CN102157075B - Method for predicting bus arrivals - Google Patents

Abstract

A bus arrival prediction method in the field of information processing technology, which judges the operation stability by analyzing the historical data of bus operation, and divides the time periods suitable for different prediction methods. In the forecast period, the Kalman filter method is used to analyze the real-time GPS data to predict the arrival time, and finally the two forecast results of historical data and real-time data are combined as the final release information through error weighting. The invention has good accuracy and fast calculation speed, and is easy to realize and popularize physically.

Description

Prediction method of bus arrival

technical field

The invention relates to a method in the technical field of information processing, in particular to a method for predicting the arrival of a bus.

Background technique

Bus arrival time prediction technology is the difficulty and focus of bus information release in urban intelligent transportation systems. Shanghai is currently vigorously promoting urban public transport, and has formed an informatization prototype of urban transport operation management and services. As an important part of the urban intelligent transportation system (ITS), the traffic information release system has developed rapidly in recent years. However, the construction of the conventional public transport information release system is still relatively backward compared with the rail transit information release system, and passenger complaints about the operation status of public transport vehicles occur frequently.

The current prediction methods can be divided into arrival time prediction methods based on historical data and arrival time prediction methods based on real-time data according to different sources of data selection. According to the different calculation methods of road section travel time, it can be divided into: time series analysis, Kalman filter, artificial neural network, and other methods.

The above prediction models have their own characteristics in terms of prediction accuracy and practical application. According to literature search, in the arrival time prediction model based on historical data, it is assumed that the actual driving conditions of buses fluctuate slightly around the historical driving conditions. The model is based on a large amount of historical data. This type of model is easy to understand, easy to operate, and widely used. However, when an emergency causes the actual driving situation of the bus to deviate greatly from the historical situation, the prediction effect will be unsatisfactory, and its prediction accuracy is limited. , is not practical. The artificial neural network model has an absolute advantage in prediction accuracy, and it is currently a highly respected bus arrival time prediction model. However, the selection of the training function, learning function and some parameters of the neural network requires experience or trials, and the network training takes a long time. Therefore, it is not easy to realize online real-time training and dynamic prediction. The Kalman filter model uses continuous approximation to obtain higher prediction accuracy, especially when predicting the travel time one step ahead, the model has good prediction performance. However, its ability continues to decline with the increase of steps.

Contents of the invention

The present invention aims at the above-mentioned deficiencies in the prior art, and provides a method for predicting the arrival of buses, which has good accuracy, fast calculation speed, and is easy to physically implement and popularize.

The present invention is realized through the following technical solutions. The present invention judges the operation stability by analyzing the historical data of bus operation, and divides time periods suitable for different prediction methods. In the forecast period, the Kalman filter method is used to analyze the real-time GPS data to predict the arrival time, and finally the two forecast results of historical data and real-time data are combined as the final release information through error weighting.

The present invention specifically comprises the following steps:

1) Data collection: collect bus operation data through three methods, and provide real-time operation status information of buses.

2) Data processing: Use ArcGIS geographic information system software to eliminate erroneous data, determine the real-time position of the bus and the time of data return.

3) Stability analysis of bus operation: Collect the historical data of bus operation within half a year, divide it into 18 situations, and conduct stability analysis for each situation.

4) Selection of bus arrival prediction method: According to the stability analysis results, select the appropriate arrival prediction method.

5) Establish the arrival time prediction model: comprehensively consider the historical data of bus operation and real-time GPS data, respectively predict the arrival time, and carry out weighted average of the two prediction results according to the error in the prediction process, and release it as the final arrival time On the electronic stop sign for passengers' reference.

The method of the invention selects different prediction modes on the basis of analyzing the operation stability of the bus according to the historical operation data of the bus. After separating the time period suitable for predicting the arrival time, the historical data and real-time GPS data are used to predict respectively, and finally the method of error weighting is used to predict the arrival time of the bus. Therefore, considering the historical law of bus operation and real-time road conditions and traffic flow information, the bus arrival time can be predicted more accurately.

Description of drawings

Figure 1 shows the results of bus GPS data processing.

Figures 2 and 3 show the stability analysis of bus operation, where Xujiahui is selected as the station, the bus operation direction is down, and the date is Saturday and Sunday.

Fig. 4 is a flowchart of a bus arrival prediction method based on historical and real-time data error weighting.

Figure 5 is a schematic diagram of data acquisition.

Fig. 6 is a schematic diagram of data transmission.

Figure 7 and Figure 8 are schematic diagrams of distribution of real-time GPS data reporting points.

Detailed ways

The embodiments of the present invention are described in detail below. This embodiment is implemented on the premise of the technical solution of the present invention, and detailed implementation methods and specific operating procedures are provided, but the protection scope of the present invention is not limited to the following implementation example.

Example

As shown in Figure 4, it is the prediction process of this embodiment, and the specific implementation is roughly divided into five parts: through the vehicle-mounted RFID beacon device/card reader, the annular electromagnetic induction coil, and the vehicle-mounted GPS transceiver terminal, the bus operation is collected. Data to provide real-time operating status information of buses; use ArcGIS geographic information system software to eliminate erroneous data, determine the real-time position of the bus and the time of data return; analyze the stability of the bus operation by time and road section according to the collected historical data of bus operation According to the stability analysis results, select the corresponding bus arrival prediction method; for the road sections and time periods suitable for the arrival time prediction, establish the arrival time prediction model, comprehensively consider the bus operation history data and real-time GPS data, respectively. For the arrival time prediction, the weighted average of the two prediction results is carried out according to the error in the prediction process, and the final arrival time is published on the electronic stop board for passengers' reference.

The specific description is as follows:

Data collection: Collect bus operation data through vehicle-mounted RFID beacon equipment/card reader, electromagnetic induction coil, and vehicle-mounted GPS transceiver terminal, and provide real-time bus operation status information. Among them, when the vehicle-mounted RFID beacon device enters and exits the bus station, it triggers the card reader at the entrance and exit of the station, and the downstream station obtains the real-time location information of the bus, thereby displaying the location of the station on the electronic stop sign; the electromagnetic induction coil Collect the location information of the bus for real-time prediction of the distance between the bus and the target site; the vehicle-mounted GPS transceiver terminal provides real-time GPS data for analyzing the time and real-time speed of the bus at the reporting data point, and combined with the vehicle odometer calculation The time the bus arrives at the destination stop. The schematic diagram of bus data collection is shown in Figure 5.

Data processing: Mainly process the real-time GPS data recorded by the GPS transceiver terminal on the bus to analyze the arrival time of the bus. Use ArcGIS geographic information system software to eliminate erroneous data, including points where the GPS data error is too large, GPS return points when the bus enters and exits the garage sooner or later, and GPS return points where the internal code meter of the bus is zero, etc., to determine the real-time position of the bus and the moment of data return. According to the time of GPS data reporting, it is divided into the following situations, that is, working days are Monday and Friday, working days are Tuesday, Wednesday and Thursday, and rest days are Saturday and Sunday. In each date, the time of the return point is divided into time periods according to the number of hours, so as to analyze the operation conditions in each time period.

Stability analysis of bus operation: The historical bus operation data used for stability analysis is not the longer the time span, the better the effect. Collect the bus operation history data within half a year and divide it into 18 situations, namely (i) the direction of bus travel: Uplink, downlink; (ii) Date: Monday, Friday, Tuesday, Wednesday, Thursday, Saturday, Sunday; (iii) Time of day: morning peak, midday peak, evening peak, stability analysis for each case. For the bus operation, the average running time between stations, the average running speed, and the variance and standard deviation of the running time between stations are all indicators to measure the stability of its operation. When calculating the mean value, the form of censored mean value is uniformly adopted, the largest and smallest 10% GPS return point data are removed, and the remaining data is then calculated for the arithmetic mean value, which can effectively avoid the influence of wrong data on the calculation results. According to the statistical results and comprehensive consideration of passengers' patience for waiting time, the standard deviation of the running time is set to 1 minute, that is, when the final arrival time is predicted, the deviation of the referenced historical data is allowed within 1 minute.

Selection of bus arrival prediction method: According to the stability analysis results, select the appropriate arrival prediction method. (i) Predict the number of stops ahead for the bus. This prediction method is suitable for the most unstable time period or road section of the bus operation. During the prediction process, hardware equipment is used to collect the information of the key positions of the bus on the upstream road section, such as entering and exiting the station, and directly transmit it to the target through the information transmission channel. The electronic stop sign of the site. (ii) Predict the distance of the bus from the target stop. It is suitable for sections where the distance between stations is short and the bus operation is relatively stable. Using sensor technology, wireless beacons combined with GPS technology, the distance of the bus upstream from the target station can be known. As the bus gets closer to the target site, the distance information published on the electronic stop board is constantly updated, and passengers can know the actual distance between the bus and the target site. (iii) Predict how long it will take for the bus to reach the destination stop. This prediction method is the most intuitive, and it is suitable for situations where there is a certain distance between stations, the traffic flow complexity of the road section is low, and the bus operation is stable. Using the historical data of bus operation or real-time GPS data, the model of bus operation status is established, and the time when the bus arrives at the target station is calculated by establishing a prediction model and a prediction algorithm.

Establish the arrival time prediction model: comprehensively consider the historical data of bus operation and real-time GPS data, respectively predict the arrival time, and carry out weighted average of the two prediction results according to the error in the prediction process, and publish it on the electronic website as the final arrival time The stop signs are for passengers' reference. Among them, the arrival time is predicted according to the historical data, arranged according to the order of the actual uplink and downlink stations, and the distance between two adjacent stations can be obtained by subtracting the odometer value of the latter station from the odometer value of the adjacent previous station, and the calculation Average travel time and average speed of buses between stops. In order to reduce the error as much as possible, the process of calculating the mean value is in the form of censored mean value. The drift in the GPS measurement process causes the error of the distance between two adjacent stations to be 30 meters, and the variance of the running time and running speed between the stations is counted during the calculation process. According to the real-time GPS data to predict the arrival time, two data reporting methods are adopted, that is, reporting data by time interval and reporting data by distance interval, as shown in Figure 7 and Figure 8, respectively establishing a Kalman iterative model. The real-time speed and time of the bus at the next data return point predicted by the iterative algorithm are transformed into the time when the bus arrives at the target station, and the variance of the running time and running speed generated during the prediction process is calculated according to the error correlation matrix. Finally, the forecast results obtained by predicting the arrival time of the two data sources respectively are weighted and averaged according to the running time variance generated during the forecasting process, as the final forecast results published on the electronic stop sign. Because in the final arrival time prediction result, with each update of real-time data, the historical data prediction part should also be continuously updated. The assumption of the historical data update process is that the bus keeps running at a constant speed between stations. Since the Kalman filtering iterative algorithm requires the initial state to be set, this paper sets the initial state as the average value of historical statistics, that is, the impact of historical data and real-time GPS data on the final prediction results is equal. Different data reporting methods are used, and the proportions of the arrival time predicted based on historical data and real-time GPS data in the final forecast results will also vary.

Claims (1)

1. A predictive method for bus arrival, characterized in that, by analyzing the bus operation history data to judge the operation stability, the bus operation history data within half a year is collected, divided into 18 situations, and each situation carries out stability analysis respectively; Selection of bus arrival prediction method: According to the stability analysis results, select the appropriate arrival prediction method: (i) predict the number of stops ahead of the bus; (ii) predict the distance between the bus and the target station; (iii) predict How long will it take for the bus to reach the target stop; In the prediction period, the Kalman filter method is used to analyze the real-time GPS data to predict the arrival time, and the real-time speed and time of the bus at the next data return point predicted by the Kalman iterative algorithm are converted into the time when the bus arrives at the target station , and according to the variance of the running time and running speed generated in the error correlation matrix statistical prediction process, and finally combine the two kinds of prediction results of historical data and real-time data through error weighting as the final release information. The method specifically includes the following steps: The first step, data collection: collect bus operation data through vehicle-mounted RFID beacon equipment/card reader, electromagnetic induction coil and vehicle-mounted GPS transceiver terminal, and provide real-time bus operation status information; The second step, data processing: use ArcGIS geographic information system software to eliminate wrong data, determine the real-time position of the bus and the time of data return; The third step, bus operation stability analysis: Collect the bus operation history data within half a year, divide it into 18 situations, and conduct stability analysis for each situation; that is (i) bus driving direction: up, down; (ii) Date: Monday, Friday, Tuesday, Wednesday, Thursday, Saturday, Sunday; (iii) Daily hours: morning peak, midday peak, evening peak; for bus operation, the average running time between stations, the average running speed and The variance and standard deviation of the running time between stations are indicators to measure the stability of its operation; The fourth step, the selection of bus arrival prediction method: according to the stability analysis results, select the appropriate arrival prediction method; The fifth step is to establish the arrival time prediction model: comprehensively consider the historical data of bus operation and real-time GPS data, respectively predict the arrival time, and carry out weighted average of the two prediction results according to the error in the prediction process, as the final arrival time The time is published on the electronic stop board for passengers' reference. The arrival time is predicted based on historical data, arranged in the order of the actual uplink and downlink stations, and can be obtained by subtracting the odometer value of the next station from the odometer value of the adjacent previous station. The distance between two adjacent stations, calculate the average running time and average speed of the bus between the stations; predict the arrival time based on real-time GPS data, and use two data reporting methods, that is, report data by time interval and by distance For interval return data, Kalman iterative models are established respectively. After that, the prediction results obtained by predicting the arrival time of the two data sources are weighted and averaged according to the running time variance generated during the prediction process, and published on the electronic stop sign prediction results.

Images