JP4767214B2 - Navigation device, navigation method and program - Google Patents

Description

  The present invention relates to a route search of a navigation device.

  Conventionally, a search algorithm such as the Dijkstra method is widely known as a route search technique in a car navigation apparatus. The Dijkstra method is a method for calculating a route by calculating which link on the map will be the minimum cost, using the starting point and the destination as base points.

  The inventors of the present invention pay attention to the fact that the evaluation function for calculating the cost of the route varies depending on the situation, and calculates the cost using an appropriate evaluation function depending on the situation to obtain the route. (Patent document 1). In Patent Document 1, a plurality of cost tables (evaluation functions) are prepared in advance. The cost table is a table used for calculating the cost of the route. For example, information indicating the driver's age group, gender, driving day of the week, driving time zone, etc. is used as a route search condition, and a cost table that meets the route search condition from a plurality of cost tables. Select (Evaluation function). Specifically, probability inference using a Bayesian network model is calculated, and the cost table with the highest probability is selected. With this configuration, it is possible to recommend an appropriate route according to the route search condition.

In the invention of Patent Document 1, a Bayesian network model for selecting a cost table is learned. A cost table selection model is learned so that a cost table for calculating a recommended route closest to the actually traveled route is selected.
JP 2007-10571 A

  For example, in daily commuting, there are many drivers who follow a fixed path. In this way, when there is a way that always passes, if learning of a model for selection is performed in the invention described in Patent Document 1, learning is performed many times by the same teacher signal, and a plurality of cost tables are stored. One of them can be extremely easy to select.

  An object of the present invention is to provide a navigation device that performs a route search by changing an evaluation function according to a search condition, and that performs learning so that an appropriate evaluation function can be selected. To do.

  The navigation apparatus according to the present invention calculates a cost of the route using an evaluation function using a plurality of road attributes of links constituting the route as parameters, and searches for an optimum route from the departure point to the destination based on the cost. A plurality of evaluation functions based on a predetermined relationship between an evaluation function storage unit storing a plurality of evaluation functions for calculating the cost and a route search condition and an evaluation function to be used. An evaluation function selection unit that selects one evaluation function corresponding to a route search condition from the above, a cost of the route using the evaluation function, and a recommendation from the starting point to the destination based on the cost of the route A recommended route search unit that searches for a route, a travel route acquisition unit that acquires a travel route of a vehicle from a departure point to a destination, and the number of times a link constituting the travel route has been passed in the past Learning to correct the predetermined relationship based on the route search condition and the travel route information when it is determined whether or not the predetermined relationship is to be corrected. A part.

  In this way, the learning unit determines whether or not to correct the predetermined relationship between the route search condition and the evaluation function to be used. Therefore, when the travel route includes many links that have passed many times in the past, It can be determined that the relationship is not corrected. As a result, excessive learning for a predetermined relationship for obtaining the evaluation function is not performed using a travel route through which the driver often passes as a teacher signal. A predetermined relationship for obtaining the evaluation function can be appropriately learned using an appropriate travel route (teacher signal) selected based on the past number of passages. Therefore, an appropriate evaluation function can be selected according to the route search condition during the route search.

  In the navigation device of the present invention, the learning unit calculates a plurality of routes corresponding to each evaluation function, and the evaluation function for calculating each route based on a difference between the plurality of routes and the travel route is The probability of being selected may be updated.

  Thus, by updating the probability that an evaluation function is selected using the difference between a plurality of routes and a travel route, the selection probability of a plurality of evaluation functions can be updated using one travel route, and learning is performed. Can speed up.

  The navigation device of the present invention includes a learned route storage unit that stores, as a learned route, a route that has passed a predetermined number of times, and the recommended route search unit stores a route obtained based on a cost in the learned route storage unit. The recommended route may be searched with correction by the stored learned route.

  With this configuration, when there is a learned route around the route from the departure point to the destination, a recommended route including the learned route can be searched.

  The navigation device of the present invention may delete the learned route from the learned route storage unit when the learned route stored in the learned route storage unit does not pass for a predetermined period.

  With this configuration, it is possible to prevent an old learned route that has not passed for a predetermined period from affecting the search for a recommended route.

  The navigation device of the present invention may set the predetermined period until the learned route is deleted according to the distance from the home of the learned route.

  Places far away from home are less frequent than places near home, so by changing the predetermined period until the learning route is erased depending on the distance from home, it is possible to use it in places close to home. The timing for deleting the learning route can be matched.

  In the navigation device according to the present invention, when the learned route storage unit stores a learned route having a branch, the learned route storage unit may store information related to a passing record of the branch.

  With this configuration, it is possible to use the branch passage record for route search. For example, in a route that merges again after branching, if both routes have the same link cost, it is possible to determine which route is recommended as the recommended route by using the branch passage record. In addition, you may memorize | store the passage performance of a branch as a probability value, for example.

  In the navigation device of the present invention, the learned route storage unit may store information indicating a route search condition when passing through the learned route in association with the learned route.

  With this configuration, when the recommended route is corrected with the learned route stored in the learned route storage unit, an appropriate learned route to be used for correction can be selected based on the route search condition.

  The navigation device of the present invention may include a learned route display unit that displays the learned route stored in the learned route storage unit on a map.

  With this configuration, the driver can easily grasp the path stored as the learning route.

  In the navigation device of the present invention, the road attribute includes information indicating the number of times the link has been passed in the past, and each evaluation function is set so that the cost decreases as the number of times the link passes. May be.

  In this way, by reducing the cost as the number of times of passing through the link in the past is increased, a recommended route including a route that passes frequently is searched by a route search based on the cost.

  The information on the number of times the link has passed in the past may be information indicating the number of times the link has passed in the most recent predetermined period.

  With this configuration, a recommended route including a route that frequently passes during the most recent predetermined period is searched instead of a route that has been frequently passed before.

  In the navigation device of the present invention, the road attribute includes information indicating a time required to pass through the link in the past, and the cost of each evaluation function is reduced as the time required is shorter. May be set.

  Thus, by reducing the cost as the required time of the link is shorter, a recommended route with a shorter required time is searched by the route search based on the cost.

  According to the navigation method of the present invention, the cost of the route is calculated using an evaluation function using a plurality of road attributes of links constituting the route as parameters, and the optimum route from the starting point to the destination is searched based on the cost. A method of preparing an evaluation function storage unit that stores a plurality of evaluation functions for calculating the cost, and based on a predetermined relationship between a route search condition and an evaluation function to be used, Selecting one evaluation function corresponding to a route search condition from a plurality of evaluation functions, obtaining a cost of the route using the evaluation function, and recommending from the departure point to the destination based on the cost of the route A step of searching for a route, a step of obtaining a travel route of the vehicle from the departure point to the destination, and a time when a link that constitutes the travel route has been passed in the past. Whether or not the predetermined relationship is to be corrected, and when it is determined that the predetermined relationship is to be corrected, the predetermined relationship is corrected based on the route search condition and the travel route information. And a step of performing.

  Thereby, similarly to the navigation device of the present invention, it is possible to appropriately learn a predetermined relationship for obtaining an evaluation function using an appropriate travel route (teacher signal) selected based on the past number of passages. . Therefore, an appropriate evaluation function can be selected according to the route search condition during the route search. Note that various configurations of the navigation device of the present invention can be applied to the navigation method of the present invention.

  The program of the present invention calculates the cost of the route by using an evaluation function having a plurality of road attributes of links constituting the route as parameters, and searches for the optimum route from the departure point to the destination based on the cost. A program for searching for a route from an evaluation function storage unit storing a plurality of evaluation functions for calculating the cost based on a predetermined relationship between a condition of the route search and an evaluation function to be used. Extracting one evaluation function corresponding to a condition; obtaining a cost of a route using the evaluation function; searching a recommended route from a departure place to a destination based on the cost of the route; Whether to correct the predetermined relationship based on the step of obtaining the travel route of the vehicle from the vehicle to the destination and the number of times the link that constitutes the travel route has passed in the past Or it determines, when it is determined that modifying the predetermined relationship, and a step of modifying the predetermined relationship based on the information of the condition and the travel path of the route search.

  Thereby, similarly to the navigation device of the present invention, it is possible to appropriately learn a predetermined relationship for obtaining an evaluation function using an appropriate travel route (teacher signal) selected based on the past number of passages. . Therefore, an appropriate evaluation function can be selected according to the route search condition during the route search. Various configurations of the navigation device of the present invention can be applied to the program of the present invention.

  According to the present invention, it is possible to appropriately learn a predetermined relationship for obtaining an evaluation function using an appropriate travel route (teacher signal) selected based on the past number of passages, and at the time of route search This has the effect that an appropriate evaluation function can be selected according to the route search conditions.

Hereinafter, a navigation device according to an embodiment of the present invention will be described with reference to the drawings.
(First embodiment)
FIG. 1 is a diagram illustrating a configuration of a navigation device 1 according to the first embodiment. The navigation device 1 includes a map information storage unit 10 that stores road map information, a vehicle position detection unit 18 that determines the current position of the vehicle, a switch 20 that receives an operation from a driver, and a display 22 that displays a map and the like. And a control unit 24 that performs route search and the like.

  The map information storage unit 10 stores various facility information such as tourist facilities and restaurants in addition to road information. The map information storage unit 10 stores information on road attributes.

  FIGS. 2A and 2B are diagrams showing examples of road attribute information. As shown in FIG. 2A, the road is expressed by nodes A to E, links Lab, Lbd, and the like. As shown in FIG. 2B, the road attribute is information such as the road width, distance, and number of passages of each link. In the present embodiment, the number of passages stores the number of passages through the link in the last three months. In the present embodiment, the information on the number of times of passage is stored in the map information. For example, when a read-only recording medium such as a CD-ROM is used as the map information storage unit 10. In addition, the information on the number of passages may be stored in the memory of the navigation device 1 separately from the map information storage unit 10.

  The vehicle position detection unit 18 has a function of detecting the current position of the vehicle. As the vehicle position detector 18, for example, a geomagnetic sensor, a gyroscope, a distance sensor, a GPS receiver, a steering rotation sensor, a vehicle sensor for each rolling wheel, or the like can be used. The vehicle position detection unit 18 may be configured using some of these sensors. Since these sensors have errors of different properties, the current position of the vehicle can be detected with high accuracy by complementing with a plurality of sensors.

  The display 22 has a function of displaying a map around the current position and a route to the destination. The display 22 includes a touch panel and receives input of information necessary for destination setting and the like.

  Next, a configuration for performing a route search by the navigation device 1 will be described. The navigation device 1 searches for a route (hereinafter referred to as “minimum cost route”) that minimizes the cost from the departure point to the destination. Then, a recommended route is obtained with respect to the minimum cost route in consideration of the route information that the driver often passes.

  The navigation device 1 calculates the cost of the route using the evaluation function. In the present embodiment, the navigation device 1 stores a plurality of evaluation functions, and searches for the minimum cost route using an appropriate evaluation function according to the search conditions.

  The evaluation function used by the navigation device 1 of the present embodiment is expressed by link cost = Σ (weight coefficient × road attribute parameter). The navigation device 1 obtains the total link cost of the links constituting the route as the route cost.

  The road attribute information stored in the road information storage unit is used as the road attribute parameter. However, when the road attribute information is used for link cost calculation, the road attribute is replaced with a link cost calculation parameter. For example, the road attribute “road width” is easier for a vehicle to pass and is preferred by the driver, so the cost is reduced. For example, the parameter is replaced with “3” when the road width is less than 5 m, the parameter is replaced with “2” when the road width is 5 m or more and less than 10 m, and the parameter is replaced with “1” when the road width is 10 m or more.

  The navigation apparatus 1 includes a weight coefficient storage unit 12 that stores weight coefficients of evaluation functions. The weight coefficient storage unit 12 stores weight coefficients of a plurality of evaluation functions.

  FIG. 3 is a diagram illustrating an example of data stored in the weighting coefficient storage unit 12. As shown in FIG. 3, the weight coefficient storage unit 12 stores a plurality of weight coefficient tables TA to TC. The weighting coefficient tables TA to TC store weighting coefficients that are multiplied by road attribute parameters in the evaluation function. One weight coefficient table corresponds to one evaluation function. In FIG. 3, for example, in the weighting factor table TA, the distance weighting factor is 0.9 and the road width weighting factor is 0.3. Therefore, the influence of the distance on the link cost is greater than the road width. The weighting coefficient table TA is an evaluation function that searches for a route with more importance on the distance than the road width.

  The navigation device 1 includes a table selection model storage unit 14 that stores a table selection model as a configuration for determining an evaluation function used for performing a route search according to a search condition. The table selection model is a Bayesian network model.

  FIG. 4 is a diagram illustrating an example of the table selection model. The table selection model is a link in which the node N1 of the age group, the node N2 of the sex, the node N3 of the day of the week, the node N4 of the time zone, and the nodes N5 to N7 of the weighting factor tables TA to TC represent a probabilistic dependency. Connected with. In FIG. 4, the direction of the arrow of the link indicates the direction of dependence. For example, the probability value of the node N5 in the weight coefficient table TA depends on the values of the nodes N1 to N4. Age node N1, gender node N2, day of week node N3, and time zone node N4 are observation nodes. These nodes N1 to N4 include driver age, gender information, day of the week at the time of route search, Set the time zone. Each link has information of a conditional probability table (CPT) (not shown), and the navigation apparatus 1 obtains probability values of the weight coefficient tables TA to TC according to the conditional probability table.

  The learned route storage unit 16 stores a route through which the vehicle often passes. The learned route storage unit 16 may store information on links that constitute a route that passes frequently, or information on nodes that constitute the route. In addition, the learned route storage unit 16 may store only nodes that are points for turning right and left in a route that passes frequently, and may not store nodes that correspond to roads. Thereby, the amount of information to be stored in the learned route storage unit 16 can be reduced.

  Next, the control unit 24 of the navigation device 1 will be described. The control unit 24 includes a table selection unit 26, a route search unit 28, a travel route acquisition unit 30, a learning availability determination unit 32, a model learning unit 34, and a route learning unit 36. The table selection unit 26 executes a process of selecting a weighting factor table from a plurality of weighting factor tables stored in the weighting factor storage unit 12 according to a search condition. Using the table selection model shown in FIG. 3, the probability values of the weight coefficient tables TA to TC are calculated, and the weight coefficient table having the highest probability value is selected.

  The route search unit 28 calculates the cost of the route from the departure point to the destination using the selected weighting coefficient table, and executes processing for searching for the minimum cost route. The route search unit 28 searches for the minimum cost route using a search method such as the Dijkstra method. In addition, the route search unit 28 uses the learned route stored in the learned route storage unit 16 to correct the route with the lowest cost and calculate a recommended route.

  FIG. 5A and FIG. 5B are diagrams illustrating an example of a method for correcting a route with the minimum cost using a learned route. FIG. 5A shows the minimum cost route R 1 from the departure point to the destination and the learned route R 2 stored in the learned route storage unit 16. As shown in FIG. 5A, when there is a learning route R2 that connects two predetermined points of the minimum cost route R1, the minimum cost route R1 is corrected by the learning route R2, and recommended as shown in FIG. 5B. The route R3 is obtained. Although an example of the method for obtaining the recommended route R3 using the learned route R2 is shown here, the method for obtaining the recommended route using the learned route is the method shown in FIGS. 5 (a) and 5 (b). Not limited to. For example, when there is a learned route in the vicinity of the minimum cost route, a route connecting the minimum cost route and the learned route may be drawn as a recommended route even if the minimum cost route and the learned route node do not overlap.

  The travel route acquisition unit 30 executes processing for acquiring a route on which the vehicle actually traveled from the departure point to the destination. The travel route acquisition unit 30 acquires the travel route by acquiring the vehicle position detected by the vehicle position detection unit 18 as needed.

  The learning availability determination unit 32 determines whether to learn the table selection model. The learning availability determination unit 32 determines whether or not the actually traveled route matches the recommended route, and determines that learning is not performed if they match. If the travel route and the recommended route do not match, it is determined whether or not the travel route often passes. If the travel route is a path that often passes, it is determined that learning is not performed. If the travel route is not a frequent path, it is determined that learning of the table selection model is performed based on the travel route.

  Here, a method for determining whether or not the travel route is a well-passed road will be described. A parameter of the number of passages of the road attribute is used to determine whether or not the travel route is a path that often passes. For example, the learning possibility determination unit 32 obtains an average value of the number of times of passing through all the links constituting the travel route, and when the average value is equal to or greater than a predetermined threshold (for example, 5 times), the travel route is a route that passes well. You may determine that there is. In addition, when the percentage of the links that exceed the predetermined threshold (for example, 5 times) exceeds the predetermined threshold (for example, 70%) is obtained and the percentage of all the links constituting the travel route is determined. The travel route may be determined to be a route that passes frequently. Here, an example of a method for determining whether or not a road is frequently used has been described. However, in addition to the above method, various methods are conceivable as a method for determining whether or not a road is frequently passed. It is done.

  The model learning unit 34 executes processing for learning the table selection model. As described above, in the table selection model shown in FIG. 4, each of the nodes N1 to N7 is connected by a link indicating a conditional probability dependency relationship, and each link has information on the conditional probability table. The model learning unit 34 updates the information in the conditional probability table according to the search conditions (driver age group, gender, day of the week, time zone) when the travel route and travel route information are acquired.

  The model learning unit 34 calculates a route from the departure point to the destination using the weighting factor table stored in the weighting factor storage unit 12. Hereinafter, for convenience of explanation, a route calculated using each weight coefficient table is referred to as a “candidate route”. Next, the model learning unit 34 obtains a difference between the travel route and each candidate route, and ranks the candidate routes based on the difference. Various methods can be used to calculate the difference. For example, the ratio of links that do not match the travel route out of all the links constituting the recommended route may be calculated as the difference.

  The model learning unit 34 learns the table selection model using the rank information given to the candidate route and the search condition when the travel route is acquired. For example, the conditional probability table is updated so as to increase the probability value of the weighting coefficient table that calculates the high-ranking candidate route and to decrease the probability value of the weighting coefficient table that calculates the low-ranking candidate route. An example of this method is shown below.

  The conditional probability table stores a probability value at which each weighting factor table is selected under each search condition. For example, in the situation (search condition) of gender: male, age group: 40s, day of the week: Monday, time zone: 8:00 to 10:00, the probability value (7/10) that the weight coefficient table TA is selected, the weight coefficient table Information on conditional probabilities such as a probability value (2/10) for selecting TB and a probability value (1/10) for selecting weighting factor table TC is stored for all combinations of search conditions. This conditional probability value can be considered to be based on the past performance that the weighting coefficient table TA was selected 7 times in 10 trials. In this situation, as a result of obtaining the difference between the actual travel route and each candidate route, the rank based on the difference is 1st: weighting factor table TA, 2nd: weighting factor table TB, 3rd: weighting factor table TC Suppose there was. In this case, if the probability value is updated on the assumption that the first weighting coefficient table TA is selected, the weighting coefficient table TA is selected 8 times out of 11, so the probability value of the weighting coefficient table TA is (8 / 11). Similarly, the probability value of the weight coefficient table TB is (2/11), and the probability value of the weight coefficient table TC is (1/11).

  The conditional probability table may be updated according to the rank. For example, if the probability value is updated assuming that the weighting factor table TA is selected twice in three times and the weighting factor table TB is selected once in three times according to the ranking, the probability value of the weighting factor table TA is (9/13), the probability value of the weighting coefficient table TB is (3/13), and the probability value of the weighting coefficient table TC is (1/13). By updating the conditional probability table of the table selection model using the difference from the travel route, it is possible to increase the probability that the weighting factor table that calculates the candidate route close to the travel route is selected.

  The route learning unit 36 obtains a difference route between the travel route and the recommended route, and performs route learning based on the difference route. For example, the difference route between the travel route and the recommended route is stored in the learned route storage unit 16. When the same differential route is written in the learned route storage unit 16 a predetermined number of times, the differential route is stored as a learned route. The learned route stored in the learned route storage unit 16 may be shown on a map displayed on the display 22. Thereby, the driver can easily grasp the path stored as the learning route.

Next, operation | movement of the navigation apparatus 1 of 1st Embodiment is demonstrated.
FIG. 6 is a diagram illustrating a route search operation by the navigation device 1. The navigation device 1 selects a weighting coefficient table used for route search (S10). The navigation device 1 sets the driver's age, sex, driving day of the week, and time zone in the observation node of the table selection model, obtains the probability value of each weight coefficient table by probability inference, and weight coefficient table based on the probability value Select.

  Next, the navigation apparatus 1 searches for the minimum cost route from the departure point to the destination input by the driver using the obtained weighting factor table (S12). Subsequently, the navigation device 1 reads the learned route information from the learned route storage unit 16 (S14), and corrects the minimum cost route using the learned route information (S16). The navigation device 1 displays the searched route as a recommended route on the display 22 and performs route guidance.

  FIG. 7 is a diagram illustrating an operation of learning the table selection model by the navigation device 1. First, the vehicle starts to travel (S20), and the travel route of the vehicle is acquired using the vehicle position information detected by the vehicle position detector 18 while the vehicle is traveling (S22). It is detected that the vehicle has finished traveling (S24). Here, the end of traveling of the vehicle means, for example, that the vehicle has arrived at the destination and stopped. In the present embodiment, the navigation device 1 detects the end of traveling of the vehicle when the engine of the vehicle is turned off.

  Next, the navigation device 1 searches for a recommended route from the departure point to the destination (S26). Even when the route search is not instructed from the driver, the navigation device 1 searches for a recommended route in order to determine whether or not to perform learning. The route search process (S26) is the same as the route search process described with reference to FIG. The navigation apparatus 1 compares the searched recommended route with the travel route (S28). If the recommended route matches the searched route (YES in S28), the navigation device 1 does not learn.

  If the recommended route does not match the searched route (NO in S28), it is determined whether or not the travel route is a route that the driver often passes (S30). The learning possibility determination unit 32 determines whether or not the traveling route is a route that often passes based on the attribute of the number of times the links constituting the route pass. When it is determined that the travel route is a path that often passes (YES in S30), the navigation device 1 does not learn. If it is determined that the travel route is not a frequent path (NO in S30), the navigation device 1 learns the table selection model (S32).

  FIG. 8 is a diagram illustrating the learning operation of the table selection model. The model learning unit 34 reads all the weight coefficient tables from the weight coefficient storage unit 12 (S40). Here, the weight coefficient tables TA to TC are read out. Next, the model learning unit 34 searches for the minimum cost route from the departure point to the destination using any one of the weight coefficient tables (S42), and calculates the difference between the minimum cost route and the travel route (S44). ). The model learning unit 34 determines whether or not the above-described minimum cost path search and difference calculation have been performed for all the weight coefficient tables read from the weight coefficient storage unit 12 (S46). As a result of this determination, if the calculation has not been completed for all weight coefficient tables (NO in S46), the process proceeds to the process for calculating the minimum cost path using the uncalculated weight coefficient table (S42).

If all the weight coefficient tables have been calculated (YES in S46), the model learning unit 34 ranks the weight coefficient tables based on the obtained differences (S48). The model learning unit 34 gives higher ranks in order from the weighting coefficient table in which routes with small differences are calculated. The model learning unit 34 learns the table selection model using the given rank (S50).
The configuration and operation of the navigation device 1 according to the present embodiment have been described above.

  The navigation device 1 according to the present embodiment learns the table selection model when the travel route is a route that often passes even if the actual travel route from the departure point to the destination is different from the recommended route. Not performed. As a result, learning is not performed excessively by using a path that passes frequently as a teacher signal. Since the table selection model is appropriately learned using the appropriate travel route (teacher signal) selected based on the past number of passages, an appropriate evaluation function is selected according to the route search conditions during the route search. it can.

  In addition, the navigation device 1 according to the present embodiment searches for the minimum cost route using an appropriate weighting coefficient table according to the search condition, and then corrects the minimum cost route using the learned route learned as a path that passes frequently. Therefore, it is possible to search for a path that the driver is accustomed to running and use it as a recommended route.

(Second Embodiment)
Next, a navigation device according to a second embodiment of the present invention will be described. The navigation device according to the second embodiment recommends a path that passes frequently by performing the route search by including the attribute of the number of passages in the cost calculation of the route instead of correcting the minimum cost route by the learned route. It is searched as a route.

  FIG. 9 is a diagram illustrating a configuration of the navigation device 2 according to the second embodiment. The basic configuration of the navigation device 2 of the second embodiment is the same as that of the navigation device 1 of the first embodiment, but the learning route storage unit 16 and route learning described in the first embodiment. The part 36 is not provided. Moreover, the navigation apparatus 2 of 2nd Embodiment differs from the navigation apparatus 1 of 1st Embodiment in the content of the weighting coefficient table memorize | stored in the weighting coefficient memory | storage part 12. FIG.

  FIG. 10 is a diagram illustrating an example of a weighting coefficient table stored in the weighting coefficient storage unit 12. In the second embodiment, a weighting factor for the attribute of the number of passes is stored in the weighting factor table. That is, in the second embodiment, the evaluation function for obtaining the link cost uses the number of times the link passes as one of the parameters. The parameter of the number of passes is replaced so that the cost decreases as the number of passes increases. For example, when the number of times of passing the link is 10 times or more, the parameter is set to “1”, and when the number of times of passing the link is 4 times or more and less than 10 times, the parameter is set to “2”. In the case of less than 3 times, the parameter is set to “3”. As a result, the cost of the link decreases as the number of times the link passes, so that a route including the link having a higher number of passes is searched for as the minimum cost route.

  FIG. 11 is a diagram illustrating an operation of learning a table selection model by the navigation device 2 according to the second embodiment. The basic operation of the navigation device 2 of the second embodiment is the same as that of the navigation device 1 of the first embodiment. However, in the second embodiment, the recommended route search process (S17) is performed. It is different from the navigation device 1 of the first embodiment. The navigation device 1 according to the second embodiment calculates the cost of the route using the number of times the link has passed as one of the parameters, and searches for the minimum cost route. When the minimum cost route is obtained, correction by the learned route is not performed, and the minimum cost route is used as a recommended route search result.

  In the learning operation of the navigation device 2 of the second embodiment, after learning the table selection model (S22), the learning process is terminated without performing the route learning process.

  The configuration and operation of the navigation device 2 according to the second embodiment have been described above with a focus on differences from the navigation device 1 according to the first embodiment.

  As in the first embodiment, the navigation device 1 according to the second embodiment appropriately learns the table selection model using an appropriate travel route (teacher signal) selected based on the past number of passages. Therefore, an appropriate evaluation function can be selected according to the route search condition during the route search.

  Further, the navigation device 2 according to the present embodiment calculates the cost of the route using the number of times of passing the link as a parameter when searching for the minimum cost route using a weighting factor table appropriate for the situation. You can search for a path you are used to running and use it as a recommended route.

  The navigation apparatus and the navigation method of the present invention have been described in detail with reference to the embodiment, but the present invention is not limited to the above-described embodiment.

  In the above embodiment, the navigation device and the navigation method have been described. However, a program that causes the navigation device to execute each process of the navigation method is also included in the scope of the present invention.

  In the first embodiment described above, when the learned route storage unit 16 stores the learned route, the learned route storage unit 16 may associate and store information indicating a situation when the learned route is learned. Thereby, when searching for a recommended route, an appropriate learned route is read from the learned route storage unit 16 according to the situation, and the minimum cost route can be corrected using the read learned route.

  In the first embodiment described above, a configuration may be adopted in which information on a learned route is deleted when the learned route once stored in the learned route storage unit 16 is not passed for a predetermined period. As a result, it is possible to prevent a disadvantage that a recommended route is searched for by a route that is no longer used. When adopting a configuration in which learned route information is deleted, a predetermined period until the learned route information is deleted may be changed according to the distance from the home. For example, for a learning route in a region less than 10 km radius from home, the learning route information is deleted if it does not pass for one month, and for a learning route in a region more than 10 km radius from home, it passes for three months. If there is not, the learning route information is deleted. Places far from home are less frequent than places close to home, so it is a common path by changing the criteria for determining whether to use a learning route according to the distance from home. It is possible to appropriately determine whether or not.

  When there is a branch in the learned route stored in the learned route storage unit 16, information indicating where the branch has passed in the past may be stored. For example, if the number of right turns at a branch is 10 and the number of left turns is 15, information such as 40% right turn and 60% left turn is stored. Thereby, when searching for a recommended route using the learned route stored in the learned route storage unit 16, the recommended route corrected using an appropriate learned route can be searched based on the past passage record.

  In the second embodiment described above, an example in which the cost of a link is calculated using the number of times that the link has passed in the past has been described. However, the time required for passing through the link in the past is further used. The cost of the link may be calculated.

  FIG. 12A is a diagram showing an example of road attribute information stored in the map information storage unit 10 in an embodiment in which cost calculation is performed using the time required for passing a link, and FIG. 12B is a weighting factor storage. 6 is a diagram illustrating an example of a weighting coefficient table stored in a unit 12. FIG. As shown in FIG. 12A, as the road attribute, data of “required passage time” indicating the time required when the link was passed in the past is stored. When the link passes through the link a plurality of times, the required travel time may be, for example, an average value of the required time when the link is passed in the past or a mode value. As shown in FIG. 12 (b), a weighting factor for the required passage time is set in the weighting factor table, and the link cost is calculated using this weighting factor. Thereby, a route including a link having a short transit time is easily searched as a recommended route.

  In the above-described embodiment, the first embodiment for searching for a recommended route by correcting the minimum cost route with a learned route, and the second embodiment for obtaining the minimum cost route using the number of times of passing of a link as a parameter will be described. However, you may employ | adopt the structure which combined 1st Embodiment and 2nd Embodiment. That is, after obtaining the minimum cost route using the number of times of passing the link as a parameter, the recommended route may be obtained by performing correction using the learned route stored in the learned route storage unit 16.

  ADVANTAGE OF THE INVENTION According to this invention, learning is performed so that an evaluation function can be selected appropriately according to route search conditions, and it is useful as a car navigation device or the like that performs route search.

It is a figure which shows the structure of the navigation apparatus of 1st Embodiment. (A) It is a figure which shows the example of the node and link which comprise a road. (B) It is a figure which shows the example of the attribute information of a road. It is a figure which shows the example of a weighting coefficient table. It is a figure which shows the example of a table selection model. (A) It is a figure which shows the example of a cost minimum path | route and a learning route | root. (B) It is a figure which shows the example of a recommendation path | route. It is a figure which shows the operation | movement of the route search performed with a navigation apparatus. It is a figure which shows the operation | movement of the whole learning of the table selection model performed with a navigation apparatus. It is a figure which shows the detailed operation | movement of learning of the table selection model performed with a navigation apparatus. It is a figure which shows the structure of the navigation apparatus of 2nd Embodiment. It is a figure which shows the example of a weighting coefficient table. It is a figure which shows the operation | movement of the whole learning of the table selection model performed with a navigation apparatus. (A) It is a figure which shows the example of the attribute information of the road memorize | stored in the map information storage part. (B) It is a figure which shows the example of the weighting coefficient table memorize | stored in the weighting coefficient memory | storage part.

Explanation of symbols

1, 2 Navigation device 10 Map information storage unit 12 Weight coefficient storage unit 14 Table selection model storage unit 16 Learning route storage unit 18 Vehicle position detection unit 20 Switch 22 Display 24 Control unit 26 Table selection unit 28 Route search unit 30 Travel route acquisition Unit 32 Learning availability determination unit 34 Model learning unit 36 Route learning unit

Claims (12)

A navigation device that calculates a cost of the route using an evaluation function having a plurality of road attributes of links constituting a route as parameters, and searches for an optimal route from a starting point to a destination based on the cost,
An evaluation function storage unit storing a plurality of evaluation functions for calculating the cost;
An evaluation function selection unit that selects one evaluation function corresponding to the condition of the route search from the plurality of evaluation functions based on a predetermined relationship between the condition of the route search and the evaluation function to be used;
A recommended route search unit that obtains a route cost using the evaluation function and searches for a recommended route from a starting point to a destination based on the cost of the route;
A travel route acquisition unit that acquires the travel route of the vehicle from the departure point to the destination;
It is determined whether or not the predetermined relationship is to be corrected based on the number of times that the link constituting the travel route has been passed in the past, and when it is determined that the predetermined relationship is to be corrected, the route search condition and the A learning unit for correcting the predetermined relationship based on information on a travel route;
Equipped with a,
The learning unit calculates a plurality of routes corresponding to each evaluation function, ranks the plurality of routes based on a difference between the plurality of routes and the travel route, and evaluates a plurality of evaluations based on the ranking. A navigation device that updates the rate at which functions are selected . A learning route storage unit that stores, as a learning route, a route that has passed a predetermined number of times,
The navigation device according to claim 1 , wherein the recommended route search unit searches for the recommended route by correcting a route obtained based on a cost with a learned route stored in the learned route storage unit. The navigation device according to claim 2 , wherein the learned route is deleted from the learned route storage unit when the learned route stored in the learned route storage unit is not passed for a predetermined period. The navigation apparatus according to claim 3 , wherein the predetermined period until the learned route is deleted is set according to a distance from the home of the learned route. The learning route storage unit, when storing the learning route having branching, navigation device according to claim 3 or 4 for storing information about the passage performance of the branch. The learning route storage unit, a navigation device according to any one of claims 3-5 for storing information indicating the condition of the route search when through the learning route in association with the learned routes. The navigation device according to any one of claims 3-6 comprising a learning route display unit that displays the learned routes stored in the learning route storage unit on a map. The road attribute includes information indicating the number of times the link has been passed in the past,
Each evaluation function, the navigation device according to any one of claims 1 to 7, the cost as the number of times passing through the link is set to be smaller. The navigation device according to claim 8 , wherein the information on the number of times the link has passed in the past is information indicating the number of times the link has passed in the most recent predetermined period. The road attribute includes information indicating the time required to pass the link in the past,
Each evaluation function, the navigation device according to any one of claims 1 to 9, wherein the required time is set as the cost shorter decreases. A navigation method for calculating a cost of the route using an evaluation function having a plurality of road attributes of links constituting the route as a parameter, and searching for an optimum route from a starting point to a destination based on the cost,
Preparing an evaluation function storage unit storing a plurality of evaluation functions for calculating the cost;
Selecting one evaluation function corresponding to the route search condition from the plurality of evaluation functions based on a predetermined relationship between the route search condition and the evaluation function to be used;
Obtaining a cost of a route using the evaluation function, and searching for a recommended route from a starting point to a destination based on the cost of the route;
Obtaining a vehicle travel route from the departure point to the destination;
It is determined whether or not the predetermined relationship is to be corrected based on the number of times that the link constituting the travel route has been passed in the past, and when it is determined that the predetermined relationship is to be corrected, the route search condition and the Correcting the predetermined relationship based on travel route information;
Equipped with a,
In the learning step, a plurality of routes corresponding to the respective evaluation functions are calculated, the plurality of routes are ranked based on a difference between the plurality of routes and the travel route, and a plurality of evaluations are performed based on the ranking. A navigation method that updates the rate at which functions are selected . A program for calculating a cost of the route by using an evaluation function having a plurality of road attributes of links constituting the route as parameters, and for searching for an optimum route from a starting point to a destination based on the cost. To the computer,
Based on a predetermined relationship between a route search condition and an evaluation function to be used, one evaluation function corresponding to the route search condition is extracted from an evaluation function storage unit storing a plurality of evaluation functions for calculating the cost. And steps to
Obtaining a cost of a route using the evaluation function, and searching for a recommended route from a starting point to a destination based on the cost of the route;
Obtaining a vehicle travel route from the departure point to the destination;
It is determined whether or not the predetermined relationship is to be corrected based on the number of times that the link constituting the travel route has been passed in the past, and when it is determined that the predetermined relationship is to be corrected, the route search condition and the Correcting the predetermined relationship based on travel route information;
Was executed,
In the learning step, a plurality of routes corresponding to the respective evaluation functions are calculated, the plurality of routes are ranked based on a difference between the plurality of routes and the travel route, and a plurality of evaluations are performed based on the ranking. A program that updates the rate at which functions are selected .

Images