A path search method for dynamic traffic network

Abstract

The application provides a path search method of a dynamic traffic network, relates to the fields of traffic network graph search and traffic modeling, and particularly relates to a path search method in a real large-scale traffic network environment. The application provides a high-efficiency path search method suitable for a real large-scale traffic network, which comprises importing and storing original traffic network graph data, constructing a road network preprocessing method capable of accepting road segment and intersection turning function impedance and a path query method; secondly, a road segment and intersection turning impedance function calculation formula corresponding to the path search method is provided; finally, a dynamic updating strategy of the road network preprocessing method in a dynamic traffic network background is provided. Compared with the prior art, the application has the advantages of high path search efficiency, good road network impedance editability and dynamic impedance road network support.

Description

Technical Field

[0001] This invention relates to the fields of traffic network graph search and traffic modeling, and in particular to a path search method for real large-scale traffic network environments. Background Technology

[0002] In recent years, with the increase in the frequency of travel per capita, the question of how to provide better route recommendation methods at the level of urban traffic management has received widespread attention.

[0003] Currently, large-scale traffic network path search still faces the following challenges: (1) In terms of scalability, the algorithm efficiency is low when using methods such as arc marking and line graphs to simulate intersection turning impedance, and it also poses certain challenges to data storage; (2) In terms of real-time data processing, the use of preprocessing-free methods based on Bellman-Ford, Dijkstra, etc., results in low efficiency in the path query stage; (3) In terms of accuracy, heuristic acceleration methods such as A* have low accuracy in querying and low adaptability to traffic networks with different structures. Therefore, how to improve the adaptability of path search methods under the premise of real large-scale road networks, and how to use more efficient search algorithms, more flexible road network impedance, and more proactive real-time strategies to handle real path search needs, has become a problem that needs to be solved in this field. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a path search method that is more efficient in querying large-scale transportation networks, easier to express various impedance information, and supports dynamic updates of road network impedance compared to the existing technology. This makes it better suited to the following requirements of path planning in large-scale transportation networks: (1) the requirement of large-scale networks for the efficiency of path search algorithms; (2) the requirement of variable impedance of various parts of the road network due to the variability of traffic flow and user demand in real scenarios; and (3) the requirement of the timeliness of path search due to the dynamic changes in road network impedance.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A path search method for a dynamic traffic network, comprising the following steps:

[0007] Obtain raw traffic network map data, including road network topology, basic road network attributes, road segment impedance, and intersection turning impedance;

[0008] The hierarchy of nodes in the original road network is obtained. Using the principle of hierarchical shrinkage, all nodes in the original road network are shrunk according to the node hierarchy order. Each shrinkage process obtains shortcut edges and performs lazy updates of the node hierarchy.

[0009] After the shrinking process is completed, the order in which the nodes shrink is recorded as the final level used for querying and added to the original road network along with all the generated shortcut edges to generate the road network for querying.

[0010] When the local impedance information of the road network changes, the road network used for querying is dynamically updated. Specifically, this includes: finding all shortcut edges whose impedance has changed; for the shortcut edges found, removing the original path mapping corresponding to the shortcut edges; and updating the original path mapping using Dijkstra's algorithm that takes into account turning impedance.

[0011] Based on the road network used for the query, a bidirectional Dijkstra graph search algorithm based on node level, road segment impedance, and intersection turning impedance is used to search for the optimal path.

[0012] As a preferred technical solution, information processing is performed on the original traffic network map, specifically including:

[0013] Construct edge numbering attributes using the connection relationships of nodes and add them to the basic attributes of the road network;

[0014] The ratio of edge length to edge free-flow velocity is used as the basic impedance value based on the road segment travel time and added to the basic attributes of the road network, while the edge length and edge free-flow velocity attributes are removed.

[0015] As a preferred technical solution, the method employs the Dijkstra graph search algorithm and adds search logic considering turning impedance during the contraction and query phases of hierarchical contraction, specifically as follows:

[0016] In the preprocessing stage, the impedance value is calculated based on the impedance function expression and added to the properties of the edge or node;

[0017] Regarding the road segment impedance, an impedance function with a variable number of parameters is input;

[0018] Regarding turning impedance, the turning impedance is taken into account when considering node neighbors in Dijkstra's algorithm:

[0019] Determine if the current node is an intersection node;

[0020] If so, obtain the impedance table and predecessor node number corresponding to the node, and when considering the neighbors of the node, use the predecessor node-current node-successor node to query and obtain the corresponding turning impedance in the above impedance table.

[0021] If not, the turning impedance is recorded as 0;

[0022] Calculate the sum of the segment and turning impedances from the current node to its neighbors. If the sum is smaller than the known minimum impedance, update the shortest path estimate for that neighbor.

[0023] As a preferred technical solution, the method adds consideration logic for turning impedance when adding or augmenting shortcut edges during the hierarchical contraction stage, specifically:

[0024] When adding or augmenting shortcut edges, search for the predecessor and successor nodes of the current target node in the original road network;

[0025] When determining whether to add the turning impedance at the intersection when combining the shortcut edges and calculating the total impedance, the specific steps are as follows:

[0026] Determine if the current target node is an intersection node. If so, add the turning impedance to the total impedance of the shortcut edge, obtain the impedance table corresponding to the current target node, and use the predecessor node-current node-successor node query in the impedance table to obtain the corresponding turning impedance and add it to the total impedance.

[0027] If not, then no steering impedance is added.

[0028] As a preferred technical solution, the method expresses the road network impedance in terms of the overall travel delay, specifically as follows:

[0029] The delays at each road segment and intersection are calculated and added to the corresponding attributes in the road network to express the delay impedance. The delays of individual turning arcs are accumulated to obtain the total delay of the entire route.

[0030]

[0031] d ij For mileage loss:

[0032]

[0033] ε ij The average congestion index:

[0034]

[0035] v ij The average of the desired driving speed:

[0036]

[0037] Among them, v mi ε represents the ideal driving speed for road segment i; i ,ε j Indicates the congestion index corresponding to road segment i and road segment j; v min The value 't' represents the minimum speed required to pass through the intersection. This indicates the expected travel speed of the entrance and exit lanes after considering delays.

[0038] As a preferred technical solution, the dynamic update follows the following rules:

[0039] The order of contraction determined by hierarchical contraction will not change, that is, the number of shortcut edges generated, the source and sink of each shortcut edge will not change;

[0040] When the local impedance information of the road network changes, it is necessary to find all shortcut edges whose impedance has changed in a certain way and update them.

[0041] Dynamic updates are replaced by lazy updates, which only respond in real time when the network impedance changes.

[0042] As a preferred technical solution, when the object of impedance change is a road segment and the direction of impedance change is increasing, the quick edge lookup that needs to be updated in the dynamic update is specifically as follows:

[0043] Query all shortcut edges related to the changed road segment, and update the original path mapping and impedance of the shortcut edges.

[0044] As a preferred technical solution, when the object of impedance change is a road segment and the direction of impedance change is decreasing, the quick edge lookup that needs to be updated in the dynamic update is specifically as follows:

[0045] Find the nearest intersection node in the topology on both sides of the source and sink of the road segment, query all shortcut edges and their superpaths related to the intersection node, and update the original path mapping and impedance of the shortcut edges.

[0046] The quick edge search is specifically defined by the following types: quick edges with the intersection node on the source side of the road segment as the source point and quick edges with the intersection node on the sink side of the road segment as the sink point.

[0047] As a preferred technical solution, when the object of impedance change is a turning point at an intersection, and the direction of impedance change is increasing, the quick edge lookup that needs to be updated in the dynamic update is specifically as follows:

[0048] Query all shortcut edges related to the turning arc segment, and update the original path mapping and impedance of the shortcut edges.

[0049] As a preferred technical solution, when the object of impedance change is a turning point at an intersection, and the direction of impedance change is decreasing, the quick edge lookup that needs to be updated in the dynamic update is specifically as follows:

[0050] Query the topologically nearest intersection nodes on both sides of the turning arc segment. If intersection nodes are successfully found on both sides, query all shortcut edges and their superpaths related to the intersection nodes, and update the original path mapping and impedance of the shortcut edges.

[0051] The shortcut edge search is accurate to the following types: shortcut edges with the source point of the intersection node on the inlet side of the arc segment as the source point, and shortcut edges with the sink point of the intersection node on the outlet side of the arc segment as the sink point.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1) This invention utilizes the hierarchical contraction approach to improve a path search method suitable for real large-scale transportation networks, thereby significantly reducing path query time and improving path search efficiency.

[0054] 2) By functionalizing various traffic network impedances, this invention enables the corresponding path search method to be applied to real and diverse demand scenarios, and has strong scalability.

[0055] 3) Based on graph theory, this invention adds a dynamic update strategy applicable to dynamic impedance road network graphs to the static road network preprocessing method, thereby enhancing the applicability of the path search method in dynamic traffic networks. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the method flow in an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the road network map data management method in an embodiment of the present invention;

[0058] Figure 3 This is an adaptive improvement strategy for the Dijkstra graph search algorithm in the embodiments of the present invention;

[0059] Figure 4 This is an adaptive improvement strategy for quick edge synthesis in the embodiments of the present invention;

[0060] Figure 5 This is a schematic diagram of the mileage loss calculation method in an embodiment of the present invention;

[0061] Figure 6 This is a schematic diagram of the quick edge update lookup in an embodiment of the present invention when the impedance of certain road segments increases;

[0062] Figure 7 This is a schematic diagram of a quick edge update lookup for certain road segments when the impedance is reduced, as described in an embodiment of the present invention.

[0063] Figure 8 This is a schematic diagram illustrating the quick edge update lookup in an embodiment of the present invention when the turning resistance at certain intersections increases;

[0064] Figure 9 This is a schematic diagram of a quick edge update lookup in an embodiment of the present invention for situations where the turning impedance at certain intersections is reduced.

[0065] Figure 10 This is a schematic diagram of the road network used for testing the path search method in this embodiment of the invention;

[0066] Figure 11 This is a schematic diagram illustrating the differences between the original diagram and the line diagram topology in an embodiment of the present invention;

[0067] Figure 12 This is a diagram showing the error test results based on the test road network in an embodiment of the present invention;

[0068] Figure 13 This is a graph showing the efficiency test results based on the test road network in an embodiment of the present invention. Detailed Implementation

[0069] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0070] This invention proposes a path search method for the field of large-scale traffic network graph search. Compared with existing technologies, this method has higher query efficiency in large-scale traffic networks, is easier to express various impedance information, and supports dynamic updates of road network impedance. This makes it better suited to the following requirements of large-scale traffic network path planning: (1) the requirement of large-scale networks for the efficiency of path search algorithms; (2) the requirement of variable impedance of various parts of the road network due to the variability of traffic flow and user demand in real scenarios; and (3) the requirement of timeliness of path search due to the dynamic changes in road network impedance.

[0071] The objective of this invention can be achieved through the following technical solutions:

[0072] This paper provides a highly efficient path search method applicable to real-world large-scale transportation networks, including the following:

[0073] Obtain the original traffic network map data;

[0074] Based on the original traffic network map data, a road network preprocessing method and a path query method that can accept the functionalized impedance of road segments and intersections are constructed.

[0075] Formulas for calculating the turning impedance functions of road segments and intersections corresponding to the path search method are proposed. In the context of dynamic traffic networks, a dynamic update strategy for the query map obtained from the road network preprocessing method is also proposed.

[0076] The path lookup method is as follows: Based on the road network used for the query, a bidirectional Dijkstra graph search algorithm based on node level, road segment impedance, and intersection turning impedance is used to search for the optimal path.

[0077] Specifically, for the original traffic network map data, the additional node attributes and edge attributes required for path search in this invention are calculated, and the original traffic network map data is expanded using the additional attributes to obtain expanded data; the expanded data is then stored in a computer for computer programs to read.

[0078] The process of the road network preprocessing method is as follows: In the initial road network containing topology, road segment impedance and intersection turning impedance, the node hierarchy is obtained using a heuristic method. Then, the principle of hierarchy shrinkage is used to shrink all nodes on the initial road network according to the node hierarchy order. Each shrinkage process obtains shortcut edges and performs lazy updates of the node hierarchy.

[0079] After the entire shrinking process is completed, the order in which the nodes shrink is recorded as the final level used for querying, and all generated shortcut edges are added to the initial road network to generate the query road network. The network is then stored in the computer according to a suitable data structure for the computer program to read.

[0080] The impedance function is a general impedance function that can express various positive impedance values ​​on road sections and at intersections.

[0081] The dynamic update strategy of the road network preprocessing method is a local, inert, fast edge update strategy when partial impedance changes occur in the road network, in order to avoid a complete restart of the road network preprocessing method. The strategy includes:

[0082] When the object is a road segment and the impedance change direction is increasing, query all shortcut edges related to the road segment with the change, and update the original path mapping and impedance of the shortcut edges.

[0083] When the object is a road segment and the impedance change direction is decreasing, query the nearest intersection node on both sides of the road segment in terms of topology, query all shortcut edges and their superpaths related to the intersection node, and perform original path mapping update and impedance update of the shortcut edges.

[0084] When the object is a turn at an intersection and the impedance changes in the direction of increase, query all shortcut edges related to the turn segment, and update the original path mapping and impedance of the shortcut edges.

[0085] When the object is a turning point at an intersection and the impedance change direction is decreasing, query the topologically nearest intersection node on both sides of the turning arc segment, query all shortcut edges and their superpaths related to the intersection node, and perform original path mapping update and impedance update for the shortcut edges.

[0086] Example

[0087] like Figure 1As shown, this embodiment provides a path search method suitable for large-scale dynamic traffic networks. The input of this method is the original traffic network map data and impedance information, and the output is the network map data with added node hierarchy and shortcut edges. Since the path search method supports dynamic impedance information, a corresponding dynamic update strategy will be adopted to update the output network map data after the impedance information changes.

[0088] As shown in Table 1, this embodiment requires the following information from the original traffic network map: network topology, basic network attributes, and dynamic network attributes.

[0089] Table 1. Information Categories That Should Be Included in the Original Traffic Network Map

[0090]

[0091] This embodiment processes the information from the original traffic network map into data that can be directly used in subsequent steps. The specific method is as follows:

[0092] (1) Construct the edge numbering attribute using the connection relationship of nodes, i.e. (starting node number, ending node number), and add it to the basic attributes of the road network;

[0093] (2) The ratio of edge length and edge free flow velocity is used as the basic impedance value based on the road segment travel time and added to the basic attributes of the road network, while the edge length and edge free flow velocity attributes are removed.

[0094] This embodiment utilizes a structure to store basic information about nodes and edges, while using a hash table to manage the attributes of nodes and edges for subsequent steps. The management method is as follows: Figure 2 As shown.

[0095] This embodiment is based on the principle of hierarchical contraction. It adds algorithm logic that considers turning impedance in its preprocessing and query stages, and performs functional processing on the road network impedance to support the path search method that can accept the functional impedance of road segments and intersections.

[0096] The improvement strategy in this embodiment consists of two parts:

[0097] (1) The Dijkstra graph search algorithm involved in the shrinking and querying phases of hierarchical shrinkage is impedance-functionalized, and search logic considering turning impedance is added, such as... Figure 3 As shown. During the initialization phase of the preprocessing computer program, the impedance value is calculated based on a general impedance function expression and added to the attributes of the edge or node. The general impedance function is:

[0098]

[0099] Where f is the segment impedance, c0 is the basic impedance value, and c1,...,c n (n≥1) represents the additional impedance value, α i (i = 1, 2, ..., n) represents the coefficients of the impedance values. The specific improvement strategy is as follows: For segment impedance, an impedance function with a variable number of parameters is passed into the preprocessing computer program; for turning impedance, the turning impedance is considered when Dijkstra's algorithm considers node neighbors: Known information is used to determine if the current node is an intersection node. If so, the impedance table and predecessor node number are obtained, and when considering the node's neighbors, the corresponding turning impedance is retrieved from the impedance table using the predecessor node-current node-successor node approach; otherwise, the turning impedance is recorded as 0. The sum of the segment and turning impedances from the current node to its neighbors is then calculated. If this sum is smaller than the known minimum impedance, the shortest path estimate for that neighbor is updated.

[0100] (2) For impedance calculations involved when adding or augmenting quick edges during the hierarchical contraction phase, add logic considering turning impedance, such as... Figure 4 As shown. The specific strategy for improvement is as follows: When adding or augmenting shortcut edges, search for the predecessor and successor nodes of the current target node in the original road network, and determine whether it is necessary to add the turning impedance at the intersection when synthesizing the shortcut edge and calculating the total impedance. The relevant judgment and implementation method is as follows: Use known information to determine whether the current target node is an intersection node. If so, it is considered that the turning impedance needs to be added to the total impedance of the shortcut edge. At this time, obtain the impedance table corresponding to the current target node, look up the corresponding turning impedance in the impedance table using predecessor node-current node-successor node, and then add it to the total impedance; otherwise, do not add the turning impedance.

[0101] To facilitate subsequent verification in real-world scenarios, this embodiment will take comprehensive travel delays as an example to demonstrate the method for expressing road network impedance based on this specific demand scenario.

[0102] The comprehensive delay defined in this embodiment refers to the difference between the actual travel time of a vehicle and the travel time of a vehicle traveling under ideal conditions throughout the entire journey. It mainly consists of two parts: first, the time loss caused by the difference between the actual speed and the ideal speed (maximum speed limit) when the vehicle is traveling through road segments and intersections, i.e., travel delay; second, the waiting time loss caused by traffic signal timing when the vehicle is passing through intersections, i.e., waiting delay.

[0103] The definition of comprehensive delay is the ratio of mileage loss incurred when a vehicle's actual travel speed is lower than the ideal speed to the speed under ideal conditions. The calculation of this mileage loss for a single turning arc along the entire route is illustrated below. Figure 5As shown in the diagram, a single turning arc segment is defined as: when passing through an intersection, the portion of the road segment containing the approach lane and exit lane that is associated with the intersection (hereinafter referred to as road segment I and road segment J, both being portions of their respective road segments closest to the intersection), and the turning portion formed by traveling from road segment I to road segment J, together constitute the arc segment. Therefore, by accumulating the delays of individual turning arc segments, the overall delay value for the entire travel process can be obtained.

[0104] Figure 5 In the middle, v mi ,v mj Indicates the ideal driving speed for road segment i and road segment j. This represents the expected speed after considering delays, v. min This represents the minimum speed (v) required to pass through the intersection. min ≥0), t represents the duration of maintaining the minimum speed at the intersection, v min When t is 0, it includes waiting delays. Therefore, d ri ,d rj ,d i ,d j This can represent the mileage loss caused by fluctuations in the actual speed during the single steering arc.

[0105] To express the expected travel speed after considering delays, the concept of a congestion index needs to be defined. In this embodiment, the congestion index ε is defined as a coefficient between 0 and 1: if the maximum speed limit for a certain road segment is v... m The corrected expected travel speed is:

[0106]

[0107] In practice, the congestion index is usually obtained from real road vehicle tests: the platform collects the average speed of vehicles passing through a certain road segment and matches it with the speed limit of that road segment to obtain the congestion index value, and updates it at a certain frequency.

[0108] Therefore, the expression of the overall delay in the road network impedance should be divided into two parts. The first part is the road segment part, and the impedance of the road segment can be expressed as an impedance function that includes the congestion index. For the road segment part of a single turning arc, the simplest expression is:

[0109]

[0110] Among them, l i ,l j ε represents the lengths of road segments i and j. i ,ε jThe first part represents the congestion index corresponding to road segment i and road segment j. The second part is the turning part of the intersection. The turning impedance can be expressed as an impedance function that includes the congestion index and mileage loss. For a single turning process, assume that the acceleration values ​​during deceleration and acceleration tend to be constant, and their absolute values ​​are a and a, respectively. i ,a j The mileage loss can then be estimated as follows:

[0111]

[0112] Among them, v min And t can be set based on reference values: for non-signaled intersections, v min According to the "Urban Road Intersection Planning Code" (GB50647-2011), the speed limit for straight-ahead traffic can be set to 0.7 times the maximum speed limit, for left turns to 0.5 times the maximum speed limit, and for right turns to be set to 30 km / h or 20 km / h depending on the intersection level (only when the maximum speed limit is higher than this value); for signalized intersections... min When there is no parking behavior, it is the same as that of a non-signaled intersection; when there is parking behavior, it is set to 0. t can be set to 5s at a non-signaled intersection, and for signaled intersections, it depends on the specific signal timing. Traffic timing at signalized intersections can usually be obtained from actual traffic data. However, for intersections where it is difficult to obtain accurate timing information, the following simulation method can be used for calculation: First, determine whether it is a signalized intersection based on the in-degree of the intersection's center node and the intersection's gradation. If so, determine its phase type as well. Then, calculate the cycle according to the Webster method, using a formula with added correction coefficients, i.e.:

[0113]

[0114] In the formula, k is the number of phases; L s The default value for the start-up loss time is 4 seconds; AR is the full red light time, with a default value of 0; Y is the sum of the flow ratios, i.e., the sum of the maximum flow ratios of all signal phases constituting the cycle, with a default value of 0.85. Then, the green light time for each phase is calculated based on the flow ratio relationship, and the stopping time for each phase, i.e., the minimum speed duration t, is obtained using the following formula:

[0115]

[0116] If the acceleration and deceleration times and efficiencies during the steering process are similar, then the average expected driving speed can be estimated as the average expected driving speed of the approach and exit lanes, that is:

[0117]

[0118] Similarly, the mean congestion index can be estimated as follows:

[0119]

[0120] The simplest expression for the delay of a single steering process is:

[0121]

[0122] The delays at each road segment and intersection are calculated using the aforementioned expression and added to the corresponding attributes in the road network, thus completing the expression of the delay impedance. The total delay for the entire route is obtained by summing the delays of each individual turning segment.

[0123]

[0124] This embodiment proposes a dynamic update strategy for the road network preprocessing method in the context of a dynamic traffic network. To ensure the accuracy of the dynamic update strategy, the following rules need to be followed: (1) The hierarchical shrinkage order determined during the first preprocessing should not change, that is, the number of generated shortcut edges, the source point and sink point of each shortcut edge should not change. Under this premise, the accuracy of the search results after dynamic update can be ensured; (2) When the local impedance information of the road network changes, it is necessary to find all shortcut edges whose impedance may change in a certain way and update them. For different change situations (impedance increase or decrease, change of road segment impedance or intersection turning impedance), the search strategy for the affected shortcut edges is different; (3) Dynamic update is essentially an inertial update, and only when the road network impedance changes does it respond in real time.

[0125] Once the shortcut edge search is complete, the update strategy is consistent: remove the original path mapping corresponding to the shortcut edge (i.e., the complete path actually traversed by the shortcut edge from its source to its sink on the original traffic network), and then update the original path mapping using the Dijkstra algorithm considering turning impedance. Analysis shows that the shortcut edge search strategy differs for different delay variations. Therefore, this embodiment provides corresponding shortcut edge search strategies for four different scenarios.

[0126] (1) When the impedance of certain road segments increases, only the shortcut edges involving those segments (i.e., those segments included in the original path mapping) may be affected. In this case, query all shortcut edges involving those segments. A schematic diagram is shown below. Figure 6 As shown. For Figure 6 As shown in the diagram, the shortcut edge to be updated is the shortcut edge that contains edge AB in the original path mapping.

[0127] (2) When the impedance of certain road segments decreases, the shortcut edges involving those segments remain valid. In this case, attention should be paid to shortcut edges that might lead to detours through those segments after the impedance decreases. The method for finding such shortcut edges is as follows: from the source and sink of the road segment, find the nearest intersection node in the topology on the other side. Depending on the location of the road segment in the traffic network, there may be 0, 1, or 2 such intersection nodes. If the intersection node is successfully found on the source or sink side, find the shortcut edge that uses it as an endpoint. In a directed real road network graph, to further save computational resources, the shortcut edge search can be more precise to the following two types: shortcut edges with the intersection node on the source side of the road segment as the source and shortcut edges with the intersection node on the sink side of the road segment as the sink. This is because, in a directed graph, among these two types of shortcut edges using intersection nodes as endpoints, the other types cannot lead to detours through the aforementioned road segments with reduced impedance. At the same time, it is also necessary to find the superpaths corresponding to the shortcut edges found above (i.e., shortcut edges in the original path mapping that contain the original path mapping of the above shortcut edges), and treat them as shortcut edges to be updated as well. A schematic diagram of the solution in the directed graph is shown below. Figure 7 As shown. For Figure 7 In the example of edge AB, the nearest intersection nodes on both sides of edge AB are found to be A' and B'. The shortcut edge to be updated is the shortcut edge with A' as the source or B' as the sink and its superpath.

[0128] (3) When the turning impedance at certain intersections increases, only the shortcut edges involving the arc segment where the turn occurs (i.e., the original path mapping includes this arc segment) may be affected. In this case, query all shortcut edges involving this arc segment. A schematic diagram of the solution is shown below. Figure 8 As shown. For Figure 8 The arc segment AOB that passes through the intersection node O shown is the shortcut edge to be updated, which is the shortcut edge that includes route AOB in the original path mapping.

[0129] (4) When the turning impedance of some intersections decreases, similar to the strategy for finding shortcut edges to be updated when the road segment impedance decreases, the nearest intersection node in the topology is searched on both sides of the arc segment involving the turning point. Depending on the location of the arc segment in the traffic network, there may be 0, 1, or 2 such intersection nodes. If the intersection nodes are successfully found on both sides, a shortcut edge is searched that takes it as one endpoint. In a directed real road network graph, to further save computing power, the shortcut edge search can be more precise to the following two types: shortcut edges with the source node of the arc segment's entrance road segment as the source node, and shortcut edges with the sink node of the arc segment's exit road segment as the sink node. At the same time, it is also necessary to find the hyperpath corresponding to the shortcut edge found above and treat it as a shortcut edge to be updated. The schematic diagram of the scheme in the directed graph is as follows. Figure 9As shown. For Figure 9 For the arc segment AOB passing through intersection node O, the nearest intersection nodes on the topology of A and B facing outwards from the arc segment are A' and B'. At this time, the shortcut edge to be updated is the shortcut edge with A' as the source or B' as the sink and its superpath.

[0130] This embodiment establishes a verification test based on a real traffic network for the path search method applicable to large-scale dynamic traffic networks, including query result error testing, query process efficiency testing, and dynamic update strategy effectiveness testing. The real traffic network used is the road network of Shanghai's inner ring road and surrounding area, including the Shanghai Inner Ring Elevated Road and the main urban road network inside, extending outwards by about 2 kilometers. A schematic diagram of the test road network is shown below. Figure 10 .

[0131] The verification test described in this embodiment will also use the line graph corresponding to the test road network diagram. A line graph is a concept in graph theory, defined as follows: for an original graph G, a vertex of its line graph L(G) corresponds to an edge of G, and vertices of L(G) are adjacent if and only if they are adjacent on their corresponding edges in G. The topological relationship between the original graph and the line graph is as follows: Figure 11 As shown in the figure, the structural differences between a simple directed graph containing 5 nodes and several road segments and its corresponding line graph are illustrated. Some important parameters of the test road network map and its line graph are shown in Table 2.

[0132] Table 2. Some important parameters of the test road network map and its alignment diagram.

[0133] Number of nodes 8 210 19 536 Number of road sections 19 536 52 933 Diameter 74 75 Average in-degree / out-degree 2.38 6.45

[0134] In this embodiment, the complete impedance information (including road segment impedance and intersection turning impedance) used in the test road network diagram is based on the comprehensive delay, meaning that the selection of the optimal path follows the principle of shortest time priority. To facilitate testing, the comprehensive delay information of the road network is provided through simulation calculation: the congestion index of road segments is uniformly initialized to 0.4, the turning delay of each intersection is calculated using the simulation method, the U-turn impedance is taken as 10 seconds, and all impedance values ​​used in the calculation are integers.

[0135] The reason for conducting query result error testing in this embodiment is that, when searching for the shortest path, the graph search algorithm needs to satisfy the property of no aftereffect to ensure that the result is necessarily the shortest path under the current conditions. The meaning of no aftereffect is that when a graph search algorithm that satisfies this property finds a node, it should not use information related to its predecessor node when considering its successor nodes. However, when the turning impedance that connects to the predecessor node is added to the search, the property of no aftereffect is violated. Therefore, error testing is considered.

[0136] In error testing, the search results from the line graph based on the original road network map are used as a control group. The reason for using the line graph as a control group is that the impedance at the vertex in the line graph (i.e., the edge impedance in the original graph) is a uniquely determined value and is no longer related to its predecessor node, thus satisfying the no-aftereffect property in the search process.

[0137] The specific implementation steps of the error test are as follows: (1) A large number of random (10,000 groups) routes with starting point S and ending point T that need to be searched are given; (2) The path length P of each group is obtained by using the hierarchical contraction considering turning impedance proposed in this invention. w (2) Based on the specific path in (2) above, obtain the successor node S of its starting point. s and the predecessor node T of the endpoint p And find the edge (S,S) in the online graph that matches the edge in the original graph. s ), edge (T) p The node S corresponding to T) L T L (3) Use the shortest path length obtained in the line graph as the starting and ending points of the search; (4) Combine the shortest path length obtained in the line graph with the edge (S,S) s ), edge (T) p Add the impedances of P and T to obtain the shortest path length when the search start and end points are exactly the same as the original graph, and then add it to P. w Compare and calculate the relative errors. A histogram of the relative error distribution of a large number of random test results is shown below. Figure 12 As shown, the horizontal axis represents the relative error value, and the vertical axis represents the count within the corresponding error value interval. To ensure the display effect of the graph, only the test group that generated relative error is shown, and the top 95% of the data with the smallest error are selected. After further calculation, in the test road network map, the mean relative error of the above search algorithm is 0.00373, and the median is 0. It can be considered that the error in the actual path search is small and acceptable.

[0138] This embodiment uses the comparison of the final search space size to test the efficiency of the query process, with Dijkstra's algorithm as the comparison object. Given extremely similar query logic, comparing search space is a more objective strategy than comparing query time.

[0139] The specific implementation steps of the efficiency test are as follows: (1) A large number of random (100,000 groups) routes with starting point S and ending point T that need to be searched are given; (2) The hierarchical contraction considering turning impedance and the Dijkstra algorithm considering turning impedance proposed in this invention are used as experimental algorithms (hereinafter referred to as CH experimental algorithm and Dijkstra experimental algorithm, the CH experimental algorithm is performed on the road network map after corresponding preprocessing by default), and the path query is performed on each group to obtain the specific path and the predecessor node hash table generated during the query process (containing all nodes considered in the query process); (3) The final number of road segments involved in the entire path is obtained using the specific path, and the number of nodes considered in the query stage is obtained using the predecessor node hash table; (4) The node skip rate η is calculated using the number of nodes considered, that is

[0140]

[0141] In the formula, n is the number of nodes considered, and N is the total number of nodes in the road network diagram. The higher the node skip rate, the smaller the search space and the higher the efficiency of the algorithm when performing path query; (5) Summarize the final number of road segments and node skip rate obtained by the two experimental algorithms and plot them as follows. Figure 13 The scatter plot shown represents the average node skip rate for each final road segment. A significant portion of the samples have a final road segment count below 80. This indicates that Dijkstra's algorithm performs better in the search space when the total search impedance is extremely low (meaning the start and end points are very close). However, as the total search impedance increases, Dijkstra's algorithm shows a significant downward trend in search space performance, while CH's algorithm shows a more stable performance. In regions with a large number of samples, its node skip rate is consistently above 96% with minimal fluctuations. Therefore, CH's algorithm demonstrates superior and more stable performance in large-scale road networks.

[0142] This embodiment uses an example to test the effectiveness of the dynamic update strategy. Point 4139 in the northeast (near Dongbaoxing Road Station of Shanghai Metro Line 3) and point 2 in the southwest (near Xujiahui Station of Shanghai Metro Line 9) in the test network are selected. First, the CH experimental algorithm is executed in the test network containing the initial complete impedance information to obtain the optimal path A and the corresponding optimal time consumption. Then, the changes in route parameters in the following four cases are studied: (1) By modifying the congestion index, the impedance of some road segments in path A is increased, the road network map is dynamically updated, and then the CH experimental algorithm is executed. (1) Obtain the optimal path B1 at this time; (2) Restore all the increased impedance in (1), dynamically update the road network map, and then execute the CH experimental algorithm to obtain the optimal path B2 at this time; (3) By modifying the mileage loss of turning, increase the impedance of the relevant turning at a certain intersection in path A, dynamically update the road network map, and then execute the CH experimental algorithm to obtain the optimal path B3 at this time; (4) Restore all the increased impedance in (3), dynamically update the road network map, and then execute the CH algorithm to obtain the optimal path B4 at this time.

[0143] The specific testing process is as follows: (1) In the initial road network, the optimal route A is planned, and the route is roughly as follows: road near Dongbaoxing Road Station - North-South Elevated Road Auxiliary Road - North-South Elevated Road - ramp - Yan'an Elevated Road - ramp - road near Xujiahui Station; (2) Based on (1), the congestion index of the relevant sections of Yan'an Elevated Road is changed to 0.2 (that is, the driving speed on it is changed to 50% of the original), and the optimal route B1 is planned. At this time, the route changes significantly. Instead of going through Yan'an Elevated Road, it continues south along the North-South Elevated Road and then reaches the destination via Zhaojiabang Road, etc.; (3) (2) (3) The changes in (4) are restored, and the optimal path B2 is planned, which is consistent with path A; (4) Based on (1), the north entrance of an intersection near the end of path A (Tianping Road-Guangyuan Road intersection) is modified to a congested state (i.e., the delay of turning north-west, north-south, and north-east is modified to a maximum value), and the optimal path B3 is planned. At this time, the path changes to a certain extent, and the turning occurs earlier at the intersection before the intersection of Tianping Road-Guangyuan Road (Tianping Road-Kangping Road intersection); (5) The changes in (4) are restored, and the optimal path B4 is planned, which is consistent with path A. It can be seen that when the impedance information changes, the expected changes occur when the optimal path is re-obtained in the road network map after the local update of the dynamic update strategy. It can be considered that the dynamic update strategy is effective.

[0144] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A path search method for a dynamic traffic network, characterized by the following steps: include: Obtain raw road network map data, including road network topology, basic road network attributes, road segment impedance, and intersection turning impedance; Obtain the hierarchy of the original road network nodes. Using the principle of hierarchical shrinkage, shrink all nodes on the original road network according to the node hierarchy order. For each shrinkage process, obtain shortcut edges and perform lazy updates of the node hierarchy. After the shrinking process is completed, the order in which the nodes shrink is recorded as the final level used for querying and added to the original road network along with all the generated shortcut edges to generate the road network for querying. When the local impedance information of the road network changes, the road network used for querying is dynamically updated. Specifically, this includes: finding all shortcut edges whose impedance has changed; for the shortcut edges found, removing the original path mapping corresponding to the shortcut edges; and updating the original path mapping using Dijkstra's algorithm that takes into account turning impedance. Based on the road network used for the query, a bidirectional Dijkstra graph search algorithm based on node level, road segment impedance, and intersection turning impedance is used to search for the optimal path. The method employs the Dijkstra graph search algorithm and adds search logic that considers turning impedance during the contraction and query phases of hierarchical contraction. Specifically: In the preprocessing stage, the impedance value is calculated based on the impedance function expression and added to the properties of the edge or node; Regarding the road segment impedance, an impedance function with a variable number of parameters is input; Regarding turning impedance, the turning impedance is taken into account when considering node neighbors in Dijkstra's algorithm: Determine if the current node is an intersection node; If so, obtain the impedance table and predecessor node number corresponding to the node, and when considering the neighbors of the node, use the predecessor node-current node-successor node to query and obtain the corresponding turning impedance in the above impedance table. If not, the turning impedance is recorded as 0; Calculate the sum of the segment and turning impedances from the current node to its neighbors. If the sum is smaller than the known minimum impedance, update the shortest path estimate for that neighbor. The method adds logic to consider turning impedance when adding or augmenting shortcut edges during the hierarchical contraction phase, specifically: When adding or augmenting shortcut edges, search for the predecessor and successor nodes of the current target node in the original road network; When determining whether to add the turning impedance at the intersection when combining the shortcut edges and calculating the total impedance, the specific steps are as follows: Determine if the current target node is an intersection node. If so, add the turning impedance to the total impedance of the shortcut edge, obtain the impedance table corresponding to the current target node, and use the predecessor node-current node-successor node query in the impedance table to obtain the corresponding turning impedance and add it to the total impedance. If not, then no steering impedance is added.

2. The path search method for a dynamic traffic network according to claim 1, characterized in that, The original road network map is processed, specifically including: Construct edge numbering attributes using the connection relationships of nodes and add them to the basic attributes of the road network; The ratio of edge length to edge free-flow velocity is used as the basic impedance value based on the road segment travel time and added to the basic attributes of the road network, while the edge length and edge free-flow velocity attributes are removed.

3. The path search method for a dynamic traffic network according to claim 1, characterized in that, The method expresses network impedance in terms of overall travel delay, specifically as follows: The delays at each road segment and intersection are calculated and added to the corresponding attributes in the road network to express the delay impedance. The delays of individual turning arcs are accumulated to obtain the comprehensive delay of the entire route. D : For mileage loss: The average congestion index: The average of the desired driving speed: in, Indicates road segment i Ideal driving speed; Indicates road segment i Road section j The corresponding congestion index; Indicates the minimum speed required to pass through the intersection; t Indicates the duration during which the minimum speed is maintained at the intersection; , This indicates the expected travel speed of the entrance and exit lanes after considering delays.

4. The path search method for a dynamic traffic network according to claim 1, characterized in that, The rules followed by the dynamic update include: The order of contraction determined by hierarchical contraction will not change, that is, the number of shortcut edges generated, the source and sink of each shortcut edge will not change; When the local impedance information of the road network changes, it is necessary to find all shortcut edges whose impedance has changed in a certain way and update them. Dynamic updates are replaced by lazy updates, which only respond in real time when the network impedance changes.

5. The path search method for a dynamic traffic network according to claim 4, characterized in that, When the object of impedance change is a road segment, and the direction of impedance change is increasing, the quick edge lookup that needs to be updated in the dynamic update is specifically as follows: Query all shortcut edges related to the changed road segment, and update the original path mapping and impedance of the shortcut edges.

6. The path search method for a dynamic traffic network according to claim 4, characterized in that, When the object of impedance change is a road segment, and the direction of impedance change is decreasing, the quick edge lookup that needs to be updated in the dynamic update is specifically as follows: Find the nearest intersection node in the topology on both sides of the source and sink of the road segment, query all shortcut edges and their superpaths related to the intersection node, and update the original path mapping and impedance of the shortcut edges. The quick edge search is specifically defined by the following types: quick edges with the intersection node on the source side of the road segment as the source point and quick edges with the intersection node on the sink side of the road segment as the sink point.

7. The path search method for a dynamic traffic network according to claim 4, characterized in that, When the object of impedance change is a turning point at an intersection, and the direction of impedance change is increasing, the specific quick edge lookup that needs to be updated in the dynamic update is as follows: Query all shortcut edges related to the turning arc, and update the original path mapping and impedance of the shortcut edges.

8. The path search method for a dynamic traffic network according to claim 4, characterized in that, When the object of impedance change is a turning point at an intersection, and the direction of impedance change is decreasing, the specific quick edge lookup that needs to be updated in the dynamic update is as follows: Query the topologically nearest intersection nodes on both sides of the turning arc segment. If intersection nodes are successfully found on both sides, query all shortcut edges and their superpaths related to the intersection nodes, and update the original path mapping and impedance of the shortcut edges. The shortcut edge search is accurate to the following types: shortcut edges with the source point of the intersection node on the inlet side of the arc segment as the source point, and shortcut edges with the sink point of the intersection node on the outlet side of the arc segment as the sink point.

Images