Data storage method, data storage device and storage medium

Abstract

To provide a data storage method that enables effective learning while suppressing the amount of data of driving data to be stored.SOLUTION: A data storage method executed by a processor 46 to store driving data, which is acquired while a vehicle is driving, in a storage medium 43 includes: acquiring the driving data; and storing the acquired driving data in the storage medium in association with a scenario defined based on a scene classification item and a vehicle behavior classification item, which are used to specify a driving scene. A plurality of the scenarios are classified in a tree structure based on at least one of the scene classification item and the vehicle behavior classification item.SELECTED DRAWING: Figure 8

Description

[Technical Field]

[0001] The present disclosure relates to a data storage method, a data storage device, and a storage medium for storing data used in learning vehicle control. [Background technology]

[0002] Patent Document 1 discloses a technology for collecting vehicle parameters and parameters related to the external environment in accordance with a collection policy for a target scenario. Patent Document 1 also discloses generating an automatic control model based on the collected parameters. The contents of the prior art documents are incorporated by reference as explanations of the technical elements in this specification. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] U.S. Patent No. 11,400,952 Summary of the Invention [Problem to be solved by the invention]

[0004] If data collected while a vehicle is traveling (hereinafter referred to as "in-travel data") is collected from a large number of vehicles for use in learning autonomous driving control, the amount of data will be enormous, and the cost of storage media for storing the data will become an issue. Therefore, it is necessary to reduce the amount of in-travel data to be stored.

[0005] As disclosed in Patent Document 1, when collecting data according to a collection policy for a target scenario, it may be possible to reduce the amount of data to be stored by reducing the number of target scenarios or making it more difficult for the conditions of the target scenario to be met. Patent Document 1 lists examples of target scenarios, such as a rainy weather scenario, a parking scenario, a lane change scenario, a vehicle-following scenario, an urban scenario, and a highway scenario. It also discloses creating hybrid scenarios, such as a highway scenario in rainy weather. However, the target scenarios are merely examples. For effective learning, it is desirable to acquire sufficient driving data for actual scenarios that occur when the vehicle is actually driving.

[0006] The present disclosure has been made based on this situation, and its purpose is to provide a data storage method, data storage device, and storage medium that enable effective learning while making it possible to reduce the amount of driving data to be stored. [Means for solving the problem]

[0007] The above object is achieved by the combination of features recited in the independent claims, and the subclaims define further advantageous specific examples. The reference numerals in parentheses in the claims correspond to specific aspects described in the following embodiments as one aspect, and do not limit the technical scope of the disclosure.

[0008] One disclosure for achieving the above object is a data storage method executed by a processor (46) to store in a storage medium (43) driving data collected while the vehicle is driving, the data storage method acquiring driving data, storing the acquired driving data in the storage medium in association with scenarios defined based on scene classification items and vehicle behavior classification items for identifying driving scenes, and classifying multiple scenarios in a tree-like manner based on at least one of the scene classification items and vehicle behavior classification items.

[0009] Another disclosure for achieving the above object is a data storage device (45) that stores in-travel data collected while a vehicle is traveling in a storage medium (43), the data storage device including a processor (46), the processor acquiring the in-travel data and storing the acquired in-travel data in the storage medium in association with a scenario defined based on scene classification items and vehicle behavior classification items for identifying a traveling scene, the data storage device classifying multiple scenarios in a tree structure based on at least one of the scene classification items and vehicle behavior classification items.

[0010] Since the multiple scenarios are classified in a tree structure, the multiple scenarios are less likely to be insufficient or overlap with the actual scenarios that occur when the vehicle is actually traveling. Therefore, the data storage method and the data storage device described above enable effective learning while suppressing the amount of data stored during traveling.

[0011] Another disclosure for achieving the above object is a storage medium in which driving data acquired while a vehicle is driving is stored in association with one of a plurality of scenarios defined based on a plurality of types of scene classification items and vehicle behavior classification items for identifying driving scenes, and the plurality of scenarios form a scenario tree defined based on at least one of the scene classification items and vehicle behavior classification items. This storage medium is a storage medium in which driving data is stored by the above data storage method. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a diagram showing the overall configuration of a data collection system that collects data while traveling. [Figure 2] FIG. 1 shows the configuration of the development vehicle system. [Figure 3] FIG. 1 is a diagram showing the configuration of a mass-produced vehicle system. [Figure 4] FIG. 1 is a diagram showing the configuration of a processing system. [Figure 5] FIG. 4 is a diagram illustrating scene classification items. [Figure 6] FIG. 1 is a diagram illustrating a scenario tree. [Figure 7] FIG. 1 is a diagram illustrating a tree-shaped folder structure. [Figure 8] FIG. 4 is a diagram illustrating the procedure of a method for storing data during driving. [Figure 9] FIG. 9 is a diagram specifically illustrating the data reduction procedure of FIG. 8. [Figure 10] FIG. 9 is a diagram specifically showing the data storage procedure of FIG. 8. [Figure 11] FIG. 4 is a diagram showing an in-vehicle system update process. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, several embodiments will be described based on the drawings. Note that corresponding components in each embodiment and modification are designated by the same reference numerals, and redundant description may be omitted. When only a portion of the configuration is described in each embodiment and modification, the configurations of other previously described embodiments and modifications may be applied to the remaining portions of the configuration. Furthermore, in addition to the combinations of configurations explicitly stated in the description of each embodiment and modification, configurations of several embodiments and modifications may be partially combined together even if not explicitly stated, as long as there is no particular problem with the combination. Furthermore, combinations of configurations described in several embodiments and modifications that are not explicitly stated are also considered to be disclosed by the following description.

[0014] 1 is a diagram showing the overall configuration of a data collection system 10 that collects data while driving. The data collection system 10 includes a development vehicle system 20 provided in a development vehicle 11, a mass-production vehicle system 30 provided in a mass-production vehicle 13, and a processing system 40 provided in a data center 15.

[0015] The development vehicle 11 and the mass-produced vehicle 13 collect driving data while they are traveling. The development vehicle 11 is capable of collecting driving data with a larger amount of information than the mass-produced vehicle 13.

[0016] The driving data includes multiple types of data. The driving data includes scene identification data for identifying the driving scene and behavior data that indicates the vehicle behavior of the vehicle collecting the driving data (hereinafter referred to as the data collection vehicle). The driving scene represents the surrounding conditions of the data collection vehicle. The scene identification data is data for identifying the driving scene, and therefore represents the surrounding conditions of the data collection vehicle.

[0017] [Configuration of development vehicle system 20] The development vehicle system 20 provided in the development vehicle 11 is a system for collecting detailed driving data. As shown in Fig. 2, the development vehicle system 20 includes a sensor group 21, a control device 22, a communication device 23, and a storage medium 24.

[0018] The sensor group 21 includes a plurality of sensors that acquire information outside the vehicle and the behavior of the development vehicle 11. The sensor group 21 can be equipped with one or more of a camera, LiDAR, radar, and sonar as sensors that acquire information outside the vehicle. Because the driving scene can be identified from information outside the vehicle, the sensors that acquire information outside the vehicle are sensors that acquire scene-identifying data.

[0019] The sensor group 21 may include one or more of an acceleration sensor, an angular velocity sensor, a speed sensor, and a steering angle sensor as sensors for acquiring the behavior of the data collection vehicle. The sensors for acquiring the behavior of the data collection vehicle are sensors that acquire behavior data included in the driving data. Since changes in the position of the data collection vehicle indicate the behavior of the data collection vehicle, the behavior of the data collection vehicle may include the position of the data collection vehicle. The sensor group 21 may include a device for acquiring the position of the data collection vehicle, such as a GNSS receiver. The sensor group 21 may be more accurate than the sensors included in the mass-produced vehicle system 30. The sensor group 21 may also be a dedicated sensor for acquiring driving data as learning data.

[0020] The control device 22 can be realized by a computer including a processor 25 and a memory 26. The memory 26 stores a program to be executed by the processor 25. The processor 25 controls the sensor group 21, the communication device 23, and the storage medium 24 by executing the program stored in the memory 26.

[0021] The control device 22 acquires in-travel data and stores it in the storage medium 24. The specific types of data that the control device 22 acquires as in-travel data are set in advance by an operator or the like operating the processing system 40. The behavior data included in the in-travel data may include control parameters for controlling the behavior of the data collection vehicle. The specific driving conditions under which the development vehicle system 20 collects in-travel data are determined by development vehicle trigger conditions. The control device 22 collects in-travel data based on the establishment of the development vehicle trigger conditions. An example of a development vehicle trigger condition is the occurrence of any one of sudden braking, sudden acceleration, and sudden steering. The development vehicle trigger conditions can be set by an operator or the like of the processing system 40. The development vehicle trigger conditions set by an operator or the like of the processing system 40 will be described later.

[0022] The storage medium 24 is computer-readable. This storage medium 24 may be non-volatile or volatile. The development vehicle system 20 provides the driving data stored in the storage medium 24 to the processing system 40. The communication device 23 can be used to provide the driving data to the processing system 40. The control device 22 may wirelessly transmit the driving data stored in the storage medium 24 from the communication device 23 to the processing system 40. Alternatively, the storage medium 24 may be portable, and the storage medium 24 may be transported to the data center 15 and connected to the processing system 40 by wire, thereby providing the driving data to the processing system 40. Alternatively, the storage medium 24 may be connected to another portable storage medium, the driving data may be transferred to the other storage medium, and the other storage medium may be connected to the processing system 40 by wire, thereby providing the driving data to the processing system 40.

[0023] [Configuration of mass-produced vehicle system 30] The mass-produced vehicle system 30 provided in the mass-produced vehicle 13 is also a system for collecting data while the vehicle is traveling. As shown in FIG. 3, the mass-produced vehicle system 30 includes a sensor group 31, a control device 32, a communication device 33, and a storage medium .

[0024] The sensor group 31 includes a plurality of sensors that acquire information outside the vehicle and the behavior of the mass-produced vehicle 13. The sensors that acquire information outside the vehicle and the sensors that acquire the behavior of the mass-produced vehicle 13 that are included in the sensor group 31 may be of the same type as or different from those included in the sensor group 21. The sensor group 31 may also be used when the mass-produced vehicle 13 performs driving control.

[0025] The control device 32 can be realized by a computer including a processor 35 and a memory 36. The memory 36 stores a program to be executed by the processor 35. The processor 35 controls the sensor group 31, the communication device 33, and the storage medium 34 by executing the program stored in the memory 36.

[0026] The control device 32 acquires driving data and stores it in the storage medium 34. The specific type of data that the control device 32 acquires as driving data is set in advance by an operator or the like operating the processing system 40. The driving conditions under which the control device 32 acquires driving data are determined by the mass-produced vehicle trigger conditions. The control device 32 collects driving data based on the mass-produced vehicle trigger conditions being satisfied. An example of a mass-produced vehicle trigger condition is the occurrence of one of sudden braking, sudden acceleration, and sudden steering. The mass-produced vehicle trigger conditions can be set by an operator or the like of the processing system 40. The mass-produced vehicle trigger conditions set by an operator or the like of the processing system 40 will be described later. The mass-produced vehicle trigger conditions and the development vehicle trigger conditions may be collectively referred to as trigger conditions. The trigger conditions are data collection conditions under which the data collection vehicle collects driving data.

[0027] The storage medium 34 is computer-readable and may be non-volatile or volatile. The mass-produced vehicle system 30 also provides the driving data stored in the storage medium 34 to the processing system 40. The control device 32 may wirelessly transmit the driving data stored in the storage medium 34 to the processing system 40 from the communication device 33. Alternatively, the storage medium 34 or another storage medium onto which the driving data has been transferred from the storage medium 34 may be transported to the data center 15 and connected to the processing system 40 by wire, thereby providing the driving data to the processing system 40.

[0028] [Configuration of processing system 40] 4, the processing system 40 includes an information presentation device 41, an input device 42, a storage medium 43, a communication device 44, and a control device 45. The information presentation device 41 is a device that presents various information to an operator who operates the processing system 40. The information presented by the information presentation device 41 includes visual information. To present the visual information, the information presentation device 41 includes a display device.

[0029] The input device 42 is a device through which an operator inputs instructions to the processing system 40. The input device 42 may be a keyboard, a mouse, or the like.

[0030] The storage medium 43 is computer-readable and stores driving data acquired from the development vehicle system 20 and the production vehicle system 30. The driving data stored in the storage medium 43 is associated with a scenario. A scenario is defined by a combination of a driving scene and a vehicle behavior. A driving scene is specified by one or more scene classification items, and a vehicle behavior is specified by one or more vehicle behavior classification items.

[0031] Scene classification items are items for classifying driving scenes. Scene classification items will be explained using FIG. 5. As shown in FIG. 5, driving scenes can be classified by static information, environmental information, and dynamic information. These static information, environmental information, and dynamic information are classification items of higher concepts. All other words shown in FIG. 5 are examples of scene classification items.

[0032] Static information is information about objects that can be considered not to change over time. An example of static information is road shapes. Road shapes include, for example, straight lines, curves, and intersections. Static information may also include information about structures around the road and information indicating road markings. Traffic rules may also be included in the static information.

[0033] The dynamic information includes information for classifying or identifying moving objects around the data collection vehicle. FIG. 5 shows examples of information for classifying moving objects, such as road moving objects and surrounding moving objects. Road moving objects are moving objects that exist on roads, such as automobiles, motorcycles, bicycles, pedestrians, and animals. These may be further classified. For example, automobiles may be classified into passenger cars, buses, emergency vehicles, and the like. Objects lying on the road may also be included in the road moving objects.

[0034] A nearby moving object is a moving object that exists around the road. Examples of nearby moving objects include vehicles and people. Note that a moving object can be mobile, and of course it can also be stationary. Furthermore, objects lying on the road or around the road do not need to move themselves, but can be moved, so they may be treated as dynamic information. Also, fallen objects may be included in static information. Almost all road moving objects can move around the road. Therefore, nearby moving objects may be classified in the same way as road moving objects.

[0035] Environmental information includes weather, sunlight intensity, etc., and although it changes over time, the change in time during driving is not as great as that of dynamic information. Weather may be sunny, rainy, snowy, etc. Sunshine intensity may be classified into specific numerical ranges or may be classified qualitatively, such as daytime or nighttime.

[0036] As shown in Fig. 5, static information, environmental information, and dynamic information are each classified into more specific categories at one or more levels below each other. In other words, static information, environmental information, and dynamic information are each classified in a tree structure.

[0037] A driving scene is identified by one or more of static information, environmental information, and dynamic information. Whether all, one, or two of the static, environmental, and dynamic information are used depends on how specifically the driving scene is identified. Furthermore, a driving scene may be identified by using multiple classification items for each of the static information, environmental information, and dynamic information.

[0038] Next, we will explain the vehicle behavior classification items, which are the other item that defines a scenario. The vehicle behavior classification items are items that classify the vehicle behavior of the data collection vehicle. The vehicle behavior classification items can include items that classify how the data collection vehicle traveled on the road. For example, examples of vehicle classification items include "the data collection vehicle followed the road," "the data collection vehicle stopped," "the data collection vehicle turned right or left," "the data collection vehicle made an evacuation maneuver," and "the data collection vehicle parked." Like the scene classification items, the vehicle behavior classification items may also be classified in a tree structure.

[0039] Since at least one of the scene classification items and the vehicle behavior classification items is classified in a tree-like manner, the scenarios are also classified in a tree-like manner, i.e., the scenarios form a scenario tree as shown in Fig. 6.

[0040] As described above, the driving data is stored in association with a scenario. More specifically, as shown in Fig. 7, the driving data is stored in folders 70 for each scenario. Each folder 70 is associated with a scenario. Since the scenarios are categorized in a tree structure, the folders 70 that store the driving data for each scenario are also categorized in a tree structure.

[0041] The communication device 44 communicates wirelessly with the communication device 23 included in the development vehicle system 20 and the communication device 33 included in the mass-produced vehicle system 30 .

[0042] The control device 45 can be realized by a computer including a processor 46 and a memory 47. The memory 47 stores a program executed by the processor 46. The processor 46 controls the information presentation device 41, the input device 42, the storage medium 43, and the communication device 44 by executing the program stored in the memory 47. Since the control device 45 controls the storage medium 43, it corresponds to a data storage device. In the following description, the statement that the control device 45 executes can be read as the processor 46 executes.

[0043] [How to store data while driving] Next, a method for storing data during travel using the processing system 40 will be described with reference to Fig. 8. In S10, the control device 45 acquires the data during travel collected by the data collection vehicle and stores it temporarily in the memory 47.

[0044] In S20, data organization is performed on the driving data acquired in S10. The procedure of S20 is specifically shown in FIG. 9. In FIG. 9, in S21, the control device 45 performs scene analysis on the driving data acquired in S10. In the scene analysis, the driving scene indicated by the driving data acquired in S10 is analyzed. In the scene analysis, scene identification data included in the driving data is used. In the scene analysis, objects and the environment around the data collection vehicle are recognized. In object recognition, in addition to the position and size of the object, the type of object may be identified. A large language model (LLM) may be used for this recognition. In addition, in S21, the vehicle behavior of the data collection vehicle may be identified based on the behavior data.

[0045] In the next step S22, the control device 45 determines whether annotation by an operator is necessary. For example, if the number of objects whose types cannot be recognized in the processing of S21 is equal to or greater than a predetermined number, or if the ratio of objects whose types cannot be recognized is equal to or greater than a predetermined ratio, it may be determined that annotation by an operator is necessary. If the determination result in S22 is YES, the process proceeds to S23. In S23, the control device 45 outputs, i.e., notifies, from the information presentation device 41, information indicating that annotation by an operator is necessary.

[0046] In S24, the operator, who has recognized from the notification in S23 that annotation is necessary, performs the annotation. Then, the process proceeds to S25. On the other hand, if the determination result in S22 is NO, the process proceeds to S25 without going through S23 and S24.

[0047] In S25, the control device 45 anonymizes the driving data obtained by the scene analysis in S21. The anonymization is performed by processing the portion of the driving data that relates to personal information so that the personal information cannot be identified.

[0048] After the data reduction in S20 is performed, S30 shown in Fig. 8 is performed, that is, the data storage process is performed. The procedure of the data storage process is specifically shown in Fig. 10.

[0049] 10, in S31, the control device 45 identifies a scenario indicated by the running data analyzed in S21 based on the analysis result in S21. Furthermore, the control device 45 determines whether the identified scenario exists in the scenario corresponding to the folder 70 in which the running data is saved in the storage medium 43. If the determination result in S31 is YES, the process proceeds to S32, and if the determination result in S31 is NO, the process proceeds to S33.

[0050] In addition, if the driving scene is a scene that has not been identified before, or if the vehicle behavior is a behavior that has not been identified before, it may not be possible to identify the scenario. Even if the scenario cannot be identified, the determination result in S31 will be NO.

[0051] In S32, the control device 45 stores the driving data analyzed in S21 in the folder 70 of the storage medium 43 that corresponds to the scenario identified in S31. Then, the process proceeds to S36.

[0052] It should be noted that an upper limit may be set for the amount of data stored in folder 70. In other words, an upper limit may be set for the amount of driving data stored in association with one scenario. The driving data stored in folder 70 is learning data. It may be found that a certain amount of learning data allows learning to be performed with sufficient accuracy. Therefore, an upper limit may be set for the amount of data. This upper limit may be a different value for each folder.

[0053] When the folder 70 for storage has been determined, but the amount of running data stored in that folder 70 has reached an upper limit, past running data such as the oldest running data can be deleted and the current running data can be stored. Also, the current running data may not be stored in the folder 70.

[0054] If the determination result in S32 is NO, S33 is executed. In S33, the control device 45 causes the information presentation device 41 to output to the operator information indicating that the scenario indicated by the current driving data does not correspond to the scenario corresponding to the driving data already stored in the storage medium 43.

[0055] In the next step S34, the operator analyzes the driving data and modifies some of the scenarios that make up the scenario tree. For example, the operator can add a scenario to the scenario tree. Alternatively, or in addition, the operator may modify the classification items that identify existing scenarios. The operator modifies some of the scenarios that make up the scenario tree so that a scenario corresponding to the current driving data exists. This generates a folder 70 in which to save the current driving data.

[0056] In S35, the current driving data is stored in the folder 70 corresponding to the scenario indicated by the driving data, either automatically or based on the operation of the operator.

[0057] In S36, the control device 45 automatically or the operator determines whether to modify one or both of the development vehicle trigger conditions and the production vehicle trigger conditions. For example, if a scenario is modified or a scenario is added, the trigger conditions may be modified to make it easier to collect driving data corresponding to the modified or added scenario.

[0058] Immediately after a scenario is modified or added, there is often a lack of in-vehicle data corresponding to that scenario. A trigger condition that makes it easy to collect in-vehicle data corresponding to a modified or added scenario can be said to be a trigger condition for collecting in-vehicle data corresponding to a scenario with insufficient data.

[0059] Furthermore, even if the scenario tree is not modified, one or both of the development vehicle trigger conditions and the production vehicle trigger conditions may be modified. For example, when the amount of data stored in one folder 70 reaches a predetermined amount of data sufficient for learning the corresponding scenario, one or both of the development vehicle trigger conditions and the production vehicle trigger conditions may be modified to suppress the collection of running data corresponding to that scenario. The modified development vehicle trigger conditions and the production vehicle trigger conditions can be said to be trigger conditions that suppress the data collection vehicle from collecting running data corresponding to a scenario for which there is sufficient running data stored in the storage medium 43. The modified trigger conditions may be determined automatically or by operator operation.

[0060] Another example of trigger condition modification is described below. When a new scenario is added, for example, and the amount of data stored in the folder 70 corresponding to that scenario is small, the development vehicle trigger conditions can first be modified so that the development vehicle system 20 collects driving data corresponding to that scenario. Then, after a certain amount of driving data corresponding to that scenario has been stored, the production vehicle trigger conditions are modified so that the production vehicle system 30 collects driving data corresponding to that scenario. At this time, the development vehicle trigger conditions may be modified so that the development vehicle system 20 suppresses collection of driving data corresponding to that scenario.

[0061] In S37, if it is determined in S36 that the development vehicle trigger conditions need to be modified, the control device 45 outputs the modified development vehicle trigger conditions to the communication device 44. The communication device 44 transmits the modified development vehicle trigger conditions to the communication device 23 of the development vehicle system 20. Furthermore, if it is determined in S36 that the production vehicle trigger conditions need to be modified, the control device 45 outputs the modified production vehicle trigger conditions to the communication device 44. The communication device 44 transmits the modified production vehicle trigger conditions to the communication device 33 of the production vehicle system 30.

[0062] [In-vehicle system update processing] Next, the in-vehicle system update process will be described. The in-vehicle system update process is a process for updating a machine learning model used in a system that controls the autonomous driving of mass-produced vehicles 13. Fig. 11 shows the in-vehicle system update process.

[0063] In S40, machine learning is performed using the driving data stored in the storage medium 43 as learning data. In S50, model compression and optimization processing is performed on the machine learning model generated by learning in S40. Specific processing includes, for example, pruning, quantization, and distillation. In S60, system verification and validation are performed. Thereafter, in S70, the in-vehicle system is updated.

[0064] [Summary of implementation] In the present embodiment described above, the control device 45 acquires in-travel data (S10). Then, the acquired in-travel data is stored in a folder 70 corresponding to the scenario indicated by the in-travel data among a plurality of scenarios (S32, S35). In this way, the acquired in-travel data is stored in association with the scenario.

[0065] The scenarios are classified in a tree structure based on scene classification items and vehicle behavior classification items, so that the multiple scenarios are unlikely to be insufficient or overlap with the actual scenarios that occur when the vehicle is actually traveling. Therefore, by storing the traveling data in association with the scenarios classified in this way, it is possible to reduce the amount of traveling data to be stored while enabling effective learning.

[0066] It is also possible to set an upper limit on the amount of data to be stored in each folder 70. In other words, it is also possible to set an upper limit on the amount of driving data to be stored in association with one scenario. In this way, it is possible to suppress the amount of data to be stored while suppressing a decrease in learning accuracy.

[0067] Furthermore, if the scenario corresponding to the acquired driving data does not correspond to the scenario corresponding to the created folder 70 (S31: NO), information indicating that there is no corresponding scenario is output to the operator (S33). This indicates that the scenario needs to be revised. Therefore, the scenario can be revised promptly.

[0068] The control device 45 modifies the trigger conditions to prevent the data collection vehicle from collecting driving data corresponding to a scenario for which there is sufficient driving data stored in the storage medium 43. This not only reduces the storage capacity of the storage medium 43 included in the processing system 40, but also reduces the storage capacity of the storage media 24, 34 installed in the data collection vehicle.

[0069] On the other hand, the control device 45 modifies the trigger conditions so that the data collection vehicle can easily collect the in-travel data corresponding to a scenario for which there is a shortage of in-travel data stored in the storage medium 43. This allows the in-travel data corresponding to the scenario for which there is a shortage of data to be collected more quickly.

[0070] Furthermore, in this embodiment, data is collected during travel by the development vehicle 11 and the mass-produced vehicle 13, and it is possible to separately set a development vehicle trigger condition, which is the condition under which the development vehicle 11 collects data during travel, and a mass-produced vehicle trigger condition, which is the condition under which the mass-produced vehicle 13 collects data during travel. This makes it possible to deal with both situations where detailed data during travel is required and situations where a large amount of data during travel, but not detailed, is required.

[0071] In this embodiment, the control device 45 performs a scene analysis (S21). Based on the results of the scene analysis, the control device 45 then identifies a scenario indicated by the traveling data and determines whether the identified scenario exists in the scenario corresponding to the folder 70 (S31). This reduces the operator's effort in determining the scenario of the traveling data.

[0072] The above-mentioned scene analysis (S21) includes an object recognition process, and the control device 45 determines whether annotation by the operator is necessary based on whether an object that cannot be recognized exists in the scene analysis (S22). This reduces the operator's effort in determining which driving scenes require annotation.

[0073] Although the embodiments have been described above, the disclosed technology is not limited to the above-described embodiments, and the following modifications are also included within the scope of the disclosure. Furthermore, various modifications other than those described below can be implemented without departing from the spirit of the invention. In the following description, elements having the same reference numerals as those used up to that point are the same as the elements having the same reference numerals in the previous embodiments, unless otherwise specified. Furthermore, when only a portion of the configuration is described, the previously described embodiment can be applied to the other portions of the configuration.

[0074] (Variation 1) In the embodiment, in S31, a binary decision is made as to whether or not a corresponding scenario exists. However, there may be cases where it is difficult to automatically decide whether or not a corresponding scenario exists.

[0075] Therefore, the control device 45 may output the acquired driving data and candidate scenarios to be associated with the driving data to the operator from the information presentation device 41 as images, etc. In this way, even if the accuracy of the determination of the correspondence between the driving data and the scenario is low, the operator can decide whether to store the driving data in association with the presented scenario as a candidate. Therefore, the accuracy of the association between the driving data and the scenario is improved for the driving data stored in the storage medium 43.

[0076] Candidate scenarios are determined, for example, as follows: Classification items identified from the driving data are compared with classification items defining the scenarios, and scenarios whose matching rate is equal to or greater than a predetermined rate can be selected as candidates.

[0077] Furthermore, the control device 45 may output, for all acquired driving data, the driving data and candidate scenarios to be associated with the driving data from the information presentation device 41. Alternatively, for only driving data for which the accuracy of determining the correspondence with a scenario is low among the acquired driving data, the driving data and candidate scenarios to be associated with the driving data may be output from the information presentation device 41. The determination accuracy is determined, for example, by comparing the classification items identified from the driving data with the classification items that define the scenarios, and based on the degree of match.

[0078] (Variation 2) In the embodiment, the storage medium 43 is provided in the data center 15 together with the control device 45. However, the storage medium 43 may be provided outside the data center 15. That is, the storage medium 43 may be provided in a building different from that of the control device 45. Cloud storage may also be used as the storage medium 43.

[0079] (Variation 3) The control device and the method described herein may be implemented by a special-purpose computer having a processor programmed to perform one or more functions embodied in a computer program. Alternatively, the control device and the method described herein may be implemented by a special-purpose hardware logic circuit. Alternatively, the control device and the method described herein may be implemented by one or more special-purpose computers configured by a combination of a processor executing a computer program and one or more hardware logic circuits. The hardware logic circuit may be, for example, an ASIC or an FPGA.

[0080] Furthermore, the storage medium for storing the computer program is not limited to a ROM, and the program may be stored in any computer-readable, non-transitory storage medium as instructions to be executed by a computer. For example, the program may be stored in a flash memory.

[0081] Furthermore, a fleet trigger condition may be set in place of or in addition to the development vehicle trigger condition and the mass-produced vehicle trigger condition. The fleet vehicles to which the fleet vehicle trigger condition is applied here are fewer in number than the development vehicles 11, and are capable of collecting driving data with a larger amount of information than the mass-produced vehicles 13. Examples of fleet vehicles include manned or unmanned taxis and buses.

[0082] (Disclosure of technical ideas) This specification discloses multiple technical ideas described in the following multiple clauses. Some clauses may be written in a multiple dependent form, with the subsequent clause referring to the preceding clause as an alternative. Furthermore, some clauses may be written in a multiple dependent form, referring to another multiple dependent clause. These multiple dependent clauses define multiple technical ideas.

[0083] (Technical thought 1) A data storage method executed by a processor (46) for storing driving data collected while a vehicle is driving in a storage medium (43), comprising: Acquire the driving data, The acquired driving data is stored in the storage medium in association with a scenario defined based on scene classification items for identifying a driving scene and vehicle behavior classification items; the plurality of scenarios are classified in a tree-like manner based on at least one of the scene classification items and the vehicle behavior classification items; Data storage method.

[0084] (Technical thought 2) An upper limit is set for the amount of data during driving that is stored in association with one of the scenarios. A data storage method according to Technical Idea 1.

[0085] (Technical Thought 3) outputting information indicating that there is no corresponding scenario based on the fact that the scenario corresponding to the acquired running data does not correspond to the scenario corresponding to the running data already stored in the storage medium; A data storage method according to Technical Idea 1 or 2.

[0086] (Technical Thought 4) outputting the acquired traveling data and the scenarios that are candidates to be associated with the traveling data to an operator who operates the processor; A data storage method according to any one of Technical Ideas 1 to 3.

[0087] (Technical Thought 5) determining the scenario that is filled with the running data stored in the storage medium; determining a data collection condition that inhibits the vehicle from collecting the traveling data corresponding to the determined scenario; outputting the determined data collection conditions for provision to the vehicle; A data storage method according to any one of Technical Ideas 1 to 4.

[0088] (Technical Thought 6) determining the scenario in which the running data stored in the storage medium is insufficient; determining a data collection condition for the vehicle to collect the traveling data corresponding to the determined scenario; outputting the determined data collection conditions for provision to the vehicle; A data storage method according to any one of Technical Ideas 1 to 5.

[0089] (Technical Thought 7) The data storage method according to Technical Idea 5 or 6, in which the data collection conditions are determined by dividing them into conditions for mass-produced vehicles (13) and conditions for development vehicles (11) from which the driving data can be acquired, which has a larger amount of information than the mass-produced vehicles.

[0090] (Technical Thought 8) performing a scene analysis on the acquired driving data to determine whether the acquired driving data corresponds to any one of a plurality of scenarios corresponding to the driving data stored in the storage medium; A data storage method according to any one of Technical Ideas 1 to 7.

[0091] (Technical Thought 9) the scene analysis includes processing for recognizing objects around the vehicle based on the driving data; determining whether an annotation by an operator is required based on whether an unrecognized object exists in the scene analysis; A data storage method according to Technical Idea 8. [Explanation of symbols]

[0092] 10: Data collection system 11: Development vehicle 13: Mass-produced vehicle 15: Data center 20: Development vehicle system 21: Sensor group 22: Control device 23: Communication device 24: Storage medium 25: Processor 26: Memory 30: Mass-produced vehicle system 31: Sensor group 32: Control device 33: Communication device 34: Storage medium 35: Processor 36: Memory 40: Processing system 41: Information presentation device 42: Input device 43: Storage medium 44: Communication device 45: Control device (data storage device) 46: Processor 47: Memory 70: Folder

Claims

1. A data storage method executed by a processor (46) for storing driving data collected while a vehicle is driving in a storage medium (43), comprising: Acquire the driving data, The acquired driving data is stored in the storage medium in association with a scenario defined based on scene classification items for identifying a driving scene and vehicle behavior classification items; the plurality of scenarios are classified in a tree-like manner based on at least one of the scene classification items and the vehicle behavior classification items; Data storage method.

2. an upper limit is set for the amount of data during driving that is stored in association with one of the scenarios; 2. The data storage method of claim 1.

3. outputting information indicating that there is no corresponding scenario based on the fact that the scenario corresponding to the acquired running data does not correspond to the scenario corresponding to the running data already stored in the storage medium; 2. The data storage method of claim 1.

4. outputting the acquired traveling data and the scenarios that are candidates to be associated with the traveling data to an operator who operates the processor; 2. The data storage method of claim 1.

5. determining the scenario that is filled with the running data stored in the storage medium; determining a data collection condition that inhibits the vehicle from collecting the traveling data corresponding to the determined scenario; outputting the determined data collection conditions for provision to the vehicle; 2. The data storage method of claim 1.

6. determining the scenario in which the running data stored in the storage medium is insufficient; determining a data collection condition for the vehicle to collect the traveling data corresponding to the determined scenario; outputting the determined data collection conditions for provision to the vehicle; 2. The data storage method of claim 1.

7. 7. The data storage method according to claim 5, wherein the data collection conditions are determined by dividing them into conditions for mass-produced vehicles (13) and conditions for development vehicles (11) from which the amount of data during driving, which is greater than that of the mass-produced vehicles, can be acquired.

8. performing a scene analysis on the acquired driving data to determine whether the acquired driving data corresponds to any one of a plurality of scenarios corresponding to the driving data stored in the storage medium; 2. The data storage method of claim 1.

9. the scene analysis includes processing for recognizing objects around the vehicle based on the driving data; determining whether an annotation by an operator is required based on whether an unrecognized object exists in the scene analysis; 9. The data storage method according to claim 8.

10. A data storage device (45) that stores running data collected while a vehicle is running in a storage medium (43), a processor (46) comprising: acquiring the driving data; and storing the acquired driving data in the storage medium in association with a scenario defined based on scene classification items for identifying a driving scene and vehicle behavior classification items; the plurality of scenarios are classified in a tree-like manner based on at least one of the scene classification items and the vehicle behavior classification items; Data storage device.

11. A storage medium in which driving data acquired while a vehicle is driving is stored in association with one of a plurality of scenarios defined based on a plurality of types of scene classification items and vehicle behavior classification items for identifying driving scenes, and the plurality of scenarios form a scenario tree defined based on at least one of the scene classification items and the vehicle behavior classification items.

Images