Route estimation device, route estimation method, and program

Abstract

To provide a route estimation device capable of estimating a movement route of a terminal device with good spatial and temporal resolution, even if sensors are not necessarily installed densely.SOLUTION: A state transition model storage section stores a transition probability between units from a discretized time point ti to a time point ti+1 as a state transition model where the state of moving from segment to segment in a walking space is taken for a unit. The data acquisition section acquires log data, which is a set of sensor identification information, signal strength of the received wireless signal, and the time point of reception, when a wireless signal transmitted from a wireless terminal device is received by sensors installed at multiple locations. An estimation processing section probabilistically estimates the movement route of the wireless terminal device as a series of segments based on the state transition model, the time series of the log data, and the relation between the receiving position of the sensor associated with the sensor identification information contained in the log data and the position of each segment.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a route estimation device, a route estimation method, and a program. [Background technology]

[0002] There is a need to estimate the paths of people (pedestrians) moving in pedestrian spaces using relatively simple devices. If pedestrian paths can be estimated, it will be useful for spatial planning and design.

[0003] In recent years, active efforts to realize smart cities have been underway, aiming to solve social problems by utilizing ICT (Information and Communication Technology).In the field of civil engineering planning, there are also efforts aimed at understanding the actual flow of people and improving mobility.

[0004] Non-Patent Document 1 creates people flow statistical data by aggregating mobile GPS data provided by mobile communication carriers. These statistics are created using movement tracking data from GPS data measured every five minutes. These statistics combine attribute data such as age, gender, and place of residence with behavioral data such as travel route, means of transportation, and stops, and because they are measured using GPS, they have a spatial resolution of about a few meters.

[0005] Non-Patent Document 2 describes a study that estimated pedestrian OD traffic volume using Wi-Fi packet sensors. The method described in Non-Patent Document 2 uses partial OD traffic volume and route traffic volume observed by Wi-Fi packet sensors to estimate pedestrian OD traffic volume from pedestrian traffic volume survey results. This method focuses on tracking people's movement routes and migratory behavior within urban areas of approximately several kilometers in size.

[0006] The research described in Non-Patent Document 3 proposes a method for indoor locations to identify the location of wireless devices carried by pedestrians using the principle of triangulation based on the received signal strength indicator (RSSI) of Wi-Fi and Bluetooth (registered trademark).

[0007] The research described in Non-Patent Document 4 proposes a method for measuring movement changes that takes into account the placement of measurement equipment. This method takes into account the placement positions of beacons at the time of the survey and measures the displacement vector within the measurement area from the acquired RSSI. This method has an accuracy of within an average error of 2.4 meters, providing sufficient resolution. [Prior art documents] [Non-patent literature]

[0008] [Non-Patent Document 1] Zenrin DataCom Co., Ltd., Zenrin DataCom Congestion Statistics, Internet, https: / / www.zenrin-datacom.net / solution / congestion, accessed March 1, 2024. [Non-patent document 2] Yuta Sueki and Kuniaki Sasaki, Estimating Pedestrian OD Traffic Volume in Urban Areas Using Data Obtained from Wi-Fi Packet Sensors, City Planning Institute of Japan, Journal of Urban Planning, Vol. 54, No. 3, October 2019. [Non-patent document 3] Teruaki Kitasuka, Tsuneo Nakanishi, and Akira Fukuda, "Indoor Positioning Method Using Wireless Communication Networks," Transactions of the Information Processing Society of Japan: Computing Systems, Vol. 44, No. SIG 10 (ACS2), July 2003. [Non-patent document 4] Tatsuya Furudate, Miyoshi Horikawa, Kazuyuki Hashimoto, Daiki Kudo, and Azuma Okamoto, "Proposal of a Pedestrian Positioning Method Using Bluetooth Low Energy Beacons," FIT2016 (15th Forum on Information Science and Technology), M-028, pp. 249-250. Summary of the Invention [Problem to be solved by the invention]

[0009] However, the above-mentioned prior art techniques each have the following problems.

[0010] The statistics in Non-Patent Document 1 are based on positioning every five minutes. However, positioning every few minutes provides too coarse a time resolution to track the trajectories of pedestrians in a walking space.

[0011] The technology described in Non-Patent Document 2 focuses on tracking people's movement routes and wandering behavior in urban areas spanning several kilometers, but it is not capable of tracking routes with fine accuracy. For example, the technology described in Non-Patent Document 2 cannot track people's movements with a spatial resolution of around 10 meters in a walking space of within a few hundred meters.

[0012] The technology described in Non-Patent Document 3 is based on the triangulation method, which requires that signals are always received simultaneously from sensors in three or more locations. In other words, to use the technology described in Non-Patent Document 3, the sensors must be arranged quite densely.

[0013] In order to estimate a travel path in a pedestrian space using the technology described in Non-Patent Document 4, the sensors must receive signals continuously over time to determine position, so the sensors must be densely arranged.

[0014] The present invention has been made based on the recognition of the above-mentioned problems, and aims to provide a route estimation device, a route estimation method, and a program that can estimate the movement route of a pedestrian holding a terminal device with a spatial resolution of several meters to several tens of meters and a temporal resolution of several seconds to several tens of seconds, without necessarily having to place sensors densely (even if signal reception is interrupted in some areas). [Means for solving the problem]

[0015] [1] In order to solve the above problem, a path estimation device according to one aspect of the present invention uses a state of moving from segment to segment in a walking space made up of connected segments as a unit, and calculates a time t i From t i+1 a data acquisition unit that acquires log data that is a combination of sensor identification information, which is information that identifies a sensor when the sensor, which is installed at each of a plurality of positions, receives a wireless signal transmitted from a specific wireless terminal device, the signal strength of the received wireless signal, and the time of reception; and an estimation processing unit that probabilistically estimates the route traveled by the wireless terminal device as a series of segments based on the state transition model read out from the state transition model storage unit, the time series of the log data acquired by the data acquisition unit, and the relationship between the reception position of the sensor associated with the sensor identification information contained in the log data and the position of each of the segments.

[0016] [2] Furthermore, one aspect of the present invention is that in the route estimation device of [1] above, the estimation processing unit probabilistically estimates the route traveled by the wireless terminal device by referring to the probabilistic relationship between the distance from the receiving position of the sensor to the segment and the signal strength, based on results of prior actual measurements.

[0017] [3] Furthermore, one aspect of the present invention is that in the path estimation device of the above [1] or [2], the state transition model is configured based on the transition probability based on statistics obtained by observing the actual movement of the wireless terminal device or a person holding the wireless terminal device in the pedestrian space.

[0018] [4] Furthermore, in one aspect of the present invention, in the route estimation device according to any one of [1] to [3] above, the estimation processing unit i The state of moving between the segments at time t i From the next time ti+1 In this method, the state transition probability between units is used as the weight for connecting the units, and a Boltzmann machine model in which the state value of the unit is probabilistically 0 or 1 is used to calculate the probability that the state value of each unit will be 1 using simulated annealing, and the route taken by the wireless terminal device is determined based on the series of units that maximizes the sum of the probability state values ​​of the units.

[0019] [5] In addition, one aspect of the present invention is a route estimation device according to any one of [1] to [4] above, further comprising an estimation result output unit that performs processing to plot on a map the route estimated by the estimation processing unit based on information on the position of the segment on the map.

[0020] [6] In addition, one aspect of the present invention is a route estimation method for estimating a moving route of a wireless terminal device that transmits a signal, in which a state transition model storage unit calculates a state of moving from segment to segment in a walking space formed by connecting segments as a unit, and calculates a time t i From t i+1 a data acquisition unit acquires log data that is a combination of sensor identification information, which is information for identifying a sensor when the sensor, which is installed at each of a plurality of positions, receives a wireless signal transmitted from a specific wireless terminal device, the signal strength of the received wireless signal, and the time of reception; and an estimation processing unit probabilistically estimates the route traveled by the wireless terminal device as a series of segments based on the state transition model read out from the state transition model storage unit, the time series of the log data acquired by the data acquisition unit, and the relationship between the reception positions of the sensors associated with the sensor identification information contained in the log data and the positions of each of the segments.

[0021] [7] In addition, one aspect of the present invention is to consider the state of moving from segment to segment as a unit in a walking space made up of connected segments, and to consider the time ti From t i+1 a data acquisition unit that acquires log data that is a combination of sensor identification information that identifies a sensor when the sensor, which is installed at each of a plurality of positions, receives a wireless signal transmitted from a specific wireless terminal device, the signal strength of the received wireless signal, and the time of reception; and an estimation processing unit that probabilistically estimates a route traveled by the wireless terminal device as a series of segments, based on the state transition model read from the state transition model storage unit, the time series of the log data acquired by the data acquisition unit, and the relationship between the reception positions of the sensors associated with the sensor identification information contained in the log data and the positions of each of the segments. [Effects of the Invention]

[0022] According to the present invention, it is possible to estimate the movement path of a wireless terminal device probabilistically based on wireless signals received by sensors that are not necessarily densely arranged. [Brief explanation of the drawings]

[0023] [Figure 1] 1 is a block diagram showing a schematic functional configuration of a route estimation system according to an embodiment of the present invention; [Figure 2] FIG. 2 is a schematic diagram showing an example of a pedestrian space model handled by the embodiment. [Figure 3] 10 is a graph created based on the measurement results of signal strength in the embodiment, showing the relationship between the distance from the wireless terminal device to the sensor and the RSSI value. [Figure 4] 1 is a schematic diagram showing a state transition model of a pedestrian in the same embodiment, showing only mutual connections with positive weights between units whose state transition probability is greater than 0. FIG. [Figure 5] FIG. 10 is a schematic diagram showing a state transition model of a pedestrian in the embodiment, showing only mutual connections with negative weights between units whose state transition probability is 0. [Figure 6] 10 is a schematic diagram showing locations where the RSSI of a Bluetooth signal was measured in a demonstration experiment of the embodiment. FIG. [Figure 7] 10 is a graph showing the results of measuring RSSI in a demonstration experiment of the same embodiment. [Figure 8] 10 is a bar graph showing a probability distribution of the intervals of the distance between the wireless terminal device and the sensor according to the intervals of the range of the RSSI value measured in the demonstration experiment of the same embodiment. [Figure 9] 10 is a graph showing the state of temperature decay in fast simulated annealing in this embodiment. [Figure 10] FIG. 10 is a schematic diagram showing the positions in a walking space where sensors were installed in a demonstration experiment of the same embodiment. [Figure 11] 10 is a schematic diagram showing a first route search result actually obtained by using the embodiment, which is a matrix of state units (vertical direction) and elapsed time (horizontal direction). FIG. [Figure 12] FIG. 12 is a schematic diagram showing the routes (correct solution and estimated results) corresponding to FIG. 11 on a map. [Figure 13] 10 is a schematic diagram showing a second route search result actually obtained by using the embodiment, illustrating a matrix of state units (vertical direction) and elapsed time (horizontal direction). FIG. [Figure 14] FIG. 14 is a schematic diagram showing the routes (correct solution and estimated results) corresponding to FIG. 13 on a map. [Figure 15] 4 is a schematic diagram showing an example of the configuration of data (part of map information) indicating the installation positions of sensors according to the embodiment. FIG. [Figure 16] 3 is a schematic diagram showing an example of the configuration of data (part of map information) representing the position of a route (segment) according to the embodiment. FIG. [Figure 17] 3 is a schematic diagram showing an example of the configuration of information stored in a state transition model storage unit according to the embodiment. FIG. [Figure 18] 10 is a schematic diagram showing the configuration of signal intensity log data collected from a plurality of sensors by a data acquisition unit according to the embodiment. FIG. [Figure 19] FIG. 2 is a block diagram showing an example of the internal configuration (computer) of the route estimation device according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0024] Next, an embodiment of the present invention will be described with reference to the drawings.

[0025] [Summary] This embodiment realizes a path estimation system for estimating the spatiotemporal path of a person walking in a pedestrian space. The path estimation system 1 of this embodiment divides the pedestrian space into "segments" each having a length of about several tens of meters, and estimates the movement trajectory of the pedestrian in a discretized space-time with the time axis at intervals of about several seconds to a dozen seconds.

[0026] In the route estimation system 1, pedestrian movement is considered as a chain of state transitions, where pedestrians move from a certain segment to that segment itself or to an adjacent segment. The route estimation system 1 constructs and uses a "movement state transition model" that reflects the overall flow of pedestrians. Note that the "movement state transition model" may also be simply called a "state transition model."

[0027] The path estimation system 1 regards the distribution of RSSI (Received Signal Strength Indicator) according to the distance from the wireless terminal device 3 carried by the pedestrian to the sensor 5 as a posterior probability distribution, and constructs an "observation model" of the RSSI captured by the sensor 5 when the pedestrian is present in a certain segment. The wireless terminal device 3 is a device that emits a Wi-Fi wireless signal or a Bluetooth wireless signal. The sensor 5 receives the wireless signal emitted by the wireless terminal device 3 and measures its signal strength (RSSI).

[0028] In the path estimation system 1 of this embodiment, the above-mentioned "movement state transition model" and "observation model" are expressed by an Ising model, and a Boltzmann machine, which is a type of interconnected neural network, is constructed. Note that the Ising model is a lattice model in statistical mechanics that is composed of lattice points that take two configuration states and considers interactions only between nearest-neighbor lattice points.

[0029] In the route estimation system 1, multiple sensors 5 appropriately placed within a walking space measure the RSSI of signals from wireless terminal devices 3 having specific identification information. In the route estimation system 1, the time series of RSSI from these multiple sensors 5 is used as input values ​​to a Boltzmann machine, and an approximate solution is obtained for each segment as to the probability that a pedestrian is present at each time, and the most likely travel route from the starting point to the end point is obtained as the estimated result value.

[0030] The following explains the terminology used in this embodiment. A path is a general term for a route that a pedestrian can walk within a pedestrian space. A path is composed of a series of multiple segments. In this case, a segment in a path (series) is connected to the next segment in the path (series). A segment is a path that does not branch from one node to another. A segment itself may be composed of only straight lines or may include curves. In this embodiment, a segment is the smallest unit of a path. Segments may be connected to each other at a node. There may be multiple nodes. Typically, a pedestrian can move in both directions through a segment. A state unit is the smallest unit when a pedestrian moves along a path. One state unit is movement from one segment to another. Movement in opposite directions between different pairs of segments is treated as a different state unit. In other words, movement from segment S1 to S2 (represented as S1 → S2, etc.) and movement from segment S2 to S1 (represented as S2 → S1, etc.) are different state units. Additionally, movement from one segment to the same segment (for example, movement from segment S1 to S1, expressed as "S1 → S1") is also one state unit.

[0031] The resulting path estimated by the path estimation system of this embodiment is a sequence of units, which are sequentially connected at nodes and have a consistent direction.

[0032] [Functional configuration of the route estimation system] FIG. 1 is a block diagram showing a schematic functional configuration of a route estimation system according to this embodiment. As shown in the figure, the route estimation system 1 includes a route estimation device 2, a wireless terminal device 3, and a sensor 5. Although only one wireless terminal device 3 is shown in the figure, multiple wireless terminal devices 3 may be present in the pedestrian space targeted by the route estimation system 1. The route estimation device 2 can individually estimate the routes of multiple pedestrians carrying wireless terminal devices 3 based on signals from each wireless terminal device 3. Furthermore, although three sensors 5 are shown in the figure, any number of sensors 5 may be used to detect wireless signals emitted in the pedestrian space targeted by the route estimation system 1.

[0033] Devices such as the route estimation device 2, the wireless terminal device 3, and the sensor 5 are realized using, for example, electronic circuits. Each device may also include internal storage means such as a semiconductor memory or a magnetic hard disk drive, as needed. At least some of the functions of each device may also be realized by a computer and software (programs).

[0034] The route estimation device 2 estimates the movement route of a pedestrian (in other words, the wireless terminal device 3 carried by the pedestrian) based on the log data passed from the sensor 5. In other words, the route estimation device 2 executes a route estimation method.

[0035] The wireless terminal device 3 is a device capable of transmitting a Wi-Fi wireless signal or a Bluetooth wireless signal. The wireless terminal device 3 is carried by a pedestrian (person). Alternatively, the wireless terminal device 3 is attached to the body of the pedestrian. The wireless terminal device 3 may be, for example, a smartphone, a wristwatch-type terminal device, other wearable terminal devices, earphones, a speaker, a video display device, a tag that emits a wireless signal (for example, a tag for preventing loss of an item), or other devices. The wireless terminal device 3 can transmit and receive data using wireless signals.

[0036] The wireless signal emitted by the wireless terminal device 3 includes identification information for uniquely identifying each wireless terminal device 3. As long as the pedestrian carrying the wireless terminal device 3 does not change during the journey, the identification information for uniquely identifying the wireless terminal device 3 can also be used to identify the pedestrian.

[0037] 1 shows only one wireless terminal device 3, there may be multiple wireless terminal devices 3 that are simultaneously emitting wireless signals. Even in a situation where multiple wireless terminal devices 3 are simultaneously emitting wireless signals, the wireless terminal devices 3 can be individually identified by the above-mentioned identification information.

[0038] The sensor 5 has a function of receiving a wireless signal and measuring the signal strength. In this embodiment, a plurality of sensors 5 are appropriately arranged in the walking space.

[0039] In this case, the sensors 5 do not need to be arranged very densely within the pedestrian space. In other words, it is not necessary for multiple sensors 5 to simultaneously receive a wireless signal transmitted from one wireless terminal device 3. Also, there may be no sensor 5 that temporarily receives a wireless signal transmitted from one wireless terminal device 3. In other words, the multiple sensors 5 do not need to completely cover the pedestrian space. Also, there may be parts of the pedestrian space that are not covered at all by sensors 5.

[0040] Furthermore, the sensor 5 does not necessarily have to constantly receive a wireless signal from the wireless terminal device 3. For example, there may be a time period during which the sensor 5 receives a wireless signal from the wireless terminal device 3, such as once every few seconds or once every 10-odd seconds.

[0041] The sensor 5 outputs a record of receiving a wireless signal from the wireless terminal device 3 as log data. The content and format of the log data will be explained later. The sensor 5 can pass the output log data to the path estimation device 2 online or offline (for example, via an information recording medium).

[0042] Next, we will explain the internal functional configuration of the route estimation device 2. As shown in the figure, the route estimation device 2 includes a data acquisition unit 21, an estimation processing unit 22, a map information storage unit 25, a state transition model storage unit 26, and an estimation result output unit 28.

[0043] The data acquisition unit 21 acquires log data, which is a set of sensor identification information for identifying the sensor 5 when the sensors 5 installed at multiple locations receive a wireless signal transmitted from a specific wireless terminal device 3, the signal strength of the received wireless signal, and the time of reception. The data acquisition unit 21 passes the acquired log data to the estimation processing unit 22 in an appropriate format. An example of a specific configuration of the log data will be described later with reference to FIG. 18.

[0044] The estimation processing unit 22 probabilistically estimates the route traveled by the wireless terminal device 3 as a series of segments based on the state transition model (information on the probability of transition from unit to unit) read from the state transition model storage unit 26, the time series of the log data acquired by the data acquisition unit 21, and the relationship between the receiving positions of the sensors associated with the sensor identification information contained in the log data and the positions of each segment. As described above, the estimation processing unit 22 estimates the route as a series of segments. However, as a processing procedure, the estimation processing unit 22 may first probabilistically estimate the route as a series of units and then convert the information on the series of units into information on the series of segments. As described above, a unit is information that can be expressed as "S1 → S2," etc., and a sequence of units is expressed, for example, as ... → "S1 → S2" → "S2 → S3" → ..., etc. Here, in a sequence of units, the final segment of a certain unit is the initial segment of the next unit (e.g., S2 above). Therefore, route information expressed as a sequence of units is equivalent to route information expressed as a sequence of segments. Note that, unless the person holding the wireless terminal device 3 changes, the estimation processing unit 22 estimating the route traveled by the wireless terminal device 3 is equivalent to estimating the route traveled by the person holding the wireless terminal device 3.

[0045] Here, the "relationship between the reception position of the sensor associated with the sensor identification information in the log data and the position of each of the segments" is as follows: The log data has sensor identification information. The sensor identification information is information for identifying the sensor 5. Once the sensor 5 is identified, the location where the sensor 5 is installed is determined. In other words, the location where the sensor 5 receives a wireless signal is determined. The relationship between the position of the sensor 5 and the position of the segment includes the distance from the sensor 5 to the segment. If the wireless terminal device 3 is present in the segment, the distance is the distance from the wireless terminal device 3 to the sensor 5. Meanwhile, the log data includes information on the signal strength of the received wireless signal. As will be described later, given this signal strength, a probability distribution of the distance from the wireless terminal device 3 to the sensor 5 is determined. In other words, the "relationship between the reception position of the sensor associated with the sensor identification information in the log data and the position of each of the segments" and the strength obtained from the log data are information that the estimation processing unit 22 can use to probabilistically infer the segments that make up the route.

[0046] In addition, the estimation processing unit 22 may probabilistically estimate the route taken by the wireless terminal device by referring to the probabilistic relationship between the distance from the receiving position of the sensor to the segment and the signal strength, based on results of previous measurements.

[0047] Furthermore, the estimation processing unit 22 i The state of moving between the segments at time t i From the next time t i+1In this case, the state transition probability between units is used as a weight for connecting the units, and a Boltzmann machine model in which the state value of the unit is probabilistically 0 or 1 is used to calculate the probability that the state value of each unit will be 1 by simulated annealing, and a sequence of units that maximizes the sum of the unit state values ​​may be determined as the route traveled by the wireless terminal device 3. Details of this processing procedure will be described later. The route determined (estimated) by the estimation processing unit 22 is constructed by connecting the units in order. How a route is constructed by connecting units will be described later with reference to examples in FIGS. 11 and 13.

[0048] The map information storage unit 25 stores map information. The map information storage unit 25 also stores information on the positions of the sensors 5 in association with this map (see FIG. 15). The map information storage unit 25 also stores information on the positions of segments that may constitute a route in association with this map (see FIG. 16). The relationship between segments and units has already been explained. The information stored in the map information storage unit 25 can be read by the estimation processing unit 22.

[0049] The state transition model storage unit 26 regards the state of moving from segment to segment in a walking space formed by connecting segments as a unit, and calculates the state transition model at time t i to the next time t i+1 The state transition model storage unit 26 stores information on the transition probabilities between units to as a state transition model. The state transition model may be configured based on transition probabilities based on statistics obtained by observing the actual movement of the wireless terminal device 3 or the person carrying the wireless terminal device 3 in the pedestrian space. The information stored in the state transition model storage unit 26 can be read by the estimation processing unit 22. The state transition model will be further described later with reference to FIGS. 4, 5, and 17.

[0050] The estimation result output unit 28 outputs information about the route (estimation result) estimated by the estimation processing unit 22. Note that the estimation result output unit 28 may output the route information in the form of matrix data, as will be shown later in Fig. 11 or 13, for example. The estimation result output unit 28 may also perform processing to draw the route estimated by the estimation processing unit 22 on a map, based on information about the positions on the map of multiple segments that make up the route.

[0051] [Pedestrian space model] FIG. 2 is a schematic diagram showing an example of a pedestrian space model handled by this embodiment. In the illustrated example, the pedestrian space is overlaid on a road map. This pedestrian space is divided into segments. Each of the circles shown in the figure represents a segment. Segments Z1 to Z8 are segments corresponding to the starting and ending points of pedestrians. Segments A to F are segments other than the starting and ending points that cover the pedestrian space. This model shows a graph structure in which segments are nodes and movement between segments is represented by an arc. When a pedestrian can move directly from one segment to another, there is an arc connecting those segments.

[0052] The path estimation system 1 of this embodiment estimates a path from a pedestrian state transition model (a model representing the transition from one unit to another). A pedestrian moving within this pedestrian space (a pedestrian whose path is to be estimated) is present in one of the units within the pedestrian space at any time.

[0053] [RSSI observation model] In the route estimation system 1 of this embodiment, in a situation where there is no guarantee that a wireless signal (such as a Bluetooth signal) transmitted from one wireless terminal device 3 can be received simultaneously by multiple sensors 5, the distance band in which the wireless terminal device 3 is located is estimated from the RSSI of the signal received by one sensor 5.

[0054] To this end, the distance from the wireless terminal device 3 to the sensor 5 is varied, the RSSI of the signal received by the sensor 5 is measured, and the average and variance are calculated in advance. From the results of measuring the RSSI at various distances, an approximate curve of a graph showing the relationship between distance and RSSI can be created. Note that the RSSI measurements performed to create this approximate curve do not necessarily have to be performed in an actual walking space.

[0055] FIG. 3 is a graph created based on the measurement results, showing the relationship between the distance from the wireless terminal device 3 to the sensor 5 and the RSSI value. The horizontal axis of this graph corresponds to the distance from the wireless terminal device 3 to the sensor 5, and the vertical axis corresponds to the measured RSSI value. The figure shows a graph of the average RSSI value against distance (solid line) and graphs of RSSI (average value + σ) and (average value - σ) (dashed lines). To create this graph, the probability of being in each distance band is calculated for each RSSI rank divided by a certain width. In this figure, the RSSI value is I s From I s+1 The distance is L h From L h+1 The probability that it exists in the interval (rank h) up to w hs By summing up and normalizing this probability by RSSI rank, a probability distribution of the distance range (distance from the wireless terminal device 3 to the sensor 5) in which a pedestrian exists when that RSSI rank is obtained can be obtained.

[0056] The signal output strength on the wireless terminal device 3 side is within the range defined by the Wi-Fi standard or the Bluetooth standard.

[0057] [Boltzmann machine-based spatiotemporal route estimation model] As a method for estimating a spatiotemporal route, the route estimation system 1 of this embodiment creates a walking space model created as a graph structure and a state transition model of the pedestrian's movement state in discretized time at regular time intervals, and performs Boltzmann machine estimation.

[0058] 4 and 5 are schematic diagrams showing a state transition model of a pedestrian. The state transition model of a pedestrian is represented as a set of mutual connections between state units. One mutual connection has either a positive weight or a negative weight. In FIG. 4 and FIG. 5, for ease of viewing, mutual connections with positive weights and total connections with negative weights are shown in separate diagrams. That is, each solid line between state units in FIG. 4 represents a mutual connection with a positive weight. Also, each dashed line between state units in FIG. 5 represents a mutual connection with a negative weight.

[0059] In each of Figs. 4 and 5, at time t i The state unit at time t i+1 It represents the state unit at time t i and t i+1 is a discrete time. Time t i and time t i+1 Although not shown in Figures 4 and 5, at time t i Before, ,time t i-3 , time t i-2 , time t i-1 There exists a discrete time. Also, at time t i+1 After that, time t i+2 , time t i+3 , time t i+4 ,··· there are discrete times.

[0060] At some time t i In this case, there is only one state unit for each pedestrian. i The state units belonging to are interconnected with negative weights (dashed lines). i+1 Similarly, at the same time t i+1 The state units belonging to are interconnected with negative weights (dashed lines).

[0061] time t i and the state units belonging to the next time t i+1 Between the state units belonging to There are either mutual connections with positive weights (solid lines) or mutual connections with negative weights (dashed lines). i A state unit belonging to the next time t i+1 If the relationship between a state unit belonging to is a rational path (if it is a probabilistic transition), those state units are mutually connected with a positive weight (solid line). A mutual connection with a positive weight (solid line) is assigned an appropriate state transition probability. For example, in Figure 4, at time t i The state unit "Z1→A" at time t i+1 The mutual connection (positive weight, solid line) between the state unit "A → Z3" at time t i The state unit "Z1→A" at time t i+1 The mutual connection (positive weight, solid line) between state units "A → B" in is given a state transition probability of 0.2. The other overall connections with positive weights are also given similar state transition probability values. Conversely, in Figure 5, For example, at time t i The state unit "Z1→A" at time t i+1 The mutual connection between the state unit "A←Z3" in has a negative weight.

[0062] The estimation of the travel path by the Boltzmann machine can be performed by calculating the following equation (1): Note that, hereinafter, the state unit may be simply referred to as a "unit."

[0063]

number

[0064] In equation (1), W ij is the weight of the connection between units ij. i is the state value of unit i, and the state value is one of the elements of the set {0,1}. Also, θ iis the bias value of unit i, and T is the temperature parameter. Annealing is performed from ∞ to 0 with respect to the temperature parameter T. Then, p i is the probability that the state value of unit i transitions to 1. This probability p i Calculating this is the estimation of the pedestrian's path.

[0065] Bias value θ i Regarding the RSSI rank, let us assume that there are n observation opportunities during the time Δt. Among them, the RSSI rank s is within the distance band L from the representative position of unit i. hk sensor k in the range of m ks When unit i is observed times, the result calculated using the following formula (2) is applied to formula (1). Note that when unit i represents a movement state from segment S1 to S2, the position of unit i may be a position belonging to segment S1 or S2. Typically, the position of unit i is appropriate to be the junction of segments S1 and S2, etc.

[0066]

number

[0067] Regarding the temperature parameter T, annealing is performed using Fast Simulated Annealing. Annealing is performed using the following equation (3).

[0068]

number

[0069] In equation (3), T(n) is the temperature at the nth annealing, and T0 is the initial temperature.

[0070] In this embodiment, for example, the initial temperature of the temperature parameter may be set to 10°C, and the number of annealing cycles may be set to 1000. During this annealing process, the number of times that each unit is in the state of 1 is recorded, and the value obtained by dividing this number by the number of annealing cycles is used as the accuracy of path search.

[0071] [Route search model] In the process of path search, the path estimation device 2 (estimation processing unit 22) uses the number of transitions to state 1 in the process of annealing for each unit by the Boltzmann machine, and when only one unit belonging to each time t is selected, i The accuracy C of the unit j selected in j and time t i+1 The accuracy C of the unit k selected in k and the weight W of the connection between units j and k jk The value V obtained by multiplying each of jk (=W jk ×C j ×C k The combination that maximizes the sum of the πs is calculated using an algorithm such as Dijkstra's algorithm. At this time, the route that maximizes the sum of the πs is selected while applying the following rules. Note that Dijkstra's algorithm is an existing method for finding a route that minimizes the total cost (or maximizes the total value) in a graph where the cost (value) of each arc is defined.

[0072] Rule 1: The starting and ending points of a pedestrian must be between Z1 and Z8. Rule 2: Units that go backwards from a unit that has already passed through may not pass through.

[0073] [Demonstration experiments and surveys] A demonstration experiment and survey of the embodiment was conducted. Specifically, the area from Jimbocho Station in Chiyoda Ward, Tokyo to Surugadai-shita intersection was the target of processing. The pedestrian space shown in Figure 2 was the target of verification.

[0074] The model of the pedestrian space used in the empirical study was a graph structure as shown in Figure 2. The pedestrian space is represented by taking the points where pedestrian crossings are located as road nodes and other points as starting and ending points.

[0075] A receiver (sensor 5) was installed at a predetermined location, and the RSSI of the received Bluetooth signal was measured. When the distance between the wireless terminal device 3 and sensor 5 was 0 meters, 5 meters, 10 meters, 20 meters, 30 meters, 40 meters, and 50 meters, respectively, a pedestrian holding the wireless terminal device 3 stopped at that location for one minute, and the RSSI of the signal received by sensor 5 during that one minute was measured.

[0076] Figure 6 is a schematic diagram showing the locations where the RSSI of Bluetooth signals was measured in this study. The location shown is a portion of the location shown in Figure 2. Between points A and B, there is a straight section of road (sidewalk), and the distance between them is approximately 50 meters. Between points A and B, the distance between the wireless terminal device 3 and the sensor 5 was changed, and the RSSI was measured.

[0077] Figure 7 is a graph showing the results of RSSI measurements. As shown in the figure, multiple measurements were taken at distances of 0 meters, 5 meters, 10 meters, 20 meters, 30 meters, 40 meters, and 50 meters. The small circles on the graph indicate the actual measured values, and the large circles indicate the average measured values ​​at each distance. The three curves on the graph are approximate curves for the average value + σ, the average value, and the average value - σ (where σ is the standard deviation) at each distance. Statistics from the measurement results show that the shorter the distance, the larger the RSSI value, and the longer the distance, the smaller the RSSI value. Furthermore, the greater the distance, the smaller the difference in RSSI values ​​between adjacent distances (in other words, the slopes of the three curves become gentler).

[0078] Table 1 below shows the probability of existence of each distance band for each RSSI strength, calculated based on the above measurement results. For example, in Table 1, the number (probability) written as "9.985.E-01" is 9.985 x 10 -1 In other words, according to the tabulation results shown in Table 1, the probability of distance can be determined based on the range of RSSI values.

[0079] [Table 1]

[0080] Fig. 8 is a bar graph showing the tabulation results shown in Table 1. That is, the bar graph in Fig. 8 shows the probability distribution of the intervals of the distance between the wireless terminal device 3 and the sensor 5 according to the intervals of the range of the measured RSSI values.

[0081] As shown in Table 1 and Figure 8, there were no cases where the measurement results showed an RSSI value in the range of less than 0 [dB] and greater than or equal to -10 [dB]. However, based on the data trend, when the RSSI value is in the range of less than 0 [dB] and greater than or equal to -10 [dB], the probability of the distance being greater than or equal to 0 meters and less than 10 meters can be considered to be 1.000, just like when the RSSI value is in the range of less than -10 [dB] and greater than or equal to -20 [dB] or less than -20 [dB] and greater than or equal to -30 [dB].

[0082] The above measurement results were used as the bias values ​​of the state units.

[0083] When performing estimation using a Boltzmann machine, we verified the temperature decay using fast simulated annealing. Fast simulated annealing is also described in the references below. References: Harold Szu, FAST SIMULATED ANNEALING, January 1984.

[0084] Figure 9 is a graph showing the temperature decay of fast simulated annealing when T0 = 10. In Figure 9, the horizontal axis corresponds to the number of annealing passes, and the vertical axis corresponds to the temperature. This graph shows that the temperature decays significantly for the first 100 or so passes from the start of annealing. On the other hand, after the 200th annealing pass, the temperature decay becomes very gradual. This gradual temperature decay means that the path estimation may fall into a local solution. Therefore, we decided to reset the number of annealing passes after a certain number of passes. In this study, we performed four sets of estimation using the Boltzmann machine by resetting the number of annealing passes to the initial state every 250 passes. This prevented the estimation results from falling into a local solution.

[0085] Figure 10 is a schematic diagram showing the locations where sensors 5 were installed in this survey. Measurement of the verification data was carried out using four sensors 5 (S1, S2, S3, and S4 in the figure) installed in the walking space also shown in Figure 2. Each sensor 5 received a Bluetooth packet signal and measured the RSSI.

[0086] The sensor 5 receives Bluetooth signals from the wireless terminal device 3 carried by the pedestrian (the subject in this study) and records the reception time (acquisition time), the identification information of the wireless terminal device 3, and the measured RSSI value. The pedestrian traveled along multiple routes, starting from Z1 and ending at Z8. The actual travel patterns included a route that passed through only one of routes A, B, C, and E, and a route that passed through all of routes A, B, C, and E. The pedestrian traveled along both the outbound and inbound routes for a total of four routes. In other words, the pedestrian traveled a total of eight trips. For each of these trips, the traffic light wait status, passing status, and other information were recorded, and this was used as the correct data for verification. Then, based on the measurement data for the eight trips measured by five sensors 5, a route search was performed using the processing method described in this embodiment. The data was compared with the actual route (correct data) to confirm consistency.

[0087] Of the route search results for the above eight trips, two representative cases will be explained with reference to the drawings.

[0088] FIG. 11 is a schematic diagram showing a matrix of the first route search results. In the matrix in this figure, the vertical direction corresponds to state units, and the horizontal direction corresponds to elapsed time. The elapsed time is expressed in the format "mm:ss" (minutes·seconds). In other words, this figure shows the results of estimating state transitions every 15 seconds from elapsed time 0 minutes 00 seconds to 5 minutes 00 seconds. In this figure, elements (cells) indicated by a thick frame and a "●" mark represent the route actually taken by the pedestrian, i.e., the correct answer. Furthermore, elements indicated by hatching represent the estimation results according to this embodiment.

[0089] Fig. 12 is a schematic diagram showing the routes (correct and estimated) corresponding to Fig. 11 on a map. In Fig. 12, the thick solid line represents the route actually taken by the pedestrian, i.e., the correct route. The thick dashed line represents the route estimated by this embodiment.

[0090] In Figure 12, the correct route and the estimated route are almost identical. However, when we check Figure 11, we can see that there are some areas where the passage times do not match.

[0091] Fig. 13 is a schematic diagram showing a matrix of the second route search results. In Fig. 13, as in Fig. 11, elements (cells) indicated by a thick frame and a "●" mark represent the route actually taken by the pedestrian, i.e., the correct answer. Furthermore, elements indicated by hatching represent the estimation results according to this embodiment.

[0092] Fig. 14 is a schematic diagram showing the routes (correct and estimated) corresponding to Fig. 13 on a map. In Fig. 14, the thick solid line also represents the route actually taken by the pedestrian, i.e., the correct route. The thick dashed line represents the route estimated by this embodiment.

[0093] In Figure 14, the starting and ending points of the estimated result match the correct route. However, the route along the way is significantly different from the correct answer.

[0094] [Considerations regarding empirical research] Analysis of the measured RSSI values ​​revealed that an RSSI below -80 [dB] was recorded at the point where the device was actually waiting for the traffic light. In this case, when applying the existence probability shown in Figure 8, the probability of the device being in the distance range of 10 to 20 meters was highest. As a result, it is believed that state units in this distance range were most likely to transition to State 1.

[0095] In addition, in the case shown in Figure 14, the estimated route goes around once and forms a loop. This is thought to be because there was a time when there was no reception record, and as a result of estimation using the Boltzmann machine, it was found that there was no difference in utility regardless of which state unit it was in, so a route that included the sidewalks on both sides of the roadway and had a loop was searched for.

[0096] [Improvement proposal] As shown in the graph in Figure 7, when the distance is short (for example, when the distance is 10 meters or less), the RSSI changes significantly with respect to the change in distance. Therefore, for short distance ranges, an improvement can be made by further dividing the distance range into smaller ranges and calculating the probability of the distance range for the range of RSSI values. This is expected to increase the frequency with which state units that are closer to reality transition to State 1 during estimation.

[0097] Furthermore, when setting the bias value for estimation using a Boltzmann machine, one possible improvement is to apply a negative coupling bias to state units that exist at locations that are clearly far away (for example, locations at a distance equal to or greater than an appropriately set threshold) from the location where the signal was actually observed (sensor 5). This would lower the probability of transitioning to a location other than near the observation location, which is thought to improve the estimation results.

[0098] [summary] This embodiment demonstrates that a spatiotemporal route can be estimated using records of received Wi-Fi or Bluetooth signals, for example, for the purpose of understanding people flow. In this embodiment, a walking space model of the section for which route estimation is desired is created, and a state transition model of movement based on this model is used. In this embodiment, for route search, estimation is performed using a biased Boltzmann machine based on the RSSI value and distance of the signal. The number of times each state unit transitioned to state 1 was recorded. Then, the route with the largest total number of state unit transitions was searched for using the Dijkstra algorithm. The searched route is the estimation result. The estimated route was generally close to the correct route.

[0099] However, there were cases where the estimated route was significantly different from the correct answer. It is believed that such errors can be improved by adjusting the distance-based existence probability distribution of RSSI values ​​and the bias value settings during estimation by the Boltzmann machine.

[0100] [More detailed embodiment examples] More detailed example embodiments are described below.

[0101] The map information storage unit 25 stores map information. The map information is data representing a map such as that shown in FIG. 2. Points on the map can be represented by two-dimensional coordinate values ​​on the map. These coordinate values ​​may be, for example, a combination of latitude and longitude, or may be coordinate values ​​according to another system. A route on the map can be represented, for example, as a series of a finite number of points on the map. In other words, a route can be represented as a finite number of series of the above-mentioned two-dimensional coordinate values.

[0102] The map information storage unit 25 may store additional data on the map. For example, the map information storage unit 25 may store information on each of the locations where multiple sensors 5 are installed. In other words, the installation location of one sensor may be expressed by two-dimensional coordinate values. The map information storage unit 25 may also store information on a pedestrian space model. The pedestrian space model may be expressed as a set of segments. In the example shown in FIG. 2, the pedestrian space model is Z1-A, A-Z3, Z3-C, C-Z5, Z5-E, E-Z7, EF, Z8-F, F-Z6, Z6-D, D-Z4, Z4-B, B-Z2, and BA. Note that each segment may be represented as the above-mentioned route (a finite series of two-dimensional coordinate values).

[0103] FIG. 15 is a schematic diagram showing an example of the configuration of data indicating the locations where sensors 5 are installed. This data is in tabular format and has fields for sensor identification information and location. The information in the location field consists of an x-coordinate value and a y-coordinate value. One row of data in this table represents the location of one sensor 5. In the example shown, the data in this table represents the locations of sensors 5 having sensor identification information S00001, S00002, S00003, S00004, S00005, ... Each location corresponds to a point on the map. The x-coordinate value and y-coordinate value may be numerical values ​​representing longitude and latitude, respectively, or may be other coordinate values.

[0104] FIG. 16 is a schematic diagram showing an example of the structure of data representing the positions of a route (segment). This data is in a tabular format and has items of segment identification information and position. In this data, one piece of segment identification information is associated with information on multiple positions. In other words, one segment has information on the series of multiple positions. In the example shown, this data includes K0001 and K0002 as segment identification information. The segment identified by K0001 is associated with information on the series of two positions. Furthermore, the segment identified by K0002 is associated with information on the series of four positions. These positions are the locations through which each segment passes.

[0105] As mentioned above, the state transition model represents the probability of transitions from one state unit to another. Each state unit corresponds to a segment (route segment) with its own direction and to a movement to the same point (e.g., movement from A to A in Figure 2). In other words, in the example shown in Figure 2, the state units are Z1 → A, A → Z3, Z3 → C, C → Z5, Z5 → E, E → Z7, E → F, Z8 → F, F → Z6, Z6 → D, D → Z4, Z4 → B, B → Z2, and B → A, as well as their respective reverse directions: Z1 ← A, A ← Z3, Z3 ← C, C ← Z5, Z5 ← E, E ← Z7, E ← F, Z8 ← F, F ← Z6, Z6 ← D, D ← Z4, Z4 ← B, B ← Z2, and B ← A, and as movement to the same point: A → A, B → B, C → C, D → D, E → E, and F → F. The state transition model represents the transition probability between state units between adjacent discrete times.

[0106] The transition probability values ​​of the state transition model are determined in advance by observing people walking in a pedestrian space and taking statistics.

[0107] 17 is a schematic diagram showing an example of the configuration of information stored in the state transition model storage unit 26. The state transition model has already been described using FIGS. 4 and 5. As shown in FIG. 17, the state transition model storage unit 26 stores, for example, a state transition model at a discrete time t i All state units at discrete time t i+117 holds the state transition probabilities between all state units in the array. That is, each column of the array shown in FIG. 17 stores an appropriate numerical value. Note that state units are coupled between any discrete time intervals. Since the state transition probabilities between adjacent discrete times can be obtained by referring to the state transition model storage unit 26, the coupling weights of state units between any discrete time intervals can be calculated based on these values. If a transition from a first state unit to a second state unit never occurs, the transition probability between those state units may be set to, for example, 0 or a negative number (such as -1.0). Note that if the data shown in FIG. 17 is redundant, the state transition model storage unit 26 may store the information in a compressed form as appropriate.

[0108] 18 is a schematic diagram showing the configuration of signal strength log data collected by the data acquisition unit 21 from multiple sensors 5. As shown in the figure, the signal strength log data can be represented, for example, as data in a table format. This table of signal strength log data has data items such as sensor identification information, wireless terminal device identification information, measurement date and time, and RSSI value. One row of data in this table corresponds to one wireless signal measurement.

[0109] The sensor identification information is information (ID) for identifying the sensor 5. The wireless terminal device identification information is identification information (ID) for individually identifying the wireless terminal device 3. This wireless terminal device identification information is included in the wireless signal emitted from the wireless terminal device 3. Normally, one wireless terminal device 3 is held by only one pedestrian. Estimating the movement path of a wireless terminal device 3 with specific wireless terminal device identification information essentially estimates the movement path of one pedestrian. The measurement date and time is the date and time when a specific sensor 5 receives and measures a signal from a specific wireless terminal device 3. The measurement date and time is expressed in a format such as "YYYY / MM / DD hh:mm:ss" (year, month, day, hour, minute, second). The sensor 5 measures the signal strength, for example, once every few seconds or once every 10 seconds, and records the measurement results as a log. The RSSI value is the measured signal strength value and is expressed, for example, as a dB (decibel) value.

[0110] The data acquisition unit 21 can pass necessary records from the collected signal strength log data to the estimation processing unit 22. The estimation processing unit 22 receives, for example, a specific piece of wireless terminal device identification information (corresponding to a specific pedestrian), analyzes the log data within a predetermined time range, and estimates the pedestrian's route.

[0111] The estimation processing unit 22 estimates the walking route of a specific pedestrian (corresponding to a specific wireless terminal device 3) based on the received signal strength log data. Specifically, the estimation processing unit 22 calculates the probability that the state value of the state unit i becomes "1" using the above-described formula (1). Here, W in formula (1) ij is the weight of the connection between units i and j, i.e., w ij The value of θ can be obtained by referring to the state transition model (Fig. 17). i is the bias value, which is obtained by equation (2). hks is the distance h from the unit representative position to the sensor k is the probability that a pedestrian is in the movement state indicated by the unit when the signal measured by the sensor 5 in the unit corresponds to the RSSI rank s. ks As mentioned above, is the number of times a signal with RSSI rank s is observed by sensor k in n observation opportunities.

[0112] Through the above calculation, the estimation processing unit 22 counts the number of times the state value of each state unit was "1" during the annealing process in each time period (every 15 seconds in the examples shown in FIGS. 11 and 13). Then, the estimation processing unit 22 uses Dijkstra's algorithm to calculate a route that maximizes the total number of times the state values ​​of the traversed state units (route, segment) transition to "1". The route calculated by the estimation processing unit 22 here is the route estimated for the pedestrian. Examples of estimation results have been described with reference to FIGS. 11, 12, 13, and 14.

[0113] The estimation result output unit 28 outputs information about the route estimated by the estimation processing unit 22 to the outside. The estimation result output unit 28 may output information in the form of a matrix as shown in FIG. 11 or 13, for example. The hatched elements in the matrices shown in FIG. 11 or 13 represent the estimated route. Alternatively, the estimation result output unit 28 may output information about the estimated route in a format in which the information is displayed superimposed on a map, as shown in FIG. 12 or 14, for example. The positions of state units constituting the estimated route are represented by the positions of the segments corresponding to the state units, as shown in FIG. 16.

[0114] FIG. 19 is a block diagram showing an example of the internal configuration of the route estimation device 2. The route estimation device 2 can be realized using a computer. As shown in the figure, the computer includes a central processing unit 901, a RAM 902, an input / output port 903, input / output devices 904 and 905, and a bus 906. The computer itself can be realized using existing technology. The central processing unit 901 executes instructions contained in a program read from the RAM 902 or the like. In accordance with each instruction, the central processing unit 901 writes data to the RAM 902, reads data from the RAM 902, and performs arithmetic and logical operations. The RAM 902 stores data and programs. Each element included in the RAM 902 has an address and can be accessed using the address. RAM stands for "random access memory." The input / output port 903 is a port through which the central processing unit 901 exchanges data with external input / output devices. The input / output devices 904 and 905 exchange data with the central processing unit 901 via the input / output port 903. The bus 906 is a common communication path used within the computer. For example, the central processing unit 901 reads and writes data from and to the RAM 902 via the bus 906. Also, for example, the central processing unit 901 accesses the input / output port 903 via the bus 906.

[0115] At least some of the functions of the route estimation device 2, wireless terminal device 3, and sensor 5 in the above-described embodiment can be implemented by a computer and a program. In this case, a program for implementing each function may be recorded on a computer-readable recording medium, and the program may be loaded into a computer system and executed. Note that the term "computer system" as used herein includes hardware such as an OS and peripheral devices. Furthermore, the term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, CD-ROMs, DVD-ROMs, and USB memory, as well as storage devices such as hard disks built into computer systems. In other words, a "computer-readable recording medium" may be a non-transitory computer-readable recording medium. Furthermore, the term "computer-readable recording medium" may also include a medium that temporarily and dynamically stores a program, such as a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line, or a medium that stores a program for a certain period of time, such as volatile memory within a computer system that serves as a server or client in such a case. The program may be designed to implement some of the above-described functions, or may be capable of implementing the above-described functions in combination with a program already stored in the computer system.

[0116] Although the embodiment has been described above, the present invention can also be implemented in the following modified examples. Note that multiple modified examples may be implemented in combination as long as they can be combined.

[0117] The data acquiring unit 21 may acquire log data information from the sensor 5, for example, online in real time. Alternatively, the data acquiring unit 21 may acquire log data information from the sensor 5, for example, offline, after the pedestrian's movement has finished. Furthermore, the estimation processing unit 22 may perform the process of estimating the route immediately after the data acquiring unit 21 acquires the log data, or may perform the process of estimating the route after a certain time has elapsed since the data acquiring unit 21 acquired the log data.

[0118] As a modified example, the object for which a path is to be estimated may be something other than a pedestrian. For example, the path estimation device 2 may be configured to estimate the path of a person traveling on a moving object that moves at a relatively slow speed, such as an electric kick scooter. Alternatively, the path estimation device 2 may be configured to estimate the moving path of an autonomously moving robot, cart, or the like.

[0119] As a modified example, the route estimation device 2 may simultaneously or in parallel perform the process of estimating routes for a plurality of pieces of terminal identification information.

[0120] As a modified example, the route estimation device 2 may perform a process of estimating a route based on other numerical values ​​that represent the strength of the signal received by the sensor 5, instead of the RSSI.

[0121] As a modified example, the route estimation device 2 may use a Kalman filter or a Kalman smoother to estimate the movement route of the wireless terminal device 3. Alternatively, the route estimation device 2 may express the movement route as an Ising model and then estimate the movement route by processing using a quantum computer.

[0122] Although an embodiment of the present invention has been described above in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs within the scope of the gist of the present invention.

[0123] As described above, according to this embodiment, it is possible to estimate the route of a pedestrian, etc. That is, according to this embodiment, it is possible to estimate the route of a pedestrian, etc. with a spatial resolution of about 10 meters and a temporal resolution of about several seconds to a dozen seconds, based on log data (time-series information on the reception strength of signals emitted from each wireless terminal device 3) acquired by sensors 5 sparsely placed in a walking space spanning, for example, several hundred meters to several kilometers. [Industrial Applicability]

[0124] The present invention can be used, for example, to estimate the movement path of pedestrians, etc., and can be used, for example, in spatial planning and spatial design of cities, etc. However, the scope of use of the present invention is not limited to the examples given here. [Explanation of symbols]

[0125] 1. Route Prediction System 2. Route estimation device 3. Wireless terminal equipment 5 Sensor 5 21 Data Acquisition Section 22 Estimation processing unit 25 Map information storage unit 26 State transition model memory unit 28 Estimation result output section 901 Central Processing Unit 902 RAM 903 Input / Output Ports 904,905 Input / Output Devices 906 Bus

Claims

1. In a walking space made up of connected segments, the state of moving from segment to segment is taken as a unit, and the time t ｉ From time t ｉ＋１ a state transition model storage unit that stores information on the transition probabilities between units as a state transition model; a data acquisition unit that acquires log data that is a combination of sensor identification information, which is information for identifying a sensor when the sensor installed at each of a plurality of locations receives a wireless signal transmitted from a specific wireless terminal device, the signal strength of the received wireless signal, and the time of reception; an estimation processing unit that probabilistically estimates a route traveled by the wireless terminal device as a series of segments based on the state transition model read out from the state transition model storage unit, the time series of the log data acquired by the data acquisition unit, and the relationship between the reception positions of sensors associated with the sensor identification information held in the log data and the positions of each of the segments; A route estimation device comprising:

2. the estimation processing unit probabilistically estimates the route traveled by the wireless terminal device by referring to a probabilistic relationship between the distance from the receiving position of the sensor to the unit and the signal strength, which is based on results of actual measurements taken in advance; The route estimation device according to claim 1 .

3. the state transition model is configured based on the transition probability based on statistics obtained by observing actual movements of the wireless terminal device or a person holding the wireless terminal device in the pedestrian space. The route estimation device according to claim 1 .

4. The estimation processing unit ｉ The state of moving between the segments at time t ｉ From the next time t ｉ＋１ In the method, the state transition probability between units is used as a weight for connecting the units, and a Boltzmann machine model in which the state value of the unit is probabilistically 0 or 1 is used to calculate the probability that the state value of each unit will be 1 by simulated annealing, and the route taken by the wireless terminal device is determined based on the sequence of units such that the sum of the state values ​​of the units is maximized. The route estimation device according to claim 1 .

5. an estimation result output unit that performs processing to plot, on a map, the route estimated by the estimation processing unit based on information about the position of the segment on the map; The route estimation device according to claim 1 , further comprising:

6. A route estimation method for estimating a moving route of a wireless terminal device that transmits a signal, comprising: The state transition model storage unit regards the state of moving from segment to segment in a walking space formed by connecting segments as a unit, and calculates the time t ｉ From time t ｉ＋１ The information on the transition probability between the units to is stored as a state transition model, a data acquisition unit acquires log data that is a set of sensor identification information, which is information for identifying a sensor when the sensor installed at each of a plurality of positions receives a wireless signal transmitted from a specific wireless terminal device, the signal strength of the received wireless signal, and the time of reception; an estimation processing unit probabilistically estimates a route traveled by the wireless terminal device as a series of segments based on the state transition model read out from the state transition model storage unit, the time series of the log data acquired by the data acquisition unit, and the relationship between the reception positions of sensors associated with the sensor identification information held in the log data and the positions of each of the segments; Route estimation method.

7. In a walking space made up of connected segments, the state of moving from segment to segment is taken as a unit, and the time t ｉ From time t ｉ＋１ a state transition model storage unit that stores information on the transition probabilities between units as a state transition model; a data acquisition unit that acquires log data that is a combination of sensor identification information, which is information for identifying a sensor when the sensor installed at each of a plurality of locations receives a wireless signal transmitted from a specific wireless terminal device, the signal strength of the received wireless signal, and the time of reception; an estimation processing unit that probabilistically estimates a route traveled by the wireless terminal device as a series of segments based on the state transition model read out from the state transition model storage unit, the time series of the log data acquired by the data acquisition unit, and the relationship between the reception positions of sensors associated with the sensor identification information held in the log data and the positions of each of the segments; A program for causing a computer to function as a route estimation device comprising:

Images