JP2004213098A - Congestion prediction system, congestion prediction method and congestion prediction program - Google Patents

Abstract

A congestion prediction system for updating a prediction model in real time and accurately predicting future congestion is provided.
A congestion prediction system (1) transmits via a communication network (25) a plurality of receivers (20a, 20b, ..., 20e) in an observation area (26) in which a pedestrian (22) carrying a transmitter (21) moves. A receiver control unit 13 for receiving device information, an external event capturing unit 12 for generating a mobile event based on transmitter information, and a mobile event generated from a mobile model used to predict the behavior of the mobile. Simulated event generator 18, input switcher 14 for switching and outputting the moving event sent from external event capturing unit 12 and simulated event generator 18, and congestion in the future in parallel with updating the moving body information And an event processing unit 16 for performing prediction.
[Selection diagram] Fig. 1

Description

時刻Ｔから次の来場時刻までの時間は、本式をｔについて解けばよい。式（１）を変形し、一様乱数ｕを導入して次式を得る。
【００７０】
【数２】

来場は午前と午後で分け、それぞれの来場時刻分布を正規分布で近似する。これによると、来場密度λ（ｔ） は、μを平均来場時刻、σを標準偏差、AMを午前、PMを午後として、次式で表される。
【００７１】
【数３】

来場モデルの同定としては、実際の午前、午後の来場時刻分布に正規分布をフィッティングさせることにより、例えば、次のような来場モデルが得られる。平均値は９：００を０秒としたときの経過時間を示す。
【００７２】
μ_AM＝3855, σ_AM＝2703
μ_PM＝18838, σ_PM＝3573 ………（４）
来場者の来場時刻を推定するには、午前来場総数Ｎ_AMおよび午後来場総数Ｎ_PMが必要だが、これについては別途与えられるものとする。
【００７３】
＜移動モデル＞
来場者の移動モデルは、次の２つのモデルを合成して定式化する。
【００７４】
・退場モデル
・次会場選択モデル
即ち、来場者がある会場での見学を終了したとき、まず退場モデルを用いて退場するか否かを決定する。もし退場しない結論が出た場合、次会場選択モデルを用いて次に訪問する会場を決定する。
【００７５】
退場モデルは、退場時刻をｔ_L、来場時刻をｔ_A、滞在時間をｔ_S、閉場時効をＴ_Cとして、次式で定義する。
【００７６】
【数４】

Ｄ_L（ｔ）がｔｒｕｅの場合、退場を選択する。ｆａｌｓｅの場合、次会場選択モデルにより次の移動先会場を決定する。滞在時間は次式で定義される確率分布（正規分布）により決める。
【００７７】
【数５】

平均値μ_Sと標準偏差σ_Sは来場者の来場時刻ｔ_Aの一次関数として与える。これは、午前の来場者は午前中に、午後の来場者は閉場時刻までに、それぞれ見学を完了しようとする傾向を表現している。
【００７８】
【数６】

退場時刻ｔ_Lは来場者の状態に依存しない。
【００７９】
次会場選択モデルは、マルコフ連鎖モデルを用いて定式化する。従って、過去の見学履歴は考慮しないモデルとなる。マルコフ連鎖モデルの詳細は、「金子秀二著、“展示会における人の流れのシミュレーション”, アドバンス, No. 6, pp. 33-37, 東芝アドバンストシステム, 1998.」に記載されている。
【００８０】
各来場者の会場間移動を表すマルコフ連鎖をＸ_kとする。来場のための受付をひとつの会場とみなし、受付を含めた展示会場数がＭだとする。受付を会場１、その他の会場を会場２〜会場Ｍと呼ぶ。来場者の状態数はＭである。いま、会場ｉに滞在する状態をｂ_iで表し、時点ｋおいて会場ｉに滞留する確率分布を次式のように書く。
【００８１】
【数７】

これはＭ次元ベクトルとして、以下のように書ける。
【００８２】
p[k] = [p₁[k], p₂[k], …,p_M[k]] ………（９）
全来場者の初期分布は必ずｐ[0]=[1,0,…,0]である。推移確率行列をＭ×Ｍ次元行列Ｐで表す。推移確率行列Ｐのｉ行ｊ列要素は、会場ｉから会場ｊへの推移確率である。
【００８３】
p_ij = P(X_k+1 = b_j | X_k = b_i) ………（１０）
各来場者の時点ｋでの確率分布は次式で表される。
【００８４】
p[k] = p[0]P^k ………（１１）
本モデルでは推移確率行列Pを、対象となる展示会開催中、不変とみなす。
【００８５】
退場モデルの同定としては、閉場時刻Ｔ_Cを１７：００とし、会場時刻９：００からの秒数で表すと次式となる。
【００８６】
T_C = 28800 ………（１２）
滞在時間ｔ_Sの分布については、以下のように求める。
【００８７】
１．サンプルをそれぞれの来場時刻によって、午前と午後の２グループに分割する。
【００８８】
２．午前のグループから全日見学者を除く。
【００８９】
３．両グループについて、一時間ごとに滞在時間の平均値と標準偏差を計算する。
【００９０】
４．両グループについて、平均値と標準偏差の時間変化を一次関数で同定する。この際、午後グループには以下のサンプルを追加しておく。
【００９１】
μ_S(T_C) = 0, σ_S(T_C) = 0 ………（１３）
標準偏差の計算は実際の平均による固定値とする。この結果、例えば、滞在時間ｔ_Sの確率分布は以下のパラメータ（単位は秒）で表す正規分布として得られる。
【００９２】
α_AM = -0.3521, α_PM = -0.6696
β_AM = 9752.4, β_PM = 19904.4
ξ_AM = 0, ξ_PM = -0.2233
ζ_AM = 3704.4, ξ_PM = 6508.8 ………（１４）
次会場選択モデルの同定としては、推移確率行列P は、実測したOD 表から、次の手順で導出する。
【００９３】
１．Ｓ列を削除し13×12 次元行列とする。
【００９４】
２．各行について正規化する。
【００９５】
３．行列を転置する。
【００９６】
４．先頭行に1 行挿入し、13×13 次元正方行列とする。挿入する行ベクトルは[0,…,0]である。
【００９７】
処理結果を図１１に示す。
【００９８】
＜滞留モデル＞
各会場における滞在時間は、ある時不変な確率分布にしたがうものとして定式化する。滞在時間の対数変換結果が正規分布にしたがうものと考える。即ち、会場ｉにおける滞在時間Ｔ_Siは次式で与えられる。
【００９９】
【数８】

ｘ_iは次式で定義される確率分布により決める。
【０１００】
【数９】

各会場の滞在時間記録を対数変換し、それぞれについて正規分布にフィッティングした結果の一例を図１２に示す。又、平均値を逆変換すると図１３を得る。
【０１０１】
上記のように作成された来場モデル、移動モデル、滞留モデルを参照し、移動体モデルを構築する。
【０１０２】
（予測シミュレーション）
上記で得られた移動体モデルに基づき、移動イベントを生成し、会場の混雑予測は行われる。以下に、シミュレーションによる会場混雑の再現を行うアルゴリズムを示す。
【０１０３】
1. 午前と午後の来場者数を、それぞれN_AM，N_PM として設定する。
【０１０４】
2. 時刻の初期化t=0（午前9:00）；来場者カウンタの初期化
i=0
3. while (i < N_AM) {
4. i=i+1
5. 来場モデルにより来場者i 発生；時刻t_i=t+発生間隔
6. 退場モデルにより来場者i の退場時刻t_L[i]決定
7. while (t_i < t_L[i]) {
8. 次会場選択モデルにより訪問会場決定，移動
9. 滞留モデルにより滞留時間t_S 決定；t_i=t_i+t_S
10. }
11. 退場処理
12. }
13. 2〜12 の処理を午後について同様に実施；t=10800（正午）
14. 全来場者の行動データを時刻順に並べ替え
15. 集計処理
シミュレーションプログラムはJava（登録商標）等で作成可能である。
【０１０５】
シミュレーションにより得られた歩行者情報、ＯＤ表、滞留時間表等は、移動体状態保持部１５に保存されても良く、シミュレーション結果保持部（図示せず）等に保持されても良い。あるいは、表示部７等に表示されても良い。これらのシミュレーション結果を用いて得られた来場時刻分布の一例を図１４に示す。図１４では、Ｎ_AM＝３５４、Ｎ_PM＝６６０を実際の来場者数として用いている。又、同様に、滞留時間表等のシミュレーション結果を用いて得られた見学時間分布を図１５に示す。又、時間毎の各エリアの混雑推移（滞在者数推移）を図１６に示す。Ｒ、Ｑ、…、Ａは、図１０の展示会場エリアに相当する。更に、シミュレーション結果から得られたＯＤ表を図１７に示す。各数値は、出発会場から到着会場に移動した歩行者の数を示す。これらのシミュレーション結果から、会場の混雑を予測し、集客施設の構造、運用時のレイアウトや案内表示の適切な設計が可能となる。
【０１０６】
（その他の実施の形態）
本発明は上記の実施の形態によって記載したが、この開示の一部をなす論述及び図面はこの発明を限定するものであると理解すべきではない。この開示から当業者には様々な代替実施の形態、実施例及び運用技術が明らかとなろう。
【０１０７】
例えば、本発明の実施の形態において、観測エリアを移動する移動体は、歩行者であると説明したが、これに限らないことは勿論である。
【０１０８】
又、本発明の実施の形態において、外部イベント取込部１２で時刻情報を付加すると説明したが、受信機２０ａ、２０ｂ、…、２０ｅが受信機制御部１３へ情報を送出する際に、時刻情報を付加しても構わない。但し、受信機２０ａ、２０ｂ、…、２０ｅよりも、外部イベント取込部１２で時刻情報を付加するほうが、受信機２０ａ、２０ｂ、…、２０ｅ間の時刻の誤差を考慮する必要がないという利点がある。
【０１０９】
又、本発明の実施の形態において、移動体状態保持部１５、環境情報保持部１７を分けて備えると説明したが、一つの保持部で共用しても構わない。又、本発明の実施の形態に係る混雑予測システムは、状態バッファやイベントキューを保持する状態バッファ保持部やイベントキュー保持部を別個に備えても構わない。このとき、状態バッファ保持部やイベントキュー保持部は、移動体状態保持部１５等と同様に、ＲＡＭ等の内部記憶装置を用いても良く、ＨＤやＦＤ等の外部記憶装置を用いても良い。
【０１１０】
このように、本発明はここでは記載していない様々な実施の形態等を含むことは勿論である。従って、本発明の技術的範囲は上記の説明から妥当な特許請求の範囲に係る発明特定事項によってのみ定められるものである。
【０１１１】
【発明の効果】
本発明によると、予測モデルをリアルタイムで更新し、精度良く、将来の混雑予測を行う混雑予測システム、混雑予測方法及び混雑予測プログラムを提供することができる。
【図面の簡単な説明】
【図１】本発明の実施の形態に係る混雑予測システムの構成ブロック図である。
【図２】本発明の実施に形態に係る発信機に備えられた無線タグの構成ブロック図である。
【図３】本発明の実施の形態に係る発信機情報の一例である。
【図４】本発明の実施の形態に係る移動体情報の一例である。
【図５】本発明の実施の形態に係る観測エリアの一例である。
【図６】本発明の実施の形態に係る動線観測方法のフローチャートである。
【図７】本発明の実施の形態に係る混雑予測方法のフローチャートである。
【図８】本発明の実施の形態に係る混雑予測方法の２つのシミュレーションインスタンスを示す図である。
【図９】本発明の実施の形態に係る移動体モデルの概念図である。
【図１０】展示会場のレイアウトの一例である。
【図１１】本発明の実施の形態に係る移動体モデルを構築する際に用いる推移確率行列である。
【図１２】本発明の実施の形態に係る滞留モデルの同定に用いる滞在時間分布である。
【図１３】本発明の実施の形態に係る滞留モデルの滞留滞在時間平均値を示す図である。
【図１４】本発明の実施の形態に係る混雑予測方法によって得られる来場時刻分布の一例である。
【図１５】本発明の実施の形態に係る混雑予測方法によって得られる滞在時刻分布の一例である。
【図１６】本発明の実施の形態に係る混雑予測方法によって得られる混雑推移の一例である。
【図１７】本発明の実施の形態に係る混雑予測方法によって得られるＯＤ表の一例である。
【符号の説明】
１ 混雑予測システム
５ 記録ファイル
６ 動線記録ファイル
７ 表示部
８ 移動体モデル
９ 入力部
１１ 記録イベント取込部
１２ 外部イベント取込部
１３ 受信機制御部
１４ 入力切替部
１５ 移動体状態保持部
１６ イベント処理部
１７ 環境情報保持部
１８ 模擬イベント発生部
２０ａ、２０ｂ、…、２０ｅ 受信機
２１ 発信機
２２ 歩行者
２５ 通信ネットワーク
２６ 観測エリア
２７ 仕切り
３０ 無線タグ
３１ 識別情報保持部
３２ 送信部
３３ アンテナ[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a congestion prediction system, a congestion prediction method, and a congestion prediction program for predicting the behavior of a moving object in a certain observation area.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, a plurality of receivers are installed in a facility for attracting customers, such as an event venue, and a pedestrian moving in the facility has a transmitter to estimate the position of the pedestrian. Radio waves transmitted from the transmitter are detected by receivers in the surrounding area, and each receiver transmits information obtained from the transmitter wirelessly or by wire to one position detection system. As described above, a position detection system and a position detection method capable of detecting the position of a moving object such as a pedestrian using a mobile communication system are disclosed (for example, see Patent Document 1).
[0003]
On the other hand, in a customer attracting facility, management of pedestrian traffic lines is an important issue. Here, the “flow line” indicates a movement history of the moving object, and can be expressed by a combination of a place and a time. The structure of the attracting facilities, the layout during operation, and the guidance display are designed in consideration of pedestrian traffic lines. Therefore, it is necessary to predict a pedestrian's flow line, and a prediction simulation may be used. As a simulation device, a device that expresses, analyzes, and predicts human collective behavior in a form close to an actual one has been disclosed (for example, see Patent Document 2).
[0004]
[Patent Document 1]
JP-A-11-178041 (pages 5-6, FIG. 1)
[0005]
[Patent Document 2]
JP-A-2000-20499 (page 4-6, FIG. 1)
[0006]
[Problems to be solved by the invention]
However, since the prediction model used in the conventional simulation does not use the raw data of the customer attracting facility, there is a problem in its reliability. In addition, since a new arrival is not taken into account or a simple Poisson arrival model is used even if it is taken into account, it is not possible to take into account changes in arrival patterns depending on the time zone of the customer attraction facility, and it is not possible to make highly accurate predictions.
[0007]
In view of the above problems, an object of the present invention is to provide a congestion prediction system, a congestion prediction method, and a congestion prediction program for updating a prediction model in real time and accurately predicting future congestion.
[0008]
[Means for Solving the Problems]
In order to achieve the above object, a first feature of the present invention is that (a) a transmitter, which is transmitted from a plurality of receivers in an observation area where a mobile body carrying the transmitter moves via a communication network; A receiver control unit that receives the transmitter information including the ID of the transmitter, and (b) the transmitter ID, the ID of the receiver that transmitted the transmitter information, and the arrival of the transmitter in the detection area of the receiver based on the transmitter information. Alternatively, an external event capturing unit that generates a moving event including a detection flag indicating time of departure and time information, and (c) a simulated event that generates a moving event based on a moving object model used to predict the behavior of the moving object A generating unit, (d) an input switching unit for switching and outputting a moving event sent from the external event capturing unit and the simulated event generating unit, and (e) transmitting from the external event capturing unit via the input switching unit. In parallel with extracting moving events, updating the moving object information expressed in the past moving history of the moving object, and estimating the current position of the moving object, input switching is performed when congestion prediction is requested. The present invention provides a congestion prediction system including an event processing unit that performs future congestion prediction using a movement event transmitted from a simulation event generation unit via a unit.
[0009]
According to the congestion prediction system according to the first aspect of the present invention, a moving body model is updated in real time from information obtained by observing a flow line of a moving body moving in a certain observation area. Predict future congestion.
[0010]
In addition, the congestion prediction system according to the first feature is a movement that holds moving object information, an OD table indicating a moving state of the moving object between detection areas, and a stay time table indicating a time of staying in the detection area for each detection area. It may further include a body condition holding unit.
[0011]
Further, the event processing unit of the congestion prediction system according to the first feature is configured such that, when the congestion prediction is requested, the event processing unit temporarily generates a moving object newly arriving at the observation area using the unsteady Poisson process model. And a moving body model may be created. According to the congestion prediction system, highly accurate prediction can be performed in consideration of a change in an arrival pattern depending on a time zone to an observation area.
[0012]
A second feature of the present invention is to receive, via a communication network, transmitter information including a transmitter ID, which is transmitted from a plurality of receivers in an observation area where a mobile body carrying the transmitter moves. In the congestion prediction system, (a) based on the transmitter information, includes a transmitter ID, an ID of a receiver that transmitted the transmitter information, a detection flag indicating arrival or departure of the transmitter from a detection area of the receiver, and time information. (B) updating the moving body information represented by the past moving history of the moving body based on the moving event; and (c) using the moving body information, E) estimating the current position; and (d) creating a mobile model used to predict the behavior of the mobile using the mobile information at the time of the congestion prediction request when congestion prediction is requested. Comprising the steps of, and summarized in that a (e) using the moving events generated based on the mobile body model, congestion prediction method and performing a future congestion prediction.
[0013]
According to the congestion prediction method according to the second feature, the mobile object model is updated in real time from the information obtained by observing the flow line of the mobile object moving in a certain observation area, so that the congestion of the future can be accurately and accurately determined. You can make predictions.
[0014]
The step of creating a moving object model in the congestion prediction method according to the second feature includes the steps of: an OD table indicating a movement state of the moving object between detection areas at the time of the congestion prediction request; A moving object model may be created using a stay time table indicating the time spent in the detection area for each area. According to this congestion prediction method, a mobile body model closer to the actual operation can be created, and congestion prediction can be performed with high accuracy.
[0015]
Further, the step of creating a moving object model in the congestion prediction method according to the second feature includes temporarily generating a moving object newly arriving at the observation area using the unsteady Poisson process model, and creating a moving object model. Is also good. According to the congestion prediction method, highly accurate prediction can be performed in consideration of a change in an arrival pattern depending on a time zone to an observation area.
[0016]
A third feature of the present invention is to receive, via a communication network, transmitter information including a transmitter ID, which is transmitted from a plurality of receivers in an observation area where a mobile body carrying a transmitter moves. The congestion prediction system includes (a) a transmitter ID, an ID of a receiver that transmitted the transmitter information, a detection flag indicating arrival or departure of the transmitter with respect to a detection area of the receiver, and time information based on the transmitter information. (B) updating the moving body information represented by the past moving history of the moving body based on the moving event; and (c) using the moving body information to identify the moving body. A procedure for estimating the current position, and (d) when a congestion prediction is requested, a procedure for creating a moving body model used for predicting the behavior of the moving body using the moving body information at the time of the congestion prediction request And (ho Using a moving event generated based on the mobile body model, and summarized in that a congestion prediction program for executing a procedure for future congestion prediction.
[0017]
The procedure for creating a moving object model in the congestion prediction program according to the third feature includes an OD table indicating a moving state of the moving object between detection areas at the time of the congestion prediction request, and a detection at the time of the congestion prediction request. A moving object model may be created using a stay time table indicating the time spent in the detection area for each area.
[0018]
Further, the procedure for creating a moving object model in the congestion prediction program according to the third feature includes the step of temporarily generating a newly arrived moving object to the observation area using the non-stationary Poisson process model, and creating a moving object model. Is also good.
[0019]
By reading the congestion prediction program according to the third feature of the present invention, it is possible to cause a congestion prediction system or the like to execute the above procedure.
[0020]
BEST MODE FOR CARRYING OUT THE INVENTION
Next, an embodiment of the present invention will be described with reference to the drawings. In the following description of the drawings, the same or similar parts are denoted by the same or similar reference numerals. However, it should be noted that the drawings are schematic.
[0021]
The congestion prediction system according to the embodiment of the present invention updates a prediction model in real time from information obtained by observing a flow line of a moving body such as a pedestrian moving in a certain observation area, and updates future congestion. Make predictions. Also, the arrival of a pedestrian that does not yet exist at the start of the prediction is actually temporarily generated using the non-stationary Poisson process model, and is used for the prediction.
[0022]
(Receiver and transmitter in observation area)
Hereinafter, a system will be described in which a pedestrian moving in a certain observation area is taken as an example of a moving body, and a flow line of the pedestrian is observed to predict a future congestion. The congestion prediction system 1 according to the embodiment of the present invention observes the flow line of the pedestrian 22 in a certain observation area 26 as shown in FIG. In the observation area 26, receivers 20a, 20b,..., 20e are installed for each of a plurality of areas whose positions are to be detected. On the other hand, the pedestrian 22 to be observed carries the transmitter 21.
[0023]
Specifically, this transmitter 21 can use a wireless tag or the like. The wireless tag includes, for example, a so-called RF-ID (Radio Frequency Identification) or the like. As illustrated in FIG. 2, the identification information holding unit 31 holds a different ID for each wireless tag 30 and the identification information holding unit 31 holds the ID. A transmission unit 32 that generates a transmission signal of a predetermined frequency according to the ID being provided, and a transmission antenna 33 are provided. The transmission unit 32 generates a transmission signal corresponding to the ID stored in the identification information storage unit 31 and transmits the transmission signal via the transmission antenna 33.
[0024]
The interval at which the wireless tag 30 transmits the signal corresponding to the ID is determined according to the moving speed of the pedestrian 22 and the arrangement interval of the receivers 20a, 20b,..., 20e. In this case, the movement is performed once every few seconds. Alternatively, when the user walks in the hall, the position of the pedestrian 22 is detected by narrowing the arrangement interval of the receivers 20a, 20b,..., 20e and shortening the interval of transmitting the signal corresponding to the ID. Accuracy can be improved. The transmission interval can be adjusted by controlling the interval at which the transmission unit 32 transmits a signal corresponding to the ID.
[0025]
On the other hand, each of the receivers 20a, 20b,..., 20e also has a unique ID. The receivers 20a, 20b,..., 20e detect radio waves including an ID unique to the transmitter from the transmitter 21. The range in which the receivers 20a, 20b,..., 20e detect radio waves can be adjusted, and the range of the detection area, that is, the area A, the area B,. The receivers 20a, 20b,..., 20e send information on the detection of the transmitter 21 to the receiver control unit 13 in the congestion prediction system 1 via the communication network 25.
[0026]
Next, a method of processing information received by the receivers 20a, 20b,..., 20e from the transmitter 21 will be described. The receivers 20a, 20b,..., 20e hold a list of information on transmitters (hereinafter, “transmitter information”). As shown in FIG. 3, the transmitter information includes a transmitter ID (tag ID), a flag indicating whether or not the detection is currently being performed, and a counter indicating the elapsed time from the time when the detection flag is changed.
[0027]
When the receivers 20a, 20b,..., 20e detect the radio wave of the transmitter 21, they search the transmitter information list using the transmitter ID as a key, and if they are found, update the transmitter information. Create new transmitter information for For example, when the radio wave of the transmitter 21 of the transmitter ID “XXX” is detected and the transmitter information is not in the list, as shown in FIG. 3, the detection flag of the transmitter ID “XXX” is set to true, Is set to 0, a new list of the transmitter ID “XXX” is created. Thereafter, the counter is set to 0 each time the transmitter information of the transmitter ID "XXX" is detected, and when the detection flag is false, the detection flag is changed to true. The receivers 20a, 20b,..., 20e check the table of FIG. 3 at predetermined time intervals t, and add 1 to the counter for all transmitters. The receivers 20a, 20b,..., 20e check the counter of the transmitter whose detection flag is true in the table of FIG. 3, and if the counter value exceeds a predetermined value cth, the detection flag is false. And set the counter to 0. For example, the transmitter ID “YYY” in FIG. 3 indicates a state immediately after the detection flag is changed to “false”, while the transmitter ID “ZZZ” indicates a state in which 15 steps have elapsed since the detection flag was changed to “false”. Indicates that it has passed. That is, when the detection flag is true, it is understood that the transmitter 21 corresponding to the transmitter ID has been detected in the corresponding area. When the detection flag is false, it is understood that the transmitter 21 corresponding to the transmitter ID has not been detected for the time indicated by the counter value in the corresponding area.
[0028]
Next, the timing at which the receivers 20a, 20b,..., 20e send information to the receiver control unit 13 in the congestion prediction system 1 via the network 25 will be described. The receivers 20a, 20b,..., 20e transmit the transmitter information and the ID of the receiver that transmitted the transmitter information to the receiver controller 13 as “receiver information” in the following two cases.
[0029]
1. When the detection flag changes to true
2. When the detection flag changes to false
The point in time when the detection flag changes to true indicates a point in time when the pedestrian 22 arrives in a certain area. The point in time when the detection flag changes to false indicates the point in time when the pedestrian 22 has left from a certain area. The time when the detection flag changes to false is the time when the counter value in the true state has passed the predetermined value cth, so that the pedestrian 22 has definitely left the area, not a change in the detection area or a mixed radio wave. It is the time when it can be considered. By this processing, the arrival of the pedestrian to a certain area and the departure from the same area are notified to the congestion prediction system 1. By integrating information from all areas, the movement history of the pedestrian 22 in the observation area 26 can be obtained.
[0030]
(Configuration of congestion prediction system)
As shown in FIG. 1, a congestion prediction system 1 according to an embodiment of the present invention includes a plurality of receivers 20a in an observation area 26 where a pedestrian 22 carrying a transmitter 21 moves via a communication network 25. , 20b,..., 20e, a receiver control unit 13 for receiving the transmitter information, an external event capturing unit 12 for generating a mobile event based on the transmitter information, and for predicting the behavior of the mobile object. A simulated event generating unit 18 for generating a moving event based on the moving object model used for the moving object model; an input switching unit 14 for switching and outputting a moving event sent from the external event capturing unit 12 and the simulated event generating unit 18; An event processing unit 16 that updates the body information and estimates the future congestion in parallel with estimating the current position of the pedestrian is provided. Further, the congestion prediction system 1 according to the embodiment of the present invention includes a recorded event capturing unit 11, a moving object state storage unit 15, and an environment information storage unit 17.
[0031]
The receiver control unit 13 controls the operation of the receivers 20a, 20b,..., 20e in the observation area 26 via the communication network 25, and receives the receiver information transmitted from the receivers 20a, 20b,. Is sent to the external event capturing unit 12. What added ID of the receiver which transmitted transmitter information to transmitter information is called "receiver information".
[0032]
The external event capturing unit 12 adds time information to the receiver information transmitted from the receiver control unit 13 and delivers the information to the input switching unit 14. Here, the information obtained by adding the time information to the receiver information is referred to as a “movement event”.
[0033]
The recording event capturing unit 11 reads the movement event described in the recording file 5 and sends the read event to the input switching unit 14. The recording file 5 is a file in which a past movement event is recorded, and may be stored in a recording medium such as an FD. While the external event capturing unit 12 captures current information in the observation area 26, the recorded event capturing unit 11 captures past information in the observation area 26.
[0034]
The simulated event generation unit 18 generates a movement event based on the moving object model 8 and sends it to the input switching unit 14. The moving body model 8 is a prediction model used to predict the behavior of a pedestrian (moving body) in the observation area 26. The method of constructing the moving object model 8 will be described later in detail.
[0035]
The input switching unit 14 switches the movement event transmitted from the external event capturing unit 12, the recorded event capturing unit 11, and the simulation event generating unit 18 according to the mode, and transmits the event to the event processing unit 16. The congestion prediction system 1 has the following three operation modes.
[0036]
1. Monitor mode: Observes the current situation in the observation area. The movement event sent from the external event capturing unit 12 is used.
[0037]
2. Reproduction mode: Observes the situation at that time from information recorded in the observation area in the past. The movement event sent from the recording event capturing unit 11 is used.
[0038]
3. Simulation mode: A state of a virtual pedestrian in a virtual environment is predicted using a prediction model. The movement event sent from the simulation event generator 18 is used. However, even in the simulation mode, observation in the current observation area is continued in parallel.
[0039]
Switching of the mode is performed via a user interface such as the input unit 9. The input unit 9 specifically refers to devices such as a keyboard and a mouse. When an input operation is performed from the input unit 9, the corresponding key information is transmitted to the input switching unit 14.
[0040]
The moving body state holding unit 15 holds the current state and the flow line of all the pedestrians 22. Specifically, it includes a moving object information and a status buffer indicating the current position of the moving object shown in FIG. For example, as shown in FIG. 5, a detection area A by the receiver 20a, a detection area B by the receiver 20b, a detection area C by the receiver 20c, a detection area D by the receiver 20d, a detection area E by the receiver 20e, reception It is assumed that the pedestrian has moved in the observation area 26 having the detection area F by the aircraft 20f. At this time, the mobile unit state holding unit 15 holds mobile unit information for each transmitter ID from information obtained from each of the receivers 20a, 20b,..., 20f. In FIG. 4, the target transmitter 21 arrives at area A at 15:30 and stays for 30 minutes, then arrives at area B at 16:05 and stays for 10 minutes, then area C and area E. , Area F.
[0041]
Furthermore, the moving object state holding unit 15 holds an OD table indicating a pedestrian's movement state between areas, a staying time table indicating the time spent in each area, and the like. The OD table is, for example, an N × N dimensional matrix having integers as elements, where N is the number of stay areas. If i and j are integers from 1 to N, the i-th row and j-th column element is the cumulative number of people who have moved from the stay area i to the stay area j. Each time a pedestrian movement event occurs, the corresponding element is incremented by one. The residence time table is composed of, for example, three N-dimensional vectors or a 3 × N-dimensional matrix. The three N-dimensional vectors are referred to as a parameter vector, a cumulative vector, and a squared cumulative vector, and a new residence time in the retention area i is x_i, The i-th element n of the parameter vector_i, The i-th element a of the cumulative vector_i, The ith element of the squared cumulative vector_iIs expressed as follows.
[0042]
n_i= n_i + 1
a_i= a_i + ln (x_i + 1)
s_i= s_i + (ln (x_i + 1)}^Two
Only events that satisfy the same conditions as those in the OD table are recorded.
[0043]
The moving object state holding unit 15 may use an internal storage device such as a RAM or an external storage device such as an HD or FD.
[0044]
The environment information holding unit 17 holds environment information regarding the observation area 26. For example, the positional information, the distance, and the passage information between the detection areas are held. In the case of the observation area 26 as shown in FIG. 5, information that a partition 27 exists between the area A and the area C is held. If a recommended travel route is presented in the observation area 26, the route is maintained. Further, a route which is often passed as a whole from the flow lines of many transmitters is maintained. Using these pieces of environment information, the event processing unit 16 updates the information of the moving object state holding unit 15. The environment information holding unit 17 may use an internal storage device such as a RAM or the like, or may use an external storage device such as an HD or FD, similarly to the moving object state holding unit 15.
[0045]
The event processing unit 16 updates the information held in the moving object state holding unit 15 from the movement event, and estimates the position of the pedestrian. The environment information held in the environment information holding unit 17 is used for this update process. The event processing unit 16 includes an event queue (queue) and a handler, as shown in FIG. The movement events supplied from the input switching unit 14 are sequentially stored at the end of the event queue. When processing the movement event, the handler takes out the movement event from the head of the event queue and performs the processing sequentially. Then, each time a moving event is processed, the moving body information, the OD table, the staying time table and the like held in the moving body state holding unit 15 are created and updated.
[0046]
The flow line for each pedestrian obtained by the event processing unit 16 is output to the flow line recording file 6 or the display unit 7. Here, the flow line recording file 6 is a file in which a pedestrian's flow line, that is, a pedestrian's movement history that can be expressed by a combination of location and time is recorded. The display unit 7 refers to a screen such as a monitor, and a liquid crystal display (LCD), a light emitting diode (LED) panel, an electroluminescence (EL) panel, or the like can be used. The display unit 7 may be a printer or the like.
[0047]
Further, when the congestion prediction is requested, the event processing unit 16 uses the moving body information, the OD table, the residence time table, and the like held in the moving body state holding unit 15 at the time of the request as initial values, and uses the prediction model and Is updated in real time. Then, using the movement event generated by the updated moving body model 8, a future congestion is predicted. In addition, when the moving object model 8 is generated, a new arrival is tentatively generated using the non-stationary Poisson process model. Therefore, a change in the arrival pattern due to the time zone to the observation area is considered, and a highly accurate prediction is made. be able to.
[0048]
Although not shown, the congestion prediction system according to the embodiment of the present invention may include a program holding unit for storing a program for updating a prediction model or executing a congestion prediction. The program holding unit may use an internal storage device such as a RAM or an external storage device such as an HD or FD.
[0049]
According to the congestion prediction system according to the embodiment of the present invention, the moving body model is updated in real time from information obtained by observing the trajectory of a pedestrian moving in a certain observation area, so that it is accurate and future Can be predicted.
[0050]
(Traffic line observation method)
Next, a flow line observation method will be described with reference to FIG. The event processing unit 16 sequentially processes movement events received from the input switching unit 14 and generates a flow line for each transmitter 21.
[0051]
(A) First, in step S101, the event processing unit 16 extracts one moving event from the head of the event queue.
[0052]
(B) Next, in step S102, the mobile body information and state buffer corresponding to the transmitter ID of the mobile event are searched from the mobile body information and state buffer held in the mobile body state holding unit 15.
[0053]
(C) Next, in step S103, if the mobile unit information and the state buffer cannot be searched, the process proceeds to step S104, where the mobile unit information and the state buffer for the transmitter ID are newly generated and newly registered. .
[0054]
(D) Next, in step S105, based on the movement event, the state buffer is updated in consideration of the environment information held in the environment information holding unit 17. For example, if the determination process is performed after narrowing down the possible routes between the two areas by using the positional relationship, the distance, and the passage information between the areas, it is not necessary to estimate a movement that is impossible in reality. Also, in an environment where the recommended movement route is presented, the possibility of moving in the direction opposite to the direction of the route is considered to be lower than in the forward direction. Position estimation is possible. Furthermore, a more reliable existence position can be estimated by comparing the flow path of the transmitter to be estimated with the flow path that often passes from the flow line of another transmitter as a whole. To perform “more probable” estimation means that when it is not possible to determine which area is correct, one area is stochastically selected, and the probability of using that area is determined by the route and flow line. That is. Further, it creates and updates the OD table and the residence time table held in the moving object state holding unit 15. Alternatively, another device such as a CCD camera may be installed in the detection area, and the update processing may be performed in consideration of the detection result.
[0055]
(E) Next, in step S106, the position of the transmitter, that is, the area where the pedestrian is present is estimated based on the updated state buffer. Then, the estimated presence area is stored in the area portion or the like of the mobile object information.
[0056]
(F) Next, in step S107, if the status buffer or the existing area has changed in steps S105 and S106, the process proceeds to step S108, and the flow line recording file 6 is output. Alternatively, flow line information is displayed on the display unit 7. If there is no change, the process returns to step S101, and the above processing is repeated.
[0057]
According to the flow line observing method according to the embodiment of the present invention, the position of a pedestrian can be accurately detected and the pedestrian's flow line can be observed without interruption.
[0058]
(Congestion prediction method)
Next, a congestion prediction method will be described with reference to FIGS. When a congestion prediction request is made, the congestion prediction system 1 according to the embodiment of the present invention switches to the simulation mode. The event processing unit 16 copies the event queue and handler at the time of the prediction request, the moving object information of the moving object state holding unit 15, the OD table, the residence time table, and the like, and newly generates a simulation instance for prediction. The state in which the flow line observation is being performed exists as it is as a monitoring instance, and the flow line observation is continued. That is, the event processing unit 16 and the moving object state holding unit 15 have two instances virtually. Here, “instance” indicates one state in which a certain process is performed.
[0059]
(A) First, in step S201 in FIG. 7, a request for predicting future congestion is transmitted to the event processing unit 16. This prediction request may be transmitted to the event processing unit 16 via the input switching unit 14 using the input unit 9 shown in FIG. At this time, how much time has elapsed since the present time is predicted is transmitted to the event processing unit 16 through the input unit 9 or the like. When a prediction request is made, the event processing unit 16 generates a new simulation instance for prediction based on the current state, as shown in FIG. In the simulation mode, the input switching unit 14 supplies the movement event from the external event acquisition unit 12 to the monitoring instance, and supplies the movement event from the simulation event generation unit 18 to the prediction simulation instance.
[0060]
(B) In step S202, the monitor instance continues the flow line observation process. That is, according to the procedure shown in the flowchart of FIG. 6, as shown in FIG. 8, the movement event sent from the input switching unit 14 is stored in the event queue, and the handler processes the movement event. Then, for each moving event, the moving body information, the OD table, the staying time table and the like of the moving body state holding unit 15 are updated.
[0061]
(C) On the other hand, in step S203, the simulation instance for prediction copies the moving body information, the OD table, and the residence time table at the time of the prediction request, and calculates the parameters of the moving body model 8 using these as initial values.
[0062]
(D) Next, in step S204, the mobile object model 8 is constructed. At this time, the arrival of a pedestrian that does not yet exist at the start of the prediction is actually temporarily generated using the non-stationary Poisson process model. By using the non-stationary Poisson process model, it is possible to make a highly accurate prediction in consideration of the change in the arrival pattern due to the time zone to the observation area. The method of constructing the moving object model 8 will be described later in detail. The simulation instance for prediction sends the constructed mobile object model 8 to the simulation event generation unit 18. The simulated event generation unit 18 generates a movement event based on the moving object model 8 and sends it to the input switching unit 14. The input switching unit 14 supplies the movement event to the simulation instance for prediction.
[0063]
(E) Next, in step S205, the simulation instance for prediction performs a prediction simulation after a lapse of a specified time based on the movement event. This prediction simulation will be described later in detail.
[0064]
According to the congestion prediction method according to the embodiment of the present invention, the moving body model is updated in real time from the information obtained by observing the flow line of the pedestrian moving in a certain observation area, so that it is accurate and future Can be predicted.
[0065]
(How to build a mobile model)
Here, a method of constructing a moving object model based on the moving object information, the OD table, the residence time table, and the like will be described.
[0066]
The moving object model used in the congestion prediction system according to the embodiment of the present invention includes the following sub-models.
[0067]
・ Visiting model: Determines the arrival time of pedestrians
・ Movement model: Select the venue to visit
・ Residence model: Determines the tour time at each venue
FIG. 9 shows a schematic diagram of the moving object model. Pedestrians arrive according to the arrival model, select and move to a venue according to the movement model until the time of leaving, and stay at each venue to be observed according to the retention model. Hereinafter, an example of constructing each model will be described. Here, the flow line of the pedestrian when observing the exhibition hall as shown in FIG. 10 is predicted.
[0068]
<Visiting model>
Since it is inappropriate to assume that the visitor's visitor density is constant, a non-stationary Poisson process (NHPP) is introduced into the formulation. Details of NHPP are described in "DP Heyman, MJ Sobel, Eds .:" Stochastic Models ", Handbooks in Operations Research and Management Science, Vol. 2, Elsevier Science Publishers BV (North-Holland), 1990." . According to the NHPP, a Poisson arrival model having a time-varying visit density can be created. The next visit probability distribution at time T is given by the following equation, where λ is the visit density.
[0069]
(Equation 1)

The time from the time T to the next visit time may be obtained by solving this equation for t. Equation (1) is modified to introduce a uniform random number u to obtain the following equation.
[0070]
(Equation 2)

The visits are divided into morning and afternoon, and each visit time distribution is approximated by a normal distribution. According to this, the visit density λ (t) is represented by the following equation, where μ is the average visit time, σ is the standard deviation, AM is AM, and PM is PM.
[0071]
(Equation 3)

For the identification of the visit model, for example, the following visit model is obtained by fitting a normal distribution to the actual visit time distribution in the morning and afternoon. The average value indicates the elapsed time when 9:00 is set to 0 second.
[0072]
μ_AM= 3855, σ_AM= 2703
μ_PM= 18838, σ_PM= 3573 ……… (4)
To estimate the visitor's arrival time, the total number of visitors in the morning N_AMAnd N in the afternoon_PMIs required, but will be given separately.
[0073]
<Moving model>
The visitor's movement model is formulated by combining the following two models.
[0074]
・ Exit model
・ Next venue selection model
That is, when a visitor finishes a tour at a certain venue, it is first determined whether or not to leave using a leaving model. If it is decided not to leave, the next venue to be visited is determined using the next venue selection model.
[0075]
The exit model uses the exit time as t_L, Visit time_A, Stay time t_S, T_CIs defined by the following equation.
[0076]
(Equation 4)

D_LIf (t) is true, exit is selected. In the case of false, the next destination venue is determined by the next venue selection model. The stay time is determined by a probability distribution (normal distribution) defined by the following equation.
[0077]
(Equation 5)

Average value μ_SAnd standard deviation σ_SIs the visitor's arrival time t_AIs given as a linear function of. This expresses the tendency of visitors in the morning to complete the tour in the morning and visitors in the afternoon to close by the closing time.
[0078]
(Equation 6)

Exit time t_LDoes not depend on the state of the visitors.
[0079]
The next venue selection model is formulated using a Markov chain model. Therefore, the model does not consider the past visit history. The details of the Markov chain model are described in Shuji Kaneko, "Simulation of the flow of people in an exhibition", Advance, No. 6, pp. 33-37, Toshiba Advanced System, 1998.
[0080]
X represents a Markov chain representing each visitor's movement between venues_kAnd Assume that the reception for the visit is one venue, and the number of exhibition venues including the reception is M. The reception is called venue 1 and the other venues are called venues 2 to M. The number of visitor states is M. Now, the state of staying at venue i is b_iAnd the probability distribution of staying at the venue i at the time point k is written as the following equation.
[0081]
(Equation 7)

This can be written as an M-dimensional vector as follows.
[0082]
p [k] = [p₁[k], p_Two[k],…, p_M[k]] ……… (9)
The initial distribution of all visitors is always p [0] = [1,0, ..., 0]. The transition probability matrix is represented by an M × M-dimensional matrix P. The i-th row and j-th column element of the transition probability matrix P is the transition probability from the venue i to the venue j.
[0083]
p_ij = P (X_{k + 1}= b_j | X_k = b_i) ……… (10)
The probability distribution of each visitor at time k is represented by the following equation.
[0084]
p [k] = p [0] P^k                         ……… (11)
In this model, the transition probability matrix P is regarded as invariable during the target exhibition.
[0085]
As the identification of the leaving model, the closing time T_CIs 17:00, and the number of seconds from the venue time 9:00 is as follows.
[0086]
T_C = 28800 ……… (12)
Stay time t_SIs obtained as follows.
[0087]
1. The samples are divided into two groups, morning and afternoon, according to their arrival times.
[0088]
2. Exclude all-day visitors from the morning group.
[0089]
3. For both groups, calculate the average value and standard deviation of the stay time every hour.
[0090]
4. For both groups, the time change of the mean value and the standard deviation is identified by a linear function. At this time, the following samples are added to the afternoon group.
[0091]
μ_S(T_C) = 0, σ_S(T_C) = 0 (13)
The calculation of the standard deviation is a fixed value based on the actual average. As a result, for example, the stay time t_SIs obtained as a normal distribution represented by the following parameters (unit is seconds).
[0092]
α_AM = -0.3521, α_PM = -0.6696
β_AM = 9752.4, β_PM = 19904.4
ξ_AM = 0, ξ_PM = -0.2233
ζ_AM = 3704.4, ξ_PM = 6508.8 ……… (14)
To identify the next venue selection model, the transition probability matrix P is derived from the actually measured OD table in the following procedure.
[0093]
1. The S column is deleted to make a 13 × 12 dimensional matrix.
[0094]
2. Normalize each row.
[0095]
3. Transpose the matrix.
[0096]
4. One row is inserted in the first row to make a 13 × 13-dimensional square matrix. The row vector to be inserted is [0, ..., 0].
[0097]
FIG. 11 shows the processing result.
[0098]
<Residence model>
The staying time at each venue is formalized as following a certain probability distribution. It is assumed that the logarithmic conversion result of the stay time follows a normal distribution. That is, the stay time T at the venue i_SiIs given by the following equation.
[0099]
(Equation 8)

x_iIs determined by the probability distribution defined by the following equation.
[0100]
(Equation 9)

FIG. 12 shows an example of the result of logarithmic conversion of the staying time record at each venue and fitting to a normal distribution for each. FIG. 13 is obtained by inversely converting the average value.
[0101]
With reference to the visit model, the movement model, and the staying model created as described above, a moving object model is constructed.
[0102]
(Predictive simulation)
Based on the moving object model obtained above, a moving event is generated, and the congestion of the venue is predicted. The algorithm for reproducing the congestion of the venue by simulation is shown below.
[0103]
1. Set the number of visitors in the morning and afternoon as N_AM and N_PM, respectively.
[0104]
2. Time initialization t = 0 (9:00 am); visitor counter initialization
i = 0
3.while (i <N_AM) {
4. i = i + 1
5. Visitor i occurs according to the visit model; time t_i = t + occurrence interval
6. Departure time t_L [i] of visitor i is determined by the exit model
7. while (t_i <t_L [i]) {
8. Determine the visiting venue by the next venue selection model and move
9. Determination of residence time t_S by residence model; t_i = t_i + t_S
Ten. }
11. Exit processing
12.}
13. Perform steps 2 to 12 in the same way in the afternoon; t = 10800 (noon)
14. Sort the behavior data of all visitors in chronological order
15. Aggregation processing
The simulation program can be created by Java (registered trademark) or the like.
[0105]
The pedestrian information, the OD table, the residence time table, and the like obtained by the simulation may be stored in the moving object state storage unit 15, or may be stored in a simulation result storage unit (not shown). Alternatively, it may be displayed on the display unit 7 or the like. FIG. 14 shows an example of the visit time distribution obtained by using these simulation results. In FIG. 14, N_AM= 354, N_PM= 660 is used as the actual number of visitors. Similarly, FIG. 15 shows a tour time distribution obtained using simulation results such as a residence time table. FIG. 16 shows a change in congestion (change in the number of visitors) in each area at each time. R, Q,..., A correspond to the exhibition hall area in FIG. FIG. 17 shows an OD table obtained from the simulation results. Each numerical value indicates the number of pedestrians who moved from the departure venue to the arrival venue. From these simulation results, it is possible to predict congestion in the venue and to appropriately design the structure of the customer attracting facility, the layout at the time of operation, and the guidance display.
[0106]
(Other embodiments)
Although the present invention has been described with the above embodiments, it should not be understood that the description and drawings forming part of this disclosure limit the present invention. From this disclosure, various alternative embodiments, examples, and operation techniques will be apparent to those skilled in the art.
[0107]
For example, in the embodiment of the present invention, the moving body moving in the observation area is described as a pedestrian, but is not limited to this.
[0108]
Also, in the embodiment of the present invention, it has been described that the time information is added by the external event capturing unit 12. However, when the receivers 20 a, 20 b,. Information may be added. However, when the time information is added by the external event capturing unit 12, it is not necessary to consider the time error between the receivers 20a, 20b,..., 20e than the receivers 20a, 20b,. There is.
[0109]
Further, in the embodiment of the present invention, it has been described that the moving object state holding unit 15 and the environment information holding unit 17 are separately provided, but may be shared by one holding unit. Further, the congestion prediction system according to the embodiment of the present invention may separately include a state buffer holding unit and an event queue holding unit for holding a state buffer and an event queue. At this time, the state buffer holding unit and the event queue holding unit may use an internal storage device such as a RAM or the like, or may use an external storage device such as an HD or FD, similarly to the moving object state holding unit 15 or the like. .
[0110]
As described above, the present invention naturally includes various embodiments and the like not described herein. Therefore, the technical scope of the present invention is determined only by the invention specifying matters according to the claims that are appropriate from the above description.
[0111]
【The invention's effect】
According to the present invention, it is possible to provide a congestion prediction system, a congestion prediction method, and a congestion prediction program for updating a prediction model in real time and accurately predicting future congestion.
[Brief description of the drawings]
FIG. 1 is a configuration block diagram of a congestion prediction system according to an embodiment of the present invention.
FIG. 2 is a configuration block diagram of a wireless tag provided in the transmitter according to the embodiment of the present invention.
FIG. 3 is an example of transmitter information according to the embodiment of the present invention.
FIG. 4 is an example of moving object information according to the embodiment of the present invention.
FIG. 5 is an example of an observation area according to the embodiment of the present invention.
FIG. 6 is a flowchart of a flow line observing method according to the embodiment of the present invention.
FIG. 7 is a flowchart of a congestion prediction method according to the embodiment of the present invention.
FIG. 8 is a diagram showing two simulation instances of the congestion prediction method according to the embodiment of the present invention.
FIG. 9 is a conceptual diagram of a moving object model according to the embodiment of the present invention.
FIG. 10 is an example of a layout of an exhibition hall.
FIG. 11 is a transition probability matrix used when constructing the moving object model according to the embodiment of the present invention.
FIG. 12 is a stay time distribution used for identifying a staying model according to the embodiment of the present invention.
FIG. 13 is a diagram showing an average staying time of the staying model according to the embodiment of the present invention.
FIG. 14 is an example of a visit time distribution obtained by the congestion prediction method according to the embodiment of the present invention.
FIG. 15 is an example of a stay time distribution obtained by the congestion prediction method according to the embodiment of the present invention.
FIG. 16 is an example of a congestion transition obtained by the congestion prediction method according to the embodiment of the present invention.
FIG. 17 is an example of an OD table obtained by the congestion prediction method according to the embodiment of the present invention.
[Explanation of symbols]
1 Congestion prediction system
5 Record files
6 Traffic flow record file
7 Display
8 Moving object model
9 Input section
11 Recorded event capture unit
12 External event capture unit
13 Receiver control unit
14 Input switching unit
15 Moving body state holding unit
16 Event processing part
17 Environmental Information Storage Department
18 Simulated event generator
20a, 20b, ..., 20e Receiver
21 transmitter
22 pedestrian
25 Communication Network
26 Observation area
27 Partition
30 wireless tags
31 Identification information holding unit
32 transmission unit
33 Antenna

Claims (9)

Via a communication network, a receiver control unit that receives transmitter information including an ID of the transmitter, which is transmitted from a plurality of receivers in an observation area in which a moving object carrying the transmitter moves,
Based on the transmitter information, a mobile event including the transmitter ID, the ID of the receiver that transmitted the transmitter information, a detection flag indicating the arrival or departure of the transmitter with respect to the detection area of the receiver, and time information including the time information. An external event capture unit to generate,
A simulated event generator that generates the movement event based on a moving body model used to predict the behavior of the moving body,
An input switching unit that switches and outputs the movement event sent from the external event capture unit and the simulation event generation unit,
The moving events sent from the external event capturing unit are sequentially taken out via the input switching unit, and the moving object information represented by the past moving history of the moving object is updated, and the current position of the moving object is estimated. In parallel with performing a congestion prediction, an event processing unit that performs a future congestion prediction using a movement event sent from the simulation event generating unit via the input switching unit when a congestion prediction is requested. A congestion prediction system characterized by the following. The apparatus further includes a moving body state holding unit that holds the moving body information, an OD table indicating a moving state of the moving body between the detection areas, and a staying time table indicating a time of staying in the detection area for each of the detection areas. The congestion prediction system according to claim 1, wherein: The event processing unit, when requested to predict congestion, tentatively generates a moving object newly arriving at the observation area using a non-stationary Poisson process model, and creates the moving object model. Item 3. The congestion prediction system according to item 1 or 2. In a congestion prediction system that receives transmitter information including an ID of the transmitter, which is transmitted from a plurality of receivers in an observation area where a mobile body carrying the transmitter moves via a communication network,
Based on the transmitter information, a mobile event including the transmitter ID, the ID of the receiver that transmitted the transmitter information, a detection flag indicating the arrival or departure of the transmitter with respect to the detection area of the receiver, and time information including the time information. Sequentially taking out;
Based on the movement event, updating the moving body information expressed in the past movement history of the moving body,
Estimating a current position of the moving object using the moving object information;
When a congestion prediction is requested, using the mobile information at the time of the congestion prediction request, creating a mobile model used to predict the behavior of the mobile,
Performing a future congestion prediction using a movement event generated based on the moving object model. The step of creating the moving object model includes: an OD table indicating a moving state of the moving object between the detection areas at the time of the congestion prediction request; 5. The congestion prediction method according to claim 4, wherein the mobile object model is created by using a residence time table indicating the time spent. 6. The method according to claim 4, wherein the step of creating the moving object model includes temporarily generating a moving object newly arriving at the observation area using an unsteady Poisson process model, and creating the moving object model. Congestion prediction method described in 1. Via a communication network, transmitted from a plurality of receivers in the observation area where the mobile body carrying the transmitter moves, the congestion prediction system to receive the transmitter information including the ID of the transmitter,
Based on the transmitter information, a mobile event including the transmitter ID, the ID of the receiver that transmitted the transmitter information, a detection flag indicating the arrival or departure of the transmitter with respect to the detection area of the receiver, and time information including the time information. Steps to take out sequentially
Based on the movement event, a procedure for updating moving body information expressed in a past movement history of the moving body,
A procedure of estimating a current position of the moving body using the moving body information;
When congestion prediction is requested, using the mobile information at the time of congestion prediction request, a procedure to create a mobile model used to predict the behavior of the mobile,
A congestion prediction program for performing prediction of future congestion using a movement event generated based on the moving object model. The procedure for creating the moving object model includes: an OD table indicating a moving state of the moving object between the detection areas at the time of the congestion prediction request; 8. The congestion prediction program according to claim 7, wherein the mobile object model is created by using a residence time table indicating the time taken. 9. The method according to claim 7, wherein the step of creating the moving object model includes temporarily generating a moving object newly arriving at the observation area using a non-stationary Poisson process model, and creating the moving object model. Congestion prediction program described in 1.

Images