JP5315065B2 - Traveling pattern prediction apparatus and traveling pattern prediction method - Google Patents

Description

  The present invention relates to a travel pattern predicting apparatus and a travel pattern predicting method, and more particularly to a travel pattern predicting apparatus and a travel pattern predicting method for predicting a travel pattern of a host vehicle based on the acquired travel pattern of the host vehicle.

  Conventionally, by predicting the traveling pattern of the host vehicle with a navigation device, etc., and setting the destination and obtaining the route to the destination without the driver's operation, the route from the home to the company or department store, A technique has been proposed for reducing the annoyance that a driver has to set a destination each time a driver travels along the route.

  For example, Patent Document 1 stores past travel pattern data including a departure place, a departure time (day of the week + time), a destination, and a travel route between the destination, and when starting the current travel, A navigation device that automatically estimates a destination from past driving pattern data based on a starting point detected by a position detection unit and a starting time obtained by a clock, and simultaneously acquires a driving route to the estimated destination Proposed.

  In this navigation device, for example, when leaving home at 8:10 am on Monday, the travel pattern data is searched, and the company, hospital, Analogize the most frequent company of banks as the destination. If the analogized destination matches the driver's intention, the navigation apparatus acquires a past travel route from the home to the destination from the travel pattern data.

JP 2003-294477 A

  However, in the above technique, the past travel pattern is accumulated for each day of the week, and the travel pattern for the day of the week is estimated from the travel pattern accumulated for each day of the week. Therefore, the travel pattern data for the day of the week is accumulated to some extent. Up to this point, there is a drawback that the accuracy of predicting the running pattern is low, and it takes time until the prediction accuracy increases. That is, in the above technique, a long learning period is required to predict a running pattern with high reliability.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a travel pattern prediction apparatus and a travel pattern prediction method capable of predicting a travel pattern with high accuracy while having a shorter learning period. is there.

The present invention relates to a travel pattern acquisition unit that acquires a travel pattern of a host vehicle, a similarity calculation unit that calculates a similarity between travel patterns of the host vehicle acquired at different times by the travel pattern acquisition unit, and a similarity calculation unit Predicting means for predicting the traveling pattern of the host vehicle based on the similarity calculated by the vehicle, an index calculating means for calculating an index of the traveling pattern based on the similarity calculated by the similarity calculating means, Threshold setting means for setting a threshold for predicting a pattern, and the similarity calculation means is the same as the day in the same time zone as the day in which the prediction means predicts the traveling pattern of the host vehicle. The driving pattern acquired by the driving pattern acquisition means for any one of the day, day of the week, week, month and year and the driving pattern for any of the day, day of the week, week, month and year different from the one day. Calculating a similarity between the running pattern over down acquisition means has acquired, predicting means, indices index calculation means has calculated, to predict the travel pattern depending on whether exceeds the threshold value threshold setting unit has set, It is a running pattern prediction device.

  According to this configuration, the similarity calculating unit calculates the similarity between the traveling patterns of the host vehicle acquired at different times by the traveling pattern acquiring unit, and the predicting unit is based on the similarity calculated by the similarity calculating unit. Therefore, for example, it is not always necessary to accumulate travel pattern data on the same day of the week, and the travel pattern is predicted from the similarity of the travel patterns of the own vehicle acquired on different days of the week. Therefore, the traveling pattern can be predicted with high accuracy while the learning period is shorter.

In addition , the similarity calculation means is a driving pattern in any one of the same day, day of the week, week, month and year as the one day in the same time zone as the day in which the prediction means predicts the driving pattern of the host vehicle. The travel pattern is predicted in order to calculate the similarity between the travel pattern acquired by the acquisition means and the travel pattern acquired by the travel pattern acquisition means for any one of the day, day of the week, week, month, and year different from the one day. Even if data such as the same day of the week in the same time zone as the day of the day is not accumulated, the running pattern in the time zone of the day is predicted from the similarity to the running pattern acquired on the day of the week different from the day of the week. Thus, the traveling pattern can be predicted with high accuracy while the learning period is shorter.

In addition , since the predicting unit predicts the running pattern based on whether the index calculated by the index calculating unit based on the similarity exceeds the threshold set by the threshold setting unit, the prediction unit can be used in a small learning period according to the similarity. A traveling pattern can be predicted with high accuracy.

  In this case, the index calculating means calculates the product of the similarity of the driving pattern of any one of the day, day of the week, week, month and year different from one day and the frequency of the driving pattern, the day, day of the week, It is preferable to calculate the running pattern index based on the value added for each week, month, and year.

  According to this configuration, the index calculating means calculates the product of the similarity between the travel pattern frequency of any one of the day, day of the week, week, month, and year that is different from one day and the frequency of the travel pattern from each day. Because the driving pattern index is calculated based on the value added for each day of the week, week, month, and year, the driving pattern index is valid according to the similarity and frequency of the driving pattern on different days of the week, etc. It becomes possible to calculate to a correct value.

  On the other hand, the threshold setting means sets the first threshold until the frequency of the running pattern reaches a predetermined number or more, and after the frequency of the running pattern becomes the predetermined number of times or more, the second lower than the first threshold. It is preferable to set a threshold value.

  In this configuration, the threshold setting unit sets the first threshold until the frequency of the running pattern reaches a predetermined number of times, and is lower than the first threshold after the frequency of the running pattern reaches the predetermined number of times. Since the second threshold is set, for a driving pattern that is frequently performed more than a predetermined number of times, such as commuting, for example, five days a week, the index drops due to reasons such as vacation and business trips, and commuting and driving only on one day It is possible to prevent a situation in which the threshold for predicting the pattern cannot be exceeded.

  Furthermore, the travel pattern can be any of a set of a departure point and a destination of the own vehicle, a route of the own vehicle, and a stop time of the own vehicle.

  In this configuration, it is possible to predict any of a set of a departure point and a destination of the own vehicle, a route of the own vehicle, and a stop time of the own vehicle.

On the other hand, the present invention provides a step of acquiring a traveling pattern of the host vehicle, a step of calculating a similarity between the traveling patterns of the host vehicle acquired at different times, and a traveling pattern of the host vehicle based on the calculated similarity. based on the process and, the calculated similarity to predict, a step of calculating an index of the running pattern, seen including a step of setting a threshold for predicting the travel pattern, calculating a similarity, the self The same day, day, week, month, and year as the one day in the same time zone as the day in which the driving pattern of the vehicle is predicted, and a different day from the one day, The step of calculating the similarity with the travel pattern acquired for any of the day of the week, the week, the month, and the year, and predicting the travel pattern is based on whether or not the calculated index exceeds a set threshold. To measure a travel pattern prediction method.

  In this case, the step of calculating the travel pattern index is performed by calculating the product of the similarity of the travel pattern of any one of the day, day of the week, week, month, and year and the frequency of the travel pattern. It is preferable to calculate a running pattern index based on a value added for each of day, day of the week, week, month, and year different from each other.

  On the other hand, in the step of setting a threshold for predicting the travel pattern, the first threshold is set until the frequency of the travel pattern reaches a predetermined number of times, and after the frequency of the travel pattern reaches the predetermined number of times, It is preferable to set a second threshold value lower than the first threshold value.

  Furthermore, the traveling pattern can be any of a set of a departure point and a destination of the own vehicle, a route of the own vehicle, and a stop time of the own vehicle.

  According to the traveling pattern predicting apparatus and the traveling pattern predicting method of the present invention, it is possible to predict a traveling pattern with high accuracy while having a shorter learning period.

It is a block diagram which shows the structure of the navigation apparatus which concerns on embodiment. It is a flowchart which shows operation | movement of the navigation apparatus which concerns on embodiment. It is a figure which shows an example of action pattern data. It is a figure which shows the method of calculating the similarity degree of work attendance on Monday and work attendance on Tuesday. It is a figure which shows the method of calculating the similarity degree of work attendance on Monday and work attendance on Wednesday.

  Embodiments of the present invention will be described below with reference to the accompanying drawings.

  FIG. 1 is a block diagram illustrating a configuration of the navigation device according to the embodiment. In the present embodiment, an example is shown in which the travel pattern prediction device of the present invention is realized as a navigation device. As shown in FIG. 1, in the navigation device 10 of this embodiment, a GPS 22, sensors 24, a behavior pattern database 26, and a display unit 40 are connected to a navigation processing unit 30.

  The GPS (Global Positioning System) 22 receives a plurality of signals from GPS hygiene by a GPS receiver, and measures the position of the host vehicle from the difference between the signals. Positioning by the GPS 22 makes it possible to acquire the departure point, the destination, and the route of the host vehicle.

  Specifically, the sensors 24 are sensors such as a vehicle speed sensor, an acceleration sensor, a yaw rate sensor, a steering sensor, and a gyro sensor, and are used for detecting the traveling state of the host vehicle. The GPS 22 and the sensors 24 are used for acquiring a traveling pattern of the host vehicle such as a set of a departure point and a destination of the host vehicle (hereinafter sometimes referred to as OD (Origin-destination)), a route, and a stop time. And functions as a running pattern acquisition means.

  The behavior pattern database 26 displays driving patterns such as OD, route, and stop time of the host vehicle acquired by the GPS 22 and sensors 24 for each item such as departure time zone, day, day of the week, week, month, and year. It is for organizing and recording.

  The navigation processing unit 30 includes a similarity calculation unit 31, a frequency acquisition unit 32, an index calculation unit 33, an action determination unit 34, and a threshold setting unit 35. The similarity calculation unit (similarity calculation means) 31 uses the method described later, and the same day, day of the week, week, month, Similarity between the driving pattern acquired by the GPS 22 and the sensors 24 for any of the years and the driving pattern acquired by the GPS 22 and the sensors 24 for any of the day, day of the week, week, month, and year different from the one day It is for calculating.

  The frequency acquisition unit 32 is for acquiring, from the behavior pattern database 26, the frequency at which the various traveling patterns acquired by the GPS 22 and the sensors 24 are performed every day, day of the week, week, month and year.

  The index calculation unit (similarity calculation means) 33 is for calculating a running pattern index based on the similarity calculated by the similarity calculation unit 31. The behavior determination unit (prediction means) 34 is for predicting a running pattern depending on whether or not the index calculated by the index calculation unit 33 exceeds the threshold set by the threshold setting unit 35. The threshold setting unit (threshold setting means) 35 is for setting a threshold for the behavior determining unit 34 to predict a running pattern.

  The display unit 40 is specifically a liquid crystal display or an audio speaker, and is for displaying route guidance based on the traveling pattern of the host vehicle predicted by the navigation processing unit 30 to the driver.

  Hereinafter, the operation of the navigation device 10 of the present embodiment will be described. In the following example, a case will be described in which it is determined as a traveling pattern to determine whether or not it is a work behavior that goes to work from Monday to Friday morning. FIG. 2 is a flowchart showing the operation of the navigation device according to the embodiment. As shown in FIG. 2, when the host vehicle is about to leave from home at around 8:00 am, the similarity calculation unit 31 of the navigation processing unit 30 determines the commuting behavior that is the behavior pattern (running pattern) to be determined. In the following example, the similarity between days of the week is calculated for the action patterns in the same OD and the same departure time zone (S1).

  FIG. 3 is a diagram illustrating an example of behavior pattern data. FIG. 3 shows the behavior pattern of the host vehicle acquired by the GPS 22 and the sensors 24 in the past and recorded in the behavior pattern database 26. In the following example, it is assumed that the OD is from “home” to “work location” and the behavior occurrence state at the departure time around 8:00 is as shown in FIG.

  In FIG. 3, “departure time around 8 o'clock” means, for example, the departure time of the vehicle is AM7: 30 to AM8: 30 and hourly hours, or AM7: 00 to AM9: 00 and hourly every two hours. It shall refer to the case corresponding to the belt. In FIG. 3, a circle indicates a case where the behavior pattern matches the commuting behavior, and a cross indicates a case where the behavior pattern does not match the commuting behavior. The case where the behavior pattern matches the commuting behavior can be, for example, a case where the difference stays within 10 to 50 m if it is OD and a case where the difference stays within 70% if it is a route.

  In the case illustrated in FIG. 3, the similarity calculation unit 31 calculates the similarity between the attendance behavior on Monday and the attendance behavior on Tuesday as follows based on the occurrence pattern of the commuting behavior. FIG. 4 is a diagram illustrating a method of calculating the similarity between the attendance behavior on Monday and the attendance behavior on Tuesday. As shown in FIG. 4, the attendance behavior on Monday and the attendance behavior on Tuesday are different in the first week because the attendance behavior is performed on Monday and the attendance behavior is not performed on Tuesday. In the second week, there is no attendance behavior on Monday, and attendance behavior is performed on Tuesday. In the third week, work attendance was done on Monday and work attendance was done on Tuesday.

  In this case, since the coincidence status of the occurrence of attendance behavior (the coincidence status of the circle and the cross mark) is the same only once in the past three weeks, the similarity calculation unit 31 calculates the attendance behavior on Monday and the attendance behavior on Tuesday. The degree of similarity K (Monday, Tuesday) = 1/3 = 33%.

  The same applies to other cases. For example, it is assumed that the attendance behavior on Monday and the attendance behavior on Wednesday are as shown in FIG. As shown in FIG. 5, the attendance behavior on Monday and the attendance behavior on Wednesday are different in the first week because the attendance behavior is performed on Monday and the attendance behavior is not performed on Wednesday. In the second week, there is no attendance behavior on Monday, and no attendance behavior is performed on Wednesday. In the third week, work attendance is performed on Monday and work attendance is performed on Wednesday.

  In this case, since the coincidence status of occurrence of attendance behavior is the same only twice in the past three weeks, the similarity calculation unit 31 determines the similarity K (month, water) = 2 between the attendance behavior on Monday and the attendance behavior on Wednesday. / 3 = 66%. Similarly, the similarity calculation unit 31 calculates the similarity K (month, day) between the Monday attendance action and the Sunday attendance action from the similarity K (month, month) between the Monday attendance action and the Monday attendance action. Calculate up to. Note that the degree of similarity K (month, month) with respect to the subject day itself, such as the degree of similarity K (month, month) between Monday's attendance behavior and Monday's attendance behavior, is 100%, that is, the degree of similarity K ( (Month, month) = 1 = 100%.

  The index calculation unit 33 calculates an index for commuting behavior determination based on the similarity between days of the week calculated by the similarity calculation unit 31 as described above (S2). In the following example, a method for calculating an index for determining the attendance behavior will be described using the attendance behavior on Monday as an example. The frequency of the commuting behavior acquired by the frequency acquisition unit 32 from the behavior pattern database 26 is, for example, 3 times Monday, 2 Tuesdays, 0 Wednesdays, 1 Thursdays, 2 Fridays, 1 Saturdays, 0 Sundays. Suppose there was. Further, the similarity K between days of the week for the attendance behavior calculated by the similarity calculation unit 31 is, for example, K (month, month) = 100%, K (month, Tuesday) = 80%, K (month, water) = 50%, K (Monday, Thursday) = 30%, K (Monday, Friday) = 10%, K (Monday, Saturday) = 100%, K (Monday, Sun) = 70%.

  In the above case, the index calculation unit 33 sets the index I for commuting behavior determination as a value obtained by adding the product of the similarity K and occurrence frequency between each day of the week as in the following formula (1). calculate.

  I = K (Monday, Monday) x 3 times + K (Monday, Tuesday) x 2 times + K (Monday, Wednesday) x 0 times + K (Monday, Thursday) x K (Monday, Friday) x 2 times + K (Monday, Saturday) ) × 1 + K (month, day) × 0 times = 6.1 (1)

  When the commuting behavior that is the behavior pattern to be determined and the behavior pattern of the same OD are not already confirmed as commuting behavior (S3), for example, the behavior of the same departure time and the same OD stored in the behavior pattern database 26 When the frequency of the pattern is less than the predetermined number of times 1 to 10, or when the number of times that the behavior determination unit 34 determines that the commuting behavior is less than the predetermined number of times 1 to 10, the threshold setting unit 35 is high. A threshold value (first threshold value) is set, and the behavior determination unit 34 determines commuting behavior depending on whether or not the index I calculated by the index calculation unit 33 exceeds a high threshold value (S4).

  For example, if the threshold setting unit has set 6 as a high threshold, if the index calculation unit 33 calculates the index I = 6.1 as shown in the above equation (1), the vehicle is About the action pattern which departs from home at about 8:00 am, the action determination part 34 determines with a commuting action.

  On the other hand, when the commuting behavior that is the behavior pattern to be determined and the behavior pattern of the same OD have already been confirmed as commuting behavior (S3), that is, for example, the same departure time and the same OD stored in the behavior pattern database 26 When the frequency of the behavior pattern is 1 to 10 times or more, or when the number of times the behavior determination unit 34 determines to be commuting behavior is 1 to 10 times or more, the threshold setting unit 35 Sets a low threshold (second threshold), and the behavior determination unit 34 determines commuting behavior based on whether or not the index I calculated by the index calculation unit 33 exceeds the low threshold (S5).

  For example, if the threshold setting unit has set 4 as a high threshold, if the index calculation unit 33 calculates the index I = 6.1 as shown in the above equation (1), the vehicle is About the action pattern which departs from home at about 8:00 am, the action determination part 34 determines with a commuting action.

  As described above, if the behavior determination unit 34 determines that the behavior pattern of the vehicle departing from home at about 8:00 am on Monday is commuting behavior, the display unit 40 can be moved from home even if there is no particular input by the driver. The route guidance to the office and the traffic situation of the route are presented to assist the driver in driving. The behavior pattern database 26 updates data for each limited period, for example, every month, the similarity calculation unit 31 calculates the similarity K of various behavior patterns in various time zones, and the frequency acquisition unit 32 The frequency of the behavior pattern is acquired, and the index calculation unit 33 calculates the index I of each behavior pattern.

  According to the present embodiment, the similarity calculation unit 31 calculates the similarity K between the traveling patterns of the host vehicle acquired by the GPS 22 and the sensors 24 at different times, and the behavior determination unit 34 calculates the similarity calculation unit 31. Therefore, for example, it is not always necessary to accumulate data of the driving pattern on the same day of the week, and the driving patterns of the own vehicle acquired on different days of the week are not necessarily stored. Since the travel pattern can be predicted from the similarity K, the travel pattern can be predicted with high accuracy while the learning period is shorter.

  In particular, according to the present embodiment, the similarity calculation unit 31 includes the GPS 22 and the sensors on the same day of the week as the day of the week in the same time zone of the day when the behavior determination unit 34 predicts the traveling pattern of the host vehicle. In order to calculate the similarity K between the travel pattern acquired by the vehicle 24 and the travel pattern acquired by the GPS 22 and the sensors 24 on the day of the week different from the day, the same time zone as the time zone of the day on which the travel pattern is predicted Even if the data for the same day of the week is not accumulated, it is possible to predict the running pattern in the time zone of the day from the similarity K with the running pattern acquired on the day different from the day of the week. The driving pattern can be predicted with high accuracy.

  Further, according to the present embodiment, the behavior determination unit 34 determines the traveling pattern depending on whether or not the index I calculated by the index calculation unit 33 based on the similarity K exceeds the threshold set by the threshold setting unit 35. In order to predict, according to the similarity K, a driving | running pattern can be accurately estimated in a small learning period.

  In addition, according to the present embodiment, the index calculation unit 33 is a product of the similarity K of the driving pattern on the day of the week when the behavior determination unit 34 predicts the driving pattern of the host vehicle and the frequency of the driving pattern. Is calculated for each day of the week that is different from the day, the travel pattern index I is calculated according to the similarity K and frequency of the travel pattern on different days of the week. It becomes possible to calculate to an appropriate value.

  Furthermore, according to the present embodiment, the threshold setting unit 35 sets a high threshold until the frequency of the running pattern reaches the predetermined number of times or more, and sets the low threshold after the frequency of the running pattern becomes the predetermined number of times or more. In order to set a driving pattern that is frequently performed more than a predetermined number of times, such as commuting behavior, which is performed five days a week from Monday to Friday, the index I decreases from the cause of vacation, business trip, etc., and the first day of the week Only the threshold for predicting the commuting behavior and the running pattern can be prevented.

  Further, according to the present embodiment, it is possible to predict any of the OD of the host vehicle, the route of the host vehicle, and the stop time of the host vehicle.

  It should be noted that the present invention is not limited to the above-described embodiment, and it is needless to say that various modifications can be made without departing from the gist of the present invention. For example, in the above embodiment, the similarity of commuting behavior between days of the week around 8:00 am is calculated, but in the present invention, the similarity of behavior patterns of different days, weeks, months, and years in the same time zone is calculated. You can do it. Alternatively, in the present invention, in the same weather or season as the day's weather or season for predicting the travel pattern of the host vehicle, the travel pattern acquired on the same day, day of the week, week, month, or year as the one day. In addition, the degree of similarity with the travel pattern acquired for any one of the day, the day of the week, the week, the month, and the year may be calculated.

DESCRIPTION OF SYMBOLS 10 ... Navigation apparatus, 22 ... GPS, 24 ... Sensors, 26 ... Action pattern database, 30 ... Navigation processing part, 31 ... Similarity calculation part, 32 ... Frequency acquisition part, 33 ... Index calculation part, 34 ... Action determination part , 35 ... threshold setting unit, 40 ... display unit.

Claims (8)

Traveling pattern acquisition means for acquiring a traveling pattern of the host vehicle;
Similarity calculation means for calculating the similarity between the travel patterns of the host vehicle acquired at different times by the travel pattern acquisition means;
Prediction means for predicting the travel pattern of the host vehicle based on the similarity calculated by the similarity calculation means;
Index calculation means for calculating an index of the running pattern based on the similarity calculated by the similarity calculation means;
Threshold setting means for setting a threshold for the prediction means to predict the travel pattern;
Equipped with a,
The similarity calculation means is the same day, day of the week, week, month and year as the first day in the same time zone as the first day when the prediction means predicts the travel pattern of the host vehicle. The degree of similarity between the travel pattern acquired by the travel pattern acquisition means and the travel pattern acquired by the travel pattern acquisition means for any of a day, a day of the week, a week, a month, and a year different from the one day is calculated. ,
The travel pattern predicting apparatus, wherein the predicting unit predicts the travel pattern based on whether or not the index calculated by the index calculating unit exceeds the threshold set by the threshold setting unit . The index calculating means calculates the product of the similarity of the travel pattern and the frequency of the travel pattern of any one of the day, day of the week, week, month and year different from the one day, a day different from the one day, day, week, based on a value obtained by adding for each one of month and year, and calculates an indication of the travel pattern, driving pattern prediction apparatus according to claim 1. The threshold value setting means sets the first threshold value until the frequency of the running pattern reaches a predetermined number or more, and after the frequency of the running pattern becomes equal to or more than the predetermined number of times, The travel pattern prediction apparatus according to claim 1 or 2 , wherein a low second threshold is set. The travel according to any one of claims 1 to 3 , wherein the travel pattern is one of a set of a start point and a destination of the host vehicle, a route of the host vehicle, and a stop time of the host vehicle. Pattern prediction device. Obtaining a running pattern of the host vehicle;
Calculating the similarity between the travel patterns of the host vehicle acquired at different times;
Predicting the travel pattern of the host vehicle based on the calculated similarity;
Calculating an indicator of the travel pattern based on the calculated similarity;
Setting a threshold for predicting the running pattern;
Only including,
The step of calculating the similarity is acquired on the same day, day of the week, week, month and year as the first day in the same time zone as the first day for predicting the travel pattern of the host vehicle. Calculating the degree of similarity between the travel pattern and the travel pattern acquired on any one of the day, day of the week, week, month and year different from the day 1,
The step of predicting the travel pattern is a travel pattern prediction method in which the travel pattern is predicted based on whether or not the calculated index exceeds the set threshold value . The step of calculating an indicator of the travel pattern includes the product of the similarity of the travel pattern and the frequency of the travel pattern on any one of the day, day of the week, week, month and year different from the one day. The travel pattern prediction method according to claim 5 , wherein an index of the travel pattern is calculated based on a value added for each of day, day of the week, week, month, and year different from each other. The step of setting a threshold for predicting the travel pattern sets a first threshold until the frequency of the travel pattern reaches a predetermined number of times, and after the frequency of the travel pattern reaches the predetermined number of times or more. The travel pattern prediction method according to claim 5 or 6 , wherein a second threshold value lower than the first threshold value is set. The travel according to any one of claims 5 to 7 , wherein the travel pattern is one of a set of a start point and a destination of the host vehicle, a route of the host vehicle, and a stop time of the host vehicle. Pattern prediction method.

Images