JPH08235483A - Time series data predicting device - Google Patents

Abstract

PURPOSE: To obtain the future predictive value of time series data with high accuracy as well as the predictive value of another time series data obtained by using such predictive value by storing the time series data and predicting the future data of a propagation destination sensor by propagating sensor data relating to the time series data. CONSTITUTION: A propagation origin/destination sensor selection part 13 selects a sensor trying to predict future vehicle occupancy ratio data out of all sensors 11a-11c installed at target road blocks as the propagation destination sensor, and a sensor that becomes the propagation origin sensor of the occupancy ratio data as the propagation origin sensor, respectively. In such a case, it is desirable to select the propagation origin and destination sensors with high correlation in the occupancy ratio data. Sensor data prediction parts 14a-14c take out time series occupancy ratio data for two sensors selected by the propagation origin/destination sensor selection part 13 from a sensor data storage part 12, and predict the future occupancy ratio data of the propagation destination sensor.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a time-series data prediction method, and in particular, for monitoring and controlling vehicle traffic flow on roads and road networks, or for providing information to drivers, vehicle traffic on roads and road networks. The present invention relates to a method and a device for predicting a flow state and travel time of a target section.

[0002]

2. Description of the Related Art For predicting the state of automobile traffic flow on roads and road networks and the travel time of a specific section, information from vehicle detectors is used, for example, as disclosed in Japanese Patent Laid-Open No. 5-114095. There is a method that predicts the traffic volume, detection time, and the mixture rate of large vehicles by an autoregressive model based on the above, calculates the speed using those predicted values, and calculates the travel time of the section using this speed. It was

[0003]

However, the above-mentioned conventional example has the following drawbacks. That is, in the above-mentioned conventional example, the predicted values of the traffic volume, the detection time, and the mixture rate of large vehicles are obtained by using the autoregressive model. In this case, it becomes difficult to determine the parameters in a stable manner, and even if the parameters are determined, there may be variations in the accuracy of the traffic volume, the detection time, the predicted value of the mixture ratio of large vehicles, and the predicted value of the section travel time. Will be more likely.

Therefore, the present invention accurately predicts the traffic volume, the time occupancy rate of the vehicle, or the average speed based on the information obtained from a plurality of traffic flow sensors installed in the target road section, and the predicted value thereof. It is an object of the present invention to provide a time-series data prediction device capable of accurately predicting time-series data such as a section travel time obtained by using the.

[0005]

As a means for solving the above-mentioned problems, a time-series data predicting apparatus of the present invention stores sensors for measuring time-series data installed at a plurality of places and data obtained by the sensors. Data storage means for
Means for selecting a propagation source sensor as a source of propagation of time series data and a propagation destination sensor as a destination of propagation, and predicting future data of the propagation destination sensor using the propagation source sensor data and the propagation destination sensor data And sensor data predicting means.

Furthermore, the time-series data prediction device of the present invention is
As means for solving the above-mentioned problems, in the sensor data predicting means, means for calculating a delay time when the data of the propagation source sensor is similar to data of the propagation destination sensor, and propagation source sensor data by the amount of the delay time Error by comparing means for shifting to the future, data dividing means for dividing the shifted data into past and future data, and past data of the divided data and the propagation destination sensor data from the data storage means And a future data correction means for correcting the future data by using the error and the future data divided by the dividing means to obtain a predicted value of the future data of the propagation destination sensor.

[0007]

In the time-series data device of the present invention configured as described above, the sensor data having a high correlation with each other is propagated, and further, the accurate correction is made by the past performance value, so that the future time with high accuracy can be obtained. Series data is predictable,
Further, it is possible to highly accurately obtain a predicted value of another time series data obtained by using the predicted value.

[0008]

EXAMPLES Specific examples will be shown below.

FIG. 1 shows a block diagram according to an embodiment of the present invention, and the flow of processing of the present invention will be described below using this block diagram.

In this embodiment, an example in which a traffic flow sensor, particularly an ultrasonic sensor, is used as a sensor for measuring time series data is shown. However, the sensor is not necessarily limited to the ultrasonic sensor, This includes not only sensors that measure traffic flow such as loop coil type sensors, image sensors, and microwave sensors, but also sensors that measure other time-series data.

In FIG. 1, passing vehicle information obtained by a plurality of ultrasonic sensors 11 installed on a target road section about a vehicle 10 passing through the sensor installation point is transmitted to a sensor data storage unit 12 to pass the point. When sensor data such as traffic volume, time occupancy rate, average speed, etc. is stored every moment as time series data and obtained by two sensors selected by the propagation source / destination sensor selection unit 13. Taking out the series sensor data from the sensor data storage unit 12,
It transmits to the sensor data prediction unit 14 to predict future sensor data at the propagation destination sensor installation point, and transmits the predicted values of all sensor data in the target section to the section travel time prediction unit 15 to send future section travel. The time is predicted, and the prediction result obtained by the sensor data prediction unit 14 and the section travel time prediction unit 15 is displayed on the prediction result display unit 16. Further, the flow of processing in the sensor data prediction unit 14 will be described with reference to FIG. In FIG. 2, first, the delay time calculation unit 140 calculates in advance the delay time required for the sensor data of the propagation source sensor selected by the propagation source / destination sensor selection unit 13 to propagate as the sensor data of the propagation destination sensor. Then, using this delay time and the sensor data of the propagation source sensor stored in the sensor data storage unit 12, the data shift unit 141 shifts the sensor data of the propagation destination sensor to the future by the delay time and propagates to the propagation destination sensor. The estimated value of the data is used, and the estimated value is divided by the data dividing unit 142 into past data and future data. The past data among them and the transmission destination sensor data obtained from the sensor data storage unit 12 are transmitted to the past data comparison unit 143, the error is calculated, and the error and the future data divided by the data division unit 142 are calculated. The data is transmitted to the future data correction unit 144 to correct future data of the propagation destination sensor data, and the corrected future data is transmitted to the section travel time prediction unit 15 and the prediction result display unit 16.

Next, the detailed configuration, processing, and functions of the present invention will be described with specific examples.

The ultrasonic sensor 11 is, for example, a device as shown in FIG. 3, and the ultrasonic transmission / reception device 32 irradiates a vehicle 34 passing directly below the transmission / reception device with ultrasonic waves to generate reflected waves. The presence / absence information (time series presence / absence pulse train) of the receiving vehicle is obtained. FIG. 4 shows an example of the obtained time series presence / absence pulse train. A vehicle 34 is provided directly below the ultrasonic wave transmitting / receiving device 32.
When is passed, the recognition output 1 is output only while passing,
The recognition output 0 is output while the vehicle is not passing. Therefore, it indicates that the number of pulses is the number of passing vehicles and the width of the pulse is the time required for passing. The number of vehicles and the vehicle occupancy time (or time occupancy rate) can be obtained from the pulse train thus obtained. In the example of FIG. 4, there are five passing vehicles. Depending on the type of sensor, there are those that directly measure the speed, such as a microwave sensor or a double sensor in which two ultrasonic sensors are installed slightly apart from each other. Further, a generally known formula based on the passing traffic volume per unit time and the occupied time obtained by an ultrasonic sensor, an optical sensor, a loop coil type sensor, and the like,

[0014]

[Equation 1]

The average speed v per unit time can also be calculated by Here, L is the average vehicle length, Q is the passing traffic volume per unit time, and Ot is the time occupancy rate.

The sensor data storage unit 12 stores passing vehicle information from a plurality of ultrasonic sensors 11 as time series sensor data such as the above-mentioned traffic volume per unit time, time occupancy rate, and average speed. is there.

Hereinafter, an example in which the time occupancy rate data is applied as the sensor data will be described, but it is not limited to the time occupancy rate and includes traffic volume and average speed.

The propagation source / destination sensor selection unit 13 selects, as a propagation destination sensor, a sensor for which future occupancy rate data is to be predicted among all the sensors installed in the target road section, and the occupancy rate data The sensor that is the propagation source is being selected as the propagation source sensor. Which of the propagation source sensor and the propagation destination sensor is on the upstream side depends on the property of the target road. Further, it is desirable to select a sensor in which the occupation ratio data of the propagation source sensor and the propagation destination sensor have a high correlation with each other. However, when the source sensor data and the destination sensor data have a high correlation, the source occupancy data must be older in time than the destination occupancy data, but for one destination sensor There is no problem even if there are multiple propagation source sensors.

An example of this will be described with reference to FIGS. 5 and 6. FIG. 5 shows the ultrasonic sensors 51 to 56 installed on the road 57 and the traffic flow flowing from left to right on this road. If the target road section is changed from point A to point B,
In this case, the ultrasonic sensors to be predicted are three ultrasonic sensors 53, 54, 55, and these are selected as propagation destination sensors. The propagation source sensor is basically selected from the sensors other than these three propagation destination sensors, that is, the ultrasonic sensors 51, 52, or 56 in FIG. 5, but even if the propagation destination sensor is selected. It doesn't matter.
For example, if the ultrasonic sensors 51 and 53 have the correlation shown in FIG. 6, the propagation source sensor of the propagation destination sensor 53 is selected as 51. Remaining propagation destination sensor 5
Similarly, the propagation source sensors are selected for 4 and 55.

The sensor data prediction unit 14 retrieves the time-series occupation rate data for the two sensors selected by the propagation source / destination sensor selection unit 13 from the sensor data storage unit 12, and determines the future occupation rate of the propagation destination sensor. We are predicting the data.

Here, the detailed internal structure, processing, and function of the sensor data prediction unit 14 will be described with reference to FIGS. 1, 2 and 6.

The delay time calculation unit 140 predetermines the delay time required for the time series occupancy data of the propagation source sensor selected by the propagation source / destination sensor selection unit 13 to propagate as the occupancy data of the propagation destination sensor. It is about to be calculated. The following is an example of calculating this delay time. Of the propagation source / destination sensors, the time-series time occupancy data of the propagation source sensor and the propagation destination sensor are respectively denoted as f (t) and g (t) (where t represents the time zone), and then The following arithmetic processing is performed. That is,

[0023]

[Equation 2]

Is calculated to obtain τ that maximizes the similarity E (τ), and this τ is used as the delay time between the two sensors. However, t0 to t3 are search time ranges when the similarity is calculated. In the example of the ultrasonic sensors 51 and 53 of FIG. 6, the delay time 61 is (t2-t1).

The data shift unit 141 shifts the propagation source occupation rate data transmitted from the sensor data storage unit 12 to the future by the delay time obtained by the delay time calculation unit 140 and estimates the propagation destination sensor data. It is about to be.

The data division unit 142 is the data shift unit 1.
The estimated value of the transmission destination sensor data obtained by shifting in 41 is compared with the reference time (current time) and is divided into past data and future data.

Data shift section 141 and data division section 14
An example of the processing content of No. 2 will be described with reference to FIG. The data shift unit 141 shifts from the upper graph of FIG. 7 to the lower graph by using the propagation source occupation rate data and the delay time. Here, 71 is the time series occupancy data of the propagation source sensor, 61 is the delay time obtained by the delay time calculation unit 140,
Reference numerals 72 and 73 denote estimated values of the time-series occupation rate data of the propagation destination sensor obtained by shifting the propagation source occupation rate data 71 to the future by the delay time 61, and the estimated values are calculated by the data dividing unit 142. The estimated value past data 72 and the estimated value future data 73 are divided.

The past data comparison unit 143 compares the past data of the propagation destination occupancy ratio data estimated value divided by the data division unit 142 with the propagation destination occupancy ratio data transmitted from the sensor data storage unit 12, and determines the error. I am asking for it.

The future data correction unit 144 uses the error calculated by the past data comparison unit 143 to divide the data into the data division unit 14
The future data of the propagation destination occupancy data estimated value divided by 2 is corrected and is obtained as a future predicted value of the propagation destination occupancy data.

Here, an example of the processing contents of the past data comparison unit 143 and the future data correction unit 144 will be described with reference to FIG. In the graph of FIG. 8, the past data comparison unit 143 compares the past data 72 of the propagation destination occupancy rate data estimated value and the actual value 81 of the occupancy rate data of the propagation destination sensor to determine, for example, the average error for the last five hours. Ask. The average error is not limited to the difference between the past data 72 of the propagation destination occupancy rate data estimated value and the actual value 81 of the occupancy rate data of the propagation destination sensor, and may be an average ratio of the two data. Next, the future data correction unit 144 uses the average error (or average ratio) obtained by the past data comparison unit 143 to divide the data into the data division unit 142.
It is possible to obtain a future predicted value of the propagation destination occupancy data by adding it to the future data of the propagation destination occupancy data estimated value (added when the average ratio is used) to correct. further,

[0031]

(Equation 3)

According to the above, it is possible to obtain a future predicted value of the average velocity of the transmission destination sensor. Here, v is an average speed prediction value, α is a negative coefficient, Ot is a future prediction value of the propagation destination occupancy rate, and vmax is the maximum speed.

When both the occupancy rate data and the predicted value of the traffic data are obtained as the sensor data, the future predicted value of the average speed of the propagation destination sensor can be obtained by (Equation 1), while the sensor data is obtained. It is obvious that the processing of (Equation 1) or (Equation 3) is not necessary when using the average velocity directly.

The section travel time prediction unit 15 uses the average speed data prediction value obtained by the future data correction unit 144 in the sensor data prediction unit 14 regarding the sensors of all the target road sections to calculate the travel time of the target road section. I'm about to predict.

The prediction result display unit 16 displays the occupancy rate data prediction result obtained by the future data correction unit 144 of the sensor data prediction unit 14 and the travel time of the target road section obtained by the section travel time prediction unit 15. It is about to display the prediction result. Examples of the display means include a monitor of a traffic control center, an electronic bulletin board, an electronic bulletin board for drivers installed on the side of the road, and a vehicle-mounted terminal via a beacon or FM multiplex. By doing so, future sensor data from all the sensors in the target road section can be predicted, and the section travel time can also be predicted by using the predicted value. This is useful not only for the driver traveling on the road to know the travel time to the destination in advance, but also for the driver to change the route if the information that he / she is about to enter the congestion area will be obtained. Moreover, it is also very effective for optimal signal control in the traffic control center.

[0036]

According to the present invention, various time series data,
In particular, it is possible to accurately predict what the travel time and road condition of the target road section will be in the near future. Moreover, if this prediction result is used, it is effective for traffic control means such as signal control and information provision.

Further, it is effective in terms of cost since the conventional ultrasonic sensor or the like already installed can be used.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a time series data prediction device of the present invention.

FIG. 2 is a block diagram showing a detailed internal configuration of a sensor data prediction unit.

FIG. 3 is a layout diagram of an ultrasonic sensor and its periphery.

FIG. 4 is a diagram showing an example of a time-series presence / absence pulse train obtained by an ultrasonic sensor.

FIG. 5 is a diagram showing an example of a road, an ultrasonic sensor installed on the road, a vehicle traveling on the road, and a target section.

FIG. 6 is a diagram illustrating an example in which time series occupancy data obtained by two ultrasonic sensors have a high correlation.

FIG. 7 is a diagram illustrating an example in which time-series occupancy data of a propagation source sensor is shifted to the future by a delay time.

FIG. 8 illustrates an example in which an error between the past data of the time series occupancy data of the shifted propagation source sensor and the same data of the propagation destination is obtained, and the future data of the same data of the propagation destination is predicted using the error. Fig.

[Explanation of symbols]

10, 34 ... Vehicles, 11, 51, 52, 53, 54, 5
5, 56 ... Ultrasonic sensor, 12 ... Sensor data storage unit,
13 ... Propagation source / destination sensor selection unit, 14 ... Sensor data prediction unit, 15 ... Section travel time prediction unit, 16 ... Prediction result display unit, 31 ... Post, 32 ... Ultrasonic transmission / reception device, 33,
57 ... Road, 35 ... Control device, 36 ... Road end, 61 ... Delay time, 71 ... Time series occupancy data of propagation source sensor, 7
2 ... Past data of the data obtained by shifting the time-series occupation rate data of the propagation source sensor to the future by the delay time, 73
... Future data out of the data obtained by shifting the time-series occupation rate data of the propagation source sensor to the future by the delay time, 81 ...
Occupancy rate data actual value of propagation destination sensor, 82 ... Occupancy rate data predicted value of propagation destination sensor, 140 ... Delay time calculation unit, 1
41 ... Data shift unit, 142 ... Data division unit, 143
… Past data comparison unit, 144… Future data correction unit.

Front page continuation (72) Inventor Yoshiyuki Sato 5-2-1 Omika-cho, Hitachi-shi, Ibaraki Hitachi Ltd. Omika factory

Claims (12)

[Claims] 1. A sensor installed at a plurality of points for measuring time series data, a data storage means for storing data obtained by the sensor, a propagation source sensor for propagation of the time series data, and a propagation destination. A time-series data predicting apparatus comprising: a means for selecting a propagation destination sensor to be a target and a sensor data predicting means for predicting future data of the propagation destination sensor using the propagation source sensor data and the propagation destination sensor data. 2. The sensor data predicting means according to claim 1,
A means for calculating a delay time when the data of the propagation source sensor is similar to the data of the propagation destination sensor, a means for shifting the propagation source sensor data to the future by the delay time, and a means for shifting the shifted data in the past and the future. Data dividing means for dividing into divided data, means for obtaining an error by comparing past data of the divided data with the transmission destination sensor data from the data storing means, and a future divided by the error and the dividing means. And a future data correction unit that corrects future data by using the data and sets it as a predicted value of future data of the propagation destination sensor. 3. The time-series data prediction device according to claim 1, wherein a sensor that measures a traffic flow is used as the sensor that measures the time-series data. 4. An ultrasonic sensor, an optical sensor, an image sensor, or a loop coil type sensor is used as a sensor for measuring time-series data or traffic flow, according to claim 1, 2, or 3. Series data prediction device. 5. The time series data prediction device according to claim 1, wherein a microwave sensor or an ultrasonic double sensor is used as a sensor for measuring time series data or traffic flow. 6. The traffic volume of a vehicle per unit time is measured as time-series data using the time-series data of claim 3, 4 or 5 or a sensor for measuring traffic flow. Alternatively, the time series data prediction device described in 2. 7. The time occupancy rate of a vehicle per unit time is measured as time series data using the time series data or the sensor for measuring traffic flow according to claim 3, 4, or 5. Item 1. The time series data prediction device according to item 1 or 2. 8. The average speed of the vehicle per unit time is measured as time series data using the time series data of claim 3 or 6 or a sensor for measuring traffic flow. The time series data prediction device described. 9. The sensor data predicting means or the future data correcting means obtains the predicted value of the average speed of the vehicle from the predicted value of the time occupancy rate.
The time series data prediction device described. 10. The time-series data according to claim 1, wherein the sensor data predicting means or the future data correcting means obtains the predicted value of the average speed of the vehicle from the predicted values of the time occupancy rate and the traffic volume. Prediction device. 11. A section travel time prediction means for obtaining a travel time prediction value of a target section by using the sensor data prediction value obtained by the sensor data prediction means or the future data correction means. Alternatively, the time series data prediction device described in 2. 12. A traffic flow control system, comprising prediction result display means for displaying a prediction result obtained by a sensor data prediction means, a future data correction means, or a section travel time prediction means. Claim 1, 2, or 1
1. The time series data prediction device according to 1.