JP2008234593A - Vehicle behavior predicting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a vehicle behavior predicting device for of accurately predicting the behavior of a vehicle in various scenes. <P>SOLUTION: This vehicle behavior predicting device 1 has a behavior learning part 14 learning an occurrence frequency of each possible phenomenon in the vehicle in response to a vehicle-situation, and a behavior predicting part 16 predicting the behavior of the vehicle from the vehicle-situation based on the learnt occurrence frequency. This behavior predicting device 1 can accurately predict the behavior of the vehicle, by using the occurrence frequency of its learnt state, when the vehicle to be predicted is placed in a certain state, by learning the occurrence frequency of each possible phenomenon in the vehicle in response to the vehicle-placed state. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a vehicle behavior prediction apparatus.

There is known an apparatus that monitors a traveling vehicle on a main line with a roadside device, predicts a lane change in the vicinity of a merging portion, and generates a traveling plan for smoothly joining an autonomous driving vehicle (for example, Patent Document 1). reference).
Japanese Patent No. 2969175

  However, actual vehicle behavior is complex, and it is difficult to predict behavior in various situations.

  The present invention has been made in view of the above-described circumstances, and an object thereof is to provide a vehicle behavior prediction apparatus capable of accurately predicting vehicle behavior in various scenes.

  According to the vehicle behavior prediction apparatus of the present invention, a learning unit that learns an occurrence frequency for each event that can occur in the vehicle according to a situation where the vehicle is placed, and the vehicle is placed based on the learned occurrence frequency. Behavior prediction means for predicting the behavior of the vehicle from the situation.

  In this behavior prediction device, since the occurrence frequency can be learned for each event that can occur in the vehicle according to the situation in which the vehicle is placed, when the vehicle to be predicted is placed in a situation, the learned situation By using the occurrence frequency, it is possible to accurately predict the behavior of the vehicle.

  The situation where the vehicle is placed may be specified by a running state of the vehicle and a surrounding condition of the vehicle. In this way, it is possible to suitably specify the situation where the vehicle is placed.

  The event may be specified by a behavior taken by the vehicle and / or a transition to a different running state. If it does in this way, the event which can occur in vehicles can be specified suitably for every situation where vehicles were put.

  The behavior prediction means may further predict future behavior from a situation in which the vehicle is placed after a predetermined interval time according to the predicted behavior. In this way, it is possible to predict more accurately even the earlier action.

  According to the present invention, it is possible to accurately predict the behavior of the vehicle in various scenes.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted.

  FIG. 1 is a block diagram showing a schematic configuration of the behavior prediction apparatus according to the embodiment of the present invention. FIG. 2 is a diagram showing a basic scene of behavior prediction by this behavior prediction device. As shown in FIG. 2, the behavior prediction apparatus 1 according to the present embodiment is mounted on a host vehicle M that travels on an automobile-only road such as an expressway, and the behavior of another vehicle X such as a manually driven vehicle or a non-communication vehicle. Is used to predict

  As shown in FIG. 1, the behavior prediction apparatus 1 includes a behavior prediction ECU (Electronic Control Unit) 10. The behavior prediction ECU 10 is connected to a peripheral sensor 20 such as a radar and a camera, a navigation device 30 for acquiring road information, and a communication device 40 for performing vehicle-to-vehicle communication, road-to-vehicle communication, and the like. In addition, a travel control plan generation ECU 50 is connected to the behavior prediction ECU 10. The travel control plan generation ECU 50 generates a travel control plan for the host vehicle M, and automatically controls the host vehicle M based on the plan, or proposes the plan to the driver.

  The behavior prediction ECU 10 includes a desired vehicle speed estimation unit 12, a behavior learning unit (learning unit) 14, and a behavior prediction unit (behavior prediction unit) 16.

  The desired vehicle speed estimation unit 12 learns in advance the desired vehicle speed of the specific vehicle according to the road condition, and estimates the desired vehicle speed according to the road condition of the specific vehicle using the learning result. The learning procedure of the desired vehicle speed of the specific vehicle by the desired vehicle speed estimation unit 12 is performed as follows.

  As shown in FIG. 3, first, an assumed traveling speed pattern of the learning target vehicle for each road condition is generated (step S301). As road conditions, whether the road is a straight road or a curved road, or a curved road, the road is divided for each curve R (curve diameter).

  In the generation of the assumed traveling speed pattern, first, an assumed traveling speed serving as a reference is calculated. For example, three types of personal preference (for example, an average speed calculated from a travel history) and a speed limit + α (for example, +0 km / h, +10 km / h as α) are prepared. Next, the assumed traveling speed in the overtaking lane is added. For example, 3 × 2 = 6 types of the above-described three assumed traveling speeds + β (for example, β is +0 km / h, +20 km / h) are prepared. Furthermore, an assumed traveling speed that takes into account the degree of deceleration according to the curve R is added. For example, 6 × 2 = 12 patterns are prepared by adding the upper limit lateral acceleration G to the above-described six assumed traveling speeds (for example, there is no limit as the upper limit lateral G and 0.2 G).

  FIG. 4 shows the assumed traveling speed pattern generated in this way. As shown in FIG. 4, twelve assumed traveling speed patterns are generated for each road condition. For example, in the case of the curve 300R, No. 1 is an average speed calculated from the travel history, and No. 2 is a speed limit. And No. 3 is obtained by adding +10 km / h to No. 2. No.4-6 is the result of adding + 20km / h to No.1-3 considering the overtaking lane. No.7-12 was calculated by considering the curve R and taking into account deceleration by the maximum lateral G to No.1-6. In the case of a straight line, No. 1-6 is entered at No. 7-12.

  When the assumed travel speed pattern of the learning target vehicle for each road condition is thus generated, the actual speed Vr of the learning target vehicle is acquired (step S302). This may be detected by the peripheral sensor 20, or may be acquired from the infrastructure equipment or other peripheral vehicles via the communication device 40.

  Next, the curve R is acquired as the road condition at that time (step S303). This may be detected by the peripheral sensor 20, may be acquired from the navigation device 30, or may be acquired via the communication device 40 from an infrastructure facility or the like.

  Next, the 12 speeds of the assumed traveling speed pattern in the road condition are compared with the actual speed Vr (step S304). For example, the difference between the 12 speeds and the actual speed Vr is taken. Then, the learning correlation value at a speed close to the actual speed Vr is increased (step S305), while the learning correlation value at a speed away from the actual speed Vr is decreased (step S306). Here, the speed learning correlation value indicates which speed pattern of the speed of the learning target vehicle has a high correlation among the 12 speed patterns, and is set to 50 at the beginning of learning, for example. ing.

  By repeating this learning, the speed learning correlation value converges at a substantially constant ratio among the 12 speed patterns. In this way, learning ends when the speed learning correlation value converges at a substantially constant rate. For example, in the example of FIG. 4, it is assumed that the correlation value of No. 1 is 20, the correlation value of No. 5 is 20, and the correlation value of No. 12 has converged to 90. This is a tendency to run at your favorite speed (No.1) in any road conditions (20), run at the speed limit in the driving lane and speed up by 20 km / h in the overtaking lane, but do not decelerate on the curve There is a tendency to travel at a speed limit of +10 km / h on the driving lane, speed up by 20 km / h on the overtaking lane, and decelerate by 0.2 G on the curve (No. 12). 90.

  The desired vehicle speed estimation unit 12 estimates the actual desired vehicle speed of the vehicle using the learning result. For example, when the curve R is 500R as the road condition, the desired vehicle speed is estimated using the assumed traveling speed pattern and the speed learning correlation value of the curve 500R in FIG. Specifically, No. 1 speed 85 km / h correlated at a rate of 20, No. 5 speed 120 km / h correlated at a rate of 20, and No. 12 correlated at a rate of 90 Using the speed of 130 km / h, it can be estimated that (85 × 20 + 120 × 20 + 130 × 90) / (20 + 20 + 90) = 121.5.

  The behavior learning unit 14 learns the occurrence frequency for each event that can occur in the vehicle according to the situation where the vehicle is placed. As shown in FIG. 5, the situation where the vehicle is placed is specified by the traveling state of the vehicle and the surrounding conditions of the vehicle. The traveling state can be divided into, for example, a follow-up traveling state, a single traveling state, a left lane change state, and a right lane change state. Further, the peripheral conditions include the conditions shown in the condition column of FIG. 5 such as whether the preceding vehicle is faster or slower than the desired speed.

  On the other hand, an event that can occur in the vehicle is specified by a behavior that the vehicle takes and / or a transition to a different driving state. Examples of the behavior include acceleration, deceleration, overtaking flag ON, and overtaking flag OFF. In the table shown in FIG. 5, the “learning coefficient” indicates the occurrence frequency for each event that can occur in the vehicle according to the situation where the vehicle is placed, and the initial value (for example, Large 75%, medium 50%, small 25%). The “correction coefficient” is used to correct the learning coefficient according to the situation in which the vehicle is placed when calculating the probability that the event will occur. For example, under conditions where the preceding vehicle is faster than the desired speed in the following traveling state (situation S1), the probability of transitioning to the single traveling state is calculated by multiplying the learning coefficient by the speed difference from the preceding vehicle. Further, under the condition where there is a merging vehicle in the following traveling state (situation S7), the probability of transition to the right lane change state is calculated by multiplying the learning coefficient by the interference probability with the merging vehicle. The interference probability means the probability that the vehicle will collide with a merging vehicle if the vehicle travels as it is. On the other hand, in the condition where there is a following vehicle in a single traveling state (situation S12), the probability of acceleration is calculated by dividing the learning coefficient by the inter-vehicle time with the following vehicle. In addition, an event for which the “learning coefficient” is indispensable is always an event that can occur, and no “correction coefficient” is set. The speed difference, the interference probability, and the inter-vehicle time as the “correction coefficient” are normalized.

  The vehicle behavior learning procedure by the behavior learning unit 14 is performed as follows. That is, referring to the flowchart of FIG. 6, first, an occurrence counter indicating the number of times each situation S1-32 has occurred and the total occurrence of the situation actually occurred are initialized to 0 (step S601). Further, the learning coefficient is set to an initial value (for example, large 75%, medium 50%, small 25%) (step S602).

  Next, as described above, the desired vehicle speed estimation unit 12 estimates the desired traveling speed of the learning target vehicle from the speed learning correlation value and the road condition at that time (step S603).

  Next, the presence of the surrounding vehicle and the traveling speed of the learning target vehicle are acquired from information such as the surrounding sensor 20 (step S604). Further, information on road lane reduction and curve R is acquired from information on the navigation device 30 and the communication device 40 (step S605).

  Next, it is determined whether or not the situation where the learning target vehicle is placed satisfies the conditions and conditions shown in FIG. 5 (in any situation S1-32) (step S606). If the state and condition are not satisfied, the process returns to step S603. On the other hand, when the state and the condition are satisfied, it is determined whether or not the learning coefficient in the situation is essential (step S607). And when it is indispensable, it returns to Step S603. On the other hand, if it is not essential, the status occurrence counter is incremented by 1 (step S608).

  Next, in that situation, it is determined whether or not the learning target vehicle has executed the behavior or transition (step S609), and if the learning target vehicle has executed the behavior or transition and the event has occurred, correction is performed. The reciprocal of the coefficient is added to the total occurrence of the situation (step S610), and the process proceeds to step S611. On the other hand, if the learning target vehicle does not execute the behavior or transition and the event does not occur, the process proceeds to step S611 without adding the reciprocal of the correction coefficient to the total occurrence of the situation.

  In step S611, it is determined whether the occurrence counter is greater than or equal to a threshold value (for example, 10). If it is equal to or greater than the threshold value, a learning coefficient is obtained by dividing the total occurrence by the occurrence counter (step S612). On the other hand, if it is smaller than the threshold value, the learning coefficient is determined using the total of the occurrences divided by the occurrence counter and the initial value of the learning coefficient (for example, the average value of both) (step S613).

  The behavior prediction unit 16 predicts the behavior of the prediction target vehicle from the situation where the vehicle is placed based on the learning coefficient learned and stored in the behavior learning unit 14. The vehicle behavior prediction by the behavior prediction unit 16 is performed as follows. That is, as shown in FIG. 7, first, as an initialization, the vehicle condition parameter is set in accordance with the current vehicle condition (step S701). The vehicle situation parameter is a parameter related to the prediction target vehicle, and includes an occurrence probability of a certain event, a situation where the vehicle is placed, a position, a speed, an acceleration, a yaw angle, and the like. Here, the occurrence probability is set to 100% as an initial value as shown in FIG. 8 (step S702).

  Next, as described above, the desired vehicle speed estimation unit 12 estimates the desired travel speed of the prediction target vehicle from the speed learning correlation value and the road condition at that time (step S703).

  Next, the presence of the surrounding vehicle and the traveling speed of the prediction target vehicle are acquired from information such as the surrounding sensor 20 (step S704). Further, information on road lane reduction and curve R is acquired from information of the navigation device 30 and the communication device 40 (step S705).

  Next, it is determined whether or not the situation where the prediction target vehicle is placed satisfies the conditions and conditions shown in FIG. 5 (in any situation S1-32) (step S706). If the state and condition are not satisfied, the process proceeds to step S713. On the other hand, when the condition and the condition are satisfied, it is determined whether or not the learning coefficient in the situation is essential (step S707). If it is essential, the behavior or transition is executed and the event is predicted to occur (step S708). On the other hand, if not essential, the occurrence coefficient of the situation is multiplied by the correction coefficient, and the probability that the action or transition is executed and the event occurs is calculated (step S709).

  Next, the current vehicle status parameters are copied, and two vehicle status parameters A and B are generated (step S710). Next, the vehicle value A is multiplied by the probability that the event will occur to obtain a multiplication value A '(step S711). Further, the vehicle condition parameter B is multiplied by (100% −probability of the event) to obtain a multiplication value B ′ (step S712).

  Next, by using these multiplication values A ′ and B ′, vehicle status parameters after a predetermined interval time (for example, 50 ms) are generated for cases where the event occurs and cases where the event does not occur (step S713). Next, parameter generation processing after a specified interval time is performed for all vehicle status parameters (step S714). If generation processing is not completed, steps S703 to S713 are repeated. On the other hand, if the generation is completed for all parameters, as shown in FIG. 8, the future behavior is predicted from the situation where the prediction target vehicle is placed after a predetermined interval time, and the event occurs or does not occur Vehicle status parameters are generated for and. That is, step S703 to step S714 are repeated. Then, it is determined whether or not the behavior prediction and the generation of the vehicle condition parameters are completed until a predetermined prediction time (for example, 10 seconds) indicating the upper limit of the time to be predicted (step S715). In this way, it is possible to predict the behavior of the prediction target vehicle and predict various parameters such as the existence position, existence probability, speed, acceleration, and yaw angle for each specified interval time.

  As described above, in the behavior prediction apparatus 1 according to the present embodiment, since the occurrence frequency can be learned for each event that can occur in the vehicle according to the situation where the vehicle is placed, when the prediction target vehicle is placed in a situation. The behavior of the vehicle can be accurately predicted by using the learned occurrence frequency of the situation.

  In addition, the behavior prediction unit 16 can further predict future behavior from the situation in which the vehicle is placed after a predetermined interval time according to the predicted behavior, so that it can more accurately predict further behavior. .

  The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the gist of the present invention.

It is a block diagram which shows schematic structure of the action prediction apparatus which concerns on embodiment. It is a figure which shows the basic scene of the action prediction by an action prediction apparatus. It is a flowchart which shows the procedure which learns the desired vehicle speed of a vehicle. It is a figure which shows an assumption driving speed pattern. It is a figure which shows the occurrence frequency for every event which can occur in the said vehicle according to the condition where the vehicle was put. It is a flowchart which shows the procedure which learns the action of a vehicle. It is a flowchart which shows the procedure of the action prediction of a vehicle. It is a figure which shows the specific example of action prediction.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Action prediction apparatus, 10 ... Action prediction ECU, 12 ... Desired vehicle speed estimation part, 14 ... Action learning part (behavior learning means), 16 ... Action prediction part (behavior prediction means), 20 ... Peripheral sensor, 30 ... Navigation apparatus 40 ... communication device, 50 ... running control plan generation ECU.

Claims (4)

Learning means for learning the frequency of occurrence of each event that can occur in the vehicle according to the situation where the vehicle is placed;
Based on the learned occurrence frequency, behavior predicting means for predicting the behavior of the vehicle from the situation where the vehicle is placed;
A vehicle behavior prediction apparatus comprising:   The vehicle behavior prediction apparatus according to claim 1, wherein the situation in which the vehicle is placed is specified by a traveling state of the vehicle and a surrounding condition of the vehicle.   The vehicle behavior prediction device according to claim 2, wherein the event is specified by a behavior taken by the vehicle and / or a transition to a different running state.   The vehicle behavior prediction according to any one of claims 1 to 3, wherein the behavior prediction means further predicts a future behavior from a situation in which the vehicle is placed after a predetermined interval time according to the predicted behavior. apparatus.

Images