. In this embodiment, the CPU 102 calculates an evaluation score BP of the driving skill as the evaluation variable B. The evaluation score BP takes a value from zero to 100.

[0051] The mapping is a trained model that has been machine-learned before the evaluation process is performed. As described above, there are four mappings. Each mapping is trained individually for the evaluation target scene to which it is applied. Here, a method for training a certain mapping will be described. When training a mapping, multiple data sets are prepared in advance. One data set is configured as a set of one training data and one teacher data. One training data includes four variables, namely, a recognition period A1P, an acceleration difference integrated value A2P, a deceleration difference integrated value A3P, and a steering angle difference integrated value A4P. One teacher data is configured only with the evaluation score BP of the driving skill. When creating one data set, first, the recognition period A1P, the acceleration difference integrated value A2P, the deceleration difference integrated value A3P, and the steering angle difference integrated value A4P are set to appropriate values. For example, the recognition period A1P is set to a first value, the acceleration difference integrated value A2P is set to a second value, the deceleration difference integrated value A3P is set to a third value, and the steering angle difference integrated value A4P is set to a fourth value. The values ​​of these four variables may be appropriately set as values ​​within the ranges that each of them can take. Next, when the above four variables take the above values, the optimal value is determined as the evaluation score BP of the driving skill. This optimal value may be determined, for example, by a person with specialized knowledge of driving skills. Now, assume that the above optimal value is the fifth value. In this case, one data set is created with the first value, the second value, the third value, and the fourth value as training data and the fifth value as teacher data. Through the same process, multiple data sets are prepared. That is, the recognition period A1P, the acceleration difference integrated value A2P, the deceleration difference integrated value A3P, and the steering angle difference integrated value A4P are set to various values, and the evaluation score BP at each time is determined. Then, various patterns of data sets are created as combinations of the above four variables and the corresponding evaluation scores BP. The mapping is learned using the multiple data sets prepared as described above. That is, for each data set, the input side variables and output side variables that define the mapping are adjusted so that the difference between the driving skill evaluation score BP output by the mapping using the training data as input and the teacher data is a predetermined value or less. Then, when the difference becomes a predetermined value or less, the learning is considered to be completed.In learning each mapping for each corresponding evaluation scene target, the above-mentioned multiple data sets may be prepared for each evaluation scene target.

[0052] The mapping is basically trained so that the shorter the recognition period A1P, the higher the evaluation score BP of the driving skill. The mapping is also trained so that the smaller the acceleration difference integrated value A2P, the higher the evaluation score BP. The mapping is also trained so that the smaller the deceleration difference integrated value A3P and the steering angle difference integrated value A4P are, like the acceleration difference integrated value A2P, the higher the evaluation score BP.

[0053] <Display processing> The CPU 102 can execute a display process as part of the evaluation process. The CPU 102 displays the evaluation score BP calculated in the calculation process on the display screen 150A of the display device 150. At the same time, in the display process, the CPU 102 reproduces the driving state of the vehicle 20T of the evaluation subject in the actual driving environment in the evaluation target scene on the display screen 150A of the display device 150 based on the simulation data DM.

[0054] <Specific steps of evaluation process> A user who is registered with the car sharing service can receive an evaluation of his / her driving skills as a privilege. When a user requests that his / her driving skills be evaluated, the business operator 300 evaluates the driving skills of the user using the above-mentioned evaluation process, with the user as the evaluation subject. At that time, the business operator 300 inputs information required for executing the evaluation process to the server 101 via the input device 160. Here, in one evaluation process, the evaluation of the driving skills is performed for one of a merging scene, a scene with a preceding vehicle, a scene with a sign, and a scene with a pedestrian. The above-mentioned necessary information includes an instruction as to which of these four scenes is to be evaluated. The above-mentioned necessary information also includes the user's name. After inputting the above-mentioned necessary information, the business operator 300 inputs an instruction to execute the evaluation process to the server 101.

[0055] The CPU 102 of the server 101 starts the evaluation process in response to an execution instruction from the business operator 300. As shown in Fig. 11, when the CPU 102 starts the evaluation process, first, the CPU 102 executes the process of step S10.

[0056] In step S10, the CPU 102 performs an extraction process. Specifically, the CPU 102 extracts various data required for the subsequent process from the data stored in the storage device 104. As a specific process of step S10, the CPU 102 first identifies the identification number of the evaluation target 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 evaluation target from the information of the user linked to this identification number. After that, the CPU 102 extracts the driving data D2 corresponding to the identification number of the vehicle 20T of the evaluation target from the data stored in the storage device 104. Then, the CPU 102 determines, based on the driving data D2, a period during which the evaluation target drove the vehicle 20T and which satisfies a predetermined driving condition as the analysis target period. The driving condition is that the evaluation target continued driving for a certain period, such as one hour, including temporary stops. When the analysis target period is determined, the CPU 102 extracts data for the analysis target period from the running data D2 extracted above as the subject running data DA. The CPU 102 also extracts, from the data stored in the storage device 104, the physical data D3 corresponding to the identification number of the vehicle 20T of the evaluation target person, which is the physical data D3 during the analysis target period, as the subject physical data DC. The CPU 102 also extracts, from the data stored in the storage device 104, the running data D2 corresponding to the identification numbers of all the vehicles 20 other than the identification number of the vehicle 20T of the target person during the analysis target period, as the first other object data DB1. The CPU 102 also extracts, from the data stored in the storage device 104, the road camera data D4 of all the road cameras 80 during the analysis target period, as the second other object data DB2. The CPU 102 generates the above-mentioned line of sight data DY based on the subject physical data DC as one of the processes of this step S10. After this, the CPU 102 advances the process to step S20.

[0057] In step S20, the CPU 102 performs a simulated data generation process. Specifically, the CPU 102 refers to the subject driving data DA, the first other object data DB1, and the second other object data DB2 extracted in step S10, as well as the basic data D1 of each vehicle 20, the map data DF, and the position coordinates of each road camera 80. Then, the CPU 102 applies each of these data to a program for generating the simulated data DM to generate the simulated data DM. As described above, the generated simulated data DM is a simulation in the virtual world of the transition of the driving state of the vehicle 20T of the subject of evaluation and the transition of other moving objects in the actual driving environment of the real world over the entire analysis target period. After this, the CPU 102 advances the process to step S30.

[0058] In step S30, the CPU 102 performs a scene specification process. That is, the CPU 102 judges whether or not the scene to be evaluated designated by the business entity 300 exists during the analysis period. When making the judgment in step S30, the CPU 102 first judges whether the scene to be evaluated is a merging scene, a scene with a vehicle ahead, a scene with a sign, or a scene with a pedestrian, based on the information input by the business entity 300. Then, the CPU 102 judges whether or not the judged scene to be evaluated exists during the analysis period. In this case, the CPU 102 uses the simulated data DM generated in step S20. Specifically, the CPU 102 sets a virtual circle CR centered on the vehicle 20T of the subject in the virtual world constructed by the simulated data DM, and has a diameter corresponding to the judged scene to be evaluated. Then, the CPU 102 judges that the scene to be evaluated exists when the following two conditions (a) and (b) are satisfied.

[0059] (i) While the subject of evaluation's vehicle 20T was traveling on the driving road R1, the recognized object moved from outside to inside the virtual circle CR. (b) The perceived object was present within the imaginary circle CR from the time the perceived object entered inside the imaginary circle CR until the response time had elapsed.

[0060] The response time corresponds to the scene to be evaluated determined above. If the scene to be evaluated does not exist (step S30: NO), CPU 102 advances the process to step S200. In this case, CPU 102 displays on display screen 150A of display device 150 that the designated scene to be evaluated does not exist. Then, the series of processes of the evaluation process is terminated.

[0061] On the other hand, in step S30, if an evaluation target scene exists (step S30: YES), CPU 102 specifies the timing at which condition (i) is satisfied as reference timing TD1 for subsequent processing. CPU 102 also specifies a scene whose existence is confirmed by satisfying both conditions (i) and (ii) as an evaluation target scene instructed by business entity 300. If a situation in which conditions (i) and (ii) are satisfied exists multiple times during the analysis target period, CPU 102 specifies the first occurrence as the evaluation target scene to be analyzed in subsequent processing. CPU 102 then proceeds to step S40.

[0062] In step S40, CPU 102 performs a recognition timing determination process. That is, CPU 102 determines recognition timing TD2, which is the timing when the person to be evaluated recognizes the evaluation object. In determining recognition timing TD2, CPU 102 synchronizes the eye line data DY with the simulation data DM. Then, CPU 102 determines, in the eye line data DY, the time later than the reference timing TD1 when the line of sight of the person to be evaluated turns toward the recognition object as recognition timing TD2. After this, CPU 102 advances the process to step S50.

[0063] In step S50, the CPU 102 performs a first pre-processing. That is, the CPU 102 calculates a recognition period A1P, which is an input variable to the mapping. When calculating the recognition period A1P, the CPU 102 refers to the reference timing TD1 specified in step S30 and the recognition timing TD2 specified in step S40. Then, the CPU 102 calculates the length of time from the reference timing TD1 to the recognition timing TD2 as the recognition period A1P. After this, the CPU 102 advances the process to step S60.

[0064] In step S60, the CPU 102 performs an ideal driving data generation process. Specifically, the CPU 102 performs a driving simulation of the vehicle 20 using the automatic driving model DU for the evaluation target scene identified in step S30. The CPU 102 calculates various input information required for performing this simulation based on the simulation data DM. For example, in the case of a merging scene, the various input information includes the positional relationship between the other vehicle 20H and the vehicle 20T of the evaluation target person, the vehicle speed SP of the other vehicle 20H, and the like. The CPU 102 generates the ideal driving data DI by performing the above simulation. As described above, the ideal driving data DI is a time series of various driving parameters that represent an ideal driving state in the evaluation target scene. After generating the ideal driving data DI, the CPU 102 advances the process to step S70.

[0065] In step S70, the CPU 102 performs a second pre-processing. Specifically, the CPU 102 calculates an acceleration difference integrated value A2P, a deceleration difference integrated value A3P, and a steering angle difference integrated value A4P, which are input variables to the mapping. When calculating these variables, the CPU 102 synchronizes the subject person driving data DA and the ideal driving data DI. Then, the CPU 102 calculates the acceleration difference integrated value A2P by integrating the absolute value of the difference between the ideal longitudinal acceleration AC and the actual longitudinal acceleration AC at each timing for a period in which the ideal longitudinal acceleration AC is equal to or greater than zero among the evaluation target scenes identified in step S30. In addition, the CPU 102 calculates the deceleration difference integrated value A3P by integrating the absolute value of the difference between the ideal longitudinal acceleration AC and the actual longitudinal acceleration AC at each timing for a period in which the ideal longitudinal acceleration AC is less than zero among a series of periods in the evaluation target scenes. In addition, the CPU 102 calculates the steering angle difference integrated value A4P by accumulating the absolute value of the difference between the ideal steering angle ST and the actual steering angle ST at each timing for the entire series of periods of the evaluation target scene. After calculating the acceleration difference integrated value A2P, the deceleration difference integrated value A3P, and the steering angle difference integrated value A4P, the CPU 102 advances the process to step S80.

[0066] In step S80, CPU 102 performs a selection process. Specifically, CPU 102 selects a map corresponding to the evaluation target scene identified in step S30 from among the four maps in map data DS. In effect, CPU 102 selects a map based on information input by business entity 300. After this, CPU 102 advances the process to step S90.

[0067] In step S90, as a pre-processing for using the mapping selected in step S80, CPU 102 assigns the values ​​of each variable acquired in the processing of steps S50 and S70 to input variables x(1) to x(4) for input to the mapping. Specifically, CPU 102 assigns recognition period A1P to input variable x(1). CPU 102 assigns acceleration difference integrated value A2P to input variable x(2). CPU 102 assigns deceleration difference integrated value A3P to input variable x(3). CPU 102 assigns steering angle difference integrated value A4P to input variable x(4). After this, CPU 102 advances the processing to step S100.

[0068] In step S100, the CPU 102 performs a calculation process. That is, the CPU 102 calculates an output variable y by inputting input variables x(1) to x(4) into the mapping selected in step S80. As described above, the output variable y is the evaluation score BP of the driving skill.

[0069] The mapping is 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 4) 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 hyperbolic tangent "tanh(x)". The neural network also includes output side coefficients wSj (j=0 to n) and 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. In this embodiment, the activation function f(x) is exemplified as hyperbolic tangent "tanh(x)". The value n indicates the dimension of the intermediate layer. The input side 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." In addition, the output coefficient wS0 is a bias parameter.

[0070] After executing the process of step S100, CPU 102 advances the process to step S110. In step S110, CPU 102 performs display processing. Specifically, CPU 102 generates a video signal that combines the subject's running data DA, the simulation data DM, the line of sight data DY, the subject's body data DC, and the evaluation score BP calculated in step S100. Then, CPU 102 outputs the video signal to display device 150. As a result, CPU 102 displays the evaluation result of the driving skill as an image on display screen 150A of display device 150. Specifically, CPU 102 displays five images, a first image E1, a second image E2, a third image E3, a fourth image E4, and a fifth image E5, on display screen 150A as shown in FIG.

[0071] The CPU 102 displays, as the first image E1, on the display screen 150A, the time change of each parameter before and after the scene to be evaluated identified in step S30. Specifically, the CPU 102 displays the movement of the line of sight of the person to be evaluated in a line graph with the horizontal axis representing the time axis TQ based on the line of sight data DY. The CPU 102 also displays the time change of each of the accelerator operation amount AP, the brake operation amount BK, the longitudinal acceleration AC, the vehicle speed SP, and the steering angle ST in a line graph with the horizontal axis representing the time axis TQ based on the subject's driving data DA. Note that each line graph is based on the same time axis TQ. In the example shown in FIG. 12, the first image E1 shows that the line of sight of the person to be evaluated moved at a recognition timing TD2 that is later than the reference timing TD1. The first image E1 also shows that the accelerator pedal was released at a response timing TD3 that is later than the recognition timing TD2, and that the longitudinal acceleration AC and the vehicle speed SP decreased accordingly. Then, it shows that the brake pedal was depressed thereafter. In the example of FIG. 12, the steering angle ST is constant.

[0072] The CPU 102 displays plots showing the reference timing TD1, the recognition timing TD2, and the response timing TD3 on the display screen 150A as the first image E1 in accordance with the time change of each parameter. Furthermore, the CPU 102 displays a plot showing a recognition period A1P, which is the length of time from the reference timing TD1 to the recognition timing TD2, on the display screen 150A. The CPU 102 also displays a plot showing a reaction period HD, which is the length of time from the recognition timing TD2 to the response timing TD3, on the display screen 150A. The CPU 102 also displays a display time plot TS, which is a plot showing a specific time on the time axis TQ, on the display screen 150A as the first image E1. This display time plot TS indicates the time shown by each image of the second image E2, the third image E3, and the fourth image E4, which will be described later.

[0073] The CPU 102 reproduces the situation at the recognition timing TD2 of the evaluation target scene on the display screen 150A based on the simulation data DM as the second image E2. At this time, the vehicle 20T of the evaluation target person may be surrounded by a circle FA, for example, so that the vehicle 20T of the evaluation target person can be distinguished from other vehicles 20H. Furthermore, with respect to the second image E2, the CPU 102 displays an arrow FB pointing in the direction of the line of sight of the evaluation target person on the image reproducing the situation at the recognition timing TD2 based on the line of sight data DY. In the example shown in FIG. 12, the evaluation target scene is a merging scene.

[0074] The CPU 102 displays, as the third image E3, on the display screen 150A, an image of the evaluation subject at the recognition timing TD2 based on the subject physical data DC. The CPU 102 displays, as the fourth image E4, the scenery seen from each window of the vehicle 20T of the subject at the perception timing TD2 based on the simulation data DM. Note that the windows are the left front window, the right front window, the left rear window, the right rear window, and the rear window.

[0075] CPU 102 displays, as the fifth image E5, the evaluation score BP of the driving skill calculated in step S100 on display screen 150A. After displaying the above content on display screen 150A, CPU 102 ends the series of steps in the evaluation process.

[0076] <Operation of the embodiment> The operator 300 of the car sharing service has the server 101 execute an evaluation process to evaluate the driving skill of the user. Then, the display device 150 displays the evaluation result on the display screen 150A. The operator 300 grasps the evaluation score BP of the driving skill in the evaluation target scene by looking at the fifth image E5 displayed on the display screen 150A. In addition, the operator 300 checks other images on the display screen 150A as reference information. For example, the operator 300 looks at the time changes of each element of the target person driving data DA and the line of sight data DY in the first image E1. In this way, the operator 300 grasps the driving operation situation before and after the evaluation target scene, such as speeding or inattentive driving. In addition, the operator 300 looks at a reproduced image of the evaluation target scene in the second image E2. In this way, the operator 300 grasps what kind of situation the vehicle 20T of the evaluation target person was actually in in the evaluation target scene. In addition, the operator 300 looks at an image of the evaluation target person in the third image E3. This allows the business operator 300 to understand, for example, the posture of the person being evaluated while driving. The business operator 300 also views the scenery from each window in the fourth image E4. This allows the business operator 300 to understand the scenery outside the vehicle as seen from inside the vehicle. Based on this information, the business operator 300 notifies the user, who is the person being evaluated, of information on the evaluation score BP of the driving skill and advice on improvements to driving operation, etc.

[0077] <Effects of the embodiment> (1) In this embodiment, the evaluation score BP of the driving skill is calculated using a mapping. Then, the recognition period A1P is included as one of the multiple input variables input to the mapping. In a situation where a recognition object exists while the vehicle 20 is traveling, the driver first recognizes the recognition object and then takes action against the recognition object. At this time, if the recognition object can be recognized as early as possible, the driver can perform driving operations with ample time thereafter. Therefore, the recognition period A1P, that is, the time required to recognize the recognition object, is a very important driving skill in a situation where the recognition object exists. Therefore, by including the recognition period A1P in the input variables, the reliability of the evaluation score BP of the driving skill can be guaranteed. In addition, when the mapping is used, if appropriate teacher data and training data can be prepared, an appropriate evaluation score BP can be easily obtained. Therefore, even if the relationship between the recognition period A1P and the evaluation score BP changes significantly due to input variables other than the recognition period A1P, an appropriate evaluation score BP can be obtained regardless of the experience and skill of the business operator 300.

[0078] (2) In order to properly evaluate driving skills in a situation where a recognized object is present, it is necessary to include in the evaluation factors the way in which the subject performs driving operations after recognizing the recognized object, that is, how to deal with the recognized object. Here, if the subject can accelerate the vehicle 20 at an appropriate timing as necessary after recognizing the recognized object, the vehicle 20 runs smoothly and the driving skill is high. In the above configuration, the acceleration difference integrated value A2P, which indicates the difference between the ideal acceleration state and the actual acceleration state, is included in the input variables. Therefore, as a way to deal with the recognized object, the evaluation score BP can be calculated by taking into account the driving skill related to the acceleration of the vehicle 20, that is, the skill related to the operation of the accelerator pedal.

[0079] (3) If the vehicle 20 can be decelerated at an appropriate timing as necessary in response to a recognized object, the vehicle 20 will travel smoothly and the driving skill will be high. In the above configuration, the deceleration difference integrated value A3P, which indicates the difference between the ideal deceleration state and the actual deceleration state, is included in the input variables. Therefore, the evaluation score BP can be calculated by taking into account the driving skill related to the deceleration of the vehicle 20, that is, the skill related to the operation of the accelerator pedal and the brake pedal, in response to a recognized object.

[0080] (4) If the driver can turn the vehicle 20 left or right at the appropriate timing as necessary to deal with the recognized object, the vehicle 20 will run smoothly and the driving skill will be high. In the above configuration, the steering angle difference integrated value A4P, which indicates the difference between the ideal turning state and the actual turning state, is included in the input variables. Therefore, the evaluation score BP can be calculated by taking into account the driving skill related to turning the vehicle 20, that is, the skill related to operating the steering wheel, as a measure against the recognized object.

[0081] (5) In a merging scene, a scene with a vehicle ahead, a scene with a sign, and a scene with a pedestrian, the time required to recognize the recognized object and the appropriate way to deal with the recognized object are different. Therefore, if the evaluation result of the driving skill is calculated for these multiple evaluation target scenes using the same criteria, that is, if the evaluation score BP of the driving skill is calculated using the same mapping, there is a risk that an appropriate evaluation score BP cannot be obtained. In this regard, in the present embodiment, a mapping is prepared for each evaluation target scene. Therefore, an appropriate evaluation score BP can be obtained for each evaluation target scene.

[0082] <Example of change> This embodiment can be modified as follows: This embodiment and the following modifications can be combined with each other to the extent that there is no technical contradiction.

[0083] The owner of the driving skill evaluation system 100 and the use of evaluating driving skills through the evaluation process 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.

[0084] The contents displayed on the display screen 150A in the display process are not limited to the examples in the above embodiment. Any index based on the evaluation variable B output from the mapping may be displayed on the display screen 150A. The contents displayed on the display screen 150A may be any information that is visually useful for the evaluator in evaluating the driving skill.

[0085] In the above embodiment, the position of the display time plot TS displayed on the display screen 150A may be made movable on the time axis TQ. In addition, it may be possible to instruct a change in the position of the display time plot TS via the input device 160. Then, an image at the instructed time may be displayed in the second image E2, the third image E3, and the fourth image E4.

[0086] The second image E2, the third image E3, and the fourth image E4 may be displayed as moving images. When displaying more than one of the second image E2, the third image E3, and the fourth image E4 as moving images, it is preferable that they are synchronized with each other over time. It is also preferable that the display time plot TS moves as time passes within the moving images.

[0087] Display processing is not essential. Instead of displaying the evaluation score BP of the driving skill calculated in step S100 on the display screen 150A, the evaluation score BP may be stored in the storage device 104.

[0088] The method of determining the analysis target period is not limited to the example of the above embodiment. As long as the period before and after the scene to be evaluated can be determined as the analysis target period, the method of determining the analysis target period is not important. The business entity 300 may determine the analysis target period in advance and input the period to the server 101 when the evaluation process is executed.

[0089] In the above embodiment, the evaluation score BP of the driving skill for one evaluation target scene is calculated for one evaluation process. However, it is also possible to calculate the evaluation score BP for each of multiple evaluation target scenes in one evaluation process. For example, by repeating the processes from step S30 to step S100 and identifying a different evaluation target scene in step S30 each time, it is possible to calculate the evaluation score BP for multiple evaluation target scenes in one evaluation process.

[0090] The contents of the input variables input to the mapping are not limited to the examples of the above embodiment. The contents of the processes from step S50 to step S70 may be changed as appropriate according to the contents of the input variables. For example, the variable adopted as the recognition variable A1 may be changed from the recognition period A1P. The recognition variable A1 may be a variable indicating the length of time from when the recognition object appears in the virtual circle CR until the evaluation subject recognizes the evaluation object. For example, the recognition distance LC indicating the distance between the vehicle 20T of the evaluation subject and the recognition object at the recognition timing TD2 may be adopted as the recognition variable A1. A long recognition distance LC means that the time from when the recognition object appears in the virtual circle CR until the evaluation subject recognizes the evaluation object is short.

[0091] For example, in the case of a merging scene, as shown in FIG. 13, the distance between the vehicle 20T of the subject and the reference position LB at the recognition timing TD2 may be set as the recognition distance LC. The reference position LB may be, for example, a point where the shoulders of the travel road R1 and the cross road R2 intersect. The recognition distance LC may be a length in a direction along the travel road R1. In calculating the recognition distance LC, the position coordinates PN of the subject's vehicle 20T and the position coordinates of the reference position LB may be specified based on the simulation data DM, and the distance between these coordinates may be calculated. When specifying the position coordinates PN of the subject's vehicle 20T, the subject's travel data DA may be referenced. Also, when specifying the reference position LB, the map data DF may be referenced. Note that the virtual circle CR is omitted from FIG. 13.

[0092] As in the merging scene, in the case of a scene where a preceding vehicle is present, as shown in FIG. 14, the distance between the vehicle 20T of the evaluation subject and another vehicle 20H at the recognition timing TD2 may be calculated as the recognition distance LC. Note that the imaginary circle CR is omitted in FIG. 14. Also, as shown in FIG. 14, in the scene where a sign is present, the distance between the vehicle 20T of the evaluation subject and a road sign RS at the recognition timing TD2 may be calculated as the recognition distance LC as the length in the direction along the traveling road R1. Also, as shown in FIG. 15, in the scene where a pedestrian is present, the distance between the vehicle 20T of the evaluation subject and a pedestrian 20M at the recognition timing TD2 may be calculated as the length in the direction along the traveling road R1.

[0093] The variable adopted as the acceleration difference variable A2 is not limited to the example of the above embodiment. The acceleration difference variable A2 is a variable indicating the amount of deviation between the time change of the longitudinal acceleration AC in an ideal driving state and the time change of the actual longitudinal acceleration AC, and may be a variable indicating the amount of deviation when the longitudinal acceleration AC in the ideal driving state is equal to or greater than zero. For example, the amount of change in the longitudinal acceleration AC per unit time is calculated at each timing for each of the ideal longitudinal acceleration AC and the actual longitudinal acceleration AC. Then, the difference between the ideal longitudinal acceleration AC and the actual longitudinal acceleration AC may be integrated with respect to the amount of change per unit time at each timing. In this case, the integration may be performed only when the longitudinal acceleration AC in the ideal driving state is equal to or greater than zero. The integrated value thus obtained may be adopted as the acceleration difference variable A2.

[0094] The variable used as the deceleration difference variable A3 is not limited to the example of the above embodiment. The deceleration difference variable A3 may be a variable indicating the amount of deviation between the time change of the longitudinal acceleration AC in an ideal driving state and the time change of the actual longitudinal acceleration AC, and may be a variable indicating the amount of deviation when the longitudinal acceleration AC in the ideal driving state is less than zero. For example, similar to the acceleration difference variable A2, the amount of deviation between the amount of change in the longitudinal acceleration AC per unit time between the ideal longitudinal acceleration AC and the actual longitudinal acceleration AC may be used as the deceleration difference variable A3.

[0095] The variable adopted as the steering angle difference variable A4 is not limited to the example of the above embodiment. The steering angle difference variable A4 may be any variable that indicates the amount of deviation between the time change of the steering angle ST in an ideal driving state and the time change of the actual steering angle ST. For example, the absolute value of the difference between the ideal steering angle ST and the actual steering angle ST may be integrated by distinguishing between the case where the vehicle 20T is turned to the left side from the driver's perspective and the case where the vehicle 20T is turned to the right side. Then, the integrated value when the vehicle 20T is turned to the left side from the driver's perspective and the integrated value when the vehicle 20T is turned to the right side may be set as the steering angle difference variable A4. For example, in a merging scene, when another vehicle 20H merges with the vehicle 20T of the subject of evaluation from the left side, the vehicle 20T of the subject of evaluation will change lanes to the right side as necessary. In evaluating the driving skill in such a situation where the vehicle is turned to the left or right, the steering angle difference variable A4 that distinguishes between the left and right as described above is also effective.

[0096] The types of input variables are not limited to those in the above embodiment. Other input variables may be used instead of or in addition to the input variables used in the above embodiment. There may be two or more input variables. One of the input variables may include the recognition variable A1. The number of input variables may be changed depending on the scene to be evaluated. For example, the vehicle speed SP, accelerator operation amount AP, and brake operation amount BK at the recognition timing TD2 may be used as input variables.

[0097] A difficulty variable may be adopted as an input variable. The difficulty variable is a variable indicating the difficulty of driving operation in the evaluation target scene. For example, in a merging scene, when another vehicle 20H merges from the cross road R2 to the traveling road R1, the other vehicle 20H may merge into the traveling road R1 ignoring the stop law, or the other vehicle 20H may suddenly accelerate. The difficulty of driving operation changes depending on such irregular events. Therefore, various events that may occur in a certain evaluation target scene are determined in advance, and a difficulty value that quantifies the difficulty of driving operation for each event is set. Then, the difficulty value may be adopted as the difficulty variable. When the difficulty value is adopted as an input variable, after identifying the evaluation target scene, the event that occurred in the evaluation target scene is further identified. Then, the difficulty value according to the event may be input into the mapping as an input variable. When the difficulty value is adopted as an input variable, basically, the mapping may be trained so that the higher the difficulty value, that is, the higher the difficulty of the turning operation, the higher the evaluation score BP of the driving skill.

[0098] A dedicated input variable may be adopted for each evaluation target scene. For example, a pedestrian variable may be adopted as an input variable dedicated to a pedestrian presence scene. The pedestrian variable is a variable indicating a response state to a pedestrian 20M when the vehicle 20T of the evaluation target passes beside the pedestrian 20M in a pedestrian presence scene. For example, as shown in FIG. 16, a separation distance LJ, which is a distance between the vehicle 20T of the evaluation target and the pedestrian 20M when the vehicle 20T and the pedestrian 20M are side by side, may be adopted as a pedestrian variable. In calculating the separation distance LJ, for example, the position coordinate PN of the vehicle 20T of the evaluation target and the position coordinate of the pedestrian 20M may be specified based on the simulation data DM. When the separation distance LJ is adopted as an input variable, a mapping may be basically learned so that the evaluation score BP of the driving skill increases as the separation distance LJ increases.

[0099] As the pedestrian variable, the vehicle speed SP of the vehicle 20T of the subject when the vehicle 20T of the subject is aligned side by side with the pedestrian 20M may be used. In this case, basically, a mapping may be learned so that the lower the vehicle speed SP, the higher the evaluation score BP of the driving skill.

[0100] The variable adopted as the evaluation variable B is not limited to the example in the above embodiment. For example, driving skills may be divided into a plurality of levels, and a value indicating such a level may be adopted as the evaluation variable B.

[0101] There may be multiple output variables. In other words, in addition to the evaluation variable B, other variables may be used as output variables. Any variable that has a relationship with the input variables may be used as an output variable.

[0102] As in the above example of changing evaluation variable B, the variable adopted as the input variable may be one that indicates a value obtained by dividing the value of a certain parameter into multiple levels. The same applies to the case where a variable other than evaluation variable B is adopted as the output variable.

[0103] In the above embodiment, a mapping is prepared for each scene to be evaluated. Instead of this form, only one mapping may be used in all scenes to be evaluated. Here, for example, between the recognition period A1P and the evaluation score BP of the driving skill, there is a relationship that the shorter the recognition period A1P, the higher the evaluation score BP, regardless of the difference in the scene to be evaluated. As with the above recognition period A1P, there is a common relationship between the input variables other than the recognition period A1P adopted in the above embodiment and the evaluation score BP in all scenes to be evaluated. In such a case, even if the same mapping is used for all scenes to be evaluated, a certain degree of accuracy can be expected for the output result of the mapping.

[0104] The configuration of the mapping is not limited to the example of the above embodiment. For example, a recurrent neural network may be used. The mapping may be any type that can output an appropriate evaluation result for the driving skill for the input variables used.

[0105] The scene to be evaluated is not limited to the example of the above embodiment. Instead of or in addition to the scene to be evaluated employed in the above embodiment, other scenes may be used as the scene to be evaluated. Furthermore, the number of scenes to be evaluated may be any number. The scene to be evaluated may be a scene in which a recognition object exists. The recognition object may be anything that the person to be evaluated needs to recognize while driving, such as a traffic light, a crosswalk, a railroad crossing, or a motorcycle.

[0106] The method of identifying the scene to be evaluated is not limited to the example of the above embodiment. For example, the target range may be set to a sector or a rectangle instead of a circle. The target range may be any suitable range for the scene to be identified. In addition, in identifying the scene to be evaluated, the condition that the object to be evaluated exists within the target range from the reference timing TD1 until the action time has elapsed may be eliminated. In other words, the recognized object may exist within the target range only for a part of the period identified as the scene to be evaluated. As long as a scene in which the recognized object exists within the target range can be identified and is based on the subject driving data DA, the method of identifying the scene to be evaluated does not matter.

[0107] The simulation data DM may not be used to identify the scene to be evaluated. For example, if video data consisting of a time series of external image information JM captured by the external camera 37 is used, the scene to be evaluated can be identified without using the simulation data DM.

[0108] The method of identifying the recognition timing TD2 is not limited to the example of the above embodiment. That is, the recognition timing TD2 may be identified using information other than the movement of the line of sight. The method of identifying the recognition timing TD2 may be based on the subject's physical data DC. As in the above embodiment, if video data consisting of a time series of the internal image capture information JN can be used as the subject's physical data DC, the change in the facial expression of the evaluation subject may be captured based on the video data. Then, the recognition timing TD2 may be determined based on a change in facial expression, such as a surprised expression. The movement of the evaluation subject may also be captured based on the video data. Then, the recognition timing TD2 may be determined based on the fact that a prior action has been taken to deal with the recognition object, such as changing the grip position of the steering wheel.

[0109] The subject's physical data DC is not limited to the above embodiment. Furthermore, the physical data D3 on which the subject's physical data DC is based is not limited to the above embodiment. In the above embodiment, the element of the physical data D3 was only the internal image information JN acquired by the in-vehicle camera 38. However, the physical data D3 may include multiple elements. The physical data D3 may be anything that shows the state of the driver's body when the vehicle 20 is driven in a real driving environment in the real world. For example, a thermograph that detects the driver's body temperature may be installed in the vehicle 20, and information on the driver's temperature acquired by the thermograph may be included as an element of the physical data D3. Also, for example, if the driver is wearing a device that can measure data such as pulse, blood pressure, respiratory rate, sweat rate, and blood alcohol concentration, these data may be received by wireless communication or the like and included as an element of the physical data D3. The subject's physical data DC may not include all the elements included in the physical data D3.

[0110] The subject driving data DA is not limited to the above embodiment. Furthermore, the driving data D2 on which the subject driving data DA is based is not limited to the above embodiment. The elements of the driving data D2 may be any number of elements as long as they include the transition of the position coordinate PN of the vehicle 20, and may also include any other elements. The driving data D2 may be data indicating the driving state of the vehicle 20 when the vehicle 20 is driven in a real driving environment in the real world. The driving data D2 may include, for example, the operation position of the shift lever and the rotational angular velocity around the vertical axis of the vehicle 20. In addition, for example, in the case of a hybrid vehicle, the driving data D2 may include the rotational speed of the motor generator and the rotational speed of the internal combustion engine. In addition, for example, the driving data D2 may include the oil temperature of the hydraulic oil of the transmission mechanism and the temperature of the battery. The subject driving data DA may not include all the elements included in the driving data D2.

[0111] Similarly to the subject person running data DA, the first other object data DB1 is not limited to the example in the above embodiment. The configuration of the vehicle 20 is not limited to the example of the above embodiment. The type of information acquisition device 30 mounted on the vehicle 20 is not limited to that shown in the above embodiment. For example, the vehicle 20 may be equipped with a radar that detects the situation around the vehicle 20 as the information acquisition device 30. Of the information acquired by each information acquisition device 30, necessary information may be transmitted to the server 101 as vehicle transmission data WP.

[0112] The second other object data DB2 is not limited to the example of the above embodiment. For example, the second other object data DB2 may include location information of a mobile terminal owned by a person who has been registered in advance. In this case, the server 101 may acquire location information of each mobile terminal. Then, the information may be included in the second other object data DB2.

[0113] The data used to generate the simulated data DM is not limited to those shown in the above embodiment. For example, the data used to generate the simulated data DM may include weather data. Any data may be used as long as the driving state of the subject's vehicle 20T in the actual driving environment can be simulated in the virtual world.

[0114] It is not necessary to generate the simulated data DM. As described in the above modification, the scene to be evaluated can be identified based on the external image capture information JM without using the simulated data DM. In addition, to calculate an input variable such as the recognition period A1P, the eye gaze data DY is sufficient, and the simulated data DM is not necessarily required.

[0115] The storage device 104 and the analysis device 107 may be configured as parts of separate processing devices. For example, as shown in FIG. 17, 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 in the above embodiment. The first processing device 501 receives vehicle transmission data WP from a plurality of vehicles 20. The first processing device 501 also receives road transmission data WR from a plurality of road cameras 80. The storage device 104 of the first processing device 501 stores the received data. 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 in the above embodiment. That is, the analysis device 107 of the second processing device 502 can execute the evaluation process. 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 the information stored in the storage device 104 of the first processing device 501 when performing the evaluation process.

[0116] 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]

[0117] 20…Vehicle 100...Driving Skills Evaluation System 104...Storage device 102…CPU 103...ROM 107…Analysis device

Claims

[Claim 1] A driving skill evaluation system for evaluating the driving skill of an evaluation subject, A storage device and an analysis device, The storage device prestores mapping data that specifies a mapping for outputting an evaluation variable, which is a variable indicating an evaluation of the driving skill of the evaluation subject, when a plurality of input variables are input, an autonomous driving model that is a program for causing a control device of the vehicle of the evaluation subject to autonomously drive the vehicle without operation by the evaluation subject, and map data that indicates position coordinates of roads; the storage device stores, over time, vehicle running data indicating a running state of the vehicle including a change in a position of the vehicle when the subject of evaluation drives the vehicle, body data of the subject of evaluation, and other object data regarding a change in a position of a moving object other than the vehicle of the subject of evaluation; The analysis device includes: A scene identification process that identifies a scene in which a recognition object that the evaluation subject needs to recognize is present within a predetermined target range including the vehicle during a period in which the evaluation subject drives the vehicle based on the driving data; a recognition timing specification process for specifying a timing at which the subject of evaluation recognizes the recognition object in the specified scene based on the body data; A first pre-processing step of calculating a recognition variable which is a variable indicating a length of time from when the recognition object appears within the target range until the evaluation subject recognizes the recognition object; an ideal driving data generation process for generating ideal driving data indicating a driving state of the vehicle when the vehicle of the evaluation subject is assumed to be driven autonomously by the control device under the same circumstances as the identified scene, based on the map data, the driving data, and the other object data, by a driving simulation using the automatic driving model; A second pre-processing step of calculating a difference variable which is a variable indicating a deviation between a state of the vehicle indicated by the ideal driving data and a state of the vehicle indicated by the running data; and executing a calculation process of calculating the evaluation variable by inputting a plurality of the input variables including the cognitive variable and the difference variable into the mapping. Driving skills evaluation system.

Images