JP5482323B2 - Driving support device and program - Google Patents

Description

  The present invention relates to a driving support apparatus and program, and more particularly, to a driving support apparatus and program for determining the possibility of a collision with another traveling body existing around the host vehicle.

  Conventionally, the possibility of a collision between the host vehicle and another vehicle, a two-wheeled vehicle, a bicycle, a pedestrian, or the like existing around the host vehicle is determined, and a warning is given to the driver based on the determination result. Controlling vehicle movement has been done.

  For example, a risk recognition system that adaptively recognizes information related to the degree of risk (risk) included in the environment from the detection result of the external environment of a moving body such as an automobile has been proposed (see Patent Document 1). In the risk recognition system of Patent Document 1, the driver is a teacher in learning of the recognizer, the risk information is extracted from the driving operation of the driver, and the relationship between the risk information based on the driving operation and the image information obtained from the camera is obtained. Learning is done using artificial intelligence technology. For example, when the driver performs an operation action that avoids a pedestrian, the situation is judged to be dangerous, and it is learned that the image obtained at that time is dangerous. Then, when a similar image is obtained next, it is output that it is dangerous, and a warning is given to the driver.

JP 2009-96365 A

  However, in the invention described in Patent Document 1, the risk level of the situation ahead of the host vehicle is estimated using the image feature amount. However, it is difficult to define the dangerous state using only the image feature amount. In some cases, the situation ahead and the dangerous state are associated with each other in an uncertain relationship. Therefore, there is a problem that complicated processing such as adjustment is required when using the learning result with accuracy for performing driving support. In determining the possibility of a collision, it is important to predict the behavior of another traveling body. However, in the invention described in Patent Document 1, the behavior of another traveling body from the viewpoint of another traveling body. Is not taken into consideration, and there is a problem that the determination accuracy is lowered.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a driving support device and a program that can perform highly accurate collision determination in consideration of the situation of other traveling body viewpoints. To do.

In order to achieve the above object, the driving support device of the present invention includes an acquisition unit that acquires a traveling environment including a traveling state of the host vehicle and a traveling state of another traveling body existing around the host vehicle, and the acquiring unit. Based on the acquired traveling state and traveling environment of the host vehicle, the host vehicle vector indicating the current state of the host vehicle and the state of another traveling body based on the host vehicle is calculated, and the other traveling A vector calculation means for calculating a running body vector indicating a current situation of the body and a situation of the own vehicle with reference to the other running body; a host vehicle vector calculated by the vector calculation means; and a plurality of host vehicle vectors predetermined for each and the free-vehicle vector running state and running environment of the vehicle used in the calculation of the learned and each change of behavior state of the vehicle after a predetermined time of the acquired time as the learning data Based on the vehicle change model for predicting the change in the behavior condition of the vehicle after between, as well as predict changes in behavior state of the vehicle after a predetermined time, travel calculated by the vector calculating unit and body vector, behavioral changes states of a plurality of run Gyotai vectors each with said traveling body vector vehicle running state and running environment other traveling body after the predetermined time of the acquired time used for calculation of the other based behavioral changes state of the running body on the traveling body change model to predict, other action state of the traveling body after the predetermined time of each said predetermined time after the learning as the learning data Predicting means for predicting the change of the vehicle, the traveling state and traveling environment of the host vehicle acquired by the acquiring unit, the change in the behavior state of the host vehicle after the predetermined time predicted by the predicting unit, and the predetermined A relative distance between the host vehicle and the other traveling body after the lapse of the predetermined time is estimated based on a change in the behavior state of another traveling body after a while, and the own traveling distance is estimated based on the estimated relative distance. Determining means for determining the possibility of a collision between the vehicle and the other traveling body.

  According to the driving support apparatus of the present invention, the acquisition unit acquires the traveling environment including the traveling state of the host vehicle and the traveling states of other traveling bodies existing around the host vehicle, and the vector calculating unit acquires the traveling unit. Based on the traveling state and traveling environment of the subject vehicle, the present vehicle vector indicating the current situation of the subject vehicle and the situation of the other traveling body based on the own vehicle is calculated, and the current state of the other traveling body is calculated. A traveling body vector indicating the situation and the situation of the host vehicle with respect to another traveling body is calculated. By calculating the traveling body vector in this way, it is possible to consider the behavior of other traveling bodies from other traveling body viewpoints.

Then, the prediction unit determines the time when the host vehicle vector calculated by the vector calculation unit, each of the plurality of host vehicle vectors, and the travel state and the travel environment of the host vehicle used to calculate the host vehicle vector are acquired. Based on the own vehicle change model for predicting the change in the behavior state of the own vehicle after the predetermined time learned as the learning data , each change in the behavior state of the own vehicle after the time, with predicting the change in the behavior condition of the vehicle, a traveling body vector calculated by the vector calculating means, a plurality of run Gyotai vector running state and running environment of each and of the vehicle used in the calculation of the running body vector of travel for but to predict changes in the other action state of the traveling body after the predetermined learned and each of the other action state of the traveling body after a predetermined acquired time time change as the learning data time Based on the change model to predict changes in the action state of the other traveling body after a predetermined time. That is, for other traveling bodies, it is possible to predict a change in behavior state using traveling body vectors from other traveling body viewpoints.

  Then, the determination means determines the traveling state and traveling environment of the own vehicle acquired by the acquiring means, the change in the behavior state of the own vehicle after a predetermined time predicted by the predicting means, and the behavior of another traveling body after the predetermined time. Based on the change in state, the relative distance between the host vehicle and another traveling body after a predetermined time has elapsed is estimated, and the possibility of a collision between the host vehicle and another traveling body is determined based on the estimated relative distance. To do.

  In this way, not only the vehicle vector indicating the situation of the own vehicle and the situation of the other traveling body based on the own vehicle, but also the situation of the other vehicle and the situation of the own vehicle based on the other traveling body. Since the traveling body vector to be shown is also calculated and the change in the behavior state of the other traveling body after a predetermined time is predicted, it is possible to perform the collision determination with high accuracy in consideration of the situation of the other traveling body viewpoint.

  The vector calculation means includes at least the position, speed, acceleration, and orientation of the host vehicle as the situation of the host vehicle, and at least the position, speed, and other position of the other running body as the situation of the other running body. The vehicle vector and the traveling body vector including the direction can be calculated. In addition, the acquisition unit acquires a shape of a traveling path where the host vehicle and the other traveling body exist as a driving environment, and the vector calculation unit further determines a predetermined traveling direction of the host vehicle as the situation of the host vehicle. The own vehicle vector and the traveling body vector including the radius of curvature of the traveling road ahead of the distance and further including the curvature radius of the traveling road ahead of the predetermined distance in the traveling direction of the other traveling body are calculated as the state of the other traveling body. To be able to. Thus, the accuracy of the collision determination can be improved by calculating the host vehicle vector and the traveling body vector using various information.

  The vector calculating means selects a predetermined number of other traveling bodies when there are a plurality of other traveling bodies existing around the own vehicle, and determines the current situation of the own vehicle and the own vehicle. Calculate the own vehicle vector indicating the status of each of the other selected traveling bodies as a reference, and for each of the selected other traveling bodies, the current status of the selected other traveling body and the other It is possible to calculate a traveling body vector indicating the situation of the host vehicle with respect to the traveling body of the vehicle and each of the other selected traveling bodies. In that case, the vector calculation means can use the own vehicle vector and the elements of the traveling body vector in order of increasing distance from the own vehicle and each of the selected other traveling bodies. Thus, even in a complicated situation where there are a plurality of other traveling bodies, the host vehicle vector and the traveling body vector can be appropriately calculated.

  Further, the host vehicle change model is determined at a rate for each change in action state in each region obtained by dividing the vector space onto which the host vehicle vector is projected, and the running body change model is defined as the running body. The vector space onto which the vector is projected can be determined at a rate corresponding to the change in the behavior state in each of the divided areas. By using such a host vehicle change model and a traveling body change model, it is possible to flexibly predict changes in the behavior state of the host vehicle and other traveling bodies.

  Further, the predicting means predicts a rate of change in behavior state of the host vehicle after the predetermined time, and predicts a rate of change in behavior state of other traveling bodies after the predetermined time. The means estimates the relative distance for each combination of a change in the behavior state of the host vehicle and a change in the behavior state of another traveling body, and the host vehicle in a combination in which the relative distance is smaller than a predetermined threshold distance The possibility of a collision between the host vehicle and the other traveling body is determined based on an occurrence ratio obtained by multiplying the ratio according to the change in behavior state of the vehicle and the ratio according to the change in behavior state of the other traveling body. Can be.

  Further, the change in the behavior state can be acceleration, constant speed, and deceleration.

  Further, the driving support device of the present invention is based on the traveling state and traveling environment history of the host vehicle acquired by the acquiring unit, and changes in the behavior state of the host vehicle after the predetermined time, and after the predetermined time. Change calculation means for calculating a change in the behavior state of another traveling body, the own vehicle vector calculated by the vector calculation means, and a change in the behavior state of the own vehicle after the predetermined time calculated by the change calculation means The vehicle change model is updated based on the correspondence between the vehicle and the behavior vector of the other vehicle after the predetermined time calculated by the vector calculator and the change calculator. Updating means for updating the traveling body change model based on the correspondence with the change. As described above, by updating the own vehicle change model and the traveling body change model based on the actually acquired data, individual differences between the vehicle and the driver can be absorbed.

  The other traveling bodies include other vehicles, two-wheeled vehicles, bicycles, and pedestrians.

The driving support program according to the present invention includes a computer that acquires a traveling environment including a traveling state of the host vehicle and a traveling state of another traveling body existing around the host vehicle, and the driver acquired by the acquiring unit. Based on the traveling state and the traveling environment of the vehicle, the present vehicle vector indicating the current state of the own vehicle and the state of the other traveling body based on the own vehicle is calculated, and the current state of the other traveling body is calculated. A vector calculation means for calculating a running body vector indicating the situation and the situation of the own vehicle with reference to the other running body, the own vehicle vector calculated by the vector calculation means, each of a plurality of own vehicle vectors and the own vehicle vector ; the predetermined time that has learned and each change of behavior state of the vehicle after a predetermined time at the time the running state and running environment of the vehicle is acquired using the calculated vehicle vector as the learning data Based on the vehicle change model for predicting the change in the behavior condition of the vehicle after, as well as predict changes in behavior state of the vehicle after a predetermined time, the running body calculated by the vector calculating unit vector and, of a plurality of run Gyotai vectors each with said traveling body vector calculated vehicle running state and running environment of the behavioral changes the state of the other traveling member after the predetermined time of the acquired time using the Based on a traveling body change model for predicting a change in the behavior state of another traveling body after the predetermined time learned as learning data, each of the behavioral states of the other traveling bodies after the predetermined time Predicting means for predicting a change, traveling state and traveling environment of the own vehicle acquired by the acquiring means, change in behavior state of the own vehicle after the predetermined time predicted by the predicting means, and the predetermined A relative distance between the host vehicle and the other traveling body after the lapse of the predetermined time is estimated based on a change in the behavior state of another traveling body after a while, and the own traveling distance is estimated based on the estimated relative distance. It is a program for functioning as a determination means for determining the possibility of a collision between a vehicle and the other traveling body.

  The storage medium for storing the program of the present invention is not particularly limited, and may be a hard disk or a ROM. Further, it may be a CD-ROM, a DVD disk, a magneto-optical disk or an IC card. Furthermore, the program may be downloaded from a server or the like connected to the network.

  As described above, according to the driving support apparatus and program of the present invention, not only the own vehicle vector indicating the situation of the own vehicle and the situation of the other running body based on the own vehicle, but also the situation of the other running body. In addition, it also calculates the traveling body vector indicating the situation of the host vehicle with reference to the other traveling body, and predicts the change in the behavior state of the other traveling body after a predetermined time. The effect that the highly accurate collision determination can be performed is obtained.

It is a block diagram which shows schematic structure of the driving assistance device of this Embodiment. It is an image figure which shows an example of a scene. It is a figure which shows an example of the own vehicle vector and traveling body vector which were memorize | stored in the scene vector memory | storage part, and action state change. It is the schematic for demonstrating the production | generation of the own vehicle change model and a traveling body change model. It is a figure which shows an example of the own vehicle change model and a traveling body change model. It is a figure for demonstrating an example of the calculation method of the acceleration at the time of estimating the relative distance of the own vehicle and another traveling body. It is a figure which shows the combination of the behavior state change of the predicted own vehicle, and the behavior state change of another traveling body. It is a figure which shows an example of the prediction result of the relative distance of the own vehicle and another traveling body. It is a flowchart which shows the content of the collision determination processing routine in the driving assistance device of this Embodiment. It is a flowchart which shows the content of the learning process routine in the driving assistance device of this Embodiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  As shown in FIG. 1, the driving support apparatus 10 according to the present embodiment irradiates the front of the host vehicle while scanning the laser in one dimension (horizontal direction), and the laser is irradiated by the reflection of the laser. A laser radar 12 that detects the two-dimensional position of an object, a CCD camera that images the periphery of the vehicle, and the like, an imaging device 14 that outputs image data of the captured image, and a satellite signal from a GPS satellite; Vehicle including GPS device 16 that outputs GPS data such as pseudorange data, a vehicle speed sensor that detects the vehicle speed of the host vehicle, an acceleration sensor that detects the acceleration of the host vehicle, and a yaw angle sensor that detects the yaw angle of the host vehicle A collision between the vehicle 18 and another traveling body with the sensor 18, an alarm device 22 including a display device for displaying an alarm according to the determination result, a speaker for outputting an alarm sound, and the like. Includes a computer 30 that executes a process of determining potential, a.

  The computer 30 temporarily stores a ROM that stores a program for executing a driving support processing routine including a CPU that controls the entire driving support device 10, a collision determination processing routine and a learning processing routine, which will be described later, data, and the like. RAM, a memory storing various information, and a bus connecting them.

  If the computer 30 is described by functional blocks divided for each function realizing means determined based on hardware and software, as shown in FIG. 1, the measurement data measured by the laser radar 12 and the imaging device 14 capture the image. Depending on the scene (situation) based on the image data, the GPS data output from the GPS device 16, the detection value of the vehicle sensor 18, and the map information stored in the electronic map storage device 20 connected via the network The environment recognition unit 32 for calculating the own vehicle vector and the traveling body vector, and the behavior state change after a predetermined time of the own vehicle and the other traveling body, the own vehicle vector, the behavior state change of the own vehicle, and the traveling body vector, A scene vector storage unit 34 in which the behavior state changes of other traveling bodies are stored correspondingly and recorded in the scene vector storage unit 34. A learning unit 36 that learns the generated information and generates a change model for predicting changes in the behavioral state of the host vehicle and other traveling bodies; and a change model storage unit 38 that stores the generated change model; Based on the behavior prediction unit 40 for predicting the behavior of the host vehicle and the other traveling body based on the host vehicle vector, the traveling body vector, and the change model, and based on the predicted behavior of the host vehicle and the other traveling body, A collision determination unit 42 that determines the possibility of a collision between the host vehicle and another traveling body can be used. The environment recognition unit 32 is an example of an acquisition unit, a vector calculation unit, and a change calculation unit of the present invention, and the learning unit 36 is an example of an update unit of the present invention.

  The environment recognition unit 32 acquires GPS data from the GPS device 16 and calculates the position coordinates (S_x, S_y) of the host vehicle. In addition, the detection values of the vehicle speed sensor, the acceleration sensor, and the yaw angle sensor of the vehicle sensor 18 are acquired, and an XY coordinate system is assumed in which the traveling direction of the host vehicle is the X axis and the direction orthogonal to the X axis is the Y axis. The speed (S_vx, S_vy) of the host vehicle, the acceleration (S_ax, S_ay) of the host vehicle, and the direction (S_dir) of the host vehicle are calculated. The speed of the host vehicle may be calculated based on the Doppler frequency of the GPS data, or may be calculated by differentiating the time-series position coordinates of the host vehicle. Further, the acceleration of the host vehicle may be calculated by differentiating the time-series speed of the host vehicle.

  Further, the environment recognition unit 32 acquires the image data output from the imaging device 14 and the measurement data of the laser radar 12, and based on the result of image recognition of the image data and the result of object detection by the laser radar 12, Other traveling bodies existing around the host vehicle are extracted to generate another traveling body list. For example, in the scene shown in FIG. 2, traveling body A and traveling body B are extracted.

  Moreover, the environment recognition part 32 can be represented by the structure further including the movable object identification part 32a and the runway identification part 32b.

  The movable object identifying unit 32a recognizes the acquired image data and identifies the attribute c (another vehicle, a two-wheeled vehicle, a bicycle, a pedestrian, etc.) of the traveling body included in another traveling body list. For example, in the example of FIG. 2, the attribute of the traveling object A is identified as “pedestrian” and the attribute of the traveling object B is identified as “other vehicle”. Further, the direction (O_dir) of each of the other traveling bodies is calculated from the result of the image recognition. The direction dir is an angle (degree) in which the positive direction of x is 0 deg, the positive direction of y is 90 deg, and the negative direction of x is 180 deg.

  In addition, when describing about the direction of the traveling body A, it is set as "OA_dir", and when describing about the direction of the traveling body B, it is set as "OB_dir." Hereinafter, when notifying each other traveling body without distinguishing each of the other traveling bodies, it is referred to as “O_”, and when describing information on a specific other traveling body, “OA_”, “OB_” "And so on.

  Further, the movable object identification unit 32a calculates the relative distance between the host vehicle and each of the other traveling bodies based on the acquired measurement data of the laser radar 12, and calculates the differential value of the relative distance to the other traveling with respect to the host vehicle. The relative speed of each body is calculated, and the differential value of the relative speed is calculated as the relative acceleration of each of the other traveling bodies with respect to the host vehicle. Based on the position coordinates (S_x, S_y) of the own vehicle, the speed (S_vx, S_vy) of the own vehicle, the acceleration (S_ax, S_ay) of the own vehicle, the relative distance, the relative speed, and the relative acceleration, another traveling body Position coordinates (O_x, O_y), speeds (O_vx, O_vy) of other traveling bodies, and accelerations (O_ax, O_ay) of other traveling bodies are calculated.

  The travel path identification unit 32b acquires map information from the electronic map storage device 20, collates the map information with GPS data, recognizes the nearest lane or sidewalk that can be traveled, and the own vehicle and other traveling bodies. Identify the center line of the runway of Then, a radius of curvature (S_WR) ahead of a predetermined distance (S_WL) when the center line of the traveling path of the host vehicle is approximated by a clothoid curve is calculated. Similarly, a curvature radius (O_WR) ahead of a predetermined distance (O_WL) on the traveling direction traveling path of another traveling body is calculated. The predetermined distance WL is assumed to be a Z-distance according to the attribute in consideration of predicting the behavior of the host vehicle and other traveling bodies after Z seconds (for example, after 2 to 3 seconds). A value is determined in advance. For example, if the attribute is the host vehicle or another vehicle, it can be determined as WL = 50 m.

  In addition, the environment recognition unit 32 selects N other traveling bodies from the other traveling bodies included in the other traveling body list in order of increasing distance from the host vehicle. In addition, with respect to the other selected traveling body, a traveling body for which data already exists with reference to past data is assigned the same identification number as the identification number unique to the traveling body already assigned. A new identification number is assigned to a newly appearing traveling body.

Moreover, the environment recognition unit 32, the vehicle and for each of the traveling body past data exists, the value of the current absolute value of the rate of synthesis of x-direction velocity vx and the y-direction of the velocity vy V _T and Z seconds ago A behavior state change indicating a change in the behavior state is calculated from an increase / decrease in the value V _TZ . In the present embodiment, the behavior state changes are acceleration, constant speed, and deceleration. _{Specifically,} <accelerated if _{ε, V T-Z -V T} > V T-Z -V T deceleration if epsilon, calculates the behavior state changes more as a constant velocity. For example, ε can be set to 1 m / s.

  The environment recognizing unit 32 also calculates the attribute c, the direction dir, the position (x, y), the speed (vx, vy), the acceleration (ax, ax) calculated for each of the host vehicle and the selected N other traveling bodies. ay) Using the distance WL and the radius of curvature WR, the vehicle vector and the traveling body vector in the current scene are generated. For example, the own vehicle vector in the scene Q as shown in FIG.

Scene Q vehicle vector
= (S_c, S_x, S_y, S_vx, S_vy, S_ax, S_ay, S_dir, S_WR, S_WL,
OA_c, OA_x, OA_y, OA_vx, OA_vy, OA_ax, OA_ay, OA_dir, OA_WR, OA_WL,
(OB_c, OB_x, OB_y, OB_vx, OB_vy, OB_ax, OB_ay, OB_dir, OB_WR, OB_WL)
... (1)

  Here, the own vehicle vector is the order of the own vehicle, the traveling body having the closest distance to the own vehicle, the traveling body having the next closest distance to the own vehicle,... A value indicating each situation is a vector element. Moreover, the traveling body vector of the other traveling body can be generated by using the same value as that of the host vehicle vector and changing the order of the elements in the order in which the distance from the other traveling body is closer. The traveling body vector of the traveling body A in the scene Q is shown in the following equation (2).

Traveling body vector of traveling body A in scene Q
= (OA_c, OA_x, OA_y, OA_vx, OA_vy, OA_ax, OA_ay, OA_dir, OA_WR, OA_WL,
OB_c, OB_x, OB_y, OB_vx, OB_vy, OB_ax, OB_ay, OB_dir, OB_WR, OB_WL,
(S_c, S_x, S_y, S_vx, S_vy, S_ax, S_ay, S_dir, S_WR, S_WL)
... (2)

  By generating the host vehicle vector and the traveling body vector in the scene Q as described above, the host vehicle vector can capture the situation of the scene Q with reference to the host vehicle. That is, the situation of the scene Q from the viewpoint of another traveling body can be captured.

  The scene vector storage unit 34 stores the generated own vehicle vector, the traveling body vector, and the calculated action state change. When storing, the own vehicle vector and the traveling body vector, and the behavior state change calculated at a time Z seconds after the time when the own vehicle vector and the traveling body vector were generated (the time when the data was acquired) Remember to correspond.

  FIG. 3 shows an example of the host vehicle vector and the traveling body vector stored in the scene vector storage unit 34, and the behavior state change. “ID” in the figure is a unique identification number assigned to the host vehicle and other traveling bodies, and the host vehicle is fixed at ID = 0. In the behavioral state change in the figure, acceleration is indicated by “1”, constant speed is indicated by “0”, and deceleration is indicated by “−1”. In addition, elements indicating the status of other traveling bodies in the host vehicle vector are “O1_” to “ON_” in the order of the closest distance to the host vehicle for the selected N other traveling bodies. Similarly, the elements indicating the situation of the host vehicle and the other traveling bodies in the traveling body vector are the reference for (N-1) other traveling bodies and the own vehicle excluding other reference traveling bodies. “O1_” to “ON_” are set in order of increasing distance from other traveling bodies. In the traveling body vector, the reference traveling body is “OS_”. Further, the data acquisition time interval dt can be set to 0.1 seconds, for example.

  The learning unit 36 can be expressed by a configuration including a host vehicle scene classification unit 36a and a traveling body scene classification unit 36b.

  The own vehicle scene classification unit 36a learns the own vehicle vector stored in the scene vector storage unit 34 and the corresponding action state change stored as learning data, and predicts the action state change of the own vehicle. A vehicle change model is generated. Specifically, as shown in FIG. 4, an action corresponding to the projected own vehicle vector is made for each block obtained by discretizing each axis of the vector space on which the own vehicle vector is projected and dividing the vector space into predetermined regions. The number of data for each state change (acceleration: AC, constant speed: CR, deceleration: DC) and the total number M thereof are calculated, and the ratio for each action state change (acceleration rate: P_AC, constant speed rate: P_CR, deceleration) Ratio: P_DC). In order to cope with the case where the corresponding data does not exist in the scene vector storage unit 34, the initial values are set such that CR = N (N> 0, for example, N = 1) and AC = DC = 0. be able to. The ratio for each action state change for each block using the discretization standard as a header is stored in the change model storage unit 38 as the own vehicle change model.

As another method, the following three groups are learned by logistic regression using the host vehicle vector and the traveling body vector.
i: A model F1 that binarizes AC and the rest (1 for AC, 0 otherwise) and regresses this
ii: Model F2 that binarizes DC and other values (1 for DC, 0 for other) and regresses this
iii: A model F3 that binarizes CR and other values (1 for CR, 0 for other values) and regresses this

Then, using the virtual input vector generated by discretizing each axis of the vector space using F1, F2, and F3, the values P_F1, P_F2, and P_F3 of F1, F2, and F3 are calculated for each corresponding block. Then, the ratios P_AC, P_CR, P_DC of AC, CR, and DC can be calculated as follows.
P_AC = P_F1 / (P_F1 + P_F2 + P_F3)
P_CR = P_F2 / (P_F1 + P_F2 + P_F3)
P_DC = P_F3 / (P_F1 + P_F2 + P_F3)

  The traveling body scene classification unit 36b generates a traveling body change model using the traveling body vector by the same method as the own vehicle scene classification unit 36a, and stores the generated traveling body change model in the change model storage unit 38. FIG. 5 shows an example of the own vehicle change model and the traveling body change model.

  The behavior prediction unit 40 can be represented by a configuration that further includes a host vehicle behavior prediction unit 40a and a traveling body behavior prediction unit 40b.

  The own vehicle behavior prediction unit 40a collates the current own vehicle vector generated by the environment recognition unit 32 with the own vehicle change model stored in the change model storage unit 38, and the behavior of the own vehicle after Z seconds. Predict changes in state. For example, in the own vehicle change model shown in FIG. 5, when the own vehicle vector corresponds to block A, the action state of the own vehicle after Z seconds is 0% when “acceleration”, and “constant speed”. Is predicted to be 80% and “decelerated” is 20%.

  The traveling body behavior prediction unit 40b collates the traveling body vector with the traveling body change model by the same method as the own vehicle behavior prediction unit 40a, and predicts a change in the behavior state of another traveling body after Z seconds. . For example, in the traveling body change model shown in FIG. 5, when the traveling body vector corresponds to block C, the behavior state of another traveling body after Z seconds is 80% when “acceleration”, “etc.” “Speed” is predicted to be 20%, and “Deceleration” is predicted to be 0%.

The collision determination unit 42 determines the trajectories of the host vehicle and the other traveling bodies based on the current traveling state of the host vehicle and the other traveling bodies and the change in the behavior state after Z seconds predicted by the behavior prediction unit 40. calculate. For example, for the host vehicle, after the current speed and acceleration of the host vehicle continues for Z seconds, a predicted position at each time when a behavioral state change (for example, “acceleration”) occurs after Z seconds is calculated. The acceleration in the behavior state change is determined in advance, for example, as acceleration of acceleration 1 m / s ² and deceleration as deceleration of acceleration −1 m / s ² , for example. Further, in the case of another traveling body, acceleration in the behavior change is, for example, acceleration of acceleration 1 m / s ² , and deceleration is deceleration of acceleration −2 m / s ² , for example. It may be determined. Different values may be determined according to the attributes of the traveling body. Moreover, you may calculate the acceleration after Z second progress as follows.

Referring to FIG. 6, when S_vx + S_ax · s = 0, there exists s satisfying Z>s> 0 and S_vy + S_ay · s = 0, or when S_vy + S_ay · s = 0. , Z>s> 0, and S_vx + S_ax · s = 0, Z is calculated as s. From now to Z seconds S_x ′ = S_x + S_vx · Z + 0.5 · S_ax · Z ²
S_y ′ = S_y + S_vy · Z + 0.5 · S_ay · Z ²
S_vx ′ = S_vx + S_ax · Z
S_vy ′ = S_vy + S_ay · Z

In the case of deceleration, the upper limit of t is [(S_vx ′) ² + (S_vx ′) ² ] ^0.5 ÷ (−α), and the acceleration after (Z + t) seconds is obtained as follows.
S_x ″ (t) = S_x ′ + S_vx ′ · t + 0.5 · azx · t ²
S_y ″ (t) = S_y ′ + S_vy ′ · t + 0.5 · azy · t ²

  In addition, the collision determination unit 42 determines a behavior state change in which the predicted behavior state change rate is other than 0% based on the trajectory of the own vehicle and the other traveling body calculated for each predicted behavior state change. For all combinations, the relative distance between the host vehicle and another traveling body is calculated. In the case of the example in FIG. 5, as shown in FIG. 7, (1) When the behavior state change of the own vehicle is “constant speed” and the behavior state change of the other traveling body is “acceleration”, (2 ) When the behavior state change of the own vehicle is “deceleration” and the behavior state change of the other traveling body is “acceleration”, (3) The behavior state change of the own vehicle is “constant speed”, and the behavior state change of the other traveling body Is “constant speed”, (4) there are four combinations in which the state change of the host vehicle is “deceleration” and the state change of other traveling bodies is “constant speed”. An example of the relative distance between the host vehicle and another traveling body for each of the combinations (1) to (4) is shown in FIG.

  If there is a combination in which the calculated relative distance becomes smaller than the threshold value dcr after Z seconds have elapsed from the current time, the ratio of the combination according to the change in the behavior state of the own vehicle and the other The rate of occurrence of the combination is calculated by multiplying the rate according to the behavior state change of the traveling body, and when the rate of occurrence is larger than a predetermined threshold value (for example, 30%), the host vehicle and the other running It is determined that there is a possibility of collision with the body, and the alarm device 22 performs control so that an alarm is output.

  Next, a collision determination processing routine executed by the computer 30 will be described with reference to FIG.

  In step 100, measurement data of the laser radar 12, image data output from the imaging device 14, GPS data output from the GPS device 16, and detection values detected by the vehicle sensor 18 are acquired.

  Next, in step 102, the position coordinates (S_x, S_y) of the host vehicle are calculated based on the GPS data. Further, based on the detection values of the vehicle speed sensor, the acceleration sensor, and the yaw angle sensor of the vehicle sensor 18, an XY coordinate system is assumed in which the traveling direction of the host vehicle is the X axis and the direction orthogonal to the X axis is the Y axis. The speed (S_vx, S_vy) of the host vehicle, the acceleration (S_ax, S_ay) of the host vehicle, and the direction (S_dir) of the host vehicle are calculated.

  Next, in step 104, the image data is image-recognized, an object around the own vehicle is detected from the measurement value of the laser radar 12, and the object is present around the own vehicle based on the result of the image recognition and the object detection result. The other traveling body to be extracted is extracted, and another traveling body list is generated.

  Next, in step 106, the image data is image-recognized to identify a traveling body attribute c (other vehicle, two-wheeled vehicle, bicycle, pedestrian, etc.) included in the other traveling body list. Further, the direction (O_dir) of each of the other traveling bodies is calculated from the result of the image recognition.

  Next, in step 108, the relative distance between the own vehicle and each of the other traveling bodies is calculated based on the measurement data of the laser radar 12, and the differential value of the relative distance is calculated for each of the other traveling bodies with respect to the own vehicle. The relative speed is calculated, and the differential value of the relative speed is calculated as the relative acceleration of each of the other traveling bodies with respect to the host vehicle.

  Next, in step 110, the position coordinates (S_x, S_y) of the own vehicle calculated in step 102, the speed (S_vx, S_vy) of the own vehicle, the acceleration (S_ax, S_ay) of the own vehicle, and calculated in step 108 above. Based on the relative distance, the relative speed, and the relative acceleration, the position coordinates (O_x, O_y) of the other traveling body, the speed (O_vx, O_vy) of the other traveling body, and the acceleration (O_ax, O_ay) is calculated.

  Next, in step 112, map information is acquired from the electronic map storage device 20, and then in step 114, the map information and GPS data are collated to recognize the nearest lane, sidewalk, etc. that can be run. Identifies the center line of the runway on which the host vehicle and other traveling bodies run.

  Next, in step 116, a curvature radius (S_WR) ahead of a predetermined distance (S_WL) when the center line of the traveling path of the vehicle identified in step 114 is approximated by a clothoid curve is calculated. Similarly, a curvature radius (O_WR) ahead of a predetermined distance (O_WL) on the traveling direction traveling path of another traveling body is calculated.

  Next, in step 118, N other traveling bodies are selected from the other traveling bodies included in the other traveling body list generated in step 104 in order of increasing distance from the host vehicle. In addition, with respect to the other selected traveling body, a traveling body for which data already exists with reference to past data is assigned the same identification number as the identification number unique to the traveling body already assigned. A new identification number is assigned to a newly appearing traveling body.

Next, in step 120, the vehicle and For each of the traveling body which historical data is present, the velocity vx and y directions of x direction of the resultant velocity of the velocity vy absolute value of the current value V _T and Z seconds before Based on the increase / decrease of the value V _T−Z , if V _T−Z −V _T <ε, acceleration is calculated; if V _T−Z −V _T > ε, deceleration is performed; To do.

  Next, in step 122, the attribute c, the direction dir, the position (x, y), the speed (vx, vy), the acceleration (ax, ay) calculated for each of the host vehicle and the selected N other traveling bodies. ), The distance WL and the radius of curvature WR, for example, the own vehicle vector in the current scene Q as shown in the above equation (1), and the traveling body in the current scene Q as shown in the above equation (2) Generate a vector. In the above equation (2), the traveling body vector of the traveling body A is shown, but the traveling body vector is similarly calculated for other traveling bodies. Then, for learning processing to be described later, each of the generated own vehicle vector and the traveling body vector, and Z seconds after the time when the own vehicle vector and the traveling body vector are generated (the time when the data is acquired) The behavioral state change calculated at the time is stored in the scene vector storage unit 34 in association with each other.

  Next, in step 124, the current own vehicle vector generated in step 122 and the own vehicle change model stored in the change model storage unit 38 are collated to change the behavior state of the own vehicle after Z seconds. Predict another percentage. Similarly, each of the traveling body vectors and the traveling body change model are collated, and the ratio of each traveling state of each other traveling body after Z seconds is predicted.

  Next, in step 126, based on the current running state of each of the host vehicle and the other traveling bodies, and the ratio of the behavior state change after Z seconds predicted in step 124, the predicted behavior state change is performed. The trajectory of each of the own vehicle and the other traveling body is calculated every time, and the own vehicle and each of the other traveling bodies are calculated with respect to all combinations of the behavior state changes in which the predicted ratio by behavior state change is other than 0%. The relative distance di (i is an identification number for each combination, (1) to (4) in the example of FIG. 7) is calculated.

  Next, in step 128, it is determined whether or not there is a combination i in which the relative distance di is smaller than the threshold value dcr. If it exists, the process proceeds to step 130, where the occurrence ratio is calculated by multiplying the ratio of the own vehicle of the combination i by the action state change and the ratio of the other traveling body by the action state change, and the occurrence ratio Is greater than a predetermined threshold. If the occurrence ratio is larger than the threshold value, the process proceeds to step 132, and it is controlled that the alarm device 22 outputs an alarm on the assumption that there is a possibility of collision between the host vehicle and another traveling body. finish.

  On the other hand, if it is determined in step 128 that d ≧ dcr, and if it is determined in step 130 that the occurrence rate is equal to or less than the threshold value, the processing is terminated.

  Next, a learning process routine executed by the computer 30 will be described with reference to FIG.

  In step 200, the host vehicle vector and the corresponding action state change stored from the scene vector storage unit 34 are read out as learning data.

  Next, in step 202, for each block in which each axis of the vector space on which the host vehicle vector is projected is discretized and the vector space is divided into predetermined regions, a behavior state change corresponding to the projected host vehicle vector (acceleration: AC, constant speed: CR, deceleration: DC) Calculate the number of data for each and the total number M, and rate by action state change (acceleration rate: P_AC, constant speed rate: P_CR, deceleration rate: P_DC) Is calculated.

  Next, in step 204, the ratio for each behavioral state change for each block using the discretization criterion as a header is stored in the change model storage unit 38 as the own vehicle change model.

  Next, in step 206, the running body vector and the corresponding action state change stored from the scene vector storage unit 34 are read out as learning data. Next, in steps 208 and 210, as in steps 202 and 204 above. Processing is performed to generate a traveling body change model, which is stored in the change model storage unit 38, and the process ends.

  This routine is executed every predetermined time, and the data newly stored in the scene vector storage unit 34 within the predetermined time is added to the number of data for each block at the time of the previous processing, so that the action state for each block Each of the own vehicle change model and the traveling body change model is updated by calculating the ratio for each change.

  As described above, according to the driving support device of the present embodiment, not only the own vehicle vector indicating the situation of the own vehicle and the situation of the other running body based on the own vehicle, but also the situation of the other running body. In addition, a traveling body vector indicating the situation of the host vehicle and the other traveling body with respect to the other traveling body is also calculated, and a change in the behavior state of the other traveling body after a predetermined time is also predicted. It is possible to perform collision determination with high accuracy in consideration of the situation of the traveling body viewpoint.

  In the above embodiment, as elements of the own vehicle vector and the traveling body vector, attributes, directions, positions, speeds, accelerations, a predetermined distance in the traveling direction, and a predetermined distance ahead in the traveling direction as the elements of the own vehicle and other traveling bodies. However, it is not necessary to include all of these as elements, and the vehicle vector includes at least the direction, position, speed, and acceleration, and the traveling body vector includes at least the direction and position. Anything that includes speed is acceptable.

  In the above embodiment, the order of the elements included in the host vehicle vector is a predetermined number (N) in the order of the distance from the host vehicle, and the order of the elements included in the traveling body vector is set to another traveling body. Although the case where the predetermined number (N) is set in order from the shortest distance to the vehicle has been described, the priority order based on the attributes of the own vehicle or other traveling body (for example, the order of the vehicle, the two-wheeled vehicle, the bicycle, the pedestrian) Or in order of increasing relative speed with the host vehicle or another traveling body. In the current scene, when the number of other traveling bodies is less than N, all corresponding values may be set to 0 or a predetermined value.

  In the above-described embodiment, the case where the own vehicle change model and the traveling body change model are determined at a rate for each behavior state change has been described. However, one behavior state for one own vehicle vector and one traveling body vector. It is good also as an own vehicle change model and a traveling body change model which predict a change.

  In the above-described embodiment, the case where the own vehicle change model and the running body change model are generated and updated using the actually detected situation of the own vehicle and the other running bodies has been described. The data detected in the other vehicle may be acquired, and this data may be used together to generate and update the own vehicle change model and the running body change model. It is also possible to use a host vehicle change model and a running body change model generated using data detected by the above method.

  In the above embodiment, the ratio of the change in the behavior state of the host vehicle and the ratio of the change in the behavior state of the other traveling body are multiplied by the predicted relative distance between the host vehicle and the other traveling body being smaller than the threshold. In the case of determining the possibility of collision based on whether or not the occurrence rate is larger than the threshold, if there is a combination whose relative distance is smaller than the threshold, it is determined that there is a possibility of collision, and The occurrence ratio may be output as a determination result.

  In the above-described embodiment, a case where a warning is output when it is determined that there is a possibility of a collision has been described. However, based on the predicted relative distance, vehicle motion is controlled so as to avoid a collision. You may do it. Moreover, you may make it display an alarm by performing display, such as enclosing another traveling body with the possibility of the collision on the image imaged with the imaging device with a window.

  Moreover, although the case where the laser radar was used was demonstrated in the said embodiment, you may use a millimeter wave radar etc.

  Moreover, although the case where map information is acquired from the outside has been described in the above embodiment, the map information may be stored in the computer of this apparatus.

DESCRIPTION OF SYMBOLS 10 Driving support device 12 Laser radar 14 Imaging device 16 GPS device 18 Vehicle sensor 20 Electronic map storage device 22 Alarm device 30 Computer 32 Environment recognition unit 32a Movable object identification unit 32b Runway identification unit 34 Scene vector storage unit 36 Learning unit 36a Own vehicle Scene classification unit 36b traveling body scene classification unit 38 change model storage unit 40 behavior prediction unit 40a own vehicle behavior prediction unit 40b traveling body behavior prediction unit 42 collision determination unit

Claims (12)

An acquisition means for acquiring a traveling environment including a traveling state of the host vehicle and a traveling state of another traveling body existing around the host vehicle;
Based on the traveling state and traveling environment of the host vehicle acquired by the acquiring unit, a host vehicle vector indicating a current state of the host vehicle and a state of another traveling body based on the host vehicle is calculated. A vector calculating means for calculating a running body vector indicating a current situation of the other running body and a situation of the host vehicle based on the other running body;
The own vehicle calculated by the vector calculation means, each of the plurality of own vehicle vectors, and the own vehicle after a predetermined time from the time when the running state and the running environment of the own vehicle used for calculating the own vehicle vector are acquired. Based on the vehicle change model for predicting a change in the behavior state of the host vehicle after the predetermined time learned using each of the changes in the behavior state as learning data, the behavior of the host vehicle after the predetermined time with predicting the change in state, the traveling body vector calculated by the vector calculating means, the vehicle traveling state and running environment used for calculating the respective and said traveling body vectors of a plurality of run Gyotai vector is obtained It travels for predicting the time the predetermined time other behavioral changes the state of the other traveling member after the predetermined time has learned and each change in the action state of the traveling body as the learning data after of Based on the change model, prediction means for predicting a change in the behavior condition of the other traveling member after the predetermined time,
The traveling state and traveling environment of the host vehicle acquired by the acquiring unit, the change in the behavior state of the host vehicle after the predetermined time predicted by the predicting unit, and the behavior state of another traveling body after the predetermined time. The relative distance between the host vehicle and the other traveling body after the lapse of the predetermined time is estimated based on the change in the vehicle, and the collision between the host vehicle and the other traveling body is based on the estimated relative distance. Determining means for determining the possibility of
A driving support device including   The vector calculation means includes at least the position, speed, acceleration, and orientation of the host vehicle as the situation of the host vehicle, and at least the position, speed, and orientation of the other running body as the situation of the other running body. The driving support device according to claim 1, wherein the vehicle vector and the traveling body vector included are calculated. The acquisition means acquires a shape of a traveling path where the host vehicle and the other traveling body exist as a traveling environment,
The vector calculating means further includes a radius of curvature of a traveling path ahead of a predetermined distance in the traveling direction of the host vehicle as the situation of the host vehicle, and further includes a predetermined direction of the traveling direction of the other traveling body as the situation of the other traveling body. The driving support device according to claim 2, wherein the host vehicle vector and the traveling body vector including a radius of curvature of a travel path ahead of a distance are calculated.   The vector calculation means selects a predetermined number of other traveling bodies when there are a plurality of other traveling bodies existing around the own vehicle, and uses the current situation of the own vehicle and the own vehicle as a reference. Calculating a host vehicle vector indicating a situation of each of the selected other traveling bodies, and, for each of the selected other traveling bodies, a current situation of the selected other traveling body and the other traveling The driving support device according to any one of claims 1 to 3, wherein a driving body vector indicating a situation of the host vehicle with respect to the body and a situation of each of the other selected traveling bodies is calculated.   The driving support device according to claim 4, wherein the vector calculation means uses the host vehicle vector and the traveling body vector as elements in order of increasing distance between the host vehicle and each of the selected other traveling bodies.   The own vehicle change model is determined by a ratio for each change in behavior state in each region obtained by dividing the vector space onto which the own vehicle vector is projected, and the running body change model is defined as The driving support device according to any one of claims 1 to 5, wherein the driving support device is defined by a ratio according to a change in a behavior state in each region obtained by dividing the projected vector space. The predicting means predicts a rate for each change in behavior state of the host vehicle after the predetermined time, and predicts a rate for change in the behavior state of another traveling body after the predetermined time,
The determination means estimates the relative distance for each combination of a change in the behavior state of the host vehicle and a change in the behavior state of another traveling body, and the combination of the relative distance is smaller than a predetermined threshold distance. The possibility of a collision between the host vehicle and the other traveling body is determined based on an occurrence ratio obtained by multiplying the ratio according to the change in the behavior state of the own vehicle and the ratio according to the change of the behavior state of the other traveling body. The driving support device according to claim 6 which judges.   The driving support device according to any one of claims 1 to 7, wherein the change in the behavior state is acceleration, constant velocity, and deceleration. Based on the travel state and travel environment history of the host vehicle acquired by the acquisition means, the change in the behavior state of the host vehicle after the predetermined time and the change in the behavior state of other traveling bodies after the predetermined time. Change calculating means for calculating;
Updating the host vehicle change model based on the correspondence between the host vehicle vector calculated by the vector calculator and the change in the behavior state of the host vehicle after the predetermined time calculated by the change calculator; The traveling body change model is updated based on the correspondence between the traveling body vector calculated by the vector calculating means and the change in behavior state of another traveling body after the predetermined time calculated by the change calculating means. Update means;
The driving support device according to claim 1, comprising:   The driving support device according to any one of claims 1 to 9, wherein the other traveling body includes another vehicle, a two-wheeled vehicle, a bicycle, and a pedestrian. Computer
An acquisition means for acquiring a traveling environment including a traveling state of the host vehicle and a traveling state of another traveling body existing around the host vehicle;
Based on the traveling state and traveling environment of the host vehicle acquired by the acquiring unit, a host vehicle vector indicating a current state of the host vehicle and a state of another traveling body based on the host vehicle is calculated. A vector calculating means for calculating a running body vector indicating a current situation of the other running body and a situation of the host vehicle based on the other running body;
The own vehicle calculated by the vector calculation means, each of the plurality of own vehicle vectors, and the own vehicle after a predetermined time from the time when the running state and the running environment of the own vehicle used for calculating the own vehicle vector are acquired. Based on the vehicle change model for predicting a change in the behavior state of the host vehicle after the predetermined time learned using each of the changes in the behavior state as learning data, the behavior of the host vehicle after the predetermined time with predicting the change in state, the traveling body vector calculated by the vector calculating means, the vehicle traveling state and running environment used for calculating the respective and said traveling body vectors of a plurality of run Gyotai vector is obtained travel for predicting the time the predetermined time other behavioral changes the state of the other traveling member after the predetermined time has learned and each change in the action state of the traveling body as the learning data after of Based on a change model, a predicting means for predicting a change in the behavioral state of another traveling body after the predetermined time, a traveling state and a traveling environment of the own vehicle acquired by the acquiring means, and the predicting means Based on the predicted change in the behavior state of the host vehicle after the predetermined time and the change in the behavior state of the other travel body after the predetermined time, the host vehicle and the other travel body after the lapse of the predetermined time A driving support program for causing the vehicle to function as a determination unit that determines the possibility of a collision between the host vehicle and the other traveling body based on the estimated relative distance.   The driving assistance program for functioning a computer as each means which comprises the driving assistance device of any one of Claims 1-10.

Images