Map topology-based routing method, device, equipment and storage medium

Abstract

The application discloses a map topology-based routing method and device, equipment and a storage medium. The method comprises the following steps: constructing a map topology relationship; determining a routing rule corresponding to a target event according to the map topology relationship and a vehicle wireless communication V2X message; predicting the intersection order in the routing rule according to a first preset model, determining the intersection range influenced by the target event, and routing the message corresponding to the target event to all communication objects in the intersection range. In this way, the traffic events, signs, vehicle-end basic safety information and other information can be finely routed dynamically, the influence range is expanded, the message can be sent to the traffic participants who are not the nearest neighbors but still need to receive the message, and the overall traffic safety or traffic efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation, and particularly relates to a routing method and device based on a map topology, equipment and a storage medium. BACKGROUND

[0002] Vehicle to Everything (V2X) is a key technology of future intelligent transportation systems. It enables communication between vehicles, vehicles and base stations, and base stations, thereby obtaining real-time traffic conditions, road information, pedestrian information and other traffic information, thereby improving driving safety, reducing congestion, improving traffic efficiency, and providing vehicle entertainment information.

[0003] In related technologies, most V2X message focuses on technical innovation at the communication level, network topology design, etc., and the routing method of V2X message lacks the combination of actual traffic business scenarios, especially some message types do not only affect the participants within a certain communication distance range around them, but may have a larger influence range. Therefore, the current routing rules for V2X messages cannot meet the actual traffic business scenarios, which is not conducive to vehicle traffic control. SUMMARY

[0004] The present application provides a routing method and device based on a map topology, equipment and a storage medium, which can expand the influence range so that the message can be sent to the traffic participants who are not the nearest adjacent intersection but still need to receive the message, thereby improving overall traffic safety or traffic efficiency.

[0005] To achieve the above purpose, the technical scheme of the present application is as follows:

[0006] In a first aspect, the present application provides a routing method based on a map topology, which comprises:

[0007] constructing a map topology relationship;

[0008] determining a routing rule corresponding to a target event according to the map topology relationship and the obtained V2X message;

[0009] predicting the intersection order in the routing rule according to a first preset model, determining the intersection range affected by the target event, and routing the message corresponding to the target event to all communication objects in the intersection range.

[0010] Further, when the target event is a congestion event, the intersection order in the routing rule is predicted according to the first preset model to determine the intersection range affected by the target event, comprising:

[0011] obtaining a congestion feature variable corresponding to the congestion event;

[0012] inputting the congestion feature variable into the first preset model, and outputting a predicted congestion dissipation time by the first preset model;

[0013] performing reverse depth-first search according to a position of the congestion event to determine a sequence of intersection levels conforming to the predicted congestion dissipation time;

[0014] determining a range of intersections affected by the congestion event according to the sequence of intersection levels.

[0015] Further, for the first preset model, comprising:

[0016] obtaining a plurality of sample congestion events;

[0017] performing data processing on the plurality of sample congestion events to determine sample congestion feature variables;

[0018] inputting the sample congestion feature variables into the first regression model for model training, and determining the trained first regression model as the first preset model when a first loss function meets a convergence condition.

[0019] Further, the performing reverse depth-first search according to the position of the congestion event to determine the sequence of intersection levels conforming to the predicted congestion dissipation time comprises:

[0020] performing reverse depth-first search according to the position of the congestion event, and determining an accumulated weight value of an edge corresponding to at least one path based on a map topological relationship;

[0021] stopping the reverse depth-first search when the accumulated weight value of the edge corresponding to the at least one path exceeds the predicted congestion dissipation time, and determining at least one intersection level of reverse backtracking;

[0022] determining the sequence of intersection levels conforming to the predicted congestion dissipation time according to the at least one intersection level of reverse backtracking.

[0023] Further, the method further comprises:

[0024] when performing the reverse depth-first search according to the position of the congestion event, judging whether there is an unvisited intersection node in the map topological relationship;

[0025] when the judgment result indicates that there is the unvisited intersection node in the map topological relationship, backtracking to the position of the congestion event, and continuing to perform the step of performing the reverse depth-first search according to the position of the congestion event;

[0026] when the judgment result indicates that there is no unvisited intersection node in the map topological relationship, stopping the reverse depth-first search, and determining the at least one intersection level as the sequence of intersection levels conforming to the predicted congestion dissipation time.

[0027] Further, the method further comprises:

[0028] determining, based on a map matching algorithm, a nearest location of occurrence of the target event;

[0029] after binding the target event to the nearest location, determining a start intersection corresponding to the target event, and taking the start intersection as a main intersection of the target event; wherein the main intersection is used for intersection depth search in a routing rule.

[0030] Further, the method further comprises: predicting, according to a second preset model, intersection turning information in the routing rule, determining predicted turning information of a to-be-predicted intersection, and routing the predicted turning information to a target communication object of the to-be-predicted intersection.

[0031] Further, the predicting, according to the second preset model, of the intersection turning information in the routing rule to determine the predicted turning information of the to-be-predicted intersection comprises:

[0032] obtaining a turning feature variable corresponding to the target event;

[0033] inputting the turning feature variable into the second preset model, and outputting, by the second preset model, the predicted turning information of the to-be-predicted intersection.

[0034] Further, for the second preset model, comprising:

[0035] obtaining historical driving data;

[0036] performing data processing on the historical driving data to determine a sample turning feature variable;

[0037] inputting the sample turning feature variable into a second regression model for model training, and determining the trained second regression model as the second preset model when a second loss function meets a convergence condition.

[0038] Further, the method further comprises:

[0039] performing preloading according to full-quantity map data to determine a map topological relationship; or,

[0040] in a case where the map topological relationship has a local change, performing incremental updating of the map topological relationship according to local map data.

[0041] In a second aspect, the application provides a routing device based on a map topology, comprising a construction unit, a determination unit and a prediction unit, wherein:

[0042] the construction unit is configured to construct a map topological relationship;

[0043] The determining unit is configured to determine a routing rule corresponding to the target event according to the map topological relationship and the obtained V2X message.

[0044] The predicting unit is configured to predict an intersection order in the routing rule according to a first preset model, determine an intersection range affected by the target event, and route a message corresponding to the target event to all communication objects in the intersection range.

[0045] In a third aspect, an embodiment of the present application provides an electronic device, which comprises a memory and a processor, wherein:

[0046] The memory is configured to store a computer program capable of running on the processor.

[0047] The processor is configured to execute the method according to any one of the first aspect when running the computer program.

[0048] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by at least one processor to implement the method according to any one of the first aspect.

[0049] The routing method, device, equipment and storage medium based on a map topology provided by the present application construct a map topological relationship, determine a routing rule corresponding to a target event according to the map topological relationship and obtained V2X messages, predict an intersection order in the routing rule according to a first preset model, determine an intersection range affected by the target event, and route a message corresponding to the target event to all communication objects in the intersection range. In this way, traffic events, signs, basic safety messages (BSM) and other information can be dynamically and finely routed, so that messages can be sent to non-nearest intersections but still necessary to be received by traffic participants. In addition, through the binding logic optimization, the target event is no longer directly bound to the nearest intersection, but is bound to the nearest intersection through map matching. By introducing a classification model to predict vehicle driving intention and intersection turning phase information, the message routing direction can be more accurately calculated. In this way, more fine routing scenarios can be provided for V2X information, and the overall traffic safety or traffic efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 A flowchart of a routing method based on a map topology provided by an embodiment of the present application Figure 1 ;

[0051] Figure 2 A map topological relationship diagram provided by an embodiment of the present application

[0052] Figure 3 A target event and routing rule configuration schematic diagram provided for an embodiment of the present application;

[0053] Figure 4 A flowchart of a routing method based on map topology provided for an embodiment of the present application Figure 2 ;

[0054] Figure 5 A flowchart of a routing method based on map topology provided for an embodiment of the present application Figure 3 ;

[0055] Figure 6 A flowchart of a routing method based on map topology provided for an embodiment of the present application Figure 4 ;

[0056] Figure 7 A target event occurrence position schematic diagram provided for an embodiment of the present application;

[0057] Figure 8 A flowchart of a routing method based on map topology provided for an embodiment of the present application Figure 5 ;

[0058] Figure 9 A turning information schematic diagram provided for an embodiment of the present application;

[0059] Figure 10 A flowchart of a routing method based on map topology provided for an embodiment of the present application Figure 6 ;

[0060] Figure 11 A flowchart of a routing method based on map topology provided for an embodiment of the present application Figure 7 ;

[0061] Figure 12 A routing device based on map topology provided for an embodiment of the present application, and a component structure schematic diagram thereof;

[0062] Figure 13 A component structure schematic diagram of an electronic device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0063] In order to enable persons skilled in the art to better understand the features and technical contents of the embodiments of the present application, the implementation of the embodiments of the present application will be described in detail below with reference to the accompanying drawings, which are only used for reference and do not limit the embodiments of the present application.

[0064] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to be limiting of this application.

[0065] In the following description, reference is made to the "some embodiments", which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. It should also be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order of the objects. Understandably, "first\second\third" can be interchanged with a specific order or sequence as allowed, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0066] Before the technical solutions of the embodiments of the present application are described in detail, the V2X message routing method in the related art is first simply described.

[0067] Road network topology refers to the mathematical abstraction of real physical roads and intersections. Intersections or road nodes are abstracted as nodes in network topology, and roads are abstracted as edges in network topology. There is a road topology relationship between intersection A and intersection B, which means that A can pass through a certain edge to reach B intersection.

[0068] V2X message: In vehicle-road cooperation and V2X, the roadside perception computing unit can identify traffic information (road congestion, reverse, abnormal parking, etc.) in real time, and send it to nearby vehicles or other road users through the communication unit.

[0069] In the related art, there are related descriptions of V2X message routing methods and systems, for example, several routing strategies suitable for different scenarios are proposed. Among them, the vehicle driving intention prediction in vehicle-road cooperation is an important dependent condition in the message distribution scenario, and some technical solutions propose a system and method for partial driving behavior prediction. For specific traffic events (such as traffic congestion), some technical solutions propose a method for calculating the influence range.

[0070] In one possible implementation, part of the technology focuses on the implementation of technical innovation at the V2X message communication layer (radio technology, dedicated short-range communication, etc.) and the optimization design of network topology and network elements. For example, it is described how to control the message routing mode at the communication layer, such as unicast (point-to-point), multicast / broadcast mode. Generally, unicast (point-to-point) can use remote cellular communication, and multicast / broadcast can use LTE V2X and the like. Part of the technology introduces the concept of edge computing, deploys services at the roadside edge node, and reduces data latency through the LTE cellular network. The roadside unit acts as a mobile network base station, receives information from the mobile edge computing (MEC) or central platform, and broadcasts to surrounding road users through the short-range communication network. In addition, the existing technology generally uses roadside units and vehicle-mounted units for V2X interaction, and broadcasts to vehicles within the communication range (generally with an upper limit of 300-500m) in a broadcast manner, which is affected by the limitations of the deployment location of the roadside unit.

[0071] In another possible implementation, part of the technology introduces methods such as semi-static scheduling to optimize the channel design of the remote cellular communication interface between the base station and the vehicle side. Part of the technology allocates different frequency and time resources according to the speed, density, direction, and location of the vehicle. This way of wireless access network makes the vehicle end information reception not limited by the distance range of the roadside unit. The remote cellular communication air interface to a certain extent solves the problem of long-distance and larger range communication, but because it is essentially unicast, there are the following problems: there is a weak coverage space, it does not support a large number of users, and the latency cannot guarantee the V2X scenario requirements.

[0072] In yet another possible implementation, part of the technology mostly focuses on technical innovation at the communication layer and network topology design, and lacks the combination of actual traffic business scenarios for the V2X message routing method, and lacks quantitative models to fine-tune the message routing direction, geographical range, and target traffic participants. Moreover, some message types do not only affect the participants within a certain communication distance range around them, but may have a larger influence range. However, it is unclear which traffic participants at which locations should be reached by the message, and related technologies lack exploration, especially the lack of research on quantitative models.

[0073] In yet another possible implementation, part of the technology proposes a routing strategy for congestion event information, which broadcasts the congestion information to the vehicles traveling within its own communication range and the downstream roadside units through the 5G network communication module. The data processing module of the downstream roadside unit receives the congestion information through the 5G network communication module, and also broadcasts the congestion information to the vehicles traveling within its own communication range and its adjacent downstream roadside units. However, this method does not further discuss the specific upstream and downstream range (such as the number of crossroads, the distance along the road, etc.) of the congestion information delivery and its rationality.

[0074] Based on this, the embodiment of the present application provides a routing method based on map topology, which first constructs a map topology relationship; then determines the routing rule corresponding to the target event according to the map topology relationship and the acquired V2X message; and then predicts the intersection order in the routing rule according to the first preset model, determines the intersection range affected by the target event, and routes the message corresponding to the target event to all communication objects in the intersection range. In this way, not only can the influence range be expanded, so that the message can be sent to the non-nearest intersection, but also the traffic participants who still need to receive the message; moreover, the nearest intersection matching can be optimized based on the binding logic, so that it is no longer directly bound to the nearest intersection, but is bound to the nearest intersection through the map matching mode; in addition, the classification model can be introduced to predict the vehicle driving intention and intersection turning phase information, so as to more accurately calculate the message routing direction; in this way, a more refined routing scenario can be provided for the V2X information, thereby improving the overall traffic safety or traffic efficiency.

[0075] The various embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0076] In an embodiment of the present application, Figure 1 The flowchart of a routing method based on map topology provided by the embodiment of the present application Figure 1 . As Figure 1 shown, the method can include:

[0077] S101: Construct a map topology relationship.

[0078] It should be noted that in the embodiment of the present application, according to the description of V2X-map message in the existing standard, for a certain intersection, the map message body needs to include the first-order upstream intersection information (from the upstream intersection to the current intersection through the lane) and the first-order downstream intersection information (from the current intersection to the downstream intersection through the lane). Similarly, the intersections in the region are modeled as nodes, and the roads are modeled as connecting lines, and the map topology relationship can be established. Generally, the intersection can be a cross or T-shaped intersection; in addition, if there is a longer road with attribute changes, such as changes in lane boundary attributes, a node can also be set at the road attribute change.

[0079] For example, map message delivery transmits a series of intersection messages for a local area, which includes a series of intersections, each of which is an intersection node (NodeID), and its message contains the connection relationship with upstream and downstream intersection nodes. Figure 2 A schematic diagram of map topology is provided. For example... Figure 2 As shown, the road is a two-way road, including intersection nodes 21, 22, 23, and 24. For intersection node 22, its first-order upstream intersection is intersection node 23, and its first-order downstream intersection is intersection node 24. For intersection node 23, its first-order upstream intersections include intersection nodes 21, 22, and 24, and its first-order downstream intersections include intersection nodes 21, 22, and 24. Similarly, n-order upstream and downstream intersections refer to searching n layers upwards and downwards based on the current intersection, where layer 1 represents one connection line in the map topology. For example, for intersection node 22, its second-order upstream intersections are intersection nodes 21 and 24, and its second-order downstream intersections are intersection nodes 21 and 24.

[0080] In some embodiments, the method further includes: preloading based on full map data to determine map topology; or, in the case of local changes in map topology, incrementally updating map topology based on local map data.

[0081] S102: Determine the routing rules corresponding to the target event based on the map topology and the obtained V2X messages.

[0082] It should be noted that in this embodiment, for V2X messages, such as traffic events, signs, special vehicle location safety information, roadside traffic participant perception information, etc., target events are identified, and routing rules for the target events are set in advance. These rules can be set for a certain type of target events or customized routing rules can be set for a single target event.

[0083] S103: Based on the first preset model, predict the intersection order in the routing rules, determine the intersection range affected by the target event, and route the message corresponding to the target event to all communication objects within the intersection range.

[0084] It should be noted that, in this embodiment of the application, routing the message corresponding to the target event to all communication objects within the intersection range includes: routing the message corresponding to the target event to the roadside communication unit within the intersection range via the regional cloud server, and having the roadside communication unit send it to the corresponding vehicle; or, sending the message corresponding to the target event to the corresponding vehicle via remote cellular communication via the regional cloud server.

[0085] In a specific embodiment, the method further comprises: setting a routing rule statically according to a target event; or setting a routing rule corresponding to the target event dynamically according to roadside information of V2X; or setting a routing rule corresponding to the target event dynamically according to the target event by MEC.

[0086] It should be noted that, in the embodiments of the present application, for setting a routing rule statically according to a target event, a user can maintain a static routing rule in advance. For example, Figure 3 A configuration diagram of a target event and a routing rule is provided.

[0087] It should be further noted that, in the embodiments of the present application, the routing rule corresponding to the target event is set dynamically according to roadside information of V2X. For example, in a description field in road traffic time information event or road traffic sign information sign data protocol, an upper level (upperLevel) field and a lower level (downLevel) field are defined in a data exchange format (JavaScript Object Notatio, json) dictionary, and the field type is integer. Here, the upperLevel field and the downLevel field respectively represent the number of upstream routing nodes and the number of downstream routing nodes.

[0088] It should be further noted that, in the embodiments of the present application, the routing rule corresponding to the target event is set dynamically according to the target event by MEC. For example, when the MEC detects an event, the routing rule can be dynamically configured according to the severity, priority, event duration, etc. of the target event; and then the cloud control area platform performs routing matching and delivery in combination with a map topology relationship.

[0089] The embodiments of the present application provide a routing method based on a map topology, constructs a map topology relationship, determines a routing rule corresponding to a target event according to the map topology relationship and acquired V2X messages, predicts the number of intersections in the routing rule according to a first preset model, determines an intersection range affected by the target event, and routes a message corresponding to the target event to all communication objects in the intersection range. In this way, the influence range can be expanded, so that the message can be sent to a traffic participant who is not the nearest intersection but still needs to receive the message, thereby improving overall traffic safety or traffic efficiency.

[0090] In another embodiment of the present application, the routing method based on a map topology described in the foregoing embodiments is taken as an example of a congestion event as a target event. When the number of intersections in the routing rule is predicted according to a first preset model to determine an intersection range affected by the target event, referring to Figure 4 , the method can comprise:

[0091] S401: Obtain a congestion feature variable corresponding to the congestion event.

[0092] It should be noted that in the embodiments of the present application, the road side computing unit includes MEC, and when the MEC perceives that the congestion event occurs, the road side perception device simultaneously obtains the congestion feature variable. The congestion feature variable can include congestion level, congestion length, time, road level, upstream average congestion level index, downstream average congestion level index, etc.

[0093] S402: Input the congestion feature variable into the first preset model, and output a predicted congestion dissipation time through the first preset model.

[0094] It should be noted that in the embodiments of the present application, the regional cloud server inputs the congestion feature variable into the first preset model, and the predicted congestion dissipation time can be obtained.

[0095] In some embodiments, for the construction of the first preset model, see Figure 5 The method can include:

[0096] S501: Obtain a plurality of sample congestion events.

[0097] It should be noted that in the embodiments of the present application, based on the road side perception device (camera, laser radar, etc.) and artificial intelligence technology, the MEC can identify the congestion event, accumulate a certain number of historical congestion events, and take the historical congestion events as sample congestion events.

[0098] S502: Data processing is performed on the plurality of sample congestion events to determine sample congestion feature variables.

[0099] It should be noted that in the embodiments of the present application, from the time when the congestion event is identified by the MEC to the duration after the congestion event dissipates, the data of the sample congestion event is obtained, the data is processed, and the sample congestion feature variable is determined. For example, Table 1 provides a main description of the congestion feature variable.

[0100] Table 1

[0101]

[0102] S503: Input the sample congestion feature variable into the first regression model for model training, and when the first loss function meets the convergence condition, determine the trained first regression model as the first preset model.

[0103] It should be noted that in the embodiments of the present application, a general linear classification model assumes that the features are independent of each other, but in a congestion event, different categories of features of the same group of sample data have a mutual dependence relationship. For example, the congestion length of a road and the road traffic level thereof have a certain correlation. In order to learn the relationship between the features, a machine learning regression model is used for training to improve the effect of the first preset model. Exemplarily, the first regression model adopts a factorization machine (FM) model, introduces cross features and hidden vectors through two-by-two feature combination, and improves the model effect. The formula of the first regression model can be represented as:

[0104]

[0105] wherein n represents the number of features of the sample, w0represents a bias term parameter, X i is the value of the i th congestion feature variable, X j is the value of the j th congestion feature variable, w i represents the weight parameter corresponding to each feature X i , and <V i , V j > is the training parameter of the cross feature. is the dependent variable, that is, the predicted congestion dissipation time. A certain amount of historical congestion data is collected, each piece of sample data contains congestion feature variables and real congestion dissipation time, which is input into the first regression model as a training sample to learn the parameters of w0, w i , <V i , and V j .

[0106] It should be further noted that in the training of the machine learning model, the loss function is used to measure the difference between the predicted value of the model and the real value, so as to indicate the optimization direction of the model, and the target of the parameters is to minimize the loss function. Exemplarily, in the embodiments of the present application, the first loss function adopts the least square error as the loss function for continuous variables. In the training process of the first regression model, a stochastic gradient descent algorithm (SGD) is used for parameter learning. Along the opposite direction of the gradient vector, the gradient decreases fastest, that is, it is easier to find the minimum value of the function. The minimum value of the first loss function is solved by iterative approximation step by step in a heuristic way. When the first loss function satisfies the convergence condition, the trained first regression model is determined as the first preset model.

[0107] Through the above embodiment, first, a plurality of sample congestion events are acquired, then the plurality of sample congestion events are data processed to determine sample congestion characteristic variables, finally, the sample congestion characteristic variables are input into a first regression model for model training, and when a first loss function meets a convergence condition, the trained first regression model is determined as a first preset model. In this way, the first regression model is trained through the acquired congestion characteristic variables, the first preset model is obtained, and the predicted congestion dissipation time predicted by the first preset model is close to the real congestion dissipation time.

[0108] S403: According to the location of the congestion event, a reverse depth-first search is performed to determine a sequence of intersection orders that meet the predicted congestion dissipation time.

[0109] In a specific embodiment, the method further comprises:

[0110] According to the location of the congestion event, a reverse depth-first search is performed, and based on the map topological relationship, an accumulated weight value of an edge corresponding to at least one route is determined.

[0111] It should be noted that in the embodiments of the present application, based on the map topological relationship, according to the location of the congestion event, a node is traced back to the upstream, and a route is recorded, and the weight value of the edge is accumulated. The weight value of the edge can be set as the travel time, and the travel time is the length of the road divided by the real-time traffic flow travel speed. If the roadside sensing device cannot obtain the real-time traffic flow travel speed, the average speed mined from the history can also be used to fill in.

[0112] When the accumulated weight value of the edge corresponding to at least one route exceeds the predicted congestion dissipation time, the reverse depth-first search is stopped, and at least one intersection order of the reverse trace back is determined.

[0113] According to the at least one intersection order of the reverse trace back, a sequence of intersection orders that meet the predicted congestion dissipation time is determined.

[0114] It should be noted that in the embodiments of the present application, when the accumulated weight value of a certain route exceeds the predicted congestion dissipation time during the reverse depth-first search, the search is stopped, and the intersection order of the reverse trace back is recorded; according to the intersection order of the reverse trace back, a sequence of intersection orders that meet the predicted congestion dissipation time is determined.

[0115] In yet another specific implementation, when performing the reverse depth first search according to the location of the congestion event, it is determined whether there is an unvisited intersection node in the map topology relationship; when the determination result indicates that there is an unvisited intersection node in the map topology relationship, the location of the congestion event is rolled back, and the step of performing the reverse depth first search according to the location of the congestion event is continued; and when the determination result indicates that there is no unvisited intersection node in the map topology relationship, the reverse depth first search is stopped, and the determined at least one intersection order is taken as the intersection order sequence meeting the predicted congestion dissipation time.

[0116] It should be noted that in the embodiments of the present application, the intersection orders of the reverse backtracking are recorded; until all reachable nodes in the reverse depth first search (DFS) are completed, the intersection order sequence of the upstream backtracking and the intersection set passed by the route searched by the reverse search are obtained. When it is determined that there is an unvisited intersection node in the map topology relationship, based on the idea of the reverse depth first search, the location of the congestion event is rolled back, the unvisited intersection node is continued to be searched, after a stop condition is reached, the search is stopped, the intersection orders of the reverse backtracking are recorded; until all reachable nodes in the reverse depth first search are completed, the intersection order sequence of the backtracking is obtained, and the V2X server of the regional cloud routes the congestion event to the roadside communication unit corresponding to the intersection order sequence respectively and performs message distribution.

[0117] S404: Determine the intersection range affected by the congestion event according to the intersection order sequence.

[0118] It should be noted that in the embodiments of the present application, the intersection range affected by the congestion event is determined according to the intersection order sequence, including the need to route to the upstream n-order intersection and route to the downstream m-order intersection. Exemplarily, as shown in FIG. 22, a congestion type traffic event occurs near the intersection node 22, and vehicles near the upstream 1-2 order intersections of the intersection node 22 need to be informed in advance to make necessary re-routing navigation to avoid congestion. The upstream 1-2 order intersections of the intersection node 22 include the intersection node 23, the intersection node 24, and the intersection node 21, and the congestion message needs to be routed to the roadside communication unit of all corresponding intersections and distributed. Figure 2

[0119] ​The embodiment of the application provides a routing method based on a map topology, when a target event is a congestion event, first, a congestion feature variable corresponding to the congestion event is acquired, then the congestion feature variable is input into a first preset model, a predicted congestion dissipation time is output through the first preset model; then, reverse depth-first search is performed according to a position where the congestion event is located, a sequence of intersection orders meeting the predicted congestion dissipation time is determined; finally, according to the sequence of intersection orders, a range of intersections affected by the congestion event is determined. In this way, the upstream and downstream backtracking orders of the map topology relationship are considered, so that the message can be sent to the traffic participants of the non-nearest intersection but still necessary to receive the message, and the overall traffic safety or traffic efficiency is improved.

[0120] In another embodiment of the application, the routing method based on the map topology described in the foregoing embodiment is based on the map topology, Figure 6 The flowchart of the routing method based on the map topology provided in the embodiment of the application is shown in Figure 4 . As Figure 6 shown, the method can include:

[0121] S601: determining the nearest position where a target event occurs based on a map matching algorithm.

[0122] It should be noted that in the embodiment of the application, usually, the nearest intersection to a certain target event needs to be determined, and then the message routing is performed, but a more reasonable way needs to consider the lane position where the current target event occurs and the in-out relationship of the relative intersection to determine the nearest position, and then the message routing is performed.

[0123] S602: after the target event is bound to the nearest position, determining a start intersection corresponding to the target event, and taking the start intersection as a main intersection of the target event; wherein the main intersection is used for intersection depth search in the routing rule.

[0124] It should be noted that in the embodiment of the application, Figure 7 a target event occurrence position schematic diagram is provided, as Figure 7As shown, the map includes intersection nodes 71, 72, 73, and 74, with the black circle representing event A. Event A occurs near intersection node 73, but its lane position only affects vehicles traveling from intersection node 74 towards intersection node 73, not vehicles traveling from intersection node 73 to intersection node 74. Therefore, a more reasonable approach is to treat intersection node 74 as the main intersection and use it as a basis for upstream and downstream intersection backtracking. In other words, for event A, we cannot rely solely on straight-line distance information to determine the starting intersection for upstream and downstream intersection backtracking. Specifically, for the target event, a map matching algorithm can be used to bind it to the nearest road (Link) or lane (Lane), rather than directly binding it to the nearest intersection. After binding to a Link or Lane, its corresponding starting intersection number is determined, and this intersection is designated as the main intersection for subsequent upstream and downstream intersection depth searches and message routing.

[0125] It should also be noted that, in this embodiment of the application, the relative position of the vehicle and the traffic event is calculated by the regional cloud server, and the target event is directly sent to the vehicle in the corresponding lane based on remote cellular communication.

[0126] This application provides a map topology-based routing method. Through a map matching algorithm, target events are bound to Links or Lanes, and their corresponding starting intersections are determined as main intersections. This avoids directly binding to the nearest intersection node, resulting in more rational and efficient message delivery.

[0127] In another embodiment of this application, based on the map topology-based routing method described in the foregoing embodiments, Figure 8 A flowchart illustrating a map topology-based routing method provided in this application embodiment. Figure 5 .like Figure 8 As shown, the method also includes:

[0128] S801: Construct map topology relationships.

[0129] S802: Determine the routing rules corresponding to the target event based on the map topology and the acquired V2X messages.

[0130] S803: Based on the second preset model, predict the intersection turning information in the routing rules, determine the predicted turning information of the intersection to be predicted, and route the predicted turning information to the target communication object of the intersection to be predicted.

[0131] Table 2

[0132] Phase value Turn instruction Phase value Turn instruction Phase value Turn instruction 0 Straight north-south 6 Right turn south-north 12 Straight west-east 1 Left turn north-south 7 U-turn south-north 13 Left turn west-east 2 Right turn north-south 8 Straight east-west 14 Right turn west-east 3 U-turn north-south 9 Left turn east-west 15 U-turn west-east 4 Straight south-north 10 Right turn east-west 5 Left turn south-north 11 U-turn east-west

[0133] It should be noted that, in this embodiment of the application, V2X message routing needs to consider the lane where the message is located and the turning phase information of the intersection ahead. For example, Figure 9 A diagram illustrating turning information is provided, such as Figure 9 As shown, the rightmost lane of Link3 has a lane ID of 4. A special vehicle (such as an ambulance or fire truck) is reporting its BSM vehicle location. Its driving intention is to pass through the intersection with NodeID 1, turn right, and continue driving towards NodeID 5. The phase ID for this right turn is 6. Table 2 provides a steering phase dictionary. When routing upstream and downstream for this BSM, the phase information of its neighboring intersections can be considered: that is, for the intersection with NodeID = 1, the downstream intersection with NodeID = 8 (corresponding to NodeID = 5) only needs to send a special vehicle location information reminder to the roadside communication unit of that intersection to remind nearby vehicles to avoid it, without needing to send it to the roadside communication units near the intersections with NodeID = 3 and 4.

[0134] In one specific embodiment, the intersection turning information in the routing rules is predicted according to a second preset model to determine the predicted turning information of the intersection to be predicted, such as... Figure 10 As shown, the method also includes:

[0135] S1001: Obtain the steering feature variables corresponding to the target event.

[0136] It should be noted that, in the embodiments of this application, the steering characteristic variables may include: the lane in which the vehicle is located, the heading angle, the speed, the distance from the stop line, etc.

[0137] S1002: Input the steering feature variables into the second preset model, and output the predicted steering information of the intersection to be predicted through the second preset model.

[0138] It should be noted that, in this embodiment, the real-time steering feature variables of the roadside sensing device are input into the second preset model to obtain predicted steering information. Then, based on the map topology, downstream intersection message broadcasting routing for a specific phase is performed. Generally, the predicted phase value with the highest probability is selected. The predicted probabilities of each phase are sorted in descending order. If the first n predicted probability values ​​are similar, message routing can also be performed for the downstream intersections of the corresponding n phases.

[0139] In yet another specific embodiment, the construction of the second preset model, such as Figure 11 As shown, the method includes:

[0140] S1101: Obtain historical driving data.

[0141] It should be noted that, in this embodiment, when determining the driving intention of a special vehicle, accurate vehicle-side BSM information cannot often be collected in reality. This is mainly because the special vehicle is not a connected vehicle and cannot report its location information to the V2X connected system; the reported location information is based on GPS satellite positioning, which results in location offset; and the lane edge line at the vehicle's current position is still a dashed line. Therefore, this embodiment collects historical vehicle driving data within a certain range before the stop line at the intersection entrance. Through roadside sensing technology, multiple frames (generally 10) of trajectory data and their corresponding historical vehicle driving data are collected to obtain the phase information of the vehicle's actual intersection turning behavior.

[0142] S1102: Process historical driving data to determine the steering characteristic variables of the samples.

[0143] It should be noted that, in this embodiment, historical driving data is acquired and processed to eliminate the influence of excessively large differences in the dimensions of different variables, thereby determining the steering characteristic variables of the sample. For example, Table 3 provides a main description of the steering characteristic variables.

[0144] It should also be noted that, in this embodiment of the application, for a piece of historical data, the turning feature vector is defined as:

[0145]

[0146] Where k is the steering feature variable of the k-th frame in a segment of the vehicle's trajectory, Y 1k Let Y be the value of Y1 in the k-th frame of the vehicle. 2k Let Y be the value of the vehicle in frame k, Y2. 3k Let Y be the value of Y3 in the k-th frame for this vehicle. 4k This is the value of Y4 in the k-th frame for this vehicle. The steering characteristic variables Y1, Y2, Y3, and Y4 are detailed in Table 3.

[0147] Table 3

[0148]

[0149] S1103: Input the sample into the feature variables of the second regression model for model training, and when the second loss function meets the convergence condition, determine the trained second regression model as the second preset model.

[0150] It should be noted that, in the embodiments of this application, different types of features in the same frame of data, as well as features in different frames of data, may have interdependent relationships. In order to learn these relationships between features, a machine learning regression model is used for training to improve the performance of the second preset model.

[0151] For example, the second regression model uses the FM model. The FM model improves model performance by combining pairwise features, introducing cross features and latent vectors. The formula for the second regression model can be expressed as:

[0152]

[0153] Where n represents the number of features in the sample, s0 represents the bias term parameter, and Y i Y is the value of the i-th steering characteristic variable. j It is the value of the j-th steering characteristic variable, s i Representing each feature Y i The corresponding weight parameters, <R i ,R j >These are the training parameters for the cross features. The dependent variable is the predicted steering information. A certain amount of historical vehicle data is collected. Each sample data package contains a steering feature vector and the actual steering information. This data is used as training samples and input into the second regression model. The values ​​of s0 and s are then compared. i , <R i ,R j >Perform parameter learning.

[0154] The normalized exponential function (Softmax) is used as the activation function. The Softmax function is applied to the n-dimensional tensor (n corresponds to several phases in Table 2), rescaling it so that the elements of the n-dimensional output tensor are in the range [0, 1] and sum to 1. Each element represents the predicted probability of its corresponding class.

[0155] The second loss function uses cross-entropy, and the cross-entropy formula is as follows:

[0156]

[0157] Where K represents the total number of categories in the classification problem, i.e., the various phases shown in Table 2, h i For the true classification result, p i This represents the probability that the model predicts a certain class. Multiplying the negative number by the logarithm ensures that the second loss function is positive and that the higher the probability of correctly predicting the class, the smaller the loss value J.

[0158] Using SGD, a heuristic approach is employed to iteratively adjust parameters in the direction of decreasing the second loss function, training a second regression model. This process iteratively approximates the minimum value of the second loss function step by step. When the second loss function meets the convergence condition, the trained second regression model is designated as the second preset model.

[0159] This application provides a map topology-based routing method that considers steering optimization for road events. A second preset model is established using the vehicle's lane, heading angle, speed, and distance to the stop line as steering feature variables. The method acquires the steering feature variables corresponding to the target event and uses them to predict the vehicle's steering direction by outputting the predicted steering information of the intersection to be predicted through the second preset model. Then, it combines map topology relationships to perform downstream intersection message broadcasting routing for specific phases. In this way, by predicting vehicle driving intentions and intersection steering phase information through the second preset model, the message routing direction can be calculated more accurately, making message delivery more reasonable and efficient, thereby improving overall traffic safety or traffic efficiency.

[0160] In another embodiment of this application, based on the map topology-based routing method described in the foregoing embodiments, ... Figure 2 The application scenario shown is a two-way road with each intersection open to traffic, which will be explained in detail.

[0161] In this embodiment, based on V2X-map data, a map topology relationship between several interconnected intersections is loaded. For V2X messages (including traffic events, signs, special vehicle location safety information, roadside traffic participant perception information, etc.), rules governing the scope of their impact on the required intersections are pre-defined. These rules can be uniform for a certain type of message or customized for a single message. Based on the geographic location attributes of the V2X message, a spatial matching algorithm is used to find the nearest intersection, determining the range of intersections it directly affects. Based on the pre-loaded and real-time dynamically updated road network topology relationship and the V2X message impact range rules, routing is used to find the range of intersections it indirectly affects.

[0162] In other words, this paper proposes a V2X message routing rule and system based on map data, which can flexibly and dynamically perform fine-grained routing of information such as traffic events, signs, and BSMs (Border Management Messages), allowing messages to be sent to traffic participants who are not at the nearest intersection but still need to receive the message, thereby improving overall traffic safety or efficiency. Specifically, the detailed process of this method may include the following steps:

[0163] Step 1: Establish map topology relationships. Based on the existing standards for V2X map messages, such as... Figure 2As shown, for a given intersection, such as intersection node 22, the map message body must include its corresponding first-order upstream intersection information (access from the upstream intersection via lanes to the current intersection) and first-order downstream intersection information (access from the current intersection via lanes to the downstream intersection). Similarly, by modeling intersections within a region as nodes and roads as connecting lines, a regional road network can be established. Generally, intersections can be crossroads or T-junctions; additionally, if there are long roads with attribute changes, such as changes in lane edge attributes, intersection nodes can be set at the points where road attributes change.

[0164] like Figure 2 As shown, the road is bidirectional, so for intersection node 22, its first-order upstream intersection is intersection node 23, and its first-order downstream intersection is intersection node 24. For intersection node 23, its first-order upstream intersections include intersection nodes 21, 22, and 24, and its first-order downstream intersections include intersection nodes 21, 22, and 24. Similarly, n-order upstream and downstream intersections refer to searching n layers upwards and downwards based on the current intersection, where layer 1 represents one connection line in the road network relationship described above. For example, for intersection node 22, its second-order upstream intersections are intersection nodes 21 and 24, and its second-order downstream intersections are intersection nodes 21 and 24.

[0165] Step 2: Map data preloading and dynamic updates. For V2X information processing systems, it is necessary to support the preloading of the entire map data, as well as incremental updates when local map data changes.

[0166] Step 3.1: Configure the mapping between target events and routing rules. For a certain type of target event, its impact is not limited to its nearest intersection but may affect a wider range of intersections upstream and downstream. Different routing rules can be set accordingly, including routing to an upstream n-order intersection or a downstream m-order intersection. For example, ... Figure 2 As shown, a traffic congestion event has occurred near intersection node 22. Vehicles near the upstream 1-2 level intersections need to be notified in advance to perform necessary rerouting navigation to avoid the congestion. The upstream 1-2 level intersections of intersection node 22 include intersection nodes 23, 24, and 21. This congestion message needs to be routed to the roadside communication units of all corresponding intersections and sent out.

[0167] V2X intelligent information services include real-time information (such as intersection collision warnings, parking warnings, and lane change warnings) and non-real-time information (regular congestion events, signs, and near-field payments). Taking regular congestion events as an example, a quantitative mathematical model is established to model and predict the intersection order of upstream and downstream points in the routing rules.

[0168] Based on roadside sensing devices (cameras, LiDAR, etc.) and artificial intelligence technology, MEC can identify congestion events and accumulate a number of historical congestion events for training the first regression model. This prediction problem is modeled as a regression model, with the dependent variable being the duration of the event from its identification by the MEC to its dissipation. The congestion characteristic variables are shown in Table 1. Training the first regression model based on historical data can predict the duration of newly occurring congestion events.

[0169] General linear classification models assume that features are independent and uncorrelated. However, in congestion events, different categories of features in the same set of sample data are interdependent. For example, the length of road congestion is related to its traffic level. In order to learn the relationship between these features, the first regression model adopts the FM model, which introduces cross features and latent vectors by combining features in pairs to improve the model performance. The formula of the first regression model is shown in formula (1). A certain amount of historical congestion data is collected. Each sample data package contains congestion feature variables and the actual time for congestion to dissipate. These are used as training samples and input into the first regression model. i , <V i V j >Perform parameter learning.

[0170] In the training of the first regression model, the loss function is used to measure the difference between the model's predicted output and the true value, thus indicating the direction of model optimization. The goal of the parameters is to minimize the loss function. For the regression prediction problem of congestion events, the first loss function is the least squared error, which is used as the loss function for continuous variables. During the training of the first regression model, SGD is used for parameter learning. The gradient decreases fastest along the opposite direction of the gradient vector, making it easier to find the minimum value of the function. By iteratively approximating the minimum value of the first loss function step by step through a heuristic approach, the minimized first loss function and the corresponding first regression model parameter values ​​are obtained, thus obtaining the first preset model.

[0171] When the MEC detects a congestion event 'e', ​​the roadside sensing devices simultaneously detect the multidimensional congestion feature variables in Table 1 above. The regional cloud server inputs these multidimensional congestion feature variables into the first preset model to obtain the predicted congestion dissipation time. A reverse depth-first search is performed at the location of event 'e', ​​tracing back upstream nodes based on map topology and recording the routes taken, accumulating the edge weights. The edge weights can be set as travel time, where travel time = length ÷ real-time traffic flow speed. If real-time traffic flow speed cannot be detected, the average speed from historical data mining can be used. When the accumulated weight of a route exceeds the predicted congestion dissipation time during the reverse search, the search stops, and the intersection order 'n' of the reverse tracing is recorded. jBased on the idea of ​​reverse depth-first search, backtrack to the previous vertex and continue to search for unvisited upstream nodes; after reaching the stopping condition, stop the search and record the order n of the backtracking path. k ; until all reachable nodes in the reverse depth-first search are completed, the sequence of intersection orders (n) backtracking upstream is obtained. j n k n l The V2X server of the regional cloud routes the congestion event e to the roadside communication units corresponding to the aforementioned intersection sets for message delivery.

[0172] Step 3.2: Perform route optimization on Step 3.1. Step 3.1 mentions that for a given target event, the nearest intersection needs to be determined for message routing. A more reasonable approach would consider the lane location where the current event occurs and its relative entry / exit relationship with the intersection. For example... Figure 7 As shown, event A occurs near intersection node 73, but its lane position only affects vehicles traveling from intersection node 74 towards intersection node 73, not vehicles traveling from intersection node 73 to intersection node 74. Therefore, a more reasonable approach is to treat intersection node 74 as the main intersection and use it as a basis for upstream and downstream intersection backtracking. That is, for event A, a map matching algorithm can be used to bind it to the nearest link or lane, rather than directly binding it to the nearest intersection node; after binding to the link or lane, its corresponding starting intersection number is determined, and this intersection is designated as the main intersection, thus enabling deep searching and message routing of upstream and downstream intersections. When roadside sensing devices identify a certain type of event (such as construction information), they broadcast the construction information to nearby vehicles. Vehicles need to receive the construction warning information to calculate their relative position; if traveling in a lane affected by construction, they need to change vehicles or replan their routes in advance. Based on this, another solution is provided: the regional cloud server calculates the relative position of the vehicle and the traffic event, and sends the target event directly to the vehicle in the corresponding lane based on remote cellular communication.

[0173] Step 3.3: For routing rules corresponding to specific types of target events, users can maintain static relationships in advance, such as... Figure 3As shown. Alternatively, rules can be defined and dynamically routed within the V2X roadside information. Specifically, the description fields in the road traffic time information event or road traffic sign information data protocol are defined using a JSON dictionary, defining upperLevel and downLevel fields of integer type. Here, upperLevel and downLevel represent the order of the upstream and downstream routing nodes, respectively. Furthermore, when the MEC detects a target event, routing rules can be dynamically configured based on factors such as the severity, priority, and duration of the event; then, the cloud control area platform combines map topology data to perform message routing matching and distribution.

[0174] Step 3.4: Refine the steering information from Step 3.1. In some scenarios, V2X message routing needs to consider the lane where the message is located and the steering phase information of the intersection ahead. For example, ... Figure 9 As shown, the rightmost lane of Link3 has LaneID 4. A special vehicle (such as an ambulance or fire truck) is reporting its BSM vehicle location. Its travel intention is to pass through intersection NodeID 1, turn right, and continue towards NodeID 5. The phaseId of this right turn is 6. When routing upstream and downstream for this BSM, the phase information of its neighboring intersections can be considered: that is, for the intersection with NodeID = 1, the downstream intersection with NodeID = 8 (corresponding to NodeID = 5) only needs to send a special vehicle location information reminder to the roadside communication unit of that intersection to warn nearby vehicles to avoid it, without needing to send it to the roadside communication units near intersections with NodeID = 3 and 4.

[0175] When determining the driving intentions of special vehicles, accurate vehicle-side BSM (Balanced Sense Time) information is often not collected in reality. This may be because: the special vehicle is not a connected vehicle and cannot report its location information to the V2X network system; the reported location information is based on GPS satellite positioning, which may result in location offset; or the lane markings at the vehicle's current location are still dashed, indicating the possibility of a lane change ahead. To address these issues, an improvement plan is proposed: collect historical vehicle driving data within a certain range before the stop line at the intersection entrance. Using roadside sensing technology, collect multiple frames (typically 10) of trajectory data and their corresponding multi-dimensional time-series features to gather the actual turning phase information of the vehicle at the intersection. This problem is modeled as a multi-classification problem: the dependent variable is the actual turning phase result of the vehicle at the intersection (dictionary shown in Table 2), and the turning feature variables are several vehicle driving-related variables, as shown in Table 3. Some of the turning feature variables are normalized to eliminate the influence of excessively large differences in the dimensions of different variables.

[0176] Features of different classes within the same frame of data, as well as features of different frames of data, may be interdependent. In order to learn the relationships between these features, the second regression model is trained using the FM model, as shown in formula (3).

[0177] The difference lies in the use of Softmax as the activation function in modeling the multi-label classification problem of the turning phase. The Softmax function is applied to an n-dimensional tensor (n corresponds to several phases in Table 2), rescaling it so that the elements of the n-dimensional output tensor are in the range [0, 1] and sum to 1. Each element represents the predicted probability of its corresponding class. Cross-entropy is used as the second loss function. The expression for cross-entropy is shown in Equation (4).

[0178] Similarly, using SGD, a heuristic approach is employed to iteratively adjust parameters in the direction of decreasing the second loss function, training a second regression model to obtain a second preset model for predicting the turning phase ahead of the vehicle. Real-time multi-dimensional feature results from roadside sensing devices are input into this second preset model to obtain predicted values ​​for the turning phase ahead of the vehicle. Then, combined with map topology, downstream intersection message broadcasting routes for specific phases are performed. Generally, the predicted phase value with the highest probability is selected. The predicted probabilities of each phase are sorted in descending order; if the top n predicted probability values ​​are similar, message routing can also be performed for the downstream intersections of those n phases.

[0179] Understandably, in this application embodiment, on the one hand, message routing in related technologies generally simply selects the K nearest roadside communication units based on Euclidean distance; while this application embodiment considers the upstream and downstream topological relationships of the traffic network, making it more suitable for intelligent transportation scenarios. On the other hand, related technologies pre-set fixed-range geographic cell fences and then broadcast messages based on the fence where the user device is located; while this application embodiment models the road network based on map data and can set first-order or higher-order routing rules. More deeply, the upstream and downstream backtracking order can be determined through model training, making it suitable for complex traffic scenarios. In addition, this application embodiment optimizes the routing rules in the following aspects based on the simplest routing rules: adding map matching logic to optimize the nearest intersection query (i.e., using road binding logic to optimize the nearest intersection matching); reserving fields in the message body to dynamically fill upstream and downstream routing strategies; considering the lane position of the V2X message and the phase information in the intersection topology in the routing rules, thereby performing refined routing. In addition, a machine learning regression model is used to model the impact range of specific events (taking congestion events as an example) to quantitatively output the routing order in the upstream and downstream backtracking scenario. Machine learning classification models are introduced to predict vehicle driving intentions: intersection turning phase information. This allows for more refined routing scenarios for vehicle-to-everything (V2X) information services, making message delivery more reasonable and efficient.

[0180] In summary, the above embodiments provide a detailed explanation of the specific implementation of the aforementioned embodiments. It can be seen that the technical solutions of the aforementioned embodiments not only expand the scope of influence, allowing messages to be sent to traffic participants who are not at the nearest intersection but still need to receive the message; but also optimize the nearest intersection matching based on road-binding logic, so that it is no longer directly bound to the nearest intersection, but rather bound to the nearest intersection through map matching; furthermore, a classification model can be introduced to predict vehicle driving intentions and intersection turning phase information, facilitating more accurate calculation of message routing direction; thus, more refined routing scenarios can be provided for V2X information, thereby improving overall traffic safety or traffic efficiency.

[0181] Based on the same inventive concept as the foregoing embodiments, see [link to previous document]. Figure 12 This illustration shows a schematic diagram of the composition structure of a map topology-based routing device 120 provided in an embodiment of this application. Figure 12 As shown, the map topology-based routing device 120 may include: a construction unit 1201, a determination unit 1202, and a prediction unit 1203; wherein,

[0182] Build unit 1201 is configured to build map topology relationships;

[0183] Unit 1202 is configured to determine the routing rules corresponding to the target event based on the map topology and the acquired V2X messages.

[0184] The prediction unit 1203 is configured to predict the intersection order in the routing rules according to the first preset model, determine the intersection range affected by the target event, and route the message corresponding to the target event to all communication objects in the intersection range.

[0185] In some embodiments, see Figure 12 The map-based routing device 120 also includes an acquisition unit 1204, configured to acquire congestion feature variables corresponding to congestion events;

[0186] The prediction unit 1203 is also configured to input congestion characteristic variables into a first preset model and output the predicted congestion dissipation time through the first preset model.

[0187] The determination unit 1202 is also configured to perform a reverse depth-first search based on the location of the congestion event to determine the intersection order sequence that matches the predicted congestion dissipation time; and to determine the intersection range affected by the congestion event based on the intersection order sequence.

[0188] In some embodiments, the acquisition unit 1204 is further configured to acquire multiple sample congestion events;

[0189] The determining unit 1202 is further configured to process multiple sample congestion events to determine sample congestion feature variables; input the sample congestion feature variables into a first regression model for model training; and when the first loss function satisfies the convergence condition, determine the trained first regression model as the first preset model.

[0190] In some embodiments, the determining unit 1202 is further configured to perform a reverse depth-first search based on the location of the congestion event, and determine the cumulative weight of the edge corresponding to at least one path based on the map topology; when the cumulative weight of the edge corresponding to at least one path exceeds the predicted congestion dissipation time, stop the reverse depth-first search, determine the order of at least one intersection for reverse backtracking; and determine the sequence of intersection orders that conform to the predicted congestion dissipation time based on the order of at least one intersection for reverse backtracking.

[0191] In some embodiments, the determining unit 1202 is further configured to: determine whether there are unvisited intersection nodes in the map topology when performing a reverse depth-first search based on the location of the congestion event; when the determination result indicates that there are unvisited intersection nodes in the map topology, backtrack to the location of the congestion event and continue to perform the reverse depth-first search based on the location of the congestion event; and when the determination result indicates that there are no unvisited intersection nodes in the map topology, stop the reverse depth-first search and use the determined at least one intersection order as the intersection order sequence that conforms to the predicted congestion dissipation time.

[0192] In some embodiments, the determining unit 1202 is further configured to determine the nearest location of the target event based on a map matching algorithm; and after binding the target event to the nearest location, determine the starting intersection corresponding to the target event and use the starting intersection as the main intersection of the target event; wherein the main intersection is used for intersection depth search in the routing rules.

[0193] In some embodiments, the prediction unit 1203 is further configured to predict the intersection turning information in the routing rules according to the second preset model, determine the predicted turning information of the intersection to be predicted, and route the predicted turning information to the target communication object of the intersection to be predicted.

[0194] In some embodiments, the acquisition unit 1204 is further configured to acquire the steering feature variable corresponding to the target event;

[0195] The prediction unit 1203 is also configured to input the steering feature variables into the second preset model and output the predicted steering information of the intersection to be predicted through the second preset model.

[0196] In some embodiments, the acquisition unit 1204 is further configured to acquire historical driving data;

[0197] The determining unit 1202 is further configured to process historical driving data to determine sample steering feature variables; input the sample steering feature variables into the second regression model for model training; and when the second loss function meets the convergence condition, determine the trained second regression model as the second preset model.

[0198] In some embodiments, the construction unit 1204 is further configured to preload the map topology based on the full map data to determine the map topology; or, in the case of local changes in the map topology, to incrementally update the map topology based on the local map data.

[0199] Understandably, in this embodiment, a "unit" can be a portion of a circuit, a portion of a processor, a portion of a program or software, etc., and can also be a module or a non-modular component. Furthermore, the components in this embodiment can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional module.

[0200] If the integrated unit is implemented as a software functional module and not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0201] Therefore, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by at least one processor, implements the steps of the method described in any of the foregoing embodiments.

[0202] Based on the above-described composition of the map topology-based routing device 120 and the computer-readable storage medium, see [link to documentation]. Figure 13 This illustrates the structural composition of the electronic device 130 provided in an embodiment of this application. For example... Figure 13As shown, electronic device 130 may include: a communication interface 1301, a memory 1302, and a processor 1303; the various components are coupled together via a bus system 1204. It is understood that the bus system 1204 is used to implement communication between these components. In addition to a data bus, the bus system 1204 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 13 All buses are labeled as bus system 1204. Among them, communication interface 1301 is used for receiving and transmitting signals during information exchange with other external network elements;

[0203] Memory 1302 is used to store computer programs that can run on processor 1303;

[0204] Processor 1303, when running the computer program, is configured to perform:

[0205] Construct map topology relationships; determine the routing rules corresponding to the target event based on the map topology relationships and the acquired V2X messages; predict the intersection order in the routing rules based on the first preset model, determine the intersection range affected by the target event, and route the message corresponding to the target event to all communication objects within the intersection range.

[0206] It is understood that the memory 1302 in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate Synchronous DRAM (DDRSDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 1302 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0207] The processor 1303 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 1303 or by instructions in software form. The processor 1303 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 1302. Processor 1303 reads the information in memory 1302 and completes the steps of the above method in conjunction with its hardware.

[0208] It is understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or combinations thereof.

[0209] For software implementation, the techniques described herein can be achieved through modules (e.g., procedures, functions, etc.) that perform the functions described herein. The software code can be stored in memory and executed by a processor. The memory can be implemented within the processor or externally.

[0210] Alternatively, as another embodiment, the processor 1303 is further configured to perform the steps of the method described in any of the foregoing embodiments when running the computer program.

[0211] It should be noted that, in this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0212] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0213] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0214] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0215] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0216] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A routing method based on map topology, characterized in that, The method includes: Construct map topology relationships; Based on the map topology and the acquired V2X wireless communication messages, determine the routing rules corresponding to the target event; Based on the first preset model, the intersection order in the routing rules is predicted to determine the intersection range affected by the target event, and the message corresponding to the target event is routed to all communication objects within the intersection range. When the target event is a congestion event, the step of predicting the intersection order in the routing rules according to the first preset model to determine the intersection range affected by the target event includes: Obtain the congestion feature variables corresponding to the congestion event; The congestion characteristic variables are input into the first preset model, and the first preset model outputs the predicted congestion dissipation time. A reverse depth-first search is performed based on the location of the congestion event to determine the intersection order sequence that matches the predicted congestion dissipation time. Based on the intersection order sequence, determine the range of intersections affected by the congestion event; Determining the first preset model includes: Obtain multiple sample congestion events; The data of the multiple sample congestion events are processed to determine the sample congestion characteristic variables; The sample congestion feature variables are input into the first regression model for model training, and when the first loss function satisfies the convergence condition, the trained first regression model is determined as the first preset model.

2. The method according to claim 1, characterized in that, The step of performing a reverse depth-first search based on the location of the congestion event to determine the sequence of intersection orders that matches the predicted congestion dissipation time includes: A reverse depth-first search is performed based on the location of the congestion event, and the cumulative weight of the edge corresponding to at least one path is determined based on the map topology. When the cumulative weight of the edge corresponding to at least one path exceeds the predicted congestion dissipation time, stop the reverse depth-first search and determine the order of at least one intersection for reverse backtracking. Based on the at least one intersection order obtained through the reverse backtracking, a sequence of intersection orders that conforms to the predicted congestion dissipation time is determined.

3. The method according to claim 2, characterized in that, The method further includes: When performing a reverse depth-first search based on the location of the congestion event, determine whether there are any unvisited intersection nodes in the map topology. When the judgment result indicates that there are unvisited intersection nodes in the map topology, the process reverts to the location of the congestion event and continues to execute the step of performing a reverse depth-first search based on the location of the congestion event. When the judgment result indicates that there are no unvisited intersection nodes in the map topology, the reverse depth-first search is stopped, and the determined intersection order is used as the intersection order sequence that conforms to the predicted congestion dissipation time.

4. The method according to claim 1, characterized in that, The method further includes: Based on a map matching algorithm, the nearest location where the target event occurred is determined; After binding the target event to the nearest location, the starting intersection corresponding to the target event is determined, and the starting intersection is used as the main intersection of the target event; wherein, the main intersection is used to perform the intersection depth search in the routing rules.

5. The method according to claim 1, further comprising: The intersection turning information in the routing rules is predicted according to the second preset model, the predicted turning information of the intersection to be predicted is determined, and the predicted turning information is routed to the target communication object of the intersection to be predicted.

6. The method according to claim 5, wherein predicting the intersection turning information in the routing rules based on the second preset model to determine the predicted turning information of the intersection to be predicted includes: Obtain the steering feature variables corresponding to the target event; The steering feature variables are input into the second preset model, and the predicted steering information of the intersection to be predicted is output through the second preset model.

7. The method according to claim 5, characterized in that, The method further includes: Obtain historical driving data; The historical driving data is processed to determine the sample steering characteristic variables; The sample is shifted to feature variables and input into the second regression model for model training. When the second loss function meets the convergence condition, the trained second regression model is determined as the second preset model.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: The map topology is determined by preloading the full map data; or... In the event of local changes to the map topology, the map topology is incrementally updated based on the local map data.

9. A routing device based on map topology, characterized in that, The map-based topology routing device includes: a construction unit, a determination unit, a prediction unit, and an acquisition unit; wherein, The building unit is configured to build map topology relationships; The determining unit is configured to determine the routing rules corresponding to the target event based on the map topology and the acquired V2X messages. The prediction unit is configured to predict the intersection order in the routing rules according to the first preset model, determine the intersection range affected by the target event, and route the message corresponding to the target event to all communication objects in the intersection range. The acquisition unit is configured to acquire congestion feature variables corresponding to the congestion event; The prediction unit is further configured to input congestion characteristic variables into a first preset model and output the predicted congestion dissipation time through the first preset model; The determining unit is further configured to perform a reverse depth-first search based on the location of the congestion event to determine the intersection order sequence that matches the predicted congestion dissipation time; and to determine the intersection range affected by the congestion event based on the intersection order sequence. The acquisition unit is further configured to acquire multiple sample congestion events; The determining unit is further configured to process data from multiple sample congestion events to determine sample congestion feature variables; input the sample congestion feature variables into a first regression model for model training; and determine the trained first regression model as a first preset model when the first loss function satisfies the convergence condition.

10. An electronic device, characterized in that, Includes memory and processor, wherein: The memory is used to store computer programs that can run on the processor; The processor is configured to execute, when running the computer program, the map-based topology routing method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by at least one processor, implements the map topology-based routing method as described in any one of claims 1 to 8.

Images