A dynamic path planning method for transport vehicles based on real-time road conditions

Abstract

The present invention relates to the field of path planning technology, specifically a method for dynamic path planning of a transport vehicle based on real-time road conditions, comprising the following steps: clustering is performed based on the node coordinate data in the full map according to the spatial proximity and connectivity of the nodes. In the present invention, clustering analysis based on the spatial proximity and connectivity of the nodes is used to identify and extract connection nodes that simultaneously belong to multiple different clusters, thereby achieving a precise definition of regional boundary nodes, reducing the degree of node redundancy in the path planning process, and improving the clarity of the road network structure; by calculating the shortcut costs between different boundary nodes, a clear list of shortcut costs is established, and a topology map is constructed based on this, making path selection more accurate and efficient; when the state of the road network changes, the shortcut corresponding to the changed section is retrieved and quickly matched in real time, the cost of the passage is dynamically recalculated, and the weight of the topology map is updated in a timely manner, thereby improving the performance of path planning in terms of real-time and adaptability.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and in particular to a dynamic path planning method for a transport vehicle based on real-time road conditions. Background Art

[0002] Path planning is one of the core technologies in the fields of transportation, robot navigation, and unmanned driving. It mainly studies how to determine one or more optimized paths to achieve safe, efficient, economical, or comprehensive optimal route selection from the initial location to the target location.

[0003] Existing route planning solutions often rely on fixed nodes or static, pre-set routes during actual operation. These solutions lack the ability to effectively respond to dynamic changes in the real-time road network. This results in the inability to adjust planned routes in response to sudden traffic surges or sudden congestion or damage on certain road sections, leading to frequent route delays or failures during actual transportation. Furthermore, existing solutions fail to consider the load conditions of transport vehicles, effectively failing to distinguish between different travel costs under heavy or varying load conditions, resulting in uncertainty in transport efficiency and energy costs. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a dynamic path planning method for a transport vehicle based on real-time road conditions.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for dynamic path planning of a transport vehicle based on real-time road conditions, comprising the following steps:

[0006] Based on the node coordinate data in the full map, clustering is performed according to the spatial proximity and connectivity of the nodes, connecting nodes belonging to two or more different clusters are identified and extracted, and a hierarchical region boundary node set is generated;

[0007] Based on the hierarchical regional boundary node set, for any two boundary nodes in the hierarchical regional boundary node set, the underlying road network data is called to calculate the shortcut travel cost, and a shortcut travel cost list is established; based on the shortcut travel cost list, the boundary nodes are used as vertices, the shortcuts are used as edges, and the travel costs are assigned to the edge weights to construct a macro shortcut topology map;

[0008] Based on the macro shortcut topology map and the shortcut cost list, all shortcuts in the map that cover the changed road section are retrieved and matched to obtain the affected shortcut indexes; based on the affected shortcut indexes, the shortcut costs are recalculated and the corresponding edge weights in the macro shortcut topology map are updated to generate a dynamic shortcut weight map;

[0009] Based on the dynamic shortcut weight graph and the starting coordinates of the transport vehicle, the shortcut travel cost in the graph is calculated in combination with the load status of the transport vehicle, and a macro path from the starting area to the end area is searched to obtain a macro boundary node sequence; based on the macro boundary node sequence, a collaborative navigation path for the transport vehicle is generated.

[0010] Preferably, the step of obtaining the hierarchical region boundary node set is:

[0011] Based on the node coordinate data in the full map, the Euclidean distance between any two nodes is calculated and the connectivity status flag is used to mark whether there is a connection relationship. Node pairs that meet both the distance less than the spatial proximity threshold and the presence of a direct connection are clustered into the same cluster. The node sets and cluster sizes of all clusters are recorded to generate the node spatial proximity clustering results.

[0012] According to the node spatial proximity clustering results, the number of clusters to which each node belongs is counted, the belonging status of each node in each cluster is extracted, and the connection value of each node is calculated;

[0013] According to the connection value of each node, all nodes that meet the connection value greater than or equal to 1 are screened as candidate boundary nodes, and the mean and standard deviation of the connection value of all nodes are calculated. The nodes ranked in the top 20% of the connection value are selected to form a hierarchical regional boundary node set.

[0014] Preferably, the steps for obtaining the shortcut cost list are:

[0015] Based on the hierarchical regional boundary node set, all boundary node pairs are traversed, and the road segment lengths, average speeds of the road segments, and historical congestion ratios of the road segments recorded in the underlying road network data are called to construct all reachable paths between each pair of nodes, and valid shortcut path groups that meet the maximum length and maximum number of segments are selected to obtain a traversable shortcut path group;

[0016] According to the traversable shortcut path group, counting the number of road segments of each path and calculating the shortcut travel cost of each shortcut path;

[0017] According to the shortcut costs of all shortcut paths, the shortcut cost of each shortcut path is paired with the starting and ending boundary nodes, and a mapping table of corresponding boundary node pairs and time costs is constructed to form a shortcut cost list.

[0018] Preferably, the steps of obtaining the macro shortcut topology map are:

[0019] Based on the shortcut cost list, extracting the identifiers of the starting boundary node and the ending boundary node corresponding to each shortcut path in the shortcut cost list, deduplicating the boundary nodes one by one, and obtaining a complete boundary node identifier set containing all boundary nodes without duplication;

[0020] Based on the complete set of boundary node identifiers, a node structure of a topological graph is constructed according to the node identifiers, the boundary node identifiers are used as vertices of the graph, the shortcut travel cost corresponding to each shortcut path is extracted one by one from the shortcut generalized time cost list, and a direct connection mapping relationship between vertices is established according to the mapping relationship between the starting node and the ending node of the path between the nodes, thereby obtaining a vertex connection relationship mapping set;

[0021] Based on the vertex connection relationship mapping set, the shortcut travel costs corresponding to the shortcut paths are assigned to the connection mapping relationships between the corresponding vertices one by one as the weight values ​​of the edges between each vertex in the topology graph, the connection mapping relationships between the vertices are updated, and a macro shortcut topology graph is generated.

[0022] Preferably, the steps of obtaining the affected shortcut index are:

[0023] Based on the macro shortcut topology, the start and end boundary node identifiers corresponding to each shortcut path are extracted one by one. According to the start and end boundary node identifiers of each shortcut path, the road segment set information corresponding to each shortcut path is called in the shortcut generalized time cost list to generate a mapping relationship between the shortcut path and the corresponding road segment set;

[0024] According to the mapping relationship between shortcut paths and corresponding road segment sets, the road segment set information of each shortcut path is retrieved one by one, each road segment identifier contained in the road segment set is compared with the road segment identifier that changes in real time, and the shortcut path identifier of the successfully matched road segment is recorded to form an initial shortcut path identifier set covering the changed road segment;

[0025] Based on the initial shortcut path identifier set covering the changed road section, after deleting the shortcut path identifiers with duplicate records, the remaining shortcut path identifiers are sorted and indexed to generate an affected shortcut index.

[0026] Preferably, the steps of obtaining the dynamic shortcut weight map are:

[0027] Based on the affected shortcut index, each affected shortcut path in the shortcut generalized time cost list is called one by one, all road segment identifiers of each affected shortcut path are extracted, and the real-time average travel speed and real-time congestion ratio of each road segment are retrieved from the underlying road network data section by section to form a real-time road status data set;

[0028] Calculating the real-time travel time of each road segment in the affected shortcut path one by one based on the real-time road status data set, obtaining the shortcut travel cost of each affected shortcut path, and updating the shortcut travel cost of the corresponding path in the shortcut generalized time cost list to form a dynamic shortcut travel cost list;

[0029] Based on the dynamic shortcut cost list, the connection edges of the boundary node pairs corresponding to the affected shortcut paths in the macro shortcut topology graph are called one by one, the shortcut costs are assigned to the weight attributes of the corresponding connection edges one by one, the weights of all affected connection edges are updated, and a dynamic shortcut weight graph is generated.

[0030] Preferably, the steps of obtaining the macro boundary node sequence are:

[0031] Based on the dynamic shortcut weight graph and the starting coordinates of the transport vehicle, the boundary node to which the starting point belongs is located, and all complete path segments starting from the node and reaching any boundary node of the destination area are retrieved from the dynamic shortcut weight graph. Paths whose path length and number of path edges both meet the maximum constraint conditions are screened to generate a macro path candidate set;

[0032] According to the macro-path candidate set, the load equivalent time cost of each path is calculated by combining the basic travel cost and the load status of the transport vehicle in each path;

[0033] Based on the load equivalent time cost of each path, the path with the minimum load equivalent time cost is selected from the macro path candidate set as the optimal path. All boundary node identifiers contained in the path are extracted in sequence to generate a macro boundary node sequence.

[0034] Preferably, the steps for obtaining the collaborative navigation path of the transport vehicle are:

[0035] Based on the macro boundary node sequence, the original coordinate information between the starting point coordinates and the first boundary node is extracted, the geometric shape, traffic direction and current road segment status data of each road segment in the underlying road network are called, and all legal paths between the starting point coordinates and the first boundary node are traversed to form a micro path segment from the starting point to the first boundary node;

[0036] Based on the microscopic path segment from the starting point to the first boundary node, the node information between the last boundary node and the end point coordinates in the macroscopic boundary node sequence is extracted, the microscopic path segment between the last boundary node and the end point coordinates is constructed, and the end microscopic path segment set is generated;

[0037] Based on the microscopic path segment from the starting point to the first boundary node, the macroscopic boundary node sequence and the terminal microscopic path segment set, the three path segments are spliced ​​in sequence according to the path logic order to generate the collaborative navigation path of the transport vehicle.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are:

[0039] In the present invention, based on cluster analysis of node spatial proximity and connectivity, connection nodes belonging to multiple different clusters at the same time are identified and extracted, so as to achieve accurate definition of regional boundary nodes, reduce node redundancy in the path planning process, and improve the clarity of the road network structure; by calculating the shortcut costs between different boundary nodes, a clear shortcut cost list is established, and a topological map is constructed based on this, so that path selection is more accurate and efficient; when the road network status changes, the shortcuts corresponding to the changed sections are retrieved and quickly matched in real time, the travel costs are dynamically recalculated and the topological map weights are updated in time, thereby improving the real-time and adaptability of path planning; under the premise of considering the actual load status of the transport vehicle, the load-sensitive path cost is calculated according to the dynamic weight information, the macro boundary node sequence between the starting and ending points is determined, and it is refined to a complete navigation path, thereby improving the practicality and flexibility of the path planning results, improving the actual transportation efficiency, reducing costs and enhancing the reliability of the transportation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0042] See also Figure 1 The present invention provides a technical solution, a method for dynamic path planning of a transport vehicle based on real-time road conditions, comprising the following steps:

[0043] Based on the node coordinate data in the full map, clustering is performed according to the spatial proximity and connectivity of the nodes, connecting nodes belonging to two or more different clusters are identified and extracted, and a hierarchical region boundary node set is generated;

[0044] Based on the hierarchical regional boundary node set, for any two boundary nodes in the hierarchical regional boundary node set, the underlying road network data is called to calculate the shortcut travel cost and establish a shortcut travel cost list; based on the shortcut travel cost list, the boundary nodes are used as vertices, the shortcuts are used as edges, and the travel cost is assigned to the edge weight to construct a macro shortcut topology map;

[0045] Based on the macro shortcut topology and shortcut cost list, retrieve and match all shortcuts in the map that cover the changed road section to obtain the affected shortcut index; based on the affected shortcut index, recalculate the shortcut cost and update the corresponding edge weights in the macro shortcut topology to generate a dynamic shortcut weight map;

[0046] Based on the dynamic shortcut weight graph and the starting coordinates of the transport vehicle, the shortcut travel cost in the graph is calculated in combination with the load status of the transport vehicle. The macro path from the starting area to the end area is searched to obtain a macro boundary node sequence. Based on the macro boundary node sequence, a collaborative navigation path for the transport vehicle is generated.

[0047] The steps to obtain the hierarchical region boundary node set are:

[0048] Based on the node coordinate data in the full map, the Euclidean distance between any two nodes is calculated and the connectivity status flag is used to mark whether there is a connection relationship. Node pairs that meet both the distance less than the spatial proximity threshold and the presence of a direct connection are clustered into the same cluster. The node sets and cluster sizes of all clusters are recorded to generate the node spatial proximity clustering results.

[0049] According to the node spatial proximity clustering results, the number of clusters to which each node belongs is counted and the belonging status of each node in each cluster is extracted. The connection value of each node is calculated using the following formula:

[0050] ;

[0051] in, For the The connection value of each node, For the The total number of clusters to which the node belongs, For the The node in The Boolean mark of belonging to the cluster (1 if it belongs, 0 if it does not belong), is the total number of clusters, For the The number of nodes in a cluster, is the total number of nodes in the map;

[0052] According to the connection value of each node, all nodes that meet the connection value greater than or equal to 1 are screened as candidate boundary nodes, and the mean and standard deviation of the connection value of all nodes are calculated. The nodes ranked in the top 20% of the connection value are selected to form a hierarchical regional boundary node set.

[0053] Specifically, based on the node coordinate data in the full map, first traverse all the nodes recorded in the map data and extract the two-dimensional coordinates of each node and its directly adjacent node identifier list, then, for any two node pairs , calculate the spatial straight-line distance , while checking the node Whether it exists in the node In the adjacent node list, we can judge whether there is a direct connection between the two. Next, we set the spatial proximity threshold. The setting of this threshold refers to the map scale and road network density. The specific setting method is to count the lengths of all road segments in the map and calculate their average value. , and set the spatial proximity threshold to For example, if the average length of road segments in a city is 150 meters, the spatial proximity threshold is set to meters, then, starting from an unclustered node, create a new cluster, and recursively add all adjacent nodes that meet the "Euclidean distance less than 180 meters" and "direct connection relationship" with the current node to the same cluster through breadth-first search or depth-first search, and mark them as visited. Repeat this process until all nodes in the map are assigned to one or more clusters. After clustering is completed, the system will traverse all generated clusters, record all the node identifiers it contains for each cluster, and count the total number of nodes in each cluster. Finally, this information is integrated into a data structure containing a list of nodes in each cluster and the corresponding size to generate a node spatial proximity clustering result.

[0054] formula: The benefit of this formula is that it can accurately quantify the importance of each node in the road network as a "transportation hub" or "boundary connection point". It not only considers the number of clusters (breadth) connected by the node, but also gives higher value weights to nodes connected to smaller and more unique areas (clusters) by introducing the logarithmic term of the inverse document frequency (IDF) idea. This ensures that only nodes connecting two or more clusters have basic connection value, directly filtering out nodes with "bridge" properties. The summation term in the second part further distinguishes the importance of these "bridge" nodes. A node connecting multiple small, independent clusters receives a higher value score than a node connecting only two large, densely packed clusters. This design ensures that the identified boundary nodes are not only physical boundaries of the region, but also functional key connection points. This provides a high-quality vertex set for the subsequent construction of the macro shortcut graph, improving the efficiency and macro-level performance of path planning.

[0055] The parameter acquisition step is to traverse all clusters recorded in the node spatial proximity clustering result generated in the previous step, and select the specified Nodes, count how many different clusters their identifiers appear in, and the number of occurrences is For example, in the clustering results, the node "ID-007" appears in the node lists of "Cluster A", "Cluster C" and "Cluster F" at the same time. The value is 3;

[0056] The parameter acquisition step is as follows: the parameter is a Boolean flag, and its value is directly obtained from the node spatial proximity clustering result. nodes and clusters, the system checks Does the node set of the cluster contain The identifier of a node, if included, then Assign a value of 1, indicating a node Belong to cluster If not included, then Assign a value of 0;

[0057] The parameter is obtained as follows: This parameter represents the total number of clusters generated after this clustering operation. Its value is directly obtained by counting the number of independent clusters in the node spatial proximity clustering results. For example, if the clustering process finally divides the entire road network into 150 independent clusters, then The value of is 150;

[0058] The steps to obtain the parameter are as follows: The total number of nodes contained in a cluster. Its value has been calculated and stored when generating the node spatial proximity clustering result. It can be directly extracted from the result. For example, if you query the record of "Cluster C" in the clustering result, the corresponding "Cluster Size" field value is 258, then The value of is 258;

[0059] The parameter is obtained by counting the total number of nodes in the full map. The parameter represents the total number of nodes in the full map. It is a static value. During the initialization of the path planning method, the total number of nodes is counted by scanning the node database or file of the full map at one time. For example, after loading the traffic network data of a city, it is found that there are 85,430 intersections and road segment endpoints. The value is 85430.

[0060] Calculation process:

[0061] To calculate the 58th node in the road network (i.e. ) connection value For example, first obtain the determined values ​​of the parameters required by the formula from the relevant data.

[0062] Obtain the total number of nodes from the full map data .

[0063] Query the total number of clusters from the node spatial proximity clustering results .

[0064] From the node spatial proximity clustering results, it is found that node 58 belongs to three different clusters, namely clusters 12, 25, and 77. Therefore, the total number of clusters to which it belongs is .

[0065] According to the affiliation, the Boolean mark can be obtained , , , and for all other clusters , .

[0066] From the node space proximity clustering results, we can query the number of nodes contained in each of the three clusters and get , , .

[0067] Substitute the above parameter values ​​into the connection value calculation formula:

[0068] ;

[0069] because Only in When it is 12, 25, or 77, it is 1, and the rest are 0, so the sum can be simplified to:

[0070] ;

[0071] Substitution Calculate the specific value of:

[0072] ;

[0073] ;

[0074] ;

[0075] ;

[0076] ;

[0077] The results show that the connection value of node 58 is 9.209, which is a comprehensive importance score. Its value is much greater than 1, indicating that this node is a very critical connection point. It not only connects multiple areas, but also includes highly localized areas of relatively small scale (such as clusters 25 and 12) in the areas it connects. This makes it have high strategic value in macro path planning.

[0078] According to the connection value of each node, the screening process is started first, and all nodes and their corresponding connection values ​​are traversed. Value, perform a comparison judgment on each value, compare the connection value with the preset hard threshold 1, and all those that meet The identifiers of the nodes are automatically extracted and stored in a temporary candidate boundary node list. This is to initially filter out all internal nodes that do not have cross-region connection attributes. Subsequently, the system calculates the arithmetic mean and standard deviation of the connection value set of all nodes (regardless of whether they are selected in the candidate list). Next, it enters the ranking screening stage. The system ranks all nodes in the candidate boundary node list according to their connection value. Sort from high to low, and calculate 20% of the total size of the candidate list as the final number of selected nodes, and select the nodes with the highest value for connecting each pair of adjacent clusters, merge the two parts of selected nodes to remove duplicates. For example, if the candidate boundary node list contains 680 nodes, the final number of selected nodes is The system will select the first 136 nodes in the sorted list, which constitute the final required hierarchical region boundary node set.

[0079] The steps to obtain the shortcut cost list are:

[0080] Based on the hierarchical regional boundary node set, all boundary node pairs are traversed, and the road segment length, average speed, and historical congestion ratio of the road segment recorded in the underlying road network data are called to construct all reachable paths between each pair of nodes. The valid shortcut path groups that meet the maximum length and maximum number of segments are selected to obtain the accessible shortcut path groups.

[0081] According to the accessible shortcut path group, the number of road segments of each path is counted, and the shortcut travel cost of each shortcut path is calculated. The calculation formula is:

[0082] ;

[0083] in, For the The shortcut cost of the shortcut path, For the The shortcut path The length of the road section, is the average traffic speed of the road section, is the historical congestion ratio of the road section, For the The total number of road segments contained in shortcut paths, is the congestion impact factor, is the path complexity penalty factor;

[0084] According to the shortcut costs of all shortcut paths, the shortcut cost of each shortcut path is paired with the starting and ending boundary nodes, and a mapping table of corresponding boundary node pairs and time costs is constructed to form a shortcut cost list.

[0085] Specifically, based on the hierarchical regional boundary node set obtained in the previous step, the system starts the shortcut generation program, first extracts all node identifiers from the node set, and generates all unique boundary node pair combinations. For each boundary node pair (for example, node A and node B), the path search algorithm is called, specifically Yen's algorithm, to find K alternative paths connecting the two nodes on the underlying road network data, where the K value is dynamically set according to the network scale, for example, it is set to 5 for the urban core area and 3 for the suburbs. During the search process, the attributes accurately recorded for each road section in the underlying road network data will be called in real time, including the road section length in meters, the average travel speed of the road section in kilometers per hour obtained through historical vehicle speed data statistics, and the road section historical congestion ratio calculated between 0 and 1 through long-term traffic flow monitoring. After the search is completed, the system will screen the generated K paths for validity. The screening conditions include maximum length and maximum number of segments. The maximum length restriction is set based on the straight-line Euclidean distance between the two boundary nodes, specifically 2.5 times the distance. For example, if the straight-line distance between points A and B is 4 kilometers, any path with a total length of more than 10 kilometers will be eliminated. The maximum number of segments is set according to the average intersection spacing of the urban road network. For example, if there are 8 road segments per kilometer on average, then for a path with a maximum length of 10 kilometers, the maximum number of segments can be set to 90. All paths that meet both restrictions are classified as valid shortcuts and are summarized to obtain a group of accessible shortcut paths.

[0086] formula: The benefit of the formula is that it builds a comprehensive travel cost model that goes beyond the simple consideration of time or distance. The first part of the summation term is introduced by the historical congestion ratio. and congestion impact factors , the ideal travel time is corrected to reality, the denominator The attenuation effect of congestion on actual traffic speed is simulated, making the cost calculation closer to the real road conditions. Part II Then a penalty for path complexity is introduced, through the logarithmic function Applying a nonlinear penalty to the total number of road segments included in a route effectively prevents overly winding and turn-heavy routes while also avoiding excessive penalties for slightly longer but necessary routes. This dual optimization of time and complexity results in a faster, more pleasant driving experience, reducing operational complexity and potential risks for the transport vehicle.

[0087] The steps to obtain the parameter are as follows: The shortcut path The length of the road segment is directly obtained from the underlying road network database, which pre-stores the geometric information of the surveyed road segments. The unit is meter. For example, when querying "XX Road Section 3" included in path c, the length attribute value directly read from the database is 450 meters;

[0088] The parameter is obtained by performing a big data analysis on speed information collected from geomagnetic coils, video detectors, and ride-hailing vehicle data deployed on the road network over the past six months. The speeds during different time periods (such as morning peak, evening peak, off-peak, and nighttime) are weighted averaged to obtain a comprehensive average speed value in kilometers per hour. For example, after data analysis, the comprehensive average speed of "XX Road Section 3" on weekdays is calculated and recorded as 35 kilometers per hour.

[0089] The steps for obtaining the parameter are as follows: the parameter is the historical congestion ratio of the road segment, which is calculated based on the historical travel time data, and the calculation formula is: ,in is the historical average travel time, The free flow time is calculated based on the legal speed limit. Both time values ​​are derived from historical traffic data analysis. The ratio ranges from 0 to 1. For example, the legal speed limit for "XX Road Section 3" is 50 km / h, and its free flow time is , while the historical average travel time is , then its congestion ratio is ;

[0090] The steps to obtain the parameter are as follows: The total number of road segments included in the shortcut path is obtained by directly counting the number of road segment identifiers contained in the path data structure after the path search algorithm (such as Yen's algorithm) generates the path. This is a procedural statistic and does not require additional data collection. For example, if path c consists of "segment 1-segment 2-segment 3", then ;

[0091] The steps to obtain the parameters are: congestion impact factor It is a calibration parameter used to adjust the impact of historical congestion ratio on travel time. Its value is determined by regression analysis on a set of sample routes containing actual travel time. The goal is to minimize the root mean square error between the model prediction time and the actual time. The specific process is to select 100 routes covering different congestion levels and test For all values ​​in the interval [0.5, 1.0] with a step size of 0.05, calculate each The root mean square error between the predicted time and the actual time under the value is selected to minimize the error. For example, by calculating the value of When , the root mean square error is the smallest, so we set ;

[0092] The steps to obtain the parameters are: path complexity penalty factor The model is used to quantify the impact of route tortuosity on cost. It is based on driver preference surveys and operational efficiency analysis. By showing 20 experienced drivers multiple pairs of routes (similar in time but different in complexity) and recording their choices, a preference model is established. The goal is to set a The value is such that the minimum cost path calculated by the formula is consistent with the preference of more than 75% of drivers. For example, the analysis found that when When , the model selection matches the driver's preference the most, reaching 78%, so the final setting .

[0093] Calculation process:

[0094] To calculate a specific shortcut path The cost of travel For example, the path consists of 3 road segments (i.e. ), the parameter values ​​of each road segment are as follows:

[0095] Road segment 1 (g=1): length km, average speed km / h, historical congestion ratio .

[0096] Road segment 2 (g=2): length km, average speed km / h, historical congestion ratio .

[0097] Road segment 3 (g=3): length km, average speed km / h, historical congestion ratio .

[0098] Congestion impact factors , path complexity penalty factor .

[0099] First, calculate the congestion-corrected travel time (in hours) for each road segment in brackets:

[0100] Section 1 time: Hour.

[0101] Section 2 time: Hour.

[0102] Section 3 time: Hour.

[0103] Then, add up the time for each section to get the total basic travel time:

[0104] ;

[0105] Next, calculate the path complexity penalty factor:

[0106] ;

[0107] Finally, multiply the two parts together to get the final shortcut cost:

[0108] ;

[0109] Convert the result to minutes: minute.

[0110] The results show that after considering historical congestion and path complexity, the path The shortcut travel cost is equivalent to 6.20 minutes.

[0111] According to the shortcut costs of all shortcut paths, the list construction phase begins. The data processed in this phase is the passable shortcut path group generated for each boundary node pair in the previous step and its corresponding shortcut cost value. First, an empty data structure is initialized to store the final shortcut cost list. Then, each boundary node pair is traversed, for example (boundary node A, boundary node B), and all passable shortcuts connecting this pair of nodes and their costs are retrieved. All shortcut costs of this node pair are compared, and the cost with the smallest value is selected, as well as the lowest cost. The corresponding only shortcut path is considered to be the optimal shortcut between A and B. Subsequently, the system creates a new mapping record, which contains three core fields: the starting boundary node identifier (for example, 'A'), the ending boundary node identifier (for example, 'B'), and the selected lowest shortcut cost (for example, 6.20). This record is added to the final list. This process will be repeated for all boundary node pairs with traversable shortcuts until all node pairs are traversed, and finally a complete and non-redundant shortcut cost list is formed.

[0112] The steps to obtain the macro shortcut topology map are:

[0113] Based on the shortcut cost list, extract the identifiers of the starting and ending boundary nodes corresponding to each shortcut path in the shortcut cost list, deduplicate the boundary nodes one by one, and obtain a complete set of boundary node identifiers that includes all boundary nodes without duplication;

[0114] Based on the complete set of boundary node identifiers, the node structure of the topological graph is constructed according to the node identifiers. The boundary node identifiers are used as the vertices of the graph. The shortcut travel cost corresponding to each shortcut path is extracted one by one from the shortcut generalized time cost list. According to the mapping relationship between the starting node and the ending node of the path between the nodes, the direct connection mapping relationship between the vertices is established to obtain the vertex connection relationship mapping set;

[0115] Based on the vertex connection relationship mapping set, the shortcut travel costs corresponding to the shortcut paths are assigned to the connection mapping relationships between the corresponding vertices one by one, which are used as the weight values ​​of the edges between each vertex in the topology graph. The connection mapping relationships between the vertices are updated to generate a macro shortcut topology graph.

[0116] Specifically, based on the shortcut cost list, a hash set (HashSet) data structure is initialized to store all unique boundary node identifiers. Subsequently, the program reads the records in the shortcut cost list one by one. For each record, for example, a shortcut from the boundary node "BN-034" to "BN-152" is implemented, and the starting boundary node identifier "BN-034" and the ending boundary node identifier "BN-152" are extracted respectively, and these two identifiers are tried to be added to the hash set one after another. By utilizing the natural deduplication property of the hash set, even if "BN-034" or "BN-152" appears again in subsequent records, only one copy will be retained in the set. This process continues until all records in the shortcut cost list are processed. Finally, the program organizes all non-duplicate boundary node identifiers stored in the hash set and converts them into a list or array structure to obtain a complete boundary node identifier set containing all boundary nodes without duplication.

[0117] Based on the complete set of boundary node identifiers generated in the previous step, the basic skeleton of the graph is constructed. First, the system initializes a graph data structure in the form of an adjacency list and traverses the complete set of boundary node identifiers. For each unique boundary node identifier in the set (for example, "BN-034"), a corresponding vertex is created in the adjacency list and the adjacency list of the vertex is initialized to empty. After all vertices are created, the system traverses the shortcut cost list again. For each record in the list, for example, the shortcut connecting "BN-034" and "BN-152" is recorded. The system then searches for these two vertices in the adjacency list and establishes a bidirectional connection between them. The specific operation is to add the identifier of "BN-152" to the adjacency list of "BN-034", and also add the identifier of "BN-034" to the adjacency list of "BN-152". This process does not assign weight values ​​such as cost or distance to the connection, but only records the connectivity between nodes. By performing the same operation on all records in the shortcut cost list, the system finally constructs an unweighted graph that describes the shortcut connectivity between all boundary nodes, and obtains a set of vertex connection relationship mappings.

[0118] Based on the established vertex connection relationship mapping set and the original shortcut cost list, the system begins to assign weights to the edges of the topological graph. The specific operation is to traverse the shortcut cost list for the third time. For each record in the list, for example, a record containing the starting node "BN-034", the ending node "BN-152" and the shortcut cost "6.20 minutes", the system will locate the connection between "BN-034" and "BN-152" in the vertex connection relationship mapping set (i.e., the adjacency list) and assign the value "6.20" as the weight of this connection edge. In the adjacency list data structure, this means updating the simple identifier in the adjacency list to a composite object containing the neighbor identifier and the corresponding weight. For example, the adjacency list of "BN-034" updates "BN-152" to {neighbor: "BN-152", weight: 6.20}, and does the same for the adjacency list of "BN-152". The system processes all entries in the shortcut cost list in turn, assigns the cost of each shortcut to the corresponding edge in the graph, and finally completes the weight assignment of all connected edges, generating a macro shortcut topology graph with a complete structure and clear weights.

[0119] The steps to obtain the affected shortcut index are:

[0120] Based on the macro shortcut topology, the start and end boundary node identifiers of each shortcut path are extracted one by one. Based on the start and end boundary node identifiers of each shortcut path, the road segment set information corresponding to each shortcut path is called in the shortcut generalized time cost list to generate a mapping relationship between the shortcut path and the corresponding road segment set;

[0121] According to the mapping relationship between shortcut paths and corresponding road segment sets, the road segment set information of each shortcut path is retrieved one by one, each road segment identifier contained in the road segment set is compared with the road segment identifier that changes in real time, and the shortcut path identifier of the successfully matched road segment is recorded to form an initial shortcut path identifier set covering the changed road segment;

[0122] Based on the initial shortcut path identifier set covering the changed road section, after deleting the shortcut path identifiers with duplicate records, the remaining shortcut path identifiers are sorted and indexed to generate an affected shortcut index.

[0123] Specifically, based on the macro shortcut topology, the system starts the mapping relationship construction program, first initializes a key-value pair storage structure, such as a hash table (HashMap), which is used to store the final mapping relationship, where the key is the unique identifier of the shortcut path (defined by the start and end boundary node pair), and the value is the list of road segment identifiers contained in the shortcut path. The system traverses each edge in the macro shortcut topology, and each edge represents an optimal shortcut. For each edge, the system extracts the two vertices connected to it, namely the start and end boundary node identifiers, such as "BN-034" and "BN-152". Then, the system uses this pair of node identifiers as the query index In the detailed record version of the shortcut cost list generated in the previous step (this version needs to additionally store the specific road segment sequence contained in each optimal shortcut), the corresponding road segment set information is accurately searched and called. This information is an ordered list containing a series of road segment unique identifiers, for example, ["RD-101", "RD-105", "RD-233"]. Finally, the system uses the shortcut path identifier formed by this pair of start and end boundary node identifiers as the key and the queried road segment set as the value to store it in the hash table. This process is repeated until all shortcut edges in the graph have been processed, and a mapping relationship between shortcut paths and corresponding road segment sets is generated.

[0124] Based on the mapping relationship between the shortcut path and the corresponding road segment set generated in the previous step, the system starts to perform matching retrieval of the changed road segments. First, the system obtains a list of road segments whose current status has changed from the real-time traffic monitoring center. This list contains the unique identifiers of all road segments whose traffic capacity has been significantly reduced due to traffic accidents, temporary construction or severe congestion, such as ["RD-105", "RD-567"]. Then, the system goes through each record in the mapping relationship between the shortcut path and the corresponding road segment set one by one. For each shortcut path, the system obtains its corresponding road segment set and compares each road segment identifier in the set with all identifiers in the real-time changed road segment list one by one. For example, when the shortcut from "BN-034" to "BN-152" is retrieved, the system will check its road segment set ["RD-101", "RD-105", "RD-233"] and find that "RD-105" successfully matches "RD-105" in the real-time changed road segment list. Once the match is successful, the system immediately records the identifier of the shortcut path (i.e. "BN-034" to "BN-152") into a temporary list. This process will continue until all shortcut paths are retrieved, forming an initial set of all shortcut paths that cover at least one changed road segment, i.e., the initial shortcut path identifier set covering the changed road segment.

[0125] Based on the initial set of shortcut path identifiers covering the changed sections, the system starts the process of generating the final index. First, since the same shortcut path may be recorded repeatedly in the previous link because it contains multiple changed sections, the system needs to deduplicate this initial set. Specifically, it can remove all duplicate shortcut path identifiers by converting the list into a hash set and then back into a list to ensure that each affected shortcut is recorded only once. After the deduplication operation is completed, the system will sort the remaining, unique shortcut path identifiers. The sorting basis can be the lexicographic order of the shortcut path identifiers, or according to some attribute of the shortcut (such as the original travel cost). For example, the system can sort in ascending order according to the alphabetical and numerical order of the starting boundary node identifier. After the sorting is completed, the system assigns a continuous integer index starting from 0 to each sorted shortcut path identifier, and finally stores these indexes together with the corresponding shortcut path identifiers to generate a structured and ordered affected shortcut index.

[0126] The steps to obtain the dynamic shortcut weight graph are:

[0127] Based on the affected shortcut index, each affected shortcut path in the shortcut generalized time cost list is called one by one, and all road segment identifiers of each affected shortcut path are extracted. The real-time average speed and real-time congestion ratio of each road segment are retrieved from the underlying road network data section by section to form a real-time road status data set;

[0128] Based on the real-time road status data set, the real-time travel time of each road section in the affected shortcut path is calculated one by one, the shortcut travel cost of each affected shortcut path is obtained, and the shortcut travel cost of the corresponding path in the shortcut generalized time cost list is updated to form a dynamic shortcut travel cost list;

[0129] Based on the dynamic shortcut cost list, the connection edges of the boundary node pairs corresponding to the affected shortcut paths in the macro shortcut topology graph are called one by one, the shortcut costs are assigned to the weight attributes of the corresponding connection edges one by one, the weights of all affected connection edges are updated, and a dynamic shortcut weight graph is generated.

[0130] Specifically, based on the affected shortcut index, the data update process is initiated. Each shortcut path identifier in the index is first traversed. For each identifier, such as the shortcut from "BN-034" to "BN-152," the system queries and extracts a list of all road segment identifiers that constitute the shortcut, for example, ["RD-101," "RD-105," and "RD-233," based on the mapping relationship between the shortcut path generated in the previous step and the corresponding road segment set. Next, the system iterates over each road segment identifier in this list, using each identifier as a query key to initiate a request to the real-time traffic data center. This request retrieves the latest traffic status parameters for that road segment from the underlying road network data. These parameters include the real-time average speed (in kilometers per hour) calculated based on vehicle traffic detector data over the past five minutes, and the real-time congestion ratio (in the range of 0-1) calculated by comparing real-time traffic volume with the road's design capacity. The system obtains this real-time data for all road segments under each affected shortcut path and organizes it into a data set per shortcut path, forming a real-time road status data set.

[0131] Based on the real-time road status data set, the system starts the cost recalculation procedure. First, the system copies the original shortcut cost list as the basis for the update. Then, the system traverses each record in the real-time road status data set. Each record corresponds to the real-time status data of an affected shortcut path and all its sections. For each shortcut path, the same shortcut cost calculation formula as the initial cost calculation is called, but this time the calculation uses the real-time average speed and real-time congestion ratio just obtained instead of historical data. For example, for the shortcut "BN-034" to "BN-152", the system will calculate the real-time travel time for each section, then sum it up and multiply it by the path complexity penalty factor to obtain a new shortcut travel cost that reflects the current road conditions. After the calculation is completed, the system uses the unique identifier of the shortcut (i.e. "BN-034" to "BN-152") to find the corresponding entry in the copied shortcut travel cost list and update the old cost value to the newly calculated real-time cost value. This process is repeated for all affected shortcut paths, and finally a partially updated dynamic shortcut travel cost list that reflects the latest road conditions is generated.

[0132] Based on the dynamic shortcut cost list, the system begins to update the weights of the macro shortcut topology map. The specific process is as follows: the system traverses all updated entries in the dynamic shortcut cost list. For each updated record, for example, the shortcut connecting "BN-034" and "BN-152" has its cost updated to 9.85 minutes. The system then searches the macro shortcut topology map (usually stored in the form of an adjacency list) for the connecting edge defined by these two boundary node identifiers. Once this edge is located, the system updates its weight attribute value from the original historical cost to the new dynamic cost of "9.85". This operation is completed simultaneously in the adjacency list of "BN-034" and the adjacency list of "BN-152" to maintain the symmetry of the graph. The system processes all changed shortcuts in the dynamic shortcut cost list in turn, updating the weights of the corresponding edges in the graph one by one until the weights of all affected connecting edges are consistent with the latest real-time road conditions, generating a dynamic shortcut weight map.

[0133] The steps to obtain the macro boundary node sequence are:

[0134] Based on the dynamic shortcut weight graph and the starting coordinates of the transport vehicle, the boundary node to which the starting point belongs is located. All complete path segments starting from this node and reaching any boundary node in the destination area are retrieved from the dynamic shortcut weight graph. Paths whose path length and number of path edges meet the maximum constraints are selected to generate a macro path candidate set.

[0135] Based on the macro-path candidate set, combined with the basic travel cost and the load status of the transport vehicle in each path, the load equivalent time cost of each path is calculated. The calculation formula is:

[0136] ;

[0137] in, For the The load equivalent time cost of the macro path, For the The number of shortcut edges contained in the macro path, For the The shortcut travel cost of shortcut edges in the dynamic shortcut weight graph, To carry the trolley in The total load mass on the path, For the The unit load adaptation coefficient of the shortcut edge, is the load sensitivity coefficient, which is used to adjust the magnification of the impact of load on travel time;

[0138] Based on the load equivalent time cost of each path, the path with the minimum load equivalent time cost is selected from the macro path candidate set as the optimal path. All boundary node identifiers contained in the path are extracted in sequence to generate a macro boundary node sequence.

[0139] Specifically, based on the dynamic shortcut weight graph and the starting coordinates of the transport vehicle, the system first performs starting point positioning, and by calculating the Euclidean distance between the current GPS coordinates of the transport vehicle and the coordinates of all boundary nodes in the hierarchical area boundary node set, the nearest boundary node is selected as the macro starting point of this path planning. At the same time, according to the coordinates of the task end point, the same method is used to locate all boundary nodes in the area to which the end point belongs, forming an end point boundary node set. Subsequently, the system uses the macro starting point as the source point and executes the K shortest path search algorithm on the dynamic shortcut weight graph, implementing, for example, Yen's algorithm, where the K value is set to 5, to retrieve 5 paths leading to any point in the end point boundary node set. After the search is completed, the five macro paths of a node are constrained and screened. The maximum constraint on the path length is set to 1.8 times the cost of the shortest path between the center points of the starting and ending areas. For example, if the shortest path cost is 30 minutes, the maximum length constraint is 54 minutes. The maximum constraint on the number of path edges is set to 2 times the number of edges in the shortest path plus 1. For example, if the shortest path contains 4 shortcut edges, the maximum number of edges is constrained to 9. The system will check these five candidate paths one by one, eliminate any path with a total cost exceeding 54 minutes or containing more than 9 shortcut edges, and finally summarize all paths that meet the conditions to generate a macro path candidate set.

[0140] formula: The benefit of the formula is that by introducing the load state of the transport vehicle itself ( ) and the adaptability of the path to the load ( ), which expands the travel cost from a single time dimension to a comprehensive dimension of time and energy consumption (or driving difficulty). The additional penalty caused by increased load on travel time is simulated, and the penalty will change dynamically due to the different load adaptability of the path. For example, for shortcuts with large slopes and poor road conditions ( The cost penalty for heavy-loaded vehicles increases significantly when the value is low. This design allows the path planning results to intelligently avoid sections of road that appear to be shortcuts but are actually difficult to navigate and consume a lot of energy. This ensures timeliness while optimizing transportation economy and safety.

[0141] The steps to obtain the parameter are as follows: The number of shortcut edges contained in a macro path. Its value can be directly obtained from the data structure of the corresponding path in the macro path candidate set. When the path search algorithm generates a macro path, the path is represented as a series of ordered shortcut edges. It only needs to count the sequence to obtain For example, if the qth path is connected by 3 shortcut edges, then ;

[0142] The steps to obtain the parameter are as follows: The shortcut travel cost of each shortcut edge in the dynamic shortcut weight graph is the edge weight calculated and updated to the macro shortcut topology graph based on real-time traffic conditions in the previous step. Therefore, when calculating the load equivalent time cost, the weight value of each shortcut edge included in a specific macro path can be directly read from the dynamic shortcut weight graph.

[0143] The steps to obtain the parameters are as follows: the parameters are the parameters of the transport vehicle when executing the The total load mass during the transport task of a route is in kilograms. Its value is provided by the central dispatching system based on the current task order information, or is measured and uploaded in real time by the weighing sensor installed on the transport vehicle. For example, in a transport task, the transport vehicle needs to transport a batch of goods, and its total load mass is measured by the on-board weighing system to be 800 kilograms. ;

[0144] The parameter acquisition steps are as follows: the unit load adaptation coefficient is a comprehensive score that quantifies the friendliness of the h-th shortcut edge to load transportation. Its value is calculated from the underlying road segment attributes that constitute the shortcut. The calculation formula is: ,in is the base load factor, set to 1000, is the number of sections included in the shortcut, is the average slope of the i-th road section (in percentage), is the road bump index (0-1, obtained by analyzing the vehicle acceleration sensor data). For example, a shortcut consists of two sections. Section 1 has a slope of 2% and a bump index of 0.1, and section 2 has a slope of -1% and a bump index of 0.2. ;

[0145] The parameters are obtained as follows: the load sensitivity coefficient is a calibration parameter used to adjust the magnification of the load's impact on travel time. Its value is determined by regression analysis of actual driving data under different load levels. The specific operation is to organize the transport vehicle to drive on 10 standard test paths with different load adaptability in three states: empty, half-loaded (500kg), and fully loaded (1000kg). The actual travel time is recorded and the actual time increment is fitted with the model prediction increment by the least squares method ( ) to obtain the optimal For example, through regression analysis, we can find out the value of When , the model prediction accuracy is the highest.

[0146] Calculation process:

[0147] To calculate a path in the macro path candidate set Load equivalent time cost For example, the path consists of two shortcut edges (i.e. ), the load mass of the carrying trolley kg, load sensitivity factor .

[0148] path The shortcut edge attributes included are as follows:

[0149] Shortcut edge 1 (h=1): reads its shortcut cost from the dynamic shortcut weight graph Minutes, the unit load adaptation coefficient is calculated as .

[0150] Shortcut edge 2 (h=2): reads its shortcut cost from the dynamic shortcut weight graph Minutes, the unit load adaptation coefficient is calculated as .

[0151] Substitute the above parameters into the load equivalent time cost calculation formula:

[0152] ;

[0153] Calculate the load equivalent time cost of the two shortcut edges separately:

[0154] Cost of shortcut edge 1:

[0155] ;

[0156] Cost of shortcut edge 2:

[0157] ;

[0158] Adding the costs of the two shortcut edges gives the total load-equivalent time cost of the entire macropath:

[0159] ;

[0160] The results show that for a trolley with a load of 800 kg, The combined cost of the route is 23.46 minutes, which is higher than the base cost (10 + 12 = 22 minutes). The extra 1.46 minutes is due to the impact of load on efficiency. In particular, the cost of the second shortcut edge, which has poor load adaptability, increases by 8%. This final cost value serves as the basis for path selection and is used for comparison among all candidate paths.

[0161] Based on the load equivalent time cost of each path, the optimal path screening is started. First, the system traverses each path in the macro path candidate set and calls its load equivalent time cost calculated in the previous step. The system initializes two variables, one for recording the minimum cost value currently encountered, with the initial value set to a maximum value, and the other for storing the corresponding optimal path, with the initial value set to empty. During the traversal process, the system compares the load equivalent time cost of each candidate path with the currently recorded minimum cost value. For example, if the cost of the first path is 23.46 minutes, the minimum cost is updated to 23.46, and the optimal path is also updated to this path. When processing the first path, the system compares the load equivalent time cost of each candidate path with the currently recorded minimum cost value. When there are two paths, if its cost is 22.80 minutes, which is less than 23.46, the minimum cost is updated to 22.80, and the optimal path is also changed. This process continues until all candidate paths have been compared. At this time, the path stored in the optimal path variable is the final selected path. Subsequently, the system parses the structure of this optimal path, which consists of a series of ordered shortcut edges. The system extracts the boundary node identifiers connected by these shortcut edges in sequence. For example, if the optimal path is (BN-A, BN-B)->(BN-B, BN-C), the system will extract the node sequence [BN-A, BN-B, BN-C] to generate a macro boundary node sequence.

[0162] The steps for obtaining the collaborative navigation path of the transport vehicle are as follows:

[0163] Based on the macro boundary node sequence, the original coordinate information between the starting point coordinates and the first boundary node is extracted. The geometric shape, traffic direction and current road status data of each road segment in the underlying road network are called. All legal paths between the starting point coordinates and the first boundary node are traversed to form a micro path segment from the starting point to the first boundary node.

[0164] Based on the microscopic path segment from the starting point to the first boundary node, the node information between the last boundary node and the end point coordinates in the macroscopic boundary node sequence is extracted, the microscopic path segment between the last boundary node and the end point coordinates is constructed, and the end microscopic path segment set is generated;

[0165] Based on the microscopic path segment from the starting point to the first boundary node, the macroscopic boundary node sequence and the terminal microscopic path segment set, the three path segments are spliced ​​in sequence according to the path logic order to generate the collaborative navigation path of the transport vehicle.

[0166] Specifically, based on the macro boundary node sequence, the system first extracts the identifier and coordinates of the first boundary node from the sequence, and uses the current starting coordinates of the transport vehicle as the starting point of the micro path planning. Subsequently, the system matches the starting coordinates to the closest physical node in the road network through the nearest neighbor search algorithm in the underlying road network data, and uses this as the precise starting point for the path search. Next, the system uses the precise starting point as the source point and the first boundary node as the target point to execute the A* (A-Star) shortest path search algorithm on the underlying road network graph. In each step of the search, When expanding from a node to its adjacent node, the geometry of the road section recorded in the underlying road network data is called to obtain its precise length, and its travel direction attributes are checked to ensure that it does not violate traffic rules such as one-way streets. At the same time, the current road section status data is called to obtain the latest real-time travel speed. The road section length is divided by the real-time travel speed as the travel cost of the road section. The A* algorithm explores the path based on this dynamic cost, and finally traverses and selects a legal path with the lowest cost. This path, composed of a series of detailed road sections, forms a microscopic path segment from the starting point to the first boundary node.

[0167] Based on the microscopic path segment from the starting point to the first boundary node, the system continues to process the end portion of the path. First, it extracts the identifier and coordinates of the last boundary node from the macroscopic boundary node sequence and obtains the final destination coordinates of the task. Similar to the starting point processing method, the system matches the destination coordinates to the nearest physical node in the underlying road network as the precise destination of the microscopic path planning. Then, the system uses the last boundary node as the source point and the precise destination as the target point, and restarts the A* shortest path search on the underlying road network. This process is also strictly based on the geometry, traffic direction and current segment status data of each road segment in the underlying road network. The same real-time cost function is used to evaluate each possible path segment as when calculating the starting microscopic path segment. By traversing all legal paths between the last boundary node and the destination coordinates, the algorithm finally determines a path with the lowest comprehensive travel cost and stores this optimal path as a single element set to generate the final microscopic path segment set.

[0168] Based on the microscopic path segments from the starting point to the first boundary node, the macroscopic boundary node sequence, and the terminal microscopic path segment set, the system performs the splicing and integration of the final path. First, the system creates an empty ordered list to store the final complete navigation path. Then, the system adds the microscopic path segments (a detailed segment sequence) from the starting point to the first boundary node to this list. Then, the system traverses the macroscopic boundary node sequence, starting from the first node, and processes each pair of adjacent boundary nodes in turn (for example, the first and boundary nodes), for each pair of nodes, the system uses the identifiers of the pair of nodes as indexes, and queries and retrieves the detailed shortcut path connecting the two boundary nodes in the mapping relationship between the shortcut path generated in the previous step and the corresponding road segment set. The shortcut path itself is also an ordered road segment sequence. The system appends this road segment sequence to the end of the final navigation path list. This process is repeated until all shortcuts in the macro boundary node sequence are spliced ​​together. Finally, the system takes the only micro path segment from the terminal micro path segment set and appends it to the final position of the list. In this way, the three parts are seamlessly spliced ​​into a continuous and detailed complete path from the starting point to the end point, generating a collaborative navigation path for the transport vehicle.

[0169] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for dynamic path planning of a transport vehicle based on real-time road conditions, characterized in that: The following steps are involved: Based on the node coordinate data in the full map, clustering is performed according to the spatial proximity and connectivity of the nodes, connecting nodes belonging to two or more different clusters are identified and extracted, and a hierarchical region boundary node set is generated; Based on the hierarchical region boundary node set, for any two boundary nodes in the hierarchical region boundary node set, calling the underlying road network data to calculate the shortcut travel cost, and establishing a shortcut travel cost list; Based on the shortcut cost list, the boundary nodes are used as vertices, the shortcuts are used as edges, and the cost is assigned to the edge weight to construct a macro shortcut topology graph; Based on the macro shortcut topology map and the shortcut travel cost list, searching and matching all shortcuts covering the changed road section in the map to obtain the affected shortcut indexes; Based on the affected shortcut index, the shortcut travel cost is recalculated, and the corresponding edge weights in the macro shortcut topology graph are updated to generate a dynamic shortcut weight graph; Based on the dynamic shortcut weight graph and the starting coordinates of the transport vehicle, the shortcut travel cost in the graph is calculated in combination with the load status of the transport vehicle, and a macro path from the starting area to the end area is searched to obtain a macro boundary node sequence; based on the macro boundary node sequence, a collaborative navigation path for the transport vehicle is generated.

2. The method for dynamic path planning of a transport vehicle based on real-time road conditions according to claim 1, characterized in that: The steps for obtaining the hierarchical region boundary node set are: Based on the node coordinate data in the full map, the Euclidean distance between any two nodes is calculated and the connectivity status flag is used to mark whether there is a connection relationship. Node pairs that meet both the distance less than the spatial proximity threshold and the presence of a direct connection are clustered into the same cluster. The node sets and cluster sizes of all clusters are recorded to generate the node spatial proximity clustering results. According to the node spatial proximity clustering results, the number of clusters to which each node belongs is counted, the belonging status of each node in each cluster is extracted, and the connection value of each node is calculated; According to the connection value of each node, all nodes that meet the connection value greater than or equal to 1 are screened as candidate boundary nodes, and the mean and standard deviation of the connection value of all nodes are calculated. The nodes ranked in the top 20% of the connection value are selected to form a hierarchical regional boundary node set.

3. The method for dynamic path planning of a transport vehicle based on real-time road conditions according to claim 1, characterized in that: The steps for obtaining the shortcut cost list are as follows: Based on the hierarchical regional boundary node set, all boundary node pairs are traversed, and the road segment lengths, average speeds of the road segments, and historical congestion ratios of the road segments recorded in the underlying road network data are called to construct all reachable paths between each pair of nodes, and valid shortcut path groups that meet the maximum length and maximum number of segments are selected to obtain a traversable shortcut path group; According to the traversable shortcut path group, counting the number of road segments of each path and calculating the shortcut travel cost of each shortcut path; According to the shortcut costs of all shortcut paths, the shortcut cost of each shortcut path is paired with the starting and ending boundary nodes, and a mapping table of corresponding boundary node pairs and time costs is constructed to form a shortcut cost list.

4. The method for dynamic path planning of a transport vehicle based on real-time road conditions according to claim 1, characterized in that: The steps for obtaining the macro shortcut topology map are as follows: Based on the shortcut cost list, extracting the identifiers of the starting boundary node and the ending boundary node corresponding to each shortcut path in the shortcut cost list, deduplicating the boundary nodes one by one, and obtaining a complete boundary node identifier set containing all boundary nodes without duplication; Based on the complete set of boundary node identifiers, a node structure of a topological graph is constructed according to the node identifiers, the boundary node identifiers are used as vertices of the graph, the shortcut travel cost corresponding to each shortcut path is extracted one by one from the shortcut generalized time cost list, and a direct connection mapping relationship between vertices is established according to the mapping relationship between the starting node and the ending node of the path between the nodes, thereby obtaining a vertex connection relationship mapping set; Based on the vertex connection relationship mapping set, the shortcut travel costs corresponding to the shortcut paths are assigned to the connection mapping relationships between the corresponding vertices one by one as the weight values ​​of the edges between each vertex in the topology graph, the connection mapping relationships between the vertices are updated, and a macro shortcut topology graph is generated.

5. The method for dynamic path planning of a transport vehicle based on real-time road conditions according to claim 1, characterized in that: The steps for obtaining the affected shortcut index are: Based on the macro shortcut topology, the start and end boundary node identifiers corresponding to each shortcut path are extracted one by one. According to the start and end boundary node identifiers of each shortcut path, the road segment set information corresponding to each shortcut path is called in the shortcut generalized time cost list to generate a mapping relationship between the shortcut path and the corresponding road segment set; According to the mapping relationship between shortcut paths and corresponding road segment sets, the road segment set information of each shortcut path is retrieved one by one, each road segment identifier contained in the road segment set is compared with the road segment identifier that changes in real time, and the shortcut path identifier of the successfully matched road segment is recorded to form an initial shortcut path identifier set covering the changed road segment; Based on the initial shortcut path identifier set covering the changed road section, after deleting the shortcut path identifiers with duplicate records, the remaining shortcut path identifiers are sorted and indexed to generate an affected shortcut index.

6. The method for dynamic path planning of a transport vehicle based on real-time road conditions according to claim 1, characterized in that: The steps for obtaining the dynamic shortcut weight graph are: Based on the affected shortcut index, each affected shortcut path in the shortcut generalized time cost list is called one by one, all road segment identifiers of each affected shortcut path are extracted, and the real-time average travel speed and real-time congestion ratio of each road segment are retrieved from the underlying road network data section by section to form a real-time road status data set; Calculating the real-time travel time of each road segment in the affected shortcut path one by one based on the real-time road status data set, obtaining the shortcut travel cost of each affected shortcut path, and updating the shortcut travel cost of the corresponding path in the shortcut generalized time cost list to form a dynamic shortcut travel cost list; Based on the dynamic shortcut cost list, the connection edges of the boundary node pairs corresponding to the affected shortcut paths in the macro shortcut topology graph are called one by one, the shortcut costs are assigned to the weight attributes of the corresponding connection edges one by one, the weights of all affected connection edges are updated, and a dynamic shortcut weight graph is generated.

7. The method for dynamic path planning of a transport vehicle based on real-time road conditions according to claim 1, characterized in that: The steps for obtaining the macro boundary node sequence are: Based on the dynamic shortcut weight graph and the starting coordinates of the transport vehicle, the boundary node to which the starting point belongs is located, and all complete path segments starting from the node and reaching any boundary node of the destination area are retrieved from the dynamic shortcut weight graph. Paths whose path length and number of path edges both meet the maximum constraint conditions are screened to generate a macro path candidate set; According to the macro-path candidate set, the load equivalent time cost of each path is calculated by combining the basic travel cost and the load status of the transport vehicle in each path; Based on the load equivalent time cost of each path, the path with the minimum load equivalent time cost is selected from the macro path candidate set as the optimal path. All boundary node identifiers contained in the path are extracted in sequence to generate a macro boundary node sequence.

8. The method for dynamic path planning of a transport vehicle based on real-time road conditions according to claim 1, characterized in that: The steps for obtaining the collaborative navigation path of the transport vehicle are as follows: Based on the macro boundary node sequence, the original coordinate information between the starting point coordinates and the first boundary node is extracted, the geometric shape, traffic direction and current road segment status data of each road segment in the underlying road network are called, and all legal paths between the starting point coordinates and the first boundary node are traversed to form a micro path segment from the starting point to the first boundary node; Based on the microscopic path segment from the starting point to the first boundary node, the node information between the last boundary node and the end point coordinates in the macroscopic boundary node sequence is extracted, the microscopic path segment between the last boundary node and the end point coordinates is constructed, and the end microscopic path segment set is generated; Based on the microscopic path segment from the starting point to the first boundary node, the macroscopic boundary node sequence and the terminal microscopic path segment set, the three path segments are spliced ​​in sequence according to the path logic order to generate the collaborative navigation path of the transport vehicle.

Images