JP6316146B2 - Moving object position prediction apparatus, moving object position prediction method, and moving object position prediction computer program - Google Patents

Description

  The present invention relates to a moving object position prediction apparatus, a moving object position prediction method, and a moving object position prediction computer program that predict the position of a moving object.

  In order to reduce contact accidents between the vehicle and the vehicle or between the vehicle and the person, by detecting the moving object around the vehicle and predicting the course of the moving object, the collision between the moving object and the vehicle is avoided. In order to do this, a technique for supporting the traveling of the vehicle has been proposed (see, for example, Patent Document 1 and Non-Patent Document 1).

  For example, the moving object surrounding risk determination method disclosed in Patent Document 1 probabilistically predicts a path that an obstacle can take, while generating a path of the moving object, and moving among the paths that the obstacle can take. The collision probability between the moving body and the obstacle is calculated by obtaining the path that interferes with the path of the body. And when this method determines the risk around the moving object, the latest collision probability between the moving object and the surrounding obstacle, and the amount of time change between the latest collision probability and the collision probability before a predetermined time. Make a determination using.

  In the method disclosed in Non-Patent Document 1, the planner predicts the path of a moving obstacle detected by a sensor such as a radar, and the vehicle and the obstacle do not collide based on the prediction result. In addition, an optimal moving speed of the vehicle at each time is set so that the vehicle traveling on the known route reaches the destination within a predetermined time.

JP 2008-158969 A

Jeff Johnson et al., “Optimal Longitudinal Control Planning with Moving Obstacles”, Intelligent Vehicles Symposium (IV) 2013 IEEE, IEEE, 2013

  In the above technique, it is required to accurately predict the course of the moving object in order to prevent a collision accident between the vehicle and the moving object. In addition, the prediction of the course of a moving obstacle may be expressed by a probability because it involves uncertainty. However, in the method disclosed in Non-Patent Document 1, since speed control that handles such a probability is not performed, it is difficult to appropriately handle the uncertainty of the course prediction.

  Therefore, an object of the present invention is to provide a moving object position prediction apparatus, a moving object position prediction method, and a moving object position prediction computer program that can improve the prediction accuracy of the course of a moving object.

According to the first aspect of the present invention, a moving object position prediction apparatus is provided as one aspect of the present invention. The moving object position prediction apparatus includes a storage unit (22) that stores trajectory information representing a trajectory of a past moving object for each of a plurality of past moving objects that have moved in the monitoring target area, A sensor (21) that generates a sensor signal representing an object existing in the object, a moving object detection unit (31) that detects a moving object that moves within the monitoring target area based on the sensor signal, and a storage unit (22). A selection unit (33) for selecting at least one trajectory information representing a trajectory similar to the assumed path of the moving object from the plurality of stored trajectory information, and each of the selected at least one trajectory information , Generating a probability distribution in which the probability that a moving object is present is higher at a position closer to the locus represented by the locus information, and according to a mixture distribution including each of the probability distributions at a predetermined time A prediction information generating unit for generating a prediction information indicating the probability that there is a moving object in Oite predetermined position and (34).
The moving object position prediction apparatus according to the present invention can improve the prediction accuracy of the path of the moving object by having the above configuration.

According to a second aspect of the present invention, the selection unit (33) includes, among the plurality of pieces of locus information, a corresponding locus including a point within a predetermined range from the current position of the moving object, and a past object at the point. It is preferable to select trajectory information in which the difference between the moving direction of the moving object and the moving direction of the moving object is within a predetermined direction range.
Thereby, the selection unit can more appropriately select trajectory information having a trajectory similar to the expected course of the moving object.

Furthermore, according to the description of claim 3, each of the plurality of pieces of trajectory information includes information indicating the type of the corresponding past moving object, and the moving object detection unit (31) identifies the type of the moving object, The selection unit (33) preferably selects trajectory information on a past moving object of the same type as the moving object type from among the plurality of trajectory information.
Thereby, the selection unit can more appropriately select trajectory information having a trajectory similar to the expected course of the moving object.

Furthermore, according to the description of claim 4, each of the plurality of pieces of trajectory information includes information indicating a time when a corresponding past moving object passes through each point on the trajectory represented by the trajectory information, and the selection unit (33) is that the difference between the time when the past moving object passes a point within the predetermined range from the current position of the moving object and the detection time of the current position of the moving object among the plurality of pieces of trajectory information is within the predetermined range. It is preferable to select trajectory information.
Thereby, the selection unit can more appropriately select trajectory information having a trajectory similar to the expected course of the moving object.

According to the fifth aspect of the present invention, it is preferable that the selection unit (33) selects trajectory information using a nearest neighbor search method.
As a result, the moving object position prediction apparatus can shorten the time required for searching for trajectory information used to predict the position of the moving object.

Furthermore, according to the description of claim 6, the prediction information generation unit (34) includes a probability distribution corresponding to the trajectory information in which the point included in the corresponding trajectory is close to the current position of the moving object among the selected trajectory information. It is preferable to generate the mixture distribution by weighting the probability distribution so that the weighting factor becomes larger as the weight coefficient increases.
Thereby, the moving object position prediction apparatus can further improve the prediction accuracy of the position of the moving object.

According to the seventh aspect of the present invention, a moving object position prediction method is provided as another aspect of the present invention. According to the moving object position prediction method, a step of detecting a moving object moving in the monitoring target area based on a sensor signal representing an object existing in the monitoring target area, and a monitoring stored in the storage unit (22) are provided. At least one trajectory information that represents a trajectory similar to the assumed path of the moving object is selected from the trajectory information that represents the trajectory of the past moving object for each of a plurality of past moving objects that have moved within the target area. For each of the selected step and the selected at least one trajectory information, a probability distribution in which the probability that a moving object is present at a position closer to the trajectory represented by the trajectory information is higher is generated, and a mixture including each of the probability distributions Generating prediction information representing a probability that a moving object is present at a predetermined position at a predetermined time according to the distribution.
The moving object position prediction method according to the present invention can improve the prediction accuracy of the path of the moving object by including the above steps.

According to the eighth aspect of the present invention, a computer program for predicting a moving object position is provided as still another aspect of the present invention. The computer program for predicting the moving object position is stored in the storage unit (22) and the step of detecting the moving object moving in the monitoring target area based on the sensor signal representing the object existing in the monitoring target area. , At least trajectory information that represents a trajectory similar to the assumed path of the moving object from among trajectory information that represents the trajectory of the past moving object for each of a plurality of past moving objects that have moved within the monitoring target area. For each of the selected step and at least one selected trajectory information, a probability distribution is generated in which the probability that a moving object is present at a position closer to the trajectory represented by the trajectory information is increased. Generating prediction information representing a probability that a moving object is present at a predetermined position at a predetermined time according to a mixed distribution including: Comprising instructions to be executed by the computer.
The computer program for predicting the moving object position according to the present invention can improve the prediction accuracy of the course of the moving object by having the above-described command.

  The reference numerals in parentheses attached to the above-described parts are examples that show the correspondence with specific means described in the embodiments described later.

It is a schematic block diagram of the automatic driving assistance system by which the moving object position prediction apparatus which concerns on one Embodiment of this invention was mounted. It is a schematic block diagram of the roadside machine which is an example of the moving object position prediction apparatus which an automatic driving assistance system has. It is a functional block diagram of the control part of a roadside machine. It is a figure which shows an example of the locus | trajectory shown by the moving object in tracking and locus | trajectory information. It is a schematic block diagram of the automatic driving device which an automatic driving assistance system has. It is a functional block diagram of the control part of an automatic driving device. It is a schematic diagram which shows the relationship between the presence probability of a moving object, and the driving pattern set. It is a schematic diagram of the response characteristic of the actual acceleration with respect to the target value of the acceleration of the vehicle in the first-order lag system. It is a figure which shows the integration range of existence probability. It is a figure which shows the operation | movement sequence of an automatic driving | operation process. When there is a moving object that crosses the planned route of the vehicle, the vehicle speed is controlled by a change in the speed of the vehicle controlled by the automatic driving support system according to the present embodiment and a detection signal from a sensor mounted on the vehicle. It is a graph showing the comparison with the speed change of a vehicle. In the same situation as FIG. 11, the vehicle position change controlled by the automatic driving support system according to the present embodiment is compared with the vehicle position change when the vehicle speed is controlled by the detection signal from the sensor mounted on the vehicle. It is a graph to represent. In the same situation as FIG. 11, when the vehicle speed is controlled by the change in the distance between the vehicle controlled by the automatic driving support system according to the present embodiment and the moving object crossing the planned route, and the detection signal from the sensor mounted on the vehicle. It is a graph showing the comparison with the distance change between a vehicle and a moving object.

Hereinafter, a moving object position prediction apparatus according to an embodiment will be described with reference to the drawings.
This moving object position prediction apparatus obtains, as a prediction result of the position of the moving object being tracked, the probability (existence probability) that the moving object being tracked exists at each future time for each position. At this time, the moving object position prediction apparatus stores trajectory information having a trajectory similar to the path of the moving object currently being tracked from a storage unit that stores trajectory information representing the trajectory of the moving object that has moved in the monitoring target area in the past. The path of the moving object being tracked is predicted such that the closer to the position on the trajectory represented by the selected trajectory information, the higher the existence probability.

  FIG. 1 is a schematic configuration diagram of an automatic driving support system in which a moving object position prediction apparatus according to one embodiment is implemented. The automatic driving support system 1 includes a roadside machine 2 that is an example of a moving object position prediction apparatus and an automatic driving apparatus 4 mounted on a vehicle 3. The roadside machine 2 is moved by a sensor signal from a sensor arranged so that the monitoring target area 100 is a passage where the vehicle can travel, such as a parking space in a highway service area or a parking space in a store, for example. The object 101 is detected. Then, the roadside machine 2 predicts the path 102 of the moving object for each detected moving object 101, and for each of a plurality of sections dividing the route 103 on which the vehicle is scheduled to travel, for each of a plurality of times. The prediction information indicating the probability that the moving object 101 exists in the section is obtained. On the other hand, when the vehicle 3 enters the communicable range of the roadside device 2, the automatic driving device 4 receives the prediction information from the roadside device 2, and the probability that the vehicle 3 collides with the moving object 101 based on the prediction information. The target acceleration of the vehicle 3 at each future time is determined so as to reduce the above. Then, the automatic driving device 4 notifies the ECU of the vehicle 3 of a control signal corresponding to the target acceleration.

  In the present embodiment, the vehicle 3 is assumed to travel along a predetermined route, and the automatic driving device 4 causes the ECU to adjust the moving speed of the vehicle 3 according to the target acceleration. The route may be included in the monitoring target area, or may be off the monitoring target area, such as adjacent to the monitoring target area. Further, the number of vehicles 3 equipped with the automatic driving device 4 included in the automatic driving support system 1 is not limited to one. A plurality of vehicles 3 equipped with the automatic driving device 4 may exist within the communication range of the roadside machine 2. The moving object is, for example, a person or a vehicle. In the following description, the speed means a moving speed unless otherwise specified.

  FIG. 2 is a schematic configuration diagram of a roadside machine 2 that is an example of a moving object position prediction apparatus. The roadside machine 2 includes a sensor 21, a storage unit 22, a control unit 23, a communication unit 24, and an antenna 25.

  The sensor 21 is, for example, a laser scanner, a radar sensor, or an image sensor, and generates a sensor signal representing an object existing in the monitoring target area. And the sensor 21 is arrange | positioned so that a detection range may overlap with a monitoring object area, for example above the pillar provided in either the center or four corners of a monitoring object area.

When the sensor 21 is a laser scanner or a radar sensor, the sensor 21 measures a distance from the sensor 21 to an object at every predetermined interval (for example, 1 °) at a predetermined period (for example, 30 msec to 100 msec). The distance measurement signal representing the measured value of the distance is output to the control unit 23 as a sensor signal.
Further, when the sensor 21 is an image sensor, the sensor 21 generates an image obtained by photographing the monitoring target area every predetermined cycle, and outputs the image to the control unit 23 as a sensor signal.
The roadside machine 2 may include a plurality of sensors 21. In this case, the roadside machine 2 may have the same type of sensors as the plurality of sensors 21, or may have different types of sensors.

  The storage unit 22 includes, for example, a nonvolatile semiconductor memory circuit and a volatile semiconductor memory circuit. The storage unit 22 may further include a magnetic storage device such as a hard disk or an optical storage device. And the memory | storage part 22 memorize | stores the various information utilized for the detection of the moving object which exists in the monitoring object area, or the prediction of a course. As such information, for example, the storage unit 22 stores a trajectory database including trajectory information representing the trajectories of a plurality of past moving objects that have passed through the monitoring target area. Further, the storage unit 22 stores information representing the locus of the moving object being tracked. Further, the storage unit 22 stores information representing the planned route of the vehicle 3.

  The control unit 23 has at least one processor and its peripheral circuits. And the control part 23 controls the roadside machine 2 whole. Further, the control unit 23 detects a moving object existing in the monitoring target area based on a signal from the sensor 21 and obtains prediction information. And the control part 23 transmits the prediction information etc. to the automatic driving apparatus 4 mounted in the vehicle 3 which exists in the communicable range of the roadside machine 2 via the communication part 24. FIG. Details of processing executed by the control unit 23 will be described later.

  When the communication unit 24 receives a signal including information (for example, prediction information) to be transmitted from the control unit 23 to the automatic driving device 4, the communication unit 24 performs transmission processing such as modulation and multiplexing on the signal according to a predetermined wireless communication standard. Run. Then, the communication unit 24 generates a radio signal by superimposing the signal subjected to the transmission process on a carrier wave having a radio frequency, and outputs the radio signal to the automatic driving device 4 via the antenna 25.

  In addition, the communication unit 24 separates and demodulates the radio signal received from the automatic driving device 4 and extracts information such as a prediction information request signal superimposed on the radio signal. Then, the communication unit 24 outputs the extracted information to the control unit 23.

  The predetermined wireless communication standard may be any communication standard that prescribes wireless communication between the roadside device and the communication device mounted on the vehicle. For example, ARIB STD-T75 or ARIB STD-T109. Can do.

  FIG. 3 is a functional block diagram of the control unit 23 of the roadside machine 2. The control unit 23 includes a moving object detection unit 31, a tracking unit 32, a selection unit 33, and a prediction information generation unit 34. Each of these units included in the control unit 23 is mounted as a functional module realized by a computer program executed on a processor included in the control unit 23, for example.

  Each time the moving object detection unit 31 receives a sensor signal from the sensor 21, the moving object detection unit 31 detects a moving object present in the monitoring target area based on the sensor signal.

  For example, when the sensor signal is a ranging signal for each azimuth, the moving object detection unit 31 is stored in advance in the storage unit 22 from the sensor 21 for each azimuth when there is no moving object in the monitoring target area. Reference distance information representing a reference distance that is a distance to a near object is read. Then, the moving object detection unit 31 calculates the difference between the distance from the sensor 21 to the object in the azimuth represented by the distance measurement signal and the reference distance for each azimuth. Then, the moving object detection unit 31 determines that there is a moving object at the position of the distance from the sensor 21 represented by the distance measurement signal with respect to an orientation in which the absolute value of the difference is equal to or greater than a predetermined threshold. Note that the moving object detection unit 31 has an absolute value of a difference between a plurality of adjacent azimuths equal to or greater than a predetermined reference value, and a difference in distance from the sensor 21 for each of the plurality of azimuths within a predetermined range. When it is included in (for example, 1 m), it may be determined that a moving object exists at the position of the distance from the sensor 21 represented by the distance measurement signal in the plurality of directions.

  When the sensor signal is an image of the monitoring target area, the moving object detection unit 31 detects the moving object by executing an image recognition process for recognizing the assumed moving object on the image. To do. In this case, as the image recognition process, the moving object detection unit 31 can use, for example, a multilayer perceptron learned using adaBoost, Real adaBoost, or Deep Learning. In this example, since the moving object detection unit 31 can also identify the type of moving object (for example, a person or a vehicle), the type of the moving object may be associated with the position where the moving object is detected.

  Each pixel on the image has a one-to-one correspondence with the direction from the sensor 21. Therefore, the moving object detection unit 31 determines the position of the moving object on the monitoring target area based on the position where the moving object is detected on the image, the position and shooting direction of the sensor 21, the assumed size of the moving object, and the like. Ask for.

  When the roadside device 2 has a plurality of sensors 21, the moving object detection unit 31 detects moving object candidates by the above-described method for each sensor signal from each sensor 21, and detects the detection based on each sensor signal. If the difference between the positions of the moving object candidates is within an allowable error range (for example, 1 m to 2 m), the moving object candidate may be detected as a moving object. Alternatively, the moving object detection unit 31 may use a moving object candidate detected for any one of the sensor signals as a moving object.

  The moving object detection unit 31 outputs the position where the moving object is detected to the tracking unit 32. When the type of the moving object is associated with the position where the moving object is detected, the moving object detection unit 31 also outputs the type to the tracking unit 32.

  The tracking unit 32 tracks the moving object based on the locus of the moving object that has already been detected and the current position of the moving object detected from the latest sensor signal. Therefore, the tracking unit 32 applies a prediction filter such as a particle filter or a Kalman filter, for example, and predicts the position of the moving object at the time of obtaining the latest sensor signal, which is predicted from the locus of the already detected moving object, The current position of the moving object detected from the latest sensor signal is compared. Then, when the tracking unit 32 includes the current position of the moving object detected from the latest sensor signal within a predetermined allowable range (for example, within 1 m) from the predicted position for any of the already detected moving objects. The moving object detected from the latest sensor signal is determined to be the same as the moving object corresponding to the predicted position. Then, the tracking unit 32 further applies a prediction filter to the past trajectory and the current position of the moving object, and obtains the predicted position of the moving object when the next sensor signal is acquired. Furthermore, the tracking unit 32 may also predict the speed of the moving object when the next sensor signal is acquired.

On the other hand, when the current position of the moving object detected from the latest sensor signal has already been detected and is not included within the predetermined allowable range from any predicted position of the moving object being tracked, the tracking unit 32 The moving object corresponding to the current position is set as a newly detected moving object, and tracking of the moving object is started. The tracking unit 32 may set the moving object as a newly detected moving object only when the current position of the moving object is included within a predetermined distance from the end of the monitoring target area.
The tracking unit 32 also applies a prediction filter to a newly detected moving object, and obtains a predicted position of the moving object when the next sensor signal is acquired.
The tracking unit 32 determines that a moving object whose change in position is a predetermined value or less over a certain period of time is an object that is stationary in the monitoring target area, and excludes it from the tracking target. Good.

  For each moving object, the tracking unit 32 includes, on the trajectory of the moving object, information indicating the current position of the moving object, the speed calculated from the distance between the previous position and the current position and the elapsed time, and the corresponding time. To update the trajectory of the moving object. Then, the tracking unit 32 stores, for each moving object, the locus of the moving object and the predicted position when the next sensor signal is acquired in the storage unit 22.

Further, the tracking unit 32 determines that the current position of the moving object detected from the latest sensor signal is not included in a predetermined allowable range from the predicted position of the moving object for a predetermined number of times. When it is continuous (for example, 2 to 5 times), it is determined that the moving object has moved out of the monitoring target area. Then, the tracking unit 32 stops tracking the moving object that has been determined to have moved out of the monitoring target area, and additionally registers the trajectory information representing the trajectory of the moving object in the trajectory database.
The trajectory information representing the trajectory of the moving object is represented by the following equation, for example.
Here, d ^j (k) represents the trajectory of the jth moving object registered. k (= 1, 2,..., K) represents the number of times of tracking from the start of tracking (ie, discrete time). q _x ^j (k) and q _y ^j (k) are the position in the x-axis direction and the position in the Cartesian coordinate system set in the monitored area at the time of the k-th tracking for the j-th registered moving object. Represents the position in the y-axis direction. Further, w _x ^j (k) and w _y ^j (k) represent the velocity in the x-axis direction and the velocity in the y-axis direction at the time of the k-th tracking for the jth registered moving object. T ^j (k) represents the elapsed time from the tracking start to the k-th tracking time for the jth registered moving object. When the tracking period is constant at T (for example, 100 msec), t ^j (k) = T (k−1). Further, T may be the same as the sensor signal acquisition cycle from the sensor 21 or may be different.
If the type of the moving object is associated with the detected position of the moving object, the tracking unit 32 may include the type of the moving object in the trajectory information.

  When the selection unit 33 receives the prediction information request signal from the automatic driving device 4, the selection unit 33 receives the signal indicating that the vehicle 3 has reached the moving destination from the automatic driving device 4 until the current position of the moving object currently being tracked. Each time the is updated, trajectory information representing the trajectory of the past moving object similar to the path of the moving object is selected from the trajectory database for each of the moving objects. The selected trajectory is used to generate prediction information about the moving object. Note that the selection unit 33 receives the prediction information request signal from the automatic driving device 4 until the period when the vehicle 3 on which the automatic driving device 4 is mounted is assumed to reach the moving destination passes. Alternatively, the trajectory may be selected for each moving object every time the current position of the moving object currently being tracked is updated.

For example, the selection unit 33, for each moving object, out of the trajectory information registered in the trajectory database, within a predetermined range (for example, 2 m) from the current position of the moving object, and the difference in moving direction is a predetermined direction range The trajectory information including the position within (for example, 45 °) is selected as the trajectory of the past moving object similar to the path of the moving object. For this selection, the selection unit 33 may use a nearest neighbor search method such as KD-tree, for example. Note that the moving direction of the moving object being tracked is estimated by obtaining a difference between the current position of the moving object and the previous position. Similarly, the moving direction of the moving object shown in the trajectory information at each time t ^j (k) is the position of the moving object at the time t ^j (k) and the previous tracking time t ^j (k-1). It is obtained from the difference in position of the moving object.

  FIG. 4 is a diagram illustrating an example of a moving object being tracked and a trajectory indicated by trajectory information. A point 400 is the current position of the moving object being tracked, and an arrow 401 represents the moving direction of the moving object. Lines 411 to 414 each represent a trajectory. In this example, since the trajectory 412 and the positions 412a and 413a on the trajectory 413 are included within 2 m from the current position of the moving object and the difference in the moving direction is within 45 °, the trajectory 412 and the trajectory 413 are Selected.

  Note that the selection unit 33 may include speed in the selection condition of the trajectory information. In this case, for each moving object that is being tracked, the selection unit 33 determines that the speed of the moving object at the current time and the current position of the moving object included in the trajectory information are within a predetermined range and the difference between the moving directions is predetermined. When the absolute value of the difference from the moving object speed indicated by the trajectory information at a position within the directional range is within 5 to 10 km / h, the trajectory information may be selected. Further, when each piece of trajectory information includes information indicating the type of the moving object, the selection unit 33 refers to that information, and the trajectory of the past moving object of the same type as the moving object being tracked is displayed. Trajectory information representing can be selected. As a result, even when trajectory information about different types of moving objects such as people and vehicles is registered in the trajectory database, the selection unit 33 displays the trajectories of past moving objects of the same type as the moving object being tracked. The trajectory information to be displayed can be selected.

  Alternatively, the selection unit 33 determines that the time difference between the time when the past moving object exists when the moving object is at the position closest to the current position of the moving object and the current time is a predetermined period (for example, 1 It may be added to the selection condition that it is within (time). Thereby, when the specific tendency is seen by the time slot | zone about the flow of the moving object which moves the monitoring object area, it becomes possible for the roadside machine 2 to predict the course of a moving object more correctly.

  Furthermore, the selection unit 33 has a predetermined range (for example, 1 m to 3 m) that is different from each position included in the locus in a predetermined period (for example, 1 sec to 2 sec) including the current time of the moving object being tracked. Trajectory information including a trajectory may be selected. Thereby, the selection unit 33 can more appropriately select trajectory information having a trajectory similar to the path of the moving object being tracked.

  In addition, when the prediction information acquisition request received from the automatic driving device 4 includes information on the current position of the vehicle 3 on which the automatic driving device 4 is mounted, the selection unit 33 The trajectory information may not be selected for a moving object within a predetermined error range (for example, 1 to 2 m) from the current position of the vehicle 3. Thereby, the selection part 33 can prevent producing | generating prediction information as one of the moving objects which are tracking the vehicle 3 itself.

  The selection unit 33 passes the selected trajectory information to the prediction information generation unit 34.

  Each time the prediction information generation unit 34 receives the selected trajectory information, the prediction information generation unit 34 generates prediction information based on the trajectory information. The prediction information is an example of a prediction result of the position of the moving object being tracked. In the present embodiment, the prediction information generation unit 34 divides a planned route, which is a route that the vehicle 3 is scheduled to pass after the current time, into predetermined sections (for example, 1 m), and moves for each section currently being tracked. The probability that any of the objects exists in the section is calculated for each time of a predetermined cycle (for example, 50 msec to 200 msec). At that time, the prediction information generation unit 34 is indicated by the trajectory information on the basis of the position of the past moving object represented by the selected trajectory information that is closest to the current position of the moving object being tracked. The closer to the position after the predetermined time elapses along the trajectory, the higher the existence probability of the moving object at the time after the predetermined time elapses.

Specifically, prediction information is generated as follows.
The position and speed of the vehicle 3 included in the prediction information request signal are represented by r ⁰ = (p _x ⁰ , p _y ⁰ , v _x ⁰ , v _y ⁰ ). Note that (p _x ⁰ , p _y ⁰ ) are the position of the vehicle 3 in the x-axis direction and the position of the vehicle 3 in the y-axis direction in the orthogonal coordinate system set in the monitoring target area, respectively (v _x ⁰ , v _y ⁰ ) are the speed of the vehicle 3 in the x-axis direction and the speed of the vehicle 3 in the y-axis direction in the orthogonal coordinate system set in the monitoring target area. Further, the position of each section on the planned route of the vehicle 3 is represented by (p _x ^u , p _y ^u ) (u = 1, 2,..., U). Information representing the planned route is read from the storage unit 22. The time for predicting the position of the moving object is ω _v (v = 1, 2,..., V). For example, when the position of a moving object is predicted from the current time to 20 seconds later at 200 msec intervals, ω ₁ = 0.2, ω ₂ = 0.4,..., Ω ₁₀₀ = 20.

selected set of trajectory information to the mobile object r ^m in the m-th track contained in lambda _m, in the j-th track information, a discrete time corresponding to the position closest to the current position of the moving object r ^m k _j . In this case, at time omega _v, past the moving object represented by the j-th trajectory information is moved to the position at discrete time _{^{h v j = (k j +}} floor (ω v / T)). The floor function floor (x) is a function that outputs an integer part of a real number x. T is the tracking period.
Therefore, assuming that the position of the moving object is distributed according to a Gaussian distribution centered on the position at the discrete time h _v ^j on the trajectory, the section on the planned route at time ω _v obtained from the selected trajectory information The existence probability p (p _x ^u , p _y ^u | ω _v ) of the moving object in u is represented by a mixed Gaussian distribution represented by the following equation.
Here, M represents the total number of selected trajectory information. Also, Omega except the vehicle 3 itself, represents a collection of moving objects r ^m currently tracking, m represents the number of moving objects included in the set. Μ _v ^j is the x-axis direction at the discrete time h _v ^j shown in the j-th trajectory information included in the set of trajectory information λ _m selected for the moving object r ^{m at} time ω _v And an average of the Gaussian distribution representing the position in the y-axis direction. Q is the covariance with a standard deviation of 1 m that represents the ambiguity of the position of the moving object. N ((p _x ^u , p _y ^u ) | μ _v ^j , Q) is a Gaussian distribution with respect to (p _x ^u , p _y ^u ) obtained based on the jth trajectory information.
Note that the prediction information generation unit 34 can obtain the existence probability of the moving object for each time according to the equation (2) for any position other than each section on the planned route.

Each time the prediction information generation unit 34 obtains the existence probability p (p _x ^u , p _y ^u | ω _v ) of the moving object in each section on the planned route, the existence probability p (p _x ^u , p _y ^u | ω _v ) is transmitted as prediction information to the automatic driving device 4 of the vehicle 3 via the communication unit 24.

Next, the automatic driving device 4 will be described.
FIG. 5 is a schematic configuration diagram of the automatic driving device 4. As shown in FIG. 5, the automatic driving device 4 includes an antenna 41, a communication unit 42, a position measurement unit 43, a storage unit 44, a control unit 45, and an in-vehicle communication unit 46.

  The communication unit 42 transmits and receives wireless signals between the roadside device 2 and the automatic driving device 4 in accordance with a predetermined wireless communication standard. For example, the communication unit 42 separates and demodulates the radio signal received from the roadside device 2 and extracts information superimposed on the radio signal. Then, the communication unit 42 outputs the extracted information to the control unit 45. For example, when the vehicle 3 enters the communicable range of the roadside machine 2, the communication unit 42 transmits a radio signal including notification information used for wireless communication with the roadside machine 2 from the roadside machine 2 via the antenna 41. Received information is extracted from the radio signal and passed to the controller 45. Furthermore, when the communication unit 42 receives a radio signal including the prediction information from the roadside device 2 via the antenna 41, the communication unit 42 extracts the prediction information from the radio signal and passes it to the control unit 45.

  When the communication unit 42 receives a signal (for example, a prediction information request signal) to be transmitted from the control unit 45 to the roadside device 2, the communication unit 42 performs transmission processing such as modulation and multiplexing on the signal. Then, the communication unit 42 generates a radio signal by superimposing the signal subjected to the transmission process on a carrier wave having a radio frequency, and outputs the radio signal to the roadside device 2 via the antenna 41.

  The predetermined wireless communication standard may be the same communication standard as the communication standard that the communication unit 24 of the roadside device 2 complies with, for example, ARIB STD-T75 or ARIB STD-T109.

  The position measuring unit 43 measures the current position of the vehicle 3. For this purpose, the position measurement unit 43 includes, for example, a receiver that receives a global positioning system (GPS) signal and an arithmetic circuit that calculates the current position of the vehicle 3 from the GPS signal. Then, the position measurement unit 43 measures the three current positions of the vehicle every predetermined cycle (for example, 100 msec) and notifies the control unit 45 of the measurement result.

  The storage unit 44 includes, for example, a nonvolatile semiconductor memory circuit and a volatile semiconductor memory circuit. The storage unit 44 stores information used for the automatic driving device 4 to generate control information for the vehicle 3. As such information, for example, the storage unit 44 receives the latest speed of the vehicle 3 received from the ECU of the vehicle 3, the current position of the vehicle 3 received from the position measurement unit 43, and the latest received from the roadside machine 2. Prediction information and the like are stored.

  The control unit 45 has at least one processor and its peripheral circuits. And the control part 45 controls the automatic driving device 4 whole. Further, the control unit 45 generates a prediction information request signal or the like, or obtains control information of the vehicle 3 based on the prediction information received from the roadside machine 2. Then, the control unit 45 transmits the control information and the like to the ECU of the vehicle 3 via the in-vehicle communication unit 46. Details of processing executed by the control unit 45 will be described later.

  The in-vehicle communication unit 46 is connected to the ECU of the vehicle 3 via an in-vehicle communication network (not shown) provided in the vehicle 3 such as a control area network. Then, the in-vehicle communication unit 46 outputs the vehicle control information received from the control unit 45 to the ECU.

Next, the automatic driving process executed by the control unit 45 will be described.
FIG. 6 is a functional block diagram of the control unit 45. The control unit 45 includes a request signal generation unit 51, a target acceleration setting unit 52, and a vehicle control information generation unit 53. Each unit included in the control unit 45 is implemented as a functional module realized by a computer program executed on a processor included in the control unit 45, for example.

  When the request signal generation unit 51 receives the notification information from the roadside device 2 via the antenna 41 and the communication unit 42, the request signal generation unit 51 determines that the vehicle 3 has entered the communicable range of the roadside device 2, and requests prediction information. An information request signal is generated. The request signal generation unit 51 is the latest speed of the vehicle 3 received from the ECU, the current position of the vehicle 3 received from the position measurement unit 43, the identification information of the automatic driving device 4, and the prediction information request signal. Is included in the prediction information request signal. Then, the request signal generation unit 51 outputs an expected information request signal to the communication unit 42.

  The target acceleration setting unit 52 is set for each predetermined period after the current time so as to optimize the traveling pattern for traveling on the planned route based on the latest prediction information received from the roadside device 2 via the communication unit 42. The target acceleration of the vehicle 3 at each time is set. The target acceleration setting unit 52 may set the target acceleration every time the prediction information is received, or may set the target acceleration only once when the prediction information is first received.

  FIG. 7 is a schematic diagram showing the relationship between the existence probability of a moving object and a set travel pattern. In FIG. 7, the horizontal axis represents the elapsed time from the setting of the travel pattern, and the vertical axis represents the position of the vehicle 3 along the planned route. In FIG. 7, the darker the color, the higher the existence probability of the moving object. In this example, it is assumed that the existence probability of the moving object in each section on the planned route is obtained based on three moving objects. Therefore, the existence probability of the moving object is high in the three areas 701 to 703. In this case, the target acceleration setting unit 52 optimizes the travel pattern 711 so that the position of the vehicle 3 does not overlap with any of the regions 701 to 703. That is, the target acceleration setting unit 52 sets the target acceleration at each time so as to reduce the probability that the vehicle 3 collides with a moving object.

  As described above, even if the target speed is set for the vehicle 3 due to the weight of the vehicle 3, engine response characteristics, and the like, the actual speed of the vehicle 3 after the control of the vehicle 3 is started is the target speed. It takes some time to reach speed. Therefore, in the present embodiment, the target acceleration setting unit 52 sets the target acceleration of the vehicle 3 at each time in order to improve the response characteristics of the vehicle 3 and allow the vehicle 3 to travel according to the optimal travel pattern. Set. Thereby, the difference between the actual behavior of the vehicle 3 and the set travel pattern is reduced.

For example, when a certain target acceleration is set for the vehicle 3, the response characteristics until the vehicle 3 actually reaches the target acceleration can be evaluated by, for example, a first-order lag system. FIG. 8 is a schematic diagram of response characteristics of actual acceleration with respect to a target value of vehicle acceleration in a first-order lag system. In FIG. 8, the horizontal axis represents time, and the vertical axis represents acceleration. For example, when the target acceleration indicated by the line 800 is given, the actual acceleration of the vehicle 3 asymptotically approaches the target acceleration as indicated by the line 801.
Considering this point, the target acceleration setting unit 52 sets the target acceleration by approximating the response characteristic of the vehicle 3 with a first-order lag system. Note that the approximate expression is determined in advance by experiments, for example. Further, by integrating the acceleration of the vehicle 3 represented by the first-order lag system corresponding to the target acceleration in the range from the current time to the setting target time, the target speed of the vehicle 3 at the setting target time is obtained, The position of the vehicle 3 at the setting target time is obtained by integrating the target speed in the range from the current time to the setting target time.

  In the present embodiment, the target acceleration setting unit 52 obtains a cost for each predetermined period over the set period of the travel pattern from the current time, and calculates the sum of the costs as an evaluation function. Note that the travel pattern setting period is set to, for example, an estimated value of the period required for the vehicle 3 to reach the moving destination from the current location, for example, 5 sec to 20 sec. Further, the target acceleration setting unit 52 obtains a cost for each predetermined period until reaching a moving destination (for example, a predetermined parking space or an end point of a predetermined passage) from the current position of the vehicle 3 along the planned route. The total cost may be calculated as an evaluation function. The target acceleration setting unit 52 sets the target acceleration at each time so that the evaluation function is minimized.

In the present embodiment, the target acceleration setting unit 52 calculates the following four types of costs.
(1) Cost1: Contact probability between a moving object and the vehicle 3 In order to avoid a contact accident, it is preferable that the contact probability between the moving object and the vehicle 3 is low. Therefore, the target acceleration setting unit 52 calculates, as Cost1, the total sum of the contact probabilities between the moving object and the vehicle 3 during the travel pattern setting period. Cost1 is calculated according to the following equation.
Here, tp is a travel pattern setting period. The probability (x, t) is the existence probability of the moving object at the position x (t) and the time t, and the existence probability at the corresponding position and time included in the prediction information is input.

  FIG. 9 is a diagram illustrating an integration range of existence probabilities. Each rectangular area 900 to 903 shown in FIG. 9 represents an integration range of existence probabilities at each time. In this example, “5” shown in the integral range of the existence probability in the expression (3) represents that the length of the offset set in front of the vehicle 3 is 5 m. L represents the length of the vehicle 3 in the traveling direction.

(2) Cost2: Absolute value of the target acceleration The comfort of the passenger of the vehicle 3 increases as the vehicle 3 moves at a constant speed. Conversely, the greater the acceleration or deceleration of the vehicle 3, the lower the comfort. Therefore, the target acceleration setting unit 52 calculates the sum of absolute values of the target acceleration of the vehicle 3 at each time during the travel pattern setting period as Cost2. Cost2 is calculated according to the following equation, for example.
Here, a _desi is the acceleration of the vehicle 3 at time i. In addition, when a _desi is negative, it represents that the vehicle 3 is decelerated.

(3) Cost3: Fluctuation amount of the target acceleration For the occupant of the vehicle 3, the more constant the acceleration of the vehicle 3, the higher the comfort. Conversely, the more the acceleration of the vehicle 3 changes, the lower the comfort. Therefore, the target acceleration setting unit 52 calculates, as Cost3, the sum of absolute values of the differences between the target acceleration at each time of the vehicle 3 and the target acceleration at the immediately preceding time during the travel pattern setting period. Cost3 is calculated according to the following equation, for example.

(4) Cost4: Amount of deviation from the reference speed The smaller the amount of deviation of the vehicle 3 from the reference speed, the more the vehicle 3 can travel on the scheduled route in advance. Therefore, the target acceleration setting unit 52 calculates, as Cost4, the sum of absolute values of the differences between the target speed and the reference speed at each time of the vehicle 3 during the travel pattern setting period. Cost4 is calculated according to the following equation, for example.
Here, V _set is a reference speed, and is _set to a speed according to the monitoring target area, for example, 10 km / h to 50 km / h. V (t) is the target speed of the vehicle 3 at time t, and is calculated by integrating the target acceleration from the current time to time t.

The target acceleration setting unit 52 uses the total cost of Cost1 to Cost4 (= Cost1 + Cost2 + Cost3 + Cost4) as an evaluation function. Then, the target acceleration setting unit 52 calculates the evaluation function while changing the target acceleration at each time within the range where the speed of the vehicle 3 satisfies the constraint condition, and determines the minimum value of the evaluation function. The speed constraint condition is, for example, that the speed v of the vehicle 3 at each time is 0 km / h or more and a speed upper limit value V _limit km / h (for example, 20 km / h to 60 km / h) or less. is there. For example, the target acceleration setting unit 52 calculates the evaluation function by selecting the target acceleration at each time from 1.0, 0.5, 0.0, -0.5, -1.0, -1.5, and -2.0 m / s ² . At that time, the target acceleration setting unit 52 may obtain the minimum value of the evaluation function by calculating the evaluation function for all combinations of the target acceleration at each time, or optimization such as steepest descent method or simulated annealing. A minimum value of the evaluation function may be obtained by applying a technique.

  The target acceleration setting unit 52 may use a value obtained by multiplying each of Cost1 to Cost4 by a weighting coefficient and adding the result as an evaluation function. In this case, for example, it is preferable to make the weighting coefficient for Cost1 the heaviest. As a result, a travel pattern in which the possibility that the vehicle 3 and the moving object collide is further reduced can be obtained. Alternatively, the target acceleration setting unit 52 may use the sum of one or two of Cost2 to Cost4 and Cost1 as an evaluation function. Thereby, the calculation amount of the control part 45 regarding the setting of a running pattern is reduced.

  The target acceleration setting unit 52 outputs the target acceleration at each time when the evaluation function becomes the minimum value to the vehicle control information generation unit 53.

  The vehicle control information generation unit 53 generates, for each time, vehicle torque that can achieve the target acceleration at that time as vehicle control information to be output to the ECU.

  Theoretically, when the vehicle 3 is accelerated at the target acceleration, the vehicle 3 should reach the target speed obtained by integrating the target acceleration at each time. However, in reality, an error occurs between the actual speed of the vehicle 3 and the target speed due to air resistance or rolling resistance. Therefore, the vehicle control information generation unit 53 receives information representing the actual speed of the vehicle 3 from the ECU or a vehicle speed sensor mounted on the vehicle 3, and performs feedback control on the acceleration applied to the vehicle 3 based on the information. Correct the acceleration. Thereby, the vehicle control information generation part 53 can reduce the difference between the actual speed of the vehicle 3 and the target speed.

For example, the vehicle control information generation unit 53 calculates a command acceleration obtained by correcting the target acceleration according to the following equation at each time when the target acceleration is set.
Here, a _des is the target acceleration set by the target acceleration setting unit 52 at the time of interest, and a _cmd is the commanded acceleration at the time of interest. Further, v _des is the target speed at the time of interest, and v is the latest measured value of the actual speed of the vehicle 3. K is a proportional constant set in advance, and is set to 0.1 to 1, for example.

The vehicle control information generation unit 53 calculates torque from the commanded acceleration at each time when the target acceleration is set. For example, the vehicle control information generation unit 53 calculates torque according to the following equation.
Here, F is the torque applied to the vehicle 3, and m is the mass of the vehicle 3. And v is the latest measured value of the actual speed of the vehicle 3, and a _cmd is the commanded acceleration at the time of _interest . A, B, and C are constants, for example, values determined in advance by a running experiment or the like, and are stored in the storage unit 44 in advance.

  Each time the vehicle control information generation unit 53 calculates torque as vehicle control information, the vehicle control information generation unit 53 outputs the vehicle control information to the ECU via the in-vehicle communication unit 46. The ECU controls the vehicle 3 so that the acceleration of the vehicle 3 becomes the target acceleration by adjusting the accelerator opening or the braking amount of the brake according to the vehicle control information.

FIG. 10 shows an operation sequence of the automatic driving support process. The roadside machine 2 and the automatic driving device 4 execute the automatic driving support process according to the following operation sequence at a predetermined cycle.
The moving object detection unit 31 of the roadside machine 2 detects a moving object present in the monitoring target area based on the sensor signal from the sensor 21 (step S101). The tracking unit 32 of the roadside device 2 tracks the moving object based on the locus of the moving object that has already been detected and the current position of the moving object detected from the latest sensor signal, and the current position and speed of the moving object. Is updated (step S102).

  Further, the selection unit 33 of the roadside machine 2 selects, for each moving object currently being tracked, trajectory information representing the trajectory of the past moving object that is similar to the path of the moving object from the trajectory database (step S103). ). And the prediction information generation part 34 of the roadside machine 2 produces | generates the prediction information showing the presence probability of a moving object for every area on the planned path | route of the vehicle 3 based on the selected locus information at a predetermined time interval ( Step S104). Then, the control unit 23 of the roadside machine 2 transmits the prediction information to the automatic driving device 4 of the vehicle 3 via the communication unit 24.

  When the control unit 45 of the automatic driving device 4 receives the prediction information via the communication unit 42, the target acceleration setting unit 52 of the automatic driving device 4 prevents the vehicle 3 from colliding with the moving object based on the prediction information. Then, the target acceleration of the vehicle 3 at each time for each predetermined period after the current time is set (step S105).

  The vehicle control information generation unit 53 of the automatic driving device 4 corrects the target acceleration by feedback control according to the actual speed of the vehicle 3 and obtains a command acceleration (step S106). Then, the vehicle control information generation unit 53 converts the command acceleration into torque (step S107). The vehicle control information generation unit 53 notifies the ECU of the torque as vehicle control information via the in-vehicle communication unit 46 (step S108). And the automatic driving assistance system 1 complete | finishes an automatic driving assistance process.

  FIG. 11 shows the speed change of the vehicle 3 controlled by the automatic driving support system according to the present embodiment and the vehicle, which is a comparative example, when there is a moving object that crosses the planned route of the vehicle obtained by simulation. It is a graph showing the comparison with the speed change of the vehicle at the time of controlling a vehicle speed with the detection signal from the mounted sensor. In this comparative example, the vehicle is decelerated when a moving object is detected within 5 m ahead of the vehicle.

  In FIG. 11, the horizontal axis represents time, and the vertical axis represents vehicle speed. A graph 1101 represents a speed change of the vehicle 3 according to the present embodiment. On the other hand, a graph 1102 represents a change in vehicle speed according to the comparative example. As shown in FIG. 11, it can be seen that the speed change of the vehicle is smaller in the present embodiment than in the comparative example. Further, in this embodiment, since the speed of the vehicle 3 does not become 0 km / h, it can be seen that the vehicle 3 can avoid moving objects without stopping.

  FIG. 12 shows the position change of the vehicle 3 controlled by the automatic driving support system according to the present embodiment and the detection signal from the sensor mounted on the vehicle, which is a comparative example, in the same situation as FIG. It is a graph showing a comparison with the position change of the vehicle at the time of controlling a vehicle speed by. As in FIG. 11, in this comparative example, the vehicle is decelerated when a moving object is detected within 5 m ahead of the vehicle.

  In FIG. 12, the horizontal axis represents time, and the vertical axis represents the position of the vehicle. A graph 1201 represents a change in the position of the vehicle 3 according to the present embodiment. On the other hand, a graph 1202 represents a change in the position of the vehicle according to the comparative example. As shown in FIG. 12, when the elapsed time is longer in this embodiment than in the comparative example, it can be seen that the vehicle can travel a long distance in the same time.

  FIG. 13 is a comparative example of the distance change between the vehicle 3 controlled by the automatic driving support system according to the present embodiment and the moving object crossing the planned route in the same situation as FIG. It is a graph showing the comparison with the distance change between a vehicle and a moving object at the time of controlling a vehicle speed with the detection signal from the mounted sensor. As in FIG. 11, in this comparative example, the vehicle is decelerated when a moving object is detected within 5 m ahead of the vehicle.

  In FIG. 13, the horizontal axis represents time, and the vertical axis represents the distance between the vehicle and the moving object. A period 1300 is a period in which the moving object is located on the planned route. A graph 1301 represents a change in the distance between the vehicle 3 and the moving object according to the present embodiment. On the other hand, a graph 1302 represents a change in the distance between the vehicle and the moving object according to the comparative example. As shown in FIG. 13, it can be seen that the distance between the moving object and the vehicle can be kept larger in the present embodiment than in the comparative example during the period in which the moving object crosses the planned route.

  As described above, this automatic driving support system uses the trajectory of the past moving object, which is similar to the path of the moving object being tracked, to move the moving object at each position on the planned route of the vehicle. The estimation accuracy of existence probability can be improved. Further, the automatic driving support system controls the vehicle so that the vehicle does not collide with the moving object based on the existence probability of the moving object at each position on the planned route. At this time, since the automatic driving support system sets the target acceleration of the vehicle, the vehicle can be appropriately controlled according to the behavior that the vehicle can actually take. At that time, this automatic driving support system appropriately evaluates the uncertainty of the prediction of the position of the moving object by incorporating the existence probability of the moving object into the evaluation function for determining the target acceleration of the vehicle at each time. Thus, the vehicle can be appropriately controlled. In addition, this automatic driving support system can suppress sudden acceleration / deceleration.

  The preferred embodiments of the present invention have been described above, but the present invention is not limited to these embodiments. For example, the prediction information generation unit 34 of the roadside machine 2 includes the path of the moving object being tracked among the Gaussian distributions included in the mixed Gaussian distribution used for obtaining the existence probability of the moving object in each section on the planned route. The Gaussian distribution created from the trajectory information representing the trajectory having a high degree of similarity may be weighted so that the weighting coefficient becomes larger.

For example, for each moving object that is being tracked, the prediction information generation unit 34 creates a Gaussian distribution created from the trajectory information as the current position of the moving object is closer to the position of the past moving object included in the trajectory information. The weighting factor for may be increased. For example, it is assumed that the relationship between the current position of the moving object and the closest position in the trajectory j can be described by a Gaussian distribution. In this case, the weight coefficient λ _v ^j is obtained according to the following equation, for example. However, the distance distance ((p _x ⁰ , p _y ⁰ ), (q _x ^j (k ^j ), q _y ^j (k ^j ))) between the current position of the moving object and the closest position in the trajectory j The average value was assumed to be 0. R is a 1-m variance representing an error in the current position of the moving object by the sensor 21.
By using this weight, the prediction information generation unit 34 uses the current position of the moving object and the past movement included in the trajectory information, as shown in the following expression, for the existence probability of the moving object in the section u at the time ω _v . It can be obtained in consideration of the weight according to the distance from the position of the object.
Thereby, the accuracy of the existence probability of the moving object in each section on the planned route is further improved.

  According to another embodiment, when the automatic driving device 4 can acquire information representing the planned route of the vehicle 3 from a navigation system or the like mounted on the vehicle 3, the request signal generation unit 51 uses the predicted information request signal. Information indicating the planned route may be included. In this case, the prediction information generation unit 34 of the roadside machine 2 obtains the existence probability for each predetermined period for each section on the planned route represented by the information received from the automatic driving device 4, and the automatic driving device 4. May be sent to.

  According to another embodiment, the tracking unit 32 of the roadside device 2 may track a moving object according to another tracking method. For example, the tracking unit 32 may track a moving object using a Kalman smoother.

According to still another embodiment, the roadside device 2 generates prediction information at a fixed period, and transmits one or more unspecified radio signals within the communicable range of the roadside device 2 to superimpose the prediction information. May be sent to the other vehicle. In this case, since the moving object corresponding to the specific vehicle cannot be excluded from the tracking moving object, the selection unit 33 selects the trajectory information for all of the tracking moving objects. Then, the prediction information generation unit 34 uses each Gaussian distribution constituting the mixed Gaussian distribution used for calculating the existence probability instead of the existence probability of the moving object for each section on the planned route of each vehicle as the prediction information. And the current position of the moving object corresponding to the trajectory information used to create the Gaussian distribution is included in the prediction information.
On the other hand, the control unit 45 of the automatic driving device 4 of each vehicle is currently within a predetermined range (for example, 1 to 2 m) of the distance from the position of the host vehicle among the parameters defining the Gaussian distribution included in the prediction information. A mixed Gaussian distribution is created excluding parameters related to a moving object with a position. And the control part 45 should just calculate the presence probability of the moving object for every time in each area on the planned path | route of the own vehicle according to the mixed Gaussian distribution.
According to this modification, the roadside machine 2 only needs to generate one piece of prediction information that is commonly used by each vehicle even when a plurality of vehicles are present in the monitoring target area. The amount of computation is reduced.

According to still another embodiment, the predicted result of the path of the moving object by the moving object position prediction device may be used other than the automatic driving of the vehicle. For example, based on prediction information from a roadside machine that is an example of a moving object position prediction device, the automatic driving device determines when the vehicle arrives in any section on the planned route when the vehicle travels at the current speed. When the existence probability of a moving object in the vehicle exceeds a predetermined threshold (for example, 10%), a warning signal is sent to the driver via a monitor (not shown) or a speaker (not shown) installed in the passenger compartment. You may make it emit.
Alternatively, when the moving object position prediction device predicts the course of the moving object entering the intersection, the moving object position prediction device outputs the prediction result of the moving object route to the control device of the traffic light installed at the intersection. May be. Then, according to the prediction result of the path of the moving object, the control device determines whether the moving object has a higher probability of existence at a position within a predetermined range from the intersection in each of the two roads connected to the intersection. The traffic light may be controlled so that the signal is blue.

  Moreover, the computer program which makes a computer perform the process of each part performed by the control part 23 of said roadside machine 2 may be recorded on an optical recording medium or a magnetic recording medium, and may be distributed.

  As described above, those skilled in the art can make various modifications in accordance with the embodiment to be implemented within the scope of the present invention.

1 Automatic driving support system 2 Roadside machine (moving object position prediction device)
DESCRIPTION OF SYMBOLS 3 Vehicle 4 Automatic driving device 21 Sensor 22 Storage part 23 Control part 24 Communication part 25 Antenna 31 Moving object detection part 32 Tracking part 33 Selection part 34 Prediction information generation part 41 Antenna 42 Communication part 43 Position measurement part 44 Storage part 45 Control part 46 In-vehicle communication unit 51 Request signal generation unit 52 Target acceleration setting unit 53 Vehicle control information generation unit

Claims (8)

For each of a plurality of past moving objects that have moved in the monitoring target area, a storage unit (22) that stores trajectory information representing the trajectory of the past moving object;
A sensor (21) for generating a sensor signal representing an object existing in the monitored area;
Based on the sensor signal, a moving object detection unit (31) that detects a moving object moving in the monitoring target area;
A selection unit (33) for selecting at least one trajectory information representing a trajectory similar to the assumed path of the moving object from the plurality of trajectory information stored in the storage unit (22);
For each of the selected at least one piece of trajectory information, a probability distribution in which the probability that the moving object exists is higher at a position closer to the trajectory represented by the trajectory information is generated, and according to a mixed distribution including each of the probability distributions A prediction information generation unit (34) for generating prediction information indicating the probability that the moving object exists at a predetermined position at a predetermined time;
A moving object position prediction apparatus.   The selection unit (33) includes, among the plurality of pieces of trajectory information, the corresponding trajectory includes a point within a predetermined range from the current position of the moving object, and the moving direction of the past object at the point and the The moving object position prediction apparatus according to claim 1, wherein trajectory information in which a difference in moving direction of the moving object is within a predetermined direction range is selected. Each of the plurality of trajectory information includes information indicating a type of the corresponding past moving object,
The moving object detection unit (31) identifies the type of the moving object,
The moving object position prediction apparatus according to claim 2, wherein the selection unit (33) selects trajectory information on the past moving object of the same type as the moving object among the plurality of trajectory information. Each of the plurality of trajectory information includes information indicating a time when the past moving object corresponding to each point on the trajectory represented by the trajectory information passes,
The selection unit (33) includes a time when the past moving object passes a point within the predetermined range from the current position of the moving object and the current position of the moving object among the plurality of pieces of trajectory information. The moving object position prediction apparatus according to claim 2, wherein trajectory information whose difference from the detection time is within a predetermined range is selected.   The said selection part (33) is a moving object position prediction apparatus as described in any one of Claims 2-4 which selects the said locus | trajectory information using a nearest neighbor search method.   The prediction information generation unit (34) has a larger weighting factor for the probability distribution corresponding to the trajectory information in which the point included in the corresponding trajectory is closer to the current position of the moving object in the selected trajectory information. The moving object position prediction apparatus according to any one of claims 1 to 5, wherein the mixed distribution is generated by weighting the probability distribution as described above. Detecting a moving object moving in the monitoring target area based on a sensor signal representing an object existing in the monitoring target area;
The moving object is assumed from the trajectory information representing the trajectory of the past moving object for each of a plurality of past moving objects that have moved within the monitoring target area, stored in the storage unit (22). Selecting at least one trajectory information representing a trajectory similar to the course;
For each of the selected at least one piece of trajectory information, a probability distribution in which the probability that the moving object exists is higher at a position closer to the trajectory represented by the trajectory information is generated, and according to a mixed distribution including each of the probability distributions Generating prediction information representing a probability that the moving object is present at a predetermined position at a predetermined time;
A moving object position prediction method comprising: Detecting a moving object moving in the monitoring target area based on a sensor signal representing an object existing in the monitoring target area;
The moving object is assumed from the trajectory information representing the trajectory of the past moving object for each of a plurality of past moving objects that have moved within the monitoring target area, stored in the storage unit (22). Selecting at least one trajectory information representing a trajectory similar to the course;
For each of the selected at least one piece of trajectory information, a probability distribution in which the probability that the moving object exists is higher at a position closer to the trajectory represented by the trajectory information is generated, and according to a mixed distribution including each of the probability distributions Generating prediction information representing a probability that the moving object is present at a predetermined position at a predetermined time;
A computer program for predicting a moving object position for causing a computer to execute.