Tracking Device

Abstract

The tracking device (4) includes a state quantity estimation unit (S40, S220, S230), a model selection unit (S50), and an estimation selection unit (S210). Each time a processing cycle is repeated, the state quantity estimation unit estimates the current state quantity of each object based on observation information obtained by observing one or more object targets by the sensor (2) and at least one of the past state quantities of the one or more object targets. The model selection unit selects a motion model from a plurality of motion models for each object target based on at least one of the state of the object target and the state of the vehicle. The estimation selection unit causes the state quantity estimation unit to estimate the state quantity using the selected motion model for each object target.

Description

Technical Field

[0001] The present invention relates to a tracking device for tracking an object existing around a vehicle. Background Art

[0002] In order to calculate the state quantities such as the position and speed of the target object by the vehicle-mounted radar, the predicted value in the current processing cycle obtained based on the estimated value of the position and speed of the target object in the previous processing cycle and the filtering process of the observed value in this processing cycle are generally performed. However, when there is a deviation between the motion model used for calculating the predicted value and the actual motion of the target object, the state quantity of the target object cannot be estimated with good accuracy. In response to this, the following target tracking device is described in Patent Document 1: the state quantity of the target object is estimated by multiple filter processing units with different motion models, and the output of any filter processing unit is selected according to the convergence status of the estimation result, so as to obtain an estimation result with good accuracy.

[0003] Patent Document 1: Japanese Patent No. 4348535

[0004] However, as a result of detailed studies by the inventors, it was found that the technology described in Patent Document 1 has a problem that the processing load increases because a plurality of filter processes need to be performed in parallel on one target object. Summary of the invention

[0005] The present invention tracks a target object with high accuracy while suppressing an increase in processing load.

[0006] One aspect of the present invention is a tracking device that is mounted on a vehicle and tracks one or more target objects that exist around the vehicle. The device includes a state quantity estimation unit, a model selection unit, and an estimation selection unit.

[0007] The state quantity estimating unit is configured to estimate a current state quantity of each of the one or more objects based on at least one of observation information obtained by observing the one or more objects by a sensor and past state quantities of the one or more objects, every time a repetition period of a preset processing loop elapses.

[0008] The model selection unit is configured to select a motion model from a plurality of preset motion models for each of the one or more object targets based on at least one of a state of the one or more object targets and a state of the vehicle.

[0009] The estimation selection unit is configured to cause the state quantity estimation unit to estimate the state quantity using the one motion model selected by the model selection unit for each of the one or more target objects.

[0010] The tracking device of the present invention constructed in this way uses one motion model to estimate the state quantity for one object. Therefore, the tracking device of the present invention can suppress the occurrence of a situation where multiple motion models are executed in parallel for one object, and can suppress the increase in processing load. In addition, the tracking device of the present invention selects a motion model from multiple motion models to estimate the current state quantity of each of one or more objects. Therefore, the tracking device of the present invention can select an appropriate motion model for each of one or more objects to estimate the state quantity, and can improve the estimation accuracy of the state quantity. Therefore, the tracking device of the present invention can track objects with good accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a block diagram showing the structure of a driving assistance system.

[0012] Figure 2 It is a diagram showing the installation position and detection range of the radar device in the first embodiment.

[0013] Figure 3 This is a flowchart showing the tracking process.

[0014] Figure 4 is a flowchart showing the prediction process.

[0015] Figure 5 This is a flowchart showing the state quantity prediction processing.

[0016] Figure 6 It is a flowchart which shows the association establishment process.

[0017] Figure 7 This is a flowchart showing the model setting process according to the first embodiment.

[0018] Figure 8 : is a flowchart showing the first setting process.

[0019] Fig. 9 : is a flowchart showing the second setting process.

[0020] Fig.10 It is a flowchart which shows the registration process.

[0021] Fig.11 This figure shows a pedestrian who is moving forward to cross the road in front of the host vehicle.

[0022] Fig.12 The diagram shows a host vehicle and a preceding vehicle traveling just ahead of a curve on a road.

[0023] Fig.13 This is a diagram showing a state quantity storage unit.

[0024] Fig.14 This is a flowchart showing the model setting process according to the second embodiment.

[0025] Fig.15 It is a diagram showing the installation position and detection range of a radar device in another embodiment. DETAILED DESCRIPTION

[0026] (First Embodiment)

[0027] The following, with the attached Figure 1 A first embodiment of the present invention will be described below.

[0028] The driving assistance system 1 of this embodiment is installed in a vehicle. Figure 1 As shown, the driving support system 1 includes a radar device 2, a behavior sensor group 3, a tracking device 4, and an assistance execution unit 5. Hereinafter, a vehicle equipped with the driving support system 1 is referred to as a host vehicle.

[0029] The radar device 2 transmits radar waves such as millimeter waves or microwaves, and receives reflected radar waves. Figure 2 As shown, the radar device 2 generates information (hereinafter referred to as observation information) about a reflection point that exists in the detection range Rd and reflects the radar wave by transmitting the radar wave toward the front of the vehicle. The observation information includes the distance to the reflection point, the azimuth of the reflection point, the relative speed of the reflection point, and the intensity of the radar wave reflected by the reflection point (hereinafter referred to as reflection intensity).

[0030] like Figure 1 As shown, the behavior sensor group 3 is composed of various devices for detecting information related to the behavior of the vehicle and information related to the driving operation that affects the behavior of the vehicle. The information detected by the behavior sensor group 3 includes, for example, the operation amount of the accelerator pedal, the operation amount of the brake pedal, the steering operation angle, the vehicle speed, and the vehicle acceleration.

[0031] The tracking device 4 tracks a target object existing around the host vehicle based on the observation information generated by the radar device 2 , and generates target object information indicating a state quantity of the tracked target object.

[0032] The tracking device 4 is an electronic control device centered around a microcomputer having a CPU 11, a ROM 12, and a RAM 13. The various functions of the microcomputer are realized by the CPU 11 executing a program stored in a non-migrating physical recording medium. In this example, the ROM 12 is equivalent to a non-migrating physical recording medium storing the program. In addition, the method corresponding to the program is executed by executing the program. In addition, a part or all of the functions performed by the CPU 11 can be constituted in hardware by one or more ICs. In addition, the number of microcomputers constituting the tracking device 4 can be either one or more.

[0033] The assistance execution unit 5 includes, for example, an actuator, an audio device, a display device, etc. The assistance execution unit 5 controls the behavior of the host vehicle based on the target object information generated by the tracking device 4 or performs a report to the driver of the host vehicle.

[0034] The following describes the order of the tracking process executed by the CPU 11 of the tracking device 4. The tracking process is a process that is repeatedly executed in each preset processing cycle during the operation of the tracking device 4. The repetition period of the processing cycle is ΔT. Hereinafter, the object that is tracked by this process is referred to as a tracking object.

[0035] If tracking processing is performed, Figure 3 As shown, the CPU 11 first executes a prediction process in S10. Here, the sequence of the prediction process executed in S10 will be described.

[0036] If the prediction process is performed, Figure 4 As shown, the CPU 11 first selects one tracking object that has not been selected in the current prediction process from one or more tracking objects in S110. Hereinafter, the selected tracking object is referred to as a target object.

[0037] Then, in S120, the CPU 11 executes the state quantity prediction process. Here, the procedure of the state quantity prediction process executed in S120 will be described.

[0038] If the state quantity prediction processing is performed, then Figure 5 As shown, the CPU 11 first determines in S210 whether to set the first motion model described later or the second motion model described later for the target object. Here, in the case of setting the first motion model, the CPU 11 performs the first prediction processing described later in S220 and ends the state quantity prediction processing. On the other hand, in the case of setting the second motion model, the CPU 11 performs the second prediction processing described later in S230 and ends the state quantity prediction processing.

[0039] Next, the first prediction process executed in S220 will be described.

[0040] In the first prediction process, the CPU 11 uses the first motion model to calculate the predicted value of the state vector in the current processing cycle based on the estimated value of the state vector of the target object obtained in the previous processing cycle. The following equations (1) to (6) represent the extended Kalman filter.

[0041] X k|k－1 is the predicted value of the state vector (i.e., the prediction vector). k is the estimated value of the state vector. k is the observed value.k|k－1 is the error covariance matrix of the predicted value of the state vector. k is the estimated value of the error covariance matrix. S k is the innovation matrix. K k is the Kalman gain. f is the function given to the predicted value from the previous state. h is the function given to the observed value. Q k is the variance of the process noise. k is the state transition matrix defined by the Jacobian matrix of function f. k is the error covariance matrix of the observation noise. k It is a transformation matrix that maps the state space defined by the Jacobian matrix of the function h to the observation space. In addition, X0 is the initial value of the state vector (ie, the initial estimated vector), and P0 represents the initial value of the error covariance matrix.

[0042] [Formula 1]

[0043] X k|k-1 =f(X k-1 ) (1)

[0044]

[0045]

[0046]

[0047] P k =(IK k H k ) k|k-1 (5)

[0048] X k =X k|k-1 +K k (z k -h(X k|k-1 )) (6)

[0049] Moreover, the first motion model is a linear motion model. Fig.11 As shown, the linear motion model is a model for detecting a pedestrian PD who is moving forward to cross the road in front of the host vehicle V1.

[0050] In the first motion model, the state of the object is composed of the lateral position x, longitudinal position y, lateral velocity v x and the longitudinal velocity v y The state vector that represents these state quantities as vector elements is X = [x, y, vx, vy] T In addition, the lateral direction is the vehicle width direction of the vehicle, and the longitudinal direction is the direction perpendicular to the vehicle width direction.

[0051] In the first motion model, the predicted value of the state quantity in the current processing cycle is set as the state vector X k|k－1 =[x', y', v x ',v y '] T , the estimated value of the state quantity in the previous processing cycle is set as the state vector X k－1 =[x、y、v x 、v y ] T Moreover, the function f is set to be k|k－1 With the state vector X k－1 The relationship shown in formula (7) holds.

[0052] [Formula 2]

[0053]

[0054] That is, in the first prediction processing of S220, CPU11 uses the function f set in formula (7) to calculate the predicted value of the state vector in this processing cycle through formula (1), and calculates the error covariance matrix of the predicted value of the state vector in this processing cycle through formula (2).

[0055] Next, the second prediction process executed in S230 will be described.

[0056] In the second prediction process of S230 , the CPU 11 uses the second motion model to calculate a predicted value of the state vector in the current processing cycle based on the estimated value of the state vector of the target object obtained in the previous processing cycle.

[0057] Furthermore, the second motion model is a turning motion model. Fig.12 As shown, the turning motion model is a model for detecting a preceding vehicle V2 traveling on a curve CV of a road ahead of the host vehicle V1.

[0058] In the second motion model, the state of the target object is represented by the lateral position x, longitudinal position y, velocity v, travel direction θ, acceleration a, and yaw rate ω. The state vector representing these state as elements is X = [x, y, v, θ, a, ω] T express.

[0059] In the second motion model, the predicted value of the state quantity in the current processing cycle is set as the state vector X k|k－1 =[x',y',v',θ',a',ω'] T , the estimated value of the state quantity in the previous processing cycle is set as the state vector X k－1 =[x, y, v, θ, a, ω]T Moreover, the function f is set to be k|k－1 With the state vector X k－1 The relationship shown in formula (8) holds.

[0060] [Formula 3]

[0061]

[0062] That is, in the second prediction processing of S230, CPU11 uses the function f set in formula (8) to calculate the predicted value of the state vector in this processing cycle through formula (1), and calculates the error covariance matrix of the predicted value of the state vector in this processing cycle through formula (2).

[0063] Furthermore, the CPU 11 stores the calculated state quantity in the state quantity storage unit 13 a provided in the RAM 13 .

[0064] like Fig.13 As shown, the state quantity storage unit 13a is formed to store [x, y, v x 、v y ] or [x, y, v, θ, a, ω]. Specifically, for each of the plurality of object targets, storage areas R1, R2, R3, R4, R5, and R6 are provided. When the CPU 11 calculates the estimated value of the state vector using the first motion model, for example, x, y, v, θ, a, ω are stored in the storage areas R1, R2, R3, and R4 of the corresponding object targets, respectively. x 、v y When the CPU 11 calculates the estimated value of the state vector using the second motion model, the CPU 11 stores x, y, v, θ, a, ω in the corresponding target object storage areas R1, R2, R3, R4, R5, and R6, respectively.

[0065] If the state quantity prediction process is completed, then Figure 4 As shown, the CPU 11 determines in S130 whether all the tracking objects among the one or more tracking objects have been selected in S110. Here, if all the tracking objects have not been selected, the CPU 11 proceeds to S110. On the other hand, if all the tracking objects have been selected, the CPU 11 ends the prediction process.

[0066] If the prediction process is finished, Figure 3 As shown, the CPU 11 executes an association process in S20 to associate the reflection point detected by the radar device 2 with the tracking target object. Here, the procedure of the association process executed in S20 will be described.

[0067] If the association process is executed, Figure 6As shown, the CPU 11 first selects a reflection point which has not been selected in the current association process from one or more reflection points detected in the current processing cycle in S310. Hereinafter, the selected reflection point is referred to as a selected reflection point.

[0068] Next, in S320 , the CPU 11 selects one tracking object that has not been selected in the current association process from one or more tracking objects as a selected object. Then, in S330 , the CPU 11 calculates the distance between the selected reflection point and the selected object.

[0069] Then, in S340, the CPU 11 determines whether all the tracking objects among the one or more tracking objects are selected in S320. Here, if all the tracking objects are not selected, the CPU 11 moves to S320. On the other hand, if all the tracking objects are selected, the CPU 11 determines in S350 whether all the reflection points among the one or more reflection points detected in this processing cycle are selected in S310. Here, if all the reflection points are not selected, the CPU 11 moves to S310.

[0070] On the other hand, when all reflection points are selected, in S360, the CPU 11 selects one of the one or more tracking objects that has not been processed in S370 as the selected object. Next, in S370, the CPU 11 determines whether to associate the reflection point with the selected object based on the distance calculated in S330. Hereinafter, the reflection point associated with the selected object is referred to as an associated reflection point. As an association method, for example, when the reflection point with the smallest distance from the selected object Mj is the reflection point Pi, and when the object with the smallest distance from the reflection point Pi is the selected object Mj, the reflection point Pi is associated with the selected object Mj.

[0071] Furthermore, in S380 , the CPU 11 calculates the distance between the position of the selected object target indicated by the prediction value of the state vector calculated in S120 and the position of the relevant reflection point indicated by the observation information of the relevant reflection point as the prediction residual of the selected object target.

[0072] Then, in S390, the CPU 11 determines whether all the tracking objects among the one or more tracking objects are selected in S360. Here, if all the tracking objects are not selected, the CPU 11 proceeds to S360. On the other hand, if all the tracking objects are selected, the CPU 11 ends the association process.

[0073] If the association process is completed, Figure 3As shown, the CPU 11 performs a disappearance process in S30 to make the tracking object without the associated reflection point disappear. Specifically, a disappearance counter is set for each tracking object, and when there is an associated reflection point, the CPU 11 resets the disappearance counter. On the other hand, when there is no associated reflection point, the CPU 11 increments the disappearance counter by 1. Moreover, when the value of the disappearance counter reaches a specified value, that is, when there is no associated reflection point continuously in a specified number of processing cycles, the CPU 11 makes the tracking object disappear.

[0074] Then, in S40, the CPU 11 performs an update process of calculating the estimated value of the state vector of the tracked target in this processing cycle based on the predicted value of the state vector of the tracked target and the observation information of the associated reflection point that are associated with each other in the association process of S20. Specifically, the CPU 11 uses equations (3) to (6) to calculate the estimated value of the state vector in this processing cycle (i.e., the state vector X k ) and the estimated values ​​of the error covariance matrix (i.e., rows and columns P k ). In addition, the CPU 11 sets the predicted values ​​of the state vector and the error covariance matrix of the tracking target without the associated reflection point in S370 as the estimated values.

[0075] When the update process is completed, the CPU 11 executes a model setting process in S50. Here, the procedure of the model setting process executed in S50 will be described.

[0076] If the model setting process is executed, Figure 7 As shown, the CPU 11 first selects one tracking object which has not been selected in the current model setting process from one or more tracking objects in S410. Hereinafter, the selected tracking object is referred to as a target object.

[0077] Then, in S411, the CPU 11 determines whether the later-described elapsed cycle of the target object is less than a preset fifth determination value. In addition, a elapsed cycle counter is set for each tracking object. And, each time a repetition cycle of a processing cycle has passed since the tracking object was detected, the CPU 11 increments the corresponding elapsed cycle counter by 1 (i.e., adds 1). The elapsed cycle is the value of the elapsed cycle counter, which is equivalent to the number of processing cycles that have passed since the target object was detected.

[0078] Here, when the elapsed cycle is equal to or smaller than the fifth determination value, the CPU 11 proceeds to S510 .

[0079] On the other hand, if the elapsed cycle exceeds the fifth determination value, the CPU 11 determines in step S420 the vehicle speed V obtained from the behavior sensor group 3. nThe ground speed V is calculated using the longitudinal speed calculated using the estimated value of the state vector calculated in S40. abs . Ground speed V abs The ground speed V is calculated by adding the relative speed of the object along the direction of travel of the vehicle to the vehicle's travel speed. abs .

[0080] Specifically, when the CPU 11 calculates the predicted value of the state vector using the first motion model in S120, the ground speed V is calculated using equation (9): abs When the CPU 11 calculates the predicted value of the state vector by the second motion model in S120, the ground speed V is calculated by equation (10): abs .

[0081] V abs =V n +v y …(9)

[0082] V abs =V n +v / cosθ …(10)

[0083] Then, in S430, the CPU 11 determines whether the ground speed V calculated in S420 is abs Here, the ground speed V abs If the ground speed V is greater than the first judgment value, the CPU 11 proceeds to S520. abs If it is smaller than the first determination value, the CPU 11 determines in S440 whether the vertical position of the target object marker is equal to or larger than a preset second determination value.

[0084] Here, when the longitudinal position is greater than the second determination value, the CPU 11 moves to S520. On the other hand, when the longitudinal position is less than the second determination value, the CPU 11 determines in S450 whether the lateral position of the target object is less than a preset third determination value. Here, when the lateral position is less than the third determination value, the CPU 11 moves to S520.

[0085] On the other hand, when the lateral position is greater than the third determination value, in S460, the CPU 11 determines whether the reflection intensity of the target object is greater than a preset fourth determination value based on the observation information corresponding to the target object. Here, when the reflection intensity is greater than the fourth determination value, the CPU 11 moves to S520.

[0086] On the other hand, when the reflection intensity is smaller than the fourth determination value, in S480 , the CPU 11 calculates the turning radius of the travel path on which the host vehicle is traveling based on the steering operation angle acquired from the behavior sensor group 3 .

[0087] Then, in S490, the CPU 11 determines whether the rotation radius calculated in S480 is less than a preset sixth determination value. Here, when the rotation radius is less than the sixth determination value, the CPU 11 moves to S520. On the other hand, when the rotation radius is greater than the sixth determination value, in S500, the CPU 11 determines whether the absolute value of the vehicle acceleration is greater than a preset seventh determination value based on the vehicle acceleration acquired from the behavior sensor group 3.

[0088] Here, when the absolute value of the vehicle acceleration is greater than the seventh determination value, the CPU 11 moves to S520. On the other hand, when the absolute value of the vehicle acceleration is less than the seventh determination value, in S510, the CPU 11 executes the first setting process described later and moves to S530. In addition, if the process moves to S520, the CPU 11 executes the second setting process described later and moves to S530.

[0089] Then, when the process proceeds to S530, it is determined whether all the tracking objects among the one or more tracking objects have been selected in S410. Here, if all the tracking objects have not been selected, the CPU 11 proceeds to S410. On the other hand, if all the tracking objects have been selected, the CPU 11 ends the model setting process.

[0090] Here, the procedure of the first setting process executed in S510 will be described.

[0091] If the first setting process is executed, Figure 8 As shown, the CPU 11 first determines whether the second motion model is set for the target object in S610. Here, if the second motion model is not set, the CPU 11 ends the first setting process. On the other hand, if the second motion model is set, the CPU 11 converts the state vector represented by the state variables x, y, v, θ, a, ω and the error covariance matrix into the state vector represented by the state variables x, y, v, θ, a, ω and the error covariance matrix in S620. x 、v y Then, in S630, the CPU 11 sets a first motion model for the target object and ends the first setting process.

[0092] Next, the procedure of the second setting process executed in S520 will be described.

[0093] If the second setting process is executed, Fig. 9As shown, the CPU 11 first determines whether the first motion model is set for the target object in S710. Here, if the first motion model is not set, the CPU 11 ends the second setting process. On the other hand, if the first motion model is set, the CPU 11 sets the state variables x, y, and v in S720. x 、v y The state vector and error covariance matrix represented by are converted into the state vector and error covariance matrix represented by the state variables x, y, v, θ, a, ω. Then, at S730, the CPU 11 sets the second motion model for the target object and ends the second setting process.

[0094] If the model setting process is completed, Figure 3 As shown, the CPU 11 executes a registration process of newly registering a reflection point that is not associated with a tracking object as a new object in S60. Here, the order of the registration process executed in S60 will be described.

[0095] If the registration process is executed, Fig.10 As shown, first in S810, CPU 11 sets the reflection points detected in this processing cycle that are not associated with the tracking object as non-associated reflection points, and selects a non-associated reflection point that has not been selected in this registration process from one or more non-associated reflection points.

[0096] Next, in S820, CPU 11 calculates the initial value of the state vector represented by X = [x, y, v, θ, a, ω]T (i.e., the initial estimated vector) based on the observation information of the selected non-associated reflection point (i.e., the distance to the reflection point, the azimuth of the reflection point, and the relative speed of the reflection point) and the steering operation angle and vehicle acceleration obtained from the behavior sensor group 3.

[0097] Then, in S830 , the CPU 11 registers the selected non-correlated reflection point as a new target object together with the initial estimated vector calculated in S820 .

[0098] Then, in S840, the CPU 11 determines whether all the non-related reflection points have been selected in S810. Here, if all the non-related reflection points have not been selected, the CPU 11 moves to S810. On the other hand, if all the non-related reflection points have been selected, the CPU 11 ends the registration process.

[0099] If the registration process is completed, Figure 3As shown, in S70, the CPU 11 generates object information and outputs the generated object information to the auxiliary execution unit 5, temporarily ending the tracking process. Specifically, in S70, the CPU 11 first converts the estimated value of the state vector of one or more tracking objects and the initial value of the state vector of one or more new objects into a state vector represented by the state variable of the first motion model (i.e., represented by X=[x, y, v x 、v y ] T In addition, the estimated value of the state vector of the tracked object is calculated in S40, and the initial value of the state vector of the new object is calculated in S50. That is, the CPU 11 performs conversion processing on the estimated value of the state vector of the tracked object and the initial value of the state vector of the new object when they are represented by the state variables of the second motion model. In addition, the CPU 11 converts the error covariance matrix into the state vector represented by the state variables x, y, v x 、v y Furthermore, the CPU 11 generates as object information an estimated value of a state vector of a tracked object converted into a state variable of the first motion model, an initial value of a state vector of a new object converted into a state variable of the first motion model, and an error covariance matrix of the state variable converted into the first motion model.

[0100] The tracking device 4 configured in this manner is mounted on a vehicle and tracks one or more target objects existing around the vehicle.

[0101] Each time a predetermined processing cycle repetition period ΔT elapses, the tracking device 4 calculates the estimated value X of the state vector of the one or more target objects based on the observation information obtained by the radar device 2 observing the one or more target objects and the past state vector of the one or more target objects. k－1 , to estimate the estimated value X of the current state vector of each of one or more objects k .

[0102] For each of the one or more objects, the tracking device 4 selects a motion model from a preset first motion model and a preset second motion model based on at least one of the state of the one or more objects and the state of the vehicle.

[0103] The tracking device 4 estimates the estimated value X of the state vector for each of the one or more objects using a selected motion model. k .

[0104] Thus, the tracking device 4 estimates the estimated value of the state vector using one motion model for one object. Therefore, the tracking device 4 is prevented from executing multiple motion models in parallel for one object, and the increase in processing load can be suppressed. Furthermore, the tracking device 4 selects one motion model from the multiple motion models to estimate the estimated value X of the current state vector of each of the one or more objects. k Therefore, the tracking device 4 can select an appropriate motion model for each of one or more objects to estimate the estimated value X of the state vector k , and can improve the estimated value of the state vector X k Thus, the tracking device 4 can track the target object with good accuracy.

[0105] In addition, the first motion model is a linear motion model, and the second motion model is a turning motion model. The main targets in the radar device 2 mounted on the vehicle are vehicles and pedestrians. Since the vehicle performs a turning action corresponding to the steering operation angle, the turning motion model is suitable as a motion model. On the other hand, for example, pedestrians crossing in front of the vehicle are more likely to move in a straight line, and unlike vehicles, they do not perform turning behavior with a constant rotation radius, so the linear motion model is more suitable than the turning motion model.

[0106] The ground speed V of the target abs When it is equal to or greater than the first determination value, the tracking device 4 selects the second motion model.

[0107] When the longitudinal position of the target object is equal to or greater than the second determination value, the tracking device 4 selects the second motion model.

[0108] When the lateral position of the target object is smaller than the third determination value, the tracking device 4 selects the second motion model.

[0109] When the reflection intensity of the target object is equal to or greater than the fourth determination value, the tracking device 4 selects the second motion model.

[0110] When the passing cycle of the target object is equal to or smaller than the fifth determination value, the tracking device 4 selects the first motion model.

[0111] When the rotation radius of the host vehicle is smaller than the sixth determination value, the tracking device 4 selects the second motion model.

[0112] When the absolute value of the vehicle acceleration is equal to or greater than the seventh determination value, the tracking device 4 selects the second motion model.

[0113] The state variables of the second motion model include acceleration a. Thus, the tracking device 4 can improve the tracking accuracy of the target object.

[0114] When a negative determination is made in S430, S440, S450, S460, S470, S490, and S500, the tracking device 4 selects the first motion model. The determination conditions in S430, S440, S450, S460, S470, S490, and S500 are second motion model selection conditions set for selecting the second motion model.

[0115] The tracking device 4 also estimates an estimated value P of the error covariance matrix k .

[0116] When a motion model is changed from the last time for each of the one or more objects, the tracking device 4 converts the state vector and the error covariance matrix to correspond to the changed motion model.

[0117] The tracking device 4 is provided with a state quantity storage unit 13a for storing state vectors of one or more objects regardless of a selected motion model. The state quantity storage unit 13a has a capacity capable of storing the state vectors of the first motion model and the second motion model having the largest number of state variables. Thus, the tracking device 4 can store the state vectors of the first motion model and the second motion model in the common state quantity storage unit 13a. Therefore, the tracking device 4 does not need to provide separate data storage units in the first motion model and the second motion model in order to store the state vectors of the first motion model and the second motion model, and can suppress the increase of data storage capacity.

[0118] The tracking device 4 converts the state vector and the error covariance matrix of one or more objects into state variables x, y, v independently of the selected motion model. x 、v y Thus, the tracking device 4 does not need to convert the state vector and the error covariance matrix between the state variables of the first motion model and the state variables of the second motion model in the auxiliary execution unit 5. Therefore, the tracking device 4 can easily use the object information in the auxiliary execution unit 5.

[0119] In the above-described embodiment, S40, S220, and S230 correspond to processing as a state quantity estimation unit, S50 corresponds to processing as a model selection unit, S210 corresponds to processing as an estimation selection unit, and the state vector corresponds to the state quantity.

[0120] In addition, the ground speed of the target V abs , the longitudinal position of the object, the lateral position of the object, the reflection intensity of the object, the passing cycle of the object and the rotation radius of the vehicle are equivalent to the selection parameters.

[0121] In addition, the first determination value, the second determination value, the third determination value, the fourth determination value, the fifth determination value, the sixth determination value, and the seventh determination value correspond to the selection determination value.

[0122] In addition, the error covariance matrix corresponds to the error covariance, S620 and S720 correspond to the processing as the conversion unit, S70 corresponds to the processing as the output unit, and the state variables x, y, and v x 、v y Equivalent to a shared physical quantity.

[0123] (Second Embodiment)

[0124] The following, with the attached Figure 1 A second embodiment of the present invention will be described. In the second embodiment, the parts different from the first embodiment will be described. The same reference numerals are attached to the common structures.

[0125] The driving assistance system 1 according to the second embodiment is different from the first embodiment in that the model setting process is changed.

[0126] The model setting process of the second embodiment is different from the first embodiment in that the processes of S412 , S414 , and S416 are added.

[0127] That is, Fig.14 As shown, in S411, when the number of elapsed cycles exceeds the fifth judgment value, CPU 11 first determines in S412 whether the number of elapsed cycles of the target object is greater than the preset switching judgment cycle. In addition, a setting elapsed cycle counter is set for each tracking object. Moreover, each time a repetition cycle of the processing cycle passes since the tracking object is detected, CPU 11 increments the corresponding setting elapsed cycle counter by 1. However, each time the motion model is switched, CPU 11 resets the corresponding setting elapsed cycle counter (i.e., sets it to 0). The setting elapsed cycle is the value of the setting elapsed cycle counter, which is equivalent to the number of processing cycles that have passed since the current motion model was set.

[0128] Here, when the setting elapsed cycle is smaller than the switching determination cycle, the CPU 11 proceeds to S420 .

[0129] On the other hand, when the setting cycle is set to be more than the switching determination cycle, the CPU 11 determines in S414 whether the prediction residual of the target object is greater than the preset switching determination value. Here, when the prediction residual is less than the switching determination value, the CPU 11 moves to S420. On the other hand, when the prediction residual is greater than the switching determination value, the CPU 11 determines in S416 whether the first motion model is set for the target object. Here, when the first motion model is not set, the CPU 11 moves to S510. On the other hand, when the first motion model is set, the CPU 11 moves to S520.

[0130] The tracking device 4 configured in this way is based on the estimated value X of the past state vector of one or more target objects. k－1 To calculate the predicted value X of the current state vector of each object of one or more objects k|k－1 Furthermore, when the number of cycles set to be the switching determination cycle or more is set, and the difference between the predicted value of the state vector (in this embodiment, the position of the target object) and the observed value represented by the observation information (in this embodiment, the position of the reflection point), i.e., the predicted residual, is greater than the switching determination value, a motion model different from the one selected last time is selected. Thus, the tracking device 4 can improve the tracking accuracy of the target object.

[0131] As mentioned above, although one embodiment of the present invention has been described, the present invention is not limited to the above-mentioned embodiment, and can be implemented with various modifications.

[0132] [Modification 1]

[0133] For example, in the above-mentioned embodiment, the first, second, third, fourth, fifth, sixth, and seventh judgment values ​​are fixed values. However, the first to seventh judgment values ​​may be changed according to a selected motion model. That is, the first to seventh judgment values ​​when the first motion model is selected may be different from the first to seventh judgment values ​​when the second motion model is selected. Thus, the tracking device 4 can set a hysteresis for the threshold value for switching the first motion model and the second motion model, and can suppress the so-called tremor caused by frequent switching of the motion model between the first motion model and the second motion model.

[0134] [Modification 2]

[0135] In the above embodiment, the radar wave is transmitted toward the front of the vehicle. However, the direction of transmitting the radar wave is not limited to the front of the vehicle. Fig.15 As shown, radar waves may be transmitted not only in front of the vehicle but also to the left and right sides in front of the vehicle and to the left and right sides in rear of the vehicle.

[0136] [Variation 3]

[0137] In the above embodiment, the ground speed V of the target object is shown. abs If the speed is equal to or greater than the first determination value, the second motion model is selected. However, the relative speed may be used instead of the ground speed.

[0138] [Variation 4]

[0139] In the above embodiment, the second motion model is selected when the longitudinal position of the target object is equal to or greater than the second determination value. However, the distance between the host vehicle and the target object may be used instead of the longitudinal position.

[0140] [Variation 5]

[0141] In the above embodiment, the second motion model is selected when the lateral position of the target object is smaller than the third determination value. However, the azimuth may be used instead of the lateral position.

[0142] [Variation 6]

[0143] In the above embodiment, the plurality of motion models include the first motion model and the second motion model, but the plurality of motion models may include more than three motion models. In addition, in the above embodiment, the lateral velocity vx and the longitudinal velocity vy of the state quantity of the target object in the first motion model are relative velocities. However, the lateral velocity vx and the longitudinal velocity vy may also be ground velocities.

[0144] The tracking device 4 and the method thereof described in the present invention may also be implemented by a dedicated computer provided by a processor and a memory programmed to execute one or more functions embodied by a computer program. Alternatively, the tracking device 4 and the method thereof described in the present invention may also be implemented by a dedicated computer provided by a processor composed of one or more dedicated hardware logic circuits. Alternatively, the tracking device 4 and the method thereof described in the present invention may also be implemented by one or more dedicated computers composed of a combination of a processor and a memory programmed to execute one or more functions and a processor composed of one or more hardware logic circuits. In addition, a computer program may also be stored as an instruction executed by a computer in a non-migratable tangible recording medium that can be read by a computer. In the method for realizing the functions of each part included in the tracking device 4, it is not necessarily necessary to include software, and all of its functions may also be realized using one or more hardware.

[0145] Multiple functions of one component in the above-mentioned embodiment can be realized by multiple components, or one function of one component can be realized by multiple components. In addition, multiple functions of multiple components can be realized by one component, or one function realized by multiple components can be realized by one component. In addition, part of the structure of the above-mentioned embodiment can also be omitted. In addition, at least part of the structure of the above-mentioned embodiment can also be added to or replaced with the structure of other above-mentioned embodiments.

[0146] In addition to the above-mentioned tracking device 4, the present invention can also be implemented in various ways such as a system with the tracking device 4 as a component, a program for making a computer function as the tracking device 4, a non-migrating physical recording medium such as a semiconductor memory that records the program, and a tracking method.

Claims

1. A tracking device, mounted on a vehicle, for tracking one or more objects existing around the vehicle, the tracking device comprising: a state quantity estimating unit configured to estimate the current state quantity of each of the one or more objects based on at least one of observation information obtained by observing the one or more objects by a sensor and past state quantities of the one or more objects, each time a predetermined processing cycle is repeated; a model selection unit configured to select, for each of the one or more objects, a motion model from at least a first motion model and a second motion model that are preset based on the state of the one or more objects, wherein the first motion model is a linear motion model and the second motion model is a turning motion model; as well as an estimation selection unit configured to cause the state quantity estimation unit to estimate the state quantity using the one motion model selected by the model selection unit for each of the one or more target objects; The model selection unit sets a parameter that is preset to represent the state of the one or more target objects as a selection parameter, and selects the one motion model by judging whether the selection parameter is equal to or greater than a preset selection determination value, wherein the selection parameter is at least one of a ground speed, a relative speed, a longitudinal position, a distance, a lateral position, an azimuth, and a reflection intensity of the one or more target objects. The model selection unit selects the second motion model in any one of the following cases: When the ground speed or relative speed of the one or more target objects is the selection parameter, the selection parameter is greater than the selection determination value; When the longitudinal position or distance of the one or more object marks is the selection parameter, the selection parameter is greater than the selection determination value; When the lateral position or orientation of the one or more objects is the selection parameter, the selection parameter is less than the selection determination value; and When the reflection intensity of the one or more target objects is the selection parameter, the selection parameter is equal to or greater than the selection determination value.

2. The tracking device according to claim 1, wherein: The model selection unit selects the second motion model when a preset second motion model selection condition is satisfied, and selects the first motion model when the second motion model selection condition is not satisfied.

3. The tracking device according to claim 1, wherein: The state variables of the second motion model include acceleration.

4. The tracking device according to any one of claims 1 to 3, wherein: The state quantity estimation unit further calculates a predicted value of the current state quantity of each of the one or more target objects based on the estimated value of the state quantity of the one or more target objects in the past, The above-mentioned model selection unit also selects a motion model different from the above-mentioned motion model selected last time when the set passed cycle is greater than a preset switching judgment cycle and the prediction residual is greater than a preset switching judgment value, the above-mentioned set passed cycle is the number of the above-mentioned processing cycles passed since the setting of the current one of the above-mentioned motion models, and the above-mentioned prediction residual is the difference between the above-mentioned prediction value and the observation value represented by the above-mentioned observation information.

5. The tracking device according to claim 1, wherein: The selection determination value is changed according to the selected one of the motion models.

6. The tracking device according to any one of claims 1 to 3, wherein: The state quantity estimation unit further estimates an error covariance of the state quantity.

7. The tracking device according to claim 6, wherein: A conversion unit is provided, and the conversion unit is configured to convert at least one of the state quantity and the error covariance to correspond to the changed one motion model for each of the one or more object targets when the one motion model is changed from the last time.

8. The tracking device according to any one of claims 1 to 3, wherein: A state quantity storage unit is provided, wherein the state quantity storage unit is configured to store the state quantity of the one or more objects regardless of the selected one motion model, The state quantity storage unit has a capacity capable of storing the state quantity in the one motion model having the largest number of state variables among at least the first motion model and the second motion model.

9. The tracking device according to claim 6, wherein: An output unit is provided, and the output unit is configured to convert the state quantity and the error covariance of the one or more target objects into a common physical quantity and output the converted physical quantity regardless of the selected one motion model.

10. The tracking device according to any one of claims 1 to 3, wherein: The model selection unit further sets a parameter pre-set to represent the state of the vehicle as the selection parameter, and selects the one motion model by determining whether the selection parameter is equal to or greater than the selection determination value, wherein the selection parameter is at least one of a rotation radius and an absolute value of acceleration of the vehicle, The model selection unit selects the second motion model in any one of the following cases: When the rotation radius of the vehicle is the selection parameter, the selection parameter is smaller than the selection determination value; and When the absolute value of the acceleration of the vehicle is the selection parameter, the selection parameter is equal to or greater than the selection determination value.

11. The tracking device according to claim 1, wherein: The selection parameter is a cycle, and the cycle is the number of the processing cycles that have passed since the one or more object markers are detected. The model selection unit selects the first motion model when the selection parameter is equal to or less than the selection determination value.

Images