Simulation program, simulation method, and information processing device
The simulation program and information processing device analyze behavioral simulation data to detect causal relationships, supporting efficient space design by identifying key attributes and suggesting modifications, thus addressing the challenges of understanding evaluation results and reducing computational burden.
Patent Information
- Application Number
- PCT/JP2023/018528
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-18
- Publication Date
- 2024-11-21
AI Technical Summary
Users face difficulty in understanding the reasoning behind evaluation results from behavioral simulations for space design, requiring specific knowledge and involving extensive simulations to find optimal designs, leading to a high burden on computational resources.
A simulation program and information processing device that analyze simulation result data to detect causal relationships between attributes, identifying causal attributes affecting the space evaluation, and outputting information to support space design improvements, thereby reducing the need for multiple trials and computational load.
Enables users to efficiently search for suitable space designs with improved evaluation results by identifying key causal attributes and suggesting design modifications, reducing the number of trials and computational efforts required.
Smart Images

Figure JP2023018528_21112024_PF_FP_ABST
Abstract
Description
Simulation program, simulation method, and information processing device
[0001] The present invention relates to a simulation program, a simulation method, and an information processing device.
[0002] As a type of computer simulation, a computer may perform behavioral simulation, which simulates the autonomous behavior of multiple agents in a specified space. The behavioral simulation may be a human behavior simulation, which simulates the autonomous behavior of multiple people. The behavioral simulation may be performed to evaluate the merits of the design of a specified space.
[0003] For example, a simulation method has been proposed that supports the decision-making of airport managers in installing signboards by simulating passenger behavior and congestion conditions at airport terminals under multiple signboard installation policies.
[0004] Shintaro Utsumi, Shingo Takahashi, Kotaro Ohori and Hirokazu Anai, "Agent-based Analysis for Design of Signage Systems in Large-scale Facilities", Proc. of the 2015 Winter Simulation Conference, pp. 3134-3135, December 2015
[0005] The results of one behavioral simulation indicate the evaluation results of one space design. However, it may be difficult for a user to understand the reason for the evaluation result by looking only at the final evaluation result. Therefore, the user may be required to have specific knowledge about the space to be designed in order to prepare multiple space designs to try. Furthermore, in order to search for a space design with good evaluation results, the computer may have to run behavioral simulations on many space designs with slightly different contents.
[0006] As a result, the burden of behavioral simulation can be heavy. Therefore, in one aspect, the present invention aims to support space design for behavioral simulation.
[0007] In one aspect, a simulation program is provided that causes a computer to execute the following processes: acquire simulation result data including a plurality of records corresponding to a plurality of agents acting within a specified space, each of the plurality of records including attribute values of a plurality of attributes related to the action result of one of the plurality of agents; detect one or more causal relationships from among the plurality of attributes, each indicating that one attribute is the cause of another attribute, based on the simulation result data; determine a causal attribute from among the plurality of attributes that indicates the cause of a target attribute that indicates an evaluation of the space, based on the one or more causal relationships; and output information corresponding to the causal attribute.
[0008] In one aspect, a computer-implemented simulation method is provided, and in another aspect, an information processing device having a storage unit and a processing unit is provided.
[0009] In one aspect, space design for behavioral simulation is supported. The above and other objects, features, and advantages of the present invention will become apparent from the following description taken in conjunction with the accompanying drawings illustrating preferred embodiments of the present invention by way of example.
[0010] FIG. 1 is a diagram for explaining an information processing apparatus according to a first embodiment. FIG. 2 is a diagram illustrating an example of hardware of an information processing apparatus according to a second embodiment. FIG. 3 is a diagram illustrating an example design of an airport terminal building. FIG. 4 is a diagram illustrating an example of an agent table. FIG. 5 is a diagram illustrating an example of an environment table. FIG. 6 is a diagram illustrating an example of a join table. FIG. 7 is a diagram illustrating an example of a normalized join table. FIG. 8 is a diagram illustrating an example of an attribute table. FIG. 9 is a diagram illustrating an example of a causal graph. FIG. 10 is a diagram illustrating an example of a calculation of causal effects for each scenario. FIG. 11 is a diagram illustrating an example of an improvement measure table. FIG. 12 is a diagram illustrating an example of a causal graph after re-simulation. FIG. 13 is a diagram illustrating an example of a calculation of causal effects after re-simulation. FIG. 14 is a diagram illustrating an example of the flow of a human behavior simulation. FIG. 15 is a diagram illustrating an example of a relationship between attribute types. A block diagram illustrating an example of functions of an information processing apparatus. A flowchart illustrating an example of a procedure for determining agent behavior. A flowchart illustrating an example of a simulation procedure.
[0011] The present embodiment will be described below with reference to the drawings. [First Embodiment] The first embodiment will be described.
[0012] FIG. 1 is a diagram illustrating an information processing device according to a first embodiment. The information processing device 10 according to the first embodiment analyzes the results of an agent-based behavioral simulation to assist in the design of another untried space. The behavioral simulation may be executed by the information processing device 10 or by another information processing device. The information processing device 10 may be a client device or a server device. The information processing device 10 may also be called a computer or a simulation device.
[0013] The information processing device 10 includes a storage unit 11 and a processing unit 12. The storage unit 11 may be a volatile semiconductor memory such as a random access memory (RAM), or may be a non-volatile storage such as a hard disk drive (HDD) or a flash memory.
[0014] The processing unit 12 is, for example, a processor such as a central processing unit (CPU), a graphics processing unit (GPU), or a digital signal processor (DSP). However, the processing unit 12 may also include an electronic circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). The processor executes a program stored in a memory such as a RAM (which may be the storage unit 11). A set of processors may be called a multiprocessor or simply a "processor."
[0015] The storage unit 11 stores simulation result data 13. The simulation result data 13 is generated by an agent-based behavioral simulation. The behavioral simulation simulates the autonomous behavior of multiple agents in a specified space. An agent may represent a person, a group of people, a living thing other than a human, or an inanimate object such as a vehicle.
[0016] The behavioral simulation is provided with spatial information that indicates the spatial design. The spatial information may define the shape of the area in which the agent can move. The spatial information may also define facilities that the agent can use, such as stores, or may define installed objects such as signboards. The spatial information may also define the agent's movement rules, such as restrictions on the direction of movement. The agent acts according to a certain behavioral algorithm under the constraints indicated by the spatial information. For example, the agent performs a series of actions, such as selecting a destination facility, finding its location on a signboard, and walking toward the destination facility.
[0017] The simulation result data 13 includes multiple records corresponding to multiple agents. One record indicates the behavioral result of one agent. Each of the multiple records includes multiple attribute values corresponding to multiple attributes. The multiple attributes are indicators related to the behavioral result, and are obtained by observing the behavior of the same agent from different aspects.
[0018] The multiple attributes may include a goal attribute that indicates the goal of the agent. The goal attribute is, for example, a destination facility that the agent wants to use. The multiple attributes may also include a behavior attribute that indicates an action performed by the agent. An example of the behavior attribute is travel time. The multiple attributes may also include an environmental attribute that indicates an interaction between the agent and the space. An example of the environmental attribute is information search time or facility waiting time.
[0019] The multiple attributes may also include external factor attributes that indicate constraints that the space imposes on the agent. Examples of external factor attributes include the installation method of signboards and the number of people allowed in a facility. The multiple attributes may also include target attributes that correspond to the evaluation results of the space. Examples of target attributes include the degree to which the agent's goal is achieved. It is preferable to use an index related to the agent's satisfaction as the target attribute. The simulation result data 13 may be table data with records as rows (tuples) and attributes as columns.
[0020] The processing unit 12 executes a causal search method on the simulation result data 13. As a result, the processing unit 12 detects one or more causal relationships from among multiple attributes included in the simulation result data 13. One causal relationship indicates that one attribute is the cause of another attribute. For example, the simulation result data 13 includes attributes 14a, 14b, 14c, and 14d. The processing unit 12 detects causal relationships 15a, 15b, 15c, 15d, and 15e from among the attributes 14a, 14b, 14c, and 14d.
[0021] Attribute 14a is a purpose attribute. Attribute 14b is a behavior attribute. Attribute 14c is an environment attribute. Attribute 14d is a target attribute. Causal relationship 15a indicates that attribute 14a is the cause of attribute 14b. Causal relationship 15b indicates that attribute 14a is the cause of attribute 14c. Causal relationship 15c indicates that attribute 14b is the cause of attribute 14c. Causal relationship 15d indicates that attribute 14c is the cause of attribute 14b. Causal relationship 15e indicates that attribute 14c is the cause of attribute 14d.
[0022] A causal relationship indicates that when the attribute value of an attribute corresponding to a cause changes, the attribute value of an attribute corresponding to a result also changes. For example, the causal search method first generates a completely undirected graph including multiple nodes corresponding to multiple attributes. The causal search method detects pairs of attributes that are independent from the simulation result data 13 through testing, and deletes edges representing the detected attribute pairs from the graph. The causal search method then determines the direction of each edge from the remaining graph structure so that the graph does not become a cyclic graph.
[0023] The processing unit 12 identifies a target attribute indicating an evaluation of the space from among the multiple attributes. The processing unit 12 determines a cause attribute indicating a cause for the target attribute based on the detected one or more causal relationships. The cause attribute is, for example, an attribute of a direct cause connected to the target attribute by one causal relationship. The processing unit 12 may determine two or more cause attributes. For example, the processing unit 12 determines attribute 14c as the cause attribute from causal relationship 15e in which attribute 14d is the result.
[0024] When there are two or more attributes that are the cause of the target attribute, the processing unit 12 may narrow down the causal attributes based on the magnitude of the causal effect between each attribute and the target attribute. The causal effect represents the strength of the influence of one attribute on another attribute, and indicates the amount of change in the attribute value of one attribute when the attribute value of the other attribute is changed by a certain amount. The processing unit 12 may select an attribute whose absolute value of the causal effect exceeds a threshold, or may preferentially select an attribute whose absolute value of the causal effect is large.
[0025] The processing unit 12 outputs information 16 corresponding to the cause attribute. The information 16 may indicate that the attribute 14c is the cause attribute. The information 16 may also indicate that a change in the attribute value of the attribute 14c (e.g., a decrease in the attribute value of the attribute 14c) causes a change in the attribute value of the target attribute, attribute 14d (e.g., an increase in attribute 14d). The information 16 may also indicate the attribute type of the attribute 14c (e.g., that the attribute 14c is an environmental attribute).
[0026] Furthermore, the information 16 may indicate causal relationships 15a, 15b, 15c, 15d, and 15e between the attributes 14a, 14b, 14c, and 14d. The information 16 may also indicate a space design policy for changing the attribute value of the attribute 14c. For example, the information processing device 10 stores a knowledge database that indicates space design policies for changing the attribute value for each attribute. The processing unit 12 may search the knowledge database for a policy for changing the attribute value of the attribute 14c. As an example, a possible space design policy for reducing the information search time of an agent is to install more signboards.
[0027] The processing unit 12 may store the information 16 in a non-volatile storage, display it on a display device, or transmit it to another information processing device. The information processing device 10 or another information processing device may accept spatial information corrected based on the information 16, and may re-execute the behavioral simulation based on the corrected spatial information.
[0028] As described above, the information processing device 10 according to the first embodiment acquires simulation result data 13 including multiple records corresponding to multiple agents acting within a specified space. Each record includes attribute values for multiple attributes related to the agent's behavioral results. Based on the simulation result data 13, the information processing device 10 detects one or more causal relationships, each indicating that one attribute is the cause of another attribute. Based on the detected causal relationships, the information processing device 10 determines a cause attribute that indicates the cause of a target attribute that indicates an evaluation of the space, and outputs information 16 corresponding to the cause attribute.
[0029] This assists in the design of other spaces that may potentially improve the attribute values of the target attributes. Therefore, even if the user does not have sufficient specific knowledge about the space being designed (for example, the trends in people flow in a particular type of facility), the user can search for a suitable space design with a small number of attempts. Furthermore, the user does not need to exhaustively try out a large number of slightly different space designs, reducing the burden of computer-based behavioral simulation.
[0030] The information 16 may also include modification suggestion information indicating a method for modifying the space to change the attribute value of the causal attribute. This allows modification of the space design to indirectly change the attribute value of the target attribute through changes in the attribute values of the behavioral attribute and the environmental attribute. This allows the user to find a suitable space design with fewer attempts.
[0031] Furthermore, the information processing device 10 may extract attributes classified into any of purpose attributes, behavior attributes, environmental attributes, external factor attributes, and target attributes from the simulation result data 13, and may detect causal relationships among the extracted attributes. This clarifies the relationships among the purpose attributes, behavior attributes, environmental attributes, external factor attributes, and target attributes, and presents logic for changing the attribute values of the target attributes. This makes it easier to plan a space design that changes the attribute values of the target attributes.
[0032] Furthermore, the information processing device 10 may calculate a causal effect between the target attribute and other attributes, and may determine a causal attribute based on the detected causal relationship and the calculated causal effect, thereby presenting causal attributes that have a large influence on the target attribute.
[0033] Furthermore, the simulation result data 13 may include records with different attribute values of the external factor attribute, and the information processing device 10 may calculate the causal effect for each attribute value of the external factor attribute. This allows the information processing device 10 to distinguish between cause attributes that have a large effect on the target attribute regardless of external factors and cause attributes that have a large effect on the target attribute only under a specific external factor, and to provide information 16 that is useful for space design.
[0034] Furthermore, the information processing device 10 may receive modified spatial information after outputting the information 16, and may execute a behavioral simulation using the modified spatial information, thereby allowing the user to confirm the effect of a design change made in accordance with the information 16.
[0035] Second Embodiment Next, a second embodiment will be described. An information processing device 100 of the second embodiment executes a human behavior simulation using an agent-based model. As an example, the information processing device 100 simulates passenger behavior in an airport terminal building with the aim of improving passenger satisfaction before boarding. The information processing device 100 may be a client device or a server device. The information processing device 100 may also be called a computer or a simulation device. The information processing device 100 corresponds to the information processing device 10 of the first embodiment.
[0036] 2 is a diagram illustrating an example of hardware of an information processing apparatus according to the second embodiment. The information processing apparatus 100 includes a CPU 101, a RAM 102, a HDD 103, a GPU 104, an input interface 105, a media reader 106, and a communication interface 107, all connected via a bus. The CPU 101 corresponds to the processing unit 12 in the first embodiment. The RAM 102 or the HDD 103 corresponds to the storage unit 11 in the first embodiment.
[0037] The CPU 101 is a processor that executes program instructions. The CPU 101 loads programs and data stored in the HDD 103 into the RAM 102 and executes the programs. The information processing apparatus 100 may have multiple processors.
[0038] The RAM 102 is a volatile semiconductor memory that temporarily stores programs executed by the CPU 101 and data used in calculations by the CPU 101. The information processing device 100 may also include a type of volatile memory other than RAM.
[0039] The HDD 103 is a non-volatile storage that stores software programs such as an operating system (OS), middleware, and application software, as well as data. The information processing device 100 may also include other types of non-volatile storage, such as a flash memory or an SSD (Solid State Drive).
[0040] The GPU 104 performs image processing in cooperation with the CPU 101 and outputs an image to a display device 111 connected to the information processing device 100. The display device 111 is, for example, a CRT (Cathode Ray Tube) display, a liquid crystal display, an organic EL (Electro Luminescence) display, or a projector. Other types of output devices, such as a printer, may also be connected to the information processing device 100.
[0041] The GPU 104 may also be used as a general purpose computing on graphics processing unit (GPGPU). The GPU 104 may execute a program in response to an instruction from the CPU 101. The information processing apparatus 100 may include a volatile semiconductor memory other than the RAM 102 as a GPU memory.
[0042] The input interface 105 receives an input signal from an input device 112 connected to the information processing apparatus 100. The input device 112 is, for example, a mouse, a touch panel, or a keyboard. A plurality of input devices may be connected to the information processing apparatus 100.
[0043] The medium reader 106 is a reading device that reads programs and data recorded on the recording medium 113. The recording medium 113 is, for example, a magnetic disk, an optical disk, or a semiconductor memory. Magnetic disks include flexible disks (FDs) and HDDs. Optical disks include compact discs (CDs) and digital versatile discs (DVDs). The medium reader 106 copies the programs and data read from the recording medium 113 to other recording media such as the RAM 102 or the HDD 103. The read programs may be executed by the CPU 101.
[0044] The recording medium 113 may be a portable recording medium. The recording medium 113 may be used to distribute programs and data. The recording medium 113 and the HDD 103 may also be referred to as computer-readable recording media.
[0045] The communication interface 107 communicates with other information processing devices via the network 114. The communication interface 107 may be a wired communication interface connected to a wired communication device such as a switch or a router, or may be a wireless communication interface connected to a wireless communication device such as a base station or an access point.
[0046] Next, the space to be simulated will be described. Figure 3 is a diagram showing an example design of an airport terminal building. The airport terminal building to be simulated has an entrance 30, an exit 31, and facilities 32 to 39 and 40 to 45. Passengers enter the airport terminal building through entrance 30 and go through check-in procedures at the check-in counter. After completing check-in procedures, passengers exit through exit 31 and head to the boarding gate. However, passengers may also use facilities 32 to 39 and 40 to 45.
[0047] Facilities 32, 33, and 45 are mobile phone shops. Facilities 34, 37, and 42 are currency exchange offices. Facilities 35, 38, and 43 are ATMs (Automatic Teller Machines). Facility 36 is a bookstore. Facility 39 is a restaurant. Facilities 40 and 41 are souvenir shops. Facility 44 is an insurance agency. Passengers may take a rest in the lobby. The airport terminal building further has information boards (not shown). The information boards indicate the locations of facilities 32 to 39 and 40 to 45. For example, the airport terminal building has three information boards, A, B, and C.
[0048] Multiple agents representing passengers enter this virtual airport terminal building. The agents act autonomously according to a certain behavioral algorithm. The agents are randomly assigned a facility to use. The agents select the desired facility, search for the facility on information boards around their current location, and walk to the desired facility. If there are no information boards around their current location or if the area around the information board is crowded, it may take some time to search for information. If the desired facility is crowded, a waiting time may occur. The agents are randomly assigned a waiting policy that indicates whether or not they will wait in line.
[0049] The information processing device 100 executes a human behavior simulation under multiple scenarios with different COVID-19 countermeasures. The first scenario shows a case where no COVID-19 countermeasures are implemented. The second scenario shows a case where social distancing is maintained between passengers. The third scenario shows a case where the number of simultaneous users of each facility is reduced. COVID-19 countermeasures are external factors that restrict the behavior of agents and are included in the spatial design.
[0050] The human behavior simulation of the second embodiment tracks the behavior of each of multiple agents by advancing time in unit time (e.g., one second) steps. For example, the information processing device 100 executes the human behavior simulation using the method described in Non-Patent Document 1 mentioned above. External factors such as the installation of signboards and COVID-19 countermeasures affect the behavior of agents. The behavior of agents affects environmental conditions such as the degree of congestion in facilities. Furthermore, the environmental conditions affect the behavior of agents.
[0051] The human behavior simulation ultimately calculates an index value indicating the satisfaction level of each agent. Possible satisfaction indices include the number of facilities available and waiting time. The spatial design of an airport terminal building affects passenger satisfaction. Therefore, the user of the information processing device 100 searches for a spatial design that will increase passenger satisfaction.
[0052] However, a single human behavior simulation only shows passenger satisfaction for one spatial design and does not explain the causal chain from external factors to passenger satisfaction. Therefore, it is not easy for a user of the information processing device 100 to decide on a spatial design modification policy, and they may prepare multiple spatial designs by trial and error based on their specific knowledge of airport terminal buildings.
[0053] Therefore, the information processing device 100 executes a human behavior simulation under several scenarios, analyzes the simulation results, and determines the causal relationships among external factors, passenger objectives, passenger behavior, environmental conditions, and passenger satisfaction. Then, the information processing device 100 proposes an indirect control method for indirectly changing passenger satisfaction based on the causal relationships.
[0054] The user of the information processing device 100 can modify the design of the airport terminal building based on the presented information. The information processing device 100 re-runs the human behavior simulation using the modified spatial information and re-evaluates the causal relationships among external factors, passenger objectives, passenger behavior, environmental conditions, and passenger satisfaction. This allows the user of the information processing device 100 to understand the adequacy of the modification policy and efficiently search for an optimal space design.
[0055] Next, an analysis of the results of the human behavior simulation will be described. FIG. 4 is a diagram showing an example of an agent table. The information processing device 100 stores an agent table 131. The agent table 131 includes multiple records corresponding to multiple agents. One record indicates the behavioral results of one agent. The agent table 131 contains a mixture of records for multiple scenarios with different COVID-19 countermeasures. Therefore, the agent table 131 is generated by multiple human behavior simulations.
[0056] The agent table 131 includes multiple attributes. The attributes include ID, time, agent type, planned tasks, and area route. The attributes also include route walk time, total waiting time, facility waiting time, visited facilities, and information search time. The attributes also include active time, sign frequency, sign reference, target, and COVID-19 countermeasures.
[0057] A record corresponds to a row (tuple) of the agent table 131. An attribute corresponds to a column of the agent table 131. Each of the multiple records includes multiple attribute values corresponding to the multiple attributes described above.
[0058] The ID is an identifier that identifies an agent. The same ID is not used between different scenarios. The time indicates the period during which the agent stayed in the airport terminal building, and includes the entry time and exit time. The time is expressed, for example, as the number of seconds from the start of the simulation. The agent may be given a maximum exit time or a maximum stay time so that the agent can make it in time for the departure time of the flight.
[0059] The agent type is the task with the highest priority for the agent. The scheduled tasks are a set of tasks that the agent wants to perform. One task indicates the use of one facility. One or more tasks are randomly assigned to an agent, and the highest priority task is randomly selected from the one or more tasks.
[0060] In the second embodiment, C may represent a mobile phone shop, X a currency exchange, A an ATM, B a bookstore, R a restaurant, S a souvenir shop, I an insurance agency, and L a lobby. A travel route indicates the tendency of an agent to travel along a route when traveling to a destination facility, and is randomly assigned to the agent. A travel route is either a straight route that takes the shortest route to the destination facility, or a not straight route that takes a route slightly deviating from the shortest route.
[0061] The travel time is the time it takes for an agent to walk within the airport terminal building towards a destination facility. The travel time is expressed, for example, in seconds. The total waiting time is the time the agent spends waiting within the airport terminal building. The total waiting time includes the time spent waiting in a queue at the facility, the time spent waiting to pass through an entrance or exit, the time spent waiting to look at a sign, etc. The total waiting time is expressed, for example, in seconds.
[0062] The facility waiting time is the time spent waiting in a queue at a facility out of the total waiting time. The facility waiting time is expressed in, for example, seconds. The used facilities are the facilities that the agent actually used during the period of his / her stay in the airport terminal building. The used facilities may be a subset of the scheduled task. The agent may not be able to use some of the facilities in the scheduled task due to lack of time or congestion. The information search time is the time it takes the agent to find the location of the destination facility. The information search time is expressed in, for example, seconds.
[0063] The active time is the time that the agent was active in the airport terminal building. Waiting time is excluded from the active time. The active time is expressed, for example, as a percentage of the stay time. The signboard frequency is the frequency with which the agent looked at the signboard. The signboard frequency is either Frequent, which indicates that the number of times the agent looked at the signboard exceeds a threshold (for example, 25 times), or Not Frequent, which indicates that the number of times the agent looked at the signboard does not exceed the threshold.
[0064] The referenced board is the board that the agent has viewed most frequently. The referenced boards can be, for example, board A, board B, board C, or even. Even indicates that the difference in the number of references between boards is less than a threshold. The completion level indicates the percentage of tasks that have been completed among the scheduled tasks, and is used as an indicator of the agent's satisfaction. The completion level is 0 if the percentage of completed tasks is less than 50%, and 1 if the percentage of completed tasks is 50% or more.
[0065] COVID-19 countermeasures identify scenarios. Airport terminal buildings with different COVID-19 countermeasures represent different spatial designs. 0 indicates no COVID-19 countermeasures. 1 indicates that social distancing is being implemented as a COVID-19 countermeasure. 2 indicates that facility capacity restrictions are being implemented as a COVID-19 countermeasure.
[0066] 5 is a diagram showing an example of an environment table. The information processing device 100 stores an environment table 132. The environment table 132 indicates the facility status at each time during a human behavior simulation. One record indicates the facility status at one time. The environment table 132 contains a mixture of records for multiple scenarios with different COVID-19 countermeasures. Therefore, the environment table 132 is generated in addition to the agent table 131 by multiple human behavior simulations.
[0067] The environment table 132 includes multiple attributes. The attributes include time (Time), multiple queue lengths (Queue), multiple facility occupancy numbers (Instore), multiple congestion levels (Congested), and COVID-19 countermeasures (Covid No). Records correspond to rows (tuples) in the environment table 132. Attributes correspond to columns (columns) in the environment table 132. Each of the multiple records includes multiple attribute values corresponding to the multiple attributes described above.
[0068] The time is a relative time from the start of the simulation. The time is expressed, for example, in seconds. The queue length is the length of the queue at a certain facility, i.e., the number of agents waiting in line. The facility population is the number of agents in a certain facility. The congestion level is a flag that indicates whether a certain facility is congested. For example, if the queue length is below a threshold, the facility is not congested, and if the queue length exceeds the threshold, the facility is congested. The queue length, facility population, and congestion level are prepared for each facility. COVID-19 countermeasures identify scenarios in the same way as the agent table 131.
[0069] 6 is a diagram showing an example of a combined table. The information processing device 100 combines the agent table 131 and the environment table 132 to generate the combined table 133. The information processing device 100 selects one record from the agent table 131 and searches the environment table 132 for a record having a time corresponding to the selected record. The information processing device 100 adds the attribute value included in the searched record or a value converted from the attribute value to the selected record. The information processing device 100 may delete some attribute values from the selected record, or may delete some attribute values from the searched record.
[0070] In the example of FIG. 6, the visited facility is deleted from the agent table 131, and Visit Congested F is added to the agent table 131. Visit Congested F is a flag that indicates whether or not the agent has visited a congested facility. If the agent has visited a congested facility, the Visit Congested F is Visited. If the agent has not visited a congested facility, the Visit Congested F is Not Visited.
[0071] The information processing device 100 searches the environment table 132 for records having times that belong to the agent's stay period. The information processing device 100 extracts the congestion level corresponding to the used facility from the searched records. If the used facility is crowded, the information processing device 100 sets the crowded facility use to "Visited." If the used facility is not crowded, the information processing device 100 sets the crowded facility use to "Not Visited."
[0072] 7 is a diagram showing an example of a normalized join table. The information processing apparatus 100 converts the join table 133 into a join table 134 by normalizing the attribute values included in the join table 133 for each attribute. The facilities indicated by the scheduled tasks and agent types are converted into integers. For example, an ATM is converted to 1, a restaurant to 2, a souvenir shop to 3, a currency exchange to 4, and a lobby to 5.
[0073] Flags indicating travel route, signboard frequency, and crowded facility use are converted to 0 or 1. For example, for travel route, "Straight" is converted to 0 and "Not Straight" is converted to 1. For signboard frequency, "Frequent" is converted to 1 and "Not Frequent" is converted to 0. For crowded facility use, "Visited" is converted to 0 and "Not Visited" is converted to 1. Reference signboards are also converted to integers. For example, signboard A is converted to 1, signboard B is converted to 2, signboard C is converted to 3, and "Even" is converted to 4. Travel time, total waiting time, facility waiting time, and information search time are scaled to a value between 0 and 1 for each attribute. This normalization improves the accuracy of subsequent analyses.
[0074] 8 is a diagram showing an example of an attribute table. The information processing device 100 classifies the attributes included in the combined table 134 into five attribute types: passenger purpose, passenger behavior, environmental conditions, external factors, and passenger satisfaction. Passenger satisfaction may also be called passenger experience. However, there may be attributes that are not classified into any of the attribute types. The passenger satisfaction attribute corresponds to the target attribute in the first embodiment.
[0075] Passenger objectives are attributes that indicate the objectives given to the agent. Passenger behaviors are attributes that indicate measurements of the behaviors autonomously chosen by the agent. Environmental conditions are attributes that indicate measurements of the interaction between the agent and the environment. External factors are attributes that indicate the constraints that the agent is subjected to from the environment. Passenger satisfaction is an attribute that indicates an estimate of the agent's satisfaction and represents an evaluation of the spatial design of the airport terminal building.
[0076] The user may specify attribute types for some or all of the attributes. Alternatively, the information processing device 100 may determine the attribute types from the attribute names. For example, the information processing device 100 refers to a knowledge database that indicates the relationship between attribute names and attribute types, and classifies the attributes included in the combined table 134 into five attribute types.
[0077] The information processing device 100 may also determine the attribute type based on a human behavior simulation algorithm. For example, the information processing device 100 classifies an attribute corresponding to an agent variable whose value is set randomly in advance as passenger purpose. The information processing device 100 classifies an attribute corresponding to an agent variable whose value changes during the simulation as passenger behavior. The information processing device 100 classifies an attribute corresponding to a facility-related variable as environmental condition. The information processing device 100 classifies an attribute corresponding to a variable indicating the design of an airport terminal building as external factor. The information processing device 100 classifies an attribute corresponding to a variable whose value is calculated ex post from the behavior result as passenger satisfaction.
[0078] The information processing device 100 generates an attribute table 135 that indicates the correspondence between attributes and attribute types. In the example of FIG. 8 , the agent type is classified into passenger purpose. The travel route, travel time, total waiting time, and activity time are classified into passenger behavior. Furthermore, the facility waiting time, information search time, signboard frequency, signboard reference frequency, and facility use during congestion are classified into environmental conditions. Furthermore, COVID-19 countermeasures are classified into external factors. Furthermore, the achievement level is classified into passenger satisfaction.
[0079] The information processing device 100 extracts attribute values of attributes classified into one of five attribute types: passenger purpose, passenger behavior, environmental state, external factors, and passenger satisfaction, from the merged table 134. The information processing device 100 detects causal relationships between multiple attributes by executing a causal search method on multiple records including the extracted attribute values. The attributes to be analyzed are five types: passenger purpose, passenger behavior, environmental state, external factors, and passenger satisfaction. However, the information processing device 100 may also analyze causal relationships by including attributes other than these five types of attributes.
[0080] Examples of causal search algorithms include GFCI (Greedy Fast Causal Inference), DirectLiNGAM (Direct Linear Non-Gaussian Acyclic Model), and PC (Peter and Clark) algorithm.
[0081] GFCI is described, for example, in the following non-patent document: Juan Miguel Ogarrio, Peter Spirtes, and Joe Ramsey, "A Hybrid Causal Search Algorithm for Latent Variable Models," Proc. of the 8th International Conference on Probabilistic Graphical Models (PGM 2016), pp. 368-379, September 2016.
[0082] DirectLiNGAM is described in, for example, the following non-patent document: Shohei Shimizu, Takanori Inazumi, Yasuhiro Sogawa, Aapo Hyvarinen, Yoshinobu Kawahara, Takashi Washio, Patrik O. Hoyer and Kenneth Ballen, "DirectLiNGAM: A Direct Method for Learning a Linear Non-Gaussian Structural Equation Model," Journal of Machine Learning Research, Volume 12, pp. 1225-1248, April 2011.
[0083] The PC algorithm is described in, for example, the following non-patent document: Peter Spirtes and Clark Glymour, "An Algorithm for Fast Recovery of Sparse Causal Graphs," Social Science Computer Review, Volume 9, Issue 1, pp. 62-72, April 1991.
[0084] 9 is a diagram showing an example of a causal graph. The information processing device 100 generates a causal graph by executing a causal search method. The causal graph includes a plurality of nodes each representing an attribute and a plurality of edges each representing a causal relationship between two attributes. For example, the node contains the attribute name and attribute type. The edge has a direction from the attribute of the cause to the attribute of the result.
[0085] The causal graph shown in FIG. 9 includes nodes 140, 141, 142, 143, 144, 145, 146, and 147. Node 140 represents agent type, which is a passenger purpose attribute. Node 141 represents travel route, which is a passenger behavior attribute. Node 142 represents reference sign, which is an environmental state attribute. Node 143 represents travel time, which is a passenger behavior attribute. Node 144 represents facility waiting time, which is an environmental state attribute. Node 145 represents total waiting time, which is a passenger behavior attribute. Node 146 represents information search time, which is an environmental state attribute. Node 147 represents achievement level, which is a passenger satisfaction attribute.
[0086] There are edges representing causal relationships from node 140 to nodes 141, 142, 143, 144, 145, 146, and 147. There are edges representing causal relationships from node 142 to nodes 143, 144, and 146. There are edges representing causal relationships from node 143 to nodes 142, 144, and 147. There are edges representing causal relationships from node 144 to nodes 142, 145, and 147. There are edges representing causal relationships from node 145 to nodes 144 and 146. There are edges representing causal relationships from node 146 to nodes 145 and 147.
[0087] A causal graph shows the causal relationships between passenger objectives, passenger behavior, environmental conditions, external factors, and passenger satisfaction. A typical causal graph shows the causal chain in which passenger objectives affect passenger satisfaction through passenger behavior, environmental conditions, and external factors. Therefore, a causal graph with attribute types is one of the useful pieces of information when considering improvement measures to improve passenger satisfaction.
[0088] Furthermore, the information processing device 100 calculates the causal effect between attributes from the join table 134, separately from the causal graph, using a causal search method. The causal effect is an index showing the strength of the causal relationship, and represents the ratio of the change in the attribute value of the result to the change in the attribute value of the cause. If the attribute value of the result increases as the attribute value of the cause increases, the causal effect is positive. If the attribute value of the result increases as the attribute value of the cause decreases, the causal effect is negative. The larger the magnitude of the causal effect (the absolute value of the causal effect), the stronger the causal relationship.
[0089] The information processing device 100 calculates the causal effect related to passenger satisfaction by regarding each attribute of passenger purpose, passenger behavior, environmental state, and external factors as a cause and passenger satisfaction as a result. However, the information processing device 100 may also calculate the causal effect between passenger satisfaction and all attributes other than passenger satisfaction. When calculating the causal effect, the information processing device 100 classifies the records included in the join table 134 into multiple scenarios based on the attribute values of the external factors. The information processing device 100 calculates the causal effect for each scenario. This is because different attribute values of the external factors may result in different strengths of causal relationships. Note that when generating a causal graph, the information processing device 100 does not need to classify the records into multiple scenarios.
[0090] FIG. 10 is a diagram showing an example of calculation of the causal effect for each scenario. In the graph of FIG. 10, the horizontal axis represents the attribute value of COVID-19 countermeasures, and the vertical axis represents the causal effect. The records in the combined table 134 are classified into three scenarios. FIG. 10 lists only some attributes with large absolute values of the causal effect.
[0091] Curve 161 shows the causal effect between facility waiting time and achievement level in the three scenarios. Curve 162 shows the causal effect between activity time and achievement level in the three scenarios. Curve 163 shows the causal effect between travel time and achievement level in the three scenarios. Curve 164 shows the causal effect between information seeking time and achievement level in the three scenarios.
[0092] Based on the causal graph and the causal effects, the information processing device 100 identifies the main causes that have a significant impact on passenger satisfaction from among the multiple attributes included in the combined table 134. First, the information processing device 100 searches the causal graph for edges that are directly connected to nodes indicating passenger satisfaction and whose end points are nodes indicating passenger satisfaction. The information processing device 100 determines the attributes on the start points of the searched edges as direct causes of passenger satisfaction.
[0093] At this time, multiple direct causes may be detected. The information processing device 100 identifies a main cause from the multiple direct causes based on the causal effect. The information processing device 100 preferentially selects a direct cause with a large absolute value of the causal effect as the main cause. The information processing device 100 may narrow down the main cause to one, or may allow two or more main causes. The information processing device 100 may identify main causes by dividing by scenario.
[0094] For example, the information processing device 100 selects direct causes whose absolute values of causal effects exceed a threshold as main causes. Furthermore, for example, the information processing device 100 selects a certain number of direct causes whose absolute values of causal effects are largest as main causes. The information processing device 100 may make the above selection by dividing direct causes into those with positive causal effects and those with negative causal effects.
[0095] The information processing device 100 may determine a direct cause that satisfies the above conditions for all scenarios as a scenario-independent main cause. At this time, the information processing device 100 may output information indicating that the main cause is scenario-independent. On the other hand, the information processing device 100 may determine a direct cause that satisfies the above conditions for some scenarios as a scenario-dependent main cause. At this time, the information processing device 100 may output information indicating that the main cause is valid only in a specific scenario.
[0096] For example, the information processing device 100 identifies, as direct causes, the agent type indicated by node 140, the travel time indicated by node 143, the facility waiting time indicated by node 144, and the information search time indicated by node 146 from the causal graph shown in Fig. 9. The information processing device 100 refers to the causal effects shown in Fig. 10 and selects the facility waiting time and information search time as the main causes from among the agent type, travel time, facility waiting time, and information search time.
[0097] The information processing device 100 outputs information about the identified main cause as explanatory information explaining why the attribute value of passenger satisfaction was calculated. The explanatory information may include the attribute name and attribute type of the identified main cause. The explanatory information may also include a causal graph showing the relationship between passenger purpose, passenger behavior, environmental conditions, external factors, and passenger satisfaction. The explanatory information may also include information showing the causal effect between passenger satisfaction and other attributes.
[0098] The information processing device 100 may also output a remedy for changing the attribute value of the main cause by including it in the explanation information. This remedy is an indirect control method for indirectly changing the attribute value of passenger satisfaction. To propose a remedy, the information processing device 100 may store a knowledge database in which attributes are associated with methods for changing the attribute values of the attributes.
[0099] 11 is a diagram showing an example of a remedial action table. The information processing device 100 stores the remedial action table 136. The remedial action table 136 associates main causes with remedial actions. The remedial action indicates a design change to the airport terminal building that may change the attribute value of the main cause. The remedial action may correspond to the addition of an external factor.
[0100] One example of an improvement measure that changes the attribute values of facility waiting time or information search time is to display congestion information on a guide board that shows the current congestion level at each of multiple facilities. Displaying congestion information can reduce the bias in congestion levels over time and between facilities, and potentially reduce the average facility waiting time. Displaying congestion information can also reduce the effort required by agents to search for uncrowded facilities, potentially reducing information search time.
[0101] Possible examples of congestion information display formats include a display format in which congestion information is displayed only on sign board A among sign boards A, B, and C, and a display format in which congestion information is displayed only on sign boards B and C among sign boards A, B, and C. When attempting human behavior simulations using these different display formats, an attribute called sign board information (sign number) is added as an external factor. In this case, scenarios are distinguished by the combination of COVID-19 countermeasures and sign board information.
[0102] After outputting the explanatory information, the information processing device 100 receives corrected spatial information from the user. For example, the information processing device 100 receives spatial information from the user in which the display content of a guide board has been changed. The information processing device 100 re-executes the human behavior simulation using the corrected spatial information. The information processing device 100 then analyzes the results of the re-simulation and outputs the explanatory information again.
[0103] By checking the causal relationships between attributes after the re-simulation, the user can understand the effect of modifying the spatial information. In this way, the information processing device 100 iteratively outputs explanatory information and obtains modified spatial information, instead of initially accepting all spatial information candidates and exhaustively executing human behavior simulations. This allows the user to efficiently search for spatial designs that will improve passenger satisfaction.
[0104] 12 is a diagram showing an example of a causal graph after re-simulation. The information processing device 100 generates an updated causal graph by executing a causal search method on the re-simulation result. The causal graph shown in FIG. 12 includes nodes 150, 151, 152, 153, 154, 155, 156, 157, 158, and 159.
[0105] Node 150 represents the travel route, which is a passenger behavior attribute. Node 151 represents the activity time, which is a passenger behavior attribute. Node 152 represents the agent type, which is a passenger purpose attribute. Node 153 represents the travel time, which is a passenger behavior attribute. Node 154 represents the reference signboard, which is an environmental state attribute. Node 155 represents the facility waiting time, which is an environmental state attribute. Node 156 represents the total waiting time, which is a passenger behavior attribute. Node 157 represents the signboard information, which is an external factor attribute. Node 158 represents the information search time, which is an environmental state attribute. Node 159 represents the achievement level, which is a passenger satisfaction attribute.
[0106] There is an edge representing a causal relationship from node 150 to node 151. There is an edge representing a causal relationship from node 151 to node 153. There are edges representing causal relationships from node 152 to nodes 153, 154, 156, 158, and 159. There are edges representing causal relationships from node 153 to nodes 154, 155, and 157.
[0107] There are edges representing causal relationships from node 154 to nodes 153 and 155. There are edges representing causal relationships from node 155 to nodes 154, 156, and 157. There is an edge representing causal relationships from node 156 to node 155. There are edges representing causal relationships from node 157 to nodes 153, 155, and 158. There are edges representing causal relationships from node 158 to nodes 157 and 159. Unlike the causal graph shown in FIG. 9 , in the resimulation results, facility waiting time is no longer a direct cause of achievement.
[0108] FIG. 13 is a diagram showing an example of calculating the causal effect after resimulation. In the graph of FIG. 13, the horizontal axis represents the scenario, and the vertical axis represents the causal effect. There are six scenarios depending on the combination of the attribute values of the COVID-19 countermeasures and the attribute values of the guide board information. Curve 165 shows the causal effect between facility waiting time and achievement level in the six scenarios. Curve 166 shows the causal effect between activity time and achievement level in the six scenarios. Curve 167 shows the causal effect between travel time and achievement level in the six scenarios. Curve 168 shows the causal effect between information search time and achievement level in the six scenarios.
[0109] The information processing device 100 may again identify the main cause based on the causal graph and causal effects after the re-simulation and output explanation information. Next, a supplementary explanation will be given of the flow of the simulation executed by the information processing device 100 according to the second embodiment.
[0110] 14 is a diagram showing an example of the flow of a human behavior simulation. The information processing device 100 executes an agent-based simulation 171 for multiple scenarios. The information processing device 100 performs data preprocessing 172 on the agent table and environment table generated by the agent-based simulation 171 to generate a combined table. The data preprocessing 172 includes classifying multiple attributes included in the combined table into passenger purpose, passenger behavior, environmental state, external factors, and passenger satisfaction.
[0111] The information processing device 100 performs causal relationship detection 173 on the combined table to generate a causal graph. The information processing device 100 also performs causal effect calculation 174 on the combined table to calculate the causal effect between passenger satisfaction and other attributes. Either the causal relationship detection 173 or the causal effect calculation 174 may be performed first.
[0112] The information processing device 100 executes an indirect control method determination 175 based on the causal graph and the causal effects. The indirect control method determination 175 includes identifying a main cause that has a significant impact on passenger satisfaction. Furthermore, the indirect control method determination 175 includes searching for an improvement measure that changes the attribute value of the main cause, as an indirect control method. The information processing device 100 outputs explanatory information that includes information indicating the main cause and the improvement measure. When corrected spatial information is input in response to the explanatory information, the information processing device 100 re-executes the agent-based simulation 171.
[0113] 15 is a diagram showing an example of the relationship between attribute types. In the human behavior simulation, a passenger objective 181 is given to an agent. The agent acts with the aim of achieving the passenger objective 181. Therefore, the passenger objective 181 influences passenger behavior 182. In addition, an external factor 183 is defined as the design of the airport terminal building. The external factor 183 influences passenger behavior 182.
[0114] As a result of passenger behavior 182, the congestion level of the facility changes. Therefore, passenger behavior 182 affects environmental conditions 184. Furthermore, the congestion level of the facility may change passenger behavior choices. Therefore, environmental conditions 184 affects passenger behavior 182. In this way, passenger behavior 182 and environmental conditions 184 influence each other and change over time. Ultimately, passenger satisfaction 185 is determined by the total of passenger behavior 182 and environmental conditions 184.
[0115] Next, the functions and processing procedures of the information processing device 100 will be described. Fig. 16 is a block diagram showing an example of functions of the information processing device. The information processing device 100 has a design information storage unit 121, a result data storage unit 122, an improvement measure storage unit 123, a simulator 124, a preprocessing unit 125, a causal analysis unit 126, and an improvement measure determination unit 127. The design information storage unit 121, the result data storage unit 122, and the improvement measure storage unit 123 are implemented using, for example, the RAM 102 or the HDD 103. The simulator 124, the preprocessing unit 125, the causal analysis unit 126, and the improvement measure determination unit 127 are implemented using, for example, the CPU 101 and a program.
[0116] The design information storage unit 121 stores spatial information of an airport terminal building. The spatial information includes the shape of the airport terminal building, the locations of facilities, the locations of signboards, and the like. The spatial information may include behavioral rules that restrict the behavior of agents, such as COVID-19 countermeasures. Parameters included in the spatial information may correspond to external factors.
[0117] The result data storage unit 122 stores the results of the human behavior simulation. For example, the result data storage unit 122 stores the agent table 131 and the environment table 132 described above. The improvement plan storage unit 123 stores, for each attribute, improvements that may change the attribute value. For example, the improvement plan storage unit 123 stores the improvement plan table 136 described above.
[0118] An agent-based human behavior simulation is performed based on the spatial information stored in the design information storage unit 121. For example, the simulator 124 generates a certain number of agents and randomly sets parameter values such as arrival time, task, and facility usage waiting policy for each agent. The simulator 124 moves the agents in the virtual space according to the set parameter values. The simulator 124 saves the results of the human behavior simulation in the result data storage unit 122.
[0119] The preprocessing unit 125 performs preprocessing on the result data stored in the result data storage unit 122. The preprocessing unit 125 generates a combined table 133 by combining the agent table 131 and the environment table 132, and normalizes the combined table 133 to generate a combined table 134. The preprocessing unit 125 classifies the multiple attributes included in the combined table 134 into five attribute types, and generates an attribute table 135.
[0120] The causal analysis unit 126 performs a causal search method on the combined table 134 generated by the preprocessing unit 125. As a result, the causal analysis unit 126 generates a causal graph showing the causal relationships among the five types of attributes and calculates the causal effects of other attributes on the passenger satisfaction attribute.
[0121] The improvement measure determination unit 127 searches for direct causes connected to the passenger satisfaction attribute by a single causal relationship from the causal graph generated by the causal analysis unit 126. The improvement measure determination unit 127 identifies, from the direct causes, a main cause whose absolute value of the causal effect on the passenger satisfaction attribute is large. The improvement measure determination unit 127 searches the improvement measure storage unit 123 for an improvement measure corresponding to the main cause.
[0122] As a result, the improvement measure determination unit 127 generates explanatory information indicating an analysis of the simulation results. The explanatory information includes, for example, information identifying the identified main cause and text indicating the searched improvement. The improvement measure determination unit 127 outputs the explanatory information. The improvement measure determination unit 127 may store the explanatory information in a non-volatile storage, display it on the display device 111, or transmit it to another information processing device.
[0123] 17 is a flowchart showing an example of a procedure for determining an agent's behavior. The simulator 124 determines the agent's behavior in accordance with this flowchart. (S10) The agent selects one target facility to use from the scheduled tasks.
[0124] (S11) The agent does not know the location of the target facility, so it walks randomly. (S12) The agent determines whether there is a signboard within a certain range from the current position. If there is a signboard, the process proceeds to step S13. If there is no signboard, the process returns to step S11, and the agent continues walking randomly.
[0125] (S13) The agent acquires information indicating the location of the target facility from the guide board. (S14) The agent walks toward the target facility based on the acquired information about the target facility. Whether or not the agent moves along the shortest route depends on the setting value of the agent.
[0126] (S15) The agent determines whether a certain amount of time has passed since the last time the information on the target facility was acquired. If the certain amount of time has passed since the information was acquired, the process proceeds to step S16. If the certain amount of time has not yet passed since the information was acquired, the process proceeds to step S17.
[0127] (S16) The agent forgets the acquired information about the target facility. Then, the process returns to step S11, and the agent starts walking randomly from its current position. (S17) The agent determines whether it has arrived at the target facility. If it has arrived at the target facility, the process proceeds to step S18. If it has not arrived at the target facility, the process returns to step S14, and the agent continues walking toward the target facility.
[0128] (S18) The agent determines whether there are any waiting passengers at the target facility upon arrival, i.e., whether the length of the queue is 1 or more. If there are waiting passengers, the process proceeds to step S19. If there are no waiting passengers, the process proceeds to step S20.
[0129] (S19) The agent determines whether a setting value indicating a policy of waiting one's turn has been set. If the policy of waiting one's turn is set, the process proceeds to step S20. If the policy of waiting one's turn is not set, the process returns to step S10, and the agent abandons the selected facility and selects another facility.
[0130] (S20) The agent enters the target facility and uses it. If the agent is in a queue, the agent can enter the target facility if a spot becomes available at the target facility while the agent is at the head of the queue. The usage time depends on the type of the target facility. After using the facility, if there is an unused facility in the scheduled task and the scheduled time to leave has not yet arrived, the agent re-executes the flowchart in FIG. 17.
[0131] 18 is a flowchart showing an example of a simulation procedure. (S30) The simulator 124 acquires spatial information of the simulation target. (S31) The simulator 124 executes an agent-based human behavior simulation using multiple scenarios defined in the spatial information. Different scenarios have different attribute values for external factors. This generates an agent table and an environment table.
[0132] (S32) The preprocessing unit 125 combines the agent table and the environment table. For example, based on a time included in the agent table, the preprocessing unit 125 searches the environment table for a record that indicates the environmental state at that time, and combines the record in the agent table with the record in the environment table. The preprocessing unit 125 normalizes the combined table generated by the combination so that the attribute value of each attribute becomes an integer or a continuous value between 0 and 1.
[0133] (S33) The preprocessing unit 125 classifies the attributes included in the combined table into five attribute types: passenger purpose, passenger behavior, environmental state, external factors, and passenger satisfaction. (S34) The causal analysis unit 126 executes a causal search method on the combined table generated in step S32, and generates a causal graph showing the causal relationships among the five types of attributes.
[0134] (S35) The causal analysis unit 126 assigns attribute types to the nodes of the causal graph. (S36) The causal analysis unit 126 classifies the records included in the combined table generated in step S32 into multiple scenarios based on the attribute values of the external factor attributes.
[0135] (S37) The causal analysis unit 126 calculates the causal effect between the passenger satisfaction attribute and other attributes for each of the multiple scenarios using the records classified into that scenario. (S38) The improvement measure determination unit 127 searches the causal graph generated in step S34 for a direct cause that indicates an attribute that is directly connected to the passenger satisfaction attribute by one causal relationship.
[0136] (S39) The improvement measure determination unit 127 determines, from the direct causes, main causes that have a strong causal relationship with the passenger satisfaction attribute, using the causal effects calculated in step S37. For example, the improvement measure determination unit 127 selects, in at least some scenarios, direct causes whose absolute values of causal effects exceed a threshold, or a certain number of direct causes whose absolute values of causal effects are large.
[0137] (S40) The improvement measure determination unit 127 searches the improvement measure table 136 for an improvement measure that can change the attribute value of the main cause determined in step S39. (S41) The improvement measure determination unit 127 outputs explanatory information including the improvement measure of step S40, the causal graph of step S34, and the causal effect of step S37.
[0138] As described above, the information processing device 100 of the second embodiment uses an agent-based human behavior simulation to simulate human behavior under a certain spatial design and output attribute values of target attributes that indicate the quality of the spatial design. At this time, the information processing device 100 analyzes causal relationships between multiple attributes, including evaluation attributes and other attributes, to detect a causal chain between the purpose attribute, behavior attribute, environmental attribute, external factor attribute, and target attribute. The information processing device 100 then identifies the main cause for the target attribute and outputs explanatory information about the main cause.
[0139] This makes it easier for users to understand how to improve a space design by referring to the explanatory information, whereas it is difficult for users to understand how to improve a space design by looking only at the attribute values of the target attributes. Therefore, even if users do not have sufficient specific knowledge about the space to be designed, they can search for an appropriate space design with a small number of attempts. In addition, users do not need to exhaustively try out a large number of space design candidates, which reduces the load of the human behavior simulation.
[0140] In particular, the information processing device 100 presents an indirect control method for changing the attribute value of the main cause, making it easier for the user to modify the space design. Furthermore, the information processing device 100 repeatedly outputs explanatory information, accepts the modified space design, and re-executes the human behavior simulation. This allows the space design to be improved step by step, enabling the user to efficiently search for a suitable space design with fewer attempts.
[0141] Furthermore, the information processing device 100 identifies a main cause from among direct causes that have a causal relationship with the target attribute based on the causal effect. This narrows down the attributes that require attention in order to improve the attribute value of the target attribute, and prioritizes the execution of indirect control methods with the greatest improvement effect. Furthermore, the information processing device 100 calculates the causal effect for each of multiple scenarios with different attribute values for the external factor attribute. This allows the user to distinguish between improvement measures that are effective regardless of the attribute value of the external factor and improvement measures that are effective only in scenarios in which the external factor has a specific attribute value, thereby narrowing down the spatial designs to be attempted.
[0142] The foregoing merely illustrates the principles of the present invention. Further, since numerous modifications and changes will be apparent to those skilled in the art, the present invention is not limited to the exact construction and application shown and described above, and all corresponding modifications and equivalents are deemed to be within the scope of the present invention as defined by the appended claims and their equivalents.
[0143] REFERENCE SIGNS LIST 10 Information processing device 11 Storage unit 12 Processing unit 13 Simulation result data 14a, 14b, 14c, 14d Attributes 15a, 15b, 15c, 15d, 15e Causal relationship 16 Information
Claims
1. A simulation program that causes a computer to execute the following processes: acquiring simulation result data including a plurality of records corresponding to a plurality of agents acting within a specified space, each of the plurality of records including attribute values of a plurality of attributes related to the behavioral results of one of the plurality of agents; detecting one or more causal relationships from among the plurality of attributes based on the simulation result data, each of which indicates that one attribute is the cause of one of the other attributes; determining a cause attribute from among the plurality of attributes that indicates the cause for a target attribute that indicates an evaluation of the space, based on the one or more causal relationships; and outputting information corresponding to the cause attribute.
2. The simulation program according to claim 1, wherein the information corresponding to the cause attribute includes modification suggestion information indicating a method of modifying the space to change the attribute value of the cause attribute.
3. The simulation program of claim 1, further comprising causing the computer to execute a process of extracting, from the plurality of attributes, two or more attributes classified into one of the following types: a purpose attribute indicating the purpose of the one agent, a behavior attribute indicating the behavior performed by the one agent, an environment attribute indicating the interaction between the one agent and the space, an external factor attribute indicating the constraints that the space imposes on the one agent, and the target attribute, and the one or more causal relationships are detected from among the two or more attributes.
4. The simulation program of claim 1, further causing the computer to execute a process of calculating a causal effect between the target attribute and other attributes among the plurality of attributes based on the simulation result data, and the causal attribute is determined based on the one or more causal relationships and the causal effect.
5. A simulation program as described in claim 4, wherein the plurality of records includes records having different attribute values of an external factor attribute among the plurality of attributes that indicates the constraints that the space imposes on the one agent, and the causal effect is calculated for each record group having the same attribute value of the external factor attribute.
6. The simulation program of claim 1, further causing the computer to execute a process of accepting spatial information indicating a modified space after outputting information corresponding to the causal attribute, and simulating the actions of the plurality of agents within the modified space.
7. A simulation method in which a computer executes the following processes: acquiring simulation result data including a plurality of records corresponding to a plurality of agents acting within a specified space, each of the plurality of records including attribute values of a plurality of attributes related to the behavioral result of one of the plurality of agents; detecting one or more causal relationships from among the plurality of attributes based on the simulation result data, each of which indicates that one attribute is the cause of one of the other attributes; determining a cause attribute from among the plurality of attributes that indicates the cause of a target attribute that indicates an evaluation of the space based on the one or more causal relationships; and outputting information corresponding to the cause attribute.
8. An information processing device having: a memory unit that stores simulation result data including a plurality of records corresponding to a plurality of agents acting within a specified space, each of the plurality of records including attribute values of a plurality of attributes related to the action result of one of the plurality of agents; and a processing unit that detects one or more causal relationships from among the plurality of attributes based on the simulation result data, each of which indicates that one attribute is the cause of another attribute, determines a cause attribute from among the plurality of attributes that indicates the cause for a target attribute that indicates an evaluation of the space based on the one or more causal relationships, and outputs information corresponding to the cause attribute.
Citation Information
Patent Citations
People flow prediction device and method
JP2013116676A
Simulation program, simulation method and simulation device
JP2019125312A
Analysis program, analysis method, and analysis device
JP2020091749A
