Motion prediction device, motion prediction method, and computer program for motion prediction

Abstract

To provide an operation prediction device capable of appropriately predicting operation of a traffic participant existing in the periphery of a vehicle.SOLUTION: An operation prediction device detects a traffic participant existing in a predetermined range from peripheral data representing a situation in the predetermined range in the periphery of a vehicle, determines whether or not there is a blind spot area that is not represented in the peripheral data in the predetermined range, and when it is determined that there is the blind spot area, assumes that a virtual traffic participant exists in the blind spot area, and predicts operation of the traffic participant caused by existence of the virtual traffic participant.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present disclosure relates to an action prediction device, an action prediction method, and a computer program for action prediction that predict the actions of traffic participants existing around a vehicle.

Background Art

[0002] In order to safely operate a vehicle, it is important to appropriately predict the actions of traffic participants existing around the vehicle, and in the case of autonomous driving, control the running of the vehicle so as not to approach the future position of the traffic participant, and in the case of manual driving, alert the driver.

[0003] For example, the future behavior estimation device described in Patent Document 1 recognizes the position of a traffic participant, determines a temporary goal that the traffic participant is about to reach in the future based on the recognition result, and simulates the movement process of the traffic participant toward the temporary goal using a movement model.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] There may be blind spots around a vehicle where traffic participants cannot be detected by surrounding sensors because they are blocked by objects such as structures and other vehicles. If the future situation is predicted only for traffic participants that can be detected by the surrounding sensors, the situation caused by traffic participants existing in the blind spots may not be appropriately predicted.

[0006] An object of the present disclosure is to provide an action prediction device that can appropriately predict the actions of traffic participants existing around a vehicle.

Means for Solving the Problems

[0007] The gist of the present disclosure is as follows.

[0008] (1) A detection unit that detects traffic participants existing in the predetermined range from peripheral data representing the situation of a predetermined range among the periphery of the vehicle, a determination unit that determines whether there is a blind spot area not represented in the peripheral data in the predetermined range, a hypothesis unit that, when it is determined that there is the blind spot area, assumes that there is a virtual traffic participant in the blind spot area, a prediction unit that predicts the actions of the traffic participants caused by the existence of the virtual traffic participant, and an action prediction device including the same.

[0009] (2) The prediction unit in the above (1), predicts the actions of the virtual traffic participant according to traffic rules applied to the blind spot area where the existence of the virtual traffic participant is assumed, and predicts the actions of the traffic participants caused by the existence of the virtual traffic participant after the predicted actions.

[0010] (3) The hypothesis unit in the above (1) or (2) assumes that among a plurality of virtual traffic participants that can be assumed to exist in the blind spot area, a virtual traffic participant having a greater possibility of affecting the movement route of the traffic participant than a predetermined value exists in the blind spot area.

[0011] (4) An action prediction method including: an action prediction device that predicts the actions of traffic participants existing around a vehicle, detects traffic participants existing in the predetermined range from peripheral data representing the situation of a predetermined range among the periphery of the vehicle, determines whether there is a blind spot area not represented in the peripheral data in the predetermined range, when it is determined that there is the blind spot area, assumes that there is a virtual traffic participant in the blind spot area, and predicts the actions of the traffic participants caused by the existence of the virtual traffic participant.

[0012] (5) Detecting traffic participants existing in the predetermined range from the peripheral data representing the situation in a predetermined range of the periphery of the vehicle; Determining whether there is a blind spot area not represented in the peripheral data in the predetermined range; When it is determined that there is the blind spot area, assuming that there is a virtual traffic participant in the blind spot area; Predicting the actions of the traffic participants due to the existence of the virtual traffic participant; A computer program for action prediction to be executed by a computer mounted on the vehicle.

[0013] According to the action prediction device according to the present disclosure, the actions of traffic participants existing around the vehicle can be appropriately predicted.

Brief Description of the Drawings

[0014]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Embodiments for Carrying Out the Invention

[0015] Hereinafter, with reference to the drawings, an operation prediction device that can appropriately predict the operations of traffic participants existing around a vehicle will be described in detail. The operation prediction device detects traffic participants existing in a predetermined range from peripheral data representing the situation in a predetermined range of the periphery of the vehicle. The predetermined range is the detection range of a peripheral sensor that generates the peripheral data. Further, the operation prediction device determines whether there is a blind spot area not represented in the peripheral data in the predetermined range, and if it is determined that there is a blind spot area, it assumes that there is a virtual traffic participant in the blind spot area. Then, the operation prediction device predicts the operations of the traffic participants caused by the presence of the virtual traffic participant.

[0016] FIG. 1 is a schematic configuration diagram of a vehicle in which the operation prediction device is mounted.

[0017] The vehicle 1 includes a peripheral sensor 2, a GNSS receiver 3, a storage device 4, and an operation prediction device 5. The peripheral sensor 2, the GNSS receiver 3, the storage device 4, and the operation prediction device 5 are communicably connected via an in-vehicle network conforming to a standard such as a controller area network.

[0018] The peripheral sensor 2 generates peripheral data representing the situation around the vehicle 1. The peripheral sensor 2 has a LiDAR (Light Detection And Ranging) sensor that generates a distance image as peripheral data in which each pixel has a value corresponding to the distance to the object represented by the pixel based on the situation around the vehicle 1. The peripheral sensor 2 is attached, for example, to the upper front inside the vehicle cabin facing forward. The peripheral sensor 2 outputs a peripheral distance image as peripheral data representing the situation around the vehicle 1 via the windshield at a predetermined imaging period (for example, 1 / 30 second to 1 / 10 second). Note that the vehicle 1 may have, as the peripheral sensor 2, a sensor other than the LiDAR sensor, for example, a peripheral camera that outputs a peripheral image in which the situation around the vehicle 1 is captured. The peripheral camera includes a two-dimensional detector composed of an array of photoelectric conversion elements sensitive to visible light, such as a CCD or a C-MOS, and an imaging optical system that forms an image of an area to be imaged on the two-dimensional detector.

[0019] The GNSS receiver 3 receives GNSS signals from GNSS (Global Navigation Satellite System) satellites at a predetermined period, and measures the self-position of the vehicle 1 based on the received GNSS signals. The GNSS receiver 3 outputs, at a predetermined period, a positioning signal representing the positioning result of the self-position of the vehicle 1 based on the GNSS signals to the operation prediction device 5 via the in-vehicle network.

[0020] The storage device 4 is an example of a storage unit, and has, for example, a hard disk device or a non-volatile semiconductor memory. The storage device 4 stores map data including information on features such as lane division lines and information representing applicable traffic rules in association with positions. The traffic rules applicable to a certain position are, for example, the traffic methods required to be followed by traffic participants existing at that position. The traffic rules may include the traffic lanes in which the vehicle should travel and not interfering with the passage of pedestrians, and for pedestrians, may include walking on the sidewalk and crossing the road through the crosswalk.

[0021] The operation prediction device 5 detects traffic participants existing in a predetermined range from the peripheral data generated by the peripheral sensor 2. Further, the operation prediction device 5 determines whether there is a blind spot area not represented in the peripheral data in the predetermined range, and if it is determined that there is a blind spot area, assumes that there is a virtual traffic participant in the blind spot area. Then, the operation prediction device 5 predicts the actions of traffic participants due to the presence of the virtual traffic participant.

[0022] FIG. 2 is a hardware schematic diagram of the operation prediction device 5. The operation prediction device 5 includes a communication interface 51, a memory 52, and a processor 53.

[0023] The communication interface 51 is an example of a communication unit, and has a communication interface circuit for connecting the operation prediction device 5 to the in-vehicle network. The communication interface 51 supplies the received data to the processor 53. Further, the communication interface 51 outputs the data supplied from the processor 53 to the outside.

[0024] The memory 52 includes a volatile semiconductor memory and a non-volatile semiconductor memory. The memory 52 stores various data used for processing by the processor 53, such as parameters of an identifier for detecting traffic participants from peripheral data, etc. Further, the memory 52 stores various application programs, such as an operation prediction program for executing operation prediction processing, etc.

[0025] The processor 53 is an example of a control unit and includes one or more processors and their peripheral circuits. The processor 53 may further include other arithmetic circuits such as a logical arithmetic unit, a numerical arithmetic unit, or a graphic processing unit.

[0026] FIG. 3 is a functional block diagram of the processor 53 included in the operation prediction device 5. FIG. 4 is a diagram for explaining an example of operation prediction.

[0027] The processor 53 of the operation prediction device 5 includes, as functional blocks, a detection unit 531, a determination unit 532, a hypothesis unit 533, and a prediction unit 534. Each of these units included in the processor 53 is a functional module implemented by a program executed on the processor 53. A computer program for realizing the functions of each unit of the processor 53 may be provided in a form recorded on a computer-readable portable recording medium such as a semiconductor memory, a magnetic recording medium, or an optical recording medium. Alternatively, each of these units included in the processor 53 may be implemented in the operation prediction device 5 as an independent integrated circuit, a microprocessor, or firmware.

[0028] The detection unit 531 acquires peripheral data representing the situation within a predetermined range among the surroundings of the vehicle 1 from the peripheral sensor 2 via the communication interface 51. Then, the detection unit 531 detects traffic participants existing within the predetermined range from the peripheral data.

[0029] The detection unit 531 detects traffic participants existing within a predetermined range by inputting the surrounding data received into a discriminator that has been pre-trained to detect traffic participants such as other vehicles and pedestrians.

[0030] The discriminator can be, for example, a convolutional neural network (CNN) having a plurality of convolutional layers connected in series from the input side to the output side. By using a large number of data including traffic participants as teacher data in advance and training the CNN according to a predetermined learning method such as the error backpropagation method, the CNN operates as a discriminator that detects traffic participants from the data. Also, machine learning algorithms such as a support vector machine (SVM) and AdaBoost may be used for the discriminator. When the discriminator is an SVM, the SVM is trained to define support vectors for identifying whether various regions on the surrounding data include traffic participants, and thus the SVM operates as a discriminator that detects traffic participants.

[0031] Further, the detection unit 531 may detect the ground features around the vehicle 1 by inputting the surrounding data received into a discriminator that has been pre-trained to detect ground features such as road signs, road markings, and street trees. The discriminator can be a pre-trained CNN using teacher data including ground features. The detection unit 531 may use the pre-trained CNN using teacher data including traffic participants and teacher data including ground features as a discriminator to detect traffic participants and ground features from the surrounding data.

[0032] In the example of FIG. 4, the surrounding data represents the situation of a predetermined range A1 around the vehicle 1 traveling on the lane L1. The other vehicle 100-1 entering the intersection from the lane L2 and traveling toward the lane L3 is an example of a traffic participant detected from the surrounding data. Also, the crosswalks PX1-PX4 are examples of ground features detected from the surrounding data.

[0033] The determination unit 532 determines whether there is a blind spot area not represented in the surrounding data within the predetermined range.

[0034] For example, the determination unit 532 collates the features represented by the feature information acquired from the storage device 4 according to the self-position and orientation represented by the positioning signal received from the GNSS receiver 3 with the features detected from the surrounding data. When any of the features represented by the feature information is not included in the features detected from the surrounding data, the determination unit 532 determines that there is a dead angle area due to an obstacle between the feature not included in the features detected from the surrounding data and the self-position, and that the feature is not represented in the surrounding data. At this time, the determination unit 532 estimates, for the vehicle 1, the area behind the object represented at the position where the feature of the surrounding data should be detected within a predetermined range as the range of the dead angle area.

[0035] In the example of FIG. 4, in a predetermined range A1 around the vehicle 1, the situation behind the shielding object O1 is not represented in the surrounding data because it is shielded by the shielding object O1. Therefore, the area behind the shielding object O1 with respect to the vehicle 1 within the predetermined range A1 becomes the dead angle area BA. The determination unit 532 specifies the azimuths toward the left and right outer edges of the range in which the shielding object O1 is represented from the surrounding sensor 2, and sets, as the dead angle area BA, the area within the area sandwiched by the specified azimuths where the distance from the surrounding sensor 2 is longer than that of the shielding object O1.

[0036] Alternatively, the determination unit 532 may determine the presence or absence of a dead angle area based on the distances represented by the respective pixels of the surrounding data. For example, when the surrounding sensor 2 is a LiDAR, in the distance image output as the surrounding data, the background such as the road surface on the concentric circle centered on the self-position is represented at substantially the same distance. Among the pixels corresponding to the circumference of such a concentric circle, the pixels represented at a distance closer than the background are considered to correspond to the foreground (for example, a feature or a traffic participant) that is not the background. In particular, in the pixels represented at a distance with a large difference from the distance to the background, there is a high possibility that another object is hidden behind it (it is a dead angle area). Therefore, when the size of the continuous area of the pixels represented at a distance with a difference from the distance to the background greater than a predetermined distance threshold is greater than a predetermined area threshold, the determination unit 532 may determine that there is a dead angle area in the surrounding data.

[0037] The assumption unit 533 obtains from the determination unit 532 the determination result as to whether or not there is a blind spot area in the surrounding data, and information representing the range of the blind spot area if the blind spot area exists. When there is a blind spot area in the surrounding data, the assumption unit 533 assumes that there is a virtual traffic participant in the blind spot area. The virtual traffic participant is a traffic participant assumed to exist in a predetermined area represented in the surrounding data generated by the surrounding sensor 2 regardless of the detection result from the surrounding data.

[0038] The assumption unit 533 can assume that there are a virtual traffic participant who is a pedestrian and a virtual traffic participant who is a vehicle in the blind spot area. At this time, the position within the blind spot area where the virtual traffic participant is assumed to exist may be any type of area existing within the blind spot area, such as a sidewalk, a crosswalk, or a lane. The assumption unit 533 refers to the self-position and orientation of the vehicle represented by the positioning signal received from the GNSS receiver 3 and the map data, identifies the area corresponding to the blind spot area on the map data, and identifies the type of area such as a sidewalk included in the identified area.

[0039] The assumption unit 533 may assume the existence of a virtual traffic participant according to the traffic rules applied to the blind spot area. For example, a reference table that associates the type of area with the type of virtual traffic participant whose existence can be assumed in the area according to the traffic rules applied to the area (for example, a sidewalk and a pedestrian) is stored in advance in the memory 52. The assumption unit 533 refers to the reference table and assumes the virtual traffic participants existing in each type of area within the blind spot area. For example, when it is estimated that there is a sidewalk within the blind spot area, the assumption unit 533 refers to the reference table and assumes that there is a virtual traffic participant who is a pedestrian within the sidewalk.

[0040] The prediction unit 534 predicts the actions of traffic participants caused by the presence of virtual traffic participants. The prediction unit 534 may predict the actions of virtual traffic participants according to the traffic rules applied to the blind spot area where the presence of virtual traffic participants is assumed. In this case, the prediction unit 534 predicts the actions of traffic participants caused by the presence of virtual traffic participants after the predicted actions. By operating in this way, the action prediction device 5 can appropriately predict the actions of traffic participants based on the actions of virtual traffic participants existing in the blind spot area.

[0041] In the example of FIG. 4, the assumption unit 533 can assume the presence of a virtual traffic participant that is a vehicle at a position within the lane L3 included in the blind spot area BA according to the traffic rules applied to the blind spot area BA. Similarly, the assumption unit 533 can assume the presence of a virtual traffic participant VTP that is a pedestrian at a position within the sidewalk SW included in the blind spot area BA.

[0042] For each possible action of the virtual traffic participant, the prediction unit 534 obtains the probability of performing that action (action probability) according to the traffic rules. Further, for each assumed action of the virtual traffic participant, the prediction unit 534 multiplies the action probability by the probability that the action of the traffic participant detected in the assumed action is affected (influence probability) to predict the possibility that the presence of the virtual traffic participant after the assumed action affects the action of the detected traffic participant. The action probability and the influence probability may be stored in the storage device 4 included in the map data as information representing the traffic rules associated with the position.

[0043] For example, the prediction unit 534 predicts, according to the traffic rules, that the action probability of a virtual traffic participant that is a vehicle located within the lane L3 traveling forward in the forward direction in front of the traffic participant is, for example, 95%. Also, assume that the influence probability that the action of the traffic participant is affected by the presence of a virtual traffic participant traveling forward in the forward direction is 10%. At this time, the prediction unit 534 predicts that the possibility that the presence of the virtual traffic participant (other vehicles in front of the lane L3) after the predicted action affects the action of the other vehicle 100-1 is 95%×10% = 9.5%.

[0044] Similarly, the prediction unit 534 predicts, according to the traffic rules, that the operation probability of a virtual traffic participant, who is a pedestrian located near the crosswalk PX3 within the sidewalk SW, moving to the crosswalk PX3 is, for example, 70%. In this traffic rule, when there is a pedestrian attempting to cross the crosswalk, it is stipulated that the vehicle should stop in front of the crosswalk so as not to obstruct the pedestrian's passage, and the influence probability of the vehicle's operation being affected by the presence of the pedestrian is set at 95%. Based on this traffic rule, the prediction unit 534 predicts that the other vehicle 100-1 traveling from lane L2 to lane L3 has a 70%×95% = 67% probability of stopping in front of the crosswalk PX3 (the position of the other vehicle 100-2) (the presence of the virtual traffic participant after the predicted operation affects the operation of the other vehicle 100-1).

[0045] Similarly, the prediction unit 534 predicts, according to the traffic rules, that the operation probability of a virtual traffic participant, who is a pedestrian located at a location away from the crosswalk PX3 within the sidewalk SW, moving to the crosswalk PX3 is, for example, 20%. Then, the prediction unit 534 predicts that the presence of the virtual traffic participant (the pedestrian walking on the crosswalk PX3) after the predicted operation has a, for example, 20%×95% = 19% probability of affecting the operation of the other vehicle 100-1.

[0046] When the possibility that the predicted operation of the above-mentioned virtual traffic participant affects the movement route of the traffic participant detected by the detection unit 531 is less than a predetermined value (for example, 30%), the assumption unit 533 may exclude the assumption that the virtual traffic participant exists. By operating in this way, the assumption unit 533 can omit the treatment for virtual traffic participants with little influence on the movement route of the traffic participant and reduce the processing load.

[0047] In the above example, the possibility that the operation predicted for the virtual traffic participant, who is a pedestrian located near the crosswalk PX3 within the sidewalk SW, affects the movement route of the traffic participant (the other vehicle 100-1) is greater than the predetermined value. Therefore, the assumption unit 533 assumes that there is a virtual traffic participant who is a pedestrian located near the crosswalk PX3 within the sidewalk SW.

[0048] FIG. 5 is a flowchart of the operation prediction process. While the vehicle 1 is running, the processor 53 of the operation prediction device 5 repeatedly executes the operation prediction process described below at a predetermined cycle.

[0049] First, the detection unit 531 of the processor 53 of the operation prediction device 5 detects traffic participants existing within a predetermined range in the vicinity of the vehicle 1 from the peripheral data generated by the peripheral sensor 2 (step S1).

[0050] The determination unit 532 of the processor 53 determines whether there is a blind spot area not represented in the peripheral data within the predetermined range (step S2). If it is determined that there is no blind spot area within the predetermined range (step S2: N), the determination unit 532 ends the operation prediction process. In this case, a process of predicting the behavior of the traffic participants detected from the peripheral data without assuming virtual traffic participants may be executed.

[0051] If it is determined that there is a blind spot area within the predetermined range (step S2: Y), the assumption unit 533 of the processor 53 assumes that there is a virtual traffic participant in the blind spot area (step S3). Subsequently, the prediction unit 534 of the processor 53 predicts the behavior of the traffic participants resulting from the presence of the virtual traffic participant (step S4), and ends the operation prediction process.

[0052] By executing the operation prediction process in this way, it is possible to appropriately predict the behavior of the traffic participants existing around the vehicle.

[0053] The vehicle 1 may have a travel control device (not shown) that controls the travel of the vehicle 1 so that the distance from the vehicle 1 to the surrounding traffic participants is an appropriate length. The travel control device controls the travel of the vehicle 1 so that the distance from the vehicle 1 to the traffic participants whose actions are predicted by the operation prediction device 5 is an appropriate length.

[0054] When it is predicted that the distance to an object such as a traffic participant around the vehicle 1 is less than a predetermined interval threshold value, the vehicle 1 may have a notification device (not shown) that notifies the driver of the vehicle 1 via notification devices (not shown) such as a display, a speaker, and a lamp mounted on the vehicle 1. The notification device notifies the driver of the vehicle 1 when the distance to the traffic participant whose operation is predicted by the operation prediction device 5 is less than the interval threshold value.

[0055] In the present embodiment, an example of predicting the operation of a traffic participant by assuming a virtual traffic participant crossing a crosswalk at an intersection has been described with reference to FIG. 4. Needless to say, the operation prediction device 5 of the present disclosure is also applicable to predicting the operation of traffic participants in other road situations.

[0056] For example, assume that the vehicle 1 and a preceding vehicle traveling in front of the vehicle 1 are about to pass by two oncoming vehicles stopped in the oncoming lane. Among the plurality of two oncoming vehicles stopped in the oncoming lane, the vehicle on the near side is a truck loaded with a container, and behind the truck is a passenger car, and the vehicle 1 is traveling on the side of the truck. The preceding vehicle is, for example, in a position straddling the lane in which the vehicle 1 is traveling and the oncoming lane in an attempt to pass between the truck and the passenger car in order to enter a parking lot of a store on the oncoming lane side.

[0057] The detection unit 531 of the operation prediction device 5 detects a truck, a passenger car, and a preceding vehicle existing in a predetermined range around the vehicle 1 from the surrounding data generated by the surrounding sensor 2 mounted on the vehicle 1.

[0058] The detection range of the surrounding sensor 2 includes the container of the truck, and features (such as guardrails and street trees) on the sidewalk side of the oncoming lane behind the container cannot be detected from the surrounding data. Therefore, the determination unit 532 of the operation prediction device 5 determines that there is a dead angle area in a predetermined range around the vehicle 1.

[0059] The assumption unit 533 of the motion prediction device 5 refers to the current position and orientation of the vehicle 1 represented in the map data and the latest positioning information, and identifies the type of area included in the blind spot area. For example, it is identified that an area from the passenger car to the road edge (the sidewalk side end of the lane) within the blind spot area is a bicycle passing zone where traffic rules prioritizing bicycle travel are applied. The assumption unit 533 assumes, in accordance with the traffic rules, that there is a virtual traffic participant that is a bicycle in the bicycle passing zone within the blind spot area.

[0060] The prediction unit 534 of the motion prediction device 5 predicts that the probability of a virtual traffic participant that is a bicycle, whose existence is assumed, traveling in the forward direction (the same direction as the traveling direction of the oncoming lane) along the bicycle passing zone in accordance with the traffic rules is 90%. Then, when the virtual traffic participant that is a bicycle approaches between the stop positions of the truck and the passenger car from the side of the passenger car by traveling in the forward direction along the bicycle passing zone, the prediction unit 534 of the motion prediction device 5 predicts that the probability of the leading vehicle stopping at a position straddling the lane in which the vehicle 1 is traveling and the oncoming lane is 70%. The prediction unit 534 multiplies the probability of influence by the probability of operation to predict that the probability that the existence of the virtual traffic participant (bicycle) after the assumed operation affects the operation of the traffic participant (leading vehicle) is 63%. Based on the prediction of the operation of the traffic participant (leading vehicle) by the motion prediction device 5, the vehicle 1 can suppress acceleration or decelerate in preparation for a sudden stop of the leading vehicle.

[0061] It should be understood that those skilled in the art can make various changes, substitutions, and modifications to this without departing from the spirit and scope of the present disclosure.

Explanation of Reference Numerals

[0062] 1 Vehicle 5 Motion Prediction Device 531 Detection Unit 532 Determination Unit 533 Assumption Unit 534 Prediction Unit

Claims

1. A detection unit that detects traffic participants existing in the predetermined range from the peripheral data representing the situation of a predetermined range among the periphery of the vehicle; A determination unit that determines whether there is a blind spot area not represented in the peripheral data in the predetermined range; A hypothesis unit that, when it is determined that there is the blind spot area, assumes that a plurality of virtual traffic participants exist in the blind spot area; Predict the actions of each virtual traffic participant according to the traffic rules stored in the storage unit corresponding to the position of the blind spot area, and predict the actions of the traffic participants caused by the existence of virtual traffic participants whose predicted existence after the action has a greater possibility of affecting the movement route of the traffic participants than a predetermined value among the plurality of virtual traffic participants; Comprising: The prediction unit, for each action of each predicted virtual traffic participant, multiplies the probability that each virtual traffic participant performs the action according to the traffic rules by the probability that the action of the traffic participant is affected by the action, thereby obtaining the possibility that the existence of each virtual traffic participant after the action affects the movement route of the traffic participant; An action prediction device.

2. An action prediction device for predicting the actions of traffic participants existing around a vehicle, Detects traffic participants existing in the predetermined range from the peripheral data representing the situation of a predetermined range among the periphery of the vehicle, Determines whether there is a blind spot area not represented in the peripheral data in the predetermined range, When it is determined that there is the blind spot area, assumes that a plurality of virtual traffic participants exist in the blind spot area, Predicts the actions of each virtual traffic participant according to the traffic rules stored in the storage unit corresponding to the position of the blind spot area, For each action of each predicted virtual traffic participant, multiplies the probability that each virtual traffic participant performs the action according to the traffic rules by the probability that the action of the traffic participant is affected by the action, thereby obtaining the possibility that the existence of each virtual traffic participant after the action affects the movement route of the traffic participant, Predicting the actions of the traffic participants caused by the existence of virtual traffic participants whose predicted existence after the action has a greater possibility of affecting the movement route of the traffic participants than a predetermined value among the plurality of virtual traffic participants; An action prediction method including this.

3. Detecting traffic participants existing in the predetermined range from the peripheral data representing the situation of a predetermined range among the periphery of the vehicle; Determining whether there is a blind spot area not represented in the peripheral data in the predetermined range; When it is determined that there is the blind spot area, assuming that there are a plurality of virtual traffic participants in the blind spot area; Predicting the actions of each virtual traffic participant according to the traffic rules stored in the storage unit in association with the position of the blind spot area; For each predicted action of each virtual traffic participant, multiplying the probability that each virtual traffic participant performs the action according to the traffic rules by the probability that the action of the traffic participant is affected by the action, to obtain the possibility that the subsequent existence of the action of each virtual traffic participant affects the moving route of the traffic participant; Predicting the action of the traffic participant due to the existence of a virtual traffic participant whose subsequent existence after the predicted action among the plurality of virtual traffic participants has a possibility of affecting the moving route of the traffic participant greater than a predetermined value; An action prediction computer program that causes a computer mounted on the vehicle to execute the above.

Images