CN114674335A

CN114674335A - Optimal target link matching method

Info

Publication number: CN114674335A
Application number: CN202210302682.6A
Authority: CN
Inventors: 黄文成; 尹燕辉; 于耀程
Original assignee: Southwest Jiaotong University
Current assignee: Southwest Jiaotong University
Priority date: 2022-03-24
Filing date: 2022-03-24
Publication date: 2022-06-28
Anticipated expiration: 2042-03-24
Also published as: CN114674335B

Abstract

The invention discloses a matching method of an optimal target link, which comprises the following steps: s1: acquiring path related information in a map; s2: obtaining a plurality of target links by utilizing a path search algorithm according to the path related information; s3: acquiring the shortest link in all the target links; s4: and obtaining a matching result of the optimal target link by using a weight and vehicle track matching algorithm according to the shortest link. The matching method of the optimal target link provided by the invention can ensure the authenticity of the shortest link and simultaneously ensure the high precision of the recommended link.

Description

Optimal target link matching method

Technical Field

The invention relates to the technical field of road path matching, in particular to a matching method of an optimal target link.

Background

Map matching algorithms that utilize road segment connectivity and other data (i.e., location, speed, and heading) in the map matching process are typically adapted for high frequency (1Hz or higher) positioning data from GPS. When such map matching algorithms are applied to low frequency data (e.g., data from private cars, buses or light vehicles or smartphones), the performance of these algorithms degrades to 70% of the area in terms of correct link identification, particularly in urban and suburban road networks. For certain real-time Intelligent Transportation System (ITS) applications and services, such as estimating link travel time and speed from low frequency GPS data, this level of performance may be insufficient, and thus, the accuracy of the travel link recommended by the system is not high enough when the originating and terminating nodes are known.

Disclosure of Invention

The invention aims to provide a matching method of an optimal target link, so as to ensure the authenticity of the shortest link and ensure the high precision of a recommended link.

The technical scheme for solving the technical problems is as follows:

the invention provides a matching method of an optimal target link, which comprises the following steps:

s1: acquiring path related information in a map;

s2: obtaining a plurality of target links by utilizing a path search algorithm according to the path related information;

s3: acquiring the shortest link in all the target links;

s4: and obtaining a matching result of the optimal target link by using a weight and vehicle track matching algorithm according to the shortest link.

Alternatively, the step S4 includes:

s41: obtaining the shortest link distance of the shortest link by using a distance and cost adding heuristic function;

s42: obtaining the vehicle track course difference of the shortest link, the vertical distance from the target node to the candidate link point and the bearing difference;

s43: obtaining the shortest link, the vehicle track course difference, the vertical distance from the target node to the candidate link point and the optimal weight coefficient corresponding to the bearing difference according to the shortest link distance of the shortest link, the vehicle track course difference, the vertical distance from the target node to the candidate link point and the bearing difference;

s44: obtaining a new weight and vehicle track matching algorithm according to the optimal weight coefficient and the weight and vehicle track matching algorithm;

s45: and obtaining a matching result of the optimal target link by using the new weight and vehicle track matching algorithm according to the actual target node.

Alternatively, the step S41 includes:

s411: acquiring all node segment-related information in all the target links, wherein the node segment-related information comprises node segment distance information, conductivity and steering limits;

s412: judging whether the initial node section of the current target link is conducted, if so, entering a step S413, otherwise, entering a step S418;

s413: judging whether the current target link is limited by steering, if so, entering a step S418, otherwise, entering a step S414;

s414: obtaining historical positioning data of a node section of a current target link which is limited by steering to obtain distance information of the current target link;

s415: according to the distance information of the current target link, a heuristic function of adding a cost to the distance is utilized to obtain the lowest function value of the current target link;

s416: judging whether the current target link is the last target link, if so, entering a step S417, otherwise, entering a step S419;

s417: comparing the lowest function values of all the target links, and taking the target path with the minimum lowest function value as the shortest link;

s418: obtaining the shortest link distance of the shortest link according to the distance information of all the node sections of the shortest link;

s419: enter the next target link and return to step S412.

Alternatively, in step S413, the steering limitation includes: the target path comprises a low-frequency node section, and the frequency response speed of the low-frequency node section is lower than that of a normal node section.

Optionally, in the step S415, the distance plus cost heuristic function f (x) is:

f(x)＝g(x)+h(x)

where g (x) represents a path cost function, and h (x) represents a heuristic evaluation function.

Optionally, in step S42, the vehicle track heading difference HD is:

HD＝W_hf(Δθ)

wherein, W_hF (delta theta) is a positive function and f (delta theta) ═ cos (delta theta) |, delta theta ═ theta-theta |, and f (delta theta) |, is a weight coefficient of the vehicle track heading difference_jI, theta is the vehicle track course from the starting node to the target node, theta_jAnd the vehicle track heading of the starting node to the candidate link.

Optionally, in step S42, the vertical distance PD from the target node to the candidate link point is:

PD＝W_df(d_j)

wherein, W_dIs a vertical distance weight coefficient, f (d)_j) Is a vertical distance d_jIs a positive function of

Optionally, in step S42, the bearing difference BD is:

BD＝W_bf(Δα)

wherein, W_bSaid is the bearing difference weight coefficient, f (Δ α) is a positive function of the bearing difference and f (Δ α) ═ cos (Δ α) |, Δ α ═ a-a |_jI, a is the orientation of the vehicle (relative to north), a_jIs a candidate link c_jIn the direction of the axis of rotation.

Alternatively, the step S43 includes:

s431: initializing the shortest link, the vehicle track course difference, the vertical distance from the target node to the candidate link point and the bearing difference to obtain a new shortest link, a new vehicle track course difference, a new vertical distance from the target node to the candidate link point and a new bearing difference;

s432: assigning the error minimization function to obtain the assigned error minimization function;

s433: respectively obtaining a new shortest link, a new vehicle track course difference, a vertical distance from a new target node to a candidate link point, a new bearing difference and an assigned error minimization function according to the new shortest link, the new vehicle track course difference, the new vertical distance from the new target node to the candidate link point and a new weight coefficient value corresponding to the new bearing difference;

s434: judging whether each weight coefficient value is the weight coefficient with the minimum error percentage, if so, entering step S435; otherwise, go to step S436;

s435: outputting each of the weight coefficient values;

s436: and optimizing the new shortest link, the new vehicle track course difference, the vertical distance from the new target node to the candidate link point and the new bearing difference by using a GA algorithm, and then returning to the step S432.

Optionally, in the step S432, the error minimization function MM_errorComprises the following steps:

MM_error＝0.2W_d+0.23W_b+0.3W_s+0.08W_h+0.006W_s ²-0.0007W_dW_b

wherein, W_dIs a vertical distance weight coefficient, W_bSaid is the bearing difference weight coefficient, W_hWeight coefficient of vehicle track course difference, W_SIs the shortest link weight coefficient.

The invention has the following beneficial effects:

according to the technical scheme, namely the optimal link matching method provided by the invention, low-frequency positioning data is considered in the process of determining the shortest link, so that the actual shortest link can be determined, the authenticity of the shortest link is ensured, and in addition, the high precision of the recommended link can be further ensured by utilizing a weight and vehicle track matching algorithm.

Drawings

FIG. 1 is a flow chart of a matching method for an optimal target link according to the present invention;

FIG. 2 is a schematic diagram of a shortest link search structure provided in the present invention;

FIG. 3 is a schematic diagram of an initial map matching process;

fig. 4 is a schematic view of the structure of the bearing difference.

Detailed Description

The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.

Examples

The technical scheme for solving the technical problems is as follows:

the invention provides a matching method of an optimal target link, which is shown in figure 1 and comprises the following steps:

s1: acquiring path related information in a map;

in the invention, an A-shortest path searching algorithm is adopted to search a target link between a target starting node and a target ending node in path related information, and a plurality of target links are further processed in consideration of the reasons of economy and matching with a vehicle track, wherein the processing mode is as follows:

s3: acquiring the shortest link in all the target links;

in the present invention, the shortest link is determined based on the route continuity and the minimum cost information (the transfer minimum cost or the minimum oil consumption) of the taxi or the bus by considering the route continuity, and the shortest link in all the target links is determined.

Alternatively, the step S4 includes:

alternatively, the step S41 includes:

s411: acquiring all node segment related information in all the target links, wherein the node segment related information comprises distance information, conductivity and steering limit of the node segments;

s414: acquiring historical positioning data of a node section of a current target link, which is limited by steering, to obtain distance information of the current target link;

s415: according to the distance information of the current target link, a distance plus cost heuristic function is utilized to obtain a minimum function value of the current target link;

s419: enter the next target link and return to step S412.

Alternatively, in the step S413, the steering limitation includes: the target path comprises a low-frequency node section, and the low-frequency node section has a frequency response speed lower than that of a normal node section.

Specifically, as a specific embodiment, the above method is described:

the lowest f (x) value of the path is determined using a distance plus cost heuristic, f (x). For road networks, f (x) is the sum of two functions:

(1) the path cost function g (x) is the distance from the originating node to the current node.

(2) The heuristic evaluation function h (x) is the distance from the current node to the target node.

The path segment with the lower cost (i.e., the minimum of f (x)) is selected as the shortest path and is continuously processed until the destination node is reached. For map matching, connectivity between road segments and turn restriction information at intersections need to be utilized in shortest path search. Fig. 2 illustrates the processing of the a-search algorithm in the case of the map matching processing. In the figure, the vehicle travels from node a to node F in the hypothetical network, where links BA and FC are unidirectional segments. The treatment was as follows:

the first step is as follows:

(1) by considering the connectivity:

f(AC)＝10+22，f(AB)＝18+12，f(AE)＝40+30

(2) considering the steering constraints:

f (ac) 10+22 and f (ae) 40+30

f (AB) even with the lowest f (x) value is removed from the candidate set. This is because AB is a unidirectional segment and the vehicle cannot travel from node a to node B.

The second step is that:

(1) by considering the connectivity:

f(CF)＝30+0，f(CB)＝20+12，f(AE)＝40+30

(2) by considering the steering limits, F (CF) is removed because CF is a one-way segment and the vehicle cannot travel from node C to node F.

The third step:

(1) by considering the connectivity:

f(BF)＝32+0，f(BD)＝30+10，f(AE)＝40+30

(2) neither of them is deleted in view of the steering restrictions.

Thus, the shortest path for the vehicle from node A to node F is AC → CB → BF. For the map matching algorithm, the shortest path algorithm also helps to reduce the number of candidate links, thereby improving efficiency, accuracy and reliability. If the shortest path between the candidate link and the previously matching link is not found, the candidate link should be deleted from the set of candidate links.

s43: obtaining an optimal weight coefficient corresponding to the short link, the vehicle track course difference, the vertical distance from the target node to the candidate link point and the bearing difference according to the shortest link distance of the shortest link, the vehicle track course difference, the vertical distance from the target node to the candidate link point and the bearing difference;

Optionally, in step S42, the vehicle track heading difference HD is:

HD＝W_hf(Δθ)

wherein, W_hF (delta theta) is a positive function and f (delta theta) ═ cos (delta theta) |, delta theta ═ theta-theta |, and f (delta theta) |, is a weight coefficient of the vehicle track heading difference_jI, theta is the vehicle track course from the starting node to the target node, theta_jThe vehicle track heading of the starting node to the candidate link.

PD＝W_df(d_j)

Optionally, in step S42, the bearing difference BD is:

BD＝W_bf(Δα)

wherein, W_bSaid is a bearing difference weight coefficient, f (Δ α) is a positive function of the bearing difference and f (Δ α) ═ cos (Δ α) |, Δ α ═ a-a |_jI, a is the orientation of the vehicle (relative to north), a_jIs a candidate link c_jIn the direction of the axis of rotation.

Alternatively, the step S43 includes:

s435: outputting each of the weight coefficient values;

TWS＝W_df(d_j)+W_bf(Δα)+W_sf(Δl)+W_hf(Δθ)

Of which, four weight coefficients (i.e., W)_d、W_b、W_s、W_h) For providing the relative importance of the different weight scores, the sum of the four weight coefficients should therefore be equal to a fixed value, such as 100. There is a need for a method to optimally estimate these weighting parameters, the values of which can be obtained by an optimization analysis. Therefore, the total weight fraction (TWS) is as shown in the above equation.

MM_error＝0.2W_d+0.23W_b+0.3W_s+0.08W_h+0.006W_s ²-0.0007W_dW_b

wherein, W_dIs a vertical distance weight coefficient, W_bSaid is the bearing difference weight coefficient, W_hWeight coefficient, W, of the vehicle track course difference_SIs the shortest link weight coefficient.

Specifically, because the shortest path uses a, the search algorithm between the start node and the end node needs to be calculated, and the IMP execution needs two epochs of positioning data, as shown in fig. 3, there are two consecutive positioning points, denoted as P₁And P₂Vehicle position P_i(i ═ 1,2) includes a number of attributes, such as the eastern coordinate x_iNorth coordinate y_iVelocity v of the vehicle_iVehicle bearing a_iThis can be expressed as:

P_i＝{x_i,y_i,v_i,a_i}

IMP targets second position P₂Matching to the correct route section, not in the first position P₁The correct road segment is identified. This is to avoid the need for a limited amount of information in P₁The likelihood of a false match to a map match. At P₂There are four candidate links within the confidence interval of (a).

This can be expressed as:

C_j＝{AB,BC,CD,DA}

where j is the number of candidate segments. Among them candidate link C_jThe above can find 4 candidate map matching locations:

confidence interval is represented by P₂Is determined by the GPS positioning quality. Matching locations from these candidate maps

To a first position fixed position P₁The 4 shortest paths can be determined using a half search algorithm. These 4 paths may share some common links as indicated by the red arrows in figure 3,or may be completely independent of each other. Two new weights relating to shortest path and vehicle trajectory are introduced. Once the shortest path is calculated, each executable link may be assigned 4 different weights to identify the correct link. They are respectively: (1) a weight of a vertical distance (PD); (2) weight of bearing difference; (3) a weight of shortest path distance (SD); (4) weight of vehicle track Heading Difference (HD). Each weight is expressed as a function:

PD＝W_df(d_j)

BD＝W_bf(Δα)

SD＝W_df(Δl)

HD＝W_df(Δθ)

w_ithe weight coefficients represent the relative importance of the different weights. The weight of the vertical distance PD is d_jA function of, i.e. P₂To candidate link C_jPerpendicular distance of d_jThe smaller the weight, the higher the weight. Thus, f (d)_j) Is a vertical distance C_jA positive function (ranging from 0 to 1) directly affects the weighted score of PD. Is defined as:

the smaller the vertical distance, the higher the weighted score of PD, i.e., if P₂The closer a point is to a candidate link, the greater the likelihood that the link will become the correct link. A candidate link with a vertical distance above 200m produces a weighted score of 'zero' indicating that the fixed distance segment is located farther away, and therefore, this large distance becomes less important in identifying the correct segment. f (Δ α) is a positive function of the bearing difference (ranging from 0 to 1), directly affecting the weighted score of PD:

f(Δα)＝|cos(Δα)|

wherein Δ α ═ a-a_jL. The absolute value of f (Δ α) is used here to reduce the angular difference from (0-180 °) to (0-90 °). This is different from existing map matching algorithms. The reason is briefly explained by means of fig. 4, as mentioned before, existing map matching algorithms are typically designed for positioning data with a frequency of 1 Hz.In FIG. 4a, if P₁And P₂Is derived from a set of 1Hz positioning data, P₁Match on link AB (i.e., P)₁And P₂). The vehicle is to travel from node a to node B.

Thus, the bearing of the connecting rod AB should be AB, i.e. it is

(associated with the north). Suppose the vehicle bearing is P₂Of (2)

(in relation to north), the bearing difference between the vehicle heading and the link AB is 60 °. Then, according to the existing map matching algorithm, a weight of cos (60 °) 0.5 is assigned to the link AB. However, if P₁And P₂Is derived from the low frequency location data set (as shown in fig. 4B), there is a priori an indication that the vehicle is traveling from node a to node B, or vice versa. The vehicle may pass through either of the two nodes of the AB link. The bearing of the connecting rod AB can be

Or can be

Thus, the bearing differential may be 60 ° or 240 °. The weight of the link AB may be cos (60 °) 0.5, or cos (240 °) 0.5.

The algorithm used gives a higher weight to the low load difference of the AB link, i.e. | cos (60 °) | ═ cos (240 °) | 0.5, considering that the vehicle can pass through either of the two nodes of the AB link in the case of low frequency data. Therefore, the range of the angle difference is narrowed to (0-90 °).

SD＝W_df (delta l) is the difference between the distance between two consecutive position fixes on the candidate link and the shortest path distance from the first position fix to the current map matching point, l represents the Euclidean distance between the two position fixes, l_jRepresents from

To P₁The shortest path distance of (2). Therefore, a function of the distance difference in the SD weighted score is proposed:

wherein Δ l ═ l-l_jThe function value of | and f (delta l) is between 0 and 1. For differences in which the distance is small, higher weight should be given. 1000 meters as the threshold for this function is also used for the shortest path search process. If the distance plus the cost heuristic function in the shortest path search algorithm exceeds 1000m plus the distance between two continuous fixed positions, the shortest path search algorithm stops the search process, and the weighted score function value f (delta l) of the SD is assigned to zero for the candidate link. This threshold is very useful for shortest path searching because it increases processing time and therefore the algorithm is suitable for real-time applications. This threshold may be dynamically adjusted based on the frequency of the positioning data, further increasing the processing time of the algorithm (e.g., a smaller threshold may be assigned to higher frequency data).

(1) Is theta and theta_jWeighted scoring of directional differences, where θ is P₁And P₂And therefore f (Δ θ) is a positive function (ranging between 0 and 1) and takes the form:

f(Δθ)＝|cos(Δθ)|

where Δ θ ═ θ - θ_jL. A high weight should give a smaller weight. The range of angular differences is also reduced to (0-90 deg.) similar to the bearing differences.

(2) Four weight coefficients (i.e., W)_d，W_b，W_s，W_h) The relative importance of the different weight scores is provided so that the sum of these four weight coefficients should equal a fixed value of 100. The values of these weights may be obtained by optimization analysis or empirical analysis. Thus, a Total Weight Score (TWS) is given:

TWS＝W_df(d)+W_bf(Δα)+W_sf(Δl)+W_hf(Δθ)

an optimization technique based on genetic algorithms is employed. The error in identifying the correct link in the map matching process depends largely on the values of these four weight coefficients. If we were able to access a reference data set with real inputs and outputs, a set of coefficient values would result in a certain percentage error of correct link identification. The percentage of map matching error and the corresponding weight coefficient may form a functional relationship. Thus, the GA algorithm may be employed to identify the value of the weighting factor that minimizes the percentage of map matching errors. This becomes an error minimization problem and two boundary conditions can be imposed.

Using a reference input output map matching dataset consisting of 2130 GPS epochs, the following error minimization function is obtained.

MM_error＝0.2W_d+0.23W_b+0.3W_s+0.08W_h+0.006W_s ²-0.0007W_dW_b

In the GA optimization process, 40% of crossing rate and 60% of variation rate are adopted, and the population size is 50 (uniformly distributed between 1-100). After 1230 generations, the function value (fitness value) reaches the global optimum value.

Finding the optimal weight coefficient value as W_d＝32，W_b＝21，W_s35 and W _h12. It can be seen that 50% of the total weight is slightly more than the conventional PD and BD weights, which means that both weights are very useful for all frequency data. Considering that GPS heading is poor at very low vehicle speeds (less than 3m/s), higher weight was found to be associated with PD. Only 35% of the total weight is allocated to the shortest path distance because the vehicle may not always travel on the shortest path. In this case, other weights have the ability to correct for possible errors caused by the SD weight. And giving the link with the highest weight as a correct link among the candidate links, wherein the map matching position on the candidate link is used as the map matching position on the current road.

The invention has the following beneficial effects:

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A matching method for an optimal target link is characterized by comprising the following steps:

s1: acquiring path related information in a map;

s3: acquiring the shortest link in all the target links;

2. The method for matching an optimal target link according to claim 1, wherein the step S4 comprises:

3. The method for matching an optimal target link according to claim 2, wherein the step S41 comprises:

s419: enter the next target link and return to step S412.

4. The method for matching an optimal target link according to claim 3, wherein in the step S413, the steering limitation comprises: the target path comprises a low-frequency node section, and the frequency response speed of the low-frequency node section is lower than that of a normal node section.

5. The method for matching an optimal target link according to claim 3, wherein in the step S415, the distance plus cost heuristic function f (x) is:

f(x)＝g(x)+h(x)

6. The method for matching an optimal target link according to claim 2, wherein in the step S42, the vehicle track heading difference HD is:

HD＝W_hf(Δθ)

7. The method according to claim 2, wherein in step S42, the vertical distance PD from the target node to the candidate link point is:

PD＝W_df(d_j)

8. The method for matching an optimal target link according to claim 2, wherein in the step S42, the bearing difference BD is:

BD＝W_bf(Δα)

wherein, W_bSaid is the bearing difference weight coefficient, f (Δ α) is a positive function of the bearing difference and f (Δ α) ═ cos (Δ α) |, Δ α |, f (Δ α) is a positive function of the bearing difference＝|a-a_jL, a is the orientation of the vehicle (relative to north), a_jIs a candidate link c_jIn the direction of the axis of rotation.

9. The method for matching an optimal target link according to any one of claims 2 to 8, wherein the step S43 comprises:

s435: outputting each of the weight coefficient values;

10. The method for matching the optimal target link of claim 9, wherein in step S432, the error minimization function MM is used as the error minimization function_errorComprises the following steps:

MM_error＝0.2W_d+0.23W_b+0.3W_s+0.08W_h+0.006W_s ²-0.0007W_dW_b