WO2018139294A1 - Moving object prediction device - Google Patents

Moving object prediction device Download PDF

Info

Publication number
WO2018139294A1
WO2018139294A1 PCT/JP2018/001136 JP2018001136W WO2018139294A1 WO 2018139294 A1 WO2018139294 A1 WO 2018139294A1 JP 2018001136 W JP2018001136 W JP 2018001136W WO 2018139294 A1 WO2018139294 A1 WO 2018139294A1
Authority
WO
WIPO (PCT)
Prior art keywords
prediction
moving object
unit
vehicle
prediction unit
Prior art date
Application number
PCT/JP2018/001136
Other languages
French (fr)
Japanese (ja)
Inventor
龍 稲葉
茂規 早瀬
Original Assignee
日立オートモティブシステムズ株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日立オートモティブシステムズ株式会社 filed Critical 日立オートモティブシステムズ株式会社
Publication of WO2018139294A1 publication Critical patent/WO2018139294A1/en

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
    • B60W30/08—Active safety systems predicting or avoiding probable or impending collision or attempting to minimise its consequences
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems

Definitions

  • the present invention relates to a moving object prediction apparatus for predicting the behavior of a moving object existing around a vehicle such as an automobile, in particular, a self-driving vehicle.
  • Patent Document 1 Conventionally, a system has been developed that acquires object information around the course of the host vehicle and predicts the path of each moving object from the acquired object information. (For example, see Patent Document 1)
  • an object of the present invention is to enable reduction of calculation load while ensuring the prediction accuracy of a moving object that affects the host vehicle.
  • the present invention is a moving object prediction apparatus for predicting a future position of a moving object around a host vehicle, and includes a first prediction unit that simply predicts a future position of the moving object, and a first prediction unit A second prediction unit that predicts the future position of the moving object with high accuracy, and an allocation unit that allocates the moving object that predicts the future position by the second prediction unit according to the result of the simple prediction;
  • a moving object prediction apparatus comprising:
  • 1 is an explanatory diagram of a traveling drive system and sensors of a host vehicle according to a first embodiment.
  • 1 is a block diagram of a vehicle travel control apparatus according to a first embodiment.
  • the block diagram of the moving object action prediction part which concerns on Example 1 is shown.
  • the block diagram of the prediction object precision judgment part which concerns on Example 1 is shown.
  • the relationship of the own vehicle influence determination result and precision information which concern on Example 1 is shown.
  • An example of the own vehicle which concerns on Example 1, and the surrounding moving object is shown.
  • 3 shows a control flow according to the first embodiment.
  • the relationship of the influence degree to the pedestrian and the own vehicle which concerns on Example 1 is shown.
  • An example of the own vehicle which concerns on Example 1, and its surrounding object is shown.
  • the block diagram of the prediction object precision determination part which concerns on Example 2 is shown.
  • Example 2 An example of the moving object with interaction which concerns on Example 2 is shown. The relationship with the future time and interaction probability which concern on Example 2 is shown. The relationship of the own vehicle influence determination result which concerns on Example 2, and an interaction determination result is shown. An example of the own vehicle which concerns on Example 2, and the surrounding moving object is shown. The control flow which concerns on Example 2 is shown. 10 shows a control flow according to a third embodiment. The block diagram of the moving object action prediction part which concerns on Example 4 is shown.
  • FIG. 1 is an explanatory diagram showing the overall configuration of the host vehicle 100 equipped with the vehicle travel control device 1 according to the first embodiment.
  • the FL wheel, the FR wheel, the RL wheel, and the RR wheel mean a left front wheel, a right front wheel, a left rear wheel, and a right rear wheel, respectively.
  • the own vehicle 100 includes a camera 2 at the front, laser radars 3 and 4 at the left and right sides, and a millimeter wave radar 5 at the rear as sensors for recognizing the outside world. Distance and relative speed can be detected.
  • a combination of the above-described sensors is shown as an example of the configuration of the sensor.
  • the present invention is not limited to this, and a combination with an ultrasonic sensor, a stereo camera, an infrared camera, or the like may be used.
  • the sensor signal is input to the vehicle travel control device 1.
  • the host vehicle 100 includes a vehicle travel control device 1, a steering control device 8, a braking control device 15, an acceleration control device 19, a communication device 23, and a display device 24.
  • the vehicle travel control device 1 includes, for example, a CPU, a ROM, a RAM, and an input / output device.
  • the ROM stores a flow of vehicle travel control.
  • the vehicle travel control device 1 calculates command values to the steering control mechanism 10, the brake control mechanism 13, and the throttle control mechanism 20 for controlling the traveling direction based on information from the various sensors 2, 3, 4, and 5. .
  • the steering control device 8 controls the steering control mechanism 10 based on the command value from the vehicle travel control device 1.
  • the brake control device 15 controls the brake control mechanism 13 based on the command value from the vehicle travel control device 1 to adjust the brake force distribution of each wheel.
  • the acceleration control device 19 controls the throttle control mechanism 20 based on the command value from the vehicle travel control device 1 to adjust the torque output of the engine.
  • the communication device 23 performs communication between road vehicles or between vehicles.
  • the display device 24 displays a travel plan of the host vehicle 100 and a behavior prediction of a moving object existing in the vicinity.
  • the vehicle travel control device 1 configured as described above calculates command values of the actuators 10, 13, and 20 for controlling the vehicle travel according to the generated travel plan, which will be described in detail later.
  • the control devices 8, 15, and 19 of the actuators 10, 13, and 20 receive the command value of the vehicle travel control device 1 through communication, and control the actuators 10, 13, and 20 based on the command value.
  • the pedaling force of the driver stepping on the brake pedal 12 is boosted by a brake booster (not shown), and a hydraulic pressure corresponding to the force is generated by a master cylinder (not shown).
  • the generated hydraulic pressure is supplied to the wheel cylinder 16 via the brake control mechanism 13.
  • Each of the wheel cylinders 16FL to 16RR includes a cylinder (not shown), a piston, a pad, and the like.
  • the piston is propelled by the hydraulic fluid supplied from the master cylinder 9, and the pad connected to the piston is a disk rotor. Pressed.
  • the disc rotor rotates with the wheels. Therefore, the brake torque acting on the disc rotor becomes a braking force acting between the wheel and the road surface.
  • a braking force can be generated in each wheel in accordance with the driver's brake pedal operation.
  • the braking control device 15 includes, for example, a CPU, a ROM, a RAM, and an input / output device, like the vehicle travel control device 1.
  • the brake controller 15 includes a combine sensor 14, wheel speed sensors 11FL to 11RR, a brake force command from the brake controller 15 described above, and a sensor from the steering wheel angle detector 21 via a steering controller 8 described later. Signal.
  • the combine sensor 14 detects longitudinal acceleration, lateral acceleration, and yaw rate. Wheel speed sensors 11FL to 11RR are installed on each wheel.
  • the braking control device 15 is connected to a pump (not shown) on the output side and a brake control mechanism 13 having a control valve, and applies an arbitrary braking force to each wheel independently of the driver's brake pedal operation. Can be generated.
  • the braking control device 15 estimates the spin, drift-out, and wheel lock of the host vehicle 100 based on the various information described above, and generates the braking force of the corresponding wheel so as to suppress them, thereby stabilizing the steering of the driver. It plays a role to improve sex. Furthermore, the vehicle travel control device 1 can generate an arbitrary braking force on the host vehicle 100 by communicating a brake command to the braking control device 15. It plays the role of braking. However, in the present embodiment, the present invention is not limited to the braking control device 15, and other actuators such as a brake-by-wire may be used.
  • the steering control device 8 controls the motor 9 based on the information and generates assist torque.
  • the steering control device 8 also has, for example, a CPU, a ROM, a RAM, and an input / output device, like the vehicle travel control device 1.
  • the steering control device 8 can control the steering control mechanism 10 by generating torque by the motor 9 independently of the driver's steering operation. Therefore, the vehicle travel control device 1 can control the front wheels to an arbitrary turning angle by communicating a steering force command to the steering control device 8, and automatically steers in automatic driving where no driver operation occurs. I have a role to do.
  • the present invention is not limited to the steering control device 8, and other actuators such as steer-by-wire may be used.
  • the acceleration control device 19 also has, for example, a CPU, a ROM, a RAM, and an input / output device, like the vehicle travel control device 1.
  • the acceleration control device 19 controls the engine by adjusting the throttle opening in accordance with the depression amount of the accelerator pedal 17.
  • the host vehicle 100 can be accelerated in accordance with the driver's operation of the accelerator pedal 17.
  • the acceleration control device 19 can control the throttle opening independently of the driver's accelerator operation. Therefore, the vehicle travel control device 1 can generate an arbitrary acceleration in the host vehicle 100 by communicating an acceleration command to the acceleration control device 19, and automatically in the case of an automatic driving in which no driver operation occurs. Has a role to accelerate.
  • FIG. 2 is a block diagram of the vehicle travel control apparatus 1 according to the present embodiment.
  • the sensor input processing unit 201 When information on the surrounding environment obtained based on information from various sensors 2, 3, 4, and 5 that recognize the outside world is input to the sensor input processing unit 201, a moving object that exists around the host vehicle 100 Converted to information. As specific object information, for example, attribute information of pedestrians, bicycles, and vehicles, and their current position and current speed vector are extracted.
  • the moving object includes a parked vehicle that may move in the future even if the speed obtained at the current time is zero.
  • the moving object behavior prediction unit 202 calculates future position and velocity information (object prediction information) of each moving object based on input information, which will be described in detail later.
  • the trajectory / motion planning unit 203 plans the travel trajectory of the host vehicle 100 and the speed at that time based on the object prediction information and map information that are input information. Specifically, for example, the trajectory / motion planning unit 203 generates a trajectory that avoids collision with these moving objects from the object prediction information of the future position of the traveling vehicle and the parked vehicle in front of the host vehicle 100. . At the same time, the track / motion planning unit 203 plans to drive automatically with acceleration / deceleration within a range in which the driver riding in the vehicle 100 does not feel as uncomfortable as possible.
  • the actuator command value calculation unit 204 calculates, for example, brake, steering, and accelerator operation amounts based on the trajectory / speed plan information that is input information. Specifically, since the track / speed plan information is target information in the future of the host vehicle 100, the actuator command value calculation unit 204 is based on the physical model of the host vehicle 100 when the target position and speed are input. Thus, the control amount of each actuator 10, 13, 20 is output.
  • FIG. 3 is a block diagram of the moving object behavior prediction unit 202 implemented in the vehicle travel control device 1 of the present embodiment.
  • the predicted object accuracy determination unit 301 calculates accuracy information of each object based on object information and trajectory / speed plan information at the time of the previous calculation, which will be described later in detail.
  • the object allocation unit 302 as “allocation unit” extracts the individual object information from the object information based on the accuracy information, and transmits it to the plurality of individual prediction units 303 and 304.
  • the plurality of individual prediction units 303 and 304 includes two sets of a plurality of individual prediction units 303 with low prediction accuracy and a set of individual prediction units 304 as “second prediction units” with high prediction accuracy.
  • the groups are divided into sets, and the individual prediction units 303 and 304 of both sets have different prediction models.
  • the object allocation unit 302 allocates individual object information to appropriate individual prediction units 303 and 304 based on accuracy information in which necessary prediction accuracy is considered.
  • the individual prediction results predicted by the individual prediction units 303 and 304 are made into a format in which the prediction calculation cycle and the coordinate system have the same format and are aggregated as object prediction information.
  • the individual prediction units 303 and 304 since the calculation cycle may be different for each prediction model, it is necessary to perform an interpolation calculation so that the calculation results predicted in the short cycle and the long cycle are the same prediction cycle.
  • FIG. 4 is a block diagram of the predicted object accuracy determination unit 301 installed in the vehicle travel control device 1 of the present embodiment.
  • the simple prediction unit 401 as the “first prediction unit” simply predicts the position R (X (T), Y (T)) of each object at the future time T based on the object information. Specifically, the simple prediction unit 401 sets the current position of the object OB_N to Rn0 (Xn (0), Yn (0)) and the current speed to Vn. In the case of (Vxn, Vyn), the prediction calculation is performed based on the following linear prediction equation (1).
  • Rn (Xn (T), Yn (T)) Vn (Vxn, Vyn) ⁇ T + Rn0 (Xn (0), Yn (0)) (1)
  • the calculation method assumes constant velocity linear motion in which each object moves while maintaining the current speed even in the future time. Thereby, many objects can be predicted in a short time.
  • the own vehicle influence determination unit 402 determines the influence on the own vehicle 100. Specifically, the own vehicle influence determination unit 402 calculates whether or not the probability that the own vehicle 100 and the predicted moving object collide at a future time is equal to or greater than a predetermined value. 100 is determined to have an influence, and if it is equal to or less than a predetermined value, it is determined that there is no influence on the host vehicle 100.
  • a method for calculating the collision probability an estimated collision time TTC (Time To Collation) between each object and the vehicle 100 at a future position is used.
  • TTC [s] relative distance [m] ⁇ relative speed [m / s].
  • the collision probability is 50% when it is less than or equal to the first predetermined value, and the collision probability is 70% when it is less than or equal to the second predetermined value ( ⁇ first predetermined value).
  • the accuracy information calculation unit 403 calculates accuracy information based on the own vehicle influence determination result.
  • FIG. 5 shows the relationship between the vehicle effect determination result and accuracy information.
  • the accuracy information calculation unit 403 indicates “low” as accuracy information when the vehicle influence determination result is “small”, and as accuracy information when the vehicle influence determination result is “medium” If the result of the vehicle influence determination is “large”, the accuracy information is classified into one of three levels of “high”.
  • the reason why the accuracy information is set to three levels is to make it equal to the classification level of the prediction model used in the individual prediction units 303 and 304.
  • FIG. 6 shows an example of the own vehicle 601 and the surrounding pedestrians 603 and 604 according to the present embodiment.
  • the own vehicle 601 travels automatically along a travel planned track 605 on a road standardized by road ends 602 on both sides.
  • Two pedestrians 603 and 604 as “moving objects” are walking around the host vehicle 601.
  • the pedestrian 603 is walking in the direction opposite to the host vehicle 601 on the right rear side of the host vehicle 601.
  • the pedestrian 604 is walking in a direction that intersects the travel plan trajectory of the host vehicle 601.
  • FIG. 7 shows a control flow in the situation of FIG. 6 of the present embodiment.
  • step S700 the control unit (CPU) of the vehicle travel control device 1 inputs sensor information about the environment around the host vehicle 601 obtained based on information from various sensors 2, 3, 4, and 5 that recognize the outside world.
  • the information is input to the processing unit 201 and converted into object information of a moving object existing around the host vehicle 601. Thereby, the information of the position and speed vector of both pedestrians 603 and 604 can be obtained.
  • step S701 the control unit (CPU) of the vehicle travel control device 1 performs simple prediction for each moving object individually based on the object information of the moving object obtained in step S700.
  • the positions of both pedestrians 603 and 604 up to 5 seconds later are predicted at a prediction interval of 1 second.
  • the positions 1 to 5 seconds after both pedestrians 603 and 604 are represented by reference numerals 608 and 609.
  • step S702 the control unit (CPU) of the vehicle travel control device 1 determines the influence on the host vehicle 601.
  • a pedestrian 603 is walking in a direction opposite to the travel planned track 605 of the host vehicle 601.
  • FIG. 8 shows the relationship of the degree of influence on the pedestrians 603 and 604 and the own vehicle 601.
  • the prediction information is determined to be “low” in the prediction time of 5 seconds.
  • the pedestrian 604 may cross the travel plan trajectory 605 of the host vehicle 601 between 2 seconds and 3 seconds in the future, and the collision probability (the degree of influence on the host vehicle 601) is high.
  • simple prediction requires a more detailed prediction because the prediction interval becomes longer. Therefore, “high” is selected as the prediction information.
  • step S708 if the prediction information is determined to be “high”, the process proceeds to step S703, and if it is determined to be “low”, the process proceeds to step S704.
  • step S703 the control unit (CPU) of the vehicle travel control device 1 performs the main prediction of the action using the prediction accuracy (high) model.
  • a model that can be predicted with higher accuracy than the prediction model used in step S701 is used.
  • a model using a parameter calculated by statistically processing motion as a correction term is used.
  • FIG. 9 shows an example of a prediction result when the pedestrian 902 moves to avoid the parked vehicle 901 in the potential field where the parked vehicle 901 and the road edge 903 are generated.
  • the potential method for example, assuming that a virtual dynamic potential is generated from an obstacle at the road end 903 and the parked vehicle 901, the pedestrian 902 moves to pass through the lowest potential and walks. This is a method for predicting the position of the person 902.
  • a dynamic potential to be generated for example, a Gaussian function is used or approximated by a power function.
  • the potential generated from the parked vehicle 901 is expressed by the following equation.
  • U (x, y) W ⁇ exp ( ⁇ (x ⁇ x0) / ⁇ x ⁇ 2 ⁇ (y ⁇ y0) / ⁇ y ⁇ 2) (2)
  • W is the potential weight
  • (x0, y0) is the position of the parked vehicle 901
  • ( ⁇ x, ⁇ y) is the variance in the x and y directions.
  • step S704 the control unit (CPU) of the vehicle travel control device 1 simply predicts the behavior using a prediction accuracy (low) model.
  • a model having the same accuracy as the constant velocity linear motion model used in step 701 is used.
  • the result calculated in step 701 may be stored in the ROM, and the result of the calculation may be reused.
  • the calculation may be omitted (for example, a fixed value) in order to reduce memory and communication traffic.
  • step S705 the control unit (CPU) of the vehicle travel control apparatus 1 collects the prediction results calculated in step S703 and step S704.
  • the result of prediction with the prediction accuracy (high) model has a shorter calculation cycle interval than the result of prediction with the prediction accuracy (low) model. To do.
  • the moving object behavior prediction unit 202 simply predicts the future position of the pedestrians 603 and 604 and the future position of the pedestrian 604 with higher accuracy than the simple prediction unit 401. Since the individual prediction unit 304 that performs the main prediction and the object allocation unit 302 that allocates the pedestrian 604 that performs the main prediction of the future position by the individual prediction unit 304 according to the result of the simple prediction, the prediction accuracy for the pedestrians 603 and 604 is provided.
  • the calculation load can be reduced while maintaining the above. Accordingly, the frequency of sudden acceleration / deceleration of the host vehicle 601 can be reduced, and the CPU mounted on the ECU can be replaced with a low-speed CPU, so that the cost of the ECU can be reduced.
  • the object allocation unit 302 allocates the pedestrian 604 that is highly likely to interfere with the host vehicle 601 to the individual prediction unit 304, the object allocation unit 302 predicts the pedestrian 604 that has a high influence on the host vehicle 601 with high accuracy.
  • the pedestrian 603 having a small influence can reduce the load on the prediction calculation, and can reduce the interference between the host vehicle 601 and both pedestrians 603 and 604 and the calculation load.
  • the prediction accuracy of the main prediction can be increased.
  • the prediction model differs between the preliminary prediction by the simple prediction unit 401 and the main prediction by the individual prediction unit 304, the degree of freedom to select a prediction model for each pedestrian 603, 604 is improved.
  • the simple prediction unit 401 simply predicts the future position of the moving object by the linear prediction method, the calculation load of the moving object behavior prediction unit 202 can be reduced.
  • the individual prediction unit 304 predicts the future position of the pedestrian 604 by the potential method using the potential map, the pedestrian 604 can be predicted with high accuracy.
  • the simple prediction unit 401 and the individual prediction unit 304 individually predict the future positions of the plurality of pedestrians 603 and 604, the prediction accuracy for each pedestrian 603 and 604 can be increased.
  • the calculation load on the pedestrian 603 can be reduced because the result of the simple prediction is used for the pedestrian 603 determined not to perform the main prediction according to the result of the simple prediction.
  • FIG. 10 is a block diagram of the predicted object accuracy determination unit 301 implemented in the vehicle travel control device 1 of the second embodiment.
  • an interaction determination unit 1004 is added to the predicted object accuracy determination unit 301 according to the first embodiment.
  • the interaction determination unit 1004 performs an operation for determining whether or not each moving object causes an interaction based on the simple prediction result of each predicted moving object. Specifically, when it is determined that there is a case where trajectories of the respective prediction results intersect with each other in the simple prediction calculation result, or TTC may become a predetermined value or less in the future time, there is an interaction between these objects. Suppose there is.
  • 11 and 12 show an example in which two pedestrians are walking along the x-axis direction and the y-axis direction. Hereinafter, the interaction between pedestrians will be described.
  • FIG. 11 the predicted positions from 1 second to 5 seconds after each pedestrian are represented by white circles.
  • the predicted trajectory of each pedestrian crosses in 3-4 seconds.
  • FIG. 12 shows the interaction probability at a future time. The calculation of the interaction probability standardizes, for example, the reciprocal of TTC. When the interaction probability is equal to or higher than a predetermined value, it is determined that the two pedestrians interact.
  • the accuracy information calculation unit 1005 calculates accuracy information based on the own vehicle influence determination result and the interaction determination result.
  • FIG. 13 shows the relationship between the vehicle effect determination result and the interaction determination result.
  • the accuracy information is “low”, “medium”, “ Toggle high.
  • the accuracy information is set to “low” when the own vehicle influence determination result is “small”.
  • the accuracy information is “interaction”.
  • the accuracy information is “interaction”
  • a prediction model capable of considering the interaction of moving objects is used.
  • An example in which the interaction can be considered is a prediction method using the potential method described above.
  • FIG. 14 shows an example of the own vehicle 601 of the present embodiment and the surrounding pedestrians 603 and 604 and the parked vehicle 1411.
  • the host vehicle 601 travels automatically along a travel planned track 605 on a road defined by road ends 602 on both sides.
  • Two pedestrians 603 and 604 are walking around the host vehicle 601.
  • the pedestrian 603 is walking in the direction opposite to the host vehicle 601 on the right rear side of the host vehicle 601.
  • a pedestrian 604 is walking in a direction parallel to the host vehicle 601.
  • a parked vehicle 1411 is parked in front of the host vehicle 601.
  • FIG. 15 shows a control flow in the situation of FIG. 14 of the present embodiment.
  • step S700, step S701, and step S702 the control unit (CPU) of the vehicle travel control device 1 processes the pedestrian 603, the pedestrian 604, and the parked vehicle 1411 in the same manner as in the first embodiment.
  • the vehicle influence determination is “low”.
  • the pedestrian 604 is walking in the direction parallel to the host vehicle 601 at the current time, and since the distance from the host vehicle 601 is short, the collision probability is not small. Therefore, the influence degree is determined to be “medium” here. Since the parked vehicle 1411 is stopped, the degree of influence on the own vehicle is “low”.
  • step S1506 the control unit (CPU) of the vehicle travel control device 1 determines the interaction.
  • the interaction between the moving objects is calculated.
  • the parked vehicle 1411 exists in the walking direction of the pedestrian 604. Therefore, it is determined that there is an interaction between the pedestrian 604 and the parked vehicle 1411.
  • the interaction with the pedestrian 604 and the parked vehicle 1411 is determined as “none”.
  • step S1509 the control unit (CPU) of the vehicle travel control apparatus 1 calculates prediction information based on the vehicle influence determination and the interaction determination result, and allocates the prediction information to each prediction unit.
  • the moving object behavior prediction unit 202 according to the second embodiment includes an individual prediction unit 304 that can calculate an interaction in the moving object behavior prediction unit 202.
  • step S704 behavior prediction using the prediction accuracy (low) model of step S704 is performed for the pedestrian 603, and the prediction accuracy (interaction) model of step S1507 is performed for the pedestrian 604 and the parked vehicle 1411. Predict behavior using.
  • step S705 the control unit (CPU) of the vehicle travel control device 1 collects the prediction results calculated in steps S1507 and S704.
  • the pedestrian 604 can predict a track 1414 that avoids the parked vehicle 1411.
  • the pedestrian 604 whose influence on the own vehicle 601 is in the middle is predicted to intersect the travel planned track 605 of the own vehicle 601.
  • the own vehicle 601 can change the travel plan trajectory 605 and calculate a new travel plan trajectory 1411 (indicated by a broken line in the figure) that avoids a collision with the pedestrian 604.
  • the interaction between the pedestrian 604 other than the host vehicle 701 and the parked vehicle 1411 is predicted, even if the pedestrian 604 and the parked vehicle 1411 interfere with each other, the pedestrian 604. And the prediction accuracy with the parked vehicle 1411 can be increased.
  • the future position of each of the plurality of pedestrians 603 and 604 may be predicted repeatedly and sequentially. Thereby, the future position of the pedestrians 603 and 604 can be predicted with high accuracy.
  • FIG. 16 shows a control flow of the third embodiment.
  • the control step (804) of the control by the moving object behavior prediction unit 202 according to the first embodiment is performed by the control unit (CPU) of the vehicle travel control device 1 with a prediction accuracy (medium) model.
  • Step S1604 is used to predict the action.
  • the moving object behavior prediction unit 202 of this embodiment includes an individual prediction unit 304 as a “third prediction unit”. The individual prediction unit 304 predicts the future positions of the pedestrians 603 and 604 with accuracy intermediate between the simple prediction and the main prediction.
  • the individual predicting unit 304 that predicts the future position of the pedestrian 604 with intermediate accuracy between the simple predicting unit 401 and the individual predicting unit 304 is provided, and the object allocating unit 302 depends on the result of the simple prediction.
  • Pedestrians 603 and 604 that predict future positions are allocated to the first prediction unit and the third prediction unit.
  • FIG. 17 is a block diagram of the moving object behavior prediction unit 202 according to the fourth embodiment.
  • Each individual prediction unit 303, 304 of the moving object behavior prediction unit 202 of this embodiment has a model accuracy switching unit 1704.
  • the model accuracy switching unit 1704 selects “interaction model”. A plurality of pieces of individual object information can be input to this interaction model.
  • the plurality of individual prediction units 303 and 304 need to be prepared in advance for each prediction model in the memory, there may be a case where an object larger than the prepared prediction model cannot be calculated with the prediction accuracy.
  • N individual prediction units 304 having a high-precision prediction model are prepared in the memory, it is impossible to predict N + 1 or more pieces of object information with high accuracy.
  • a model accuracy switching unit 1704 that switches prediction models used for prediction in the individual prediction units 303 and 304 based on accuracy information is provided. In this configuration, it is not necessary to secure a memory area unnecessarily, so that the memory can be used effectively.
  • the model accuracy switching unit 1704 described as “low” indicates that a low-precision prediction model is used.

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Traffic Control Systems (AREA)
  • Control Of Driving Devices And Active Controlling Of Vehicle (AREA)

Abstract

The present invention make it possible to reduce computation loads while ensuring the prediction precision of the future positions of moving objects. The present invention includes a moving object behavior prediction unit 202 that predicts the future positions of pedestrians 603, 604 in the vicinity of a host vehicle 601. The moving object behavior prediction unit 202 comprises: a simple prediction unit 401 that simply measures the future positions of the pedestrians 603, 604; an individual prediction unit 304 that predicts future positions of the pedestrian 604 with greater precision than the simple prediction unit 401; and an object allocation unit 302 that allocates the pedestrian 604 whose future positions are to be predicted by the individual prediction unit 304 in response to the results of the simple prediction.

Description

移動物体予測装置Moving object prediction device
　本発明は、自動車等の車両、特に自動運転走行中の自車両の周囲に存在する移動物体の行動を予測する移動物体予測装置に関する。 The present invention relates to a moving object prediction apparatus for predicting the behavior of a moving object existing around a vehicle such as an automobile, in particular, a self-driving vehicle.
　従来、自車両の進路周辺の物体情報を取得し、取得した物体情報から各移動物体の進路を予測するシステムが開発されている。（例えば、特許文献１を参照） Conventionally, a system has been developed that acquires object information around the course of the host vehicle and predicts the path of each moving object from the acquired object information. (For example, see Patent Document 1)
特開２０１５－２６１７９号公報Japanese Patent Laying-Open No. 2015-26179
　しかしながら、多数の移動物体が存在する場合には、それぞれの移動物体の予測演算を同一の方法で実施すると、自車両の走行計画に影響を及ぼさない移動物体に対する計算の負荷が高まり、走行計画に必要な予測演算が十分に行なえない可能性が考えられる。 However, when there are a large number of moving objects, if the prediction calculation of each moving object is performed by the same method, the calculation load on the moving objects that do not affect the travel plan of the host vehicle increases, and the travel plan is increased. There is a possibility that the necessary prediction calculation cannot be performed sufficiently.
　そこで、本発明の目的は、自車両に影響する移動物体の予測精度を確保しつつ、計算負荷の軽減を可能にすることにある。 Therefore, an object of the present invention is to enable reduction of calculation load while ensuring the prediction accuracy of a moving object that affects the host vehicle.
　本発明は、自車両の周囲の移動物体の将来位置を予測する移動物体予測装置であって、前記移動物体の将来位置を簡易予測する第一の予測部と、前記第一の予測部よりも高精度に前記移動物体の将来位置を本予測する第二の予測部と、前記簡易予測の結果に応じて前記第二の予測部によって前記将来位置を本予測する移動物体を割振る割振り部とを備える移動物体予測装置。 The present invention is a moving object prediction apparatus for predicting a future position of a moving object around a host vehicle, and includes a first prediction unit that simply predicts a future position of the moving object, and a first prediction unit A second prediction unit that predicts the future position of the moving object with high accuracy, and an allocation unit that allocates the moving object that predicts the future position by the second prediction unit according to the result of the simple prediction; A moving object prediction apparatus comprising:
　本発明によれば、移動物体の将来位置の予測精度を保ちつつ、計算負荷を軽減することができる。 According to the present invention, it is possible to reduce the calculation load while maintaining the prediction accuracy of the future position of the moving object.
実施例１に係る自車両の走行駆動系およびセンサの説明図を示す。1 is an explanatory diagram of a traveling drive system and sensors of a host vehicle according to a first embodiment. 実施例１に係る車両走行制御装置のブロック図を示す。1 is a block diagram of a vehicle travel control apparatus according to a first embodiment. 実施例１に係る移動物体行動予測部のブロック図を示す。The block diagram of the moving object action prediction part which concerns on Example 1 is shown. 実施例１に係る予測物体精度判断部のブロック図を示す。The block diagram of the prediction object precision judgment part which concerns on Example 1 is shown. 実施例１に係る自車影響判定結果および精度情報の関係を示す。The relationship of the own vehicle influence determination result and precision information which concern on Example 1 is shown. 実施例１に係る自車両およびその周囲の移動物体の一例を示す。An example of the own vehicle which concerns on Example 1, and the surrounding moving object is shown. 実施例１に係る制御フローを示す。3 shows a control flow according to the first embodiment. 実施例１に係る歩行者及び自車両への影響度の関係を示す。The relationship of the influence degree to the pedestrian and the own vehicle which concerns on Example 1 is shown. 実施例１に係る自車両およびその周囲物体の一例を示す。An example of the own vehicle which concerns on Example 1, and its surrounding object is shown. 実施例２に係る予測物体精度判定部のブロック図を示す。The block diagram of the prediction object precision determination part which concerns on Example 2 is shown. 実施例２に係る相互作用がある移動物体の一例を示す。An example of the moving object with interaction which concerns on Example 2 is shown. 実施例２に係る将来時刻および相互作用確率との関係を示す。The relationship with the future time and interaction probability which concern on Example 2 is shown. 実施例２に係る自車影響判定結果および相互作用判定結果の関係を示す。The relationship of the own vehicle influence determination result which concerns on Example 2, and an interaction determination result is shown. 実施例２に係る自車両およびその周囲の移動物体の一例を示す。An example of the own vehicle which concerns on Example 2, and the surrounding moving object is shown. 実施例２に係る制御フローを示す。The control flow which concerns on Example 2 is shown. 実施例３に係る制御フローを示す。10 shows a control flow according to a third embodiment. 実施例４に係る移動物体行動予測部のブロック図を示す。The block diagram of the moving object action prediction part which concerns on Example 4 is shown.
　幾つかの実施例を図面を参照して詳細に説明する。 Some embodiments will be described in detail with reference to the drawings.
　図１は、実施例１に係る車両走行制御装置１を搭載した自車両１００の全体構成を示した説明図である。ＦＬ輪、ＦＲ輪、ＲＬ輪およびＲＲ輪は、左前輪、右前輪、左後輪および右後輪をそれぞれ意味する。 FIG. 1 is an explanatory diagram showing the overall configuration of the host vehicle 100 equipped with the vehicle travel control device 1 according to the first embodiment. The FL wheel, the FR wheel, the RL wheel, and the RR wheel mean a left front wheel, a right front wheel, a left rear wheel, and a right rear wheel, respectively.
　自車両１００は、外界を認識するセンサとして、前方にカメラ２と、左右側方にレーザーレーダ３，４と、後方にミリ波レーダ５とを備えており、自車両１００と周囲車両との相対距離および相対速度を検出することができる。尚、本実施例では、センサの構成の一例として上記センサの組み合わせを示しているが、それに限定するものではなく、超音波センサ、ステレオカメラ、赤外線カメラなどとの組み合わせでもよい。上記センサ信号が、車両走行制御装置１に入力される。 The own vehicle 100 includes a camera 2 at the front, laser radars 3 and 4 at the left and right sides, and a millimeter wave radar 5 at the rear as sensors for recognizing the outside world. Distance and relative speed can be detected. In the present embodiment, a combination of the above-described sensors is shown as an example of the configuration of the sensor. However, the present invention is not limited to this, and a combination with an ultrasonic sensor, a stereo camera, an infrared camera, or the like may be used. The sensor signal is input to the vehicle travel control device 1.
　さらに、自車両１００は、車両走行制御装置１と、操舵制御装置８と、制動制御装置１５と、加速制御装置１９と、通信装置２３と、表示装置２４とを備えている。 Furthermore, the host vehicle 100 includes a vehicle travel control device 1, a steering control device 8, a braking control device 15, an acceleration control device 19, a communication device 23, and a display device 24.
　車両走行制御装置１は、例えば、ＣＰＵと、ＲＯＭと、ＲＡＭと、入出力装置とを有する。ＲＯＭには、車両走行制御のフローが記憶されている。車両走行制御装置１は、各種センサ２，３，４，５の情報に基づいて、進行方向を制御するためのステアリング制御機構１０、ブレーキ制御機構１３、スロットル制御機構２０への指令値を演算する。 The vehicle travel control device 1 includes, for example, a CPU, a ROM, a RAM, and an input / output device. The ROM stores a flow of vehicle travel control. The vehicle travel control device 1 calculates command values to the steering control mechanism 10, the brake control mechanism 13, and the throttle control mechanism 20 for controlling the traveling direction based on information from the various sensors 2, 3, 4, and 5. .
　操舵制御装置８は、車両走行制御装置１からの指令値に基づいて、上記ステアリング制御機構１０を制御する。制動制御装置１５は、車両走行制御装置１からの指令値に基づいて、上記ブレーキ制御機構１３を制御し、各輪のブレーキ力配分を調整する。加速制御装置１９は、車両走行制御装置１からの指令値に基づいて、スロットル制御機構２０を制御し、エンジンのトルク出力を調整する。通信装置２３は、路車間または車車間の通信を行う。表示装置２４は、自車両１００の走行計画および周辺に存在する移動物体の行動予測を表示する。 The steering control device 8 controls the steering control mechanism 10 based on the command value from the vehicle travel control device 1. The brake control device 15 controls the brake control mechanism 13 based on the command value from the vehicle travel control device 1 to adjust the brake force distribution of each wheel. The acceleration control device 19 controls the throttle control mechanism 20 based on the command value from the vehicle travel control device 1 to adjust the torque output of the engine. The communication device 23 performs communication between road vehicles or between vehicles. The display device 24 displays a travel plan of the host vehicle 100 and a behavior prediction of a moving object existing in the vicinity.
　このように構成された車両走行制御装置１は、詳細は後述する、生成した走行計画に従って車両走行を制御するための各アクチュエータ１０，１３，２０の指令値を演算する。
各アクチュエータ１０、１３、２０の制御装置８，１５，１９は、車両走行制御装置１の指令値を通信により受信し、当該指令値に基づいて、各アクチュエータ１０，１３，２０を制御する。 The vehicle travel control device 1 configured as described above calculates command values of the actuators 10, 13, and 20 for controlling the vehicle travel according to the generated travel plan, which will be described in detail later.
The control devices 8, 15, and 19 of the actuators 10, 13, and 20 receive the command value of the vehicle travel control device 1 through communication, and control the actuators 10, 13, and 20 based on the command value.
　次いで、ブレーキの動作について説明する。ドライバが自車両１００を運転している状態では、ドライバのブレーキペダル１２を踏む踏力をブレーキブースタ（不図示）で倍力し、その力に応じた油圧をマスタシリンダ（不図示）によって発生させる。発生した油圧は、ブレーキ制御機構１３を介して、ホイルシリンダ１６に供給される。ホイルシリンダ１６ＦＬ～１６ＲＲは、シリンダ（不図示）と、ピストンと、パッド等とを有しており、マスタシリンダ９から供給された作動液によってピストンが推進され、ピストンに連結されたパッドがディスクロータに押圧される。尚、ディスクロータは、車輪と共に回転している。そのため、ディスクロータに作用したブレーキトルクは、車輪と路面との間に作用するブレーキ力となる。以上により、ドライバのブレーキペダル操作に応じて、各輪に制動力が発生させることができる。 Next, the operation of the brake will be described. In a state where the driver is driving the host vehicle 100, the pedaling force of the driver stepping on the brake pedal 12 is boosted by a brake booster (not shown), and a hydraulic pressure corresponding to the force is generated by a master cylinder (not shown). The generated hydraulic pressure is supplied to the wheel cylinder 16 via the brake control mechanism 13. Each of the wheel cylinders 16FL to 16RR includes a cylinder (not shown), a piston, a pad, and the like. The piston is propelled by the hydraulic fluid supplied from the master cylinder 9, and the pad connected to the piston is a disk rotor. Pressed. The disc rotor rotates with the wheels. Therefore, the brake torque acting on the disc rotor becomes a braking force acting between the wheel and the road surface. As described above, a braking force can be generated in each wheel in accordance with the driver's brake pedal operation.
　制動制御装置１５は、車両走行制御装置１と同様に、例えば、ＣＰＵと、ＲＯＭと、ＲＡＭと、入出力装置とを有する。制動制御装置１５には、コンバインセンサ１４と、車輪速センサ１１ＦＬ～１１ＲＲと、上述の制動制御装置１５からのブレーキ力指令と、後述する操舵制御装置８を介してハンドル角検出装置２１からのセンサ信号とが入力されている。コンバインセンサ１４は、前後加速度、横加速度、ヨーレートを検出する、車輪速センサ１１ＦＬ～１１ＲＲは、各輪に設置されている。さらに、制動制御装置１５は、出力側がポンプ（不図示）と、制御バルブを有するブレーキ制御機構１３とに接続されており、ドライバのブレーキペダル操作とは独立して各輪に任意の制動力を発生させることができる。制動制御装置１５は、上記各種情報に基づいて自車両１００のスピン、ドリフトアウト、および車輪のロックを推定し、それらを抑制するように該当輪の制動力を発生させることによって、ドライバの操縦安定性を高める役割を担っている。さらに、車両走行制御装置１は、制動制御装置１５にブレーキ指令を通信することによって、自車両１００に任意のブレーキ力を発生させることができ、ドライバの操作が生じない自動運転の場合、自動的に制動を行う役割を担っている。但し、本実施例では、上記制動制御装置１５に限定するものではなく、例えば、ブレーキバイワイヤ等の他のアクチュエータを用いてもよい。 The braking control device 15 includes, for example, a CPU, a ROM, a RAM, and an input / output device, like the vehicle travel control device 1. The brake controller 15 includes a combine sensor 14, wheel speed sensors 11FL to 11RR, a brake force command from the brake controller 15 described above, and a sensor from the steering wheel angle detector 21 via a steering controller 8 described later. Signal. The combine sensor 14 detects longitudinal acceleration, lateral acceleration, and yaw rate. Wheel speed sensors 11FL to 11RR are installed on each wheel. Further, the braking control device 15 is connected to a pump (not shown) on the output side and a brake control mechanism 13 having a control valve, and applies an arbitrary braking force to each wheel independently of the driver's brake pedal operation. Can be generated. The braking control device 15 estimates the spin, drift-out, and wheel lock of the host vehicle 100 based on the various information described above, and generates the braking force of the corresponding wheel so as to suppress them, thereby stabilizing the steering of the driver. It plays a role to improve sex. Furthermore, the vehicle travel control device 1 can generate an arbitrary braking force on the host vehicle 100 by communicating a brake command to the braking control device 15. It plays the role of braking. However, in the present embodiment, the present invention is not limited to the braking control device 15, and other actuators such as a brake-by-wire may be used.
　次いで、ステアリングの動作について説明する。ドライバが自車両１００を運転している状態では、ドライバがハンドル６を介して入力した操舵トルクとハンドル角とをそれぞれ操舵トルク検出装置７とハンドル角検出装置２１とで検出する。操舵制御装置８は、それらの情報に基づいてモータ９を制御し、アシストトルクを発生させる。尚、操舵制御装置８も、車両走行制御装置１と同様に、例えば、ＣＰＵと、ＲＯＭと、ＲＡＭと、入出力装置とを有する。ステアリング制御機構１０が、ドライバの操舵トルクと、モータ９によるアシストトルクとの合力により可動すると、前輪が切られ、前輪の切れ角に応じて路面からの反力がステアリング制御機構に伝わると、その反力が路面反力としてドライバに伝わる。 Next, the steering operation will be described. In a state where the driver is driving the host vehicle 100, the steering torque and the steering wheel angle detected by the driver via the steering wheel 6 are detected by the steering torque detection device 7 and the steering wheel angle detection device 21, respectively. The steering control device 8 controls the motor 9 based on the information and generates assist torque. The steering control device 8 also has, for example, a CPU, a ROM, a RAM, and an input / output device, like the vehicle travel control device 1. When the steering control mechanism 10 is moved by the resultant force of the driver's steering torque and the assist torque by the motor 9, the front wheel is cut, and when the reaction force from the road surface is transmitted to the steering control mechanism according to the turning angle of the front wheel, The reaction force is transmitted to the driver as a road surface reaction force.
　操舵制御装置８は、ドライバのステアリング操作とは独立してモータ９によってトルクを発生し、ステアリング制御機構１０を制御することができる。従って、車両走行制御装置１は、操舵制御装置８に操舵力指令を通信することによって、前輪を任意の切れ角に制御することができ、ドライバの操作が生じない自動運転においては自動的に操舵を行う役割を担っている。但し、本実施例では、上記操舵制御装置８に限定するものではなく、ステアバイワイヤ等の他のアクチュエータを用いてもよい。 The steering control device 8 can control the steering control mechanism 10 by generating torque by the motor 9 independently of the driver's steering operation. Therefore, the vehicle travel control device 1 can control the front wheels to an arbitrary turning angle by communicating a steering force command to the steering control device 8, and automatically steers in automatic driving where no driver operation occurs. I have a role to do. However, in this embodiment, the present invention is not limited to the steering control device 8, and other actuators such as steer-by-wire may be used.
　次いで、アクセルについて説明する。ドライバの操作によるアクセルペダル１７の踏み込み量は、ストロークセンサ１８によって検出され、加速制御装置１９に入力される。尚、加速制御装置１９も、車両走行制御装置１と同様に、例えば、ＣＰＵと、ＲＯＭと、ＲＡＭと、入出力装置とを有する。加速制御装置１９は、上記アクセルペダル１７の踏み込み量に応じてスロットル開度を調節し、エンジンを制御する。以上により、ドライバのアクセルペダル１７の操作に応じて、自車両１００を加速させることができる。さらに、加速制御装置１９は、ドライバのアクセル操作とは独立してスロットル開度を制御することができる。従って、車両走行制御装置１は、加速制御装置１９に加速指令を通信することによって、自車両１００に任意の加速度を発生させることができ、ドライバの操作が生じない自動運転の場合、自動的に加速を行う役割を担っている。 Next, the accelerator will be described. The amount of depression of the accelerator pedal 17 by the driver's operation is detected by the stroke sensor 18 and input to the acceleration control device 19. The acceleration control device 19 also has, for example, a CPU, a ROM, a RAM, and an input / output device, like the vehicle travel control device 1. The acceleration control device 19 controls the engine by adjusting the throttle opening in accordance with the depression amount of the accelerator pedal 17. As described above, the host vehicle 100 can be accelerated in accordance with the driver's operation of the accelerator pedal 17. Further, the acceleration control device 19 can control the throttle opening independently of the driver's accelerator operation. Therefore, the vehicle travel control device 1 can generate an arbitrary acceleration in the host vehicle 100 by communicating an acceleration command to the acceleration control device 19, and automatically in the case of an automatic driving in which no driver operation occurs. Has a role to accelerate.
　図２は、本実施例に係る車両走行制御装置１のブロック図である。 FIG. 2 is a block diagram of the vehicle travel control apparatus 1 according to the present embodiment.
　外界を認識する各種センサ２，３，４，５の情報に基づいて得られた周囲の環境の情報がセンサ入力処理部２０１に入力されると、自車両１００の周囲に存在する移動物体の物体情報に変換される。具体的な物体情報としては、例えば、歩行者、自転車、車両の属性情報や、それらの現在位置および現在の速度ベクトルが抽出される。ここで、移動物体には、現時刻で得られた速度がゼロであったとしても、将来において動く可能性がある駐車車両が含まれる。 When information on the surrounding environment obtained based on information from various sensors 2, 3, 4, and 5 that recognize the outside world is input to the sensor input processing unit 201, a moving object that exists around the host vehicle 100 Converted to information. As specific object information, for example, attribute information of pedestrians, bicycles, and vehicles, and their current position and current speed vector are extracted. Here, the moving object includes a parked vehicle that may move in the future even if the speed obtained at the current time is zero.
　次いで、前述した物体情報と、後述する方法によって前演算ステップで演算された軌道および速度計画情報とが、「移動物体予測装置」としての移動物体行動予測部２０２に入力される。移動物体行動予測部２０２では、詳細は後述する、入力情報に基づいて、各移動物体の将来の位置および速度情報（物体予測情報）を演算する。 Next, the object information described above and the trajectory and speed plan information calculated in the pre-calculation step by a method to be described later are input to the moving object action prediction unit 202 as the “moving object prediction device”. The moving object behavior prediction unit 202 calculates future position and velocity information (object prediction information) of each moving object based on input information, which will be described in detail later.
　次いで、軌道／運動計画部２０３について説明する。軌道／運動計画部２０３は、入力情報である物体予測情報および地図情報に基づいて、自車両１００の走行軌道およびその際の速度を計画する。具体的には、軌道／運動計画部２０３は、例えば、自車両１００の前方の走行車両および駐車車両の将来位置の物体予測情報から、それらの移動物体との衝突を避けるような軌道を生成する。それと同時に、軌道／運動計画部２０３は、自車両１００に同乗しているドライバが出来る限り不快と感じない範囲の加減速で自動運転走行をするように計画する。 Next, the trajectory / motion planning unit 203 will be described. The trajectory / motion planning unit 203 plans the travel trajectory of the host vehicle 100 and the speed at that time based on the object prediction information and map information that are input information. Specifically, for example, the trajectory / motion planning unit 203 generates a trajectory that avoids collision with these moving objects from the object prediction information of the future position of the traveling vehicle and the parked vehicle in front of the host vehicle 100. . At the same time, the track / motion planning unit 203 plans to drive automatically with acceleration / deceleration within a range in which the driver riding in the vehicle 100 does not feel as uncomfortable as possible.
　次いで、アクチュエータ指令値演算部２０４について説明する。アクチュエータ指令値演算部２０４は、入力情報である軌道・速度計画情報に基づいて、例えば、ブレーキ、ステアリング、アクセルの操作量を演算する。具体的には、軌道・速度計画情報は、自車両１００の将来における目標情報であるため、アクチュエータ指令値演算部２０４は、目標位置および速度が入力されると、自車両１００の物理モデルに基づいて、各アクチュエータ１０，１３，２０の制御量を出力する。 Next, the actuator command value calculation unit 204 will be described. The actuator command value calculation unit 204 calculates, for example, brake, steering, and accelerator operation amounts based on the trajectory / speed plan information that is input information. Specifically, since the track / speed plan information is target information in the future of the host vehicle 100, the actuator command value calculation unit 204 is based on the physical model of the host vehicle 100 when the target position and speed are input. Thus, the control amount of each actuator 10, 13, 20 is output.
　図３は、本実施例の車両走行制御装置１に実装されている移動物体行動予測部２０２のブロック図である。 FIG. 3 is a block diagram of the moving object behavior prediction unit 202 implemented in the vehicle travel control device 1 of the present embodiment.
　予測物体精度判断部３０１は、詳細は後述する、物体情報および前回演算時の軌道・速度計画情報に基づいて、各物体の精度情報を演算する。「割振り部」としての物体割振り部３０２は、精度情報に基づいて、物体情報から個別物体情報を抽出し、複数の個別予測部３０３，３０４に送信する。複数の個別予測部３０３，３０４は、予測精度が低精度な複数の個別予測部３０３の組と、予測精度が高精度な「第二の予測部」としての個別予測部３０４の組との二組に分けられ、両組の個別予測部３０３，３０４は、相互に異なる予測モデルを有している。物体割振り部３０２は、必要な予測精度が考慮された精度情報に基づいて、適切な個別予測部３０３，３０４に個別物体情報を割り振る。 The predicted object accuracy determination unit 301 calculates accuracy information of each object based on object information and trajectory / speed plan information at the time of the previous calculation, which will be described later in detail. The object allocation unit 302 as “allocation unit” extracts the individual object information from the object information based on the accuracy information, and transmits it to the plurality of individual prediction units 303 and 304. The plurality of individual prediction units 303 and 304 includes two sets of a plurality of individual prediction units 303 with low prediction accuracy and a set of individual prediction units 304 as “second prediction units” with high prediction accuracy. The groups are divided into sets, and the individual prediction units 303 and 304 of both sets have different prediction models. The object allocation unit 302 allocates individual object information to appropriate individual prediction units 303 and 304 based on accuracy information in which necessary prediction accuracy is considered.
　次いで、予測結果集約部３０５について説明する。個別予測部３０３，３０４によって予測された個別予測結果を、予測演算周期および座標系を同一の形式とし、物体予測情報として集約した形式とする。特に、個別予測部３０３，３０４では、予測モデル毎に演算周期が異なる場合があるため、短周期および長周期で予測した演算結果を同一の予測周期とするような補間演算が必要となる。 Next, the prediction result aggregation unit 305 will be described. The individual prediction results predicted by the individual prediction units 303 and 304 are made into a format in which the prediction calculation cycle and the coordinate system have the same format and are aggregated as object prediction information. In particular, in the individual prediction units 303 and 304, since the calculation cycle may be different for each prediction model, it is necessary to perform an interpolation calculation so that the calculation results predicted in the short cycle and the long cycle are the same prediction cycle.
　図４は、本実施例の車両走行制御装置１に実装されている予測物体精度判定部３０１のブロック図である。 FIG. 4 is a block diagram of the predicted object accuracy determination unit 301 installed in the vehicle travel control device 1 of the present embodiment.
　「第一の予測部」としての簡易予測部４０１は、物体情報に基づいて、各物体の将来時間Ｔにおける位置Ｒ（Ｘ（Ｔ），Ｙ（Ｔ））を簡易予測する。具体的には、簡易予測部４０１は、物体ＯＢ＿Ｎの現在位置をＲｎ０（Ｘｎ（０），Ｙｎ（０））、現在速度をＶｎ
（Ｖｘｎ，Ｖｙｎ）とした場合、以下の線形予測式（１）に基づいて、予測演算する。
 
  Ｒｎ（Ｘｎ（Ｔ），Ｙｎ（Ｔ））＝Ｖｎ（Ｖｘｎ，Ｖｙｎ）×Ｔ
                                ＋Ｒｎ０（Ｘｎ（０），Ｙｎ（０））・・・（１）
 
 　演算方法は、各物体が将来時間においても現在速度を維持して移動する等速直線運動を仮定している。これにより、短時間に多くの物体の予測が可能となる。 The simple prediction unit 401 as the “first prediction unit” simply predicts the position R (X (T), Y (T)) of each object at the future time T based on the object information. Specifically, the simple prediction unit 401 sets the current position of the object OB_N to Rn0 (Xn (0), Yn (0)) and the current speed to Vn.
In the case of (Vxn, Vyn), the prediction calculation is performed based on the following linear prediction equation (1).

Rn (Xn (T), Yn (T)) = Vn (Vxn, Vyn) × T
+ Rn0 (Xn (0), Yn (0)) (1)

The calculation method assumes constant velocity linear motion in which each object moves while maintaining the current speed even in the future time. Thereby, many objects can be predicted in a short time.
　次いで、自車影響判定部４０２は、簡易予測演算結果と軌道・速度計画情報が入力されると、自車両１００への影響を判定する。具体的には、自車影響判定部４０２は、将来時刻において、自車両１００と予測移動物体とが衝突する確率が所定値以上あるか否かを演算し、所定値以上の場合は、自車両１００への影響ありと判定し、所定値以下の場合は、自車両１００への影響なしと判定する。衝突する確率の演算する方法としては、将来位置における各物体と自車両１００との衝突予想時間ＴＴＣ（Ｔｉｍｅ　Ｔｏ　Ｃｏｌｌｉｓｉｏｎ）を用いる。ここで、ＴＴＣ［ｓ］＝相対距離［ｍ］÷相対速度［ｍ／ｓ］である。ＴＴＣを用いた判定では、例えば、第１所定値以下である場合は、衝突確率５０％、第２所定値（＜第１所定値）以下である場合は、衝突確率７０％とする。 Next, when the simple prediction calculation result and the trajectory / speed plan information are input, the own vehicle influence determination unit 402 determines the influence on the own vehicle 100. Specifically, the own vehicle influence determination unit 402 calculates whether or not the probability that the own vehicle 100 and the predicted moving object collide at a future time is equal to or greater than a predetermined value. 100 is determined to have an influence, and if it is equal to or less than a predetermined value, it is determined that there is no influence on the host vehicle 100. As a method for calculating the collision probability, an estimated collision time TTC (Time To Collation) between each object and the vehicle 100 at a future position is used. Here, TTC [s] = relative distance [m] ÷ relative speed [m / s]. In the determination using the TTC, for example, the collision probability is 50% when it is less than or equal to the first predetermined value, and the collision probability is 70% when it is less than or equal to the second predetermined value (<first predetermined value).
　次いで、精度情報演算部４０３について説明する。精度情報演算部４０３は、自車影響判定結果に基づいて、精度情報を演算する。 Next, the accuracy information calculation unit 403 will be described. The accuracy information calculation unit 403 calculates accuracy information based on the own vehicle influence determination result.
　図５は、自車影響判定結果および精度情報の関係を示す。
  精度情報演算部４０３は、図５に示すように、自車影響判定結果が「小」である場合は精度情報として「低」、自車影響判定結果が「中」である場合は精度情報として「中」、自車影響判定結果が「大」である場合は精度情報として「高」の３水準の何れかに分類する。ここで、精度情報を３水準としたのは、個別予測部３０３，３０４で用いる予測モデルの分類水準と等しくするためである。 FIG. 5 shows the relationship between the vehicle effect determination result and accuracy information.
As shown in FIG. 5, the accuracy information calculation unit 403 indicates “low” as accuracy information when the vehicle influence determination result is “small”, and as accuracy information when the vehicle influence determination result is “medium” If the result of the vehicle influence determination is “large”, the accuracy information is classified into one of three levels of “high”. Here, the reason why the accuracy information is set to three levels is to make it equal to the classification level of the prediction model used in the individual prediction units 303 and 304.
　図６は、本実施例の自車両６０１およびその周囲の歩行者６０３，６０４の一例を示す。 FIG. 6 shows an example of the own vehicle 601 and the surrounding pedestrians 603 and 604 according to the present embodiment.
　図６に示した状況において、車両走行制御装置１の移動物体行動予測部２０２の作用効果を以下に説明する。自車両６０１は、両側にある道路端６０２で規格される道路上を、走行計画軌道６０５に沿って自動運転走行している。自車両６０１の周囲には、二人の「移動物体」としての歩行者６０３，６０４が歩行している。歩行者６０３は、自車両６０１の右後方を自車両６０１とは反対の方向に歩行している。歩行者６０４は、自車両６０１の走行計画軌道と交差する方向に歩行している。 In the situation shown in FIG. 6, the operation and effect of the moving object behavior prediction unit 202 of the vehicle travel control device 1 will be described below. The own vehicle 601 travels automatically along a travel planned track 605 on a road standardized by road ends 602 on both sides. Two pedestrians 603 and 604 as “moving objects” are walking around the host vehicle 601. The pedestrian 603 is walking in the direction opposite to the host vehicle 601 on the right rear side of the host vehicle 601. The pedestrian 604 is walking in a direction that intersects the travel plan trajectory of the host vehicle 601.
　図７は、本実施例の図６の状況における制御フローを示す。 FIG. 7 shows a control flow in the situation of FIG. 6 of the present embodiment.
　ステップＳ７００では、車両走行制御装置１の制御部（ＣＰＵ）は、外界を認識する各種センサ２，３，４，５の情報に基づいて得られた自車両６０１の周囲の環境の情報をセンサ入力処理部２０１に入力し、自車両６０１の周囲に存在する移動物体の物体情報に変換する。これにより、両歩行者６０３，６０４の位置および速度ベクトルの情報を得ることができる。 In step S700, the control unit (CPU) of the vehicle travel control device 1 inputs sensor information about the environment around the host vehicle 601 obtained based on information from various sensors 2, 3, 4, and 5 that recognize the outside world. The information is input to the processing unit 201 and converted into object information of a moving object existing around the host vehicle 601. Thereby, the information of the position and speed vector of both pedestrians 603 and 604 can be obtained.
　ステップＳ７０１では、車両走行制御装置１の制御部（ＣＰＵ）は、ステップＳ７００で得られた移動物体の物体情報に基づき、移動物体毎を個別に簡易予測する。ここでは、両歩行者６０３，６０４の５秒後までの位置を予測間隔１秒で予測する。両歩行者６０３，６０４の１～５秒後の位置を、符号６０８，６０９で表している。 In step S701, the control unit (CPU) of the vehicle travel control device 1 performs simple prediction for each moving object individually based on the object information of the moving object obtained in step S700. Here, the positions of both pedestrians 603 and 604 up to 5 seconds later are predicted at a prediction interval of 1 second. The positions 1 to 5 seconds after both pedestrians 603 and 604 are represented by reference numerals 608 and 609.
　ステップＳ７０２では、車両走行制御装置１の制御部（ＣＰＵ）は、自車両６０１への影響を判定する。歩行者６０３は、自車両６０１の走行計画軌道６０５と反対方向に歩行している。 In step S702, the control unit (CPU) of the vehicle travel control device 1 determines the influence on the host vehicle 601. A pedestrian 603 is walking in a direction opposite to the travel planned track 605 of the host vehicle 601.
　図８は、歩行者６０３，６０４及び自車両６０１への影響度の関係を示す。 FIG. 8 shows the relationship of the degree of influence on the pedestrians 603 and 604 and the own vehicle 601.
　ここで、図８に示すように、歩行者６０３は、衝突確率（自車両６０１への影響度）が時間経過と共に低下していく。そのため、予測時間５秒間において、予測情報は「低」と判定される。歩行者６０４は、将来時刻２秒から３秒の間に、自車両６０１の走行計画軌道６０５と交差する可能性があり、衝突確率（自車両６０１への影響度）が高い。ただし、簡易的な予測では、予測間隔が長くなるため、より詳細な予測を行なう必要がある。そのため、予測情報としては、「高」が選択される。 Here, as shown in FIG. 8, in the pedestrian 603, the collision probability (the degree of influence on the own vehicle 601) decreases with time. Therefore, the prediction information is determined to be “low” in the prediction time of 5 seconds. The pedestrian 604 may cross the travel plan trajectory 605 of the host vehicle 601 between 2 seconds and 3 seconds in the future, and the collision probability (the degree of influence on the host vehicle 601) is high. However, simple prediction requires a more detailed prediction because the prediction interval becomes longer. Therefore, “high” is selected as the prediction information.
　図７に戻る。ステップＳ７０８では、予測情報が「高」と判定された場合、ステップＳ７０３に進み、「低」と判定された場合、ステップＳ７０４に進む。 Return to Figure 7. In step S708, if the prediction information is determined to be “high”, the process proceeds to step S703, and if it is determined to be “low”, the process proceeds to step S704.
　ステップＳ７０３では、車両走行制御装置１の制御部（ＣＰＵ）は、予測精度（高）モデルを用いて行動を本予測する。ここでは、ステップＳ７０１で用いた予測モデルと比較して、高精度に予測できるモデルを用いる。具体的には、例えば、ポテンシャル法を用いたポテンシャルモデル、予測周期が短いモデル、現在の加速度を考慮した等加速度運動モデル、周囲車両およびデータセンタと通信することによって得られた予測対象の事前の動きを統計処理して算出したパラメータを補正項とするモデルを用いる。 In step S703, the control unit (CPU) of the vehicle travel control device 1 performs the main prediction of the action using the prediction accuracy (high) model. Here, a model that can be predicted with higher accuracy than the prediction model used in step S701 is used. Specifically, for example, a potential model using a potential method, a model with a short prediction cycle, an isoacceleration motion model considering current acceleration, a prediction target obtained in advance by communicating with surrounding vehicles and a data center. A model using a parameter calculated by statistically processing motion as a correction term is used.
　図９は、駐車車両９０１と道路端９０３とが発生するポテンシャル場において、歩行者９０２が駐車車両９０１を避けるように移動していく場合の予測結果の例を示す。 FIG. 9 shows an example of a prediction result when the pedestrian 902 moves to avoid the parked vehicle 901 in the potential field where the parked vehicle 901 and the road edge 903 are generated.
　ここで、個別予測部３０４が有するポテンシャル法を用いた予測モデルについて述べる。ポテンシャル法は、例えば、道路端９０３および駐車車両９０１の障害物から仮想的な力学ポテンシャルが発生していると仮定し、歩行者９０２がポテンシャルの最も低い個所を通過するように移動するとして、歩行者９０２の位置を予測する手法である。発生させる力学ポテンシャルとしては、例えば、ガウス関数を用いたり、べき乗関数で近似する。
一例としては、駐車車両９０１から発生させるポテンシャルを以下の式で表す。
 
  Ｕ（ｘ,ｙ）＝Ｗ×ｅｘｐ（－（ｘ－ｘ０）／σｘ＾２－（ｙ－ｙ０）／σｙ＾２）・・・（２）
 
  ここで、Ｗはポテンシャルの重み、（ｘ０，ｙ０）は駐車車両９０１の位置、（σｘ，σｙ）はｘ方向およびｙ方向の分散を表す。複数の物体が存在する場合には、重ね合わせの原理に基づいて演算を行なう。 Here, a prediction model using the potential method possessed by the individual prediction unit 304 will be described. In the potential method, for example, assuming that a virtual dynamic potential is generated from an obstacle at the road end 903 and the parked vehicle 901, the pedestrian 902 moves to pass through the lowest potential and walks. This is a method for predicting the position of the person 902. As a dynamic potential to be generated, for example, a Gaussian function is used or approximated by a power function.
As an example, the potential generated from the parked vehicle 901 is expressed by the following equation.

U (x, y) = W × exp (− (x−x0) / σx ^ 2− (y−y0) / σy ^ 2) (2)

Here, W is the potential weight, (x0, y0) is the position of the parked vehicle 901, and (σx, σy) is the variance in the x and y directions. When there are a plurality of objects, calculation is performed based on the principle of superposition.
　再び、図７に戻る。ステップＳ７０４では、車両走行制御装置１の制御部（ＣＰＵ）は、予測精度（低）モデルを用いて行動を簡易予測する。ここでは、ステップ７０１で用いた等速直線運動モデルと同程度の精度のモデルを用いる。若しくは、ステップ７０１で演算した結果をＲＯＭに格納し、その演算結果を再利用しても良い。または、メモリおよび通信量を削減するために、演算を省略（例えば、固定値とする）しても良い。 Return to Fig. 7 again. In step S704, the control unit (CPU) of the vehicle travel control device 1 simply predicts the behavior using a prediction accuracy (low) model. Here, a model having the same accuracy as the constant velocity linear motion model used in step 701 is used. Alternatively, the result calculated in step 701 may be stored in the ROM, and the result of the calculation may be reused. Alternatively, the calculation may be omitted (for example, a fixed value) in order to reduce memory and communication traffic.
　ステップＳ７０５では、車両走行制御装置１の制御部（ＣＰＵ）は、ステップＳ７０３およびステップＳ７０４で演算した予測結果を集約する。ここでは、予測精度（高）モデルで予測した結果は、予測精度（低）モデルで予測した結果よりも演算周期間隔が短くなっているため、短い予測周期間隔に統一できるように、例えば線形補間を行なう。 In step S705, the control unit (CPU) of the vehicle travel control apparatus 1 collects the prediction results calculated in step S703 and step S704. Here, the result of prediction with the prediction accuracy (high) model has a shorter calculation cycle interval than the result of prediction with the prediction accuracy (low) model. To do.
　以上に述べた処理によって、移動物体行動予測部２０２は、歩行者６０３，６０４の将来位置を簡易予測する簡易予測部４０１と、前記簡易予測部４０１よりも高精度に歩行者６０４の将来位置を本予測する個別予測部３０４と、簡易予測の結果に応じて個別予測部３０４によって将来位置を本予測する歩行者６０４を割振る物体割振り部３０２とを備えるので、歩行者６０３，６０４に対する予測精度を保ちつつ、計算負荷を軽減することができる。従って、自車両６０１の急加減速頻度を低減することができ、さらにＥＣＵに搭載するＣＰＵを低速度ＣＰＵへ置き換えることが可能となり、ＥＣＵを低コスト化することができる。 Through the processing described above, the moving object behavior prediction unit 202 simply predicts the future position of the pedestrians 603 and 604 and the future position of the pedestrian 604 with higher accuracy than the simple prediction unit 401. Since the individual prediction unit 304 that performs the main prediction and the object allocation unit 302 that allocates the pedestrian 604 that performs the main prediction of the future position by the individual prediction unit 304 according to the result of the simple prediction, the prediction accuracy for the pedestrians 603 and 604 is provided. The calculation load can be reduced while maintaining the above. Accordingly, the frequency of sudden acceleration / deceleration of the host vehicle 601 can be reduced, and the CPU mounted on the ECU can be replaced with a low-speed CPU, so that the cost of the ECU can be reduced.
　物体割振り部３０２は、自車両６０１と干渉する可能性が高い歩行者６０４を個別予測部３０４に割振るので、自車両６０１への影響が高い歩行者６０４の予測を高精度に行なうと同時に、影響が小さい歩行者６０３は、予測演算への負荷を低減することができ、自車両６０１と両歩行者６０３，６０４との干渉および計算負荷を低減することができる。 Since the object allocation unit 302 allocates the pedestrian 604 that is highly likely to interfere with the host vehicle 601 to the individual prediction unit 304, the object allocation unit 302 predicts the pedestrian 604 that has a high influence on the host vehicle 601 with high accuracy. The pedestrian 603 having a small influence can reduce the load on the prediction calculation, and can reduce the interference between the host vehicle 601 and both pedestrians 603 and 604 and the calculation load.
　個別予測部３０４による本予測は、簡易予測部４０１による簡易予測よりも演算周期が短いので、本予測の予測精度を高めることができる。 Since the main prediction by the individual prediction unit 304 has a shorter calculation cycle than the simple prediction by the simple prediction unit 401, the prediction accuracy of the main prediction can be increased.
　簡易予測部４０１による予備予測と、個別予測部３０４による本予測とは、予測モデルが異なるので、歩行者６０３，６０４毎に予測モデルを選択する自由度が向上する。 Since the prediction model differs between the preliminary prediction by the simple prediction unit 401 and the main prediction by the individual prediction unit 304, the degree of freedom to select a prediction model for each pedestrian 603, 604 is improved.
　簡易予測部４０１は、線形予測法によって移動物体の将来位置を簡易予測するので、移動物体行動予測部２０２の計算負荷を軽減することができる。 Since the simple prediction unit 401 simply predicts the future position of the moving object by the linear prediction method, the calculation load of the moving object behavior prediction unit 202 can be reduced.
　個別予測部３０４は、ポテンシャルマップを用いるポテンシャル法によって歩行者６０４の将来位置を予測するので、高い精度で歩行者６０４を本予測することができる。 Since the individual prediction unit 304 predicts the future position of the pedestrian 604 by the potential method using the potential map, the pedestrian 604 can be predicted with high accuracy.
　簡易予測部４０１と、個別予測部３０４とは、複数の歩行者６０３，６０４の各々の将来位置を個別に予測するので、歩行者６０３，６０４毎の予測精度を高めることができる。 Since the simple prediction unit 401 and the individual prediction unit 304 individually predict the future positions of the plurality of pedestrians 603 and 604, the prediction accuracy for each pedestrian 603 and 604 can be increased.
　簡易予測の結果に応じて本予測をしないと判断された歩行者６０３には、簡易予測の結果を流用するので、歩行者６０３に対する計算負荷を軽減することができる。 The calculation load on the pedestrian 603 can be reduced because the result of the simple prediction is used for the pedestrian 603 determined not to perform the main prediction according to the result of the simple prediction.
　図１０は、実施例２の車両走行制御装置１に実装されている予測物体精度判定部３０１のブロック図を示す。実施例２の予測物体精度判定部３０１には、実施例１の予測物体精度判定部３０１に相互作用判定部１００４が追加されている。相互作用判定部１００４は、各予測移動物体の簡易予測結果に基づいて、それぞれの移動物体が相互作用を起こすかを判定する演算を行なう。具体的には、簡易予測演算結果でそれぞれの予測結果の軌跡が交差する、若しくは将来時間においてＴＴＣが所定値以下となる場合があると判定された場合には、それらの物体間には相互作用があるとする。 FIG. 10 is a block diagram of the predicted object accuracy determination unit 301 implemented in the vehicle travel control device 1 of the second embodiment. In the predicted object accuracy determination unit 301 according to the second embodiment, an interaction determination unit 1004 is added to the predicted object accuracy determination unit 301 according to the first embodiment. The interaction determination unit 1004 performs an operation for determining whether or not each moving object causes an interaction based on the simple prediction result of each predicted moving object. Specifically, when it is determined that there is a case where trajectories of the respective prediction results intersect with each other in the simple prediction calculation result, or TTC may become a predetermined value or less in the future time, there is an interaction between these objects. Suppose there is.
　図１１および図１２には、２人の歩行者がｘ軸方向およびｙ軸方向に沿って歩行している場合の一例を示す。以下、歩行者間の相互作用について説明する。 11 and 12 show an example in which two pedestrians are walking along the x-axis direction and the y-axis direction. Hereinafter, the interaction between pedestrians will be described.
　図１１では、それぞれの歩行者の１秒後から５秒後までの予測位置を白丸で表す。それぞれの歩行者の予測軌跡は、３～４秒の間に交わっている。図１２は、将来時刻における相互作用確率を表している。相互作用確率の演算は、例えば、ＴＴＣの逆数を規格化する。相互作用確率が所定値以上となる場合には、２人の歩行者は相互作用すると判定される。 In FIG. 11, the predicted positions from 1 second to 5 seconds after each pedestrian are represented by white circles. The predicted trajectory of each pedestrian crosses in 3-4 seconds. FIG. 12 shows the interaction probability at a future time. The calculation of the interaction probability standardizes, for example, the reciprocal of TTC. When the interaction probability is equal to or higher than a predetermined value, it is determined that the two pedestrians interact.
　図１０に戻る。次いで、相互作用を含む精度情報演算部１００５について説明する。精度情報演算部１００５は、自車影響判定結果および相互作用判定結果に基づいて、精度情報を演算する。 Return to FIG. Next, the accuracy information calculation unit 1005 including the interaction will be described. The accuracy information calculation unit 1005 calculates accuracy information based on the own vehicle influence determination result and the interaction determination result.
　図１３は、自車影響判定結果および相互作用判定結果の関係を示す。 FIG. 13 shows the relationship between the vehicle effect determination result and the interaction determination result.
　具体的には、図１３に示すように、相互作用判定結果が「無し」のときは、実施例１と同様に、自車影響判定結果に基づいて、精度情報は「低」「中」「高」を切り替える。相互作用判定結果が「有り」のときは、自車影響判定結果が「小」の場合、精度情報を「低」とする。一方で、自車影響判定結果が「中」もしくは「大」の場合、精度情報を「相互作用」とする。精度情報が「相互作用」のときは、移動物体の相互作用を考慮することが可能な予測モデルを使う。尚、相互作用を考慮することが可能な例としては、前述したポテンシャル法を用いた予測方法がある。 Specifically, as shown in FIG. 13, when the interaction determination result is “none”, the accuracy information is “low”, “medium”, “ Toggle high. When the interaction determination result is “present”, the accuracy information is set to “low” when the own vehicle influence determination result is “small”. On the other hand, when the vehicle influence determination result is “medium” or “large”, the accuracy information is “interaction”. When the accuracy information is “interaction”, a prediction model capable of considering the interaction of moving objects is used. An example in which the interaction can be considered is a prediction method using the potential method described above.
　図１４は、本実施例の自車両６０１と、その周囲の歩行者６０３，６０４および駐車車両１４１１との一例を示す。 FIG. 14 shows an example of the own vehicle 601 of the present embodiment and the surrounding pedestrians 603 and 604 and the parked vehicle 1411.
　図１４において、自車両６０１は、両側にある道路端６０２で規格される道路上を、走行計画軌道６０５に沿って自動運転走行している。自車両６０１の周囲には、二人の歩行者６０３，６０４が歩行している。歩行者６０３は、自車両６０１の右後方を自車両６０１とは反対の方向に歩行している。歩行者６０４は、自車両６０１と平行方向に歩行している。さらに、自車両６０１の左前方には、駐車車両１４１１が駐車している。 In FIG. 14, the host vehicle 601 travels automatically along a travel planned track 605 on a road defined by road ends 602 on both sides. Two pedestrians 603 and 604 are walking around the host vehicle 601. The pedestrian 603 is walking in the direction opposite to the host vehicle 601 on the right rear side of the host vehicle 601. A pedestrian 604 is walking in a direction parallel to the host vehicle 601. Furthermore, a parked vehicle 1411 is parked in front of the host vehicle 601.
　図１５は、本実施例の図１４の状況における制御フローを示す。 FIG. 15 shows a control flow in the situation of FIG. 14 of the present embodiment.
　ステップＳ７００と、ステップＳ７０１と、ステップＳ７０２とにおいて、車両走行制御装置１の制御部（ＣＰＵ）は、歩行者６０３、歩行者６０４および駐車車両１４１１に対して、実施例１と同様の方法で処理を行なう。歩行者６０３は、実施例１と同様に、走行計画軌道６０５への影響は考えられないため、自車影響判定は「低」となる。歩行者６０４は、現時刻において自車両６０１と平行方向に歩行しており、自車両６０１との距離は近いことから衝突確率も小さくない。そのため、ここでは影響度は「中」であると判定される。尚、駐車車両１４１１は、停車しているため、自車への影響度は「低」となる。 In step S700, step S701, and step S702, the control unit (CPU) of the vehicle travel control device 1 processes the pedestrian 603, the pedestrian 604, and the parked vehicle 1411 in the same manner as in the first embodiment. To do. Since the pedestrian 603 cannot consider the influence on the travel planned track 605 as in the first embodiment, the vehicle influence determination is “low”. The pedestrian 604 is walking in the direction parallel to the host vehicle 601 at the current time, and since the distance from the host vehicle 601 is short, the collision probability is not small. Therefore, the influence degree is determined to be “medium” here. Since the parked vehicle 1411 is stopped, the degree of influence on the own vehicle is “low”.
　ステップＳ１５０６では、車両走行制御装置１の制御部（ＣＰＵ）は、相互作用を判定する。ここでは、各移動物体間の相互作用を演算する。ここでは、歩行者６０４の歩行方向には駐車車両１４１１が存在する。そのため、これら歩行者６０４および駐車車両１４１１間には、相互作用があると判定される。一方、歩行者６０３に関しては、歩行者６０４および駐車車両１４１１との相互作用は「なし」と判定される。 In step S1506, the control unit (CPU) of the vehicle travel control device 1 determines the interaction. Here, the interaction between the moving objects is calculated. Here, the parked vehicle 1411 exists in the walking direction of the pedestrian 604. Therefore, it is determined that there is an interaction between the pedestrian 604 and the parked vehicle 1411. On the other hand, regarding the pedestrian 603, the interaction with the pedestrian 604 and the parked vehicle 1411 is determined as “none”.
　ステップＳ１５０９では、車両走行制御装置１の制御部（ＣＰＵ）は、自車影響度判定および相互作用判定結果に基づいて、予測情報を演算し、各予測部に割振る。ここで、実施例２の移動物体行動予測部２０２は、移動物体行動予測部２０２に、相互作用を演算できる個別予測部３０４を有している。 In step S1509, the control unit (CPU) of the vehicle travel control apparatus 1 calculates prediction information based on the vehicle influence determination and the interaction determination result, and allocates the prediction information to each prediction unit. Here, the moving object behavior prediction unit 202 according to the second embodiment includes an individual prediction unit 304 that can calculate an interaction in the moving object behavior prediction unit 202.
　図１４の状況においては、歩行者６０３に対し、ステップＳ７０４の予測精度（低）モデルを用いた行動予測を行い、歩行者６０４および駐車車両１４１１に対し、ステップＳ１５０７の予測精度（相互作用）モデルを用いて行動予測する。 In the situation of FIG. 14, behavior prediction using the prediction accuracy (low) model of step S704 is performed for the pedestrian 603, and the prediction accuracy (interaction) model of step S1507 is performed for the pedestrian 604 and the parked vehicle 1411. Predict behavior using.
　ステップＳ７０５では、車両走行制御装置１の制御部（ＣＰＵ）は、ステップＳ１５０７およびステップＳ７０４で演算した予測結果を集約する。 In step S705, the control unit (CPU) of the vehicle travel control device 1 collects the prediction results calculated in steps S1507 and S704.
　以上により、自車両６０１への影響が小さく且つ相互作用がない場合には、低精度の予測モデルを使うことによって演算負荷を低減し、自車両６０１への影響が大きく且つ相互作用がある場合は、相互作用を考慮することができる予測モデルを使う。従って、より高精度の予測演算が可能となる。図１４の状況においては、相互作用を用いて演算した結果、歩行者６０４は駐車車両１４１１を避けるような軌道１４１４が予測できるようになる。これにより、相互作用を考えないで予測した場合には自車両６０１への影響が中となっていた歩行者６０４は、自車両６０１の走行計画軌道６０５と交差するように予測されることによって、自車両６０１は、走行計画軌道６０５を変更し、歩行者６０４との衝突を回避するような新たな走行計画軌道１４１１（図中、破線で示す）を演算することが可能となる。 As described above, when the influence on the own vehicle 601 is small and there is no interaction, the calculation load is reduced by using a low-precision prediction model, and the influence on the own vehicle 601 is large and there is an interaction. Use a predictive model that can account for interactions. Therefore, more accurate prediction calculation is possible. In the situation of FIG. 14, as a result of calculation using the interaction, the pedestrian 604 can predict a track 1414 that avoids the parked vehicle 1411. Thereby, when it is predicted without considering the interaction, the pedestrian 604 whose influence on the own vehicle 601 is in the middle is predicted to intersect the travel planned track 605 of the own vehicle 601. The own vehicle 601 can change the travel plan trajectory 605 and calculate a new travel plan trajectory 1411 (indicated by a broken line in the figure) that avoids a collision with the pedestrian 604.
　本実施例によれば、自車両７０１以外の歩行者６０４と駐車車両１４１１との相互作用を予測するので、歩行者６０４と駐車車両１４１１とが相互に干渉する場合であっても、歩行者６０４と駐車車両１４１１との予測精度を高めることができる。 According to the present embodiment, since the interaction between the pedestrian 604 other than the host vehicle 701 and the parked vehicle 1411 is predicted, even if the pedestrian 604 and the parked vehicle 1411 interfere with each other, the pedestrian 604. And the prediction accuracy with the parked vehicle 1411 can be increased.
　複数の歩行者６０３，６０４の各々の将来位置を、逐次繰り返して予測しても良い。これにより、歩行者６０３，６０４の将来位置を高い精度で予測することができる。 The future position of each of the plurality of pedestrians 603 and 604 may be predicted repeatedly and sequentially. Thereby, the future position of the pedestrians 603 and 604 can be predicted with high accuracy.
　図１６は、実施例３の制御フローを示す。 FIG. 16 shows a control flow of the third embodiment.
　実施例３の移動物体行動予測部２０２による制御は、実施例１の移動物体行動予測部２０２による制御のステップ８０４が、車両走行制御装置１の制御部（ＣＰＵ）によって予測精度（中）モデルを用いて行動を予測するステップＳ１６０４に置き換えられている。
即ち、この実施例の移動物体行動予測部２０２は、「第三の予測部」としての個別予測部３０４を有している。この個別予測部３０４は、簡易予測および本予測の中間の精度で歩行者６０３，６０４の将来位置を予測する。 In the control by the moving object behavior prediction unit 202 according to the third embodiment, the control step (804) of the control by the moving object behavior prediction unit 202 according to the first embodiment is performed by the control unit (CPU) of the vehicle travel control device 1 with a prediction accuracy (medium) model. Step S1604 is used to predict the action.
In other words, the moving object behavior prediction unit 202 of this embodiment includes an individual prediction unit 304 as a “third prediction unit”. The individual prediction unit 304 predicts the future positions of the pedestrians 603 and 604 with accuracy intermediate between the simple prediction and the main prediction.
　この実施例によれば、簡易予測部４０１および個別予測部３０４の中間の精度で歩行者６０４の将来位置を予測する個別予測部３０４を備え、物体割振り部３０２は、簡易予測の結果に応じて将来位置を予測する歩行者６０３，６０４を第一の予測部および前記第三の予測部に割振る According to this embodiment, the individual predicting unit 304 that predicts the future position of the pedestrian 604 with intermediate accuracy between the simple predicting unit 401 and the individual predicting unit 304 is provided, and the object allocating unit 302 depends on the result of the simple prediction. Pedestrians 603 and 604 that predict future positions are allocated to the first prediction unit and the third prediction unit.
　図１７は、実施例４に係る移動物体行動予測部２０２のブロック図である。 FIG. 17 is a block diagram of the moving object behavior prediction unit 202 according to the fourth embodiment.
　この実施例の移動物体行動予測部２０２の各個別予測部３０３，３０４は、モデル精度切替部１７０４を有している。モデル精度切替部１７０４は、「相互作用モデル」を選択する。この相互作用モデルには、複数の個別物体情報を入力することが可能である。 Each individual prediction unit 303, 304 of the moving object behavior prediction unit 202 of this embodiment has a model accuracy switching unit 1704. The model accuracy switching unit 1704 selects “interaction model”. A plurality of pieces of individual object information can be input to this interaction model.
　ここで、複数の個別予測部３０３，３０４が、事前に予測モデル毎にメモリに用意されている必要があるため、用意した予測モデル以上の物体を、その予測精度で演算できない場合がある。例えば、高精度予測モデルを有するＮ個の個別予測部３０４をメモリに用意した場合に、Ｎ＋１個以上の物体情報を高精度に予測しようとしても不可能となる。そのため、図１６に示すように、精度情報に基づいて、個別予測部３０３，３０４において予測に用いる予測モデルを切り替えるモデル精度切替部１７０４を有している。この構成では、不必要にメモリ領域を確保する必要がなくなるため、メモリを有効に利用することが可能となる。ここでは、「低」と記載されているモデル精度切替部１７０４は、低精度の予測モデルを使用することを示す。 Here, since the plurality of individual prediction units 303 and 304 need to be prepared in advance for each prediction model in the memory, there may be a case where an object larger than the prepared prediction model cannot be calculated with the prediction accuracy. For example, when N individual prediction units 304 having a high-precision prediction model are prepared in the memory, it is impossible to predict N + 1 or more pieces of object information with high accuracy. For this reason, as shown in FIG. 16, a model accuracy switching unit 1704 that switches prediction models used for prediction in the individual prediction units 303 and 304 based on accuracy information is provided. In this configuration, it is not necessary to secure a memory area unnecessarily, so that the memory can be used effectively. Here, the model accuracy switching unit 1704 described as “low” indicates that a low-precision prediction model is used.
　尚、個別予測部３０３，３０４の個数を、演算周期毎もしくはある周囲環境毎に変更できる構成としても良い。例えば、歩行者や自転車が存在する一般道を走行する場合は、自動車専用道路を走行する場合と比較して、予測する物体の個数を多くする必要がある。そのため、自車両が一般道を走行する場合は、個別予測部３０３，３０４の個数を、自動車専用道を走行する場合と比較して多く用意してもよい。これにより、事前に個別予測部３０３，３０４に対して、メモリ領域を不必要に多く確保する必要がなくなるため、不要なメモリ領域を削減することが可能となる。 In addition, it is good also as a structure which can change the number of the individual prediction parts 303 and 304 for every calculation period or every surrounding environment. For example, when traveling on a general road where pedestrians and bicycles exist, it is necessary to increase the number of objects to be predicted as compared to traveling on an automobile-only road. Therefore, when the host vehicle travels on a general road, a larger number of individual prediction units 303 and 304 may be prepared than when traveling on an automobile-only road. This eliminates the need to reserve an unnecessarily large memory area for the individual prediction units 303 and 304 in advance, thereby reducing unnecessary memory areas.
１０１…自車両、２０２…移動物体行動予測部、３０２…物体割振り部、３０３，３０４…個別予測部、４０１…簡易予測部、６０１…自車両、６０３，６０４…歩行者、１００４…相互作用判定部、１４１１…駐車車両 DESCRIPTION OF SYMBOLS 101 ... Own vehicle, 202 ... Moving object action prediction part, 302 ... Object allocation part, 303, 304 ... Individual prediction part, 401 ... Simple prediction part, 601 ... Own vehicle, 603, 604 ... Pedestrian, 1004 ... Interaction determination Part, 1411 ... parked vehicle

Claims (11)

  1. 　自車両の周囲の移動物体の将来位置を予測する移動物体予測装置であって、
    　前記移動物体の将来位置を簡易予測する第一の予測部と、
    　前記第一の予測部よりも高精度に前記移動物体の将来位置を本予測する第二の予測部と、
    　前記簡易予測の結果に応じて前記第二の予測部によって前記将来位置を本予測する移動物体を割振る割振り部とを備える移動物体予測装置。     A moving object prediction apparatus for predicting a future position of a moving object around the host vehicle,
    A first prediction unit that simply predicts the future position of the moving object;
    A second prediction unit that performs the main prediction of the future position of the moving object with higher accuracy than the first prediction unit;
    A moving object prediction apparatus comprising: an allocating unit that allocates a moving object for which the future position is predicted by the second prediction unit according to the result of the simple prediction.
  2. 　前記割振り部は、前記自車両と干渉する可能性が高い移動物体を前記第二の予測部に割振る、請求項１に記載の移動物体予測装置。 The moving object prediction apparatus according to claim 1, wherein the allocation unit allocates a moving object that is highly likely to interfere with the host vehicle to the second prediction unit.
  3. 　前記本予測は、前記簡易予測よりも演算周期が短い、請求項１又は２に記載の移動物体予測装置。 The moving object prediction apparatus according to claim 1 or 2, wherein the main prediction has a shorter calculation cycle than the simple prediction.
  4. 　前記簡易予測と、前記本予測とは、予測モデルが異なる、請求項１乃至３の何れか一項に記載の移動物体予測装置。 The moving object prediction apparatus according to any one of claims 1 to 3, wherein a prediction model is different between the simple prediction and the main prediction.
  5. 　前記第一の予測部は、線形予測法によって前記移動物体の将来位置を予測する、請求項１乃至４の何れか一項に記載の移動物体予測装置。 The moving object prediction apparatus according to any one of claims 1 to 4, wherein the first prediction unit predicts a future position of the moving object by a linear prediction method.
  6. 　前記第二の予測部は、ポテンシャルマップを用いるポテンシャル法によって前記移動物体の将来位置を予測する、請求項１乃至５の何れか一項に記載の移動物体予測装置。 The moving object prediction apparatus according to any one of claims 1 to 5, wherein the second prediction unit predicts a future position of the moving object by a potential method using a potential map.
  7. 　前記第一の予測部と、前記第二の予測部とは、複数の前記移動物体の各々の将来位置を個別に予測する、請求項１乃至６の何れか一項に記載の移動物体予測装置。 The moving object prediction apparatus according to any one of claims 1 to 6, wherein the first prediction unit and the second prediction unit individually predict future positions of the plurality of moving objects. .
  8. 　前記簡易予測の結果に応じて前記本予測をしないと判断された移動物体には、前記簡易予測の結果を流用する、請求項１乃至７の何れか一項に記載の移動物体予測装置。 The moving object prediction device according to any one of claims 1 to 7, wherein the result of the simple prediction is used for a moving object that is determined not to perform the main prediction according to the result of the simple prediction.
  9. 　複数の前記移動物体同士の相互作用を予測する相互作用判定部を備える、請求項１乃至８の何れか一項に記載の移動物体予測装置。 The moving object prediction apparatus according to any one of claims 1 to 8, further comprising an interaction determination unit that predicts an interaction between the plurality of moving objects.
  10. 　複数の前記移動物体の各々の将来位置を、逐次繰り返して予測する、請求項９に記載の移動物体予測装置。 The moving object prediction apparatus according to claim 9, wherein the future position of each of the plurality of moving objects is predicted sequentially and repeatedly.
  11. 　前記第一の予測部および前記第二の予測部の中間の精度で前記移動物体の将来位置を予測する第三の予測部を備え、
    　前記割振り部は、前記簡易予測の結果に応じて前記将来位置を予測する移動物体を前記第一の予測部および前記第三の予測部に割振る、請求項１乃至７の何れか一項に記載の移動物体予測装置。     A third prediction unit for predicting the future position of the moving object with an intermediate accuracy between the first prediction unit and the second prediction unit;
    The allocation unit according to any one of claims 1 to 7, wherein the allocation unit allocates a moving object that predicts the future position to the first prediction unit and the third prediction unit according to the result of the simple prediction. The moving object prediction apparatus described.
PCT/JP2018/001136 2017-01-30 2018-01-17 Moving object prediction device WO2018139294A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2017-014571 2017-01-30
JP2017014571A JP2018124663A (en) 2017-01-30 2017-01-30 Mobile object predictor

Publications (1)

Publication Number Publication Date
WO2018139294A1 true WO2018139294A1 (en) 2018-08-02

Family

ID=62978989

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2018/001136 WO2018139294A1 (en) 2017-01-30 2018-01-17 Moving object prediction device

Country Status (2)

Country Link
JP (1) JP2018124663A (en)
WO (1) WO2018139294A1 (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP7466407B2 (en) 2019-10-18 2024-04-12 本田技研工業株式会社 FUTURE BEHAVIOR ESTIMATION DEVICE, VEHICLE CONTROL DEVICE, FUTURE BEHAVIOR ESTIMATION METHOD, AND PROGRAM
JP7232166B2 (en) 2019-10-18 2023-03-02 本田技研工業株式会社 Prediction device, vehicle system, prediction method, and program
JP7470584B2 (en) 2020-07-03 2024-04-18 日産自動車株式会社 Driving support method and driving support device
CN113393669A (en) * 2021-06-11 2021-09-14 阿波罗智联(北京)科技有限公司 Control method, device, equipment, medium and program product for vehicle
WO2023209942A1 (en) * 2022-04-28 2023-11-02 株式会社Subaru Driving assistance system, vehicle, and recording medium with computer program recorded therein

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011248445A (en) * 2010-05-24 2011-12-08 Toyota Central R&D Labs Inc Moving object prediction device and program

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011248445A (en) * 2010-05-24 2011-12-08 Toyota Central R&D Labs Inc Moving object prediction device and program

Also Published As

Publication number Publication date
JP2018124663A (en) 2018-08-09

Similar Documents

Publication Publication Date Title
WO2018139294A1 (en) Moving object prediction device
JP6532786B2 (en) Vehicle travel control device and speed control method
JP6839770B2 (en) Mobile control system and control device
JP6294247B2 (en) Vehicle travel control device
JP7253908B2 (en) vehicle controller
JP5174609B2 (en) Driving support device
US20190106103A1 (en) Driving assistance control device
US7363155B2 (en) Method and device for warning the driver of a motor vehicle
JP6138655B2 (en) Vehicle motion control device
CN109131312B (en) ACC/ESC integrated control system and method for intelligent electric vehicle
CN111770865B (en) Vehicle travel control device
WO2021124794A1 (en) Vehicle control device, and vehicle control system
JP2003112618A (en) Braking controller for vehicle
US20210387632A1 (en) Vehicle control system
Shibata et al. Collision avoidance control with steering using velocity potential field
JPWO2017217256A1 (en) Vehicle control device
CN103827940A (en) Vehicle driving assistance system
CN112092805A (en) Integrated control method and system for collision avoidance and collision damage reduction of intelligent vehicle
WO2022004042A1 (en) Vehicle control device and vehicle control system
JP7200786B2 (en) vehicle braking controller
JP7196448B2 (en) Collision control device
US20240149889A1 (en) System
JP2018176814A (en) Deflection control apparatus
WO2023112385A1 (en) Vehicle control device
Drechsel et al. Wheel individual longitudinal control systems

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18744735

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18744735

Country of ref document: EP

Kind code of ref document: A1