Prediction program, prediction method, prediction device, and prediction system

Abstract

To provide a prediction program, prediction method, prediction device and prediction system, which improve human flow prediction accuracy.SOLUTION: A prediction program provided herein makes a computer perform steps of simulating a human flow based on intent of people on a plurality of actions (S102), evaluating the simulated human flow based on human flow observation data (S106), and executing processing for predicting the human flow on the basis of an evaluation result of the simulated human flow (S107).SELECTED DRAWING: Figure 8

Description

Technical Field

[0001] The present invention relates to a prediction program, a prediction method, a prediction device, and a prediction system.

Background Art

[0002] In situations where the flow of people concentrates, such as during an event or evacuation in case of a disaster, congestion occurs, and there is a possibility of accidents due to overcrowding or human casualties due to delayed evacuation. Therefore, if it is possible to predict the congestion situation that may occur in advance and consider countermeasures before the event occurs, it will lead to the avoidance of accidents and human casualties and risk reduction. Therefore, it becomes important to predict the future flow of people.

[0003] Agent-based simulation, which is a conventional technique that models the movement of each individual, simulates movement and interference assuming the behavior of each pedestrian, and predicts the flow of people, can predict the state of the flow of people for phenomena that have not been experienced so far.

[0004] In addition, a method of improving the prediction accuracy of simulation by incorporating observational data into simulations such as weather prevention (sometimes referred to as "data assimilation") has been studied, and the application of data assimilation to the prediction of the flow of people has also been studied.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0006] However, conventional agent-based simulations are essentially what-if scenario analyses, and since it's uncertain whether the simulated situation will actually occur, they don't accurately predict real-world conditions. Furthermore, data assimilation for pedestrian flow prediction is limited to tracking individual locations and cannot be applied to predicting pedestrian flow.

[0007] One aspect of this project is to provide prediction programs, prediction methods, prediction devices, and prediction systems that improve the accuracy of predicting human flow. [Means for solving the problem]

[0008] In one embodiment, the prediction program simulates pedestrian flow based on the intentions of multiple human actions, evaluates the simulated pedestrian flow based on observational data of pedestrian flow, and causes the computer to perform a process of predicting pedestrian flow based on the evaluation results of the simulated pedestrian flow. [Effects of the Invention]

[0009] In one respect, it can improve the accuracy of predicting pedestrian traffic. [Brief explanation of the drawing]

[0010] [Figure 1] Figure 1 is a diagram illustrating pedestrian flow and people's behavioral intentions. [Figure 2] Figure 2 shows an example of the configuration of the prediction system 1 according to Example 1. [Figure 3] Figure 3 shows an example of pedestrian flow prediction according to Example 1. [Figure 4] Figure 4 shows an example of the configuration of the prediction device 10 according to Example 1. [Figure 5] Figure 5 shows an example of pedestrian flow evaluation according to Example 1. [Figure 6] Figure 6 shows an example of the estimation results of behavioral intent according to Example 1. [Figure 7]Figure 7 shows an example of a group of density maps according to Example 1. [Figure 8] Figure 8 is a flowchart showing an example of the prediction process flow according to Example 1. [Figure 9] Figure 9 illustrates an example of a hardware configuration. [Modes for carrying out the invention]

[0011] The following describes in detail, with reference to the drawings, embodiments of the prediction program, prediction method, prediction device, and prediction system according to this embodiment. However, this embodiment is not limited to this one. Furthermore, each embodiment can be combined as appropriate within a consistent range. [Examples]

[0012] First, Figure 1 will be used to explain pedestrian flow and people's behavioral intentions. Note that people's behavioral intentions are sometimes simply referred to as "behavioral intentions." Figure 1 is a diagram for explaining pedestrian flow and people's behavioral intentions. The left side of Figure 1 shows pedestrian flow in a passageway heading in the X or Y direction. On the other hand, the right side of Figure 1 shows pedestrian flow in an area heading towards Exit A or Exit B. In situations where we want to predict pedestrian flow, there are many cases where people's behavioral intentions are limited, such as heading towards XX. More specific behavioral intentions could be, for example, in a train station, whether to transfer to line xx, line yy, or go to exit zz, or in an evacuation situation, which exit to evacuate from, when to start moving, etc., but are not limited to these. Also, multiple behavioral intentions may be combined to form a single behavioral intention, for example, heading towards the exit in 60 seconds.

[0013] Taking the behavior intention of a person as θ, the behavior intention θ cannot be observed from the flow of people. However, for example, even in a complex flow of people, it can be said that the flow of people is formed by the superposition of the behaviors of individuals who act based on several simple behavior intentions θ. For example, within the precincts of a station, there are people acting based on the behavior intention θ1 of changing trains to the xx line, people acting based on the behavior intention θ2 of changing trains to the yy line, ··· people acting based on the behavior intention θ z and the flow of people is formed by these people. And although each behavior intention θ cannot be observed and is unknown, the behaviors as the results of each behavior intention θ can be observed by tracking each person. For example, by tracking a person acting based on the behavior intention θ1 of changing trains to the xx line, it is possible to observe the behavior of that person heading in the Z direction along the passage. Since this observed behavior in the Z direction is the behavior as the result of the behavior intention θ1, if the unobservable behavior intention θ can be derived from the observed behavior, it will be possible to predict the future flow of people by simulation based on the behavior intention. Also, since agent-based simulation can be executed faster than real time, it is possible to predict the flow of people after obtaining observation data on people's behaviors, and by extension, the flow of people, and connect it to control.

[0014] Therefore, in this embodiment, one of the purposes is to simulate the flow of people from a plurality of people's behavior intentions, estimate the behavior intention by evaluating the simulation result with observation data, and predict the flow of people based on the estimation result.

[0015] [System Configuration of Prediction System 1] FIG. 2 is a diagram showing a configuration example of the prediction system 1 according to Example 1. As shown in FIG. 2, the prediction system 1 is a system in which a prediction device 10 and observation devices 100-1 to 100-n (n is an arbitrary integer. Hereinafter, collectively referred to as "observation device 100") are communicably connected to each other via a network 50.

[0016] The network 50 can adopt various communication networks such as an intranet used within a predetermined area such as inside a station, regardless of whether it is wired or wireless. Also, the network 50 is not a single network, and for example, an intranet and the Internet may be configured via a network device such as a gateway or other devices (not shown).

[0017] The prediction device 10 receives the data observed by the observation device hundred from the observation device hundred. Here, the observed data may be, for example, video or images captured by the observation device hundred, or a density map of the flow of people at a predetermined time generated from the video or images, but is not limited thereto. The prediction device 10 simulates the flow of people from a plurality of people's action intentions, estimates the action intentions by evaluating the simulation results with the observed data received from the observation device hundred, and predicts the flow of people based on the estimation results.

[0018] The observation device 100 is a device for observing the flow of people within a predetermined area, and is, for example, a camera device. Also, although the prediction system 1 shown in FIG. 2 includes a plurality of observation devices 100, it may be one. The observation device 100 observes the flow of people, for example, every second and generates observed data. Also, the observation device 100 can acquire the position information of the flow of people by means of GPS or the like.

[0019] [Functional overview] FIG. 3 is a diagram showing an example of the prediction of the flow of people according to Example 1. As shown in FIG. 3, for example, the prediction device 10 simulates the actions of individual people based on various action intentions θ such as 500 patterns for people in a predetermined area such as inside a station or a passage. Since the set of the simulated actions of individual people becomes the flow of people, the prediction device 10 evaluates the simulated flow of people using the observed data of the flow of people in the predetermined area obtained by the observation device 100. Although details will be described later, the observed data is, for example, a density map of the flow of people at a predetermined time, and the prediction device 10 can evaluate the simulated flow of people by converting the simulated flow of people into a density map of the same format and comparing the two density maps.

[0020] The prediction device 10 then estimates the behavioral intention θ that formed the basis of the simulated pedestrian flow, which is close to actual observational data, from the evaluation results of the simulated pedestrian flow, and predicts future pedestrian flow based on the estimated behavioral intention θ. Such pedestrian flow prediction may involve estimating the behavioral intention θ that indicates what percentage of people want to go to the left in a left-right passage, and then further simulating and predicting pedestrian flow based on the estimated behavioral intention θ. The pedestrian flow simulation may be performed using existing technologies such as algorithms or simulation models for particle filters.

[0021] [Functional configuration of the prediction device 10] Next, the functional configuration of the prediction device 10, which is the main execution unit of this embodiment, will be described. Figure 4 is a diagram showing an example configuration of the prediction device 10 according to Embodiment 1. As shown in Figure 4, the prediction device 10 has a communication unit 20, a storage unit 30, and a control unit 40.

[0022] The communication unit 20 is, for example, a processing unit that controls communication with the observation device 100 via the network 50, and is a communication interface such as a network interface card.

[0023] The memory unit 30 is an example of a storage device that stores various data and programs executed by the control unit 40, such as memory or a hard disk. The memory unit 30 stores the simulation model 31, simulation result data 32, and observation data 33, etc.

[0024] The simulation model 31 is, for example, a model for simulating pedestrian flow based on people's behavioral intentions. In this embodiment, the pedestrian flow simulated from people's behavioral intentions using the simulation model 31 is evaluated with observational data 33, and the prediction accuracy is improved by further simulating pedestrian flow using the simulation model 31 based on the evaluation results. The simulation model 31 may be created for predetermined areas such as train station premises and passageways, or for predetermined scenes such as during events and evacuations in the event of a disaster.

[0025] The simulation result data 32 is, for example, data on pedestrian flow simulated based on human behavioral intentions using the simulation model 31. The simulation result data 32 may also include data on pedestrian flow further simulated based on evaluation results from observational data 33. Furthermore, the simulation result data 32 may be a predicted pedestrian flow density map at a given time.

[0026] The observation data 33 includes, for example, video or images taken by the observation device 100 of a predetermined area such as the station premises or passageways, or a density map of pedestrian flow at a predetermined time generated from said video or images. In particular, the density map may be generated by the prediction device 10 after receiving the underlying data from the observation device 100, or the prediction device 10 may receive a density map generated by the observation device 100.

[0027] The above-mentioned information stored in the memory unit 30 is merely an example, and the memory unit 30 can store a variety of other information besides the above.

[0028] The control unit 40 is a processing unit that oversees the entire prediction device 10, and is, for example, a processor. The control unit 40 comprises a simulation unit 41, an estimation unit 42, a prediction unit 43, and an output unit 44. Each processing unit is an example of an electronic circuit in the processor or an example of a process executed by the processor.

[0029] The simulation unit 41 simulates pedestrian flow based on multiple behavioral intentions of people. Furthermore, the simulation unit 41 simulates pedestrian flow based on behavioral intentions predicted by the prediction unit 43, which in turn simulates the simulated pedestrian flow. This is a so-called resimulation, where the prediction device 10 simulates pedestrian flow based on predetermined behavioral intentions the first time, and then simulates pedestrian flow based on behavioral intentions estimated by the estimation unit 42 from the second time onward. This allows for more plausible predictions and resimulations of pedestrian flow based on evaluation results with observational data, thereby improving the accuracy of pedestrian flow predictions.

[0030] The estimation unit 42 evaluates the simulated pedestrian flow based on observed pedestrian flow data. This involves, for example, calculating the likelihood of the simulated pedestrian flow based on the reciprocal of the error between the observed pedestrian flow data and the simulated pedestrian flow. The calculation of this error may be performed by generating a second density map from the positions of individual people in the simulated pedestrian flow based on the observed data, which is a first density map of people, and then calculating the likelihood based on the first and second density maps. Based on the evaluation results of the simulated pedestrian flow, the estimation unit 42 estimates the behavioral intentions that formed the basis of the simulated pedestrian flow, which is closer to the actual observed pedestrian flow data, and the current state of the pedestrian flow. The estimation of behavioral intentions can also be performed by assuming multiple behavioral intentions that explain the actually observed behaviors and estimating them. For example, in predicting evacuation behavior, the estimation unit 42 can estimate the degree to which people choose each exit from multiple exits as a behavioral intention for evacuation destination tendencies, and the degree to which people start moving at a certain time or after a predetermined time as a behavioral intention for evacuation initiation tendencies.

[0031] The prediction unit 43 simulates and predicts pedestrian flow based on the behavioral intentions and pedestrian flow conditions estimated by the estimation unit 42. The pedestrian flow simulation may be performed by the simulation unit 41. The prediction unit 43 also generates, for example, a density map for each elapsed time as a result of the pedestrian flow prediction.

[0032] The output unit 44 outputs the pedestrian flow predicted by the prediction unit 43 as the prediction result. The prediction result may be output to another information processing device that is communicatively connected to the prediction device 10, or to a display device that is communicatively connected to the prediction device 10, or simply to a log or image file.

[0033] [Function Details] Figures 1, 3, and 5-7 will be used to explain in more detail the process for predicting pedestrian flow, which improves the accuracy of pedestrian flow prediction. Here, the area containing exits A and B shown on the right side of Figure 1 will be used as an example to explain the predetermined area where the predicted pedestrian flow will occur.

[0034] First, as explained using Figure 3, the prediction device 10 uses existing algorithms and simulation models to simulate the behavior of individual people based on various randomly set behavioral intentions θ for people in a predetermined area. Furthermore, the prediction device 10 compares the simulation results, which represent the positions of individual people, with the observed data using a density map (people / m²) in the same format as the observed data. 2 It converts the data to a format. When the prediction device 10 receives observation data of the flow of people in a predetermined area from the observation device 100, it uses the observation data to evaluate the simulation results.

[0035] Figure 5 shows an example of pedestrian flow evaluation according to Example 1. Shown on the left side of Figure 5 is simulation result data 32, which shows the simulation result of pedestrian flow in a predetermined area 50 seconds after the start of pedestrian flow observation by the observation device 100. On the other hand, shown on the right side of the figure is observation data 33 of pedestrian flow 50 seconds after the start of pedestrian flow observation by the observation device 100.

[0036] As shown in Figure 5, the simulation result data 32 and the observation data 33 can be represented by density maps in the same format, and the prediction device 10 calculates the error between both density maps to evaluate the simulation result data 32. In other words, the smaller the error between the simulation result data 32 and the observation data 33, the more accurately the prediction device 10 can predict the actual positions of individual people and the pedestrian flow as a whole.

[0037] As an example of an evaluation value for the simulation result data 32, the prediction device 10 calculates the likelihood by taking the reciprocal of the total error amount of both density maps. As a result, the smaller the error, the higher the likelihood, and the prediction device 10 can evaluate that it was able to predict pedestrian flow more accurately the higher the likelihood.

[0038] The prediction device 10 then estimates the current state of pedestrian flow and the behavioral intention θ based on the likelihood of a set of multiple simulation result data 32 obtained by simulating based on various behavioral intentions θ. The estimated state of pedestrian flow and behavioral intention θ are used in the next simulation.

[0039] In this way, the prediction device 10 can improve the accuracy of predicting pedestrian flow by evaluating the results of the pedestrian flow prediction simulation, estimating the current pedestrian flow and behavioral intention θ from simulations that are close to actual pedestrian flow observation data, and repeating simulations based on the estimation results.

[0040] Figure 6 shows an example of the behavioral intention estimation results according to Example 1. Figure 6 is a graph showing the progression of the estimated behavioral intention θ when the correct result is behavioral intention θ = 0.7 (for example, 70% of people headed towards exit A). As shown in Figure 6, at the first simulation with an elapsed time of 0 seconds, the simulation is performed using a pre-set randomly determined behavioral intention θ, so the estimated value of behavioral intention θ is approximately 0.5, which is far from the correct value of 0.7. However, as time elapses, observational data 33 is acquired from the observation device 100, and the simulation result data 32 is evaluated using the observational data 33, and the estimated value of behavioral intention θ gradually approaches the correct value.

[0041] Next, we will show the changes using density maps. Figure 7 is a diagram showing an example of a group of density maps for Example 1. The density map groups shown in the upper part of Figure 7 are the observation data 33-1, 33-2, and 33-3 of human flow 50 seconds, 100 seconds, and 150 seconds after the start of human flow observation by the observation device 100, respectively. On the other hand, the density map groups shown in the lower part of Figure 7 are the simulation result data 32-1, 32-2, and 32-3, showing the simulation results of human flow 50 seconds, 100 seconds, and 150 seconds after the start of human flow observation by the observation device 100, respectively.

[0042] As shown in Figure 7, the prediction device 10 compares the simulation result data 32 with the observed data 33 at each elapsed time to evaluate the simulation result data 32. Then, it estimates the current state of pedestrian flow and behavioral intention θ from the simulation result data 32 and uses this for the simulation and subsequent prediction for the next elapsed time. In this way, the prediction device 10 can gradually bring the simulation result data 32 closer to the correct observed data 33 as time passes, thereby improving the accuracy of pedestrian flow prediction.

[0043] [Process flow] Next, the flow of the pedestrian flow prediction process by the prediction device 10 will be explained using the flowchart in Figure 8. Figure 8 is a flowchart showing an example of the prediction process flow according to Example 1.

[0044] First, the prediction device 10 generates an initial set of simulations by randomly setting simulation parameters such as the location information of people in a predetermined area, such as a station premises or passageway, where pedestrian flow is predicted, as well as their movement speed and behavioral intention θ (step S101). Note that the behavioral intention θ may be randomly set from various patterns, such as 500 patterns.

[0045] Next, the prediction device 10 performs a simulation of pedestrian flow prediction by inputting each of the simulation parameters set in step S101 into an algorithm or simulation model of an existing technology (step S102). This simulation may be performed for each behavioral intention, for example, on a regular or irregular basis. The predicted pedestrian flow may be for a predetermined elapsed time, such as 30 seconds or 50 seconds later.

[0046] Next, the prediction device 10 outputs the simulated pedestrian flow for each elapsed time as the simulation result (step S103).

[0047] Next, if the prediction device 10 has not received the elapsed time observation data simulated in step S102 from the observation device 100 (step S104: No), it returns to step S102 and predicts the flow of people for the next elapsed time.

[0048] On the other hand, if observational data for the simulated elapsed time is received (step S104: Yes), the prediction device 10 converts the simulated human flow output in step S103 into, for example, a density map corresponding to the received observational data (step S105). If the format of the density map of the observational data is known in advance, the prediction device 10 may, for example, convert the simulated human flow into a density map and output it when outputting the simulation results in step S103. In this case, naturally, the conversion to a density map in step S105 does not need to be performed again.

[0049] Next, the prediction device 10 calculates and evaluates the likelihood of the simulation results for the observed data (step S106). This process involves, for example, calculating the likelihood as the reciprocal of the error between the density map of the observed data and the density map of the simulated human flow, and then evaluating the simulation results based on the likelihood. Furthermore, this evaluation of the simulation results may be performed for each behavioral intention that formed the basis of the simulated human flow.

[0050] Next, the prediction device 10 reconstructs the simulation group based on the likelihood of the simulation results (step S107). This is a process in which, for example, simulation results are selected based on likelihood, the underlying behavioral intention θ is estimated, and then the next simulation parameters are set. More specifically, for example, if the prediction device 10 initially set the behavioral intention θ between 0.1 and 1.0 in step S101, and the behavioral intention θ with the highest likelihood was 0.6, then the next simulation parameter's behavioral intention θ is set between 0.5 and 0.7. Note that each value of behavioral intention θ described here is merely an example. In addition, other simulation parameters such as people's location information and movement speed may be reconstructed based on observational data.

[0051] Next, the prediction device 10 outputs simulation parameters such as the location information of people in the reconstructed simulation group, their movement speed, and their behavioral intention θ (step S108). Then, returning to step S102, the prediction device 10 inputs the simulation parameters output in step S108 into existing technology algorithms and simulation models to perform a re-simulation of pedestrian flow prediction. If no further simulations are to be performed, the prediction process shown in Figure 8 ends without returning to step S102.

[0052] [effect] As described above, the prediction device 10 simulates pedestrian flow based on the intentions of multiple human actions, evaluates the simulated pedestrian flow based on the pedestrian flow observation data 33, and predicts pedestrian flow based on the evaluation results of the simulated pedestrian flow.

[0053] As a result, the prediction device 10 can simulate pedestrian flow based on the intentions of multiple human actions, evaluate the simulation results with observational data to estimate the state of pedestrian flow and the intentions of actions, and predict pedestrian flow based on the evaluation results of the simulation, thereby improving the accuracy of pedestrian flow prediction.

[0054] Furthermore, the process performed by the prediction device 10 to evaluate the simulated pedestrian flow calculates the likelihood of the simulated pedestrian flow as an evaluation result, based on the error between the observed data 33 and the simulated pedestrian flow.

[0055] This allows the prediction device 10 to improve the accuracy of predicting pedestrian flow.

[0056] Furthermore, the prediction device 10 generates a second density map from the positions of individual people in the simulated human flow based on observational data 33, which is a first density map of people, and calculates an error based on the first density map and the second density map.

[0057] This allows the prediction device 10 to improve the accuracy of predicting pedestrian flow.

[0058] Furthermore, the prediction device 10 estimates the state of pedestrian flow and the intention of action based on the evaluation results, and the process performed by the prediction device 10 to simulate pedestrian flow based on the intention of action includes the process of simulating pedestrian flow based on the predicted intention of action.

[0059] This allows the prediction device 10 to further improve the accuracy of its pedestrian flow predictions.

[0060] [system] The processing procedures, control procedures, specific names, and various data and parameters shown in the above documents and drawings may be changed at will unless otherwise specified. Furthermore, the specific examples, distributions, and numerical values ​​described in the embodiments are merely examples and may be changed at will.

[0061] Furthermore, the components of each illustrated device are functionally conceptual and do not necessarily need to be physically configured as shown. In other words, the specific forms of distribution and integration of each device are not limited to those shown. That is, all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads and usage conditions. Moreover, all or any part of the processing functions performed by each device can be realized by a CPU (Central Processing Unit) or GPU (Graphics Processing Unit) and programs that are analyzed and executed by the CPU or GPU, or they can be realized as hardware using wired logic.

[0062] [Hardware] Figure 9 illustrates an example of a hardware configuration. As shown in Figure 9, the prediction device 10 includes a communication interface 10a, an HDD (Hard Disk Drive) 10b, memory 10c, and a processor 10d. Furthermore, the components shown in Figure 9 are interconnected by a bus or similar means.

[0063] The communication interface 10a is a network interface card or similar device that communicates with other servers. The HDD 10b stores the programs and databases that operate the functions shown in Figure 4.

[0064] The processor 10d is a hardware circuit that operates a process that performs the functions described in Figure 4 by reading a program that performs the same processing as each processing unit shown in Figure 4 from the HDD 10b or the like and loading it into memory 10c. In other words, this process performs the same functions as each processing unit of the prediction device 10. Specifically, the processor 10d reads a program that has the same functions as the simulation unit 41 and the estimation unit 42 from the HDD 10b or the like. Then, the processor 10d executes a process that performs the same processing as the simulation unit 41 and the estimation unit 42.

[0065] Thus, the prediction device 10 operates as an information processing device that performs operation control processing by reading and executing a program that performs the same processing as each processing unit shown in Figure 4. Furthermore, the prediction device 10 can also achieve the same functionality as the above-described embodiment by reading a program from a recording medium using a media reader and executing the read program. Note that the program referred to in this other embodiment is not limited to being executed by the prediction device 10. For example, this embodiment can be similarly applied when another computer or server executes a program, or when they collaborate to execute a program.

[0066] Furthermore, programs that perform the same processing as the processing units shown in Figure 4 can be distributed via networks such as the Internet. These programs can also be recorded on computer-readable recording media such as hard disks, flexible disks (FDs), CD-ROMs, MOs (Magneto-Optical disks), and DVDs (Digital Versatile Discs), and executed by being read from these media by a computer. [Examples]

[0067] Now, although embodiments of the present invention have been described, the present invention may be implemented in various other forms besides those described above.

[0068] With regard to embodiments including the above examples, the following additional information is disclosed.

[0069] (Note 1) Human flow is simulated based on the intentions of multiple human actions, Based on the observed data of the human flow, the simulated human flow is evaluated. Based on the results of the simulated evaluation of the human flow, predict the human flow. A prediction program characterized by having a computer perform the processing.

[0070] (Note 2) The process for evaluating the simulated pedestrian flow is as follows: Based on the error between the observed data and the simulated pedestrian flow, the likelihood of the simulated pedestrian flow is calculated as the evaluation result. The prediction program described in Appendix 1, characterized by including processing.

[0071] (Note 3) Based on the observation data which is the first density map of the people, a second density map is generated from the positions of the individual people in the simulated human flow. The error is calculated based on the first density map and the second density map. The prediction program described in Appendix 2, characterized in that it causes the computer to perform the processing.

[0072] (Note 4) Based on the evaluation results, the computer is made to perform a process to estimate the intention of the action. The process of simulating the flow of people based on the intent of the aforementioned actions is: The flow of people is simulated based on the predicted intentions of the actions. A prediction program according to any one of the appendices 1 to 3, characterized by including processing.

[0073] (Note 5) Based on the intentions of multiple people's actions, pedestrian flow is simulated. Based on the observed data of the human flow, the simulated human flow is evaluated. Based on the results of the simulated evaluation of the human flow, predict the human flow. A prediction method characterized by the processing being performed by a computer.

[0074] (Note 6) The process for evaluating the simulated pedestrian flow is as follows: Based on the error between the observed data and the simulated pedestrian flow, the likelihood of the simulated pedestrian flow is calculated as the evaluation result. The prediction method according to Appendix 5, characterized by including processing.

[0075] (Note 7) Based on the observation data which is the first density map of the person, a second density map is generated from the position of each person in the simulated human flow. The error is calculated based on the first density map and the second density map. The prediction method according to Appendix 6, characterized in that the processing is performed by the computer.

[0076] (Note 8) Based on the evaluation results, the computer performs a process to estimate the intention of the action. The process of simulating the flow of people based on the intent of the aforementioned actions is: The flow of people is simulated based on the predicted intentions of the actions. A prediction method according to any one of appendices 5 to 7, characterized by including processing.

[0077] (Note 9) Based on the intentions of multiple people's actions, pedestrian flow is simulated. Based on the observed data of the human flow, the simulated human flow is evaluated. Based on the results of the simulated evaluation of the human flow, predict the human flow. A prediction device characterized by having a control unit that performs processing.

[0078] (Note 10) The process for evaluating the simulated human flow is as follows: Based on the error between the observed data and the simulated pedestrian flow, the likelihood of the simulated pedestrian flow is calculated as the evaluation result. The prediction device according to Appendix 9, characterized by including processing.

[0079] (Note 11) Based on the observation data which is the first density map of the people, a second density map is generated from the positions of the individual people in the simulated human flow. The error is calculated based on the first density map and the second density map. The prediction device according to Appendix 10, characterized in that the control unit performs the processing.

[0080] (Note 12) Based on the evaluation results, the control unit performs a process to estimate the intention of the action. The process of simulating the flow of people based on the intent of the aforementioned actions is: The flow of people is simulated based on the predicted intentions of the actions. A prediction device according to any one of appendices 9 to 11, characterized by including processing.

[0081] (Note 13) An observation device that observes human flow and generates human flow observation data, Upon receiving the aforementioned observation data, The flow of people is simulated based on the intentions of multiple human actions. Based on the aforementioned observational data, the simulated human flow is evaluated, Based on the results of the simulated evaluation of the human flow, predict the human flow. A predictive device that performs processing and A prediction system characterized by having the following features.

[0082] (Note 14) The process for evaluating the simulated human flow is as follows: Based on the error between the observed data and the simulated pedestrian flow, the likelihood of the simulated pedestrian flow is calculated as the evaluation result. The prediction system according to Appendix 13, characterized by including processing.

[0083] (Note 15) Based on the observation data which is the first density map of the person, a second density map is generated from the position of each person in the simulated human flow. The error is calculated based on the first density map and the second density map. The prediction system according to Appendix 14, characterized in that the prediction device performs the processing.

[0084] (Note 16) Based on the evaluation results, the prediction device performs a process to estimate the intention of the action. The process of simulating the flow of people based on the intent of the aforementioned actions is: The flow of people is simulated based on the predicted intentions of the actions. A prediction system according to any one of appendices 13 to 15, characterized by including processing.

[0085] (Note 17) Processor and, Memory that is operablely connected to the processor and A prediction device equipped with a processor, By simulating pedestrian flow based on the intentions behind multiple human actions, Based on the observed data of the human flow, the simulated human flow is evaluated. Based on the results of the simulated evaluation of the human flow, predict the human flow. A prediction device characterized by performing processing. [Explanation of symbols]

[0086] 10 Prediction device 10a communication interface 10b HDD 10c memory 10d processor 20 Communications Department 30 Storage section 31 Simulation Models 32 Simulation result data 33 Observational data 40 Control Unit 41 Simulation Department 42 Estimation part 43 Prediction Section 44 Output section 50 Networks 100 observation devices

Claims

1. For each of the multiple intended actions of a person, the flow of people in a designated area is simulated. For each of the aforementioned multiple action intentions, the likelihood is calculated as the reciprocal of the error between the actual pedestrian flow observation data in the predetermined area and the simulated pedestrian flow data, and the action intention with the highest likelihood is selected. The simulation is repeated in the predetermined area to predict the flow of people based on the intent of the selected action. A prediction program characterized by having a computer perform the processing.

2. A second density map is generated using the positions of individual people in the simulated human flow, in the same format as the observation data, which is a first density map of people in a predetermined area where the human flow is simulated. The error is calculated based on the first density map and the second density map. The prediction program according to claim 1, characterized in that it causes the computer to perform the processing.

3. The process to be selected is, For each of the aforementioned intentions of the multiple actions, a second density map is generated using the position of each individual in the simulated human flow, in the same format as the observation data which is the first density map of the person in the predetermined area. The error is calculated based on the first density map and the second density map. The reciprocal of the aforementioned error is calculated as the likelihood, Select the intention of the action with the highest likelihood. The prediction program according to feature 1.

4. For each of the multiple intended actions of a person, the flow of people in a designated area is simulated. For each of the aforementioned multiple action intentions, the likelihood is calculated as the reciprocal of the error between the actual pedestrian flow observation data in the predetermined area and the simulated pedestrian flow data, and the action intention with the highest likelihood is selected. The simulation is repeated in the predetermined area to predict the flow of people based on the intent of the selected action. A prediction method characterized by the processing being performed by a computer.

5. For each of the multiple intended actions of a person, the flow of people in a designated area is simulated. For each of the aforementioned multiple action intentions, the likelihood is calculated as the reciprocal of the error between the actual pedestrian flow observation data in the predetermined area and the simulated pedestrian flow data, and the action intention with the highest likelihood is selected. The simulation is repeated in the predetermined area to predict the flow of people based on the intent of the selected action. A prediction device characterized by having a control unit that performs processing.

6. An observation device that observes the actual flow of people in a designated area and generates observational data of that flow, Upon receiving the aforementioned observation data, For each of the multiple intended actions of a person, the flow of people in the predetermined area is simulated. For each of the aforementioned multiple action intentions, the likelihood is calculated as the reciprocal of the error between the actual pedestrian flow observation data in the predetermined area and the simulated pedestrian flow data, and the action intention with the highest likelihood is selected. The simulation is repeated in the predetermined area to predict the flow of people based on the intent of the selected action. A predictive device that performs processing and A prediction system characterized by having the following features.

Images