Abnormality determination device

Abstract

To suitably determine presence or absence of abnormality of a periphery recognition model of an automatic operating system.SOLUTION: An abnormality determination device which determines presence or absence of abnormality of a periphery recognition model which is a machine learning model used for periphery recognition of an automatic operating system of one's own vehicle includes: a prediction position estimation unit which, on the basis of a detection result of an external sensor of the own vehicle, uses the periphery recognition model to estimate a prediction position of a moving body around the own vehicle at a predetermined point in time; and an abnormality determination unit which, on the basis of a result of comparison between the prediction positions of the plurality of moving bodies at the predetermined point in time and detection positions of the plurality of moving bodies at the predetermined point in time, determines the presence or absence of the abnormality of the periphery recognition model.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to an abnormality determination device. [Background technology]

[0002] Conventionally, JP 2022-032109 A is known as a technical document related to an abnormality determination device. This publication describes a collision prevention system that uses detection results obtained from multiple different types of detectors to determine the reliability of the movement characteristics of an object that is to be determined as a collision, and determines that the error in the calculated posture or velocity vector of the target object is large and the reliability is reduced based on a predetermined index (the deviation between the posture and velocity vector of the target object). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2022-032109 A Summary of the Invention [Problem to be solved by the invention]

[0004] By the way, the use of a machine learning model for the surrounding recognition of the vehicle by the autonomous driving system is being considered. In this case, when an abnormality occurs in the surrounding recognition of the autonomous driving system, it is required to appropriately determine whether the abnormality is in the sensor or the machine learning model. [Means for solving the problem]

[0005] One aspect of the present invention is an abnormality determination device that determines whether or not there is an abnormality in a surrounding recognition model, which is a machine learning model used for surrounding recognition of an autonomous driving system of a host vehicle, and includes a predicted position estimation unit that estimates a predicted position of a moving body around the host vehicle at a predetermined time using the surrounding recognition model based on detection results from an external sensor of the host vehicle, and an abnormality determination unit that determines whether or not there is an abnormality in the surrounding recognition model based on a comparison result between the predicted positions of multiple moving bodies at a predetermined time and the detected positions of the multiple moving bodies at a predetermined time.

[0006] According to an abnormality determination device of one embodiment of the present invention, when an abnormality occurs in the surrounding recognition model of an autonomous driving system, it is expected that a characteristic effect will occur in the comparison result between the predicted positions of multiple moving bodies at a specified time and the detected positions of multiple moving bodies at a specified time, and therefore it is possible to appropriately determine whether or not there is an abnormality in the surrounding recognition model based on the comparison result.

[0007] In an abnormality judgment device according to one embodiment of the present invention, the abnormality judgment unit may judge that an abnormality has occurred in the surrounding recognition model when there are a certain number or more moving bodies whose difference between the predicted position and the detected position is equal to or greater than a judgment threshold, and when it is judged that a predetermined abnormality occurrence condition is satisfied based on the predicted position and the detected position. Effect of the Invention

[0008] According to one aspect of the present invention, it is possible to appropriately determine whether or not there is an abnormality in a surrounding recognition model of an autonomous driving system. [Brief description of the drawings]

[0009] [Figure 1] 1 is a block diagram showing an automatic driving system (an abnormality determination device) according to an embodiment. FIG. [Diagram 2] 13 is a flowchart illustrating an example of a predicted position estimation process. [Diagram 3] 5 is a flowchart illustrating an example of an abnormality determination process. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0011] FIG. 1 is a block diagram showing an autonomous driving system (abnormality determination device) 100 according to an embodiment. The autonomous driving system 100 shown in FIG. 1 is a system that executes autonomous driving of a host vehicle 1. Autonomous driving is vehicle control that automatically drives the host vehicle 1 along a preset route or along a road on which the host vehicle 1 is traveling, without the driver performing any driving operation. Autonomous driving corresponds to, for example, autonomous driving level 2 or higher control certified by the Society of Automotive Engineers (SAE). Autonomous driving may be limited to autonomous driving level 3 or higher control.

[0012] The autonomous driving system 100 is configured to include an abnormality determination device 50. The abnormality determination device 50 is a device that determines whether or not there is an abnormality in the surrounding recognition performance of the autonomous driving system 100. The surrounding recognition performance is performance related to surrounding recognition (recognition of the external environment) of the host vehicle 1. Note that the abnormality determination device 50 does not necessarily need to constitute a part of the autonomous driving system 100, and may be provided independently of the autonomous driving system 100.

[0013] The autonomous driving system 100 includes an autonomous driving ECU 30 [Electronic Control Unit]. The autonomous driving ECU 30 is an electronic control unit having a CPU [Central Processing Unit] and a storage unit. The storage unit is composed of, for example, a ROM [Read Only Memory], a RAM [Random Access Memory], and an EEPROM [Electrically Erasable Programmable Read-Only Memory]. The autonomous driving ECU 30 realizes various functions by executing programs stored in the storage unit with the CPU. The autonomous driving ECU 30 may be composed of multiple electronic units.

[0014] As shown in FIG. 1, the autonomous driving ECU 30 is connected to an external camera 10 and a radar sensor 11, which are external sensors. The external camera 10 is an imaging device (external sensor) that images the external situation of the host vehicle 1. The external camera 10 is provided, for example, on the back side of the windshield of the host vehicle 1, and images the area in front of the host vehicle 1. The external camera 10 may be provided on the side or rear of the host vehicle 1, and may be configured to be able to image the surroundings of the host vehicle 1. The external camera 10 transmits the captured image of the outside of the host vehicle 1 to the autonomous driving ECU 30.

[0015] The radar sensor 11 is a detection device (external sensor) that detects objects around the vehicle 1 by using radio waves (e.g., millimeter waves) or light. The radar sensor 11 includes, for example, a millimeter wave radar or a LiDAR (Light Detection and Ranging) provided in multiple directions of the vehicle 1. The radar sensor 11 detects objects by transmitting radio waves or light to the surroundings of the vehicle and receiving the radio waves or light reflected by the objects. The radar sensor 11 transmits information on the detected objects to the autonomous driving ECU 30. The objects include fixed obstacles such as guardrails and buildings, as well as moving obstacles such as pedestrians, bicycles, and other vehicles.

[0016] Next, a functional configuration of the autonomous driving ECU 30 will be described. As shown in Fig. 1, the autonomous driving ECU 30 has a periphery recognition unit 31, an abnormality determination unit 32, and a vehicle control unit 33. In this embodiment, the periphery recognition unit 31 and the abnormality determination unit 32 configure the abnormality determination device 50 described above. Note that the abnormality determination device 50 does not necessarily need to include the entire periphery recognition unit 31, and it is sufficient if it includes a predicted position estimation unit 31c described later.

[0017] The surroundings recognition unit 31 recognizes the surroundings of the vehicle 1 (recognizes the external environment) in order to execute autonomous driving of the vehicle 1. The surroundings recognition unit 31 recognizes the surroundings of the vehicle 1 based on at least one of an image captured by the external camera 10 and a detection result of the radar sensor 11. The surroundings recognition unit 31 may recognize the surroundings of the vehicle 1 by so-called sensor fusion.

[0018] The surroundings recognition unit 31 includes a surroundings recognition model 31a, a detected position recognition unit 31b, and a predicted position estimation unit 31c. The surroundings recognition model 31a is a machine learning model used for surroundings recognition of the host vehicle 1.

[0019] The surrounding recognition model 31a is a machine learning model that has been trained by deep learning so as to output surrounding recognition results from at least one of, for example, an image captured by the external camera 10 and a detection result (information on detected objects) from the radar sensor 11. The surrounding recognition results include predicted positions of moving objects around the vehicle 1. A moving object is a movable object such as another vehicle, a pedestrian, or a bicycle. The surrounding recognition results may also include detected positions of moving objects around the vehicle 1. The detected positions and predicted positions will be described later.

[0020] The surrounding recognition model 31a is configured using a neural network. As the neural network, a convolutional neural network (CNN) including a plurality of layers including a plurality of convolution layers and a pooling layer can be used. The surrounding recognition model 31a may be configured as a recurrent neural network (RNN). The configuration of the surrounding recognition model 31a is not particularly limited as long as it is capable of outputting the surrounding recognition result as a machine learning model. The surrounding recognition model 31a may be configured of a plurality of machine learning models.

[0021] The detection position recognition unit 31b recognizes the detection position of a moving object around the vehicle 1 based on at least one of the captured image of the external camera 10 and the detection result of the radar sensor 11. A moving object around the vehicle 1 is a moving object included in the capture range of the external camera 10 or the detection range of the radar sensor 11 of the vehicle 1. Hereinafter, the capture range and the detection range are referred to as the sensor field of view. The sensor field of view is determined in advance according to, for example, the specifications of each device and the user's settings. The user is a passenger or driver of the vehicle 1. The detection position of a moving object is the position of the moving object detected by the external camera 10 or the radar sensor 11. The detection position is recognized, for example, in a vehicle coordinate system based on the vehicle 1.

[0022] The detection position recognition unit 31b may recognize the detection position of the moving object using the surroundings recognition model 31a. The detection position recognition unit 31b may output the detection position of the moving object around the host vehicle 1 from the surroundings recognition model 31a using at least one of the captured image of the external camera 10 and the detection result of the radar sensor 11 as input. The detection position recognition unit 31b may recognize the traveling direction and the moving speed of the moving object in addition to the detection position.

[0023] The predicted position estimation unit 31c estimates a predicted position of a moving object around the vehicle 1 using the surrounding recognition model 31a. The predicted position estimation unit 31c estimates a predicted position of the moving object at a predetermined time in the future. The predetermined time may be 1 second ahead, 2 seconds ahead, 5 seconds ahead, or 10 seconds ahead. The predetermined time is not particularly limited. The predicted position estimation unit 31c may estimate predicted positions for a plurality of predetermined times.

[0024] The predicted position estimation unit 31c receives the detected position of the moving object by the detected position recognition unit 31b as input and outputs the predicted position of the moving object from the surroundings recognition model 31a. The predicted position estimation unit 31c may use the detected positions of the moving object multiple times within a certain period of time in the past as input. The predicted position estimation unit 31c may use the predicted position of the moving object estimated one position before as input. The predicted position estimation unit 31c may use the traveling direction and moving speed of the moving object recognized by the detected position recognition unit 31b as input.

[0025] The abnormality determination unit 32 determines whether or not there is an abnormality in the surrounding recognition model 31a. The abnormality determination unit 32 determines whether or not there is an abnormality in the surrounding recognition model 31a based on a comparison result between the predicted positions of the multiple moving bodies at a predetermined time and the detected positions of the multiple moving bodies at a predetermined time. Note that, when the surrounding recognition model 31a includes multiple machine learning models, it is assumed that the same machine learning model is used to estimate the predicted positions.

[0026] First, the narrowing down of moving objects used in the abnormality determination of the surrounding recognition model 31a (excluding moving objects exhibiting abnormal behavior) will be described. The abnormality determination unit 32 performs the abnormality determination by excluding abnormally behaving moving objects that exhibit abnormal behavior from among the moving objects in the vicinity of the host vehicle 1.

[0027] For example, when there is a moving object with an exceptionally large difference between the detected position and the predicted position, even though the average value of the differences between the detected position and the predicted position of other moving objects around the vehicle 1 is small, the abnormality determination unit 32 excludes the moving object as an abnormally behaving object. That is, when the average value of the differences between the detected position and the predicted position of other moving objects is less than the exclusion start determination threshold, the abnormality determination unit 32 excludes the moving object as an abnormally behaving object when the difference between the detected position and the predicted position of one moving object and the average value is equal to or greater than the abnormal behavior threshold. The exclusion start determination threshold and the abnormal behavior threshold can be set to any value used in determining an abnormally behaving object.

[0028] In addition, when the difference between the detected position and the predicted position of one moving object described above is not a distance that can actually be moved, the one moving object is not excluded as an abnormally behaving object because it may be related to an abnormality in the surrounding recognition model 31a. Whether or not it is a distance that can actually be moved can be determined based on, for example, the attributes of the moving object (attributes such as four-wheeled vehicle, two-wheeled vehicle, bicycle, pedestrian, etc.) and the time difference between the time when the prediction was made and the time at the predicted position. An upper limit value of the expected speed change is set for each attribute of the moving object. In addition, the abnormality determination unit 32 may exclude a moving object in an overtaking scene as an abnormally behaving object. An overtaking scene is a scene in which the target moving object is a vehicle, a low-speed vehicle or an obstacle is present in front of the target moving object, and the adjacent lane is empty.

[0029] When the abnormality determination unit 32 has excluded the abnormally behaving object, the abnormality determination unit 32 may notify the user of the presence of the abnormally behaving object. The abnormality determination unit 32 notifies the user of the position of the abnormally behaving object by at least one of an image display on the display of the vehicle 1 and an audio output from the speaker.

[0030] Next, the abnormality determination of the surrounding recognition model 31a will be described. First, the abnormality determination unit 32 determines whether or not there are a certain number or more of moving objects whose difference between the predicted position and the detected position at a certain time is equal to or greater than the determination threshold. The certain number may be 2, 5, or 10. There is no particular limit to the certain number.

[0031] The abnormality determination unit 32 calculates the difference (the distance between the predicted position and the detected position) between the predicted position of a moving object, for example, at 14:00, and the detected position of the moving object actually detected when 14:00 is reached, and counts the number of moving objects for which the difference is equal to or greater than the determination threshold. The determination threshold is a threshold value that is appropriately set for abnormality determination of the surrounding recognition model 31a. Hereinafter, a moving object for which the difference between the predicted position and the detected position is equal to or greater than the determination threshold is referred to as a target moving object.

[0032] The abnormality determination unit 32 determines that no abnormality has occurred in the surroundings recognition model 31a when there are not a certain number of moving objects whose difference between the predicted position and the detected position at a given time is equal to or greater than the determination threshold.

[0033] When there are a certain number or more of target moving objects, the abnormality determination unit 32 determines whether or not a preset abnormality occurrence condition is satisfied. For example, when the detection positions of a certain number of target moving objects are outside the field of view of the sensor, the abnormality determination unit 32 determines that the abnormality occurrence condition is satisfied. The certain number may be a value different from the above-mentioned certain number, or may be the same value. The certain number can be a number equal to or greater than 1.

[0034] The abnormality determination unit 32 may determine that the abnormality occurrence condition is satisfied when the detected positions of a predetermined number of target moving objects are biased to a predetermined range within the sensor field of view as viewed from the vehicle 1. Biased to a predetermined range refers to, for example, a case where all target moving objects or a predetermined number of target moving objects are detected collectively in a range diagonally to the right as viewed from the vehicle 1, even though the predicted positions of the target moving objects have been estimated to be located in the front, rear, left and right of the vehicle 1 at a predetermined time.

[0035] The abnormality determination unit 32 may determine that the abnormality occurrence condition is satisfied when the directions of the deviation vectors between the detected positions and the predicted positions of a predetermined number of target moving objects at a first predetermined time and the deviation vectors between the detected positions and the predicted positions of the same target moving objects at a second predetermined time are different from each other, and the magnitude of each deviation vector is equal to or greater than a vector threshold. The deviation vector is, for example, a vector calculated to connect the detected positions and the predicted positions in a planar view. The directions of the deviation vectors are different when the angle formed by the deviation vectors at multiple predetermined times for the same moving object is equal to or greater than a certain angle. The magnitude of the deviation vector corresponds to the distance between the detected positions and the predicted positions.

[0036] The abnormality determination unit 32 may determine that the abnormality occurrence condition is satisfied when a state in which a certain number or more of target moving objects are present continues for a certain period of time. The certain period of time is not particularly limited and may be any period of time.

[0037] The abnormality determination unit 32 may determine that the abnormality occurrence condition is satisfied when the increase or decrease in the number of detected positions of the target moving object at a given time relative to the number of predicted positions of the target moving object at the given time is equal to or greater than the error threshold. The error threshold may be 3, 5, or 8. The error threshold is not particularly limited. The abnormality determination unit 32 may perform the above determination under the prerequisite that the surrounding environment of the vehicle 1 is unlikely to change. An example of a case where the surrounding environment of the vehicle 1 is unlikely to change is a case where the vehicle speed of the vehicle 1 is less than a certain speed and the surrounding visibility is good.

[0038] The abnormality determination unit 32 may determine that the abnormality occurrence condition is satisfied when there are a predetermined number or more of target moving objects whose difference between the detected position and the predicted position is equal to or greater than the abnormality threshold. The abnormality threshold is not particularly limited as long as it is a value greater than the determination threshold used to extract the target moving object. Note that the abnormality determination unit 32 may be configured not to perform the above determination in a scene where the moving object needs to suddenly change its running state, such as in a construction zone.

[0039] The abnormality determination unit 32 may set the abnormality threshold to a different value depending on the attribute of the target moving object (attributes such as four-wheeled vehicle, two-wheeled vehicle, bicycle, pedestrian, etc.). The abnormality determination unit 32 may use the abnormality threshold according to the attribute of the target moving object to determine that an abnormality has occurred in the surrounding recognition model 31a when there are a predetermined number or more target moving objects whose difference between the detected position and the predicted position is equal to or greater than the abnormality threshold.

[0040] The abnormality determination unit 32 may determine that the abnormality occurrence condition is satisfied when the difference (absolute value) between the change in predicted position between a plurality of predetermined times for a predetermined number of target moving objects and the change in the detected position between the predetermined times is equal to or greater than a tolerance threshold. The tolerance threshold is not particularly limited, and different values ​​may be set according to the attributes of the target moving objects.

[0041] The abnormality determination unit 32 may determine that an abnormality occurrence condition is satisfied when the detection positions of a predetermined number of target moving objects overlap with other moving objects or structures such as walls. The abnormality determination unit 32 may determine the overlap using the size of the target moving object that is determined in advance according to the attributes of the target moving object, for example. The abnormality determination unit 32 may recognize the actual size of each target moving object based on the captured image of the external camera 10 or the detection result of the radar sensor 11, and determine the overlap using the actual size of the target moving object.

[0042] The abnormality determination unit 32 may determine that the abnormality occurrence condition is satisfied when the comparison result between the detected position and the predicted position of the target moving object resembles a preset model abnormality occurrence pattern. The model abnormality occurrence pattern is a pattern of the comparison result between the predicted position and the detected position stored in advance for abnormality determination of the surrounding recognition model 31a. The model abnormality occurrence pattern includes, for example, a pattern in which the detected position or the predicted position of the target moving object is biased to a specific range of the sensor field of view of the vehicle 1, a pattern in which a blind spot area occurs in which the predicted position will not be detected next time when it enters a specific range of the sensor field of view of the vehicle 1, and the like. The model abnormality occurrence pattern is preset based on the simulation data of the surrounding recognition model 31a.

[0043] When it is determined that the abnormality occurrence condition is satisfied, the abnormality determination unit 32 determines that an abnormality has occurred in the surrounding recognition model 31a. The abnormality determination unit 32 warns the user of the abnormality occurrence and transmits an abnormality occurrence signal to the vehicle control unit 33.

[0044] The vehicle control unit 33 performs automatic driving of the host vehicle 1 based on the surrounding recognition result (including the detected position and predicted position of the moving object) by the surrounding recognition unit 31. The vehicle control unit 33 performs automatic driving by controlling the traveling of the host vehicle 1, for example, by transmitting control signals to an engine actuator, a brake actuator, and a steering actuator (not shown).

[0045] When the abnormality determination device 50 (abnormality determination unit 32) determines that there is an abnormality in the surrounding recognition model 31a of the automatic driving system 100 during automatic driving of the vehicle 1, the vehicle control unit 33 warns the user of the occurrence of the abnormality. When the automatic driving cannot be continued, the vehicle control unit 33 notifies the user that the automatic driving cannot be continued, and then ends the automatic driving.

[0046] Next, a control method of the abnormality determination device 50 according to this embodiment will be described with reference to the drawings. Fig. 2 is a flowchart showing an example of a predicted position estimation process. The predicted position estimation process is executed as part of the automatic driving process when the host vehicle 1 is in an automatic driving state.

[0047] 2, the abnormality determination device 50 (autonomous driving system 100) recognizes the detection position of a moving object around the host vehicle 1 by the detection position recognition unit 31b in S1. The detection position recognition unit 31b recognizes the detection position of a moving object around the host vehicle 1 based on at least one of the captured image of the external camera 10 and the detection result of the radar sensor 11. After that, the abnormality determination device 50 proceeds to S2.

[0048] As S2, the abnormality determination device 50 uses the predicted position estimation unit 31c to estimate a predicted position of the moving object around the vehicle 1. The predicted position estimation unit 31c receives the detected position of the moving object by the detected position recognition unit 31b as an input and outputs the predicted position of the moving object from the surroundings recognition model 31a.

[0049] 3 is a flowchart showing an example of the abnormality determination process. The abnormality determination process is executed, for example, every time the predicted position estimation process of FIG.

[0050] 3, in S10, the abnormality determination device 50 compares the predicted position and the detected position of the moving object at each predetermined time by the abnormality determination unit 32. Thereafter, the abnormality determination device 50 proceeds to S11.

[0051] In S11, the abnormality determination device 50 uses the abnormality determination unit 32 to exclude abnormally behaving objects that are moving objects exhibiting abnormal behavior from among the moving objects around the vehicle 1. For example, when there is a moving object with an exceptionally large difference between the detected position and the predicted position even though the average value of the difference between the detected position and the predicted position of other moving objects around the vehicle 1 is small, the abnormality determination unit 32 excludes the moving object as an abnormally behaving object.

[0052] In S12, the abnormality determination device 50 determines whether or not there are a certain number or more of moving objects whose difference between the predicted position and the detected position is equal to or greater than the determination threshold by the abnormality determination unit 32. If the abnormality determination device 50 determines that there are a certain number or more of moving objects whose difference between the predicted position and the detected position is equal to or greater than the determination threshold (S12: YES), it proceeds to S13. If the abnormality determination device 50 does not determine that there are a certain number or more of moving objects whose difference between the predicted position and the detected position is equal to or greater than the determination threshold (S12: NO), it ends the current abnormality determination process.

[0053] In S13, the abnormality determination device 50 determines whether or not the abnormality occurrence condition is satisfied by the abnormality determination unit 32. If the abnormality determination device 50 determines that the abnormality occurrence condition is satisfied (S13: YES), the process proceeds to S14. If the abnormality determination device 50 does not determine that the abnormality occurrence condition is satisfied (S13: NO), the abnormality determination device 50 ends the current abnormality determination process.

[0054] In S14, the abnormality determination device 50 determines that an abnormality has occurred in the surrounding recognition model 31a by the abnormality determination unit 32. Thereafter, in S15, the abnormality determination device 50 notifies the user of the occurrence of the abnormality in the surrounding recognition model 31a by transmitting an abnormality occurrence signal to the vehicle control unit 33 by the abnormality determination unit 32. The vehicle control unit 33 may end the autonomous driving after warning the user as necessary.

[0055] According to the abnormality determination device 50 of the present embodiment described above, when an abnormality occurs in the surrounding recognition model of the autonomous driving system 100, it is considered that a characteristic effect will be produced in the comparison result between the predicted positions of the multiple moving bodies at a given time and the detected positions of the multiple moving bodies at a given time, and therefore it is possible to appropriately determine the presence or absence of an abnormality in the surrounding recognition model 31a based on the comparison result.

[0056] Although the embodiment of the present invention has been described above, the present invention is not limited to the above-mentioned embodiment. The present invention can be embodied in various forms including the above-mentioned embodiment and various modifications and improvements based on the knowledge of those skilled in the art.

[0057] The abnormality determination unit 32 does not necessarily have to exclude abnormally behaving objects. The abnormality determination unit 32 may use all detected moving objects for abnormality determination of the surrounding recognition model 31a. Moreover, the abnormality determination unit 32 does not necessarily have to extract a target moving object. The abnormality determination unit 32 may determine that an abnormality has occurred in the surrounding recognition model 31a when an abnormality occurrence condition is satisfied based on the detected position and predicted position of the moving object. [Explanation of symbols]

[0058] 1... host vehicle, 30... autonomous driving ECU, 31... surroundings recognition unit, 31a... surroundings recognition model, 31b... detected position recognition unit, 31c... predicted position estimation unit, 32... abnormality judgment unit, 33... vehicle control unit, 50... abnormality judgment device, 100... autonomous driving system.

Claims

1. An abnormality determination device that determines whether or not there is an abnormality in a surrounding recognition model, which is a machine learning model used for surrounding recognition of an autonomous driving system of a vehicle, a predicted position estimation unit that estimates a predicted position of a moving object around the host vehicle at a predetermined time using the surrounding recognition model based on a detection result of an external sensor of the host vehicle; an abnormality determination unit that determines whether or not there is an abnormality in the surrounding recognition model based on a comparison result between predicted positions of the plurality of moving bodies at the predetermined time and detected positions of the plurality of moving bodies at the predetermined time; An abnormality determination device comprising:

2. 2. The abnormality determination device according to claim 1, wherein the abnormality determination unit determines that an abnormality has occurred in the surrounding recognition model when a certain number or more of the moving bodies exist whose difference between the predicted position and the detected position is equal to or greater than a determination threshold, and when it is determined that a predetermined abnormality occurrence condition is satisfied based on the predicted position and the detected position.

3. The abnormality determination device according to claim 2 , wherein the abnormality determination unit determines that the abnormality occurrence condition is satisfied when an increase or decrease in the number of detected positions of the moving body at the specified time relative to a number of predicted positions of the moving body at the specified time is equal to or greater than an error threshold value.

4. The abnormality judgment unit determines whether or not there is an abnormality in the surrounding recognition model by excluding abnormal behaving objects among the moving objects that are exhibiting a predetermined abnormal behavior. An abnormality judgment device as described in any one of claims 1 to 3.

5. The abnormality determination device according to claim 4 , wherein the abnormality determination unit notifies a user of the host vehicle of the presence of the abnormally behaving object.

Images