Route design system, cost function learning device, design route output device, method and program
The route design system uses inverse reinforcement learning to optimize bus routes and stops based on expert-intended features and constraints, addressing the inefficiencies of manual updates by automating the process for rapid and high-quality route planning.
Patent Information
- Application Number
- JP2022578002
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2021-02-01
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-02-01
AI Technical Summary
Existing route design systems require significant manual effort and know-how to optimize bus routes and bus stop locations, making frequent updates difficult and time-consuming.
A route design system that uses inverse reinforcement learning to automatically generate high-quality routes by weighting features intended by experts, such as user numbers and new stop establishment, and constraints like proximity to existing stops, allowing efficient and quick route optimization.
Enables high-quality route design to be performed quickly and efficiently, reducing the need for manual labor and enabling regular updates.
Smart Images

Figure 0007593420000002 
Figure 0007593420000003 
Figure 0007593420000004
Abstract
Description
[Technical Field]
[0001] The present invention relates to a route design system and a route design method for designing a route that passes through relay points, a cost function learning device and a cost function learning program for learning a cost function used in route design, and a design route output device, a design route output method, and a design route output program for outputting a designed route. [Background technology]
[0002] Various factors must be taken into consideration when designing routes, such as transportation lines and communication networks. For example, when it comes to public transportation such as buses, efficient route designs are required that balance convenience for local residents with profitability. Furthermore, even for existing routes, periodic reviews are required depending on the development status of the area and the demographic trends of the residents.
[0003] Patent Document 1 describes a bus route evaluation support system that supports decision-making on route reorganization and timetable revisions based on the results of quantitative evaluation of bus routes. The system described in Patent Document 1 creates multiple evaluation and analysis projects for each route generation, and performs evaluation and analysis individually. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2011-164889 Summary of the Invention [Problem to be solved by the invention]
[0005] For example, in the case of the bus route mentioned above, the establishment of new bus stops and route changes are actually considered and updated based on petitions from residents. Furthermore, the establishment of bus stops requires the satisfaction of various constraints, such as negotiations to use the land for the bus stop and the distance from curves. Furthermore, it is necessary to set routes by taking into account the overall conflicts of interest with other modes of transportation (other buses, taxis, etc.). Given these circumstances, it is difficult to manually and frequently review routes and bus stop locations, and adjusting routes requires a great deal of know-how.
[0006] By using the system described in Patent Document 1, it is possible to individually evaluate the correlation between the evaluation data value and the analysis target (for example, within walking distance). However, the results of evaluation by the system described in Patent Document 1 indicate the factors that should be considered individually. Therefore, there is a problem in that a lot of know-how and a lot of time are required to perform overall optimization, as described above.
[0007] Therefore, the present invention aims to provide a route design system and route design method that can quickly and efficiently perform high-quality route design, a cost function learning device and cost function learning program that learns cost functions to be used in route design, and a design route output device, design route output method, and design route output program that output designed routes. [Means for solving the problem]
[0008] The route planning system of the present invention is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount that is assumed to be intended by an expert when selecting relay point candidates and designing a route, and the cost occurring in at least one of the selection of relay point candidates and the design of the route is calculated. Toa learning means for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about use of the relay point, and route performance information, which is performance data of routes passing through each relay point; a condition input means for receiving input of relay point information and constraint conditions for a network for which a new route design is to be performed; a relay point selection means for selecting relay point candidates that will minimize the cost calculated by the cost function based on the input relay point information so as to satisfy the input constraint conditions; and a designed route output means for outputting a route that will minimize the cost calculated by the cost function from among routes that pass through some or all of the relay points in the network so as to satisfy the constraint conditions. The cost function is a function expressed as a linear sum of a feature indicating the number of users at each time period at a relay point and a feature indicating whether the relay point is a newly established point. It is characterized by:
[0009] The cost function learning device according to the present invention is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount that is assumed to be intended by an expert when selecting relay point candidates and designing a route, and the cost occurring in at least one of the selection of relay point candidates and the design of a route is calculated. To and learning means for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, and route performance information, which is performance data of routes that pass through each relay point. The cost function is a function expressed as a linear sum of a feature indicating the number of users at each time period at a relay point and a feature indicating whether the relay point is a newly established point. It is characterized by:
[0010] The design route output device according to the present invention is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature quantity that is assumed to be intended by an expert when selecting relay point candidates and designing a route, and the cost occurring in at least one of the selection of relay point candidates and the design of the route is calculated. ToThe system includes a condition input means for receiving input of a cost function, which is a function for calculating a cost function, relay point information, which is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, and constraint conditions, for a network for which a new route is to be designed; a relay point selection means for selecting relay point candidates that will minimize the cost calculated by the cost function based on the input relay point information so as to satisfy the input constraint conditions; and a design route output means for outputting a route that will minimize the cost calculated by the cost function from among routes that pass through some or all of the relay points in the network so as to satisfy the constraint conditions. The cost function is a function expressed as a linear sum of a feature indicating the number of users at each time period at a relay point and a feature indicating whether the relay point is a newly established point. The condition input means is characterized in that it accepts input of a cost function learned by inverse reinforcement learning using training data including relay point information and route performance information, which is performance data of routes passing through each relay point.
[0011] The route design method of the present invention is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount that is assumed to be intended by an expert when selecting relay point candidates and designing a route, and the cost occurring in at least one of the selection of relay point candidates and the design of the route is calculated. To and a learning means for receiving an input of a cost function, which is a function for calculating the cost function, and learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, and route performance information, which is performance data of routes that pass through each relay point. The cost function is a function expressed as a linear sum of a feature indicating the number of users at each time period at a relay point and a feature indicating whether the relay point is a newly established point. It is characterized by:
[0012] The design route output method according to the present invention is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature quantity that is assumed to be intended by an expert when selecting relay point candidates and designing a route, and the cost occurring in at least one of the selection of relay point candidates and the design of the route is calculated. ToThe method accepts input of a cost function that is a function for calculating a cost of a route, relay point information that is data that associates information indicating a relay point with information about the vicinity of the relay point and information about the use of the relay point, and constraints for a network for which a new route is to be designed, selects relay point candidates that will minimize the cost calculated by the cost function based on the input relay point information so as to satisfy the input constraints, and outputs a route that will minimize the cost calculated by the cost function from among routes that pass through some or all of the relay points in the network so as to satisfy the constraints, and when accepting the input, accepts input of a cost function that has been learned by inverse reinforcement learning using training data including relay point information and route performance information that is performance data for routes that pass through each relay point. The cost function is a function expressed as a linear sum of a feature indicating the number of users at each time period at a relay point and a feature indicating whether the relay point is a newly established point. It is characterized by:
[0013] The cost function learning program according to the present invention is a program for learning a cost function that is input to a computer and is expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount that is assumed to be intended by an expert when selecting a candidate relay point and designing a route, and that represents the cost that occurs in at least one of selecting a candidate relay point and designing a route. To and executing a learning process for learning the cost function by inverse reinforcement learning using training data including a function input process for receiving input of a cost function, which is a function for calculating the cost function, and relay point information, which is data in which information indicating relay points is associated with information about the vicinity of the relay point and information about the use of the relay point, and route performance information, which is performance data of routes that pass through each relay point. The cost function is a function expressed as a linear sum of a feature indicating the number of users at each time period at a relay point and a feature indicating whether the relay point is a newly established point. It is characterized by:
[0014] The design route output program according to the present invention is a program for outputting a design route to a computer, which is a function expressed by a linear sum of terms in which the degree of importance is weighted for each feature quantity that is assumed to be intended by an expert when selecting candidate relay points and designing a route, and which represents the cost that occurs in at least one of the selection of candidate relay points and the design of the route. Toand a condition input process for receiving input of a cost function that is a function for calculating the cost of a route that is to be newly designed, relay point information that is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, and constraint conditions, based on the input relay point information, to select relay point candidates that will minimize the cost calculated by the cost function so as to satisfy the input constraint conditions, and a design route output process for outputting a route that will minimize the cost calculated by the cost function from among routes that pass through some or all of the relay points in the network so as to satisfy the constraint conditions, and the condition input process receives input of a cost function that has been learned by inverse reinforcement learning using training data including relay point information and route performance information that is performance data for routes that pass through each relay point. The cost function is a function expressed as a linear sum of a feature indicating the number of users at each time period at a relay point and a feature indicating whether the relay point is a newly established point. It is characterized by: [Effects of the Invention]
[0015] According to the present invention, high-quality route design can be performed quickly and efficiently. [Brief explanation of the drawings]
[0016] [Figure 1] 1 is a block diagram showing an example of the configuration of an embodiment of a route design system according to the present invention. [Figure 2] FIG. 2 is an explanatory diagram illustrating an example of route design data. [Figure 3] FIG. 10 is an explanatory diagram showing an example of a process for updating a route. [Figure 4] 10 is a flowchart illustrating an example of the operation of the learning device. [Figure 5] 10 is a flowchart illustrating an example of the operation of the design path output device. [Figure 6] FIG. 10 is a block diagram showing a modified example of the route design system. [Figure 7] FIG. 10 is an explanatory diagram illustrating an example of a process for designing a new route. [Figure 8] 1 is a block diagram showing an overview of a route design system according to the present invention. [Figure 9]1 is a block diagram showing an overview of a cost function learning device according to the present invention; [Figure 10] 1 is a block diagram showing an overview of a design path output device according to the present invention; DETAILED DESCRIPTION OF THE INVENTION
[0017] The route design system of the present invention is a system for designing routes that pass through relay points (for example, bus stops and relay devices) in various networks such as road networks and communication networks. In the following explanation, for simplicity's sake, an example will be given of designing a bus route that passes through bus stops that are relay points in a road network. However, routes designed with the route design system of the present invention are not limited to bus routes. For example, the route design system of the present invention can also be applied in situations where trains, ships, airplanes, network routing, etc. are to be designed.
[0018] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
[0019] 1 is a block diagram showing an example of the configuration of an embodiment of a route design system according to the present invention. The route design system 1 of this embodiment includes a route design data storage device 10, a learning device 20, a designed route output device 30, and a display device 40.
[0020] The display device 40 is a device that outputs various processing results by the route design system 1. The display device 40 is realized by, for example, a display device. Note that, although FIG. 1 illustrates one display device 40 connected to the design route output device 30, the display device 40 connected to the learning device 20 and the display device 40 connected to the design route output device 30 may be provided separately.
[0021] The route design data storage device 10 stores data on routes that have been designed in the past (hereinafter referred to as route design data). Specifically, the route design data includes data (hereinafter referred to as relay point information) that associates information indicating relay points with surrounding information about the relay points (e.g., the distance to nearby relay points, whether there is a department store), and usage information about the relay points (e.g., the number of users of the transportation facility, the number of packets passing through, and information indicating whether the relay point is new), as well as performance data of routes that pass through each relay point (hereinafter referred to as route performance information).
[0022] Fig. 2 is an explanatory diagram showing an example of route design data. The route design data shown in Fig. 2 includes relay point information P1 indicating the number of users at each relay point by time period, the distance to nearby bus stops, whether there are department stores around the relay point, and whether the relay point is newly opened, as well as route performance information P2 indicating the performance of past routes connecting those relay points.
[0023] The route design data illustrated in FIG. 2 is information from when an already operating route was designed, and is data created by an expert taking various factors into consideration, so it can be said to be data created by an expert. Note that the relay point information and route performance information illustrated in FIG. 2 are examples, and other information may be included. Examples of other information include population density by area, building information, the difficulty of setting relay points, and surrounding road information. Specific details of these will be described later.
[0024] The learning device 20 includes a cost function input unit 21, a data extraction unit 22, a waypoint learning unit 23, a route learning unit 24, a learning result output unit 25, and a storage unit 26.
[0025] The cost function input unit 21 receives input of cost functions used for learning by the relay point learning unit 23 and the route learning unit 24, which will be described later. Specifically, the cost function input unit 21 receives input of a cost function (hereinafter referred to as a first cost function) that calculates the cost incurred in selecting relay point candidates, and a cost function (hereinafter referred to as a second cost function) that calculates the cost incurred in designing a route.
[0026] The cost function input unit 21 may accept input of one type of cost function instead of two types of cost functions, the first cost function and the second cost function. Feature quantities less relevant to the selection of relay points and route design are set to low weights as a result of machine learning, which will be described later. As a result, feature quantities intended by an expert are extracted in the selection of relay points and route design.
[0027] The cost function is expressed as a linear sum of terms in which the degree of importance is weighted for each feature that is assumed to be intended by an expert when selecting relay point candidates and designing a route, as included in the route design data shown in Figure 2. Note that this degree of importance can also be said to represent the intention of the expert when designing. Therefore, the value calculated by the cost function can also be said to be an evaluation index used to evaluate the selection of relay points and route design.
[0028] The cost function input unit 21 may also receive input of constraints to be satisfied along with the cost function. The cost function and constraints are determined in advance by a designer or the like. That is, candidates for feature quantities to be considered in selecting relay points and designing routes are selected in advance by an analyst or the like and defined as a cost function. The cost function input unit 21 may receive input of a cost function whose feature quantities include, for example, at least one of the number of users and satisfaction level.
[0029] More specifically, when the number of users in each time period and whether the relay point is newly established or not are considered as items (features) intended by the expert, the cost function is expressed by the following example formula 1. ij and y i indicates the feature amount, and α and β indicate the weights (degree of importance). Specifically, x ij indicates the number of users at relay point i in time slot j, and y i indicates whether the relay point i is a newly established point.
[0030]
number
[0031] By using the time-dependent changes shown above (for example, the number of users from 6:00 to 8:00, the number of users from 8:00 to 10:00, etc.) as features, it becomes possible to automatically generate routes dynamically.
[0032] The above-described feature quantities are merely examples, and the cost function may include other feature quantities. For example, the population density by area may be cited as another feature quantity. Since a larger population means more users, the cost incurred when setting a relay point is calculated to be lower the larger the population. On the other hand, since a smaller population means a more useful niche route, the cost incurred when designing a route is calculated to be lower the smaller the population.
[0033] Another feature may be building information (presence or absence). The presence of buildings such as stations, hospitals, convenience stores, and supermarkets significantly attracts pedestrian traffic, so the more of these buildings there are, the lower the calculated cost. Another feature may be road gradient. A steeper gradient leads to reduced visibility and speed, so the steeper the gradient, the higher the calculated cost. Other features that may be used include the surrounding natural environment, the presence or absence of temples, and whether the land is public.
[0034] Furthermore, when designing an airplane route, feature values indicating weather and seasonal winds may be used.
[0035] The data extraction unit 22 extracts route design data for the range in which a route is designed from the route design data storage device 10. The range in which a route is designed is specified in advance by a designer or the like. Note that the extracted route design data is also referred to as training data because it is data used for learning by the waypoint learning unit 23 and the route learning unit 24, which will be described later.
[0036] In addition, the data extraction unit 22 may perform processes such as converting items included in the route design data into features (such as calculations or conversion to binary values), data integration processes, and data cleansing in order to match the features included in the cost function.
[0037] The waypoint learning unit 23 learns the above-mentioned cost function (specifically, the first cost function) by inverse reinforcement learning using the training data extracted by the data extraction unit 22.
[0038] Any method may be used for the inverse reinforcement learning by the waypoint learning unit 23. For example, the waypoint learning unit 23 may learn the cost function by repeating a mathematical optimization process for selecting waypoints based on the input cost function and constraint conditions, and a cost function estimation process for updating the parameters (degree of importance) of the cost function so as to reduce the difference between the generated waypoints of the expert and the training data.
[0039] The relay point learning unit 23 learns a cost function by inverse reinforcement learning using route design data, and thereby it becomes possible to extract the feature amount intended when selecting a relay point.
[0040] The path learning unit 24 learns the above-mentioned cost function (specifically, the second cost function) by inverse reinforcement learning using the training data extracted by the data extraction unit 22.
[0041] The method of inverse reinforcement learning performed by the route learning unit 24 is also arbitrary. For example, the route learning unit 24 may learn the cost function by repeating a mathematical optimization process for designing a route based on the input cost function and constraint conditions, and a cost function estimation process for updating the parameters (degree of importance) of the cost function so as to reduce the difference between the designed route of the expert and the training data.
[0042] The route learning unit 24 learns a cost function by inverse reinforcement learning using route design data, making it possible to extract the intended feature amount when designing a route.
[0043] The waypoint learning unit 23 and the route learning unit 24 may be configured as one learning means. That is, the learning means may include the waypoint learning unit 23 and the route learning unit 24. In this case, the learning means may learn the cost function by inverse reinforcement learning using training data.
[0044] The learning result output unit 25 outputs the learned cost function. Specifically, the learning result output unit 25 outputs feature amounts included in the cost functions (more specifically, the first cost function and the second cost function) in association with weights of the feature amounts. The learning result output unit 25 may store the learned cost function in the storage unit 26, or may transmit information about the cost function to the design path output device 30 and store it in the storage unit 35.
[0045] Furthermore, the learning result output unit 25 may display the contents of the cost function on the display device 40. By displaying the contents of the cost function on the display device 40, it becomes possible for an expert to visually recognize the items that are important in selecting relay points and designing a route.
[0046] The storage unit 26 stores the learned cost function. The storage unit 26 may also store various parameters used in inverse reinforcement learning by the waypoint learning unit 23 and the route learning unit 24. The storage unit 26 is realized by, for example, a magnetic disk.
[0047] The cost function input unit 21, the data extraction unit 22, the relay point learning unit 23, the route learning unit 24, and the learning result output unit 25 are realized by a computer processor (e.g., a CPU (Central Processing Unit)) that operates according to a program (learning program, route design program).
[0048] For example, the program may be stored in memory unit 26 of learning device 20, and the processor may read the program and operate, in accordance with the program, as cost function input unit 21, data extraction unit 22, waypoint learning unit 23, route learning unit 24, and learning result output unit 25. Furthermore, the functions of learning device 20 may be provided in the form of SaaS (Software as a Service).
[0049] The cost function input unit 21, the data extraction unit 22, the waypoint learning unit 23, the route learning unit 24, and the learning result output unit 25 may each be realized by dedicated hardware. Furthermore, some or all of the components of each device may be realized by a general-purpose or dedicated circuit, processor, etc., or a combination of these. These may be configured by a single chip, or by multiple chips connected via a bus. Some or all of the components of each device may be realized by a combination of the above-mentioned circuits, etc., and a program.
[0050] Furthermore, when some or all of the components of learning device 20 are implemented using multiple information processing devices, circuits, etc., the multiple information processing devices, circuits, etc. may be centrally or decentralized. For example, the information processing devices, circuits, etc. may be implemented as a client-server system, cloud computing system, or other system connected via a communication network.
[0051] The learning device 20 may learn a plurality of cost functions. For example, the learning device 20 may learn a plurality of cost functions using past performance (training data) in a plurality of districts (district A, district B, etc.).
[0052] The designed route output device 30 includes a condition input unit 31 , a relay point selection unit 32 , a route design unit 33 , a designed route output unit 34 , and a storage unit 35 .
[0053] The storage unit 35 stores various information used by the relay point selection unit 32 (described later) when selecting relay point candidates, and various information used by the route design unit 33 when designing a route. The storage unit 35 also stores map information of the target area, candidate relay point locations in the area, building information around the relay points, demographics, and other related information. The storage unit 35 may also store a cost function learned by the learning device 20. The storage unit 35 is realized, for example, by a magnetic disk or the like.
[0054] The condition input unit 31 receives input of relay point information for a network for which a new route is to be designed, constraints for selecting candidate relay points, and constraints for designing a route. The condition input unit 31 may receive input of conditions indicating legal constraints (e.g., no bus stop can be installed at a point with a sharp curve, no bus stop can be installed within 10 meters of a sidewalk, etc.) or constraints based on compatibility with other transportation modes (e.g., no bus stop can be installed near an existing bus stop, etc.) as constraints for designing a route. The condition input unit 31 may also receive input of conditions for the start point and goal point, and conditions indicating constraints on the transportation network (e.g., one-way streets) as constraints for designing a route.
[0055] Furthermore, since the learned cost function is assumed to reflect the intentions of an expert, it is assumed that the selection of relay point candidates that violate constraints and the design of routes that violate constraints are implicitly suppressed even without the input of explicit constraint conditions. For example, if a bus stop has never been installed closer than 10 m from a sidewalk in the past, it is assumed that using the learned cost function will reduce the possibility of selecting a location closer than 10 m from the sidewalk.
[0056] Furthermore, in addition to the input of constraint conditions, the condition input unit 31 also receives input of surrounding information about relay points and usage information about relay points for the network for which a new route is to be designed. Specifically, the condition input unit 31 may receive input of relevant information when designing a route. For example, if regional information used when designing a bus route has been updated, the condition input unit 31 may receive input of regional information such as population density by area, building (store) information, and redevelopment information as surrounding information about relay points that indicates the latest situation in the area. Furthermore, if the cost function takes into account the driving history of other means of transportation (for example, taxi driving records), the condition input unit 31 may also receive input of such driving history.
[0057] The condition input unit 31 may accept, as a constraint condition, an input of a list of relay points to be selected as candidates by the relay point selection unit 32 (described later). The condition input unit 31 may also accept, as a constraint condition, an input of information indicating an area within a network in which the relay point selection unit 32 (described later) can set relay points.
[0058] If the cost function is not stored in the storage unit 35, the condition input unit 31 may receive an input of the cost function.
[0059] Based on the input relay point information, the relay point selection unit 32 selects a relay point candidate that minimizes the cost calculated by the cost function (more specifically, the first cost function) so as to satisfy the input constraint conditions. Note that a list of relay point candidates to be selected may be determined in advance by a designer or the like. That is, the relay point selection unit 32 may select a relay point candidate from the list of relay point candidates. Furthermore, when a range within a network in which relay points can be set is input as a constraint condition, the relay point selection unit 32 may select a relay point candidate from the input range within the network.
[0060] Furthermore, the relay point selection unit 32 may select multiple relay point candidates by excluding the selected relay point candidate and repeating the process of selecting a further relay point candidate a predetermined number of times.
[0061] The route design unit 33 designs a route that minimizes the cost calculated by a cost function (more specifically, a second cost function) among routes that pass through some or all of the selected relay point candidates in the network so as to satisfy the input constraint conditions. Note that the designed route does not need to pass through all of the selected relay point candidates. Furthermore, the route design unit 33 may design a route that includes existing relay points in addition to the relay points selected by the relay point selection unit 32.
[0062] Specifically, the route design unit 33 may design a route by determining the combination of relay points that minimizes the total cost based on the cost incurred when traveling from a relay point candidate selected by the relay point selection unit 32 to another relay point candidate.
[0063] The route design unit 33 may use any method to determine a combination of routes that minimizes the total cost. The route design unit 33 may design a route as a combinatorial optimization problem. For example, the route design unit 33 may use a cost calculated by a cost function instead of the distance used in the Dijkstra algorithm to design a route as a problem of solving a route that minimizes the cost.
[0064] The designed route output unit 34 outputs the designed route. For example, when designing a bus route, the designed route output unit 34 may output a bus route connecting an output point, which is the route, and an arrival point, and the positions of bus stops, which are relay points.
[0065] The designed route output unit 34 may also output the forecast number of users and the forecast profitability together with the new route design proposal. Furthermore, the designed route output unit 34 may output each relay point included in the designed route and the route connecting the relay points superimposed on a map. This allows the designed route to be understood more specifically.
[0066] 3 is an explanatory diagram showing an example of a route update process. For example, assume that information indicating past route data 41 is stored as route design data in the route design data storage device 10. The route design data includes, for example, area-specific population density at the time of design, building (store) information, redevelopment information, road information, and the like. The route design data storage device 10 may also store the number of passengers and profitability of the route indicated by the past route data 41, as well as constraints used when the past route data 41 was designed.
[0067] Based on this information, the relay point learning unit 23 and the route learning unit 24 learn a model (i.e., a cost function) that indicates the policy based on which high-quality routes (successful routes) that have been manually designed in the past were created.
[0068] Next, the condition input unit 31 accepts input of constraint conditions. In the example shown in Fig. 3, a start point and a finish point are specified as constraint conditions 42. Note that the condition input unit 31 may also accept input of updated area information as another constraint condition (related information).
[0069] The relay point selection unit 32 selects relay point candidates from the input constraint conditions and related information. The route design unit 33 designs a new route from the input constraint conditions and related information. The example shown in FIG. 3 shows that a new route 43 that passes through new relay points 431 and 432 has been designed. The designed route output unit 34 then outputs the designed route 43. The designed route output unit 34 may also output, together with the route 43, forecasts of the number of users, forecasts of profitability, and the like.
[0070] The condition input unit 31, relay point selection unit 32, route design unit 33, and designed route output unit 34 are realized by a processor (for example, a CPU) of a computer that operates according to a program (designed route output program, route design program).
[0071] Next, the operation of the route design system of this embodiment will be described. FIG. 4 is a flowchart showing an example of the operation of the learning device 20 of this embodiment. The cost function input unit 21 accepts input of a cost function (step S11). The cost function input unit 21 may accept input of two types of cost functions (a first cost function and a second cost function). The data extraction unit 22 extracts training data from the route design data storage device 10 (step S12). The waypoint learning unit 23 learns a first cost function that calculates the cost incurred in selecting candidate waypoints through inverse reinforcement learning using the training data (step S13). Furthermore, the route learning unit 24 learns a second cost function that calculates the cost incurred in designing a route through inverse reinforcement learning using the training data (step S14). Then, the learning result output unit 25 outputs the learned cost function (step S15).
[0072] 5 is a flowchart showing an example of the operation of the design route output device 30 of this embodiment. The condition input unit 31 accepts input of peripheral information and utilization information about a network for which a new route is to be designed, as well as constraint conditions (step S21). The relay point selection unit 32 selects relay point candidates that minimize the cost calculated by the cost function based on the input relay point peripheral information and relay point utilization information, so as to satisfy the input constraint conditions (step S22). The route design unit 33 designs a route that minimizes the cost calculated by the cost function from among routes that pass through some or all of the relay points in the network so as to satisfy the constraint conditions (step S23). Then, the design route output unit 34 outputs the designed route (step S24).
[0073] As described above, in this embodiment, the cost function input unit 21 accepts input of a cost function, and the relay point learning unit 23 and route learning unit 24 learn the cost function through inverse reinforcement learning using training data. The condition input unit 31 accepts input of peripheral information and usage information about the network for which a new route is to be designed, as well as constraint conditions. The relay point selection unit 32 selects relay point candidates that minimize the cost calculated by the cost function so as to satisfy the input constraint conditions. The route design unit 33 then designs a route that minimizes the cost calculated by the cost function from among routes that pass through some or all of the relay points in the network so as to satisfy the constraint conditions, and the designed route output unit 34 outputs the designed route. In this way, routes designed by so-called experts can be said to be of high quality, allowing for high-quality route design to be performed quickly and efficiently.
[0074] In other words, in this embodiment, inverse reinforcement learning is used to learn route designs that are deemed to satisfy constraints from past route design data, and the resulting route design model is used to update or design new routes. This makes it possible to design public transportation routes without relying on human labor, thereby enabling efficient and high-quality route updates and designs to be carried out regularly.
[0075] Next, a modified example of the route design system of this embodiment will be described. Fig. 6 is a block diagram showing a modified example of the route design system of this embodiment. The route design system 2 illustrated in Fig. 6 includes a route design data storage device 10, a learning device 20a, a learning device 20b, a designed route output device 30, and a display device 40. The contents of the route design data storage device 10, the designed route output device 30, and the display device 40 are the same as those in the above embodiment.
[0076] The learning device 20a is a device that learns a first cost function. The learning device 20a includes a cost function input unit 21a, a data extraction unit 22, a relay point learning unit 23, a learning result output unit 25a, and a storage unit 26. The contents of the data extraction unit 22, the relay point learning unit 23, and the storage unit 26 are the same as those in the above embodiment.
[0077] The cost function input unit 21a receives input of a first cost function as a cost function to be used for learning. Other aspects are the same as those of the cost function input unit 21. Furthermore, the learning result output unit 25a outputs the learned first cost function. Other aspects are the same as those of the learning result output unit 25.
[0078] The learning device 20b is a device that learns the second cost function. The learning device 20b includes a cost function input unit 21b, a data extraction unit 22, a route learning unit 24, a learning result output unit 25b, and a storage unit 26. The contents of the data extraction unit 22, the route learning unit 24, and the storage unit 26 are the same as those in the above embodiment.
[0079] The cost function input unit 21b receives input of a second cost function as a cost function to be used for learning. Other aspects are the same as those of the cost function input unit 21. Furthermore, the learning result output unit 25b outputs the learned second cost function. Other aspects are the same as those of the learning result output unit 25.
[0080] The operation of the route design system 1 of this embodiment will be described below using a specific example. Fig. 7 is an explanatory diagram showing an example of the process of designing a new route. In this specific example, it is assumed that two types of cost functions are learned by the learning device 20 of this embodiment using training data from different areas.
[0081] Specifically, it is assumed that cost function A shown in FIG. 7 is learned using past performance in district A, and cost function B shown in FIG. 7 is learned using past performance in district B. Here, x i is a feature indicating the number of users at relay point i, and y i is a feature indicating whether the relay point i is a newly established point. Here, since the content of the cost function is expressed as a reward, the relay point selection unit 32 selects the relay point that maximizes the reward.
[0082] Under these circumstances, it is assumed that a bus route is planned for a new district C having a bus stop candidate D3 as shown in FIG. 7. Since cost function A and cost function B are learned by the learning device 20 of this embodiment, the weighting of the feature corresponds to the degree of importance. Here, when a route is designed to maximize profits (in other words, to increase the number of users), x i It can be said that a cost function with a large coefficient is a cost function intended to increase profits. Therefore, we will design a route using cost function A.
[0083] When considering the bus stop candidate D3 shown in FIG. 7, the cost function A can be maximized by adopting the relay points (bus stops) indicated by points (i), (iv), and (v) as the new route.
[0084] Next, an overview of the present invention will be described. Fig. 8 is a block diagram showing an overview of a route design system according to the present invention. A route design system 70 (for example, route design system 1) according to the present invention includes function input means 71 (for example, cost function input unit 21) that accepts input of a cost function that calculates the cost incurred in at least one of selecting candidate relay points (for example, bus stops) and designing a route (for example, a route), expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount that is assumed to be intended by an expert when selecting candidate relay points (for example, bus stops) and designing a route (for example, a route), and calculates the cost function (for example, the first relay point) by inverse reinforcement learning using training data that includes relay point information that is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about use of the relay point, and route performance information that is performance data of routes that pass through each relay point. the network for which a new route is to be designed; a condition input means 73 (e.g., a condition input unit 31) for receiving input of relay point information and constraint conditions for the network for which a new route is to be designed; a relay point selection means 74 (e.g., a relay point selection unit 32) for selecting relay point candidates that will minimize the cost calculated by the cost function based on the input relay point information so as to satisfy the input constraint conditions; and a design route output means 75 (e.g., a route design unit 33, a design route output unit 34) for outputting a route that will minimize the cost calculated by the cost function from among routes that pass through some or all of the relay points in the network so as to satisfy the constraint conditions.
[0085] Such a configuration enables high-quality route design to be carried out quickly and efficiently.
[0086] Specifically, the design route output means 75 may output a designed route by determining the combination of relay points that minimizes the total cost based on the cost incurred when moving from one relay point candidate to another relay point candidate.
[0087] Furthermore, the learning means 72 may include relay point learning means (e.g., relay point learning unit 23) that learns a first cost function that is a cost function that calculates the cost incurred in selecting relay point candidates, and route learning means (route learning unit 24) that learns a second cost function that is a cost function that calculates the cost incurred in designing a route. Then, relay point selection means 74 may select relay point candidates that minimize the cost calculated by the first cost function, and designed route output means 75 may output the route that minimizes the cost calculated by the second cost function.
[0088] The condition input means 73 may accept an input of a list of candidate relay points to be selected as a constraint condition, and the relay point selection means 74 may select a candidate relay point from the accepted list.
[0089] In addition, the condition input means 73 may accept input of information indicating the range of a network in which a relay point can be set as a constraint condition, and the relay point selection means 74 may select a relay point candidate from the input network range.
[0090] Furthermore, the designed route output means 75 may output each relay point included in the designed route and the route connecting each relay point by superimposing them on a map.
[0091] The route design system 70 may also include a learning result output means (for example, the learning result output unit 25) that outputs the feature amounts included in the cost function in association with the degrees of importance of the feature amounts.
[0092] Furthermore, the learning means 72 may learn a cost function that includes at least one of the number of users and the satisfaction level as a feature quantity.
[0093] 9 is a block diagram showing an outline of a cost function learning device according to the present invention. A cost function learning device 80 (e.g., learning device 20) according to the present invention includes function input means 81 and learning means 82. The contents of the function input means 81 and learning means 82 are similar to those of the function input means 71 and learning means 72 shown in FIG.
[0094] 10 is a block diagram showing an outline of a design route output device according to the present invention. A design route output device 90 according to the present invention (for example, design route output device 30) comprises condition input means 91, relay point selection means 92, and design route output means 93. The contents of the condition input means 91, relay point selection means 92, and design route output means 93 are the same as those of the condition input means 73, relay point selection means 74, and design route output means 75 illustrated in FIG. 8.
[0095] A part or all of the above-described embodiments can be described as, but not limited to, the following supplementary notes.
[0096] (Supplementary Note 1) A function input means for receiving an input of a cost function for calculating a cost incurred in at least one of selecting a candidate relay point and designing a route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature that is assumed to be intended by an expert when selecting the candidate relay point and designing the route; a learning means for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating relay points is associated with surrounding information of the relay points and usage information of the relay points, and route performance information, which is performance data of routes that pass through each relay point; condition input means for receiving input of the relay point information and constraint conditions for the network for which a new route design is to be performed; a relay point selection means for selecting relay point candidates that minimize the cost calculated by the cost function based on the input relay point information, so as to satisfy the input constraint conditions; and a design route output means for outputting a route that has the smallest cost calculated by the cost function among routes that pass through some or all of the relay points in the network so as to satisfy the constraints. A route design system characterized by:
[0097] (Note 2) The designed route output means outputs a designed route by determining a combination of relay points that minimizes the total cost based on the cost incurred when moving from one relay point candidate to another relay point candidate. Attachment 1: A route design system.
[0098] (Appendix 3) The learning method is a relay point learning means for learning a first cost function which is a cost function for calculating a cost incurred in selecting relay point candidates; a route learning means for learning a second cost function which is a cost function for calculating a cost incurred in designing a route, the relay point selection means selects a relay point candidate that minimizes the cost calculated by the first cost function; The design path output means outputs a path that minimizes the cost calculated by the second cost function. 10. The route design system of claim 1 or 2.
[0099] (Note 4) The condition input means receives an input of a list of candidate relay points to be selected as a constraint condition; The relay point selection means selects relay point candidates from the received list. A route design system according to any one of appendices 1 to 3.
[0100] (Supplementary Note 5) The condition input means receives input of information indicating a range of a network in which a relay point can be set as a constraint condition; The relay point selection means selects relay point candidates from the input network range. A route design system according to any one of appendices 1 to 4.
[0101] (Appendix 6) The designed route output means outputs each relay point included in the designed route and the route connecting the relay points by superimposing them on a map. A route design system according to any one of appendices 1 to 5.
[0102] (Appendix 7) A learning result output means is provided for outputting the feature values included in the cost function in association with the degree of importance of the feature values. A route design system according to any one of appendices 1 to 6.
[0103] (Appendix 8) The learning means learns a cost function that includes at least one of the number of users and satisfaction level as a feature. A route design system according to any one of appendices 1 to 7.
[0104] (Supplementary Note 9) A function input means for receiving an input of a cost function that calculates a cost incurred in at least one of selecting a candidate relay point and designing the route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature that is assumed to be intended by an expert when selecting the candidate relay point and designing the route; and a learning means for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating relay points is associated with information about the vicinity of the relay point and information about use of the relay point, and route performance information, which is performance data of routes passing through each relay point. A cost function learning device characterized by:
[0105] (Supplementary Note 10) A cost function for calculating costs incurred in at least one of selecting a candidate relay point and designing a route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature that is assumed to be intended by an expert when selecting a candidate relay point and designing a route, and a condition input means for receiving input of relay point information and constraint conditions, which are data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, with respect to a network for which a new route is to be designed; a relay point selection means for selecting relay point candidates that minimize the cost calculated by the cost function based on the input relay point information, so as to satisfy the input constraint conditions; a design route output means for outputting a route that has the smallest cost calculated by the cost function among routes that pass through some or all of the relay points in the network so as to satisfy the constraints; The condition input means receives an input of the cost function learned by inverse reinforcement learning using training data including the relay point information and route performance information, which is performance data of routes passing through each relay point. A design path output device characterized by:
[0106] (Appendix 11) Accepting input of a cost function that calculates the cost incurred in at least one of selecting the candidate relay points and designing the route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature that is assumed to be intended by an expert when selecting the candidate relay points and designing the route; and a learning means for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating relay points is associated with information about the vicinity of the relay point and information about use of the relay point, and route performance information, which is performance data of routes passing through each relay point. A route design method comprising:
[0107] (Appendix 12) Accepting input of the relay point information and constraint conditions for the network for which a new route design is to be performed; Based on the input relay point information, a relay point candidate that minimizes the cost calculated by a cost function is selected so as to satisfy the input constraint conditions; Among routes that pass through some or all of the relay points in the network so as to satisfy the constraints, a route that minimizes the cost calculated by the cost function is output. 12. The route design method according to claim 11.
[0108] (Supplementary Note 13) A cost function is calculated for calculating the cost incurred in at least one of selecting a candidate relay point and designing the route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature that is assumed to be intended by an expert when selecting a candidate relay point and designing the route, and also receives input of relay point information and constraints, which are data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, for a network for which a new route is to be designed; Based on the input relay point information, a relay point candidate that minimizes the cost calculated by the cost function is selected so as to satisfy the input constraint condition; outputting a route that has the smallest cost calculated by the cost function among routes that pass through some or all of the relay points in the network so as to satisfy the constraints; When the conditions are input, an input of the cost function learned by inverse reinforcement learning using training data including the relay point information and route performance information, which is performance data of routes passing through each relay point, is accepted. A design path output method comprising:
[0109] (Appendix 14) To the computer, a function input process for receiving an input of a cost function that calculates a cost incurred in at least one of selecting the candidate relay points and designing the route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount that is assumed to be intended by an expert when selecting the candidate relay points and designing the route; and A learning process for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating relay points is associated with information about the vicinity of the relay point and information about use of the relay point, and route performance information, which is performance data of routes that pass through each relay point. A program storage medium that stores a cost function learning program for executing the above.
[0110] (Appendix 15) To the computer, a cost function for calculating costs incurred in at least one of selecting candidate relay points and designing the route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature quantity that is assumed to be intended by an expert when selecting candidate relay points and designing the route; and a condition input process for receiving input of relay point information and constraint conditions, the relay point information being data in which information indicating the relay points is associated with information about the vicinity of the relay points and information about the use of the relay points, with respect to the network for which a new route is to be designed. a relay point selection process for selecting relay point candidates that minimize the cost calculated by the cost function based on the input relay point information so as to satisfy the input constraint conditions; and executes a design route output process for outputting a route that has the smallest cost calculated by the cost function among routes that pass through some or all of the relay points in the network so as to satisfy the constraints; In the condition input process, an input of the cost function learned by inverse reinforcement learning using training data including the relay point information and route performance information, which is performance data of routes passing through each relay point, is accepted. A program storage medium storing a design path output program for the above.
[0111] (Appendix 16) To the computer, a function input process for receiving an input of a cost function that calculates a cost incurred in at least one of selecting the candidate relay points and designing the route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount that is assumed to be intended by an expert when selecting the candidate relay points and designing the route; and A learning process for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating relay points is associated with information about the vicinity of the relay point and information about use of the relay point, and route performance information, which is performance data of routes that pass through each relay point. A cost function learning program for executing the above.
[0112] (Appendix 17) To the computer, a cost function for calculating costs incurred in at least one of selecting candidate relay points and designing the route, the cost function being expressed as a linear sum of terms in which the degree of importance is weighted for each feature quantity that is assumed to be intended by an expert when selecting candidate relay points and designing the route; and a condition input process for receiving input of relay point information and constraint conditions, the relay point information being data in which information indicating the relay points is associated with information about the vicinity of the relay points and information about the use of the relay points, with respect to the network for which a new route is to be designed. a relay point selection process for selecting relay point candidates that minimize the cost calculated by the cost function based on the input relay point information so as to satisfy the input constraint conditions; and executes a design route output process for outputting a route that has the smallest cost calculated by the cost function among routes that pass through some or all of the relay points in the network so as to satisfy the constraints; In the condition input process, an input of the cost function learned by inverse reinforcement learning using training data including the relay point information and route performance information, which is performance data of routes passing through each relay point, is accepted. A design path output program for.
[0113] Although the present invention has been described above with reference to the embodiments, the present invention is not limited to the above embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. [Explanation of symbols]
[0114] 1,2 Route design system 10 Route design data storage device 20, 20a, 20b Learning device 21, 21a, 21b Cost function input section 22 Data Extraction Section 23 Relay Point Learning Department 24 Route Learning Unit 25, 25a, 25b Learning result output section 26 Memory section 30 Design path output device 31 Condition input section 32 Relay point selection section 33 Route Planning Department 34 Design path output section 35 Storage section 40 Display device
Claims
1. a function input means for receiving an input of a cost function, which is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount assumed to be intended by an expert when selecting a candidate relay point and designing a route, and which is a function for calculating a cost incurred in at least one of selecting the candidate relay point and designing the route; a learning means for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating a relay point is associated with surrounding information of the relay point and utilization information of the relay point, and route performance information, which is performance data of a route passing through each relay point; a condition input means for receiving input of the relay point information and constraint conditions for a network for which a new route design is to be performed; a relay point selection means for selecting relay point candidates that minimize the cost calculated by the cost function based on the input relay point information so as to satisfy an input constraint condition; a design route output means for outputting a route that has a minimum cost calculated by the cost function among routes that pass through a part or all of the relay points in the network so as to satisfy the constraint condition; The cost function is a function expressed as a linear sum of a feature value indicating the number of users at each time period at the relay point and a feature value indicating whether the relay point is a newly established point. A route design system comprising:
2. The designed route output means outputs a designed route by determining a combination of relay points that minimizes a total cost calculated by a cost function that includes the number of users as a feature value, based on the cost incurred when moving from one relay point candidate to another relay point candidate. The route design system according to claim 1 .
3. The learning tools are: a relay point learning means for learning a first cost function which is a cost function for calculating a cost incurred in selecting relay point candidates; a route learning means for learning a second cost function which is a cost function for calculating a cost incurred in designing a route, the relay point selection means selects a relay point candidate that minimizes a cost calculated by the first cost function; The design path output means outputs a path that minimizes the cost calculated by the second cost function. The route design system according to claim 1 or 2.
4. The condition input means receives an input of a list of intermediate points to be selected as a constraint condition; The relay point selection means selects relay point candidates from the received list. The route design system according to any one of claims 1 to 3.
5. a function input means for receiving an input of a cost function, which is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount assumed to be intended by an expert when selecting a candidate relay point and designing a route, and which is a function for calculating a cost incurred in at least one of selecting the candidate relay point and designing the route; a learning means for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating a relay point is associated with surrounding information of the relay point and utilization information of the relay point, and route performance information, which is performance data of a route passing through each relay point; The cost function is a function expressed as a linear sum of a feature value indicating the number of users at each time period at the relay point and a feature value indicating whether the relay point is a newly established point. A cost function learning device comprising:
6. a cost function, which is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount assumed to be intended by an expert when selecting relay point candidates and designing a route, and which is a function for calculating a cost incurred in at least one of selecting the relay point candidates and designing the route, and a condition input means for receiving input of relay point information and constraint conditions, which is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, for a network for which a new route is to be designed; a relay point selection means for selecting relay point candidates that minimize the cost calculated by the cost function based on the input relay point information so as to satisfy an input constraint condition; a design route output means for outputting a route that has a minimum cost calculated by the cost function among routes that pass through a part or all of the relay points in the network so as to satisfy the constraint condition; the cost function is a function expressed as a linear sum of a feature value indicating the number of users in each time period at the relay point and a feature value indicating whether the relay point is a newly established point, The condition input means receives an input of the cost function learned by inverse reinforcement learning using training data including the relay point information and route performance information, which is performance data of a route passing through each relay point. A design path output device comprising:
7. accepting an input of a cost function, the cost function being a linear sum of terms in which the degree of importance is weighted for each feature amount assumed to be intended by an expert when selecting a candidate relay point and designing a route, the cost function being a function for calculating a cost incurred in at least one of selecting the candidate relay point and designing the route; learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating a relay point is associated with surrounding information of the relay point and utilization information of the relay point, and route performance information, which is performance data of a route passing through each relay point; The cost function is a function expressed as a linear sum of a feature value indicating the number of users at each time period at the relay point and a feature value indicating whether the relay point is a newly established point. A route design method comprising:
8. a cost function, which is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount assumed to be intended by an expert when selecting a candidate relay point and designing a route, and which is a function for calculating a cost incurred in at least one of selecting the candidate relay point and designing the route, and receiving input of relay point information and constraint conditions, which are data in which information indicating a relay point is associated with surrounding information of the relay point and utilization information of the relay point, for a network for which a new route is to be designed; Based on the input relay point information, a candidate relay point that minimizes the cost calculated by the cost function is selected so as to satisfy the input constraint condition; outputting a route that has a minimum cost calculated by the cost function among routes that pass through some or all of the relay points in the network so as to satisfy the constraint condition; When accepting the input, an input of the cost function learned by inverse reinforcement learning using training data including the relay point information and route performance information, which is performance data of a route passing through each relay point, is accepted; The cost function is a function expressed as a linear sum of a feature value indicating the number of users at each time period at the relay point and a feature value indicating whether the relay point is a newly established point. A design path output method comprising:
9. On the computer, a function input process for receiving an input of a cost function, which is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount assumed to be intended by an expert when selecting a candidate relay point and designing a route, and which is a function for calculating a cost incurred in at least one of selecting the candidate relay point and designing the route; and executing a learning process for learning the cost function by inverse reinforcement learning using training data including relay point information, which is data in which information indicating a relay point is associated with surrounding information of the relay point and utilization information of the relay point, and route performance information, which is performance data of a route passing through each relay point; The cost function is a function expressed as a linear sum of a feature value indicating the number of users at each time period at the relay point and a feature value indicating whether the relay point is a newly established point. A cost function learning program.
10. On the computer, a cost function, which is a function expressed as a linear sum of terms in which the degree of importance is weighted for each feature amount assumed to be intended by an expert when selecting relay point candidates and designing a route, and which is a function for calculating costs incurred in at least one of selecting the relay point candidates and designing the route; and a condition input process for receiving input of relay point information and constraint conditions, which is data in which information indicating a relay point is associated with information about the vicinity of the relay point and information about the use of the relay point, for a network for which a new route is to be designed; a relay point selection process for selecting relay point candidates that minimize the cost calculated by the cost function based on the input relay point information so as to satisfy the input constraint conditions; and executing a design route output process for outputting a route that has a minimum cost calculated by the cost function among routes that pass through some or all of the relay points in the network so as to satisfy the constraint condition; In the condition input process, an input of the cost function learned by inverse reinforcement learning using training data including the relay point information and route performance information that is performance data of a route passing through each relay point is accepted; The cost function is a function expressed as a linear sum of a feature value indicating the number of users at each time period at the relay point and a feature value indicating whether the relay point is a newly established point. A design path output program comprising:
Citation Information
Patent Citations
Vehicle navigation device
JP2008175773A
Travel control plan generation system
JP2008180591A
Route searching method, route searching system and navigation apparatus
JP2009042051A
System, method and program for supporting evaluation of bus route, and recording medium
JP2011164889A
