KR20070061015A - Method for searching the shortest route based on traffic prediction and apparatus thereof - Google Patents

Abstract

A method and an apparatus for searching the shortest route based on traffic prediction are provided to enhance user convenience through speed prediction based on an accumulated speed pattern and a prediction method based on a kalman filter scheme. A first directed graph is generated based on road traffic network data and a linear dual graph corresponding to the first directed graph is generated(S110). A future traffic speed according to the current traffic condition variation is predicted(S120). A cost function of an edge in the linear dual graph is generated based on the predicted traffic speed and a traffic topology(S130). The shortest path is obtained based on the cost function(S140).

Description

Method for searching the shortest route based on traffic prediction and apparatus

1 is a flowchart illustrating an outline process of a shortest route searching method based on traffic prediction according to the present invention.

2 is a flow chart showing in more detail the shortest route search method based on traffic prediction according to the present invention.

3 is a block diagram showing the configuration of the shortest route search apparatus based on traffic prediction according to the present invention.

4A is a diagram illustrating a basic graph for generating a dual graph for shortest path search according to the present invention.

4B is a diagram illustrating a dual graph generated based on FIG. 4A.

5 is a view showing a process of correcting when a current speed is missing in the shortest path method and apparatus according to the present invention.

FIG. 6A is a view illustrating a concept of a Kalman filter applied when a traffic jam occurs during a traffic prediction.

FIG. 6B is a diagram illustrating the actual implementation of the Kalman filter logic by embodying the concept of FIG. 6A.

The present invention relates to a method and apparatus for searching for the shortest route. More specifically, the present invention is based on a traffic prediction method so that the accuracy of the current traffic information can be increased by not only using speed information of a past traffic network. The present invention relates to a method and apparatus for searching for the shortest path.

Recently, LBS has provided services useful in real life with the development of location technology, wireless communication technology and vehicle industry. Also, LBS has been spotlighted as an essential technology in the field of telematics which is closely related to vehicle location information. In particular, the navigation service, one of the important functions of the LBS technology, should support the driver's convenience and safety by providing more intelligent traffic prediction information that combines GIS and ITS technology. Such intelligent traffic systems should be able to provide real help to the driver by smoothing the flow of traffic by detecting and appropriately controlling the overall traffic situation. For example, it should be able to inform the current flow information such as road delays, congestion, and traffic conditions in neighboring areas, and provide the fastest route to avoid the delays to the destination in real time. To this end, speed prediction for future time should be possible by reflecting real-time traffic information, and a system that provides the fastest route based on this is required. Most existing traffic prediction systems do not reflect real-time traffic conditions because they provide the shortest route service using only accumulated experience data. The reason is that it is very difficult to find the shortest path in the network, and many heuristic methods have been suggested.

A recent study has proposed a method for monitoring road congestion information in real time and reassigning a moving object based on the information. According to the results of this study, the dynamic route allocation considering the road congestion reduces the average travel time of all moving objects compared to the static route allocation. Therefore, based on the current traffic situation, the measurement of dynamic traffic prediction information of the future and the traffic guidance service considering this have become a very important area of interest in the transportation network.

The technical problem to be solved by the present invention is to solve the above problems, to generate a dual graph for reducing the search cost and to solve the rotation problems that may occur at the intersection and to accumulate the traffic information or The present invention provides a method and apparatus for searching a shortest distance after performing velocity prediction by applying a Kalman filter.

In order to solve the above technical problem, the shortest route searching method based on the traffic prediction according to the present invention generates a first graph based on road traffic network data and a linear dual graph corresponding to the first graph. Generating a dual graph; Predicting future traffic speeds based on current traffic conditions; Generating a cost function of each edge in the linear dual graph based on the predicted traffic speed and traffic topology; And obtaining a shortest path based on the cost function.

In order to solve the above technical problem, the shortest route retrieval apparatus based on traffic prediction according to the present invention includes a generation unit for generating a linear dual graph based on road traffic network data; A prediction unit for inputting a current traffic situation change and predicting and outputting a future traffic speed according to the change; A cost function generator for generating a cost function of an edge of the linear dual graph by inputting the predicted traffic speed and traffic topology; And a searcher for obtaining a shortest path based on the cost function.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. 1 is a flowchart showing an overview process of a shortest route searching method based on traffic prediction according to the present invention, and FIG. 2 is a flowchart showing a shortest route searching method based on traffic prediction according to the present invention in more detail. . 3 is a block diagram showing the configuration of the shortest route search apparatus based on traffic prediction according to the present invention. FIG. 4A is a diagram illustrating a basic graph for generating a dual graph for shortest path search according to the present invention, and FIG. 4B is a diagram showing a dual graph generated based on FIG. 4A. FIG. 5 is a diagram illustrating a process of correcting when a current speed is missing in the shortest path method and apparatus according to the present invention. 6A is a diagram illustrating a concept of a Kalman filter applied when a traffic jam occurs during a traffic prediction, and FIG. 6B is a diagram illustrating an actual process of performing the Kalman filter logic by embodying the concept of FIG. 6A.

A general description will be made with reference to FIG. 1 and then with reference to FIG. First, a graph as shown in FIG. 4A is generated from a node set having a node as an element and an edge set composed of edges connecting the nodes. Based on this graph, a linear dual graph is generated as shown in FIG. 4B (S110). Next, the traffic speed is predicted according to the traffic situation including the current traffic speed. In this case, as will be described later, it is divided into two cases of a normal situation and an abnormal situation, and the future traffic speed is predicted by using cumulative traffic information in the former case and Kalman filter in the latter case (S120). Now, a cost function is generated based on the predicted traffic speed and the traffic topology such as the number of lanes or the type of road (S130). Finally, the generated cost function is applied to the linear dual graph that has already been generated to search for and determine the shortest path (S140).

라고 하고, 도 4b의 듀얼 그래프를

라고 하였을 때, 아래와 같이 생성된다.This will be described in more detail with reference to FIG. 2. In order to predict the traffic speed, the current traffic information including the current, that is, the measurement position and the moving speed at the measurement time is measured in real time and a dual graph is generated (S210). Meanwhile, as mentioned above, the dual graph is constructed.

And the dual graph of FIG.

Is generated as follows.

는 노드집합

과 에지집합

로 구성된다. 에지는 노드들 사이의 이진관계,

를 나타낸다. 출발노드는

으로, 목적노드는

로 표기한다.graph

Is a set of nodes

And edge sets

It consists of. Edges are binary relationships between nodes,

Indicates. Departure node

The destination node

It is written as.

에 대하여, 듀얼그래프

는

인 정점집합을 가지며, 에지집합

을 갖도록 이루어진다.graph

About, dualgraph

Is

Have a set of vertices

It is made to have.

To this end, traffic network data is displayed as nodes and links, and the dual graph of FIG. 4B is generated from the data.

U-turns exist in the city's road network with a number of left-turn prohibitions. This is being done to minimize the disturbance of the main traffic flow due to the waiting queue of vehicles to turn left at the intersection. In this situation, it is desirable to impose the following conditions to find the optimal path.

를 에지

의 가중치로 갖는 그래프

상의 연속된 에지

와

에 대해

에서

로 가는 패널티(penalty)

이 존재한다고 할 때, 다음과 같은 통행 조건을 가진다.given

Edge

Graph with weights of

Edges on top

Wow

About

in

Penalty to

When it exists, it has the following conditions of passage.

의 노드 수 /

/는 주어진 그래프

의 에지 수 /

/ 와 같으며,

에서 에지 수는

에서 경로길이가 2인 경로의 수와 같다.

의 노드의 차수(degree)

에 대해 Cauchy-Schwarz 부등식을 기반으로 상한수(upper estimation)를 구하면 다음의 수학식 1처럼 된다.In the case of using the dual graph of FIG.

Number of nodes in /

/ Is a given graph

Number of edges of /

Is the same as /

Edge number

Is equal to the number of paths with path length equal to 2.

The degree of the node

The upper estimation is obtained based on the Cauchy-Schwarz inequality for.

를 노드

와

사이의 속도라고 가정할 때, 링크

의 보정은 링크

,

,

의 평균값으로 이루어진다.Meanwhile, a dual graph is formed based on the current traffic information measured in step S210, and correction of the missing section is performed to perform more accurate traffic prediction based on the current real-time speed (S220). That is, for accurate traffic prediction, a missing link whose speed is input in real time to measure the current traffic flow but whose speed is not measured may exist in the measured data. This is illustrated in FIG. 5 as a missing section between node a and node b. This requires a calibration process for more accurate measurements. In this case, the speed of the missing section is corrected by calculating the average speed of the adjacent links except the U-turn. Look at Figure 5

Node

Wow

Assuming the speed between

Correction of link

,

,

It consists of an average value of.

Now, the corrected current traffic information and the cumulative traffic information are compared (S230), and according to whether or not the current traffic condition is normal (S240), it is determined in which manner. This is because special attention should be given to drivers on roads where traffic jams or accidents (roadwork, accidents, events, etc.) occur. In order to find such a section, the real-time speed corrected in step S220 and the cumulative average speed are compared to deviate from a preset allowable threshold error, and if not, the prediction method by the Kalman filter and the prediction by the pattern of the cumulative speed are performed. Do each of them.

일 때, 예측시간

의 속도는

부터

시간까지의 누적평균 속도를 이용하여 구하게 된다.First, it is determined that the current traffic situation is normal and the traffic speed predicted by the cumulative traffic pattern (S260) will be described in detail. The predictive model using the cumulative speed pattern is a method of predicting future speed by analyzing the cumulative speed pattern. As mentioned earlier, this method is meaningful because it applies when the difference between the present speed and the cumulative mean speed is within the acceptable threshold, i.e. when the standard deviation of the two speeds is not large. The pattern of cumulative speed may vary depending on the type of road, but in general, the pattern shows the same pattern for the same day and time. Therefore, the current time

When is estimated time

The speed of

from

This is obtained using the cumulative average speed to time.

As another case, when the current traffic speed is abnormal, the process of performing traffic prediction by the Kalman filter will be described (S250). Invented by Rudolf E. Kalman, the Kalman filter is an efficient iterative computation method that tracks time-dependent state vectors with noise motion equations in real time using the least-square method. to be. Kalman filters are often used to find signals from noise so that a system can properly predict changes over time. The Kalman filter at discrete time is represented by the following equation.

는

시간에서 상태 벡터

의 값을 나타내고,

은 시간

에서 입력으로 주어지는 제어 값을 의미한다. 그리고

은 발생 가능한 잡음을 나타내며,

와

는 각각의 값들을 대응시키는 상수 행렬이다. here,

Is

State vector in time

Represents the value of,

Silver time

The control value given as input to the. And

Indicates possible noise,

Wow

Is a constant matrix that maps individual values.

값과 제어 값

을 이용해 다음 시간에서의 상태 값

를 계산하는 것이라고 말할 수 있다. 다음 시간에서의 상태 값

를 추정하기 위해 필요한 다른 요소는 오차를 가진 측정 값이며 다음의 수학식 3처럼 표현된다. Estimation using the Kalman filter eventually results in the previous state

Values and control values

Status value at

It can be said to calculate. State value at

Another factor required for estimating is an error measured value and is expressed as Equation 3 below.

는 상태 값

와 측정 값

의 관계를 나타내는 행렬이고,

는 측정 오차이다. here,

Is the state value

And measured value

Is a matrix representing

Is the measurement error.

The basic goal of the Kalman filter is to repeatedly correct the difference between the ideal state value and the calculated state value, ultimately converging the calculated value to the ideal state value. This concept is expressed as in Equations 4a to 4b below.

를 최소화하는 과정이 칼만 필터를 계산하는 과정이 되며, 결과적인 칼만 필터식은 다음의 수학식 5와 같다.From here

The process of minimizing is the process of calculating the Kalman filter, and the resulting Kalman filter equation is shown in Equation 5 below.

The Kalman filter computes two modules repeatedly and estimates the correct values. As shown in FIG. 6A, the "predict" and "correct" modules are repeatedly calculated. The equation actually used in each time update module and measurement update module is shown in FIG. 6B.

,

는 k번째 에지, 도로의 종류를

, 차선의 수를

, 예측된 교통속도를

라고 할 때, 다음의 수학식 6a 내지 수학식 6b처럼 생성한다.Next, a cost function for searching for the shortest distance is generated using the dual graph (S270), which will be described below. That is, as described above, after the traffic prediction process is performed, the process for searching the shortest path is performed next. In the OpenLS Route specification, the shortest route service is defined as "travel time for a route", and the present invention complies with this. For this purpose, the cost function is obtained as follows. The cost function

,

K edge, the type of road

, The number of lanes

, The estimated traffic speed

Is generated as in Equations 6a to 6b.

The cost function is applied to a linear dual graph to predict future traffic information based on past and present traffic information, and a dynamic route search that enables drivers to reach their destination in the shortest time (S280).

Now, a preferred embodiment of the shortest route search apparatus based on traffic prediction according to the present invention will be described. However, the content duplicated with the shortest path search method described above will be omitted or simply mentioned.

The first generation unit (not shown) in the generation unit 310 receives the road traffic network data collected as an input and uses an edge set composed of a node set having nodes as an element and an edge connecting each node, as illustrated in FIG. 4A. Create a graph as shown. The second generator (not shown) in the generator 310 generates and outputs the dual graph of FIG. 4B as described above based on the graph generated by the first generator.

The prediction unit 320 receives the road traffic network data and the current traffic conditions as inputs to predict future traffic speeds, and determines the two conditions mentioned above, that is, a normal situation and an abnormal situation. Predict the speed. In order to obtain more accurate results, the correction for the missing section is performed as described above. In addition, the first prediction unit (not shown) in the prediction unit 320 outputs the prediction result at a cumulative average speed from the previous time to the present time by the time to be predicted as described above for the prediction in the normal situation. . The second predictor (not shown) generates and outputs a prediction result by applying a Kalman filter in an abnormal situation.

The cost function generator 330 generates and outputs the above cost function based on the traffic information including the predicted traffic speed, the type of road, and the number of lanes output by the predictor 320.

Finally, the search unit 340 receives the dual graph output from the generator 310 and the cost function output from the cost function generator 330, respectively, and applies the cost function to the dual graph to finally search for the shortest distance. Will print.

Results of directly testing the shortest route search method and the apparatus based on the traffic prediction according to the present invention are as follows. In the test, the traffic prediction model based on the cumulative speed data and the model according to the present invention were compared and analyzed in Gangnam-gu, one of the most complex areas in Korea. The area consists of about 20,000 links. For the experiment, we created a six-month cumulative speed table and compared the traffic prediction and actual speed for four weeks. The cumulative speed data for 6 months were classified by day and time. The reason is that the experiment showed an average speed difference of less than 5 km for the same time zone (1 hour) on the same day for a particular link, while an average speed difference of 15 km was shown for different days and the same time zone. The cumulative speed patterns were similar for each day, day of the week, and time of day. There are three cases of comparative analysis.

(1) Speed prediction using only cumulative data

(2) Prediction method using real-time traffic information and cumulative speed pattern

(3) "Method (2)" + When real-time traffic information and cumulative speed pattern are different, it is estimated by applying Kalman filter

(1) Prediction results using only cumulative data

The cumulative data consisted of the days of the week from Monday to Sunday, and the actual data were entered in 5 minute increments. Experimental results showed that the standard mean error for each link was 5km. Therefore, since the link included in the 95% confidence interval was 85.3% of the prediction results, the prediction rate using only cumulative data is 85.3% within the 95% confidence interval.

(2) Prediction results using real time data and cumulative speed pattern

The cumulative speed pattern predicts the speed by using the real-time speed and the change pattern of the cumulative speed. In this test, the real-time speed was compared with the speed pattern with the accumulated speed while maintaining the data one hour before the current time. As a result, the link included in the 95% confidence interval is 86.2% of the total, which is better than the cumulative speed alone.

(3) Prediction results using Kalman filter

This method applies the prediction method using Kalman filter in case of traffic congestion or sudden situation where the speed change is big. That is, the Kalman filter is applied when the speed of the current time is more than 2 times the past cumulative average speed. As a result, 89.2% of the links included in the 95% confidence interval resulted in an improvement over the above methods (1) and (2).

The shortest route search method based on traffic prediction according to the present invention can also be embodied as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD_ROM, magnetic tape, floppy disks, and optical data storage, and may also include those implemented in the form of carrier waves (eg, transmission over the Internet). . The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, optimal embodiments have been disclosed in the drawings and the specification. Although specific terms have been used herein, they are used only for the purpose of describing the present invention and are not intended to limit the scope of the present invention as defined in the claims or the claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, the shortest route searching method and the apparatus based on the traffic prediction according to the present invention propose a new route method that can guide a destination to a destination in the fastest time using a traffic prediction model reflecting real-time traffic information. do.

To this end, a linear dual graph was generated as a structure to reduce search cost and solve rotation problems that may occur at intersections.

In addition, when predicting traffic using only the cumulative speed, only an average value of the speed can be obtained, and the change in real time is difficult to estimate the result. In other words, the speed prediction using the cumulative speed pattern applicable when the variation of the traffic speed is acceptable, and the Kalman filter prediction method which can be used in traffic congestion or sudden situation, are applied to the road information. The shortest path search can be performed.

As a result, it can provide reliable information to the driver about the near future speed and further improve the quality of telematics.

Claims (11)

(a) generating a first graph based on road traffic network data and generating a linear dual graph corresponding to the first graph; (b) predicting future traffic speeds in accordance with changes in current traffic conditions; (c) generating a cost function of an edge in the linear dual graph based on the predicted traffic speed and traffic topology; And (d) obtaining a shortest route based on the cost function; and a shortest route searching method based on traffic prediction. The method of claim 1, wherein step (a) (a1) generating the first graph using a node set having a node as an element and an edge set having an edge as an element connecting the nodes; And (a2) the linear dual graph includes generating the node set and the edge set having the same vertex set, such that the start of one element of any two elements of the edge set and the end of the other element coincide; The shortest route search method based on traffic prediction. The method of claim 1, wherein step (b) (b1) predicting a future speed at a cumulative average speed from a previous time to a current time by a time to be predicted when the current traffic situation is normal; And (b2) if it is determined that the current traffic situation is abnormal, applying a Kalman filter to predict a future speed; the shortest path searching method based on the traffic prediction, comprising: a. The method of claim 1, wherein step (b) Measuring the current traffic flow based on the real-time speed, and correcting the speed of the missing section; Shortest path search method based on the traffic prediction, characterized in that it further comprises. The method of claim 4, wherein the correction is A shortest route searching method based on traffic prediction, which is performed by obtaining an average speed of links adjacent to a current location except a U-turn. The method of claim 1, wherein the cost function Cost function

,

K edge, the type of road

, The number of lanes

, The estimated traffic speed

In this case, the shortest route searching method based on traffic prediction, which is generated as in Equation 7 below.

The method of claim 1, wherein step (d) The shortest path search method based on traffic prediction, wherein the shortest path is obtained by applying a Dijkstra algorithm to the linear dual graph. A generation unit generating a linear dual graph based on the road traffic network data; A prediction unit for inputting a current traffic situation change and predicting and outputting a future traffic speed according to the change; A cost function generator for generating a cost function of an edge of the linear dual graph by inputting the predicted traffic speed and traffic topology; And And a searcher for obtaining the shortest path based on the cost function. The method of claim 8, wherein the generation unit A first generation unit generating a first graph by using a node set having an element as a node of the road traffic network data and an edge set having an element as an edge connecting the nodes; And And a second generator configured to generate the linear dual graph such that the node set and the edge set have the same vertex set, and the start of any one of the two elements of the edge set and the end of the other element coincide with each other. Shortest route retrieval device based on traffic prediction, characterized in that. The method of claim 8, wherein the prediction unit A first prediction unit for predicting and outputting a future speed at a cumulative average speed from a previous time to a current time by a time to be predicted when the current traffic situation is normal; And And a second prediction unit for predicting and outputting a future speed by applying a Kalman filter, when the current traffic situation is determined to be an abnormal situation. (a) generating a first graph based on road traffic network data and generating a linear dual graph corresponding to the first graph; (b) predicting future traffic speeds in accordance with changes in current traffic conditions; (c) generating a cost function of each edge in the linear dual graph based on the predicted traffic speed and traffic topology; And (d) obtaining a shortest path based on the cost function; a computer-readable recording medium having recorded thereon a program capable of executing a method on a computer.

Images