JP2019102066A - Travel route estimation device, travel route estimation method, and program - Google Patents

Abstract

To efficiently estimate an action route of an agent so as to reduce an observation error even under such a situation that a time delay occurs.SOLUTION: A simulator execution unit 141 calculates, based on a simulation result, an objective function indicating an observation error between a calculated value for an agent regarding each of observation points in a predetermined number of time zones including an estimation object time zone and a time zone subsequent thereto and an observation value in an actual environment. A next input parameter determination unit 142 determines a next input parameter on the basis of an in input parameter and a calculation result of the objective function. Calculation and determination are repeated, so that an optimization control unit 143 obtains an input parameter optimizing the objective function.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a movement path estimation device, a movement path estimation method, and a program, and in particular, performs estimation of a movement path of an agent matching an observation based on the number of people observed at a certain point obtained for each time zone. The present invention relates to a movement path estimation device, a movement path estimation method, and a program.

  In order to cope with the congestion that may occur in a large-scale event etc., grasping of the flow of people using a multi agent simulator (MAS) and formulation of control measures are being carried out. Highly accurate simulations are required for more accurate understanding of human flow and development of effective control measures. For that purpose, the technique which determines the parameter which reproduces the observation data based on the observation data becomes indispensable.

Observational data refers to data that is measured and aggregated at a certain granularity in time and space. Specifically, as shown in FIG. 9, the observation time width is

When the time is

From to time

Time zone

Number of people observed at observation point k at

Is obtained.

In addition, as shown in FIG. 10, MAS is executed when an occurrence time, a departure place, a destination, and a route are specified to the agent, and the transit node and transit time of each agent are output. . The parameter is a vector in which the number of agents passing the route R _i is arranged in a certain time zone T _j as shown in FIG.

The whole time zone is connected I × J dimensional vector

It is.

Observations obtained by aggregating the output of MAS

And the actual observed values

The problem of minimizing the error with the above can be regarded as the problem of black box function optimization, and there is Bayesian optimization (Non-Patent Document 1) as a prior art of the optimization method.

  The optimization method is an approach of approximating an objective function and searching for a parameter giving a minimum value by setting in advance a parameter to be searched and an objective function and sequentially obtaining an objective function value for the parameter.

  For this problem, two naive settings can be considered depending on how to determine the input parameters and the objective function. The respective image diagrams are shown in FIG. 12 and FIG.

The first naive setting (setting 1) shown in FIG. 12 sets the input parameter to “the number of agents passing each movement route in the entire time zone.

I × J dimensional vector whose element is

The objective function is set as "all time zone, observation error at each observation point k". Here, the time zone

From time zone

Assuming that the error up to the point is expressed by the following equation (1), in setting 1, the parameter is solved by solving the optimization problem of the following equation (2)

Ask for

The second naive setting (hereinafter referred to as setting 2) shown in FIG. 13 is that the input parameter is “the number of agents passing each route R _i in one time zone T _j

I-dimensional vector whose element is

The objective function is set as “one time zone T _j , observation error at each observation point k”.

That is, in setting 2, the parameters are solved by solving the optimization problem of the following equation (3)

Ask for

J.Snoek, H. Larochelle, and R.P. Adams. Practical Bayesian optimization of machine learning algorithms. In Advances in Neural Information Processing Systems (NIPS), 2012. Nello Cristianini, John Shawa-Taylor, Takeshi Ohkita (Translation): Kernel pattern analysis (2010).

  In setting 1, Bayesian optimization is performed by using the number of agents passing all the time and all routes as input values. However, in this formulation, the dimensionality of the parameters becomes too large, and the approximation accuracy of the objective function deteriorates. Therefore, it becomes difficult to search for an efficient solution.

  On the other hand, in setting 2, since the number of input dimensions is the number of movement paths, the problem of the number of dimensions does not occur. This setting is natural when the agent arrives at the observation point within the generation time zone. The reason is that arriving in the same time zone means that agents generated at each time affect only the observation error of the same time zone. As such an example, movement between booths in a small scale venue can be considered.

  However, considering a large-scale event, it is not always possible to assume arrival within the same time zone as the occurrence time. For example, as shown in FIG. 14, consider a case where an agent generated at departure 2 at 8:02 goes to the destination 12. At this time, if the observation interval is 5 minutes, although agents are observed within the same time at observation point (3), it takes time from generation to observation at observation point (10), and it is observed in the next time zone It will As in this example, when the observation point is separated from the occurrence point, a gap (time delay) occurs between the occurrence time zone of the agent and the observation time zone. In other words, it is necessary to take into account that the observation error affected by the agent spans several time zones. In such a case, in the prior art, there is a problem that adaptation is difficult.

  The present invention has been made in view of the above points, and it is possible to efficiently estimate an agent's action route that reduces observation error even under a situation where time delay occurs. An object of the present invention is to provide an apparatus, a movement path estimation method, and a program.

  The moving path estimation apparatus according to the present invention is configured to calculate a plurality of moving paths in each time zone based on input parameters having the number of agents passing through the paths in the estimation target time zone for each of a plurality of paths in the real environment. The simulation of the movement of the agent is executed, and from the result of the simulation, the calculation regarding the agent for each of the observation points in a predetermined number of time zones including the estimation target time zone and the next time zone of the estimation target time zone A simulator execution unit that calculates an objective function representing an observation error between a value and an observation value in a real environment, a next input that determines a next input parameter based on the input parameter and a calculation result of the objective function Execution by the simulator execution unit until the parameter determination unit and the predetermined repetition condition are satisfied; It was repeated a determination by the force parameter determination unit configured to include a, and optimization controller for determining an input parameter to optimize the objective function.

  Further, in the movement path estimation method according to the present invention, the simulator execution unit is based on an input parameter whose element is the number of agents passing through the path in the estimation target time zone for each of a plurality of paths in the real environment. Simulation is executed in each time zone, and from the result of the simulation, the observation point is observed in a predetermined number of time zones including the estimation target time zone and the next time zone of the estimation target time zone. Calculating an objective function representing an observation error between the calculated value of each of the agents and the observed value in the real environment, the next input parameter determination unit calculating the input parameter, and the calculation result of the objective function Determining the next input parameter, and the optimization control unit satisfies the predetermined repetitive condition. And execution by the simulator execution unit causes repeated and decision by the next input parameter determination unit includes, a step of determining the input parameters to optimize the objective function.

  According to the movement path estimation device and the movement path estimation method according to the present invention, the simulator execution unit sets the number of agents passing through the path in the estimation target time zone for each of the plurality of paths in the real environment as an element Based on the input parameters, a simulation in which a plurality of agents move in each time zone is executed, and from the result of the simulation, a predetermined number of time zones including the estimated time zone and the next time zone of the estimated time zone And calculate an objective function representing an observation error between the calculated value of the agent for each of the observation points and the observation value in the real environment.

  Then, the next input parameter determination unit determines the next input parameter based on the input parameter and the calculation result of the objective function, and the simulator controls the simulator until the optimization control unit satisfies a predetermined repetition condition. The execution by the execution unit and the determination by the next input parameter determination unit are repeated to obtain an input parameter for optimizing the objective function.

  Thus, from the simulation results, in the predetermined number of time zones including the estimation target time zone and the next time zone of the estimation target time zone, the calculated values regarding the agent for each of the observation points and the observation values in the real environment Calculating the objective function representing the observation error with the above, determining the next input parameter based on the input parameter and the calculation result of the objective function, and finding the input parameter for optimizing the objective function Thus, even under a situation where time delay occurs, it is possible to efficiently estimate the agent's action path that reduces the observation error.

  The movement path estimation device according to the present invention further includes an iterative control unit that causes the optimization control unit to repeatedly obtain the input parameter, with each of all the time zones as the estimation target time zone, The iterative control unit can repeatedly execute repeatedly obtaining the input parameter by the optimization control unit as each of the time zones to be estimated, until the predetermined repetitive condition is satisfied. .

  In this way, it is repeatedly determined that the input parameters are repeatedly determined using each of all time zones as an estimation target time zone until each of the all time zones is an estimation target time zone until the predetermined repetition condition is satisfied. By performing the execution, even under a situation where a time delay occurs, it is possible to accurately estimate the agent's action path that reduces the observation error.

  Further, the simulator execution unit in the movement path estimation device according to the present invention may include, among all the time zones, a predetermined number of time zones before the estimation target time zone, the estimation target time zone, and the estimation target time zone. The simulation may be performed in which a plurality of agents move in a predetermined number of time zones including the next time zone of.

  Thus, among all the time zones, a plurality of agents in a predetermined number of time zones including the predetermined number of time zones before the estimation target time zone, the estimation target time zone, and the next time zone of the estimation target time zone By executing a simulation in which moves, even under a situation where a time delay occurs, it is possible to quickly estimate the agent's action path that reduces the observation error.

  A program according to the present invention is a program for functioning as each part of the above movement path estimation apparatus.

  According to the movement path estimation device, the movement path estimation method, and the program of the present invention, it is possible to efficiently estimate the action path of the agent which reduces the observation error even under the situation where time delay occurs. it can.

It is an image figure showing an outline of a 1st embodiment of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a block diagram which shows the structure of the movement path | route estimation apparatus based on the 1st Embodiment of this invention. It is a flowchart which shows the movement path | route estimation processing routine of the movement path | route estimation apparatus based on the 1st Embodiment of this invention. It is a flowchart which shows the optimization process routine of the movement path | route estimation apparatus based on the 1st Embodiment of this invention. It is an image figure which shows the outline | summary of the 2nd Embodiment of this invention. It is a block diagram which shows the structure of the movement path | route estimation apparatus based on the 2nd Embodiment of this invention. It is a flowchart which shows the movement path | route estimation processing routine of the movement path | route estimation apparatus based on the 2nd Embodiment of this invention. It is an image figure which shows the outline | summary of the 3rd Embodiment of this invention. It is an example of a road network and observation data. It is an example of input / output of MAS. It is an example of an input parameter. This is an example of a conventional naive input / output setting (setting 1). This is an example of a conventional naive input / output setting (setting 2). It is an image figure of time delay.

  Hereinafter, embodiments of the present invention will be described using the drawings.

<Overview of Moving Path Estimation Device According to First Embodiment of the Present Invention>
First, an outline of the first embodiment of the present invention will be described.

In this embodiment, in order to prevent an increase in the number of input dimensions, the number of agents passing each route R _i in one time zone T _j is the same as in the setting 2

I-dimensional vector whose element is

"

  However, the objective function is different from setting 2 and uses not only the observation error of the same time zone but also the measurement error of the next time zone which is affected by the time delay. FIG. 1 is an image diagram showing an outline of the present embodiment.

  That is, in the present embodiment, the number of agents passing each route in each time zone is determined by solving the optimization problem of the following equation (4).

  Here, D is the number of time zones that it takes at most after an agent is generated and observed. Since D represents how much the agent is delayed, it can be appropriately changed according to the situation.

Objective function

And by, it is possible to consider to the observation error of the time period that the agent which occurs in the time zone T _j affect the time delay, it becomes possible to estimate the movement path of the agent with high accuracy.

<Configuration of Moving Path Estimation Device According to First Embodiment of the Present Invention>
The configuration of the moving path estimation apparatus according to the embodiment of the present invention will be described with reference to FIG. FIG. 2 is a block diagram showing the configuration of a moving path estimation apparatus according to an embodiment of the present invention.

  The movement path estimation apparatus 100 is configured by a computer including a CPU, a RAM, and a ROM storing a program for executing a movement path estimation processing routine described later, and is functionally configured as follows: ing.

  As shown in FIG. 2, the movement path estimation apparatus 100 according to the present embodiment includes a simulator setting processing unit 110, an optimization processing unit 120, a data storage unit 130, an optimization execution unit 140, and an optimum parameter processing unit. 150 and is connected to an external device 200 such as an input device or a display device.

  The simulator setting processing unit 110 sets, in the simulator setting table 131, fields necessary for executing the simulator, which are input from the external device 200.

  The optimization processing unit 120 sets the maximum number of optimization execution times S, the number of time divisions J, and the time delay constant D input from the external device 200 in the optimization setting table 132.

  The maximum number of times of optimization execution S is the maximum number of times of optimization for one time zone, the number of time divisions J is the number of divisions of the time zone in which the data was observed, and the time delay constant D is the agent Is the maximum number of time periods it takes to be observed from the occurrence time zone.

  The data storage unit 130 includes a simulator setting table 131, an optimization setting table 132, and an optimum parameter table 133.

  The simulator setting table 131 has fields necessary for executing a human flow simulator. The required fields are, for example, the road network, the walking speed of the agent, and the like.

  The optimization setting table 132 has a maximum optimization execution number field, a maximum repetition number field, a time zone division number field, and a time delay constant field.

  Maximum optimization execution number S is set in the maximum optimization execution number field, time division number J is set in the time zone division number field, and time delay constant D is set in the time delay constant field .

  In the optimum parameter table 133, the optimum parameters input from the optimization execution unit 140 are set.

The optimization execution unit 140 reads the settings of the simulator setting table 131 and the optimization setting table 132 to obtain an objective function.

Parameters to optimize

Ask for

  Specifically, the optimization execution unit 140 includes a simulator execution unit 141, a next input parameter determination unit 142, an optimization control unit 143, an optimization result storage unit 144, and a time zone optimization result storage unit 145. Is equipped.

The simulator execution unit 141 uses, as an element, the number of agents passing through the route R _i in the estimation target time zone T _j for each of the plurality of routes R _i in the real environment.

Based on the above, a simulation in which a plurality of agents move in each time zone T _j is executed, and from the simulation results, a predetermined number of times including the estimation target time zone T _j and the next time zone T _{j + D} of the estimation target time zone Calculated values for the agent for each of the observation points in the

And observed values in the real environment

Objective function representing the observation error with

Calculate

Specifically, the simulator execution unit 141 inputs an input parameter in the time zone T _j

Run a human flow simulator (MAS) based on the objective function value

Calculate

Is added to the optimization result storage unit 144.

Here, the human flow simulator itself simulates all time zones, and the simulator execution unit 141 determines the objective function value from the execution result of the human flow simulator.

Calculate

The next input parameter determination unit 142 determines the input parameter

And the objective function

The following input parameters based on the calculation results of

Decide.

For example, in the case of using Bayesian optimization as the optimization method, the following input parameters are determined by using a probability model such as a Gaussian process. Specifically, the next input parameter determination unit 142 uses the all data in the optimization result storage unit 144 to generate parameters.

And the objective function value

For example, estimation with a Gaussian process is performed using

When it is expressed as a certain parameter

Objective function value in

Is the average value represented by the following formula (5)

And the dispersion value represented by the following equation (6)

Follow a Gaussian distribution with.

  However, superscript symbol T represents transposition of a matrix, and superscript symbol "-1" represents inverse matrix. This kernel function may change depending on the problem. Typical kernels include linear kernels and Gaussian kernels (see Non-Patent Document 2).

here,

Where k and K are parameters called kernel functions

When

Functions that define the degree of similarity with

The following equation (7) and equation (8) can be written using

Then, the next input parameter determination unit 142 sets the next parameter to be input to the human flow simulator according to the following equation (9):

Decide.

here,

Is called the acquisition function, and the parameters

Is an index to quantitatively evaluate the possibility of giving the minimum value. As the acquisition function, probability improvement (PI) or expectation value improvement (EI) may be used.

The optimization control unit 143 repeats the execution by the simulator execution unit 141 and the determination by the next input parameter determination unit 142 until the predetermined repetition condition is satisfied, with the time zone T _j as the estimation target time zone, and the objective function

Parameters to optimize

Ask for

  Specifically, first, the optimization control unit 143 initializes the optimization result storage unit 144 to an empty set.

Next, the optimization control unit 143 repeats the processing of the simulator execution unit 141 and the next input parameter determination unit 142 until the maximum optimization execution number S is exceeded in the time zone T _j .

The optimization control unit 143 causes the simulator execution unit 141 to obtain the parameters obtained by the next input parameter determination unit 142.

Is input as an input parameter to run a human flow simulator.

Then, the optimization control unit 143 minimizes the error among the input parameters stored in the optimization result storage unit 144 after the repetition.

Time zone optimization result storage unit 145,

Store Here, Data means the optimization result storage unit 144.

Then, the optimization control unit 143 initializes the optimization result storage unit 144 to an empty set, and performs the same processing with the next time zone T _{j + 1} as the estimation target time.

When the optimization control unit 143 performs the above process for all the time zones (T _{1 to} T _J ), the optimization control unit 143 puts together the data included in the time zone optimization result storage unit 145 to obtain optimum parameters.

Are calculated and stored in the optimum parameter table 133.

The optimization result storage unit 144 receives input parameters in the time zone T _j obtained by the simulator execution unit 141.

And objective function values

It is a pair with

Remember.

Further, the optimization result storage unit 144 stores it when the optimization control unit 143 acquires an instruction to be initialized.

Delete all and make it an empty set.

The time zone optimization result storage unit 145 obtains the optimum parameters in the time zone T _j obtained by the optimization control unit 143.

I remember.

The optimum parameter processing unit 150 acquires the optimum parameters acquired from the optimum parameter table 133.

Are output to the external device 200. The output process may be executed, for example, when an output request is input from the external device 200.

<Operation of Moving Path Estimation Device According to First Embodiment of the Present Invention>
FIG. 3 is a flowchart showing a movement path estimation processing routine according to the first embodiment of the present invention.

  When each setting is input to the simulator setting processing unit 110 and the optimization processing unit 120, the movement route estimation processing routine shown in FIG. 3 is executed in the movement route estimation apparatus 300.

  First, in step S100, the simulator setting processor 110 receives the field necessary for executing the simulator, which is input from the external device 200, and the optimization processor 120 performs the maximum number of optimization executions input from the external device 200. S, the time division number J, and the time delay constant D are acquired respectively.

  In step S110, the simulator setting processing unit 110 sets, in the simulator setting table 131, fields necessary for executing the simulator acquired in step S100.

  In step S120, the optimization processing unit 120 sets the maximum optimization execution number S, the number of time divisions J, and the time delay constant D acquired in step S100 in the optimization setting table 132.

  In step S130, the optimization control unit 143 initializes the optimization result storage unit 144 (Data) to an empty set.

In step S140, j = 1. j is a counter for counting the estimation target time zone T _j .

  In step S150, the optimization execution unit 140 executes optimization processing.

In step S160, the optimization control unit 143 minimizes the error among the input parameters stored in the optimization result storage unit 144.

Time zone optimization result storage unit 145,

Store

  In step S170, j = j + 1.

  In step S180, the optimization control unit 143 determines whether j is larger than the time division number J.

  If j is not larger than J (NO in step S180), in step S190, the optimization control unit 143 initializes the optimization result storage unit 144 to an empty set, and repeats the processing in steps S150 to S170 again.

On the other hand, when j is larger than J (YES in step S180), in step S200, the optimization control unit 143 puts together the data included in the time zone optimization result storage unit 145 and generates the optimum parameters.

Are stored in the optimum parameter table 133.

In step S210, the optimal parameter processing unit 150 determines the optimal parameter acquired from the optimal parameter table 133.

Are output to the external device 200.

  Here, the optimization process in step S150 will be described. FIG. 4 is a flowchart showing the optimization processing routine.

  In step S300, s = 1. s is a counter for counting the number of times of optimization.

In step S310, the simulator execution unit 141 uses, as an element, the number of agents passing the route R _i in the estimation target time zone T _j for each of the plurality of routes R _i in the real environment.

Based on the simulation in which a plurality of agents move in each time zone T _j .

In step S320, the simulator execution unit 141 determines each of the observation points in a predetermined number of time zones including the estimation target time zone T _j and the next time zone T _{j + D} of the estimation target time zone from the simulation result in step S310. Calculated value for agent about

And observed values in the real environment

Objective function representing the observation error with

Calculate

In step S330, the simulator execution unit 141 calculates in step S320.

Is added to the optimization result storage unit 144.

In step S340, the next input parameter determination unit 142 uses the all data in the optimization result storage unit 144 to generate parameters.

And the objective function value

Estimate the relationship with.

In step S350, the next-input-parameter determining unit 142 sets the next parameter to be input to the human flow simulator according to equation (9).

Decide.

  In step S360, s = s + 1.

  In step S370, the optimization control unit 143 determines whether s is greater than the maximum optimization execution number S or not.

  If s is not larger than the maximum optimization execution number S (NO in step S370), the process returns to step S310, and the processes in steps S310 to S360 are repeated.

  If it is larger than the maximum optimization execution number S (YES in step S370), the process returns.

  As described above, according to the present embodiment, from the result of simulation, in the predetermined number of time zones including the estimation target time zone and the next time zone of the estimation target time zone, the calculation regarding the agent for each of the observation points Calculate the objective function that represents the observation error of the value and the observation value in the real environment, and repeatedly determine the next input parameter based on the input parameter and the calculation result of the objective function. By obtaining input parameters that optimize the function, it is possible to efficiently estimate the agent's action path that reduces the observation error, even in a situation where a time delay occurs.

<Overview of Moving Path Estimation Device According to Second Embodiment of the Present Invention>
Next, an outline of the second embodiment of the present invention will be described.

FIG. 5 is an image diagram showing an outline of the present embodiment. In the present embodiment, according to the first embodiment, the number of agents passing through each route in all the time zones is finally determined by sequentially optimizing the time zones T ₁ , T ₂ , T ₃ . . At this time, optimization of the entire time zone is performed once, and then optimization is performed a plurality of times such as time zones T ₁ , T ₂ , T ₃ , and so on.

Since the input parameters of time zone T _j are affected by agents generated in the previous and subsequent D time zones, optimization is performed again after the previous and subsequent time zones are optimized even in the same time zone. It is possible to increase the accuracy of agent movement path estimation.

<Configuration of Moving Path Estimation Device According to Second Embodiment of the Present Invention>
The configuration of the movement path estimation apparatus according to the embodiment of the present invention will be described with reference to FIG. FIG. 6 is a block diagram showing a configuration of the movement path estimation apparatus 300 according to the embodiment of the present invention. In addition, about the structure similar to the movement path | route estimation apparatus 100 based on 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  The optimization processing unit 320 sets the maximum number of times of optimization execution S, the maximum number of repetitions R, the number of time divisions J, and the time delay constant D input from the external device 200 in the optimization setting table 332.

  The maximum number of repetitions R is the maximum number of units when the maximum number of optimization executions S is one unit.

  The optimization setting table 332 has a maximum optimization execution number field, a maximum repetition number field, a time zone division number field, and a time delay constant field. The maximum number of repetitions R is set in the maximum number of repetitions field.

  The optimization execution unit 340 includes a simulator execution unit 141, a next input parameter determination unit 142, an optimization control unit 143, an optimization result storage unit 144, a time zone optimization result storage unit 145, and an iteration control unit 346. Is equipped.

The iterative control unit 346 causes the optimization control unit 143 to repeatedly obtain input parameters, with each of all time zones (T _{1 to} T _J ) as estimation target time zones.

Specifically, when the optimization control unit 143 completes the processing for all time zones (T _{1 to} T _J ), the iterative control unit 346 executes the processing by the simulator execution unit 141 and the next input parameter determination unit 142. to greater than the maximum number of repetitions R is the number of times repeated and determination by to perform the process again from the time zone T ₁ with respect to the optimization control unit 143.

At this time, the optimization control unit 143 does not initialize the optimization result storage unit 144 to an empty set, and the optimization result storage unit 144

And its objective function value

Store

In addition, when the number of repetitions exceeds the maximum number of repetitions R, the repetition control unit 346 causes the optimization control unit 143 to combine the data included in the time zone optimization result storage unit 145 into an optimum parameter.

Are calculated and stored in the optimum parameter table 133.

<Operation of Moving Path Estimation Device According to Second Embodiment of Present Invention>
FIG. 7 is a flowchart showing a movement path estimation processing routine according to the embodiment of the present invention. In addition, about the process similar to the movement path | route estimation process routine which concerns on 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step S400, the simulator setting processing unit 110 receives the field necessary for executing the simulator, which is input from the external device 200, the optimization processing unit 320, the maximum optimization execution number S input from the external device 200, The time division number J, the maximum number of repetitions R, and the time delay constant D are acquired respectively.

  In step S420, the optimization processing unit 320 sets the maximum optimization execution number S, the number of time divisions J, the maximum number of repetitions R, and the time delay constant D acquired in step S400 in the optimization setting table 332.

  In step S435, r is set to 1. r is a counter for counting the number of times the optimization process is repeated.

  In step S170, j = j + 1. When j + 1> J, j = 1.

  In step S475, r = r + 1.

  In step S480, the iteration control unit 346 determines whether r is greater than the maximum number of iterations R.

  If r is not larger than the maximum number of repetitions R (NO in step S480), the optimization control unit 143 determines whether r is larger than the number of time divisions J in step S490.

  If r is not larger than J (NO in step S490), in step S500, the optimization result storage unit 144 is initialized to an empty set, and the processing in steps S150 to S480 is repeated again.

On the other hand, when r is larger than J (YES in step S 490), the optimization control unit 143 stores the time zone optimization result storage unit 145 in step S 510.

And its objective function value

Are stored in the optimization result storage unit 144, and the processing of steps S150 to S480 is repeated again.

  On the other hand, when j is larger than J (YES in step S 490),

  On the other hand, if r is larger than the maximum number of repetitions R (YES in step S480), the process proceeds to step S200.

  As described above, according to the present embodiment, it is determined that each of all the time zones is repeatedly assumed to be an estimation target time zone until the predetermined repetition condition is satisfied. By repeatedly executing the process as the estimation target time zone, it is possible to accurately estimate the agent's action path which reduces the observation error even under a situation where a time delay occurs.

<Movement Path Estimation Device According to Third Embodiment of the Present Invention>
Next, a third embodiment of the present invention will be described.

FIG. 8 is an image diagram showing an outline of the present embodiment. In the present embodiment, in the first embodiment or the second embodiment, when time zone T _j is optimized, agents in all time zones can be generated by generating only agents in D time zones before and after. The execution time of the MAS can be shortened compared to the case of making the action of H. As a result, the optimization can be speeded up.

  In the present embodiment, the same components as those of the movement path estimation apparatus according to the first embodiment or the second embodiment are denoted by the same reference numerals, and detailed description will be omitted.

The simulator execution unit 141 determines a predetermined number of time zones (T _{j -D} ) before the estimation target time zone among all the time zones (T _{1 to} T _J ), an estimation target time zone (T _j ), and estimation. A simulation is performed in which a plurality of agents move in a predetermined number of time zones (T _{j + D 2} ) including a time zone next to the target time zone.

Specifically, in the case of executing a human flow simulator, the simulator execution unit 141 uses parameters on an actual simulator.

Act only the agents generated by.

  As described above, according to the present embodiment, among all the time zones, the predetermined number of time zones before the estimation target time zone, the estimation target time zone, and the next time zone of the estimation target time zone are included. By executing a simulation in which a plurality of agents move in a predetermined number of time zones, it is possible to quickly estimate an agent's action path which reduces observation error even under a situation where time delay occurs. .

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made without departing from the scope of the present invention.

  Although the above-mentioned embodiment demonstrated the case where Bayesian optimization was used as an example of the optimization method, it is not limited to Bayesian optimization, You may apply another optimization method and algorithm. For example, a genetic algorithm can be applied as an optimization method.

In the present embodiment, the number of passing people measured in a certain time zone T _j at a plurality of observation sites k

The number of people who passed route R _i in each time zone from

It is possible to flexibly adapt to various observation data (the number of people passing, density of observation points per unit time, average traveling time to the destination, etc.). Thus, it is possible to cope with data obtained from various devices.

  Also, in the present specification, although the program is described as an embodiment installed in advance, the operation of each component of the movement path estimation device is constructed as a program, and installed in a computer used as the movement path estimation device It can be implemented or distributed via a network. Further, the program can be provided by being stored in a computer readable recording medium.

100 travel path estimation device 110 simulator setting processing unit 120 optimization processing unit 130 data storage unit 131 simulator setting table 132 optimization setting table 133 optimum parameter table 140 optimization execution unit 141 simulator execution unit 142 next input parameter determination unit 143 optimization Control unit 144 Optimization result storage unit 145 Time zone optimization result storage unit 150 Optimal parameter processing unit 200 External device 300 Movement path estimation device 320 Optimization processing unit 332 Optimization setting table 340 Optimization execution unit 346 Repeating control unit

Claims (7)

Performing a simulation in which a plurality of agents move in each time zone based on input parameters whose elements are the number of agents passing through the paths in the estimated time zone for each of a plurality of paths in the real environment; From the result of the simulation, in the predetermined number of time zones including the estimation target time zone and the next time zone of the estimation target time zone, the calculated value regarding the agent for each of the observation points, and the observation value in the real environment A simulator execution unit that calculates an objective function representing an observation error of
A next input parameter determination unit that determines a next input parameter based on the input parameter and the calculation result of the objective function;
An optimization control unit for obtaining an input parameter for optimizing the objective function, causing the execution by the simulator execution unit and the determination by the next input parameter determination unit to be repeated until a predetermined repetition condition is satisfied;
A movement path estimation device including And an iterative control unit that causes the optimization control unit to repeatedly obtain the input parameter as each of the all time zones as the estimated time zone.
The iterative control unit causes the optimization control unit to repeatedly and repeatedly determine the input parameter as the estimation target time zone, until the predetermined repetitive condition is satisfied. The movement path estimation device according to 1. The simulator execution unit is a predetermined number of time periods including a predetermined number of time zones before the estimation target time zone, the estimation target time zone, and a time zone next to the estimation target time zone among all time zones. The movement path estimation apparatus according to claim 1, wherein the simulation in which a plurality of agents move in a band is executed. The simulator execution unit moves a plurality of agents in each time zone based on input parameters whose elements are the number of agents passing through the paths in the estimation target time zone for each of a plurality of routes in the real environment The simulation is executed, and from the result of the simulation, in the predetermined number of time zones including the estimation target time zone and the next time zone of the estimation target time zone, the calculated values regarding the agent for each of the observation points and the real environment Computing an objective function representing the observation error with the observation at
Determining a next input parameter based on the input parameter and the calculation result of the objective function;
Obtaining an input parameter for optimizing the objective function by causing the optimization control unit to repeat the execution by the simulator execution unit and the determination by the next input parameter determination unit until a predetermined repetition condition is satisfied;
Travel path estimation method including: The iterative control unit further includes the step of causing the optimization control unit to repeatedly obtain the input parameter by setting each of all the time zones as the estimation target time zone,
The step of repeatedly causing the optimization control unit to repeatedly obtain the input parameter as the estimation target time period until the predetermined repetition condition is satisfied. The movement path estimation method of 4. The step of executing includes a predetermined number of time periods including a predetermined number of time zones before the estimation target time zone, the estimation target time zone, and a next time zone of the estimation target time zone among all time zones. The movement path estimation method according to claim 4 or 5, wherein the simulation in which a plurality of agents move in a band is executed.   The program for functioning a computer as each part of the movement path | route estimation apparatus in any one of Claims 1 thru | or 3.