Patent Application
20190063938
This application is based upon and claims the benefit of priority from the prior Japanese Patent Application No. 2017-167861, filed on Aug. 31, 2017, the entire contents of which are incorporated herein by reference.
Embodiments described herein relate to a route estimation apparatus, a route estimation method and a computer program.
A mobile body such as an electric car, an airplane, a drone, an electric motorcycle and a car equipped with a diesel engine needs to be supplied with energy at an energy supply point. A mobile body needs to use the supplied, limited energy to arrive at the next energy supply point. Since energy supply points are geographically limited, it is required to be able to predict energy consumption of the mobile body to decide the next energy supply point of the mobile body. A lot of methods therefor have been proposed.
As a method for predicting energy consumption, it is common to build an energy consumption model and make prediction using the built model. It is necessary to estimate parameters of the model to build the model. At the time of estimating the parameters, history information which includes information of energy consumption at the time of a mobile body traveling on a movement route and information of the movement route is necessary as learning data.
As a method for acquiring the history information, it is conceivable to acquire the history information directly from a mobile body. In this method, however, it is necessary to individually communicate with each mobile body to acquire the history information. Therefore, cost for communication with mobile bodies, management cost and data collection cost are high. Especially, since it is necessary to acquire the history information from a lot of mobile bodies in order to improve prediction accuracy, the costs become enormous.
According to one embodiment, a route estimation apparatus includes an information communicator and a route estimator. The information communicator is configured to acquire position history information of a mobile body from an information apparatus via a communication network. The route estimator is configured to estimate a movement route used by the mobile body to move from a first position to a second position, using the position history information and route information indicating routes among a plurality of positions and characteristics of the routes.
Embodiments of the present invention will be described below with reference to drawings. Though the description below will be made on a case where a mobile body is an electric vehicle (hereinafter referred to as an EV) as an example, the present invention can be practiced in any other mobile body than an EV. As examples of such a mobile body, an airplane, a drone, an electric motorcycle and a car equipped with a diesel engine are given.
An EV travels on a road to move to various points using electric power charged in a battery (charged energy). In the present embodiments, especially energy supply points where an EV is charged will be dealt with as the points. The points in the present embodiment, however, are not limited to the energy supply points but may be arbitrary points on a map. For example, a home, a restaurant, a store and a point arbitrarily specified by an EV user are possible.
An EV is supplied with electric power at an energy supply point and accumulates the supplied electric power in a battery. The EV moves using the electric power accumulated in the battery. The EV needs to move to the next energy supply point to be supplied with electric power there before running out of the electric power of the battery. Thus, the EV is to be charged at each energy supply point before arriving at a destination. An energy supply point is provided with one or more chargers (energy suppliers), and the EV wiredly or wirelessly connects to any of the chargers to be supplied with electric power. The present embodiment assumes such a situation.
Since the present embodiment assumes that a mobile body is an EV, energy supply refers to charging, an energy supply point to a charging stand and an energy supplier to a charger. However, a process similar to that of the present embodiment is also possible for any other mobile body than an EV, for example, by replacing the charging, charging stand and charger with fuel supply, gasoline station and fuel supplier if a mobile body is a gasoline-powered vehicle.
The inputting circuit 16 is an input interface for a user of the present apparatus (hereinafter, an operator) to input parameter design information required for an operation of the present apparatus or input an instruction.
The outputting circuit 17 is an output interface for displaying information or data generated by the present apparatus.
The information communicator 18 acquires information of one or more EVs from one or more information apparatuses 19 different from an EV via a communication network 20, and stores the acquired information into an energy supply history database (DB) in the energy supply history storage 21. The communication network 20 is a wired or wireless network or a wired/wireless hybrid network. The information apparatus 19 is, for example, a charger installed at an energy supply point (hereinafter referred to as a supply point) where an EV is charged, a terminal apparatus used by an EV user (a smartphone, a tablet apparatus, a notebook PC or the like) and an ETC2.0 apparatus. The ETC2.0 apparatus is an example of a data communication device communicable with at least one of a user's terminal apparatus and EV (for example, a car navigation device mounted on the EV) and is arranged at each of a plurality of predetermined spots in a traveling area. The predetermined spots may be an ITS spot, a position along a route, a highway service area, a building and the like. In the case of providing a gateway apparatus communicable with one or more chargers installed at energy supply positions to collect information of the one or more chargers from the gateway apparatus, the case corresponds to one form of an operation of collecting information from chargers.
The energy supply history storage 21 stores the energy supply history DB.
The energy supply history DB holds information of an EV identifier (ID), supply position IDs, start amounts of charge and end amounts of charge as energy supply history information of each EV.
The EV ID is an identifier for recognizing each of EVs (EV1, EV2, EV3, . . . , EVi). For example, an ID of EVi is “123456”.
The supply position IDs are identifiers identifying energy supply positions (hereinafter referred to as supply positions) where the EV was charged. Where on a map the supply positions exist can be separately grasped by map information. For example, the map information holds each supply position ID and coordinates on the map in association with each other. Here, it is assumed that there are “n” supply positions, and IDs of the “n” supply positions are indicated by Q1, Q2, Q3, . . . , Qn. The supply positions are positions which the EV stopped by and are an example of positions where the EV was located. The energy supply history information is an example of position history information indicating a history of the positions where the EV was located. In the case of collecting information from chargers, the energy supply history information is obtained by collecting information set for or recorded to the chargers. In the case of collecting information from smartphones, users may be allowed to input the information to applications of smartphones in any manner so that the information can be collected. For example, the users may manually input supply position IDs or may wirelessly or wiredly receive the supply position IDs from chargers or other apparatuses installed at supply positions and automatically input the received information to the applications of the smartphones. Further, information of the positions is not limited to supply position IDs. Coordinates on the map may be acquired by GPS to cause the acquired coordinates to be position information. In the case of collecting information of a position of a kind different from a supply position (for example, a home, a store or the like), coordinates of the position on the map may be acquired. If association with map information is possible, a name of a place, a facility or the like associated with the coordinates on the map may be acquired. The acquired information may be manually inputted to an application of a smartphone by a user.
The start amount of charge is an amount of battery charge at the time when each EV starts charging with a charger at a supply position. The end amount of charge is an amount of battery charge at the time when the EV ends charging with the charger at the supply position. Though the start amount of charge and the end amount of charge are expressed in units of kWh in the example of
In the example of
Instead of or in addition to the start amount of charge and the end amount of charge, a start SOC (State of Charge) and an end SOC may be held. The SOC indicates a rate of actually accumulated electric energy to a battery capacity (a full charge capacity).
Battery state information of an EV according to the present embodiment may indicate at least one of the start amount of charge, the end amount of charge, the start SOC, the end SOC and the electric energy (the amount of energy) charged in an EV. The battery state information may be any other value if a battery state of an EV can be identified by the value. For example, in the case of acquiring EV information with an application of a smartphone or an ETC2.0 apparatus, a battery state (an amount of charge or SOC) at the position or timing may be acquired.
In addition, the energy supply history DB can hold at least one of a vehicle type of each EV, a date and time of starting charging, a date and time of ending charging and a full charge capacity.
If the date and time of starting charging or the date and time of ending charging does not exist in the energy supply history DB, entries of the DB may be chronologically arranged. Further, information of the energy supply history DB may be past information collectively acquired. Otherwise, information may be acquired in real time and sequentially added to the energy supply history DB.
The route information storage 22 stores a route information database (DB).
The route information DB holds a supply position ID, an ID of an adjacent supply position (an adjacent supply position ID) and distance information as route information. That a supply position is adjacent to an adjacent supply position means that there is no other supply position on a route from the supply position to the adjacent supply position. The actual route from the supply position to the adjacent supply position can be identified from map information. An EV can move from a supply position to an adjacent supply position using a route between the supply position and the adjacent supply position. There may be a plurality of adjacent supply positions for a supply position. Further, there may be one route from a supply position to an adjacent supply position, or there may be a plurality of routes to an adjacent supply position. In description below, it is assumed that there is one route from a supply position to an adjacent supply position.
The distance information indicates a distance from a supply position to an adjacent supply position. A distance is an example of a characteristic of a route.
Though the examples in
An adjacency matrix is applicable to a case where a route from one of two supply positions to the other is different from a route from the other to the one. For example, when it is assumed that the two supply positions are Q2 and Qj, and a route from the supply position Q2 to the supply position Qj and a route from the supply position Qj to the supply position Q2 are not the same, an adjacency matrix is applicable. Up and down lanes of a highway corresponds to an example of such a case. Similarly, in the case of a table format as in
In an adjacency matrix, it is possible to have what is other than a supply position ID as a row name and a column name.
It is also possible to configure the route information using a common network structure description method.
When a traveling area is divided in a plurality of areas, the route information may be held for each divided area. Of course, it is possible to hold one piece of route information for the whole traveling area.
In the network structure in
The information acquirer 10 acquires an entry (energy supply history information) of at least one EV from the energy supply history DB. In the acquired energy supply history information, a history of energy supply at at least two supply positions is held for the EV. Further, the information acquirer 10 acquires route information from the route information DB. If route information is managed for each area, route information to which a supply position included in the acquired energy supply history information belongs to is acquired. The information acquirer 10 hands over the acquired energy supply history information and route information to the information processor 12.
The information processor 12 receives the energy supply history information and route information of the EV from the information acquirer 10. The information processor 12 estimates a movement route of the EV using these pieces of information and the route estimator 13. The movement route is a route or a combination of routes which the EV used to move from a certain supply position to another supply position.
First, the information processor 12 identifies a supply position Qj′ where charging was performed after a certain supply position Qj (hereinafter referred to as the next supply position Qj′) using the received energy supply history information. The certain supply position Qj corresponds to a first position as an example, and the next supply position Qj′ corresponds to a second potion as an example. The information processor 12 calculates a difference between an end amount of charge at the supply position Qj and a start amount of charge at the time of charging at the next supply position Qj′. The calculated difference is set as an amount of energy consumption Eq for one movement of the EV. If a date and time of starting charging and a date and time of ending charging are stored in the energy supply history DB, the next supply position Qj′ can be identified highly accurately.
The information processor 12 generates a query which includes the ID of the supply position Qj and the ID of the next supply position Qj′ and hands over the query to the route estimator 13. This query requests estimation of a movement route which the EV used to move from the supply position Qj to the next supply position Qj′. An identifier (a history ID) may be included in the query. Further, the information processor 12 hands over route information acquired from the information acquirer 10 to the route estimator 13. Instead of the information processor 12 handing over the route information, the route estimator 13 may acquire the route information from the route information DB by being triggered by receiving the query.
Using the ID of the supply position Qj, the ID of the next supply position Qj′ and the route information received from the information processor 12, the route estimator 13 estimates a movement route which is a route or a combination of routes which the EV used to move from the supply position Qj to the next supply position Qj′ (movement route estimation). If there are a plurality of movement route candidates, one of the candidates is selected. A method for the estimation will be described later.
As an example, the estimated movement route can be expressed as a list in which identifiers of a plurality of supply positions are chronologically arranged (a supply position list). For example, if the EV passed through the supply positions Q4, Q8, . . . after the supply position Qj and then reached the next supply position Qj′, the supply position list is expressed as information like “Qj→Q4→Q8→ . . . →Qj′”. That an EV passes through a supply position means that the EV travels on a route on which the supply position exists or on a route near the supply position, passing the supply position without charging there. In the case of “Qj→Q4→Q8→ . . . →Qj′”, it means that, after charging at the supply position Qj, the EV is charged at the next supply position Qj′ without charging at the supply positions Q4 or Q8.
For the estimated movement route, the route estimator calculates values of one or more items determined in advance, such as a distance of the movement route, time required to move on the movement route and gradient information of the movement route, and generates movement route estimation information which includes the calculated values, as information of the estimated movement route. The route estimator 13 returns the movement route estimation information to the information processor 12. Information identifying the estimated movement route (the above-described list in which nodes of supply positions are chronologically combined, or the like) may be included in the movement route estimation information.
The information processor 12 generates learning data in which the amount of energy consumption Eq and the movement route estimation information received from the route estimator 13 are associated with a history ID. If an item which is not necessary for model learning exists in the movement route estimation information, the item may be removed. The information processor 12 stores the generated learning data into a learning data database (DB) of the learning data storage 23. Thereby, learning data for one time is stored for the EV. The learning data storage 23 stores the learning data DB.
By the information processor 12 performing a plurality of movement route estimations for a plurality of EVs, a plurality of pieces of learning data on the plurality of EVs are stored in the learning data DB.
The query acquirer 31 acquires a query which includes a supply position Qj where a certain EV was supplied with energy, a supply position Qj′ where the EV was supplied with energy next, and a history ID from the information processor 12.
Further, the route estimator 13 acquires route information from the information processor 12 or the route information DB. In the case of dividing a traveling area in a plurality of areas and managing route information for each divided area, route information of an area to which both of the supply position Qj and the supply position Qj′ belong is acquired. It is assumed here that the route information is expressed in an adjacency matrix (see
The shortest route estimator 32 calculates the shortest movement route (the shortest route) from the supply position Qj to the next supply position Qj′ using the supply position Qj, the next supply position Qj′ and the adjacency matrix. As a method for solving a shortest route problem, the Bellman Ford algorithm, the Gabow algorithm, the Warshall-Floyd algorithm and the like, including the Dijkstra's algorithm which is common, can be used. The shortest route estimator 32 generates movement route estimation information of the determined shortest route.
The configuration of the movement route estimation information in
There is a possibility that two or more shortest routes from the supply position Qj to the next supply position Qj′ exist. In this case, one shortest route may be randomly selected from among the two or more shortest routes. Further, one shortest route may be selected based on an arbitrary condition. For example, such a shortest route that time required for movement from the supply position Qj to the next supply position Qj′ is the shortest may be selected. Otherwise, a shortest route with the smallest cumulative altitude difference may be selected. Otherwise, a shortest route where the weather is the best may be selected.
The energy consumption model builder 14 (hereinafter referred to as the model builder 14) reads learning data from the learning data DB and constructs an energy consumption model (hereinafter referred to as a model) using the read learning data. As a method for constructing the model, there are a lot of methods such as artificial intelligence, machine learning, black-box modeling and such white-box modeling that defines a physical model. The black-box modeling is a method for performing modeling using regression, a neural network, SVM or statistics when characteristics of a target are unknown. The white-box modeling is modeling performed by defining a physical model and the like when characteristics of a target are known. In the present embodiment, any model may be built with any model method.
The model builder 14 may build a model by receiving instruction information specifying necessary information from the operator through the inputting circuit 16 or may build a model according to information specified beforehand. As an example, the operator specifies a model type (a form of a regression model or the like) and an optimization algorithm used for estimation of model parameters as the instruction information.
The model builder 14 reads a basic function corresponding to the model type from a predetermined storage device (a memory, a hard disk device, an SSD or the like). The predetermined storage stores a basic function for each of a plurality of model types, and the model builder 14 reads a basic function corresponding to a relevant model type. Values of model parameters of the basic function is unknown at this point of time, and the model builder 14 estimates the values of the model parameters included in the read basic function using the learning data. The model builder 14 stores a set of the read basic function (the model type) and the estimated model parameters into an energy consumption model database (DB) as an energy consumption model (a model).
If the model type is a multiple regression model, the basic function is expressed as shown below.
[Formula 1]
y=w
0
+w
1
x
1
+w
2
x
2
+w
3
x
3
+ . . . +w
n
x
n (1)
Here, “w0, w1, w2, w3, . . . , wn” are model parameters to be estimated; “x1, x2, x3, . . . , xn” are input variables (explanatory variables); and “y” is an output variable (a purpose variable). In order to absorb differences among measurement units of the explanatory variables, the purpose variable and all the explanatory variables may be normalized to a mean value of 0 and a variance of 1 (scaling). As examples of the explanatory variables, distance, required time and outdoor temperature are given. These are items included in the learning data. For example, “x1”, “x2” and “x3” are distance, required time and outdoor temperature, respectively. The explanatory variables may be different values calculated from a plurality of items included in the learning data. For example, a velocity obtained by dividing a distance by required time may be an explanatory variable. The model parameters can be determined by a well-known optimization algorithm such as a maximum likelihood method and a least squares method.
The energy consumption model DB stores the model generated by the model builder 14.
The route estimation apparatus 1 in
The information acquirer 10 acquires energy supply history information and route information (S101).
The information acquirer 10 sends the acquired energy supply history information and route information to the information processor 12 (S102).
The information processor 12 extracts, for a certain EV (a mobile body), information identifying a supply position and the next supply position from the energy supply history information (S103). Further, an amount of energy the mobile body consumed to move from the supply position to the next supply position is calculated.
The information processor 12 sends the extracted information and the route information to the route estimator 13 (S104).
The route estimator 13 estimates, for the certain EV, a movement route used to move from the supply position to the next supply position (S105).
The route estimator 13 sends information of the estimated movement route (movement route estimation information) to the information processor 12 (S106).
The information processor 12 generates learning data based on the movement route estimation information (S107). Specifically, the information processor 12 extracts values of items required for model building from the movement route estimation information and generates learning data which includes the extracted values and the amount of energy consumption.
The model builder 14 builds an energy consumption model (a model) based on the learning data (S108). After that, the controller 15 may display the model on the outputting circuit 17. Further, information showing the movement route estimated by the route estimator 13 (a supply position list or the like) may be displayed on the outputting circuit 17.
As described above, according to the present embodiment, a first position of a mobile body and a second position, which is a movement destination after the first position, are identified based on energy supply history information (position history information) of the mobile body. A movement route used by the mobile body to move from the first position to the second position is estimated with route information (see
The present embodiment shows an example of generating route information. In the present embodiment, the route information is generated with traffic control data (traffic observation data).
The TC information DB holds data in which a supply position, an adjacent supply position and IDs of one or more TCs between the supply positions are mutually associated.
The route information generation apparatus 41 is provided with a traffic control data acquirer 42 and an adjacency matrix generator 43. The traffic control server 51 is provided with a velocity information storage 52. The velocity information storage 52 stores a velocity information database (DB).
The traffic control server 51 is connected to traffic counters TC1 to TCg arranged at “g” positions via a communication network. The traffic control server 51 communicates with the traffic counters TC1 to TCg to acquire traffic control data which is data of the traffic counters TC1 to TCg. Here, velocity information is acquired as the traffic control data. The velocity information holds a value of a velocity [km/h] measured at certain time in a certain TC (traffic counter). The traffic control server 51 stores the velocity information into the velocity information DB. Instead of the velocity information, information of an occupancy rate or a traffic volume may be acquired as the traffic control data.
The traffic control data acquirer 42 of the route information generation apparatus 41 is connected to the traffic control server 51 via a communication network and acquires velocity information from the traffic control server 51 as traffic control data. The traffic control server 51 and the route information generation apparatus 41 may be the same apparatus. In this case, the traffic control data acquirer 42 can read the velocity information from the velocity information DB. The traffic control data acquirer 42 hands over the acquired velocity information to the adjacency matrix generator 43.
The adjacency matrix generator 43 acquires adjacency relationship data stored in the route information storage 22 from the route estimation apparatus 1. Further, the adjacency matrix generator 43 acquires TC information stored in the TC information DB from the route estimation apparatus 1.
The adjacency matrix generator 43 calculates, for supply positions in an adjacency relationship, a route characteristic using the velocity information and the TC information and generates route information by adding the calculated route characteristic to the adjacency relationship data. Here, an adjacency matrix is generated as the route information.
As the route characteristic, the adjacency matrix generator 43 calculates a statistical value of velocity values of traffic counters existing between a certain supply position Qj and an adjacent supply position Qj′, a sum of tracking time or the like. As examples of the statistical value, a mean value, a median, a maximum value and a minimum value are given. For example, for all traffic counters existing on a relevant route, an average velocity is calculated for data in the velocity information DB corresponding to a predetermined time and the average velocity is made characteristic information of the route. The adjacency matrix generator 43 adds the calculated characteristic information to a relevant element of an adjacency matrix. By performing this process for all combinations of supply positions which are in an adjacency relationship, the adjacency matrix is generated. The sum of tracking time is required traveling time calculated in the case of specifying a range covered by each traffic counter and assuming that a vehicle has traveled at a velocity measured by the traffic counter (for example, a maximum value) within the covered range.
The adjacency matrix generator 43 transmits the generated adjacency matrix to the route estimation apparatus 1. The route estimation apparatus 1 stores the received adjacency matrix into the route information DB as route information. This route information becomes the route information used in the first embodiment.
For example, the route information generation apparatus 41 acquires traffic control data (velocity information) at predetermined time intervals and generates (updates) an adjacency matrix. The acquired velocity information is held in an internal storage of the route information generation apparatus 41 or an external storage accessible from the route information generation apparatus 41. Therefore, a plurality of adjacency matrices (pieces of route information) corresponding to velocity information acquisition times are accumulated in the route information DB.
The information communicator 18 (see
Though the route information generation apparatus 41 and the route estimation apparatus 1 are separate apparatuses, the route information generation apparatus 41 may be configured integrally with the route estimation apparatus 1.
Further, the controller 15 of the route estimation apparatus 1 may acquire data of TCs from the TCs or a management apparatus managing the TCs through communication and build a velocity information DB using the acquired data. The controller 15 generates route information using the generated velocity information DB and adjacency relationship data and stores the generated route information into the route information DB.
A method for selecting the shortest route is used as a route estimation method in the first embodiment. In the present embodiment, however, an estimation method is decided according to a condition from among a plurality of movement route estimation methods.
An example of performing movement route estimation mainly for one EV is shown in the first or second embodiment. In the present embodiment, however, a situation is assumed in which movement route estimation is performed for a plurality of EVs. Some examples of a condition for deciding an estimation method from among a plurality of estimation methods will be shown below.
The route estimator 13 or the information processor 12 randomly selects an estimation method according to a certain parameter “α”. The parameter “α” specifies a probability of each estimation being selected. For example, when there are two estimation methods A and B, it is specified by the parameter “α” that the estimation methods A and B are selected with a probability of 40% and a probability of 60%, respectively. The parameter “α” is specified by the operator of the present apparatus and is inputted with the inputting circuit 16.
As an example, the estimation method A is a method of selecting a movement route with the shortest distance, and the estimation method B is a method of selecting a movement route with the smallest gradient change (cumulative altitude difference or the like). A user (a driver of an EV) really does not necessarily select the shortest route, and it is conceivable that a user who selects a movement route with a small gradient change exists.
Otherwise, it is conceivable that the estimation method A is the method of selecting a movement route with the shortest distance, and the estimation method B is a method of selecting a movement route with the second or third shortest distance.
Otherwise, it is also conceivable that the estimation method A is the method of selecting a movement route with the shortest distance and the estimation method B is a method of selecting a movement route corresponding to a traffic situation (for example, a movement route which does not pass through a part where a traffic volume is equal to or above a predetermined value).
Thus, by appropriately switching an estimation method to be used, movement route estimation that is more suitable for an actual situation becomes possible. In the case of selecting an estimation method from among three or more estimation methods, the selection can be similarly performed.
An estimation method to be used is selected according to at least one of conditions of a day of the week, a time zone and a place. For example, a plurality of conditions about a day of the week, a time zone and a place, such as “up lane toward Tokyo in the evening on Sunday”, are prepared, and an estimation method is associated with each condition. At the time of estimating a movement route, a condition satisfied by an energy supply history to be processed is identified among the plurality of conditions, and an estimation method corresponding to the identified condition is used.
An estimation method to be used is selected based on a model of a car navigation apparatus. A method for supporting a user in selection of a route differs according to a model of a car navigation apparatus. Therefore, an estimation method is selected based on a type of a car navigation apparatus mounted on an EV. For example, an estimation method A is used in the case of the latest car navigation apparatus compatible with a certain standard or specifications, and an estimation method B is used in the case of an old car navigation apparatus which is not compatible with the standard or specifications. As an example, the estimation method A is a method of selecting a movement route with the smallest sum of tracking time, and the method B is a method of selecting a movement route with the shortest distance. Further, there may be a case where, even in the case of the same car navigation apparatus, a supporting method differs according to a model of a vehicle on which the car navigation is mounted. Therefore, an estimation method to be used may be selected based on a model of a car navigation apparatus and a model of a vehicle.
An estimation method to be used is decided based on data (for example, velocitys) of traffic counters installed on routes among a plurality of supply positions. For example, in a case where there are a plurality of candidates for a movement route from a supply position to the next supply position, a mean value of velocitys of traffic counters existing on each movement route candidate is calculated. Variation (variance) of the mean values among the candidates is calculated, and an estimation method A is used if the variation is equal to or above a threshold, and an estimation method B is used if the variation is small. The estimation method A is, for example, an estimation method of selecting a movement route with the shortest required time, and the estimation method B is an estimation method of selecting a movement route with the shortest distance. Each of the estimation methods A and B may be any other method than the examples given here.
Movement route estimation is performed with a plurality of estimation methods. The number of methods the estimated movement routes of which are the same (that is, the number of estimation methods which have outputted the same movement route) is counted, and a movement route for which a value of the counting result is the largest is selected. The plurality of estimation methods can be arbitrary methods. The estimation methods given in the above first to fourth examples may be combined.
In the present embodiment, a confidence coefficient (a weight) is given to an estimated movement route, and the given confidence coefficient is used at the time of constructing a model. As an example, the confidence coefficient is expressed by a real number. An example of calculating the confidence coefficient is shown below.
Movement route estimation is performed with a plurality of estimation methods. A movement route is selected from among estimated movement routes by an arbitrary method. How many movement routes which are the same as the selected movement route are included among all the estimated movement routes (that is, the number of estimation methods which have outputted the same movement routes at the selected movement route) is counted. A confidence coefficient is calculated from a ratio of the counted value to the number of all the estimated movement routes (that is, the number of used estimation methods). Specifically, the confidence coefficient is calculated by the following formula.
Confidence coefficient=(the number of estimation methods which have outputted selected route/the number of all estimation methods)
Another example of calculating the confidence coefficient will be shown. The number of branch points included in each of the estimated movement routes is counted. For example, in the example of
Confidence coefficient=1/the number of branch points existing before reaching next supply position
A method for utilizing the confidence coefficient will be described below.
In the case of deciding model parameters by a least squares method, a weighted least squares method using a confidence coefficient as a weight can be implemented. For example, as shown in Formula (2) below, model parameters can be decided so that a sum of squares of residuals “3” is minimized with a confidence coefficient “e”. If the multiple regression model described before is assumed, “y” corresponds to a purpose variable, and “f(x)” corresponds to linear combination of explanatory variables and coefficients (see the right-hand side of Formula (1)). Here, “i” indicates a number of learning data (a sample). As the value of the confidence coefficient is larger, a larger weight is given. Thereby, it becomes possible to calculate highly accurate model parameters.
Further, in the case of deciding model parameters by a boosting method, the confidence coefficient can be used as a weight at the time of generating an initial value of occurrence distribution of samples used in boosting.
A case is assumed where, for some EVs, actual movement histories can be acquired by an ETC2.0.
It is assumed that, at the time of estimating a movement route of a certain EV, the route estimator 13 of the route estimation apparatus 1 can identify, when another EV moves from the same supply position to the same next supply position in the same time zone as the EV, a movement route of that another EV from movement route history data. In this case, the movement route of that another EV is used as the movement route of the EV which is the estimation target. The time zone refers to, for example, any one of a plurality of time zones obtained by dividing one day. A condition about the time zone may not be set.
In the present embodiment, an example of acquiring energy supply history information to be stored into the energy supply history DB in
At the time of creating learning data from the energy supply history DB by movement route estimation, an energy supply history which satisfies each of the following conditions may be used. By adding such a condition, it is possible to build a highly accurate model and enhance accuracy of prediction of an amount of energy consumption. It is also possible to combine the following conditions.
An energy supply history is to be extracted from the energy supply history DB for the same user under the condition that a difference between a date and time of ending use of a supply position and a date and time of starting use of the next supply position should be within “X” hours. Here, “X” may be an arbitrary value. For example, “X” may be 6. If this condition is not satisfied, the energy supply history is not used to generate learning data (that is, movement route estimation is not performed). For example, in the example of
A sum of tracking time and a difference between a date and time of ending use of a charger and a date and time of starting use of the next supply position are compared, and a difference between the sum of tracking time and the difference is “Y” hours or more, the energy supply history is not used. Here, “Y” may be an arbitrary value. For example, “Y” may be 2 or 3.
In the present embodiment, an example of factor analysis using an energy consumption model will be shown. Though a regression model is used as the energy consumption model, a statistical/machine learning model and a physical model, which are used for general factor analysis such as ANOVA (analysis of variance), main component analysis and factor analysis, are also usable.
[Formula 3]
y=f(x)=wx (3)
As shown below, “y” in (Formula 3) indicates amounts of energy consumption (purpose variables) of “R” estimated movement routes.
y={y
r
: r=1, . . . ,R} [Formula 4]
As shown below, “x” indicates “E” explanatory variables thought to be related with the amounts of energy consumption of the “R” estimated move routes, respectively.
x={x
r
: r=1, . . . ,R} [Formula 5]
As shown below, “xr” indicates “E” explanatory variables for the “r-th” piece of learning data.
X
r
={x
r,e
:e=1,E} [Formula 6]
As shown below, model parameters “w” are calculated with a maximum likelihood method or least squares method as stated in the description of the first embodiment.
w={w
e
: e=1, . . . ,E} [Formula 7]
In the present embodiment, prediction of energy consumption utilizing an energy consumption model will be described.
An energy consumption model calculated in any of the first to seventh embodiments is transmitted from the route estimation apparatus 1 to a prediction server 61 and stored into the prediction server 61. The prediction server 61 predicts an amount of energy consumption for a certain EV using the energy consumption model. Further, calculation of a travelable distance (a cruising range) and the like are also performed based on the predicted amount of energy consumption.
The prediction server 61 predicts energy consumption. For example, a movement distance from a supply position where charging was performed last is calculated from the GPS information and the energy supply history, and an amount of energy consumption is predicted from the movement distance and an energy consumption model. If explanatory variables of the energy consumption model include an item other than movement distance, for example, outdoor temperature, information of outdoor temperature is also acquired from the communication terminal A1 or another server. The prediction server 61 calculates a current amount of residual energy from the predicted amount of energy consumption and calculates a movable distance (a travelable distance) with the amount of residual energy. Further, if the amount of residual energy is below a threshold, the prediction server 61 looks for a candidate for the next supply position. The candidate for the next supply position is a supply position existing within a reachable range with the amount of residual energy. The prediction server 61 predicts energy consumption required to move to each supply position and identifies a supply position which can be reached with the amount of residual energy as the candidate for the next supply position (a recommended supply position). The prediction server 61 transmits output information showing the recommended supply position, the amount of energy consumption and the travelable distance to the communication terminal A1. The communication terminal A1 displays the received output information on its screen. The operation of the prediction server 61 which has been described here is a mere example, and other operations are also possible. For example, search for a supply position candidate may be performed irrespective of whether the amount of residual energy is below the threshold.
Though an example of using GPS information and the like received from a communication terminal of a user of an EV is shown in the present embodiment, the present embodiment is not limited to the form. For example, if the explanatory variables of the energy consumption model include such that uses data of traffic counters, the data may be received from the traffic counters. Further, data of the energy supply history may be acquired from a supply position.
A hardware configuration of the route estimation apparatus 1 according to the first to ninth embodiments will be described.
The CPU (Central Processing Unit) 151 executes a computer program which realizes each of the above-described functional components of the route estimation apparatus 1 on the main memory 155. By the CPU 151 executing the computer program, each functional component is realized.
The input interface 152 is a circuit for inputting an operation signal from an input device such as a keyboard, a mouse or a touch panel to the route estimation apparatus 1.
The display device 153 displays data or information outputted from the route estimation apparatus 1. The display device 153 is, for example, an LCD (Liquid Crystal Display), a CRT (Cathode-Ray Tube) display or a PDP (a plasma display panel) but is not limited thereto. The data or information outputted from the computer apparatus 100 can be displayed by the display device 153.
The communication device 154 is a circuit for the route estimation apparatus 1 to wirelessly or wiredly communicate with an external apparatus. Energy supply history and route information can be inputted from the external apparatus via the communication device 154. Information inputted from the external apparatus can be stored into a DB. The information communicator 18 can be built on the communication device 154.
The main memory 155 stores a program for movement route estimation and energy consumption model building, data required for execution of the program, data generated by execution of the program and the like. The program is developed and executed on the main memory 155. The main memory 155 is, for example, a RAM, a DRAM or an SRAM but is not limited thereto. The storage in each embodiment may be built on the main memory 155.
The external storage device 156 stores the program for movement route estimation and energy consumption model building, the data required for execution of the program, the data generated by execution of the program and the like. The program and data are read onto the main memory 155 at the time of executing movement route estimation and energy consumption model building. The external storage device 156 is, for example, a hard disk, an optical disk, a flash memory or a magnetic tape but is not limited thereto. The storage in each embodiment may be built on the external storage device 156.
The program for movement route estimation and energy consumption model building may be installed in the computer apparatus 100 in advance or may be stored in a recording medium such as a CD-ROM. Further, the program may be uploaded on the Internet.
The computer apparatus 100 may be provided with one or more CPUs 151, one or more input interfaces 152, one or more display devices 153, one or more communication devices 154 and one or more main memories 155. Peripheral equipment such as a printer and a scanner may be connected.
Further, the route estimation apparatus 1 may be configured with a single computer apparatus 100 or may be configured as a system constituted by a plurality of computer apparatuses 100 which are mutually connected.
While certain embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel embodiments described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the embodiments described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the inventions.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2017-167861 | Aug 2017 | JP | national |
