JP7537263B2 - Remote support device, remote support method, and remote support program - Google Patents

Description

This disclosure relates to a remote assistance device, a remote assistance method, and a remote assistance program.

There is a technology that allows an autonomous vehicle to be remotely supported by an autonomous driving assistance center. For example, if an autonomous vehicle stops while driving due to some kind of obstacle, it requests assistance from the autonomous driving assistance center. In response to the assistance request, an operator at the autonomous driving assistance center determines whether the scene is on-street parking, construction work, etc., and responds to the autonomous vehicle based on the determined scene. After that, an automatic response based on the operator's response is made to the autonomous vehicle traveling through the point where the assistance request was made.

For example, Patent Document 1 describes a control device that remotely controls an autonomous vehicle. The control device includes a registration unit that, when remote control is performed, registers location data including position information acquired from a target vehicle, which is an autonomous vehicle that is the subject of remote control, and sensing data collected by sensors of the target vehicle. When remote control is performed, the control device includes a selection unit that, when location data similar to the current location data acquired from the target vehicle among the location data registered by the registration unit is available, selects a first remote control that instructs driving by remote control using the similar location data, and, when similar location data is not available, receives a command input by an operator to an interface and selects a second remote control that instructs driving by remote control based on the command.

JP 2019-185280 A

In the technology described in Patent Document 1, the same automatic response is made when the location data is similar. However, there are cases where the effective period for continuing the automatic response is not specified. In such cases, even if a different fault has actually occurred, the same automatic response as before may be made, which may cause the autonomous vehicle to behave erroneously. For this reason, it is desirable to improve the accuracy of the automatic response.

The present disclosure aims to provide a remote assistance device, a remote assistance method, and a remote assistance program that can improve the accuracy of automatic responses when providing remote assistance to an autonomous vehicle.

In order to achieve the above object, the remote assistance device (20) according to the first aspect of the present disclosure is a remote assistance device that provides remote assistance to an autonomous vehicle within an area managed by an autonomous driving assistance center, and includes: an acquisition unit (21A) that acquires assistance request information including position information of the autonomous vehicle from the autonomous vehicle; a classification unit (21B) that acquires a judgment scene most recently responded to by the operator in a specific area including a point indicated by the position information from a database (25B) that includes an operator judgment log, which is a log of judgment scenes to which the operator has responded in the past, in response to the assistance request information acquired by the acquisition unit, and classifies the scene of the autonomous vehicle into either a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait based on the acquired judgment scene; and a setting unit (21B) that sets a valid period indicating a period during which an automatic response based on the operator's most recent response is output as a valid response, to be shorter than when the scene is classified into the scene where the autonomous vehicle should wait, when the classification unit classifies the scene of the autonomous vehicle into the scene where the autonomous vehicle should overtake.

A remote assistance method according to a second aspect of the present disclosure is a remote assistance method using a remote assistance device that provides remote assistance to an autonomous vehicle within an area managed by an autonomous driving assistance center, and includes obtaining assistance request information including position information of the autonomous vehicle from the autonomous vehicle, obtaining a judgment scene most recently responded to by the operator in a specific area including a point indicated by the position information from a database including an operator judgment log, which is a log of judgment scenes to which the operator has responded in the past, in response to the obtained assistance request information, classifying the scene of the autonomous vehicle into either a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait based on the obtained judgment scene, and setting a validity period indicating a period during which an automatic response based on the operator's most recent response is output as a valid response, if the scene of the autonomous vehicle is classified as a scene where the autonomous vehicle should overtake, to be shorter than if the scene is classified as a scene where the autonomous vehicle should wait.

The remote assistance program (25A) according to the third aspect of the present disclosure is a remote assistance program for a remote assistance device that provides remote assistance to an autonomous vehicle within a management area of an autonomous driving assistance center, and causes a computer to function as an acquisition unit (21A) that acquires assistance request information including position information of the autonomous vehicle from the autonomous vehicle, a classification unit (21B) that acquires a judgment scene most recently responded to by the operator in a specific area including a point indicated by the position information from a database (25B) that includes an operator judgment log, which is a log of judgment scenes to which the operator has responded in the past, in response to the assistance request information acquired by the acquisition unit, and classifies the scene of the autonomous vehicle into either a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait based on the acquired judgment scene, and a setting unit (21B) that sets a valid period indicating a period during which an automatic response based on the operator's most recent response is output as a valid response, to be shorter than the period when the scene of the autonomous vehicle is classified into the scene where the autonomous vehicle should wait, when the classification unit classifies the scene of the autonomous vehicle into the scene where the autonomous vehicle should overtake.

The disclosed technology has the effect of improving the accuracy of automatic responses when providing remote assistance to autonomous vehicles.

1 is a diagram illustrating an example of a configuration of a remote support system according to a first embodiment. FIG. 13 is a diagram for explaining remote support by an operator according to the embodiment. 1 is a block diagram showing an example of a functional configuration of an in-vehicle device and a remote assistance device according to a first embodiment; FIG. 4 is a diagram illustrating an example of a scene classification table according to the embodiment. 11 is a diagram illustrating a valid period setting process according to the first embodiment; FIG. FIG. 11 is a diagram showing an example of a duration distribution for each determination scene according to the embodiment. 11 is a diagram illustrating a method for calculating a duration length according to an embodiment. FIG. 13 is a diagram illustrating another method for calculating a duration length according to the embodiment. FIG. 11A and 11B are diagrams for explaining a method of identity determination according to an embodiment. 11 is a flowchart showing an example of a processing flow of a remote assistance program according to the first embodiment. 11 is a diagram for explaining a valid period setting process according to a comparative example; FIG. FIG. 13 is a diagram illustrating an example of an automatic response unavailable period according to the second embodiment. FIG. 11 is a diagram illustrating a valid period setting process according to the second embodiment; 13A to 13C are diagrams illustrating a method for calculating a duration distribution for each location associated with a determination scene according to a fourth embodiment. FIG. 13 is a diagram illustrating an example of a duration model with respect to distance from a facility. 13 is a diagram illustrating a method for calculating a duration distribution for each time period associated with a determination scene according to the fourth embodiment. FIG. FIG. 13 is a diagram illustrating an example of a duration model for precipitation. 13 is a diagram illustrating a method for calculating a duration distribution for each vehicle type associated with a determination scene according to a fourth embodiment. FIG. FIG. 13 is a block diagram showing an example of a functional configuration of an in-car device and a remote assistance device according to a fifth embodiment.

Below, an example of a form for implementing the technology disclosed herein is described in detail with reference to the drawings.

[First embodiment]
FIG. 1 is a diagram showing an example of the configuration of a remote assistance system 100 according to the first embodiment.

As shown in FIG. 1, the remote assistance system 100 according to this embodiment includes an in-vehicle device 10 mounted on an autonomous vehicle and a remote assistance device 20 provided in an autonomous driving assistance center. The remote assistance device 20 provides remote assistance to autonomous vehicles within an area managed by the autonomous driving assistance center. In this embodiment, a private passenger car is used as an example of an autonomous vehicle, but the system may also be applied to other vehicles such as trucks, buses, and taxis. Furthermore, autonomous vehicles also include cases where a person is on board to control the vehicle or to take over control of the vehicle in an emergency. Furthermore, it also includes vehicles in which some of the steering of the vehicle is performed automatically.

The in-vehicle device 10 and the remote assistance device 20 are communicatively connected via a network N. Examples of the network N include the Internet and a wide area network (WAN).

An autonomous vehicle is a vehicle that can drive autonomously without the driver's operation under certain conditions. When an autonomous vehicle encounters an event such as on-street parking, traffic jams, or construction while driving, it will overtake or wait. These actions are performed, for example, based on a response from the autonomous driving assistance center.

The in-vehicle device 10 has a function to generate a driving plan including a driving route to a destination based on destination information such as an address or latitude and longitude, and a function to control the autonomous driving of the vehicle. The in-vehicle device 10 has a CPU (Central Processing Unit) 11, a memory 12, a display unit 13, a storage unit 14, a sensor group 15, a camera 16, and a communication unit 17.

The CPU 11 is an example of a processor. The term "processor" here refers to a processor in a broad sense, and includes general-purpose processors (e.g., CPUs) and dedicated processors (e.g., GPU: Graphics Processing Unit, ASIC: Application Specific Integrated Circuit, FPGA: Field Programmable Gate Array, programmable logic device, etc.). The memory 12 is composed of ROM (Read Only Memory), RAM (Random Access Memory), etc.

The display unit 13 may be, for example, a liquid crystal display (LCD) or an organic electroluminescence (EL) display. The display unit 13 may also have an integrated touch panel.

For example, a hard disk drive (HDD), a solid state drive (SSD), a flash memory, etc. is used as the storage unit 14. A control program (not shown) for controlling the automatic driving is stored in the storage unit 14.

The sensor group 15 is composed of various sensors for grasping the situation around the vehicle. The sensor group 15 includes a millimeter wave radar that transmits a search wave to a predetermined range outside the vehicle, and a LIDAR (Light Detection and Ranging/Laser Imaging Detection and Ranging) that scans at least a predetermined range in front of the vehicle. The sensor group 15 may also include a GNSS (Global Navigation Satellite System) receiver mounted on the vehicle. This GNSS receiver acquires information such as the current position of the vehicle and the current time.

The camera 16 captures an image of a predetermined range in a predetermined direction of the vehicle. Specifically, the camera 16 is provided at the front of the vehicle and captures an image of the area in front of the vehicle. There may be only one camera 16, but multiple cameras may be provided at multiple locations to obtain more information.

The communication unit 17 is a communication interface that connects to a network N such as the Internet or a WAN, and communicates with the remote support device 20.

The in-vehicle device 10 is connected to a driving device (not shown) required for autonomous driving, and performs autonomous driving by controlling this driving device. Examples of this driving device include electric power steering, electronically controlled brakes, and electronically controlled throttle.

The in-vehicle device 10 performs automatic driving by controlling the driving, steering, and braking of the vehicle so that the vehicle operates automatically according to a driving plan for the vehicle. Note that there are various known methods for automatic driving itself, and the method is not particularly limited in this embodiment.

Meanwhile, the remote assistance device 20 monitors the vehicle status of the autonomous vehicle by periodically communicating with the on-board device 10 of the autonomous vehicle. As an example, the remote assistance device 20 may be a general-purpose computer device such as a server computer or a personal computer (PC). The remote assistance device 20 includes a CPU 21, a memory 22, an operation unit 23, a display unit 24, a storage unit 25, and a communication unit 26.

The CPU 21 is an example of a processor. As mentioned above, the processor here refers to a processor in a broad sense, including general-purpose processors and dedicated processors. The memory 22 is composed of ROM, RAM, etc.

The operation unit 23 is configured as an interface for receiving operation input to the remote support device 20. For example, a liquid crystal display (LCD) or an organic EL display is used for the display unit 24. The display unit 24 may have an integrated touch panel.

For example, a HDD, SSD, flash memory, etc. are used for the storage unit 25. A remote assistance program 25A for remotely assisting an autonomous vehicle is stored in the storage unit 25. For example, the remote assistance program 25A may be pre-installed in the remote assistance device 20. The remote assistance program 25A may be realized by storing it in a non-volatile non-transitory recording medium or distributing it via the network N and installing it appropriately in the remote assistance device 20. Examples of non-volatile non-transitory recording media include a CD-ROM (Compact Disc Read Only Memory), an optical magnetic disk, a HDD, a DVD-ROM (Digital Versatile Disc Read Only Memory), a flash memory, a memory card, etc.

The communication unit 26 is a communication interface that connects to a network N such as the Internet or a WAN, and communicates with the in-vehicle device 10.

Figure 2 is a diagram used to explain remote support by an operator in this embodiment.

As shown in FIG. 2, when an autonomous vehicle 30 stops while traveling, it makes an assistance request to the autonomous driving assistance center. In response to the assistance request from the autonomous vehicle 30, an operator OP at the autonomous driving assistance center judges a scene such as on-street parking or construction, and makes a response to the autonomous vehicle 30 based on the judged scene. At this time, the operator OP registers the judged scene in response by associating it with point data on a map. Thereafter, in response to an assistance request, an automatic response based on the same judged scene is made to an autonomous vehicle traveling through the point where the judged scene has been registered.

However, as mentioned above, there are cases where the validity period for continuing the automatic response is not specified. In such cases, even if a different fault actually occurs, the same automatic response as before may be made, and the autonomous vehicle 30 may behave erroneously. For this reason, it is desirable to improve the accuracy of the automatic response.

Therefore, the CPU 11 of the in-vehicle device 10 according to this embodiment functions as each unit shown in FIG. 3 by writing the control program stored in the storage unit 14 to the RAM and executing it. Also, the CPU 21 of the remote assistance device 20 according to this embodiment functions as each unit shown in FIG. 3 by writing the remote assistance program 25A stored in the storage unit 25 to the RAM and executing it.

Figure 3 is a block diagram showing an example of the functional configuration of the in-vehicle device 10 and the remote assistance device 20 according to the first embodiment.

As shown in FIG. 3, the CPU 11 of the in-vehicle device 10 according to this embodiment functions as a request output unit 11A and a judgment acquisition unit 11B.

The request output unit 11A outputs (transmits) assistance request information, including time information, position information, sensing information, etc., when it is determined that the vehicle cannot continue traveling, to the remote assistance device 20 of the autonomous driving assistance center via the communication unit 17.

The judgment acquisition unit 11B acquires (receives) the operator response or automatic response output from the remote assistance device 20 via the communication unit 17, and sends the judgment scene contained in the acquired operator response or automatic response to a control device (not shown) that controls the vehicle of the vehicle.

Meanwhile, the CPU 21 of the remote support device 20 according to this embodiment functions as a request acquisition unit 21A, an automatic judgment unit 21B, a judgment output unit 21C, an operator presentation unit 21D, a duration calculation unit 21E, and an operator judgment registration unit 21F. Note that the request acquisition unit 21A is an example of an acquisition unit, the automatic judgment unit 21B is an example of a classification unit, a setting unit, and an instruction unit, and the duration calculation unit 21E is an example of a calculation unit.

For example, an operator judgment log database (hereinafter referred to as "operator judgment log DB") 25B and a duration distribution database (hereinafter referred to as "duration distribution DB") 25C are stored in the storage unit 25. Note that the operator judgment log DB 25B and the duration distribution DB 25C may be stored in an external storage device instead of the storage unit 25.

The operator judgment log DB25B has an operator judgment log registered therein. This operator judgment log is a log of judgment scenes to which the operator responded in the past. This judgment scene represents the result of the operator's judgment of the current scene of the autonomous vehicle as, for example, "parking on the road", "construction", "traffic jam" or the like in response to the assistance request information from the autonomous vehicle. Each operator judgment log includes, for example, a unique ID (Identification), a judgment time, an assistance request position (or an assistance request area), assistance content (including a judgment scene, control content, etc.), sensing information (including surrounding images, point clouds, recognition results, etc.), whether or not there is an identical judgment (including an ID of the same operator judgment log if there is an identical judgment), and other information. In addition, the duration distribution DB25C has a duration distribution registered therein. This duration distribution includes, for example, scene information related to the scene, which is a unit for managing the duration, duration distribution (e.g., a histogram), and other information. A method for calculating this duration distribution will be described later.

The request acquisition unit 21A acquires (receives) the assistance request information output from the autonomous vehicle via the communication unit 26, and sends the acquired assistance request information to the automatic judgment unit 21B.

The automatic judgment unit 21B acquires, from the operator judgment log DB 25B, a judgment scene most recently responded to by the operator in a specific area including the point indicated by the position information of the autonomous vehicle, in response to the assistance request information acquired by the request acquisition unit 21A. Here, the specific area is defined as, for example, a circular area with a diameter of Xm centered on the assistance request point.

Based on the acquired judgment scene, the automatic judgment unit 21B classifies the scene of the autonomous vehicle into either a scene where it should overtake or a scene where it should wait.

Figure 4 shows an example of a scene classification table 25D according to this embodiment.

The scene classification table 25D shown in FIG. 4 is stored, for example, in the memory unit 25 and can be referenced by the automatic determination unit 21B. If the determined scene is, for example, "parking on the street," "construction," or "accident," the scene of the autonomous vehicle is classified as a scene where the autonomous vehicle should overtake, and if the determined scene is, for example, "pedestrian," "waiting at a traffic light," "traffic jam," or "waiting for a bus," the scene of the autonomous vehicle is classified as a scene where the autonomous vehicle should wait.

However, in the case of a scene where overtaking is required, if a scene where waiting is actually required is classified as such, there is a problem of erroneous determination that unnecessary overtaking will occur.On the other hand, in the case of a scene where waiting is required, if a scene where overtaking is required is classified as such, there is a problem of erroneous determination that waiting will occur.

When the scene of the autonomous vehicle is classified as a scene where it should overtake, the automatic determination unit 21B sets the validity period to be shorter than when it is classified as a scene where it should wait. The validity period refers to the period during which an automatic response based on the most recent response of the operator is output as a valid response. The validity period is derived from the most recent operator response as the starting point.

Figure 5 is a diagram used to explain the validity period setting process according to the first embodiment.

As shown in FIG. 5, in the graph 40, the horizontal axis indicates duration, and the vertical axis indicates frequency and probability of occurrence. The graph 40 shows the case of a "street parking" scene as an example. The probability of occurrence 41 indicates the probability that a "street parking" scene continues to exist, and the probability decreases over time. The frequency 42 indicates the frequency with which a "street parking" scene occurs for each duration time period. For example, parking for 5 minutes or more and 10 minutes or less occurs 10 times. The duration distribution 43 is calculated, for example, for each judgment scene from the operator judgment log included in the operator judgment log DB 25B. The duration distribution 43 represents the relationship between the occurrence frequency of a judgment scene and the duration of the judgment scene. The duration distribution 43 is calculated by the duration calculation unit 21E described later.

As an example, as shown in graph 40 in FIG. 5, when the effective period when classified as a scene where waiting is taken as the average value μ of duration distribution 43, automatic determination unit 21B sets the effective period t when classified as a scene where overtaking is taken to be shorter than the average value μ, that is, the presence probability 41 is set to be relatively high. The effective period t is calculated, for example, by the following formula (1). Here, μ is the average value of duration distribution 43, and σ is the standard deviation of duration distribution 43.

t=μ-3σ...(1)

In (S1) of FIG. 5, the autonomous vehicle 30 encounters an obstacle while traveling, so it stops and requests assistance from the autonomous driving assistance center. The operator (OP) of the autonomous driving assistance center judges the scene encountered by the autonomous vehicle 30 to be "parked on the street" and responds with the judged scene that is the result of that judgment. In accordance with the operator's response, the autonomous vehicle 30 judges that it is an overtaking scene and takes action to overtake the "parked on the street" vehicle.

Next, in (S2), similar to (S1) above, the autonomous vehicle 30 encounters some kind of obstacle and stops temporarily, requesting assistance from the autonomous driving assistance center. The remote assistance device 20 at the autonomous driving assistance center automatically replies with "parked on the street" as an automatic response based on the most recent operator response. The autonomous vehicle 30 determines that this is an overtaking scene in accordance with the automatic response, and takes action to overtake the "parked on the street" vehicle.

Next, in (S3), similar to (S2) above, the autonomous vehicle 30 encounters some kind of obstacle and stops temporarily, and requests assistance from the autonomous driving assistance center. The remote assistance device 20 at the autonomous driving assistance center automatically responds with "parked on the street" as an automatic response based on the most recent operator response. The autonomous vehicle 30 determines that the scene is an overtaking scene in accordance with the automatic response, and performs the overtaking operation. However, at this point in time, there is no "parked on the street" vehicle, and the scene may have changed to one where the vehicle should wait, such as "pedestrian" or "waiting at a traffic light." In other words, performing an overtaking operation when the scene is actually one where the vehicle should wait is undesirable from a safety standpoint.

For this reason, in this embodiment, as described above, the valid period t of a scene to be overtaken is set shorter than the valid period of a scene to be waited for (for example, the average value μ). In other words, in a scene to be waited for, the scene simply waits, so it can be said that there is no problem from the perspective of safety even if the scene changes to one to be overtaken within the valid period. On the other hand, in a scene to be overtaken, if the scene changes to one to be waited for within the valid period, as described above, this may not be desirable from the perspective of safety. For this reason, the valid period t of a scene to be overtaken is set shorter.

In addition, if the period from the most recent operator response to obtaining assistance request information is within the valid period, the automatic judgment unit 21B instructs the judgment output unit 21C to output an automatic response to the autonomous vehicle. On the other hand, if the period from the most recent operator response to obtaining assistance request information exceeds the valid period, the automatic judgment unit 21B stops the automatic response, sends assistance request information to the operator presentation unit 21D, and instructs the operator to register the current judgment scene.

The operator presentation unit 21D presents the forward image, map information, sensing information, etc. included in the assistance request information to the operator, and accepts registration of the current judgment scene from the operator. The operator presentation unit 21D instructs the judgment output unit 21C to output the operator response to the autonomous driving vehicle.

The judgment output unit 21C outputs (sends) the automatic response as the judgment result in accordance with an instruction to output the automatic response from the automatic judgment unit 21B, or outputs the operator response as the judgment result in accordance with an instruction to output the operator response from the operator presentation unit 21D.

Here, as described above, the duration calculation unit 21E uses the operator judgment log included in the operator judgment log DB 25B to calculate the duration distribution for each judgment scene, and registers the calculated duration distribution in the duration distribution DB 25C. The automatic judgment unit 21B sets a validity period according to the scene of the autonomous vehicle based on the duration distribution registered in the duration distribution DB 25C.

Figure 6 shows an example of the duration distribution for each determination scene according to this embodiment.

As shown in FIG. 6, a different duration distribution may be calculated for each determined scene, such as "street parking" or "obstacle." In this case, a different duration distribution can be obtained for each determined scene. This can further improve the accuracy of the validity period.

Figure 7 is a diagram used to explain the method for calculating the duration length according to this embodiment.

As an example, as shown in FIG. 7, the duration calculation unit 21E calculates the duration length when the same scene is judged continuously from the operator judgment log acquired from the operator judgment log DB 25B. In other words, the difference between the time when the first (first) operator judgment is made and the time when the same operator judgment is made last (third time in this case) is calculated as the duration length (duration of the same judgment).

Figure 8 is a diagram used to explain another method of calculating the duration length according to this embodiment.

In FIG. 8, × indicates a driving log. As shown in FIG. 8, the duration calculation unit 21E may calculate the duration length using the driving log of the autonomous vehicle. In other words, using the duration length (same judgment duration) calculated from the operator judgment log as a reference, a predetermined period of several tens of seconds (e.g., 20 seconds) before and after the reference is determined, and driving logs for the predetermined period, logs from cameras installed as infrastructure, etc. are acquired. Then, the reference duration length is corrected according to the period in which the sensing information obtained from these logs is similar. This makes it possible to grasp the duration length more accurately.

The operator judgment registration unit 21F also determines whether the most recent operator judgment log and the current operator judgment result are the same judgment scene, and registers the current operator judgment result together with whether or not they are the same judgment in the operator judgment log DB 25B.

Figure 9 is a diagram used to explain the method of identity determination according to this embodiment.

The operator judgment registration unit 21F acquires an operator judgment log, which is a log of the operator judgment made most recently in the same area as the current operator judgment result, from the operator judgment log DB 25B. Then, as an example, as shown in FIG. 9, the operator judgment registration unit 21F determines whether the current operator judgment result and the most recent operator judgment log in the same area are the same judgment scene. If both judgment scenes are, for example, "on-street parking", they are determined to be the same. The determination of whether they are the same or not may be made by presenting the operator with a forward image or the like obtained from each autonomous vehicle to encourage the operator to make a judgment. Alternatively, it may be mechanically determined using a known method whether the features of the sensing information obtained from each autonomous vehicle are similar.

As shown in FIG. 3 above, the remote support device 20 according to this embodiment calculates in advance the duration of scenes judged by past operators, and registers the calculated duration distribution in the duration distribution DB 25C. Then, when setting the effective period, the remote support device 20 reads out the duration distribution from the duration distribution DB 25C, and sets the effective period based on the read out duration distribution.

Next, the operation of the remote support device 20 according to the first embodiment will be described with reference to FIG. 10.

Figure 10 is a flowchart showing an example of the processing flow of the remote assistance program 25A according to the first embodiment.

First, when the remote assistance device 20 is turned on, the remote assistance program 25A is started and executes the following steps.

In step S101 of FIG. 10, the CPU 21 determines whether or not assistance request information has been acquired from the in-vehicle device 10 of the autonomous vehicle. If it is determined that assistance request information has been acquired from the in-vehicle device 10 (if a positive determination), the process proceeds to step S102, and if it is determined that assistance request information has not been acquired from the in-vehicle device 10 (if a negative determination), the process waits in step S101. At this time, the location of the autonomous vehicle and an area including the location (for example, a circular area with a diameter of Xm centered on the location) are identified based on the position information included in the acquired assistance request information.

In step S102, the CPU 21 obtains from the operator judgment log DB 25B the most recent operator judgment log for the same area as the area identified in step S101. This operator judgment log includes the operator's most recent judgment scene.

In step S103, the CPU 21 classifies the scene of the autonomous vehicle based on the judgment scene included in the most recent operator judgment log acquired in step S102, by referring to the scene classification table 25D shown in FIG. 4 as an example.

In step S104, the CPU 21 determines whether the scene of the autonomous vehicle is classified as a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait, based on the result of the classification in step S103. If it is determined that the scene is classified as a scene where the autonomous vehicle should overtake (if the scene is a scene where the autonomous vehicle should wait), the process proceeds to step S105. If it is determined that the scene is classified as a scene where the autonomous vehicle should wait (if the scene is a scene where the autonomous vehicle should wait), the process proceeds to step S106.

In step S105, the CPU 21 sets the effective period t of the scene to be overtaken to be shorter than the effective period (e.g., the average value μ) of the scene to be waited for. The effective period t of the scene to be overtaken is calculated, for example, using the above-mentioned formula (1).

In step S106, the CPU 21 sets the mean value μ of the duration distribution as the effective period of the scene to be waited for, as shown in FIG. 5 above, as an example.

In step S107, the CPU 21 determines whether the period from the most recent operator response to obtaining the support request information is within the validity period (here, the validity period t or the average value μ). If it is determined that it is within the validity period (if the determination is positive), the process proceeds to step S108, and if it is determined that it will exceed the validity period (if the determination is negative), the process proceeds to step S109.

In step S108, the CPU 21 issues an instruction to output an automatic response.

In step S109, the CPU 21 stops the automatic response and instructs the operator to register the current judgment scene.

In step S110, the CPU 21 outputs to the autonomous vehicle an automatic response based on the instruction in step S108 or an operator response based on the instruction in step S109.

In step S111, the CPU 21 determines whether the end timing has arrived, such as an instruction to end remote support. If it is determined that the end timing has not arrived (if the result is negative), the process returns to step S101 and is repeated. If it is determined that the end timing has arrived (if the result is positive), the process by the remote support program 25A ends.

In this way, according to this embodiment, the validity period of the scene in which the vehicle should overtake is set to be shorter than the validity period of the scene in which the vehicle should wait. This improves the accuracy of the automatic response when providing remote assistance to an autonomous vehicle.

Second Embodiment
In the first embodiment, the validity period of a scene to be overtaken is set to be shorter than the validity period of a scene to be waited for. In the present embodiment, the validity period of a scene to be overtaken is set to be shorter than the validity period of a scene to be waited for only in the case of a scene in which unnecessary overtaking is undesirable.

The remote support device according to this embodiment has the same components as the remote support device 20 described in the first embodiment above, and only the differences will be described with reference to FIG. 3 above.

First, referring to FIG. 11, we will explain the case where the effective period t of the scene to be overtaken is set to the average value μ of the duration distribution 43.

Figure 11 is a diagram used to explain the validity period setting process in a comparative example.

In the comparative example of FIG. 11, the effective period t of the "parked on the street" scene, which is a scene to be overtaken, is set to the average value μ of the duration distribution 43. In addition, the actual duration of the "parked on the street" scene is shorter than the effective period t.

In (S11) of FIG. 11, the autonomous vehicle 30 encounters an obstacle while traveling, so it stops and requests assistance from the autonomous driving assistance center. The operator (OP) of the autonomous driving assistance center determines the scene encountered by the autonomous vehicle 30 as "parked on the street" and responds with the determined scene that is the result of this determination. In accordance with the operator's response, the autonomous vehicle 30 determines that it is an overtaking scene, and takes action to overtake the "parked on the street" vehicle.

Next, in (S12), similar to (S11) above, the autonomous vehicle 30 encounters some kind of obstacle and stops temporarily, and requests assistance from the autonomous driving assistance center. The remote assistance device 20 at the autonomous driving assistance center automatically responds with "parked on the street" as an automatic response based on the most recent operator response. The autonomous vehicle 30 determines that this is an overtaking scene in accordance with the automatic response, and performs an overtaking operation. However, at this point in time, there is no "parked on the street" vehicle, and a "traffic jam" has occurred, which is a scene in which the autonomous vehicle should wait, so the automatic response determines that this is an incorrect scene. Period T1 represents the period during which the incorrect determination occurs.

As shown in FIG. 11, erroneous judgment A shows a case where a "traffic jam" is erroneously judged as "parking on the street" in a place with two lanes on each side. In this case, there is a problem of unnecessary overtaking that goes into the adjacent lane. On the other hand, erroneous judgment B shows a case where a "traffic jam" is erroneously judged as "parking on the street" in a place with one lane on each side. In this case, there is a problem of unnecessary overtaking that goes into the oncoming lane.

Therefore, if the specific area for which an assistance request has been made is an area where the occurrence frequency of the scene to be waited for is equal to or higher than a preset threshold, or an area where the occurrence frequency of the scene to be waited for is higher than the occurrence frequency of the scene to be overtaken, the effective period t of the scene to be waited for may be set shorter than the effective period (e.g., the average value μ) by applying the above formula (1). Note that the above threshold is, for example, a value experimentally determined in advance, and as an example, 0.5 is set when the occurrence frequency is expressed in the range of 0 to 1. The occurrence frequency f of a certain scene A in the same area is calculated, for example, by the following formula (2). Here, N indicates the number of times scene A is judged in the same area, and R indicates the number of times the autonomous vehicle travels in the same area.

f=N/R...(2)

The occurrence frequency f may be calculated for each major scene classification (scenes where you should wait, scenes where you should overtake) or for each time period.

In other words, in places where congestion is likely to occur on a daily basis, it is possible to reduce erroneous judgments made by automatic responses by shortening the valid period t for scenes in which overtaking is required, such as "parked on the street."

Figure 12 shows an example of an automatic response unavailable period T2 according to the second embodiment.

As shown in FIG. 12, when the effective period t for a scene where overtaking is required, such as "on-street parking," is set shorter than the effective period of the scene where waiting is required (for example, the average value μ), unnecessary overtaking can be suppressed. However, on the other hand, when the duration of the scene where overtaking is required is actually longer than the effective period t, an automatic response unavailable period T2 occurs during which an automatic response is not possible. In this embodiment, the idea is to make the automatic response unavailable period T2 as short as possible and to lengthen the period during which an automatic response is possible.

For example, if the specific area for which assistance is requested is an area where overtaking situations such as "on-street parking" are likely to occur, setting the validity period t to a longer period often does not pose a significant safety problem.

For this reason, the automatic determination unit 21B according to this embodiment sets the effective period t of the scene to be overtaken to be equal to or greater than the effective period of the scene to be waited for (e.g., the average value μ) when the specific area for which an assistance request has been made is an area where the occurrence frequency of the scene to be overtaken is equal to or greater than a preset threshold value, or an area where the occurrence frequency of the scene to be overtaken is higher than the occurrence frequency of the scene to be waited for. Note that the above threshold value is, for example, a value experimentally determined in advance. In other words, by determining whether or not to set the effective period t short depending on the characteristics of the area in which the scene to be overtaken is occurring, it is possible to ensure a certain period for automatic response while suppressing erroneous determinations due to automatic response.

Figure 13 is a diagram used to explain the validity period setting process according to the second embodiment.

13, black circles on the map indicate points where the judgment of "street parking" is registered. For point A, from the graph showing the occurrence frequency of each scene that occurred in the past, it can be seen that the occurrence frequency of "traffic jam" is higher than that of "street parking" (or the occurrence frequency of "traffic jam" is equal to or higher than a preset threshold). In this case, if the effective period t of "street parking" is set relatively long, there is a high possibility that it will change to "traffic jam" within the effective period t, and there is a high possibility that the automatic response will be an erroneous judgment. For this reason, the effective period t is set shorter than the average value μ. On the other hand, for point B, from the graph showing the occurrence frequency of each scene that occurred in the past, it can be seen that the occurrence frequency of "street parking" is higher than that of "traffic jam" (or the occurrence frequency of "street parking" is equal to or higher than a preset threshold). In this case, even if the effective period t of "street parking" is set relatively long, there is a low possibility that it will change to "traffic jam" within the effective period t, and there is a low possibility that the automatic response will be an erroneous judgment. For this reason, the effective period t is set to the average value μ.

As described above, according to this embodiment, whether or not to set the valid period short is determined based on the characteristics of the area in which the scene to be overtaken is occurring. This ensures a certain amount of time for automatic response while suppressing erroneous judgments due to automatic response.

[Third embodiment]
In the first embodiment, the scene of the autonomous vehicle is classified based on the most recent operator's judgment scene in a specific area obtained from the operator judgment log. In the present embodiment, in addition to the judgment scene, the scene of the autonomous vehicle is classified using the similarity of the sensing information.

The remote support device according to this embodiment has the same components as the remote support device 20 described in the first embodiment above, and only the differences will be described with reference to FIG. 3 above.

The operator judgment log registered in the operator judgment log DB25B includes sensing information (hereinafter referred to as "first sensing information") obtained from the most recent autonomous vehicle that received the operator's most recent response. In addition, the assistance request information includes sensing information (hereinafter referred to as "second sensing information") obtained from the current autonomous vehicle.

The automatic determination unit 21B classifies the scene of the autonomous vehicle into either a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait, based on the similarity between the first sensing information and the second sensing information. A known method is used to calculate the similarity. For example, when images are used as the first sensing information and the second sensing information, pattern matching or the like is applied, as an example.

The automatic determination unit 21B instructs the autonomous vehicle to output an automatic response when the remaining time of the valid period from the most recent response by the operator to the acquisition of the support request information is a certain time (e.g., 20 seconds) or more and the similarity is a first threshold value (e.g., 0.6) or more. On the other hand, when the remaining time of the valid period from the most recent response by the operator to the acquisition of the support request information is a certain time (e.g., 20 seconds) or more and the similarity is less than the first threshold value (e.g., 0.6), the automatic determination unit 21B stops the automatic response and instructs the operator to register the current judgment scene. In addition, when the remaining time of the valid period from the most recent response by the operator to the acquisition of the support request information is less than a certain time (e.g., 20 seconds) and the similarity is a second threshold value (e.g., 0.9) or more that is greater than the first threshold value, the automatic determination unit 21B instructs the autonomous vehicle to output an automatic response. On the other hand, if the remaining valid period from the most recent response by the operator to the acquisition of the support request information is less than a certain time (e.g., 20 seconds) and the similarity is less than a second threshold (e.g., 0.9), the automatic determination unit 21B stops the automatic response and instructs the operator to register the current determined scene. Note that the valid period here may be the valid period t of the scene to be overtaken, or the valid period (e.g., the average value μ) of the scene to be waited for.

In addition, when the similarity is equal to or greater than a threshold (e.g., 0.6), the automatic judgment unit 21B may instruct the autonomous vehicle to output an automatic response if the period from the most recent response by the operator to obtaining the assistance request information is within the valid period, and when the similarity is less than the threshold (e.g., 0.6), the automatic response may be stopped even if the period from the most recent response by the operator to obtaining the assistance request information is within the valid period, and an instruction may be given to the operator to register the current judgment scene. Note that the valid period here may be the valid period t of the scene to be overtaken, or the valid period (e.g., the average value μ) of the scene to be waited for, as described above.

The threshold value may be set to a larger value as the remaining validity period from the most recent response by the operator to the acquisition of support request information decreases. As a result, when the remaining validity period is long, the similarity judgment is relatively lenient, and when the remaining validity period is short, the similarity judgment is relatively strict.

In this way, according to this embodiment, scenes of autonomous vehicles are classified using the similarity of the sensing information in addition to the determined scenes. This allows scenes of autonomous vehicles to be classified with high accuracy.

[Fourth embodiment]
In the first embodiment, a configuration in which a determination scene is used as a unit for managing duration distribution has been described. In the present embodiment, a configuration in which a location, a time period, and a vehicle type are used as units for managing duration distribution will be described.

The remote assistance device according to this embodiment has the same components as the remote assistance device 20 described in the first embodiment above, and only the differences will be described with reference to FIG. 3 above.

The automatic determination unit 21B calculates a duration distribution for each determined scene and each location associated with the determined scene, or for each determined scene and each time period associated with the determined scene, or for each determined scene and each vehicle type associated with the determined scene.

Figures 14(A) to 14(C) are diagrams used to explain a method for calculating a duration distribution for each location associated with a determination scene according to the fourth embodiment.

As shown in FIG. 14(A), the judged scene is further divided into grids to calculate the duration distribution. In the example of FIG. 14(A), the duration distribution and effective period are calculated that reflect the different characteristics (regionality, etc.) of areas 1 to 4. This improves the accuracy of the automatic response.

As shown in FIG. 14(B), when the determined scene is "street parking," the determined scene is further divided by the facilities it faces (convenience stores, commercial facilities, homes, parks, etc.) and the duration distribution is calculated. In the example of FIG. 14(B), the purpose of on-street parking differs for each facility, and the parking time also differs depending on the purpose. Therefore, a duration distribution and validity period that reflect the characteristics of parking time that differ for each facility are calculated. This improves the accuracy of the automatic response.

As shown in FIG. 14(C), when the determined scene is "accident", "obstacle", etc., the duration distribution is calculated by dividing it by the distance from a base that provides road service such as JAF (JAPAN AUTOMOBILE FEDERATION). In the example of FIG. 14(C), a duration distribution and effective period are calculated that reflect the characteristics of the time required for removal, which differs from area to area.

Figure 15 shows an example of a duration model versus distance from a facility.

In Figure 15, the three axes are frequency of occurrence (Y axis), duration (X axis), and distance from the base (Z axis), and the three-dimensional duration distribution is estimated using machine learning such as a Gaussian process. In this case, even if information is missing for a certain distance, it is possible to estimate it from information on other distances.

Figure 16 is a diagram used to explain a method for calculating a duration distribution for each time period associated with a determination scene according to the fourth embodiment.

As shown in FIG. 16, for example, the "street parking" judgment scene is further divided by time period (e.g., daytime: 8:00-20:00, nighttime: 20:00-8:00), and the duration distribution is calculated. Similarly, the "obstacle" (or "broken down car") judgment scene is further divided by time period (e.g., daytime: 8:00-20:00, nighttime: 20:00-8:00), and the duration distribution is calculated. For example, the duration of "street parking" is short during the day and long at night. On the other hand, the duration of "obstacle" (or "broken down car") is short during the day because there are more people on hand and towing is smooth, and long at night. In the example of FIG. 16, the duration distribution and effective period are calculated that reflect the different characteristics for each time period. This improves the accuracy of automatic responses.

In addition, in the "construction" judgment scene, the duration distribution is calculated by dividing the duration from registration by the amount of precipitation. A duration distribution and validity period are calculated that reflect the different characteristics for each period, such as the amount of precipitation. This improves the accuracy of automatic responses.

Figure 17 shows an example of a duration model for precipitation.

In Figure 17, the three axes are frequency of occurrence (Y axis), duration (X axis), and precipitation amount (Z axis), and the three-dimensional duration distribution is estimated using machine learning such as a Gaussian process. In this case, even if information on a certain amount of precipitation is missing, it is possible to estimate it from information on other precipitation amounts.

Figure 18 is a diagram used to explain a method for calculating a duration distribution for each vehicle type associated with a determination scene in the fourth embodiment.

As shown in Figure 18, in the "street parking" judgment scene, the duration distribution is calculated by further dividing by vehicle type (e.g., passenger car, taxi, truck, etc.). For example, trucks tend to park on the street for a long time. In the example of Figure 18, a duration distribution and validity period are calculated that reflect the different characteristics of each vehicle type. This improves the accuracy of the automatic response.

In this way, according to this embodiment, the duration distribution is calculated for each location, time period, or vehicle type associated with the determined scene. This improves the accuracy of the automatic response.

[Fifth embodiment]
FIG. 19 is a block diagram showing an example of a functional configuration of the in-car device 10 and the remote assistance device 20A according to the fifth embodiment.

As shown in FIG. 19, the CPU 11 of the in-vehicle device 10 according to this embodiment functions as a request output unit 11A and a judgment acquisition unit 11B.

On the other hand, the CPU 21 of the remote support device 20A according to this embodiment functions as a request acquisition unit 21A, an automatic judgment unit 21B, a judgment output unit 21C, an operator presentation unit 21D, a duration calculation unit 21E, and an operator judgment registration unit 21F. The remote support device 20A according to this embodiment differs from the remote support device 20 shown in FIG. 3 above in that it does not include a duration distribution DB 25C.

As shown in FIG. 19, the remote support device 20A according to this embodiment calculates the duration distribution and effective period of the scenes judged by the past operators in advance, and registers the calculated duration distribution and effective period in the operator judgment log DB 25B. Then, when setting the effective period, the effective period is read from the operator judgment log DB 25B and set.

In other words, the operator judgment registration unit 21F operates the duration calculation unit 21E when the same judgment scene has not been registered recently in the same area as the area for which the current support request was made. The duration calculation unit 21E then calculates the duration length and duration distribution for the judgment scene by the operator that is to be registered by the operator judgment registration unit 21F. The duration calculation unit 21E then associates the calculated duration distribution with the operator judgment log and registers it in the operator judgment log DB 25B. At this time, the effective period may be registered instead of the duration distribution, or the duration distribution and the effective period may be registered.

The above describes an example of a remote support device according to an embodiment. The embodiment may be in the form of a program for causing a computer to execute the functions of each unit of the remote support device. The embodiment may be in the form of a non-transient recording medium that stores these programs and is readable by a computer.

In addition, the configuration of the remote support device described in the above embodiment is merely an example, and may be modified according to the situation without departing from the spirit of the invention.

The processing flow of the program described in the above embodiment is also an example, and unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the program.

In the above embodiment, a case has been described in which the processing according to the embodiment is realized by a software configuration using a computer by executing a program, but this is not limited to this. The embodiment may be realized, for example, by a hardware configuration or a combination of a hardware configuration and a software configuration.

10 Vehicle-mounted device, 11, 21 CPU, 11A Request output unit, 11B Judgment acquisition unit, 12, 22 Memory, 13, 24 Display unit, 14, 25 Storage unit, 15 Sensor group, 16 Camera, 17, 26 Communication unit, 20, 20A Remote support device, 23 Operation unit, 21A Request acquisition unit, 21B Automatic judgment unit, 21C Judgment output unit, 21D Operator presentation unit, 21E Duration calculation unit, 21F Operator judgment registration unit, 25A Remote support program, 25B Operator judgment log DB, 25C Duration distribution DB, 25D Scene classification table, 100 Remote support system

Claims (12)

A remote assistance device (20) that remotely assists an autonomous vehicle within a management area of an autonomous driving assistance center,
An acquisition unit (21A) that acquires assistance request information including position information of the autonomous driving vehicle from the autonomous driving vehicle;
a classification unit (21B) that acquires a judgment scene most recently responded to by the operator in a specific area including the point indicated by the position information from a database (25B) including an operator judgment log, which is a log of judgment scenes to which the operator has responded in the past, in response to the assistance request information acquired by the acquisition unit, and classifies a scene of the autonomous vehicle into either a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait based on the acquired judgment scene;
a setting unit (21B) that sets a validity period representing a period during which an automatic response based on the most recent response of the operator is output as a valid response when the scene of the autonomous vehicle is classified by the classification unit as the scene where overtaking is required to be performed to be shorter than a validity period when the scene is classified as the scene where waiting is required;
an instruction unit (21B) that instructs the autonomous vehicle to output the automatic response if the period from the most recent response by the operator to obtaining the assistance request information is within the valid period, and that instructs the autonomous vehicle to stop the automatic response and to have the operator register the current determination scene if the period from the most recent response by the operator to obtaining the assistance request information exceeds the valid period;
A remote support device equipped with the above. The remote assistance device of claim 1, wherein the setting unit sets the effective period of the scene to be overtaken to be shorter than the effective period of the scene to be waited for when the specific area is an area where the occurrence frequency of the scene to be waited for is equal to or higher than a predetermined threshold, or when the occurrence frequency of the scene to be waited for is higher than the occurrence frequency of the scene to be overtaken. The remote assistance device according to claim 1, wherein the setting unit sets the effective period of the scene to be overtaken to be longer than the effective period of the scene to be waited for when the specific area is an area where the occurrence frequency of the scene to be overtaken is equal to or higher than a preset threshold value, or when the occurrence frequency of the scene to be overtaken is higher than the occurrence frequency of the scene to be waited for. The operator judgment log further includes first sensing information obtained from a most recent autonomous vehicle that received a most recent response from the operator,
The assistance request information further includes second sensing information obtained from the autonomous vehicle,
The remote assistance device according to claim 1 , wherein the classification unit classifies a scene of the autonomous vehicle into either the scene where the autonomous vehicle should overtake or the scene where the autonomous vehicle should wait, based on a similarity between the first sensing information and the second sensing information. The instruction unit instructs the autonomously driven vehicle to output the automatic response when the remaining time of the validity period for the period from the most recent response by the operator to acquisition of the assistance request information is equal to or greater than a certain time and the similarity is equal to or greater than a first threshold, or when the remaining time of the validity period for the period from the most recent response by the operator to acquisition of the assistance request information is less than the certain time and the similarity is equal to or greater than a second threshold that is greater than the first threshold .
The remote assistance device according to claim 4 . When the similarity is equal to or greater than a threshold, if a period from the most recent response by the operator to obtaining the assistance request information is within the valid period , the instruction unit instructs the autonomous driving vehicle to output the automatic response, and when the similarity is less than the threshold, the instruction unit instructs the autonomous driving vehicle to stop the automatic response even if a period from the most recent response by the operator to obtaining the assistance request information is within the valid period, and instructs the operator to register the current determination scene.
The remote assistance device according to claim 4 . The remote support device according to claim 6 , wherein the threshold value is set to a larger value as the remaining time of the valid period with respect to the period from the most recent response by the operator to acquisition of the support request information decreases. a calculation unit (21E) for calculating, for each determined scene, a duration distribution representing a relationship between an occurrence frequency of the determined scene and a duration of the determined scene, using the operator determination log;
The remote assistance device according to any one of claims 1 to 7 , wherein the setting unit sets the effective period according to a scene of the autonomous driving vehicle based on the duration distribution calculated by the calculation unit. The remote assistance device according to claim 8, wherein the calculation unit calculates the duration distribution for each determined scene and a location associated with the determined scene, or for each determined scene and a time period associated with the determined scene, or for each determined scene and a vehicle type associated with the determined scene. The remote support device according to claim 8 or claim 9 , wherein the calculation unit calculates the effective period based on the duration distribution, and registers the calculated effective period in the database in association with the operator judgment log. A remote assistance method by a remote assistance device that provides remote assistance to an autonomous vehicle within a management area of an autonomous driving assistance center, comprising:
Acquire assistance request information from the autonomous vehicle, the assistance request information including location information of the autonomous vehicle;
In response to the acquired assistance request information, a judgment scene most recently responded to by the operator in a specific area including the point indicated by the position information is acquired from a database including an operator judgment log, which is a log of judgment scenes to which the operator has responded in the past, and based on the acquired judgment scene, classifying the scene of the autonomous vehicle into either a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait;
When the scene of the autonomous vehicle is classified as the overtaking scene, a validity period representing a period during which an automatic response based on the most recent response of the operator is output as a valid response is set to be shorter than when the scene is classified as the waiting scene ;
If the period from the most recent response by the operator to obtaining the assistance request information is within the valid period, an instruction is given to output the automatic response to the autonomously driven vehicle, and if the period from the most recent response by the operator to obtaining the assistance request information exceeds the valid period, an instruction is given to stop the automatic response and to have the operator register the current judgment scene .
Remote assistance methods. A remote assistance program (25A) for a remote assistance device that provides remote assistance to an autonomous vehicle within a management area of an autonomous driving assistance center,
Computer,
An acquisition unit (21A) that acquires assistance request information including position information of the autonomously driven vehicle from the autonomously driven vehicle;
a classification unit (21B) that acquires a judgment scene most recently responded to by the operator in a specific area including the point indicated by the position information from a database (25B) including an operator judgment log, which is a log of judgment scenes to which the operator has responded in the past, in response to the assistance request information acquired by the acquisition unit, and classifies a scene of the autonomous vehicle into either a scene where the autonomous vehicle should overtake or a scene where the autonomous vehicle should wait based on the acquired judgment scene ;
a setting unit (21B) that sets a validity period representing a period during which an automatic response based on the most recent response of the operator is output as a valid response when the scene of the autonomous vehicle is classified by the classification unit as the scene to be overtaken, to be shorter than when the scene is classified as the scene to be waited; and
An instruction unit (21B) that instructs the autonomous vehicle to output the automatic response when the period from the most recent response by the operator to the acquisition of the assistance request information is within the valid period, and that instructs the autonomous vehicle to stop the automatic response and to have the operator register the current judgment scene when the period from the most recent response by the operator to the acquisition of the assistance request information exceeds the valid period.
A remote support program to function as a