Multimodal transport path dynamic planning method based on adaptive large-scale neighborhood search

Abstract

The invention relates to the technical field of logistics transportation, and discloses a multimodal transport path dynamic planning method based on adaptive large-scale neighborhood search, and the method comprises the steps: obtaining and preprocessing multi-source real-time data, extracting dynamic disturbance information and calculating a disturbance value, if the disturbance value exceeds a preset threshold value, adjusting the weight and state of a disturbance node, and if the disturbance value exceeds the preset threshold value, adjusting the weight and state of the disturbance node. A disturbance response node graph is generated, path connectivity verification is carried out, path timeliness is calculated in a grouping mode, a candidate switching sequence is generated, if load overrun nodes exist in the candidate switching sequence, standby nodes are used for replacing and calculating path total resource consumption, an optimized switching combination is obtained, and the distance and time of the path are calculated; and performing a disturbance influence test, if the data consistency score is lower than a preset threshold value, adjusting the node weight of the path, extracting an execution parameter, transmitting the execution parameter to a transmission control system, generating a dynamic adjustment scheduling instruction, and determining a transportation execution sequence. According to the method, environment changes can be sensed in real time, and decisions can be made quickly.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of logistics transportation, and in particular to a multi-modal transport path dynamic programming method based on adaptive large-scale neighborhood search. BACKGROUND

[0002] At present, as an important organization form of modern logistics, multi-modal transport can effectively improve transport efficiency, reduce comprehensive cost, and reduce environmental pollution to a certain extent by integrating railway, highway, water transport and air transport and other transport modes. However, under the trend of dynamic and complex supply chain environment, multi-modal transport faces problems such as frequent transport network disturbance, high node state uncertainty, and diversified constraint conditions. How to use reinforcement learning method to plan path and select node in dynamic environment has become a technical problem in the industry.

[0003] In one prior art, the multi-modal transport path is usually arranged according to the static transport plan or historical statistical data, and the transport scheme is generated according to the fixed timetable and average transport time before the transport task starts. However, this method only relies on static data and cannot timely reflect real-time disturbance events occurring in the transport environment. When traffic congestion, weather changes or port operation delays occur in actual operation, the preset scheme often deviates from the reality. As the transport process continues, without global coordination across modes and nodes, even if the local adjustment is appropriate, it will also cause the overall efficiency to decline, and even form new transport bottlenecks. At the same time, due to the non-uniform standards and inconsistent update frequencies of multi-source heterogeneous data in collection and integration, there are delays and missing information, which further weakens the dynamic perception ability of the system to the node state, so that the path adjustment is often delayed.

[0004] In summary, the prior art has the problem of being unable to perceive environmental changes in real time and make decisions quickly. SUMMARY

[0005] The present application provides a multi-modal transport path dynamic programming method based on adaptive large-scale neighborhood search to realize real-time perception of environmental changes and rapid decision-making.

[0006] In a first aspect, to solve the above technical problems, the present application provides a multi-modal transport path dynamic programming method based on adaptive large-scale neighborhood search, comprising: Obtaining multi-source real-time data and preprocessing the multi-source real-time data to obtain a refined feature vector set; Extracting dynamic disturbance information from the refined feature vector set and calculating a disturbance value, if the disturbance value exceeds a preset disturbance value threshold, adjusting the node weight and state of the disturbance node, and generating a disturbance response node graph; perform path connectivity verification according to the perturbation response node graph to obtain a potential path set; group the potential path set, calculate path timeliness of each group of paths, generate a candidate switching sequence, if there is a load over-limit node in the candidate switching sequence, replace it with a backup node and calculate total resource consumption of the path to obtain an optimized switching combination; calculate the distance and time of the path in the optimized switching combination and perform perturbation impact test to generate a disturbed path set, if the data consistency score of the disturbed path set is lower than a preset consistency threshold, adjust the node weight of the path to obtain a verified path scheme; extract the execution parameters from the verified path scheme and transmit them to a transmission control system to generate dynamic adjustment scheduling instructions and determine a transportation execution sequence.

[0007] In an optional implementation, the obtaining of the multi-source real-time data and the preprocessing of the multi-source real-time data to obtain a refined feature vector set comprises: perform time sequence alignment on the multi-source real-time data according to timestamps and frequencies to generate a time-synchronized data set; convert the time-synchronized data set into a data set of uniform structure using a preset format rule; perform interpolation filling on the data set of uniform structure to obtain a standardized real-time data set; extract key features from the standardized real-time data set, perform smoothing processing and outlier filtering on the key features to obtain a refined feature vector set.

[0008] In an optional implementation, the extraction of dynamic perturbation information from the refined feature vector set and the calculation of a perturbation value, if the perturbation value exceeds a preset perturbation value threshold, adjustment of the node weight and state of a perturbation node, generation of a perturbation response node graph, comprises: extract dynamic perturbation information containing weather mutation information or road congestion information from the refined feature vector set; perform smoothing processing on the dynamic perturbation information and calculate a perturbation value; if the perturbation value exceeds a preset perturbation value threshold, adjust the node resource allocation weight according to a preset transportation task priority to obtain a node state; arrange nodes according to the node state and the dynamic perturbation information to generate a perturbation response node graph.

[0009] In an optional implementation, the path connectivity verification according to the perturbation response node graph to obtain a potential path set comprises: group the nodes in the perturbation response node graph to obtain a grouped node path; Check the physical connection state and historical connection data record between nodes in the grouped node path, and obtain a check result; Based on the check result, exclude unfeasible paths from the grouped node path, and obtain a potential path set.

[0010] In an optional implementation, after grouping the potential path set, the path time efficiency of each group of paths is calculated to generate a candidate switching sequence. If there is a load over-limit node in the candidate switching sequence, a backup node is used to replace and the total resource consumption of the path is calculated to obtain an optimized switching combination, including: According to the potential path set, the switching of the nodes is simulated, the path time efficiency before and after switching is calculated, and the stability of the path is evaluated to obtain a candidate switching sequence; The node information of the candidate switching sequence is extracted, and it is judged whether there is a load over-limit node. If there is, the backup node is inserted from a preset backup node pool according to a preset priority rule to obtain an updated node set; According to the updated node set, the transportation path is calculated, and the path data is compared through a preset global index database. If the comparison result meets a preset similarity threshold, the total resource consumption of the path is calculated, the stability of the path is evaluated, and an optimized switching combination is obtained.

[0011] In an optional implementation, the distance and time of the path in the optimized switching combination are calculated and perturbation influence test is performed to generate a perturbed path set. If the data consistency score of the perturbed path set is lower than a preset consistency threshold, the node weight of the path is adjusted to obtain a verified path scheme, including: According to the optimized switching combination, a transportation path set is obtained from a preset path template database, and the distance and time of the shortest path are calculated to obtain a preliminary path set; The nodes of the path in the preliminary path set are perturbed to generate a perturbed path set, and the data consistency score is calculated. If the data consistency score is lower than a preset consistency threshold, the node weight of the perturbed path set is adjusted to obtain an adjusted path set; The stability of the adjusted path set is evaluated to obtain a verified path scheme.

[0012] In an optional implementation, the execution parameters are extracted from the verified path scheme and transmitted to a transportation control system to generate a dynamic adjustment scheduling instruction, determine a transportation execution sequence, including: The path stability and node weight are obtained from the verified path scheme, the execution parameters including transportation priority are extracted, and initial execution parameters are obtained; According to the initial execution parameter, the real-time data is compared with the historical data to generate a dynamically adjusted scheduling instruction, and an adjusted instruction set is determined; If the instruction update frequency of the adjusted instruction set is lower than a preset frequency threshold, the transportation priority and the node weight are optimized, the transportation sequence is rearranged, and an optimized transportation sequence is obtained; The optimized transportation sequence and the scheduling instruction are pushed to a transportation control system to generate a transportation execution sequence.

[0013] In a second aspect, the present application provides a multi-modal transport path dynamic programming system based on adaptive large-scale neighborhood search, comprising: A data processing module is configured to acquire multi-source real-time data and pre-process the multi-source real-time data to obtain a refined feature vector set; A disturbance response module is configured to extract dynamic disturbance information from the refined feature vector set and calculate a disturbance value, and if the disturbance value exceeds a preset disturbance value threshold, adjust the node weight and state of a disturbance node to generate a disturbance response node graph; A path generation module is configured to perform path connectivity verification according to the disturbance response node graph to obtain a potential path set; A path optimization module is configured to calculate the path time efficiency of each group of paths after grouping the potential path set, generate a candidate switching sequence, and if there is a load overrun node in the candidate switching sequence, replace it with a backup node and calculate the total resource consumption of the path to obtain an optimized switching combination; A path verification module is configured to calculate the distance and time of the paths in the optimized switching combination and perform disturbance influence testing to generate a disturbed path set, and if the data consistency score of the disturbed path set is lower than a preset consistency threshold, adjust the node weight of the path to obtain a verified path scheme; An instruction generation module is configured to extract execution parameters from the verified path scheme and transmit them to a transportation control system to generate a dynamically adjusted scheduling instruction and determine a transportation execution sequence.

[0014] In a third aspect, the present application further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the multi-modal transport path dynamic programming method based on adaptive large-scale neighborhood search according to any one of the above.

[0015] In a fourth aspect, the present application further provides a computer readable storage medium comprising a stored computer program, wherein the computer readable storage medium controls the device where it is located to execute the multi-modal transport path dynamic programming method based on adaptive large-scale neighborhood search according to any one of the above when the computer program runs.

[0016] Compared with the prior art, the present application has the following beneficial effects: (1) The present application obtains a refined feature vector set by acquiring multi-source real-time data and preprocessing, including time series alignment, structure unification, missing value interpolation, key feature extraction and smoothing filtering. The heterogeneous, multi-source, noisy and missing data are uniformly processed, which can convert complex network state information into high-credibility and continuous dynamic input, not only improving the accuracy of disturbance identification, but also reducing the misjudgment probability, and realizing high-precision perception of network operation state in complex and variable transportation environment.

[0017] (2) The present application extracts dynamic disturbance information from the refined feature vector set and calculates the disturbance value, adjusts the node weight and state in combination with the transportation task priority, generates a disturbance response node graph, so that the path selection is no longer dependent on the static topology, ensures that the key nodes are given priority in planning, thereby reducing the path failure caused by sudden events, and realizing the adaptive response of the transportation network under dynamic disturbance.

[0018] (3) The present application forms an optimized switching combination by path connectivity verification, path time-efficiency grouping, load constraint detection and backup node replacement on the disturbance response node graph, ensures that the generated candidate path is executable under network topology and physical constraints, so that local bottlenecks do not block the overall transportation process, and transportation tasks can be reasonably prioritized according to the degree of urgency, the total resource consumption of the path is quantitatively evaluated, the robustness and execution feasibility of the scheduling scheme are guaranteed, and efficient, stable and reliable scheduling of transportation tasks in complex dynamic environment is realized. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a flowchart of a multi-modal transport path dynamic planning method based on adaptive large-scale neighborhood search provided by the first embodiment of the present application; Figure 2 is a structural diagram of a multi-modal transport path dynamic planning system based on adaptive large-scale neighborhood search provided by the second embodiment of the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] Reference Figure 1The first embodiment of the present application provides a multimodal transport path dynamic programming method based on adaptive large-scale neighborhood search, comprising the following steps: S11, acquiring multi-source real-time data and preprocessing the multi-source real-time data to obtain a refined feature vector set; S12, extracting dynamic disturbance information from the refined feature vector set and calculating a disturbance value, if the disturbance value exceeds a preset disturbance value threshold, adjusting the node weight and state of the disturbance node, and generating a disturbance response node graph; S13, performing path connectivity verification according to the disturbance response node graph to obtain a potential path set; S14, grouping the potential path set and calculating the path time efficiency of each group of paths to generate a candidate switching sequence, if there is a load overrun node in the candidate switching sequence, replacing it with a backup node and calculating the total resource consumption of the path to obtain an optimized switching combination; S15, calculating the distance and time of the paths in the optimized switching combination and performing disturbance influence test to generate a disturbed path set, if the data consistency score of the disturbed path set is lower than a preset consistency threshold, adjusting the node weight of the path to obtain a verified path scheme; S16, extracting execution parameters from the verified path scheme and transmitting them to a transmission control system to generate a dynamic adjustment scheduling instruction and determine a transportation execution sequence.

[0022] In step S11, multi-source real-time data is acquired and preprocessed to obtain a refined feature vector set, comprising: aligning the multi-source real-time data according to the timestamp and frequency to generate a time-synchronized data set; converting the time-synchronized data set into a unified structure data set using a preset format rule; interpolating and filling the unified structure data set to obtain a standardized real-time data set; extracting key features from the standardized real-time data set, smoothing the key features and filtering outliers to obtain a refined feature vector set.

[0023] It should be noted that the system obtains real-time data from railways, highways, water transport, and air transport modes through a sensor network, including timestamps, vehicle positions, road condition information, weather changes, and port throughput. The system aligns the data at the same time point according to the timestamps of real-time data of different frequencies to generate a time-synchronized sequence. Specifically, taking 1 minute as the reference time interval, the data with a sampling interval less than 1 minute is selected for downsampling at intervals of 1 minute, and the data with a sampling interval greater than 1 minute is linearly interpolated according to the two sampling results to proportionally fill in the data between the two sampling intervals to obtain a time-synchronized data set. Specifically, the number of data points to be filled in is obtained by dividing the two sampling interval times by 1 minute, the ratio of the difference between the two sampling results and the number of data points to be filled in is calculated, and the difference of the filled data is obtained. The data points between the two sampling intervals are sequentially filled in at an interval of one incremental interpolation.

[0024] The time-synchronized data is converted according to a preset format rule. Specifically, a general JSON format including a timestamp, a data type, and a value field is defined, the data is unified into one format, and a unified structure data set is obtained, for example, railway position data can be converted to {timestamp: 2025-08-28 10:00:00, type: position, value: {longitude: 121.3, latitude: 31.5}}; road condition data can be converted to {timestamp: 2025-08-28 10:00:00, type: road condition, value: {congestion index: 2}} to obtain a standardized real-time data set.

[0025] After format conversion, it is judged whether there is a missing value in the standardized real-time data set. If there is a missing value, it is filled in by mean interpolation. Specifically, the data ten minutes before and after the missing value is used for judgment. If the missing value is a numerical value, the average value is calculated for filling. If it is not a numerical value, the data with the highest frequency within the range of ten minutes before and after the missing value is selected for filling.

[0026] Node load data is extracted from the standardized real-time data set. The node load data is the load of each node in the railway transport network, including the number of trains processed per hour or the dispatch interval time. For example, the extracted key features are {timestamp: 2025-08-28 10:00:00, type: load, value: {train number: 20, dispatch interval time: 3 minutes}}. The extracted key features are subjected to mean filtering processing. Specifically, in order to fully capture short-term fluctuations and avoid long delays, a 10-minute time window is selected to calculate the average value and standard deviation of the key features within the window. If a data exceeds the average value plus or minus twice the standard deviation, it is marked as abnormal data and excluded. The average value is calculated to replace the data points within the window after excluding the abnormal data to obtain a refined feature vector set.

[0027] In step S12, dynamic disturbance information is extracted from the set of refined feature vectors and a disturbance value is calculated. If the disturbance value exceeds a preset disturbance value threshold, the node weight and state of the disturbance node are adjusted, and a disturbance response node graph is generated, including: Dynamic disturbance information containing weather mutation information or road congestion information is extracted from the set of refined feature vectors; The dynamic disturbance information is smoothed and a disturbance value is calculated; If the disturbance value exceeds a preset disturbance value threshold, the node resource allocation weight is adjusted according to the preset transportation task priority, and the node state is obtained; The nodes are arranged according to the node state and the dynamic disturbance information to generate a disturbance response node graph.

[0028] It should be noted that according to the weather mutation information or road congestion information detected by the railway dispatching center, dynamic disturbance information corresponding to the time and related nodes is extracted from the set of refined feature vectors. The disturbance information is represented as changes in load data and cost parameters, such as the number of trains processed per hour changing from 20 to 10, and the dispatching interval time changing from 3 minutes to 5 minutes.

[0029] According to the extracted dynamic disturbance information, a 10-minute time window is set, the average value of the data in the time window is calculated as the smoothing value, and the average value of the data one hour before the disturbance starts is calculated as the standard value. The disturbance value is obtained by subtracting the standard value from the smoothing value.

[0030] If the disturbance value exceeds the standard value by 30%, the node resource allocation weight is adjusted according to the preset transportation task priority. The transportation task priority is as follows: high-speed trains have high priority and need to be prioritized for passage, with a weight of 0.8; freight trains have low priority, with a weight of 0.2. The 30% disturbance threshold is set based on a comprehensive analysis of historical operation data and expert experience. In most transportation networks, when the node processing capacity decreases by more than 30%, the overall transportation time efficiency fluctuates significantly, so it is used as the critical point for disturbance triggering. The adjustment of the weight is based on the urgency of the transportation task and the social benefit evaluation: high-speed trains carry large-scale passenger transportation tasks and need to ensure travel safety and punctuality, so they are given a higher weight of 0.8; freight trains have higher flexibility and lower sensitivity to real-time, so they are given a lower weight of 0.2. For example, after a node disturbance, the number of trains processed per hour becomes 10. Multiply the weight by the total processing quantity to get 8 high-speed trains and 2 freight trains processed per hour by the node. The node resource allocation strategy is obtained as a reference for subsequent scheduling. The node state is obtained by recording the node allocation strategy and the node disturbance.

[0031] According to the node state and dynamic disturbance information, the node load data is standardized, and the node load data is divided by the sum of the load data of all nodes. The standardized nodes are topologically sorted, the dynamic disturbance information of the nodes is recorded, and a disturbance response node graph is generated. Specifically, the system constructs a directed acyclic graph according to the railway network structure, regards each node as a vertex in the graph, and regards a transportation path as an edge; secondly, the in-degree of each node is calculated, and the nodes with zero in-degree are sequentially added to a scheduling queue; subsequently, the nodes are taken out from the queue one by one while recording the dynamic disturbance information of the nodes, the out-edges of the nodes are deleted, and the in-degrees of the related nodes are updated; if new nodes with zero in-degree appear, the new nodes are continuously added to the queue; the above steps are repeated until all nodes are sorted.

[0032] In step S13, path connectivity verification is performed according to the disturbance response node graph, and a potential path set is obtained, including: The nodes in the disturbance response node graph are grouped to obtain a grouped node path; The physical connection state and historical connectivity data record between the nodes in the grouped node path are checked respectively to obtain a checking result; Based on the checking result, unfeasible paths are excluded from the grouped node path to obtain a potential path set.

[0033] It should be noted that the system classifies the corresponding nodes of the connected path in the disturbance response node graph into a group to obtain the grouped node path. Specifically, first, all topological paths that meet the connectivity in the disturbance response node graph are extracted; wherein the "satisfy connectivity" standard means that there is an effective physical link between any two adjacent nodes in the path, and the link is recorded in the dispatch center log as available at the current time, and at the same time the path as a whole does not form a loop in the topological structure, and can realize the one-way reachability from the starting point to the terminal. On this basis, the nodes contained in the path are classified into the same node set according to their order relationship in the topological structure, thereby forming the corresponding grouping result. In other words, each topological path that meets the connectivity standard corresponds to an independent node set, and the nodes in the set have a clear dependence and transmission relationship. The physical connection state and historical connectivity data record between the nodes in the grouped node path are checked using the dispatch center log to obtain the checking result; wherein the physical connection state check means verifying whether the physical link between adjacent nodes exists and is in an available state based on the line topology information and real-time monitoring data recorded in the dispatch center log, such as checking whether a railway section is closed, a port berth is occupied, a road is temporarily interrupted due to construction or traffic accidents, etc., to confirm that the path link has physical reachability at the current time. The historical connectivity data record check means calling the historical transportation execution record and link stability data saved in the dispatch center log to statistically analyze the reliability of the connection between nodes in the past 30 days, such as checking the punctuality rate of a certain route or the average delay of a certain railway line. Among them, the reason for choosing 30 days as the statistical time window is that this length of time can cover a complete transportation cycle and has strong representativeness; on the other hand, a 30-day time span can reflect the long-term stability of the link, and also avoids the interference of historical data caused by a long time window, so that the link reliability evaluation is closer to the current running state. Through the physical connection state check, the current unavailable or failed path link is excluded; through the historical connectivity data record check, the link that performs poorly in operation can be identified and excluded, specifically, if the punctuality rate of a certain line recorded in the dispatch center log is less than 90% or the delay rate is higher than 10%, it is determined that the link has a high uncertainty risk, and it is excluded from the candidate path set. The potential path set is obtained after excluding the infeasible path. The "punctuality rate less than 90% or delay rate higher than 10%" standard is selected based on the service level requirements and industry experience data of the transportation system operation.On the one hand, when the punctuality rate is lower than 90%, it will cause large-scale missed stops of passengers, blocked transfer plans and time efficiency of goods distribution, and further affect the overall operation efficiency of the transportation system, and thus can be used as a critical value for judging the instability of the link; on the other hand, historical operation data shows that the line with a delay rate exceeding 10% usually has structural problems or is frequently affected by external disturbances, and it is difficult to ensure the continuity and reliability of the schedule.

[0034] In step S14, the path time efficiency of each group of paths is calculated after grouping the set of potential paths, a candidate switching sequence is generated, if there is a load over-limit node in the candidate switching sequence, a backup node is used to replace and the total resource consumption of the path is calculated to obtain an optimized switching combination. It includes: According to the set of potential paths, the switching of the nodes is simulated, the path time efficiency before and after switching is calculated, and the stability of the path is evaluated to obtain a candidate switching sequence; The node information of the candidate switching sequence is extracted, and it is judged whether there is a load over-limit node, if there is, the backup node is inserted into the updated node set according to the preset priority rule; According to the updated node set, the transportation path is calculated, and the path data is compared through the preset global index database, if the comparison result satisfies the preset similarity threshold, the total resource consumption of the path is calculated, the stability of the path is evaluated, and the optimized switching combination is obtained.

[0035] It should be noted that the system calculates the time spent by the path through all nodes according to the set of potential paths, replaces a certain node in the path with other nodes and re-plans the path, judges whether it can be connected, and re-calculates the time spent by the new path, and then evaluates the stability of the path. Specifically, the system multiplies the availability score of each node in the path (the ratio of the historical disturbance duration to the historical normal operation time calculated according to the schedule center log is divided by the maximum value of the ratio and then multiplied by 100 to convert to a percentage score) and the reliability score of each link (the historical delay rate and interruption rate are calculated according to the schedule center log, i.e. taking 100 points as full score, first deducting 0.5 points for every 1% increase in link delay rate, and then deducting 1 point for every 1% increase in link interruption rate, to obtain the link reliability score), and the current node or link load data is divided by the sum of all node or link load data as a weight, multiplied by the availability score or the reliability score, to obtain the overall stability score, which is used as the stability evaluation result of the path. When the stability is higher than 80 and the time spent is not more than 20% of the original path, the switching path is accepted, and the candidate switching sequence is obtained.

[0036] The load value of each node in the candidate switching sequence is checked. If the load value exceeds 20% of the maximum load value of the node design, it is judged as a load overrun node. A backup node is inserted from the preset backup node pool according to the preset priority rule to obtain an updated node set. The preset backup node pool is a set of alternative nodes or alternative road segments that have the same predecessor node and successor node as each node and have a time difference of no more than 20% from the predecessor node to the successor node. The preset priority rule is that the reciprocal of the path cost time after replacement is multiplied by the stability score, divided by the maximum value of the reciprocal of the path cost time after replacement and the stability score multiplied by 100 to convert to percentage, and then multiplied by 100 to convert to percentage. The priority score is obtained, and the selection is performed in the order of high to low priority score.

[0037] Each path in the updated node set is extracted and compared with the preset global index database. The preset global index database is the average path distance, time cost, and stability score calculated from historical data for each origin-destination pair. The database is updated in batches every week to reflect the overall trend of the transportation network under normal operating conditions. When a sudden event (such as extreme weather, major traffic accidents, key hub congestion or closure) causes significant fluctuations in node or path state, a local update mechanism is triggered immediately to recalculate the average path distance, actual time cost, and stability score for the affected origin-destination pairs and write them into the database in real time. The corresponding indicators of the path are calculated and compared with the global index database. The smaller the deviation, the greater the similarity, and the greater the deviation, the smaller the similarity. When the deviation of each indicator is less than 5%, the comparison is passed, the resource consumption of each segment of the path is calculated, and the total resource consumption of the path is obtained. Historical operation data shows that when the indicator deviation is controlled within 5%, the difference between the path and the global optimal solution in terms of actual transportation time, distance, and energy consumption can be ignored, and the scheduling optimization result will not be substantially affected. If the threshold is exceeded, the transportation time and resource utilization may decrease significantly, so 5% is set as the upper limit of the deviation passing. The total stability score of the path is calculated by weighted summation of the stability score of each link in the path. The link weight is obtained by dividing the link load data by the maximum load data and multiplying 100 to convert to percentage. The total stability score of the path is greater than 80, and the optimized switching combination is obtained. Through statistical analysis of historical stability distribution, when the score is less than 80, the link is prone to frequent delays, congestion, or interruptions, and the overall transportation continuity is difficult to guarantee. Links with a score above 80 are usually in a stable operating state and can meet the scheduling needs of large-scale transportation tasks.

[0038] In step S15, the distance and time of the paths in the optimization switching combination are calculated and the perturbation influence test is performed, a set of disturbed paths is generated, if the data consistency score of the set of disturbed paths is lower than the preset consistency threshold, the node weight of the path is adjusted, and a verified path scheme is obtained, including: According to the optimization switching combination, a set of transportation paths is obtained from the preset path template database, and the distance and time of the shortest path are calculated to obtain a preliminary path set; The path nodes in the preliminary path set are perturbed to generate a set of disturbed paths, and the data consistency score is calculated, if the data consistency score is lower than the preset consistency threshold, the node weight of the set of disturbed paths is adjusted to obtain an adjusted path set; The stability of the adjusted path set is evaluated to obtain a verified path scheme.

[0039] It should be noted that the set of transportation paths is obtained from the preset path template database and the path with the least resource consumption and the highest stability score is selected. The preset path template database is composed of recorded historical paths and corresponding resource consumption and stability score data. It is checked whether there is an unavailable or recently poor performing node in the selected path node, wherein the recently poor performing node is a node whose punctuality rate is less than 90% or delay rate is higher than 10% within 10 days. If there is, the set of alternative nodes in the optimization switching combination is used for replacement to obtain a preliminary path set. The standard of "punctuality rate less than 90% or delay rate higher than 10%" is selected according to the industry experience data of the transportation system operation. On the one hand, when the punctuality rate is less than 90%, it will cause large-scale mispoint of passengers, blockage of transfer plan and decline of goods distribution time, and thus affect the overall service reputation and operation efficiency of the transportation system, so it can be used as the critical value to determine the instability of the link. On the other hand, historical operation data shows that the line with a delay rate higher than 10% usually has structural problems or is frequently affected by external disturbances, and it is difficult to ensure the continuity and reliability of scheduling.

[0040] In the preliminary path set, a path is randomly selected, the nodes covered in the path are extracted, 1-5 nodes are randomly selected for disturbance, and the load data is randomly reduced or the scheduling interval time is randomly increased to simulate the disturbance that may occur in the actual situation, to generate a disturbed path set. Before the disturbance of part of the nodes, the time consumption from the starting point of the path to the end point of the path is calculated, and the time consumption from the starting point of the path to the end point of the path after the disturbance of part of the nodes is calculated. The difference between the two is calculated and divided by the actual consumption before the disturbance as the deviation before and after the disturbance. Since the smaller the deviation represents the stronger the anti-interference ability of the path, the deviation is subtracted by 1 and multiplied by 100 to convert it into a percentage score as a consistency score. If the consistency score exceeds 80 points, it proves that the anti-disturbance ability of the node is strong, and the node stability score increases by 5 points. If the consistency score is less than 80 points, it proves that the anti-disturbance ability of the node is weak, and the node stability score decreases by 5 points. Among them, according to the statistical analysis of the simulation results of the historical path disturbance, it is found that when the consistency score is higher than 80 points, the total time consumption deviation of the path is less than 5%, and the node can still maintain stable operation under high load or sudden disturbance. Therefore, 80 points is selected as the critical value to effectively distinguish the strength of the anti-disturbance ability of the node and guide the stability score adjustment. The path is re-evaluated for stability and the priority score of the path is calculated to obtain a verified path scheme. Among them, the priority score is obtained by multiplying the reciprocal of the path time consumption and the stability score, dividing by the maximum value of the reciprocal of the path time consumption and the stability score of all paths, and then multiplying by 100 to convert it into a percentage.

[0041] In step S16, the execution parameter is extracted from the verified path scheme and transmitted to the transportation control system to generate a dynamically adjusted scheduling instruction and determine a transportation execution sequence, including: The path stability and node weight are obtained from the verified path scheme, the execution parameter containing the transportation priority is extracted, and the initial execution parameter is obtained; According to the initial execution parameter, the real-time data and the historical data are compared, and the dynamically adjusted scheduling instruction is generated to determine the adjusted instruction set; If the instruction update frequency of the adjusted instruction set is lower than the preset frequency threshold, the transportation priority and the node weight are optimized, the transportation sequence is rearranged, and the optimized transportation sequence is obtained; The optimized transportation sequence and the scheduling instruction are pushed to the transportation control system to generate a transportation execution sequence.

[0042] It should be noted that the stability score and node information of each path in the verification path scheme are extracted and allocated according to the pre-set transportation task priority and path load capacity. Based on task priority, paths with low time consumption and high stability scores are assigned to high-priority, medium-priority, and low-priority tasks according to their path priority scores, generating initial execution parameters, i.e., the initial scheduling scheme. Specifically, paths are first assigned to high-priority tasks, selecting the path with the highest path priority score and not yet at saturation load. After all high-priority tasks are assigned, paths are then assigned to medium-priority and low-priority tasks in turn. The path priority score is calculated by multiplying the inverse of the path's time consumption by its stability score, dividing by the maximum value of the products of the inverse of the time consumption of all replaced paths and their stability scores, and then multiplying by 100 to convert it to a percentage.

[0043] Based on real-time node data and historical data from the past 10 hours monitored by the system, it is determined whether there are any disturbed nodes in the initial scheduling plan. Specifically, the average train handling volume over the past 10 hours is calculated as a standard value; if the real-time data exceeds the standard value by 30%, the node is considered disturbed. According to the path switching scheme in the verification path plan, the path is switched, and a new scheduling plan and scheduling execution instructions are generated.

[0044] The system finds from the historical scheduling log that when the scheduling update frequency is higher than 2 times per hour, it indicates that there is a large fluctuation or frequent disturbance in the transportation environment, and the system needs to frequently adjust the strategy to maintain the optimal state of path priority and resource allocation; when the update frequency is lower than 2 times per hour, the system path execution deviation is small, and the overall transportation network is in a relatively stable state, and further frequent update has limited effect on improving the scheduling performance. Based on this analysis, 2 times per hour is set as the empirical threshold for judging the stability of the scheduling strategy. If the scheduling execution instruction update frequency is less than 2 times per hour, in order to further optimize to improve the overall transportation efficiency and resource utilization, the system dynamically corrects the path priority based on real-time feedback data. The real-time feedback data is collected from multiple sources, including: the load rate sensor output of the transportation node, the link transmission delay log, the positioning and speed information of the vehicle or ship, the operation progress of the port or warehouse link, and the actual time spent on the path and the real-time load rate of the node are calculated. The actual time spent on the path is obtained by adding the transportation time of each link in the path and the node operation time. Specifically, the system first obtains the actual transportation time from the link transmission delay log, and combines the positioning and speed information of the vehicle or ship to estimate the actual transportation time according to the transportation distance and real-time speed; for the port or warehouse link, the node waiting or processing time is calculated according to the operation progress lag. The transportation time of all links in the path and the operation time of the node are added to obtain the total actual time spent on the path. The real-time load rate of the node is calculated according to the current processing capacity occupation of each node. The system obtains the real-time processing capacity of the node through the load rate sensor, and combines the maximum processing capacity of the node to evaluate the current load ratio of the node; if a node contains multiple operation links, the processing capacity and maximum capacity ratio of each link are weighted and averaged to obtain the real-time load rate of the entire node. Specifically, the system compares the real-time feedback data with the global index database to calculate the deviation degree of each transportation path in the execution process; if the average time spent on a path increases by more than 10% compared with the expected value, the path priority score is automatically reduced by 10%; if the load rate of a node exceeds 90%, the path priority score passing through the node is reduced by 15% accordingly, the transportation sequence is redistributed according to the adjusted path priority score, and the optimized transportation sequence is generated. The optimized transportation sequence and the scheduling instruction are pushed to the transportation control system to generate the transportation execution sequence.

[0045] In summary, this invention refines and extracts features from multi-source real-time data, eliminating redundant and noisy information to make subsequent calculations more efficient and accurately reflect the core state of the transportation environment. Based on this, a dynamic disturbance detection and node weight correction mechanism is introduced, enabling the path network to adaptively reflect sudden events in the environment and avoid failures caused by static modeling. Furthermore, path connectivity verification and potential path set generation ensure that feasible candidate solutions always exist in the transportation network under dynamic disturbances, thereby improving the robustness of the planning. During candidate path optimization, load constraints and resource consumption assessments are combined to ensure that transportation resources are rationally allocated, preventing local overload and ensuring that overall energy consumption remains within a controllable range. Subsequently, disturbance consistency verification and scheme readjustment mechanisms further enhance the reliability of path decisions, preventing overall scheme imbalance caused by a single disturbance. Finally, the execution parameters, optimized and verified at multiple levels, are rapidly distributed to the control system, realizing a closed loop from environmental perception to path scheduling, achieving real-time perception of environmental changes and rapid decision-making.

[0046] Reference Figure 2 The second embodiment of the present invention provides a multimodal transport path dynamic planning system based on adaptive large-scale neighborhood search, comprising: The data processing module is used to acquire multi-source real-time data and preprocess the multi-source real-time data to obtain a refined feature vector set; The disturbance response module extracts dynamic disturbance information from the refined feature vector set and calculates the disturbance value. If the disturbance value exceeds the preset disturbance value threshold, the node weight and state of the disturbance node are adjusted to generate a disturbance response node graph. The path generation module is used to perform path connectivity verification based on the disturbance response node graph to obtain a set of potential paths. The path optimization module is used to group the potential path set, calculate the path time of each group of paths, generate a candidate switching sequence, and if there is a node with overload in the candidate switching sequence, a backup node is used to replace it and the total resource consumption of the path is calculated to obtain an optimized switching combination. The path verification module is used to calculate the distance and time of the paths in the optimized switching combination and perform disturbance impact tests to generate a set of disturbed paths. If the data consistency score of the set of disturbed paths is lower than a preset consistency threshold, the node weights of the paths are adjusted to obtain a verified path scheme. The instruction generation module is used to extract execution parameters from the verification path scheme and transmit them to the transmission control system to generate dynamic adjustment scheduling instructions and determine the transportation execution sequence.

[0047] It should be noted that the multimodal transport path dynamic planning system based on adaptive large-scale neighborhood search provided in this embodiment of the invention is used to execute all the process steps of the multimodal transport path dynamic planning method based on adaptive large-scale neighborhood search in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0048] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program, such as a data processing program, stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps described in the various embodiments of the multimodal transport path dynamic planning method based on adaptive large-scale neighborhood search, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module in the above-described device embodiments, such as the data processing module.

[0049] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0050] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0051] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0052] The memory can be used to store the computer programs and modules. The processor implements various functions of the electronic device by running or executing the computer programs and modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0053] If the modules integrated into the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0054] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0055] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A multimodal transport route planning method based on adaptive large-scale neighborhood search, characterized in that, include: Acquire multi-source real-time data and preprocess the multi-source real-time data to obtain a refined feature vector set; Dynamic perturbation information is extracted from the refined feature vector set and the perturbation value is calculated. If the perturbation value exceeds the preset perturbation value threshold, the node weight and state of the perturbation node are adjusted to generate a perturbation response node graph. Based on the disturbance response node graph, path connectivity verification is performed to obtain a set of potential paths; After grouping the potential path set, the path timeliness of each group of paths is calculated to generate a candidate switching sequence. If there is a node with overload in the candidate switching sequence, a backup node is used to replace it and the total resource consumption of the path is calculated to obtain an optimized switching combination. Calculate the distance and time of the paths in the optimized switching combination and conduct disturbance impact tests to generate a set of disturbed paths. If the data consistency score of the set of disturbed paths is lower than a preset consistency threshold, adjust the node weights of the paths to obtain a verified path scheme. The execution parameters are extracted from the verification path scheme and transmitted to the transmission control system to generate dynamic adjustment scheduling instructions and determine the transportation execution sequence.

2. The multimodal transport route planning method based on adaptive large-scale neighborhood search according to claim 1, characterized in that, The process of acquiring multi-source real-time data and preprocessing the multi-source real-time data to obtain a refined feature vector set includes: The multi-source real-time data is time-aligned according to timestamps and frequencies to generate a time-synchronized dataset; The time synchronization dataset is transformed into a dataset with a uniform structure using preset format rules; The dataset with the unified structure is interpolated and imputed to obtain a standardized real-time dataset. Key features are extracted from the standardized real-time dataset, and smoothing and outlier filtering are performed on the key features to obtain a refined feature vector set.

3. The multimodal transport route planning method based on adaptive large-scale neighborhood search according to claim 1, characterized in that, The step of extracting dynamic perturbation information from the refined feature vector set and calculating the perturbation value, and adjusting the node weights and states of the perturbation nodes to generate a perturbation response node graph if the perturbation value exceeds a preset perturbation value threshold, includes: Extract dynamic disturbance information containing information on sudden weather changes or road congestion from the refined feature vector set; The dynamic disturbance information is smoothed and the disturbance value is calculated; If the disturbance value exceeds the preset disturbance value threshold, the node resource allocation weight is adjusted according to the preset transportation task priority to obtain the node status. The nodes are arranged according to their states and the dynamic disturbance information to generate a disturbance response node diagram.

4. The multimodal transport route planning method based on adaptive large-scale neighborhood search according to claim 1, characterized in that, The step of performing path connectivity verification based on the disturbance response node graph to obtain a set of potential paths includes: The nodes in the disturbance response node graph are grouped to obtain the grouped node paths; The physical connection status and historical connectivity data records between nodes in the grouped node paths are checked respectively to obtain the check results; Based on the inspection results, infeasible paths are excluded from the grouped node paths to obtain a set of potential paths.

5. The multimodal transport route planning method based on adaptive large-scale neighborhood search according to claim 1, characterized in that, After grouping the potential path set, the path timeliness of each group is calculated to generate a candidate switching sequence. If there are overloaded nodes in the candidate switching sequence, backup nodes are used to replace them, and the total resource consumption of the path is calculated to obtain an optimized switching combination, including: Based on the set of potential paths, the switching of simulated nodes is performed, the path timeliness before and after the switching is calculated, and the stability of the path is evaluated to obtain a candidate switching sequence. Extract the node information of the candidate switching sequence, determine whether there are any nodes with overload, and if so, insert the backup node from the preset backup node pool according to the preset priority rules to obtain the updated node set. The transportation path is calculated based on the updated node set. The path data is compared with the preset global index database. If the comparison result meets the preset similarity threshold, the total resource consumption of the path is calculated, the path stability is evaluated, and an optimized switching combination is obtained.

6. The multimodal transport route planning method based on adaptive large-scale neighborhood search according to claim 1, characterized in that, The process involves calculating the distance and time of paths in the optimized switching combination and conducting disturbance impact tests to generate a set of disturbed paths. If the data consistency score of the disturbed path set is lower than a preset consistency threshold, the node weights of the paths are adjusted to obtain a verified path scheme, including: Based on the optimized switching combination, a set of transportation routes is obtained from the preset route template database, and the distance and time of the shortest path are calculated to obtain a preliminary route set; The path nodes in the initial path set are perturbed to generate a perturbed path set. Then, the data consistency score is calculated. If the data consistency score is lower than a preset consistency threshold, the node weights of the perturbed path set are adjusted to obtain an adjusted path set. The stability of the adjusted path set is evaluated to obtain the verified path scheme.

7. The multimodal transport route planning method based on adaptive large-scale neighborhood search according to claim 1, characterized in that, The step of extracting execution parameters from the verification path scheme and transmitting them to the transmission control system to generate dynamic adjustment scheduling instructions and determine the transportation execution sequence includes: The path stability and node weights are obtained from the verified path scheme, and the execution parameters including transportation priority are extracted to obtain the initial execution parameters; Based on the initial execution parameters, real-time data is compared with historical data to generate dynamically adjusted scheduling instructions and determine the adjusted instruction set. If the instruction update frequency of the adjusted instruction set is lower than a preset frequency threshold, then the transportation priority and the node weight are optimized, the transportation sequence is rearranged, and an optimized transportation sequence is obtained. The optimized transportation sequence and scheduling instructions are pushed to the transportation control system to generate a transportation execution sequence.

8. A multimodal transport path dynamic programming system based on adaptive large-scale neighborhood search, characterized in that, include: The data processing module is used to acquire multi-source real-time data and preprocess the multi-source real-time data to obtain a refined feature vector set; The disturbance response module extracts dynamic disturbance information from the refined feature vector set and calculates the disturbance value. If the disturbance value exceeds the preset disturbance value threshold, the node weight and state of the disturbance node are adjusted to generate a disturbance response node graph. The path generation module is used to perform path connectivity verification based on the disturbance response node graph to obtain a set of potential paths. The path optimization module is used to group the potential path set, calculate the path time of each group of paths, generate a candidate switching sequence, and if there is a node with overload in the candidate switching sequence, a backup node is used to replace it and the total resource consumption of the path is calculated to obtain an optimized switching combination. The path verification module is used to calculate the distance and time of the paths in the optimized switching combination and perform disturbance impact tests to generate a set of disturbed paths. If the data consistency score of the set of disturbed paths is lower than a preset consistency threshold, the node weights of the paths are adjusted to obtain a verified path scheme. The instruction generation module is used to extract execution parameters from the verification path scheme and transmit them to the transmission control system to generate dynamic adjustment scheduling instructions and determine the transportation execution sequence.

Images