Control system, control program, and control method

Abstract

A purpose of the present invention is to make it possible to cause an automatic driving vehicle to travel safely when an unknown event occurs during traveling of the automatic driving vehicle, even without preparing a vast amount of training data including unknown events and training a model on said data.　A control system (100) comprises: a first planning unit (1) that obtains first travel plan information by inputting traffic information into a first model (M1) that is constructed so as to output travel plan information for dealing with a traffic situation when traffic information indicating the traffic situation is input; a second planning unit (2) that is constituted by a large language model and obtains second travel plan information by inputting traffic information into a second model (M2) that is constructed so as to output travel plan information for dealing with the traffic situation when traffic information is input; and a selection unit (3) that selects the first travel plan information or the second travel plan information on the basis of the travel state of an autonomous vehicle (V).

Description

Control system, control program, and control method

[0001] The present disclosure relates to a control system, a control program, and a control method.

[0002] In recent years, autonomous driving has reached the "Level 4" stage. Level 4 autonomous driving allows a vehicle to carry passengers without a driver under specific conditions (such as limited locations and remote human monitoring). For example, Non-Patent Document 1 below discloses the current status and challenges of such autonomous driving vehicles.

[0003] "Japan's first self-driving "Level 4" vehicle accident caused by insufficient training data," [online], November 13, 2023, Nikkei XTECH, [accessed January 10, 2024], Internet <URL: https: / / xtech.nikkei.com / atcl / nxt / column / 18 / 00001 / 08613 / ?i_cid=nbpnxt_reco_atype>

[0004] Conventional autonomous vehicle control systems using general trained models can safely handle events similar to those indicated by previously trained data when the autonomous vehicle is traveling. In other words, conventional control systems are unable to handle unknown events when they occur, potentially resulting in accidents such as those described in Non-Patent Document 1. In other words, in order for autonomous vehicles to travel safely while ensuring safety at least equivalent to that of human drivers, it is necessary to prepare a huge amount of training data corresponding to all events, including unknown events, and train the model using that data. However, preparing such a large amount of training data and training the model using that data requires a great deal of effort and time.

[0005] In order to solve the above problems, a control system according to one embodiment of the present disclosure is a control system for controlling the operation of an autonomous vehicle, and includes: a first planning unit that inputs traffic information acquired by the autonomous vehicle into a first model that is constructed so that, when traffic information indicating traffic conditions is input, it outputs driving plan information to respond to the traffic conditions, and obtains first driving plan information corresponding to the autonomous vehicle; a second planning unit that inputs the traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed so that, when the traffic information is input, it outputs driving plan information to respond to the traffic conditions, and obtains second driving plan information corresponding to the autonomous vehicle; and a selection unit that selects the first driving plan information or the second driving plan information based on the driving conditions of the autonomous vehicle.

[0006] A control program according to another aspect of the present disclosure causes a computer that controls the operation of an autonomous vehicle to execute a selection process to select, based on the driving status of the autonomous vehicle, first driving plan information corresponding to the autonomous vehicle obtained by inputting traffic information acquired by the autonomous vehicle into a first model that is configured to, when traffic information indicating traffic conditions is input, output driving plan information for dealing with the traffic conditions, or second driving plan information corresponding to the autonomous vehicle obtained by inputting the traffic information acquired by the autonomous vehicle into a second model that is configured of a large-scale language model and is configured to, when the traffic information is input, output driving plan information for dealing with the traffic conditions. Note that a computer-readable recording medium having a control program recorded thereon is also within the scope of the present disclosure.

[0007] Furthermore, a control method according to another aspect of the present disclosure is a control method for controlling the operation of an autonomous vehicle, and includes: a first planning step of inputting traffic information acquired by the autonomous vehicle into a first model that is constructed so that, when traffic information indicating traffic conditions is input, it outputs driving plan information to respond to the traffic conditions, thereby obtaining first driving plan information corresponding to the autonomous vehicle; a second planning step of inputting the traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed so that, when the traffic information is input, it outputs driving plan information to respond to the traffic conditions, thereby obtaining second driving plan information corresponding to the autonomous vehicle; and a selection step of selecting the first driving plan information or the second driving plan information based on the driving conditions of the autonomous vehicle.

[0008] FIG. 1 is a block diagram showing an example of a functional configuration of a control system according to embodiment 1 of one aspect of the present disclosure. FIG. 2 is a block diagram showing another example of a control system according to embodiment 1. FIG. 3 is a diagram explaining a modified example of a method for outputting a second evaluation value performed by a second planning unit provided in the system. FIG. 4 is a diagram explaining a modified example of a method for outputting a second evaluation value performed by a second planning unit provided in the system. FIG. 5 is a block diagram showing an example of a functional configuration of a control system according to embodiment 2 of one aspect of the present disclosure. FIG. 6 is a flowchart showing an example of the flow of a control method according to an embodiment of another aspect of the present disclosure. FIG. 7 is a block diagram showing an example of a computer required when at least some functions of the control systems according to embodiments 1 and 2 of the present disclosure are realized by software.

[0009] First Embodiment of Control System First, a control system 100 according to a first embodiment of one aspect of the present disclosure will be described in detail.

[0010] [Configuration of Control System 100] The control system 100 is a system that controls the operation of an autonomously driven vehicle V. As shown in FIG. 1 , the control system 100 includes a first planning unit 1, a second planning unit 2, and a selection unit 3. The control system 100 according to this embodiment further includes a first model M1, a second model M2, a classification model M3, an evaluation unit 4, an adjustment unit 5, a control unit 6, and a transmission unit 7. The units of the control system 100 according to this embodiment are distributed across multiple devices (the autonomously driven vehicle V and the server S). Specifically, the first planning unit 1 and the first model M1 are provided in the autonomously driven vehicle V. Meanwhile, the second planning unit 2, the second model M2, the selection unit 3, the classification model M3, the evaluation unit 4, the adjustment unit 5, the control unit 6, and the transmission unit 7 are provided in the server S.

[0011] (First model M1) The first model M1 is constructed by machine learning so that, when traffic information indicating a traffic situation is input, it outputs driving plan information for responding to the traffic situation. The first model M1 can be a model that has been used in conventional control systems for autonomous vehicles.

[0012] (First planning unit 1) The first planning unit 1 inputs traffic information acquired by the autonomously driven vehicle V into the first model M1, and acquires the driving plan information output by the first model M1 as first driving plan information corresponding to the autonomously driven vehicle V. The "traffic information acquired by the autonomously driven vehicle V" includes at least one of video, ranging information, and various types of detection information. The video is generated by a camera V1 equipped on the autonomously driven vehicle V, capturing an image of at least one of the outside and inside of the autonomously driven vehicle V. The ranging information is information indicating the distance between the autonomously driven vehicle V and an obstacle, generated by a ranging device V2 equipped on the autonomously driven vehicle V. For example, Lidar (Light Detection And Ranging) is used for ranging. The various types of detection information are information indicating what has been detected by various sensors V3 equipped on the autonomously driven vehicle V.

[0013] (Second Model M2) The second model M2 is configured as a large-scale language model (LLM) and is constructed by machine learning so that, when traffic information is input, it outputs trip plan information to respond to the traffic situation. The second model M2 according to the embodiment may be configured as, for example, any of "BERT," "Sentence Bert," "GPT4," "Gemini," and "Llama3." The second model M2 may also be configured as a combination of at least two of these. The second model M2 may also be configured as a combination of at least one of these and another model. Because the second model M2 is configured as a large-scale language model, it can output trip plan information corresponding to an unknown event even when traffic information indicating an unknown event that has not been learned is input.

[0014] (Second planning unit 2) The second planning unit 2 inputs traffic information acquired by the autonomously driven vehicle V into the second model M2, and acquires the driving plan information output by the second model M2 as second driving plan information corresponding to the autonomously driven vehicle V.

[0015] (Classification model M3) The classification model M3 is constructed by machine learning so that, when the first driving plan information and the second driving plan information are input, the classification model M3 determines whether the first driving plan information or the second driving plan information is appropriate. The classification model M3 according to this embodiment classifies the input first driving plan information and second driving plan information into either safe driving information or unsafe driving information based on the driving conditions of the autonomously driven vehicle V at that time.

[0016] (Selection Unit 3) The selection unit 3 selects the first driving plan information or the second driving plan information based on the driving conditions of the autonomously driven vehicle V. The selection unit 3 according to this embodiment selects the first driving plan information or the second driving plan information using the classification model M3. Furthermore, the selection unit 3 according to this embodiment selects the second driving plan information if driving plan information is not obtained within a predetermined time after the first planning unit 1 inputs traffic information to the first model M1. This is because, if traffic information indicating an unknown event that has not been learned by the first model M1 is input, the first model M1 may take a long time to make inference or may stop inference. Furthermore, the selection unit 3 according to this embodiment selects the first driving plan information when the autonomously driven vehicle V is driving or when the speed at which the autonomously driven vehicle V is driving is equal to or greater than a predetermined speed. This is because, due to hallucination of the second model M2 (large-scale language model), the autonomously driven vehicle V may cause an accident if it operates based on erroneous second driving plan information while driving (at high speed). Conversely, the selection unit 3 according to this embodiment selects the second driving plan information regardless of whether the first driving plan information has been obtained if the autonomously driven vehicle V has been stopped for a predetermined period of time or if the speed at which the autonomously driven vehicle V is traveling is equal to or less than a predetermined speed. This is because, due to the hallucination of the second model M2, even if the autonomously driven vehicle V operates based on erroneous second driving plan information, the autonomously driven vehicle V is unlikely to cause an accident while stopped or traveling at a low speed. Note that the selection unit 3 may be configured to select the first driving plan information or the second driving plan information (using an algorithm) without using the classification model M3.

[0017] (Evaluation unit 4) The evaluation unit 4 evaluates the selection result of the selection unit 3. The evaluation unit 4 according to this embodiment determines whether the autonomously driven vehicle V was able to travel safely (whether an emergency stop was not made) as a result of autonomous driving based on the driving plan information selected by the selection unit 3.

[0018] (Adjustment Unit 5) The adjustment unit 5 adjusts the classification model M3 based on the results of the evaluation by the evaluation unit 4. For example, suppose that the selection unit 3 selects one of the first driving plan information and the second driving plan information, and the autonomously driving vehicle V performs autonomous driving based on that one driving plan information, but is unable to drive safely. In this case, when selecting first driving plan information and second driving plan information with the same content, the adjustment unit 5 adjusts the classification model M3 so that the other driving plan information is selected.

[0019] (Control unit 6) The control unit 6 controls the autonomously driven vehicle V based on the driving plan information selected by the selection unit 3. If the driving plan information includes content to accelerate or drive at a constant speed, the control unit 6 controls the drive mechanism V4 of the autonomously driven vehicle V. As a result, the autonomously driven vehicle V accelerates or drives at the specified speed. If the driving plan information includes content to change direction, the control unit 6 controls the steering mechanism V5 of the autonomously driven vehicle V. As a result, the autonomously driven vehicle V changes its direction of travel. Furthermore, if the driving plan information includes content to decelerate or stop, the control unit 6 controls the braking mechanism V6 of the autonomously driven vehicle V. As a result, the autonomously driven vehicle V decelerates or stops.

[0020] (Transmitter 7) When the selector 3 selects the second driving plan information, the transmitter 7 transmits the traffic information that the second planner 2 inputs to the second model M2 to a control system 200 (hereinafter, another control system 200) of another autonomously driving vehicle V. The other control system 200 that has received the traffic information uses the received traffic information to, for example, fine-tune the trained model provided in the other control system 200.

[0021] [Variation of Control System 100] As shown in FIG. 2 , the control system 100 may be configured such that the second model M2 generates a second driving plan when first driving plan information is input and determines whether the input first driving plan information or second driving plan information is the appropriate driving plan information. The second planning unit 2A may also serve as the selection unit 3 of the control system 100 according to the first embodiment. That is, the second planning unit 2A (selection unit) may be configured to select the first driving plan information or the second driving plan information based on the determination result of the second model M2. In this case, the transmission unit 7 may be configured to transmit at least one of the first driving plan information, the second driving plan information, and the selection result of the selection unit (information indicating whether the first driving plan information or the second driving plan information was selected) to the remote monitoring device.

[0022] [Action and Effect of Control System 100] For example, as shown in the upper part of Fig. 3, when a parked vehicle and a person are present ahead of a traveling autonomous vehicle V, both the conventional control system and the control system 100 according to this embodiment recognize the presence of the parked vehicle and the person. Then, both the conventional control system and the control system 100 according to this embodiment stop the autonomous vehicle V in front of the person. Then, when both the conventional control system and the control system 100 according to this embodiment recognize that the person has disappeared from the path, they resume autonomous driving and control the autonomous vehicle V to avoid the parked vehicle.

[0023] On the other hand, for example, as shown in the lower part of FIG. 2 , if a fire engine and firefighters are present ahead of a traveling autonomous vehicle V and the firefighters are engaged in firefighting activities, the fire is an unknown event to conventional control systems. Therefore, conventional control systems cannot distinguish this from a case where only a parked vehicle and a person are present ahead, and as shown in the left side of the lower part of FIG. 2 , there is a possibility that the autonomous vehicle V will continue to stop until the firefighting activities are completed and the firefighters are no longer in the path of travel. Meanwhile, in the control system 100 described above, the second model M2, which is composed of a large-scale language model, distinguishes the fire engine and firefighters from ordinary parked vehicles and people (recognizing that a fire is being extinguished) and outputs second driving plan information for driving while avoiding the fire engine and firefighters (e.g., by passing through the opposite lane). The selection unit 3 then selects the second driving plan information unless there is a problem with safe driving. As a result, the control unit 6 controls the autonomous vehicle V based on the second driving plan information, allowing the autonomous vehicle V to continue driving safely.

[0024] Furthermore, as shown in FIG. 4 , if road construction (paving) is taking place ahead of an autonomously driven vehicle V while it is traveling, the road construction is also an unknown event to conventional control systems. Therefore, conventional control systems cannot distinguish roads under construction from normal roads, and there is a possibility that the autonomously driven vehicle V may enter the road under construction, as shown on the left side of FIG. 4 . Meanwhile, in the control system 100 described above, the second model M2 distinguishes the road under construction from normal roads (recognizing that the road is closed) and outputs second driving plan information for changing course (e.g., making a U-turn, turning right or left, etc.). The selection unit 3 then selects the second driving plan information. As a result, the control unit 6 controls the autonomously driven vehicle V based on the second driving plan information, allowing the autonomously driven vehicle V to continue traveling safely.

[0025] In this way, even if the autonomously driven vehicle V encounters an unknown event, the control system 100 outputs second transportation plan information that enables the second model M2 to respond to the unknown event. Therefore, according to the control system 100, when an unknown event occurs while the autonomously driven vehicle V is traveling, the autonomously driven vehicle V can be driven safely, without preparing a huge amount of learning data including unknown events and having the model learn that data. Furthermore, such an effect also contributes to the achievement of, for example, Goal 9 "Build resilient infrastructure, promote inclusive and sustainable industrialization," and Goal 11 "Make cities and towns inclusive and sustainable," of the Sustainable Development Goals (SDGs) advocated by the United Nations.

[0026] Next, a control system 100A according to a second embodiment of the present disclosure will be described. For ease of explanation, components having the same functions as those described in the first embodiment will be denoted by the same reference numerals, and the description thereof will not be repeated.

[0027] [Configuration of Control System 100A] As shown in FIG. 5 , the control system 100A according to this embodiment includes a first planning unit 1, a second planning unit 2, a selection unit 3, a first model M1, a second model M2, a classification model M3, and a control unit 6, similar to the control system 100 according to the embodiment. The components of the control system 100A according to this embodiment are also distributed across multiple devices (an autonomously driven vehicle V and a server S). However, the manner in which the components are distributed according to this embodiment differs from that of the first embodiment. Specifically, the first planning unit 1, the first model M1, the selection unit 3, the classification model M3, and the control unit 6 are provided in the autonomously driven vehicle V. Meanwhile, the second planning unit 2 and the second model M2 are provided in the server S. The control system 100A may further include at least one of an evaluation unit 4, an adjustment unit 5, and a transmission unit 7. In this case, the evaluation unit 4, the adjustment unit 5, and the transmission unit 7 may be provided in the autonomously driven vehicle V or in the server S.

[0028] [Effects and Functions of Control System 100A] The control system 100A described above provides the same effects as the control system 100 according to embodiment 1. That is, the control system 100A makes it possible to safely drive the autonomously driven vehicle V when an unknown event occurs while the autonomously driven vehicle V is traveling, without having to prepare a huge amount of learning data including unknown events and have the model learn that data.

[0029] <Control Method S100> Next, a control method S100 according to another embodiment of the present disclosure will be described. For ease of explanation, the same reference numerals will be used to designate components having the same functions as those described in the first embodiment, and the description thereof will not be repeated.

[0030] [Flow of Control Method S100] As shown in Fig. 6, the control method S100 is a method for controlling the operation of an autonomously driven vehicle V. As shown in Fig. 6, the control method S100 includes a first planning step S1, a second planning step S2, and a selection step S3. The control method S100 according to this embodiment further includes an evaluation step S4, an adjustment step S5, a control step S6, a transmission step S7, and a determination step S8.

[0031] (First Planning Step S1) In the first planning step S1, the computer inputs traffic information acquired by the autonomously driven vehicle V into the first model M1 to obtain first driving plan information corresponding to the autonomously driven vehicle V. The computer that communicates with the first model M1 may be a device that includes the first planning unit 1 in the control system 100, 100A, or may be another device.

[0032] (Second planning step S2) In the second planning step S2, the computer inputs traffic information acquired by the autonomously driven vehicle V into the second model M2 to obtain second driving plan information corresponding to the autonomously driven vehicle V. The computer that communicates with the second model M2 may be a device that includes the second planning unit 2 in the control system 100, 100A, or may be another device.

[0033] (Selection Step S3) After obtaining the first driving plan information and the second driving plan information, the process proceeds to selection step S3. In selection step S3, the computer selects the first driving plan information or the second driving plan information based on the driving conditions of the autonomously driving vehicle V. The computer that selects the driving plan information may be a device that includes the selection unit 3 in the control system 100, 100A, or may be another device.

[0034] (Evaluation Step S4) After selecting the first driving plan information or the second driving plan information, evaluation step S4 is performed as necessary. In evaluation step S4, the computer evaluates the selection result made by the computer in selection step S3. The computer that evaluates the selection result may be a device that includes the evaluation unit 4 in the control system 100, 100A, or may be another device.

[0035] (Adjustment Step S5) After the selection results have been evaluated, the process proceeds to adjustment step S5. In adjustment step S5, the computer adjusts the classification model M3 based on the results of the evaluation of the selection results in evaluation step S4. The computer that adjusts the classification model M3 may be a device that includes the adjustment unit 5 in the control system 100, 100A, or may be another device.

[0036] (Control Step S6) After selecting the first driving plan information or the second driving plan information, the process proceeds to control step S6. In control step S6, the computer controls the autonomously driven vehicle V based on the driving plan information selected in selection step S3. The computer controlling the autonomously driven vehicle V may be a device including the control unit 6 in the control system 100, 100A, or may be another device.

[0037] (Transmission Step S7) After the computer selects the second driving plan information in the selection step S3, it performs the transmission step S7 as needed. In the transmission step S7, the computer transmits the traffic information input to the second model M2 to the other control system 200. The computer that transmits the traffic information to the other control system 200 may be a device that includes the transmitter 7 in the control system 100, 100A, or may be another device.

[0038] (Determination step S8) After controlling the autonomously driven vehicle V based on the driving plan information, the process proceeds to determination step S8. In determination step S8, it is determined whether autonomous driving has ended (for example, whether the destination has been reached). If it is determined that autonomous driving has not ended (S8: NO), the process returns to step S1. That is, in the control method S100 according to this embodiment, the first planning step S1, the second planning step S2, the selection step S3, and the control step S6 are repeated until autonomous driving ends. On the other hand, if it is determined that autonomous driving has ended, the control method S100 ends.

[0039] [Effects of Control Method S100] The control method S100 described above provides the same effects as the control system 100 according to embodiment 1. That is, the control method S100 makes it possible to safely drive the autonomously driven vehicle V when an unknown event occurs while the autonomously driven vehicle V is traveling, without having to prepare a huge amount of learning data including unknown events and have the model learn that data.

[0040] <Modifications> The present disclosure is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present disclosure.

[0041] For example, each unit in the control system 100, including at least the selection unit 3, can be realized by a control program P that causes a computer C of the autonomously driving vehicle V or the server S to function as each unit. In this case, as shown in FIG. 7 , each unit includes a computer having at least one processor C1 and at least one memory C2 as hardware for executing the control program P. Each unit is realized by the processor C1 and the memory C2 executing each process (first planning process, second planning process, evaluation process, adjustment process, control process, and transmission process) including at least the selection process of the control program P. Note that the control system 100 may be configured so that one computer realizes one of the units, or so that one computer realizes two or more of the units.

[0042] Furthermore, the control program P may be stored not temporarily but on one or more computer-readable storage media. Each unit may or may not have a storage medium. In the latter case, the control program P may be supplied to each unit via any wired or wireless transmission medium.

[0043] In addition, some or all of the functions of each unit can be realized by a logic circuit. For example, an integrated circuit in which a logic circuit that functions as each unit is formed is also included in the scope of this disclosure. In addition, the functions of each unit can also be realized by, for example, a quantum computer.

[0044] [Summary] A control system according to aspect 1 of the present disclosure is a control system for controlling the operation of an autonomous vehicle, and is configured to include: a first planning unit that inputs traffic information acquired by the autonomous vehicle into a first model that is constructed so that, when traffic information indicating traffic conditions is input, it outputs driving plan information to respond to the traffic conditions, and obtains first driving plan information corresponding to the autonomous vehicle; a second planning unit that inputs the traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed so that, when the traffic information is input, it outputs driving plan information to respond to the traffic conditions, and obtains second driving plan information corresponding to the autonomous vehicle; and a selection unit that selects the first driving plan information or the second driving plan information based on the driving conditions of the autonomous vehicle.

[0045] A control system according to aspect 2 of the present disclosure may be configured such that, in the above-described aspect 1, when the first driving plan information and the second driving plan information are input, the selection unit selects either the first driving plan information or the second driving plan information using a classification model constructed to determine which of the first driving plan information and the second driving plan information is the appropriate driving plan information.

[0046] A control system according to aspect 3 of the present disclosure may be configured such that, in aspect 1 or 2 above, the second model is configured to determine, when the first driving plan information is input, whether the input first driving plan information or the second driving plan information is the appropriate driving plan information, and the selection unit selects either the first driving plan information or the second driving plan information based on the determination result of the second model.

[0047] A control system according to aspect 4 of the present disclosure may be configured such that, in any of aspects 1 to 3 above, the selection unit selects the second driving plan information if the first driving plan information is not obtained within a predetermined time after the first planning unit inputs the traffic information into the first model.

[0048] A control system according to aspect 5 of the present disclosure may be configured as in aspect 2 above, further comprising an evaluation unit that evaluates the selection result of the selection unit, and an adjustment unit that adjusts the classification model based on the result of the evaluation by the evaluation unit.

[0049] A control system according to aspect 6 of the present disclosure may be configured in any one of aspects 1 to 5 above, further including a control unit that controls the autonomous vehicle based on the driving plan information selected by the selection unit.

[0050] The control system according to aspect 7 of the present disclosure may be configured in any of aspects 1 to 6 above, further comprising a transmitting unit that transmits the traffic information input by the second planning unit to the second model to a control system of another autonomous vehicle when the selecting unit selects the second driving plan information.

[0051] A control program according to aspect 8 of the present disclosure is configured to cause a computer that controls the operation of an autonomous vehicle to execute a selection process that selects first driving plan information corresponding to the autonomous vehicle, obtained by inputting traffic information acquired by the autonomous vehicle into a first model that is constructed to output driving plan information to respond to traffic conditions when traffic information indicating traffic conditions is input, based on the driving status of the autonomous vehicle; or second driving plan information corresponding to the autonomous vehicle, obtained by inputting traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed to output driving plan information to respond to traffic conditions when the traffic information is input.

[0052] A control method according to aspect 9 of the present disclosure is a control method for controlling the operation of an autonomous vehicle, the method including: a first planning step of inputting traffic information acquired by the autonomous vehicle into a first model that is constructed so that, when traffic information indicating traffic conditions is input, it outputs driving plan information to respond to the traffic conditions, thereby obtaining first driving plan information corresponding to the autonomous vehicle; a second planning step of inputting the traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed so that, when the traffic information is input, it outputs driving plan information to respond to the traffic conditions, thereby obtaining second driving plan information corresponding to the autonomous vehicle; and a selection step of selecting the first driving plan information or the second driving plan information based on the driving conditions of the autonomous vehicle.

[0053] 100, 100A Control system 1 First planning unit 2, 2A Second planning unit 3 Selection unit 4 Evaluation unit 5 Adjustment unit 6 Control unit 7 Transmission unit M1 First model M2 Second model M3 Classification model V Self-driving vehicle V1 Camera V2 Ranging device V3 Sensor V4 Drive mechanism V5 Steering mechanism P Control program 200 Other control system S100 Control method S1 First planning step S2 Second planning step S3 Selection step S4 Evaluation step S5 Adjustment step S6 Control step S7 Transmission step

Claims

1. A control system for controlling the operation of an autonomous vehicle, comprising: a first planning unit that inputs traffic information acquired by the autonomous vehicle into a first model that is constructed so that, when traffic information indicating traffic conditions is input, it outputs driving plan information to respond to the traffic conditions, and obtains first driving plan information corresponding to the autonomous vehicle; a second planning unit that inputs the traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed so that, when the traffic information is input, it outputs driving plan information to respond to the traffic conditions, and obtains second driving plan information corresponding to the autonomous vehicle; and a selection unit that selects the first driving plan information or the second driving plan information based on the driving conditions of the autonomous vehicle.

2. The control system of claim 1, wherein when the first driving plan information and the second driving plan information are input, the selection unit selects either the first driving plan information or the second driving plan information using a classification model constructed to determine which of the first driving plan information and the second driving plan information is the more appropriate driving plan information.

3. The control system of claim 1, wherein the second model is constructed to determine, when the first driving plan information is input, whether the input first driving plan information or the second driving plan information is the appropriate driving plan information, and the selection unit selects either the first driving plan information or the second driving plan information based on the judgment result of the second model.

4. The control system of claim 1, wherein the selection unit selects the second driving plan information if the first driving plan information is not obtained within a predetermined time after the first planning unit inputs the traffic information into the first model.

5. The control system according to claim 2, further comprising: an evaluation unit that evaluates a selection result of the selection unit; and an adjustment unit that adjusts the classification model based on a result of the evaluation by the evaluation unit.

6. The control system according to claim 1, further comprising a control unit that controls the autonomous vehicle based on the driving plan information selected by the selection unit.

7. The control system of claim 1, further comprising a transmitting unit that transmits the traffic information input by the second planning unit to the second model to a control system of another autonomous vehicle when the selecting unit selects the second driving plan information.

8. A control program that causes a computer that controls the operation of an autonomous vehicle to execute a selection process that selects, based on the driving status of the autonomous vehicle, first driving plan information corresponding to the autonomous vehicle, obtained by inputting traffic information acquired by the autonomous vehicle into a first model that is constructed so that, when traffic information indicating traffic conditions is input, it outputs driving plan information to respond to the traffic conditions; or second driving plan information corresponding to the autonomous vehicle, obtained by inputting traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed so that, when traffic information is input, it outputs driving plan information to respond to the traffic conditions.

9. A control method for controlling the operation of an autonomous vehicle, comprising: a first planning step of inputting traffic information acquired by the autonomous vehicle into a first model that is constructed so that, when traffic information indicating traffic conditions is input, it outputs driving plan information to deal with the traffic conditions, thereby obtaining first driving plan information corresponding to the autonomous vehicle; a second planning step of inputting the traffic information acquired by the autonomous vehicle into a second model that is composed of a large-scale language model and is constructed so that, when the traffic information is input, it outputs driving plan information to deal with the traffic conditions, thereby obtaining second driving plan information corresponding to the autonomous vehicle; and a selection step of selecting the first driving plan information or the second driving plan information based on the driving conditions of the autonomous vehicle.

Images