Sensing system, sensing server, sensing method, and sensing program

Abstract

To obtain a sensing system, a sensing server, a sensing method, and a sensing program that can reduce adjustment work required to determine an estimated value and improve accuracy of the estimated value.SOLUTION: A sensing system according to an embodiment comprises: at least one sensor that acquires data on the number of target individuals; an acquisition unit that acquires sensor data from the at least one sensor; a generation unit that generates an additional constraint based on the sensor data; a derivation unit that derives an estimation model for estimating a value of the number of the target individuals using at least the additional constraint; and an estimation unit that performs optimization processing using the estimation model and the sensor data.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The embodiments of the present invention relate to a sensing system, a sensing server, a sensing method, and a sensing program. [Background technology]

[0002] In recent years, the importance of sensing has increased for purposes such as advanced control and comfort. However, there are situations where it is difficult to directly measure the data required by applications, or where it is necessary to introduce expensive sensor systems. To address these situations, a technology known as virtual sensing integrates already obtained data and uses software to estimate the required data from the integrated data. However, there have been cases where it is difficult to obtain highly accurate estimates, such as when sensor data is insufficient.

[0003] For example, one prior art proposed is a technology that efficiently estimates the data required by an application by utilizing multiple different types of sensor data, with smart buildings as one application. Here, it has been shown that even large-scale problems can be efficiently solved by using equations to describe the relationships between sensor data and then solving the optimization problem using this model description as a constraint. It has also been shown that it can automatically handle situations where data contradict each other due to sensor errors (hereafter referred to as "over-defined problems"), situations where values cannot be determined without appropriate assumptions due to a lack of sensor data (hereafter referred to as "under-defined problems"), and even situations where these are mixed (hereafter referred to as "complexly defined problems").

[0004] However, in under-defined problems, it can be difficult to make appropriate assumptions to determine values, and it is necessary to switch assumptions depending on the situation, making it difficult to put into practical use. At the same time, there is a need for methods to improve the accuracy of the estimates obtained. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2023-137830 [Non-patent literature]

[0006] [Non-Patent Document 1] K. Kondo et al., Equation-based modeling and optimization-based parameter estimation in multimodal virtual sensing platforms for smart buildings, Build. Environ. 243, 110620, 2023 DOI: 10.1016 / j.buildenv.2023.110620 Summary of the Invention [Problem to be solved by the invention]

[0007] When sensor data is insufficient, it is necessary to make appropriate assumptions to determine values, which poses problems such as the need for advanced knowledge and long-term adjustment work, as well as the difficulty of ensuring accuracy.

[0008] One example of the problem that the present invention aims to solve is to provide a sensing system, a sensing server, a sensing method, and a sensing program that can reduce the adjustment work required to determine estimated values and improve the accuracy of estimated values. [Means for solving the problem]

[0009] A sensing system according to an embodiment includes at least one sensor that acquires data regarding the number of target individuals, an acquisition unit that acquires sensor data from the at least one sensor, a generation unit that generates additional constraints based on the sensor data, a derivation unit that uses at least the additional constraints to derive an estimation model that estimates a value regarding the number of target individuals, and an estimation unit that performs optimization processing using the estimation model and the sensor data.

[0010] The sensing server according to the embodiment includes an acquisition unit that acquires sensor data from at least one sensor that acquires data regarding the number of target individuals, a generation unit that generates additional constraints based on the sensor data, a derivation unit that uses at least the additional constraints to derive an estimation model that estimates a value regarding the number of target individuals, and an estimation unit that performs optimization processing using the estimation model and the sensor data.

[0011] A sensing method according to an embodiment includes an acquisition step of acquiring sensor data from at least one sensor that acquires data regarding the number of target individuals, a generation step of generating additional constraints based on the sensor data, a derivation step of deriving an estimation model that estimates a value regarding the number of target individuals using at least the additional constraints, and an estimation step of performing optimization processing using the estimation model and the sensor data.

[0012] A sensing program according to an embodiment is a sensing program that estimates required parameters from sensor data from at least one sensor that acquires data regarding the number of target individuals, and includes an acquisition step that acquires sensor data, a generation step that generates additional constraints based on the sensor data, a derivation step that uses at least the additional constraints to derive an estimation model that estimates a value regarding the number of target individuals, and an estimation step that performs optimization processing using the estimation model and the sensor data. [Brief explanation of the drawings]

[0013] [Figure 1]FIG. 1 is a diagram showing an example of a schematic configuration of a sensing system according to this embodiment. [Figure 2] FIG. 2 is a diagram showing an example of an OD table related to the elevator traffic demand estimation problem that is the subject of this embodiment. [Figure 3] Figure 3 shows the relationship between operation and OD tables for the time interval for which traffic demand is to be estimated. [Figure 4] FIG. 4 is a diagram illustrating the relationship between the sensor data and the OD table. [Figure 5] FIG. 5 shows the results of expressing the relationship between sensor data and the OD table as a mathematical model. [Figure 6] FIG. 6 is a diagram for explaining a method for calculating an estimated value from the OD table in the estimation unit. [Figure 7] FIG. 7 shows an example in which the characteristics of individual operations can be attributed to some elements of the OD table being zero. [Figure 8-1] Figure 8-1 shows the characteristics of each operation in an OD table. [Figure 8-2] Figure 8-2 shows the procedure by which extracted operational characteristics are converted into additional constraints. [Figure 9] FIG. 9 shows an example of the results obtained when additional constraints are added. [Figure 10] FIG. 10 shows the results of estimation using only the base model. [Figure 11] FIG. 11 is a flowchart showing the processing procedure in the central server. [Figure 12] FIG. 12 is a flowchart showing a processing procedure for generating additional constraints based on feature extraction of sensor data. [Figure 13] FIG. 13 is a diagram showing a schematic configuration of a sensing system according to the second embodiment. [Figure 14] FIG. 14 is a diagram showing operation segments on adjacent floors, operation segments that pass through the first floor, and operation segments that pass through the second floor. [Figure 15]FIG. 15 is a diagram showing the operation segment from passing through the 3rd floor to passing through the 6th floor. [Figure 16] FIG. 16 is a diagram showing an example of the inclusion relationship of operation segments. [Figure 17] FIG. 17 is a diagram showing an example of a mechanism for successively recording information on operation segments that are passed through on the way to necessary data. [Figure 18] FIG. 18 is a diagram showing the state of the base model in the third embodiment. [Figure 19] FIG. 19 is a diagram showing the operation of separating the movement of people into those in high-priority operation segments and those in other segments. [Figure 20] FIG. 20 is a diagram showing the operation of separating the movement of people into those in high-priority operation segments and those in other segments. [Figure 21] FIG. 21 is a diagram showing the operation of separating the movement of people into those in high-priority operation segments and those in other segments. [Figure 22] FIG. 22 is a diagram showing an equation generated as an additional constraint in the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0014] A sensing system, sensing server, sensing method, and sensing program that add constraints by referencing data will be described in detail below with reference to the drawings. In the following embodiments, parts with the same numbers perform similar operations, and redundant description will be omitted. For example, when there are multiple identical or similar elements, a common symbol may be used to describe each element without distinguishing between them, or a subnumber may be used in addition to the common symbol to describe each element with distinction.

[0015] [Embodiment] (composition) Fig. 1 is a diagram showing an example of the schematic configuration of a sensing system according to this embodiment. As shown in Fig. 1, the virtual sensing system includes a plurality of sensors 102 installed in on-site equipment and facilities 101, and a central server (information processing device) 104.

[0016] The sensor 102, together with the GW (gateway) 103, functions to measure and transmit data. This function can be realized by one or more computers. The computer here also includes an embedded system that includes hardware such as a microcomputer chip, an LSI (Large Scale Integration), an ASIC (Application Specific Integrated Circuit), or an FPGA (Field-Programmable Gate Array). A plurality of sensors 102 are provided, and they acquire data on the number of target individuals.

[0017] There are various possible embodiments for the on-site equipment and facilities 101, sensors 102, and GW 103, but here, taking into account the relevance of the explanation for the central server 104, the on-site equipment and facilities 101 will be explained as an elevator installed in a building as an example.

[0018] Elevators are equipped with load sensors that measure the load on the car that carries passengers and detect situations such as full capacity, door sensors that detect passengers getting on and off to ensure safety when getting on and off, and cameras to monitor the situation inside the car. Data from these devices is sent to the elevator control panel, converted into the required data format, and transferred via GW 103 to a central server 104 such as a remote monitoring center.

[0019] The central server 104 can be realized by one or more computers and operates as a virtual sensing server. The central server 104 can communicate with the on-site devices and equipment 101 (here, elevators) via the GW 105.

[0020] The central server 104 stores data obtained from the on-site devices and facilities 101 in an acquisition unit (data temporary storage unit) 106 that temporarily stores the data. A generation unit (additional constraint generation unit) 107 generates constraint conditions that are suited to the characteristics of the data according to a processing procedure described below. Next, a storage unit (base model storage unit) 108 stores a basic base model that has been created in advance. Furthermore, a derivation unit (estimation model storage unit) 109 integrates the constraint conditions stored in the generation unit 107 with the base model in the storage unit 108 and stores the integrated model in order to derive an estimation model by performing estimation using the data. Furthermore, the system includes an estimation unit (optimization processing data estimation unit) 110 and an output unit (output interface) 111.

[0021] The acquisition unit 106, memory unit 108, and derivation unit 109 perform the necessary processing and are equipped with a memory function, and as the memory function, for example, a non-volatile memory that can be written to and read at any time, such as an EPROM (Erasable Programmable Read Only Memory), HDD (Hard Disk Drive), or SSD (Solid State Drive), can be used in combination with a non-volatile memory such as a ROM (Read Only Memory).

[0022] The GW 105 has a communication function and includes one or more wired or wireless communication modules. For example, it includes a communication module that connects to the on-site device / facility 101 by wire via a hub or the like. It may also include a communication module that connects to the on-site device / facility 101 by wireless. In other words, the GW 105 only needs to be equipped with a general communication interface that can communicate between the on-site device / facility 101 and the sensor 102 and send and receive various information.

[0023] Furthermore, the estimation unit 110 uses a model such as the constraint conditions stored in the derivation unit 109 to process the data stored in the acquisition unit 106 according to a procedure described below, and outputs an estimated value of the data. This output result is transferred via the output unit 111 to an application that uses this output result. Note that this application may be located in the central server 104, or may be located on a separate server and configured to transfer the result via a communication line.

[0024] Next, a specific format of the model stored in the storage unit 108 will be described in detail. Before describing the model, the elevator traffic demand estimation problem set as the problem to be solved will be described. It is known that if elevator traffic demand can be accurately estimated, an elevator group control system can use that data to perform appropriate control, which can lead to shorter waiting times and reduced energy consumption. Therefore, it is very useful and important to estimate traffic demand as accurately as possible while minimizing the installation of costly sensors, etc. Note that for details on how a group control system utilizes traffic demand information to perform control that leads to shorter waiting times and reduced energy consumption, the technology described below can be used, for example.

[0025] Toshiba Elevator and Building Systems Corporation, Group control systems, [Online], Internet <URL:https: / / www.toshiba-elevator.co.jp / elv / infoeng / products / gcs / >

[0026] Y. Sakamaki et al.,Design of elevator-group control system to save energy consumption by dynamically controlling the number of running cars,IEICE TRANSACTIONS on Fundamentals of Electronics,Communications and Computer Sciences,E98-A(2) pp.612-617, Feb. 2015, DOI: 10.1587 / transfun.E98.A.612

[0027] Next, Fig. 2 shows an example of an OD table related to the elevator traffic demand estimation problem targeted in this embodiment. That is, it shows an OD table (Origin and Destination Table) used to express traffic demand. The vertical axis 20 of the table indicates the boarding floor, and the horizontal axis 21 indicates the alighting floor. The corresponding boxes 22 on the vertical axis 20 and horizontal axis 21 indicate the number of people alighting. In this embodiment, the building is taken as an example to be eight stories high.

[0028] Here, we will explain the numbers shown in Figure 2. As shown in Figure 2, for example, assume that the elevator is currently waiting on the first floor with zero passengers, and that a passenger has just boarded. Here, we will refer to the period from the waiting state until all passengers have disembarked as a trip. First, six passengers boarded on the first floor and departed. Next, the elevator stopped on the second floor, and no one disembarked, but three new passengers boarded. Next, the elevator stopped on the third floor, and three of the passengers who boarded on the first floor disembarked on the third floor. Also, one passenger boarded on the third floor. The elevator passed through the fourth floor, and three passengers who boarded on the second floor disembarked on the fifth floor. Next, the elevator stopped on the sixth floor, and one passenger who boarded on the third floor disembarked there. The elevator passed through the seventh floor, and three passengers who boarded on the first floor disembarked on the eighth floor, bringing the number of passengers on board to zero, completing one trip.

[0029] For this operation, there was a demand for three passengers from the first floor to the third floor, three from the second floor to the fifth floor, one from the third floor to the sixth floor, and three from the first floor to the eighth floor. This is shown in Figure 2 on the OD table. In the OD table, the diagonal elements are the same for boarding and disembarking floors, and are always 0. Upward operations are in the upper right part of the OD table, and downward operations are in the lower left part. Since upward and downward operations are independent of the definition of operations, only upward operations will be considered below.

[0030] The traffic demand estimation problem is a problem of estimating values shown in an OD table from sensor data of a sensor 102 provided in an elevator.

[0031] The sensor data used for estimation may be, for example, the number of passengers during operation. The number of passengers can be estimated by dividing the value of the load sensor by the average weight. If the in-car camera is equipped with an image recognition function, the number of passengers can be counted by image recognition. Furthermore, values relating to the number of passengers at the boarding location, the number of passengers disembarking at the stopping location, and the number of passengers at the passing locations can be used as examples.

[0032] Specifically, for each trip, trip segments are defined based on information about the floors the train stops at. In the example shown in Figure 2, the trip is broken down into five trip segments: from floor 1 to floor 2, from floor 2 to floor 3, from floor 3 to floor 5 (floor 4 is passed through), from floor 5 to floor 6, and from floor 6 to floor 8 (floor 7 is passed through). The numbers of passengers associated with each trip segment are 6, 9, 7, 4, and 3, respectively. Assume that the number of passengers for each trip segment is obtained as sensor data.

[0033] The smallest unit of an operation segment is operation between adjacent floors. For example, in the example in Figure 2, an operation segment from floors 3 to 5 that passes through floor 4 can be decomposed into two, from floors 3 to 4 and from floors 4 to 5, in which case the number of passengers is 7 for the two operation segments. In other words, the operation segment information from floors 3 to 5 that passes through floor 4 can be converted into data indicating that the floor passed is 4 and that the number of passengers for the two operation segments from floors 3 to 4 and from floors 4 to 5 is 7, by processing the information. Below, we consider the passed floor data and the decomposed number of passengers for the operation segments between adjacent floors as sensor data.

[0034] Figure 3 shows the relationship between operations and OD tables for the time interval for which traffic demand is to be estimated. Here, it is assumed that there were 10 upward operations during the time for which traffic demand is to be estimated. Since an OD table can be defined for each operation as shown in Figure 2, traffic demand for the desired measurement time can be calculated by adding up the OD tables for the 10 operations.

[0035] Figure 3(a) shows the operation segments for each trip. Specifically, it shows operation segment information from the first trip (Trip 1) to the tenth trip. Figure 3(b) shows each trip in an OD table. Figure 3(c) shows 10 upward trips all in one OD table. Specifically, Trip 1 in Figure 3(a) is shown as Trip 1 in Figure 3(b). Similarly, Trip 2 in Figure 3(a) is shown as Trip 2 in Figure 3(b). Trip 3 in Figure 3(a) is shown as Trip 3 in Figure 3(b). Figure 3(c) shows the OD table for the time interval to be estimated, obtained by combining the 10 OD tables.

[0036] FIG. 4 is a diagram illustrating the association between sensor data and the OD table. 1,0304 is the number of passengers in the segment from the 3rd floor to the 4th floor in the first trip. Similarly, m a,0203 is the number of passengers on the second to third floor trip segment at the desired measurement time.

[0037] For example, in the example in Figure 3, there are 10 operations, so m a,0203 =Σ 10 i=0 m i,0203 It can be calculated as shown in Figure 4(a). 1,0304 is the total number of passengers who board on floors 1, 2, or 3 and disembark on floors 4, 5, 6, 7, or 8 in Trip 1, and is the sum of the areas indicated by the dotted pattern. Similarly, m a,0203 is the total number of passengers boarding on the first or second floor and disembarking on the third, fourth, fifth, sixth, seventh or eighth floor at the desired measurement time, and is the sum of the shaded areas.

[0038] FIG. 5 shows the results of expressing the relationship between sensor data and the OD table as a mathematical model. Specifically, FIG. 5(a) is a table in which variables are assigned to the parts of the OD table related to upward travel. FIG. 5(b) expresses the variables of the OD table and the sensor data of the number of passengers for each travel segment as a set of linear expressions. In other words, it expresses the values obtained from the sensor data and the sum of the variables of the OD table for each travel segment, respectively. These expressions are models stored in the storage unit 108. For example, the equation expresses that the sensor data corresponds to the sum of the variables in a certain range of the OD table.

[0039] The constraints generated by the generation unit 107 are basically in the same equation format. In the embodiment of Fig. 1, the derivation unit 109 integrates the constraint conditions stored in the generation unit 107 and the storage unit 108 and stores the result as simultaneous equations. In the embodiment of Fig. 13, the derivation unit 109 stores the constraints generated by the generation unit 107.

[0040] We want to find the variables related to the number of passengers shown in Figure 5(a). a,0102 , m a,0203 m a,0708 is an example of sensor data, and the right-hand side is the sum of variables related to the number of passengers for each sensor data.

[0041] (operation) (1) Optimization processing data estimation operation Next, a method for estimating the OD table in the estimation unit 110 by referring to the model held in the derivation unit 109 and the sensor data stored in the acquisition unit 106 will be described.

[0042] The estimation process is solved as a mathematical optimization problem with a given model as a constraint. Specifically, it is formulated as a minimization problem of the objective function shown in Figure 6. For example, m a,0203 Since there is an error between the actual value and the direct sensor data, the direct sensor data is converted into x a,0203 Then, by introducing the error ε2, m a,0203 =x a,0203 +ε2. Direct sensor data is expressed as x a,0203 It is expressed as an equation such as =36.0.

[0043] The first term of the objective function indicates that the value with the smallest error is set as the estimated value. When there are many different types of sensors 102 and discrepancies occur between the measured values due to errors, it is effective to set the value with the smallest error as the estimated value (the reason it is called over-defined).

[0044] However, in the base model shown in Figure 5, the number of equations is insufficient compared to the number of variables, and the error ε i Even if you set all of these to zero, the values in the OD table will still not be uniquely determined (which is why it is called under-defined).

[0045] Therefore, in order to determine the values of variables whose values are not uniquely determined, the second term of the objective function shown in Figure 6 is introduced. In other words, the value is determined so that the sum of squares of the values is minimized. Here, λ is a coefficient, which is set so that the influence of the second term on the first term is sufficiently small. As a result, for over-defined variables, the values are determined by minimizing the error, and for under-defined variables, the values are determined by minimizing the sum of squares of the values. For details of this process, for example, the technology described below can be used.

[0046] K.Kondo et al., Equation-based modeling and optimization-based parameter estimation in multimodal virtual sensing platforms for smart buildings, Build. Environ.243,110620,2023 DOI: 10.1016 / j.buildenv.2023.110620

[0047] Note that the example in FIG. 6 is a process using only the base model, and the generation unit 107 generates x a,0203 Only expressions such as =36.0 are generated.

[0048] (2) Additional constraint generation operation There is room for improvement in the accuracy of the estimation results when processing using only the base model, as shown in Figure 6. Therefore, in order to improve accuracy, we consider generating additional constraints by referring to the content of the data (i.e., at the time of measurement execution) (hence, we call this an on-the-fly mathematical formulation).

[0049] In the estimation using only the base model, the demand was formulated as an OD table showing the total traffic demand at the desired measurement time, but this approach makes it difficult to capture the characteristics of individual operations.

[0050] Therefore, in this embodiment, we consider adding a constraint equation for each operation. When solving the problem in FIG. 6, we obtain x a,0203 It would be possible to send measurement values accumulated over such a measurement period, but in the following, it is assumed that individual measurement values are sent for each operation.

[0051] Specifically, returning to Figure 3, the OD table showing the overall traffic demand at the desired measurement time shown in Figure 3(c) has non-zero values for most elements. On the other hand, in the OD table defined for individual operations as shown in Figure 3(b), many elements are zero (blank cells are zero), resulting in a sparse matrix overall. This is because it is rare for a single operation to have passengers with all possible combinations of boarding and disembarking floors, and which elements have non-zero values can be considered to be characteristics of individual operations.

[0052] Figure 7 shows an example where the characteristics of individual trips can be reduced to some elements of the OD table being zero. Here, we use the fourth trip in Figure 3(a) as an example. In this trip, there were five passengers traveling from the second to the fourth floor, three from the third to the seventh floor, and two from the fourth to the eighth floor.

[0053] Figure 7 explains the fourth operation in Figure 3(a) as an example. As shown in Figure 7(a), this operation starts from the second floor, so there are no passengers boarding on the first floor. Therefore, it can be seen that all elements with the first floor as the boarding floor, indicated by the black diagonal lines, are zero. Next, as shown in Figure 7(b), the fifth floor is a passing floor, so there are zero passengers boarding on the fifth floor and zero passengers disembarking on the fifth floor. Therefore, it can be seen that all elements in the black diagonal line area are zero. Similarly, as shown in Figure 7(c), the sixth floor is a passing floor, so the elements in the black diagonal line area are also zero.

[0054] Furthermore, it is assumed that the presence or absence of passengers getting on or off at the stopping floors can be detected by door sensors that detect passengers getting on and off to ensure safety when getting on and off, or by cameras that monitor the inside of the car, etc. In other words, it is assumed that at the stopping floors, it is possible to detect as sensor data whether passengers are only getting on, only getting off, or both getting on and off.

[0055] As shown in Figure 7(d), for example, the bus is stopped on the third floor, but only passengers are boarding, and the elements of the black diagonal lined area for passengers who have disembarked, i.e., the third floor as the disembarking floor, are zero. As shown in Figures 7(a) to 7(d), elements whose values are confirmed to be zero are characteristics of this operation, and can be converted into constraint equations and added.

[0056] That is, as shown in FIG. 7, the generation unit 107 divides the data set under predetermined conditions, extracts the characteristics of each divided operation, and generates an equation.

[0057] Figure 8-1 shows the characteristics of individual operations in an OD table, where the elements of the overall OD table that are zero are represented by the black diagonal lines.

[0058] The zero elements shown in the black shaded area, e.g., w 4,1 To express the fact that is zero as a constraint, w 4,1 It is conceivable to add a new equation, w = 0. By defining such an additional equation for the elements that are zero, as shown in the black diagonal line area, the characteristics of the operation can be expressed as an equation. 4,1 =0 are integrated as simultaneous equations in the derivation unit 109 and stored. In this case, the overall configuration is as shown in FIG.

[0059] On the other hand, it is also possible to eliminate elements with a value of zero from the base model formula prepared in advance and transform it into a simpler formula for processing. Generally, it is known that the fewer the number of variables and formulas in the later stages of processing, the shorter the processing time, so there are cases where eliminating elements is advantageous. Figure 8-2 is a diagram showing the procedure by which extracted operational characteristics are converted into additional constraints. In other words, it shows the procedure by which elements whose values are confirmed to be zero, extracted as operational characteristics, are converted into additional constraints. In this case, the overall configuration will be as shown in Figure 13, which will be described later.

[0060] First, because we decided to extract features for each operation, we add the relationship between the OD table for each operation and the overall OD table. For example, as shown in Figure 8-1, in the overall OD table, elements that have been determined to be zero based on Figures 7(a) to 7(d) are represented by the black diagonal lines.

[0061] The number of passengers at the desired measurement time, m, is shown in Figure 5. a,0102 and the number of passengers on each trip, m 1,0102 The relationship between variables such as m a,0102 =m 1,0102 +m 2,0102 +···+m 10,0102 In Figure 8-2, a basic equation is generated for the fourth operation as an example. That is, from the base equation shown in Figure 5(b), the fourth operation is generated as shown in Figure 8-2(a). That is, Figure 8-2(a) is a diagram showing the base equation of the base model for the fourth operation.

[0062] Next, Figure 8-2(b) shows the formula transformed by eliminating the elements that are confirmed to be zero, shown by the black diagonal lines in Figure 8-1, from the base formula in Figure 8-2(a). Furthermore, as shown in Figure 8-2(b), m 4,0405 and m 4,0506 and m 4,0607 is originally m due to the relationship between the passing floors 4,0407 The values are inherited from and are the same. Therefore, the three expressions related to these three variables are identical. Therefore, Figure 8-2(c) shows the expression generated by eliminating the duplicated identical expressions shown in Figure 8-2(b). This is the characteristic of an element that is zero. In this way, an additional constraint is generated.

[0063] The estimation model shown in Fig. 8-2(c) is expressed in the form of an equation. The estimation unit 110 executes the process of minimizing or maximizing the objective function shown in Fig. 6 using the estimation model as a constraint.

[0064] Next, as another feature, we consider constraint equations regarding the number of boardings or alightings, focusing on changes in the number of passengers between operation segments.

[0065] As shown in Figure 7, the segment from the 3rd floor to the 4th floor has 8 passengers, and the segment from the 4th floor to the 5th floor has 5 passengers. In other words, the number of passengers decreases by 3 when the train stops at the 4th floor. This is the total number of passengers getting off at the 4th floor, i.e., the total number of passengers getting off at the 4th floor, w 4,3 +w 4,9 +w 4,14 This indicates that there are at least three people. 4,3 is known to be zero, so w 4,9 +w 4,14 A constraint equation of ≧3 can be generated.

[0066] As a result, the optimization process executed by the estimation unit 110 can also handle constraints based on inequalities, so generating and adding constraint equations for such constraints is also effective in making the estimated values more accurate.

[0067] If the number of people getting off can be measured specifically using a camera monitoring the inside of the car, that value can be used. However, in the case of image recognition, people may overlap in the image, and multiple people may be counted as one person. For this reason, it is better to express the constraints using inequalities in the same way.

[0068] For example, if the image sensor measures five passengers getting off at the fourth floor stop in Figure 7, then w 4,9 +w 4,14 ≥ 5, which is a stronger constraint and is expected to produce more accurate estimates.

[0069] When the above operation is performed, elements whose values are zero are deleted from the base model stored in the storage unit 108 to generate additional constraints, and furthermore, constraint equations expressed as inequalities are additionally generated. In this case, since there is no need for integration with the base model stored in the storage unit 108 in the derivation unit 109, the schematic configuration becomes as shown in Fig. 13, which will be described later.

[0070] Next, Figure 9 shows an example of the results obtained when additional constraints are added. That is, it shows the results of actual calculations of the example problem in Figure 3 in this embodiment. The results shown here are the results when additional constraints are added, and are the results using both additional constraints based on elements whose values are confirmed to be zero, and inequality constraints when the number of people disembarking can be specifically measured using an in-car camera or the like.

[0071] Figure 9(a) shows the actual measurements in the overall OD table. That is, it is the overall OD table, showing the correct answer to the estimation problem. Figure 9(b) shows the estimated values when additional constraints are added to the base model equation and sensor data is used. Figure 9(c) shows the difference between the actual measurements in Figure 9(a) and the estimated values in Figure 9(b).

[0072] On the other hand, Figure 10 shows the results of estimation using only the base model for the same problem as Figure 9. In other words, this is the result without the additional constraint that the value of the additional constraint is zero and without the inequality constraint.

[0073] Figure 10(a) shows the actual measurements in the overall OD table. That is, it is the overall OD table, showing the correct answer to the estimation problem. Figure 10(b) shows the estimated values when using sensor data in the base model equation. Figure 10(c) shows the difference between the actual measurements in Figure 10(a) and the estimated values in Figure 10(b).

[0074] Figures 9(c) and 10(c) show that the use of additional constraints reduces the overall error, and the maximum error value for each element in the OD table is kept small. Specifically, in the case of Figure 9(c), where additional constraints are used, the total error is 26.8, while in the case of Figure 10(c), where additional constraints are not used, the total error is 63.5.

[0075] Fig. 11 is a flowchart showing the processing procedure in the central server 104. As shown in Fig. 11, the acquisition unit 106 of the central server 104 acquires sensor data from sensors 102 installed in on-site devices and facilities 101. The sensor data is direct data obtained from the sensors 102, or data that has been processed by data integration, conversion, or the like. For example, it is data related to weight, or data converted from weight to number of people (step S1).

[0076] Next, the generation unit 107 generates an equation related to the additional constraint based on the acquired sensor data. The additional constraint may be, for example, a feature that determines that the number of passengers at a floor is zero due to the passage of each operation, a feature that determines that the number of passengers getting on and off at a floor is zero due to the absence of at least one of boarding and alighting, or a feature that adds an inequality (step S2).

[0077] Next, the derivation unit 109 derives an estimation model that estimates a value related to the number of target individuals using the additional constraints generated by the generation unit 107 and the base model stored in the memory unit 108 of the central server 4. Alternatively, the generation unit 107 may generate the additional constraints through a transformation operation that deletes variables from the base model stored in the memory unit 108. This deletes zero elements and redundant expressions from the expressions of the base model (step S3).

[0078] Therefore, by deriving an estimation model after removing redundant estimates using additional constraints, it is possible to reduce the effort required for adjusting the estimates. Furthermore, the accuracy of the estimates can be improved by performing optimization processing using the estimation model and sensor data. In other words, an estimation model using additional constraints can improve the accuracy of estimating the number of target individuals without increasing the number of sensors.

[0079] Next, the estimation unit 110 performs an optimization process using the derived estimation model and the acquired sensor data. By performing the optimization process, a value related to the number of target individuals is calculated. The optimization process involves minimizing or maximizing an objective function using the estimation model as a constraint. The optimization process may also include, for example, a process of dividing a value related to the load in the sensor data by the average body weight. This may be used to estimate the number of passengers (step S4).

[0080] Next, the estimation unit 110 outputs to the OD table the estimated values for the variables of the OD table estimated from the estimation model by the optimization process (step S5).

[0081] Furthermore, the processes from step S1 to step S5 are repeated for each operation within a predetermined range of the overall OD table (step S6).

[0082] Then, the output unit 111 outputs the outputted overall OD table. The output destination may be a control panel provided in an elevator or the like, or may be the central server 104 or another server (step S7).

[0083] This allows for accurate calculation of estimated values by using additional constraints when estimating values related to the number of individuals in the OD table from the base model.

[0084] Next, Fig. 12 is a flowchart showing a processing procedure for generating additional constraints based on feature extraction of sensor data. In particular, the procedure is shown here in accordance with a method for eliminating zero elements from the base model. As shown in Fig. 12, the generation unit 107 acquires basic equations from the storage unit 108, for example. A base model may be generated by applying (combining) additional constraints to the basic equations (step S21).

[0085] Next, the generation unit 107 determines whether or not there is an element that is zero from the sensor data acquired from the acquisition unit 106 (step S22). If it is determined that there is an element that is zero (step S22: Yes), the process proceeds to step S23. On the other hand, if it is determined that there is no element that is zero (step S22: No), the process proceeds to step S24.

[0086] In step S22, if it is determined that there is an element that is zero, the generating unit 107 deletes the element that is zero from the basic formula (step S23), and then proceeds to step S24.

[0087] Similarly, the generation unit 107 determines whether or not there are any overlapping expressions from the sensor data (step S24). If it is determined that there are any overlapping expressions (step S24: Yes), the process proceeds to step S25. On the other hand, if it is determined that there are no overlapping expressions (step S24: No), the process proceeds to step S26.

[0088] In step S24, if it is determined that there is a duplicate formula, the generating unit 107 deletes the duplicate formula from the basic formula (step S25), and then proceeds to step S26.

[0089] Next, the generation unit 107 adds an inequality for the value relating to the number of passengers getting on and off (step S26).

[0090] An equation for additional constraints is generated by the processing from step S21 to step S26 (step S27).

[0091] In steps S21 to S27, the generation unit 107 acquires the basic equations from the storage unit 108 and generates the equations obtained by modifying the basic equations as equations of the base model. As another example, the generation unit 107 may generate equations of additional constraints, and then the derivation unit 9 may derive an estimated model using the equations of the additional constraints and the equations of the base model acquired from the storage unit 108.

[0092] According to this embodiment, in the central server 104, additional constraints are generated based on sensor data obtained from sensors 102 installed in elevators, which are on-site devices and facilities 101, an estimation model is derived using the additional constraints and a base model, and optimization processing is performed using the estimation model and sensor data. This makes it possible to obtain estimation values with higher accuracy than when an estimation value is calculated using only the base model.

[0093] [Second embodiment] 13, the generation unit 107 may refer to a base model stored in the storage unit 108 and modify the model to generate additional conditions. In this case, the derivation unit 109 may be configured not to perform integration processing with the base model in the storage unit 108. In this case, the configuration shown in FIG. 13 is a second embodiment obtained by modifying the schematic configuration shown in FIG. 1.

[0094] Fig. 13 is a diagram showing a schematic configuration of a sensing system according to the second embodiment. As shown in Fig. 1, the derivation unit 109 derives an estimation model using the equations of additional constraints generated by the generation unit 107 and the equations of the base model stored in the storage unit 108. Alternatively, as shown in Fig. 13, the generation unit 107 may acquire basic equations from the storage unit 108 and use the basic equations transformed into equations of additional constraints. Then, the derivation unit 109 derives an estimation model from the equations of additional constraints.

[0095] [Third embodiment] In the present embodiment and the second embodiment, it is assumed that the measurement values of the sensor data are sent separately for each operation. However, it is also assumed that there are situations where it is only possible to send the measurement values integrated over the measurement time due to the specifications of the on-site equipment. Therefore, in the third embodiment, a method of using only the integrated measurement values will be described.

[0096] (composition) The hardware configuration of the on-site devices and facilities 101 and the central server 104 and the software configuration associated with the hardware configuration are the same as those in the embodiment, and therefore a duplicated description will be omitted.

[0097] (operation) Here, we focus on the fact that elevator operation conditions vary greatly depending on the time of day, and that by prioritizing the use of data corresponding to the operation conditions at each time of day, we can improve the accuracy of data estimation.

[0098] For example, in an office building, during rush hour, the majority of traffic travels from the entrance floor to the main residential floors of each employee. It is known that there are very few traffic services that serve floors where other departments are located. Furthermore, during lunchtime, traffic flow to the cafeteria floor increases significantly. In such situations, traffic often passes through all floors except for specific departure and disembarking floors. On the other hand, during normal working hours, there are many traffic services between floors that are closely related to the occupying departments.

[0099] Taking these points into consideration, we consider accumulating passenger numbers for operation segments that include intermediate floors, in addition to the basic operation segments on adjacent floors, and utilizing this data.

[0100] Figure 14 shows an example of an eight-story building. Figure 14(a) shows the operation segments of adjacent floors that serve as basic data. Figure 14(b) shows the operation segment of a one-floor skip, passing through one floor. Figure 14(c) shows the operation segment of a two-floor skip, passing through two floors.

[0101] Figure 15(a) shows a segment of travel that involves passing three floors with a three-floor skip. Figure 15(b) shows a segment of travel that involves passing four floors with a four-floor skip. Figure 15(c) shows a segment of travel that involves passing five floors with a five-floor skip. Figure 15(d) shows a segment of travel that involves passing six floors with a six-floor skip.

[0102] Next, for example, m a,0106 represents the passenger total value within a specified time associated with the trip segment from floor 1 to floor 6.4,0405 and m 4,0506 and m 4,0607 is originally m due to the relationship between the passing floors 4,0407 We explained that "values are inherited from the data", but it is also possible to inherit passenger data and accumulate data in the same way.

[0103] FIG. 16 shows an example of the inclusion relationship of operation segments. a,0106 The number of passengers associated with a given service segment is a,0103 It is also the number of passengers in the operating segments such as a,0103 The data for the basic operation segment m a,0102 That is, in the elevator, which is a field device, the data is accumulated as the operation occurs, and the value is inherited by m a,0106 If x passengers board or disembark at m a,0106 The following m including a,0103 ~m a,0206 x number of people is also added to the basic operation segment of the adjacent floor, which is even lower.

[0104] FIG. 17 shows an example of a mechanism for recording information on the operation segments that are passed through in the middle of a train. a,0106 If x names occur in m a,0106 The following m including a,0103 ~m a,0206 This shows an example of how to execute the process of adding x names to the operation record. Figure 17(a) and Figure 17(b) show the parent-child relationship of the operation record in a graph. a,0106 When higher-level data like this arrives, the graph structure can be traced and the data can be registered recursively. a,0105 From Fig. 17(c) m a,0104 and m a,0205 Also, m in Figure 17(b) a,0206 From Fig. 17(c) m a,0306 and m a,0205 The branching from FIG. 17(c) to FIG. 17(d) also branches in the same way.

[0105] In this way, the sensor data collected by the elevators of the on-site equipment is sent via a communication line to the central server 104. In this case, the data stored in the central server 104 does not include data for each individual operation, but passenger data for the operation segments of the floors that the passengers pass through is available.

[0106] The relationship between the passenger count data for each operation segment and the elements of the OD table is the same as in the first embodiment shown in Figure 4. In Figures 14 and 15, the sum of the elements enclosed in the OD table corresponds to the passenger count data.

[0107] Here, we define an index for determining the priority of operation segments. If the priority is α, the total number of passengers in each operation segment is p, and the number of corresponding squares in the OD table is n. The priority is expressed by the following formula.

[0108] α=p / (n+1)

[0109] A large number of passengers per square indicates that the corresponding operation segment accounts for a large proportion of the total. Here, to generate additional constraints, we consider estimating the OD table by dividing the operations with this large index from the remaining operations, and then adding them up at the end.

[0110] Fig. 18 is a diagram showing the state of the base model in the third embodiment. Fig. 19, Fig. 20, and Fig. 21 are diagrams showing the sequence of operations from Fig. 18, in which the movement of people in high-priority operation segments is separated from the movement of people in other segments.

[0111] First, in Figure 19, m a,0204 Since this has a higher priority, the number of people traveling on the segment of the trip that passes through the third floor from the second floor to the fourth floor is separated from those traveling on other trips.

[0112] For example, for people who travel from the first floor to the fourth floor, there are those who used the operation segment that passes through the third floor on the way from the second floor to the fourth floor, and those who used the operation that stops at every floor. Therefore, the data for people who traveled from the first floor to the fourth floor is shown in Figure 19(b). a,0204 190 and basic data m a,0204 It will be separated into two parts: 191 minus 191.

[0113] In Figure 20, the divided m a,0204 The further division of 200 is shown in Figure 20(c). a,0205 Figure 20(d) shows the division of 201. Similarly, the operation is carried out sequentially, and Figure 21 shows the final division state.

[0114] Fig. 22 shows an equation generated as an additional constraint in the third embodiment. That is, it shows the result of outputting the constraint corresponding to the final division state obtained by dividing the object up to the state shown in Fig. 21 as an equation.

[0115] According to the third embodiment, even in a situation where it is only possible to send measurement values accumulated over the measurement time due to the specifications of the on-site equipment and facilities 101, it is possible to generate additional constraints in accordance with the situation of the operation that has occurred, using data on operation segments including passing floors and the corresponding number of passengers.

[0116] In this third embodiment, it is possible to improve accuracy compared to when only the base model is used, but because individual operational data cannot be used in detail, the degree of improvement in accuracy is lower than in this embodiment.

[0117] [Variations] It should be noted that the present invention is not limited to the above-described embodiment. For example, it is possible to define characteristics of operation segments other than those shown in the first and second embodiments, convert them into constraint equations, and use them as additional constraints.

[0118] Furthermore, while detailed examples have been explained using elevators in buildings as an example, it is possible to solve the problem of estimating the movement of people between bus stops from data on boarding and alighting from a route bus and data on the number of passengers inside the bus (for example, it is conceivable to estimate the load from the pressure of the bus's air springs, and then count the number of people from that).

[0119] As described above, the sensing system of this embodiment comprises at least one sensor that acquires data regarding the number of target individuals, an acquisition unit that acquires sensor data from the at least one sensor, a generation unit that generates additional constraints based on the sensor data, a derivation unit that derives an estimation model that estimates a value regarding the number of target individuals using at least the additional constraints, and an estimation unit that performs optimization processing using the estimation model and the sensor data.

[0120] This allows, for example, the labor required for adjusting the estimated values to be reduced by deriving an estimation model after removing redundant estimated values using additional constraints. Furthermore, the accuracy of the estimated values can be improved by performing optimization processing using the estimation model and sensor data.

[0121] In this embodiment, the estimation model is generated using the additional constraints and the base model.

[0122] This allows, for example, an estimation model to be generated based on a base model, thereby reducing the effort required for adjustment work to determine estimated values and improving the accuracy of the estimated values.

[0123] In this embodiment, the estimation model is expressed in the form of an equation, and the estimation unit executes the process of minimizing or maximizing the objective function using the estimation model as a constraint.

[0124] This allows the estimation unit to, for example, perform processing to minimize or maximize the objective function, thereby reducing errors in the optimization processing.

[0125] In addition, in this embodiment, the sensors are installed in transportation equipment such as elevators and buses, and the sensor data includes at least one of a value related to the number of passengers, a value related to the number of passengers at the boarding location, a value related to the number of passengers disembarking at the stopping location, or a value related to the number of passengers at the passing location.

[0126] This makes it possible to calculate, for example, the estimated number of people getting on or off the transportation device.

[0127] In addition, in this embodiment, the estimation unit estimates a data set that represents traffic volume related to the number of people getting on and off the transportation device.

[0128] This allows the estimation unit to estimate traffic volume related to the number of passengers getting on and off, for example, and allows the group management system of transportation equipment to perform appropriate control.

[0129] In addition, in this embodiment, the data set is characterized as an OD table in which the boarding location or the stopping location is set on the vertical axis or the horizontal axis, respectively, and the traffic volume value is shown at the corresponding position on the table from the vertical axis and the horizontal axis.

[0130] This makes it possible to easily grasp the traffic volume within a given time period from the OD table, and by estimating the values in the OD table, the group management system of traffic equipment can perform appropriate control.

[0131] Furthermore, this embodiment is characterized in that the sensor data corresponds to the sum of variables in the partial areas of the OD table, as expressed by an equation.

[0132] This allows, for example, the variables in an OD table to be estimated by expressing sensor data in an equation.

[0133] Furthermore, in this embodiment, the generation unit divides a data set under a predetermined condition, extracts features for each of the divided data, and generates an equation.

[0134] This allows, for example, to simplify equations by allowing features to be extracted from the split data.

[0135] In addition, the sensing server of this embodiment includes an acquisition unit that acquires sensor data from at least one sensor that acquires data regarding the number of target individuals, a generation unit that generates additional constraints based on the sensor data, a derivation unit that derives an estimation model that estimates a value regarding the number of target individuals using at least the additional constraints, and an estimation unit that performs optimization processing using the estimation model and the sensor data.

[0136] In addition, the sensing method of this embodiment includes an acquisition step of acquiring sensor data from at least one sensor that acquires data regarding the number of target individuals, a generation step of generating additional constraints based on the sensor data, a derivation step of deriving an estimation model that estimates a value regarding the number of target individuals using at least the additional constraints, and an estimation step of performing optimization processing using the estimation model and the sensor data.

[0137] In addition, the sensing program of this embodiment is a sensing program that estimates required parameters from sensor data of at least one sensor that acquires data regarding the number of target individuals, and includes an acquisition step that acquires sensor data, a generation step that generates additional constraints based on the sensor data, a derivation step that derives an estimation model that estimates a value regarding the number of target individuals using at least the additional constraints, and an estimation step that performs optimization processing using the estimation model and the sensor data.

[0138] The present invention is not limited to the above-described embodiments, and various modifications can be made in the implementation stage without departing from the spirit of the invention. Furthermore, the embodiments may be implemented in combination as appropriate as possible, and in such a case, the combined effects can be obtained. Furthermore, the above-described embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining the disclosed multiple constituent elements.

[0139] Although several embodiments of the present invention have been described above, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0140] 101...On-site equipment and facilities 102...Sensor 104...Central server 106...Acquisition unit (temporary data storage unit) 107...Generation section (additional constraint generation section) 108...Memory unit (base model storage unit) 109...Derivation unit (estimation model holding unit) 110...Estimation unit (optimization processing data estimation unit) 111...Output unit (output interface)

Claims

1. at least one sensor that acquires data regarding the number of individuals of interest; an acquisition unit that acquires sensor data from at least one of the sensors; a generation unit that generates additional constraints based on the sensor data; a derivation unit that derives an estimation model that estimates a value related to the number of target individuals using at least the additional constraint; an estimation unit that performs optimization processing using the estimation model and the sensor data; A sensing system comprising:

2. the estimation model is generated using the additional constraints and the base model. The sensing system of claim 1 .

3. The estimation model is expressed in the form of an equation, the estimation unit executes a process of minimizing or maximizing an objective function using the estimation model as a constraint condition. The sensing system of claim 1 .

4. The sensor is installed in transportation equipment such as an elevator or a bus, The sensor data includes at least one of a value related to the number of passengers, a value related to the number of passengers at a boarding location, a value related to the number of passengers at a stop location, or a value related to the number of passengers at a passing location. The sensing system of claim 1 .

5. The estimation unit estimates a data set representing traffic volume related to the number of passengers getting on and off the transportation device. The sensing system according to claim 4 .

6. The data set has boarding locations and stopping locations set on the vertical axis and horizontal axis, respectively; The OD table indicates the traffic volume value at a position corresponding to the vertical axis and the horizontal axis. The sensing system according to claim 5 .

7. The sensor data corresponds to a sum of variables in a partial area of the OD table, and the sum is expressed by an equation. The sensing system of claim 6 .

8. The generation unit divides the data set under a predetermined condition, extracts features for each of the divided data, and generates an equation. The sensing system of claim 7 .

9. an acquisition unit that acquires sensor data from at least one sensor that acquires data regarding the number of individuals of interest; a generation unit that generates additional constraints based on the sensor data; a derivation unit that derives an estimation model that estimates a value related to the number of target individuals using at least the additional constraint; an estimation unit that performs optimization processing using the estimation model and the sensor data; A sensing server comprising:

10. acquiring sensor data from at least one sensor that acquires data regarding the number of individuals of interest; generating additional constraints based on the sensor data; a derivation step of deriving an estimation model that estimates a value related to the number of individuals of interest using at least the additional constraints; an estimation step of performing optimization processing using the estimation model and the sensor data; A sensing method comprising:

11. A sensing program that estimates a required parameter from sensor data of at least one sensor that acquires data on the number of target individuals, an acquisition step of acquiring the sensor data; generating additional constraints based on the sensor data; a derivation step of deriving an estimation model that estimates a value related to the number of individuals of interest using at least the additional constraints; an estimation step of performing optimization processing using the estimation model and the sensor data; A sensing program that includes:

Images