A method of recommending a cargo source to a driver

Abstract

The application discloses a method for recommending a cargo source to a driver, and realizes intelligent freight scheduling by constructing a dynamic cargo source splicing map. First, the method integrates multi-source real-time data, constructs a cargo source relationship network, and evaluates the risk coefficient of the connection edge. Based on this network, an optimization algorithm is used to generate an anti-interference set containing a main scheme and multiple backup schemes, and the scheme resilience is evaluated through simulation. During the execution process, the transportation state is continuously monitored, and when interference events such as road closure and vehicle failure are identified, the system can automatically switch to the optimal backup scheme to ensure task continuity. At the same time, through a closed-loop feedback mechanism, execution data is collected, and the risk model and decision parameters are continuously calibrated. The method significantly improves the overall resilience, resource utilization and operation efficiency of freight scheduling in complex environments, and realizes the transformation from passive response to active optimization.

Description

Technical Field

[0001] This invention relates to the field of recommendation algorithm technology, and in particular to a method for recommending cargo to drivers. Background Technology

[0002] In the traditional logistics industry, especially in scenarios such as long-haul full truckload and coastal bulk cargo, there is a large amount of planned logistics demand. Manufacturers typically issue periodic logistics plans, while carriers often accept orders on a one-way basis. This model leads to a lack of continuity in transportation tasks; after completing one shipment, vehicles often need to drive empty to the next loading point, resulting in a significant waste of transportation resources, high empty-run rates, and directly increasing overall operating costs. Furthermore, many aspects of the logistics process, such as vehicle on-the-go monitoring, handling of anomalies, and financial settlement, heavily rely on manual support. This makes the entire system unable to respond quickly to emergencies, resulting in low decision-making efficiency, and posing a significant competitive disadvantage, especially for small and medium-sized enterprises (SMEs).

[0003] To improve matching efficiency, existing technologies have proposed several solutions. For example, some methods calculate a matching score between cargo and driver plans by setting time and location matching tags, aiming to filter out transportation tasks that can be well-connected. Another approach is to recommend based on the behavioral similarity between drivers, calculating the matching score by analyzing the communication records between drivers and cargo owners, thereby recommending cargo owners contacted by similar drivers to the current driver. However, these methods mostly focus on static pre-matching and lack effective mechanisms to deal with dynamic risks during transportation execution. For example, they struggle to respond in real time to disruptive events such as road closures, vehicle breakdowns, or sudden weather changes, making it easy for pre-planned transportation plans to be interrupted during actual execution, failing to guarantee the continuity and resilience of transportation tasks. In addition, some methods fail to fully consider the resource competition problems that may arise in multi-vehicle collaborative scenarios, making it difficult to achieve optimal global scheduling for the system. Summary of the Invention

[0004] This application provides a method for recommending cargo to drivers, enabling rapid response to emergencies and automatic switching of solutions in complex transportation environments.

[0005] This application provides a method for recommending cargo to drivers, including:

[0006] S1 collects real-time data streams of cargo sources, transportation capacity, and environment. The data streams are then filtered and standardized through events, and output in a unified format with timestamps and cargo source credibility scores. When the cargo source credibility score is higher than the threshold, it is considered as valid cargo source information. S2, construct and update a dynamic cargo source mosaic based on the processed data stream, generate nodes for each cargo source, calculate the spatiotemporal connectivity between nodes to establish feasible edges, and evaluate the interference risk coefficient for each edge; S3. Based on the dynamic cargo splicing map, a set of cargo splicing schemes is generated through an optimization algorithm. The cargo splicing scheme group includes a main recommended scheme and multiple backup schemes that are pre-constructed after analyzing the vulnerabilities in the main recommended scheme. Anti-interference simulation evaluation is performed on all schemes in the group. S4 monitors the real-time processed data stream to identify interference events and automatically selects the most suitable solution from the backup solution library based on the event type and the current transportation status of the transport vehicles. S5 collects data on the results of the scheme implementation and dynamically adjusts and updates the risk coefficient in the mosaic diagram through feedback learning.

[0007] Preferably, the dynamic construction and updating of the cargo source mosaic map based on real-time data stream specifically includes: creating a corresponding graph node for each valid cargo source; calculating the spatial proximity and temporal continuity between any two nodes based on the spatiotemporal attributes of the nodes; establishing a directed edge representing the feasibility of transportation connection between a set of nodes if and only if a set of nodes simultaneously satisfies both the spatial proximity and temporal continuity conditions; evaluating a quantified interference risk coefficient for each directed edge by comprehensively considering the historical interruption frequency of the associated path and the current environmental risk information; continuously monitoring changes in the data stream and incrementally and dynamically updating the cargo source mosaic map, including creating, updating, or deleting nodes according to newly added, changed, or invalid cargo source information, as well as recalculating the associated edges and reassessing the risk coefficient.

[0008] Preferably, the step of analyzing the vulnerabilities in the main recommended solution and pre-constructing multiple backup solutions includes: starting from the vehicle's current position based on the mosaic map, selecting subsequent nodes iteratively according to a multi-objective evaluation function to generate a main recommended solution; performing vulnerability analysis on the main recommended solution to identify links in its sequence with high interference risk or strict time window constraints as vulnerabilities; for each identified vulnerability, applying time elasticity strategies and spatial redundancy strategies to generate local or global alternative paths to pre-construct multiple backup solutions; performing Monte Carlo simulations on all generated solutions to quantitatively evaluate the task completion rate and cost change rate of each solution by simulating random interference events, and calculating the stability score of the solution accordingly.

[0009] Preferably, the anti-interference simulation evaluation of all solutions within the group specifically includes: real-time monitoring of standardized data streams to identify interference events and classifying them; immediately initiating an evaluation of the current transport vehicle status upon identification of an interference event to obtain its location, load, and task progress information; based on the type of interference event and the evaluation results of the current transport vehicle status, selecting all logically feasible candidate solutions from the backup solution library; performing multi-dimensional matching degree scoring on each feasible candidate solution, and selecting the solution with the highest comprehensive matching degree as the final execution solution based on the scoring results. Preferably, step S3, which involves performing anti-interference simulation evaluation on all schemes, further includes: S31, by accessing historical data streams to extract spatiotemporal pattern factors of interference events, and constructing an enhanced predictive mosaic that integrates historical patterns and short-term trends in real time; S32, based on the enhanced predictive stitching map, generates a forward-looking cargo stitching scheme that covers multiple time windows and has different strategy orientations; S33 combines historical scenario backtesting with multi-perturbation Monte Carlo simulation to conduct a dual resilience assessment of the primary recommended scheme and the set of backup schemes.

[0010] Preferably, the stitching scheme includes: loading an enhanced prediction stitched image and obtaining the real-time status of the vehicle; setting a planning period containing multiple consecutive time windows; starting from the vehicle's current position, within the first time window, performing path searches using multiple strategies in parallel based on the enhanced prediction stitched image to generate initial path segments corresponding to different strategies; using the endpoint status of the planned path in the previous time window as the virtual starting point of the next time window, iteratively performing multi-strategy path searches to generate a complete strategy scheme covering the planning period; and outputting scheme summary information for each generated complete strategy scheme.

[0011] Preferably, the enhanced predictive mosaic includes: combining real-time vehicle coordinates, current date and time, weather codes, and calendar marker parameters to form a structured query request; extracting complete historical records of the same period and region from historical data storage based on this request to form a highly correlated dataset; sending the dataset to a time-series analysis service for real-time scanning and calculation, statistically analyzing the probability and average duration of delays on each road segment during the same period in history, comparing recent traffic flow with the historical average to calculate the regional trend percentage, and analyzing historical event chains to identify the probability of influence of related locations, thereby generating a spatiotemporal pattern factor with confidence; obtaining a basic cargo source mosaic, dynamically injecting the spatiotemporal pattern factor through a graph traversal program, adjusting the risk weight of corresponding edges and the priority of fine-tuning points in the graph according to the factor type, adding virtual connection edges between nodes where chain reactions are predicted, and outputting the enhanced predictive mosaic. Preferably, the dual resilience assessment of the primary recommended solution and the set of backup solutions includes: S321, based on the enhanced prediction stitching map, constructs a spatiotemporal path reliability map covering the next few hours on the current driving path of the vehicle involved in the main recommended plan and the backup plan, and calculates the road travel time; S322, when enhancing the stitching scheme of the prediction stitching map, dynamically binds the moving segments between each node to the spatiotemporal path reliability map, and immediately marks the risk of moving segments with road travel time below the threshold or insufficient reliability. S323 monitors the progress and route risk status of vehicles during execution. When a change in the reliability of the bound route is detected, it actively triggers local route replanning and splices the nodes in the main recommended route that need route replanning with the local routes in the backup route.

[0012] Preferably, the step of splicing the nodes requiring route replanning in the main recommended plan with the local paths in the backup plan includes: continuously tracking the vehicle's real-time location and speed while it is executing the recommended plan, dynamically querying the latest predicted travel time and reliability index of the current and subsequent road segments' bound paths, and estimating the estimated arrival time in real time. Continuously assessing core risk indicators to determine the intervention timing, when the reliability index of the bound path is detected to be lower than a dynamic threshold, or the deviation between the real-time estimated arrival time and the original predicted arrival time exceeds the allowable range, immediately triggering the replanning of the local path for the current or next travel segment. Using the vehicle's latest location or the original planned starting point as the starting point for path planning, and the originally planned next node as the endpoint, a new, comprehensively optimal path is searched again in the latest spatiotemporal path reliability map. The evaluation criteria integrate predicted travel time, path reliability, and the urgency of remaining time. From the pre-generated backup plan library, all backup plan segments containing the original next node and compatible with the vehicle's current status and the new route are automatically retrieved; the matching degree between the segment and the new route is quantitatively evaluated, and based on the time window coincidence, route reliability and route similarity, the backup segment with the highest matching degree is selected and seamlessly logically spliced ​​with the new route to generate a revised complete transportation plan and push the update.

[0013] Preferably, the step of instantly marking the risk of moving segments with road travel time below a threshold or insufficient reliability includes: discretizing the road segments and time axis of vehicle travel into reservable spatiotemporal resource blocks, and initializing their capacity and global status table; after generating a cargo splicing scheme for the cargo and determining the specific path, parsing the resource requirements of the path, mapping it into a spatiotemporal resource block sequence and generating a structured reservation request; upon receiving the reservation request, immediately querying the global resource status table to check the resource block capacity, and performing real-time monitoring and flexible adjustment of the reservation execution process.

[0014] One or more technical solutions provided in this application have at least the following technical effects or advantages: By constructing a dynamic cargo source mosaic and generating a set of interference-resistant optimized solutions, the system can respond in real time to sudden interference events such as road closures and vehicle breakdowns during transportation, automatically switching to the optimal backup solution, effectively ensuring the continuity and completion rate of transportation tasks. Simultaneously, through a closed-loop feedback learning mechanism, the risk model and decision parameters are continuously calibrated, enabling the recommended strategy to have self-evolutionary capabilities, thereby significantly improving the overall resilience, resource utilization, and operational efficiency of freight scheduling in complex environments.

[0015] By deeply mining the spatiotemporal patterns of historical data and constructing multi-scale prediction models, this method can proactively identify and mitigate foreseeable transportation risks such as seasonality and cyclicality, achieving a shift from passive response to proactive planning. The method generates multiple forward-looking strategy solutions covering short-term, medium-term, and long-term scenarios, and rigorously evaluates their resilience through simulations. This significantly improves the long-term stability, strategic adaptability, and overall risk resistance of cargo sourcing recommendations in complex and volatile environments, effectively ensuring the continuous and efficient completion of transportation tasks.

[0016] By constructing a spatiotemporal path reliability map covering several hours into the future, and deeply integrating road travel time and risk prediction into cargo splicing decisions, high-delay-risk road sections can be proactively avoided during the plan generation phase. Furthermore, the optimal reliable path is assigned to each segment of the journey based on dynamic road conditions. During transportation execution, real-time monitoring of path status changes triggers local replanning, and intelligent splicing with backup plan segments enables dynamic and seamless adjustments to the transportation plan. This significantly improves the on-time completion rate, overall resilience, and adaptability to complex road conditions.

[0017] By introducing spatiotemporal resource blocks and a collaborative reservation mechanism, this method achieves refined management of road capacity in the spatiotemporal dimensions, enabling global optimization of multi-vehicle collaboration. When generating freight recommendation schemes, this method not only plans reliable routes for individual vehicles but also proactively coordinates the occupation of key spatiotemporal resources by multiple vehicles through the reservation mechanism, fundamentally avoiding hidden congestion and delays caused by route conflicts. During transportation execution, the system monitors and dynamically adjusts the reservation status in real time. When vehicles deviate from the plan or resource status changes abruptly, it can quickly trigger resource renegotiation and reallocation. This significantly improves the overall efficiency, resource utilization, and robustness of multi-vehicle collaborative transportation in complex road networks. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a method for recommending cargo to drivers according to an embodiment of the present invention. Detailed Implementation

[0019] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, the purpose of providing embodiments is to enable a more thorough and complete understanding of the disclosure of the present invention.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terminology used herein includes any and all combinations of one or more of the associated listed items.

[0021] Example 1: Figure 1 This is a flowchart illustrating a method for recommending cargo to drivers according to an embodiment of the present invention.

[0022] like Figure 1 As shown, a method for recommending cargo to drivers includes the following steps: S1 collects real-time data streams of cargo sources, transportation capacity, and environment. The data streams are then filtered and standardized through events, and output in a unified format with timestamps and cargo source credibility scores. When the cargo source credibility score is higher than the threshold, it is considered as valid cargo source information.

[0023] Among them, the source of goods refers to the set of scheduling demand data associated with the goods that need to be transported. Its core attributes include: identifier (unique source ID), goods attributes (type, weight, volume, special requirements (such as temperature control, dangerous goods)), location (precise geographical coordinates of loading and unloading locations), time window (the allowed time window for loading and unloading respectively), and priority / urgency (a level set by the consignor / sender or rules (such as high, medium, low) to determine the scheduling weight).

[0024] Transport capacity refers to the data set of vehicles available for transporting goods and their current status. Its core attributes include: identifier (unique vehicle ID, which can be obtained by license plate number), static attributes (vehicle type, rated load, volume, and commonly transported goods type), dynamic status (real-time GPS location, speed, direction, vehicle status (such as empty, fully loaded, or faulty)), and task context (currently accepted cargo information, planned route, and progress).

[0025] The environment refers to the set of external conditions data that affect transportation efficiency and safety. Its core attributes include: traffic conditions (real-time average vehicle speed of major roads, congestion index, and information on road closures due to accidents / construction), weather conditions (current and forecast rainfall, snow, fog, wind speed, etc., and the warnings they trigger (such as orange rainstorm warnings)) and infrastructure conditions (bridge weight limits, toll station status, availability of charging stations / gas stations, etc.).

[0026] The formula for calculating the credibility score of the source of goods is:

[0027] It is the credibility score of a single product source link (the threshold is limited according to the specific application scenario, and the threshold is different for different product sources and different distances), with a value range of [0,1]. It is the data source authority coefficient, which is mapped to a static value based on the preset authority level, with a value range of [0,1]. It is the freshness decay function of the goods, and it is a monotonically decreasing function of the time difference Δt between the time the goods are shipped and the current time. The range of scores for the rationality of the source of goods is [0,1]. These are the non-negative weight coefficients for the corresponding dimensions, used to adjust the degree of influence of different evaluation dimensions on the final score, and the sum of all weight coefficients is greater than zero.

[0028] S2 constructs and updates a dynamic cargo source mosaic based on the processed data stream, generates nodes for each cargo source, calculates the spatiotemporal connectivity between nodes to establish feasible edges, and evaluates the interference risk coefficient for each edge.

[0029] Specifically, the standardized data stream from step S1 is monitored in real time, and the node generation process is immediately triggered when it contains valid cargo information. A graph node is created for each valid cargo, and the attribute structure of this node includes loading and unloading time windows, latitude and longitude coordinates of loading and unloading locations, cargo type, and priority.

[0030] After node generation, edge calculation is performed to establish connections between nodes, and pairwise analysis is conducted on all existing nodes. For any pair of predecessor and successor nodes to be evaluated, the shortest path distance on the Earth's surface between the unloading point of the predecessor node and the loading point of the successor node is first calculated. This calculation is based on the great circle distance formula. If the obtained distance value is less than or equal to a preset geographical proximity threshold, the spatial connection condition is satisfied. The difference between the end time of the unloading time window of the predecessor node and the start time of the loading time window of the successor node is calculated. If this time difference is greater than or equal to zero and less than or equal to a preset time continuity threshold, the time connection condition is satisfied. Only when a pair of nodes simultaneously satisfies the above spatial and temporal conditions can a directed edge be established in the direction from the predecessor node to the successor node, indicating that the transportation connection sequence has basic feasibility.

[0031] For each established feasible edge, a quantified interference risk coefficient is assessed. This coefficient is calculated by combining historical statistical data with real-time environmental information. Specifically, the historical interruption frequency of the path associated with the edge is used as a benchmark to confirm the historical risk value. Real-time environmental indicators such as the current weather warning level and traffic congestion index on the path are obtained and quantified into real-time risk values ​​using predefined mapping rules. The historical and real-time risk values ​​are then weighted and fused using a preset weight to generate a final risk coefficient between zero and one, which is assigned to the edge. The reference formula is as follows: ,in To integrate weights, historical risks The risk can be calculated based on the historical interruption frequency of the associated path; real-time risk. It is obtained by mapping environmental data such as the current weather warning level and real-time traffic congestion index.

[0032] It's important to note that the entire process is dynamic and continuous. As new goods are continuously input into the data stream, and old goods information is updated or invalidated, incremental updates are performed accordingly. When a new goods arrive, nodes and edges are created; when existing goods information changes, the corresponding node attributes are updated and related edges are recalculated; when a goods become invalid, the corresponding node and all its associated edges are deleted. Each time an edge is added, deleted, or its associated real-time environmental information changes, its risk coefficient is reassessed, ensuring that the entire goods mosaic always reflects the latest status and risk level.

[0033] S3. Based on the dynamic cargo stitching map, a set of cargo stitching schemes is generated through an optimization algorithm. The cargo stitching scheme group includes a main recommended scheme and multiple backup schemes that are pre-constructed after analyzing the vulnerabilities in the main recommended scheme. Anti-interference simulation evaluation is performed on all schemes in the group.

[0034] Specifically, the algorithm for generating the main solution based on the current mosaic graph starts from the vehicle's current position or a specified starting point, treating all nodes in the current mosaic graph as a candidate set. The core iterative logic of the algorithm is that at each decision point, from all feasible outgoing edges pointing to subsequent nodes of the current node, the next connecting node is selected based on a multi-objective evaluation function. This evaluation function simultaneously optimizes vehicle utilization and priority satisfaction rate, and its specific form is a weighted summation:

[0035] in, This represents the empty driving distance required to travel from the current node i to the candidate node j (the mileage that a vehicle must travel to its loading location after completing the previous task, without generating direct transportation revenue). The numerical priority of candidate node j (the higher the value, the higher the priority). and These are the normalized weight coefficients for empty driving distance and priority, respectively. The algorithm continues to select nodes and expand paths until there are no feasible successor nodes that satisfy the spatiotemporal constraints, ultimately forming an ordered sequence of nodes, i.e., the main recommendation scheme.

[0036] Vulnerability analysis is performed on the generated master scheme to identify its high-risk components. Vulnerabilities mainly refer to the connecting edges in the scheme sequence that have a high interference risk coefficient. For transportation segments or nodes with extremely strict time window requirements, the Dijkstra algorithm is applied to generate local or global alternative paths for each identified vulnerability. A time-flexibility strategy allows for bounded floating of time windows for specific non-critical nodes, such as extending the original time window... Relaxed to ,in A floating threshold is set, and then feasible connecting nodes are searched again in the stitched graph based on the relaxed constraints to generate variant solutions. The spatial redundancy strategy targets node pairs connecting high-risk edges, within their geographical proximity (distance)... Find alternative nodes with similar functions, and use these nodes as hubs to recalculate the connections between them to form a detour plan. This process is carried out independently or in combination for different vulnerability points, thereby pre-generating multiple backup plans.

[0037] Monte Carlo simulations are performed on all generated primary and backup schemes. In each simulation, various types of disturbances are randomly generated, such as randomly canceling a node in the scheme or randomly interrupting a path represented by an edge in the scheme. The task completion rate of a scheme under disturbances is defined as the ratio of the actual successfully completed quantity of goods to the original planned quantity, and the cost change rate is defined as the ratio of the total cost after disturbance (e.g., distance, time) to the original cost. Each scheme undergoes N independent simulations, and its stability score is quantified by calculating the expected ability of its task completion rate and cost change rate to remain stable under disturbances.

[0038] S4 monitors the real-time processed data stream to identify interference events and automatically selects the most suitable solution from the backup solution library based on the event type and the current transportation status of the transport vehicles.

[0039] Specifically, by continuously monitoring the standardized real-time data stream from step S1, interference events affecting the execution of the transportation plan are automatically identified and classified. Specifically, an event listener is set up to parse the content of the data stream in real time. When the data content meets a preset abnormal pattern, it is determined to be an interference event and subsequent processing is triggered. Typical interference event patterns include: changes in cargo status to canceled or significant changes in time and location; vehicles reporting GPS locations that are stuck on the expected path for an extended period or reporting a malfunction; receiving information from traffic information sources that key nodes on the current planned route are closed or severely congested; and receiving high-level weather warnings for the areas traversed. Once an interference event is identified, a status assessment of the current transportation context is immediately initiated. This assessment quickly extracts the following key information: the vehicle's latest location, speed, and orientation at the time of the interference; the list of currently loaded cargo and their respective unloading locations and time windows; the remaining uncompleted subsequent nodes and their attributes in the currently executing main transportation plan sequence; and the vehicle's remaining available load capacity and volume.

[0040] Based on the identified types of interference events and the assessment results of the current transportation context, an automated solution matching process is initiated. From the backup solution library generated in step S3, all logically feasible candidate solutions are selected. Selection criteria include: the starting point of the candidate solution must be compatible with the vehicle's current location or status; the solution must be able to accommodate the vehicle's already loaded cargo that still needs to be delivered; and the resource requirements of the solution must not exceed the vehicle's remaining capacity. Each feasible candidate solution is scored for multi-dimensional matching. The score comprehensively considers the additional cost of implementing the solution, the estimated completion time, the degree of protection for high-priority cargo, and its own stability score. Through weighted calculation, the solution with the highest overall matching degree is selected as the final execution solution.

[0041] Once the plan is selected, specific execution instructions and instructions are immediately generated. These instructions include: the new destination sequence, navigation route, estimated arrival time, and necessary operational change instructions. This information is pushed via the vehicle terminal or driver's app, clearly explaining the reason for the change (e.g., due to road closure ahead, you have been switched to an alternative route; estimated arrival time at the next loading point is delayed by 25 minutes). The final adopted plan and the reason for the switch are updated in the global transportation plan log, replacing the original plan sequence. According to rules, status change notifications are automatically sent to affected shippers or the dispatch center.

[0042] S5 collects data on the results of the scheme implementation and dynamically adjusts and updates the risk coefficient in the mosaic diagram through feedback learning.

[0043] Adjusting the risk coefficient in the stitched image involves calibrating the interference risk coefficient model for the stitched image. By comparing the deviation between the predicted interference risk coefficient and the actual frequency of interruption events on each path, optimization methods such as gradient descent are used to dynamically adjust the fusion weights of historical and real-time risks. or update real-time risks The mapping rules make the predicted risks more realistic.

[0044] The technical solutions described in the embodiments of this application above have at least the following technical effects or advantages: By constructing a dynamic cargo source mosaic and generating a set of interference-resistant optimized solutions, the system can respond in real time to sudden interference events such as road closures and vehicle breakdowns during transportation, automatically switching to the optimal backup solution, effectively ensuring the continuity and completion rate of transportation tasks. Simultaneously, through a closed-loop feedback learning mechanism, the risk model and decision parameters are continuously calibrated, enabling the recommended strategy to have self-evolutionary capabilities, thereby significantly improving the overall resilience, resource utilization, and operational efficiency of freight scheduling in complex environments.

[0045] Example 2: In Example 1, the method uses real-time data to drive the construction of a cargo source mosaic map and generate anti-interference solutions, enabling a short-term response to sudden changes and ensuring transportation continuity. However, this method's decision-making relies heavily on current and recent real-time status information. When facing transportation environments with strong periodicity, trends, or complex correlations, its strategy generation lacks in-depth guidance from historical patterns and proactive prediction of future trends. Because it fails to systematically mine and utilize the spatiotemporal patterns inherent in historical data (such as seasonal route bottlenecks or regional economic freight volume growth trends), the generated solutions are essentially local optimizations for the current optimal situation. They cannot automatically identify and avoid foreseeable long-term risks, nor can they conduct strategic capacity deployment and route planning spanning several days to weeks. This over-reliance on real-time data results in insufficient effectiveness and overall resilience of the method's backup plans when facing significant disturbances or structural changes exceeding the short-term forecast window.

[0046] In some embodiments, anti-interference simulation evaluation is performed on all schemes, and step S3 further includes: S31 extracts spatiotemporal pattern factors of interference events by accessing historical data streams, and constructs an enhanced predictive mosaic that integrates historical patterns and short-term trends in real time.

[0047] Specifically, parameters such as real-time vehicle coordinates, current date and time, weather codes, and calendar markers are combined to form a structured query request. This request is then used to index historical data storage, quickly extracting complete historical records from the past few years, including cargo details, vehicle trajectories, weather events, and road conditions, for a single season, the same week, and similar geographical locations and weather patterns. This results in a highly correlated dataset for analysis.

[0048] The dataset is fed into a pre-built time-series analysis service for real-time scanning and calculation. This service calculates the probability and average duration of delays on various road segments during the same historical period, considering the current time and surrounding routes; compares recent traffic flow with the historical average for the same period to calculate the percentage of relative growth or decline in the current area; and analyzes historical event chains to identify the probability that other related locations will be affected in the next few hours when an event occurs at a certain location. The calculation results are then packaged into a set of spatiotemporal pattern factors with confidence levels and validity periods.

[0049] A basic source map reflecting the instantaneous state is obtained, and the graph traversal program dynamically injects spatiotemporal pattern factors. The risk weights of corresponding edges in the graph are adjusted according to the factor type; for example, the real-time risk value of an edge is mixed with the historical delay probability. The priority values ​​of nodes in specific areas are fine-tuned to reflect trends. Virtual connection edges labeled with the probability of influence are temporarily added between nodes where chain reactions are predicted. After all attribute updates are completed, an enhanced predictive map is output, with a real-time network structure, but the connection weights have incorporated historical pattern recognition.

[0050] S32 generates forward-looking cargo stitching schemes that cover multiple time windows and have different strategy orientations based on enhanced predictive stitching maps.

[0051] Specifically, the enhanced predictive mosaic map generated by S31 is loaded, and the real-time status parameters of the target vehicle are acquired simultaneously, including its precise geographical location, current load and volume, details of accepted but uncompleted transportation tasks, and the vehicle's fixed attributes such as vehicle type and frequently transported cargo categories. Next, a multi-layered time frame is set for this planning, from near to far. For example, a planning period spanning seven days is defined and divided into three consecutive execution windows: the upcoming 24 hours, the following 48 hours (i.e., days 2 to 3), and the subsequent four days (i.e., days 4 to 7).

[0052] For the first 24-hour window, starting from the vehicle's actual location, a multi-branch parallel search process is initiated in the enhanced prediction mosaic map. This search simultaneously explores three strategy paths: The first path rigorously selects nodes and connecting edges with the lowest historical risk coefficients and the highest predictive stability in the map for extension, constructing a robust solution; the second path prioritizes nodes and regions injected with positive trend predictive factors (e.g., areas predicting freight volume growth or efficiency improvement corridors), constructing a trend-adaptive solution; the third path proactively associates a similar alternative node with relatively relaxed time requirements at key logistics hubs along the baseline or positive path, using this alternative node as a flexible anchor point for that segment, constructing a flexible buffer solution.

[0053] After completing the route planning for the first 24-hour window, the expected vehicle status (including location, time, and load) corresponding to the planned route endpoint is used as a virtual starting point and substituted into the second time window (days 2 and 3). Based on the prediction weights pre-calculated for this future period in the enhanced prediction stitching map, the same multi-strategy parallel search and evaluation process is repeated to generate continuous paths for the second window. This iterative process is repeated continuously, using the output of the previous window as the input for the next window, until the entire preset seven-day planning cycle is covered. Ultimately, a complete multi-segment transportation plan covering the medium and long term is generated for each strategy, consisting of multiple short-term window paths seamlessly connected.

[0054] All complete plans representing different strategy orientations generated through the above process are compiled and output. The planning process automatically generates a summary description for each plan, which clearly lists the core strategy followed by the plan, the specific dates and time windows it covers, a list of essential key nodes arranged in sequence, and an overall risk aversion effectiveness score calculated based on the injected predictive factors.

[0055] S33 combines historical scenario backtesting with multi-perturbation Monte Carlo simulation to conduct a dual resilience assessment of the primary recommended scheme and the set of backup schemes.

[0056] Specifically, at the start of the evaluation process, based on the planned execution time period of the scheme to be evaluated, characteristic parameters such as its start date, seasonal information, weekly cycle, major geographical areas traversed, and weather patterns are extracted. Historical databases are retrieved for all historical periods over the past three to five years that highly match these characteristics. For each matching historical scenario, the candidate scheme is placed in the complete data environment of that historical moment for virtual simulation execution. This simulation follows the scheme's preset node sequence and time plan, while also being subject to the impact of all events actually recorded within that historical period. Key results such as the proportion of cargo successfully completed by the scheme in each historical scenario, and the deviation of the overall transportation time from the original plan are recorded. After all matching historical scenario simulations are completed, the scheme's performance under all scenarios is analyzed, calculating the fluctuation range of its task completion rate and average delay level, thereby quantifying the scheme's stability performance in the face of historical periodic fluctuations.

[0057] Parallel multi-perturbation Monte Carlo simulations are initiated. This process begins by pre-setting a pool of disturbance events, defining the probability and impact of various negative events, such as the probability of a specific cargo being canceled or the probability distribution of delays on certain roads. For each scenario to be evaluated, thousands of independent simulation copies are created. In each simulation, a unique sequence of various disturbance events is randomly generated based on pre-set probabilities and applied to the dynamic execution of the scenario. When the scenario encounters random disturbances in the simulation environment, it responds according to its own logic (such as switching to an alternative route). After each simulation, the actual performance and resource consumption of the scenario under the random disturbance are recorded. After all simulations are completed, statistical analysis is performed on all results to calculate the average probability of successful task completion and the average extent of cost exceeding the budget when facing various random disturbances. These two indicators together define the robustness of the scenario.

[0058] Finally, a comprehensive evaluation and selection process is conducted. For each scheme, the stability index obtained from historical backtesting and the robustness index obtained from Monte Carlo simulation are standardized and weighted to generate a unified comprehensive resilience score. This score integrates the scheme's adaptability to known historical patterns and its resistance to unknown random risks. Based on this score, all candidate schemes are ranked, and an evaluation report is generated, clearly indicating the specific values ​​of each scheme's quantitative stability index, comprehensive score, and ranking.

[0059] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By deeply mining the spatiotemporal patterns of historical data and constructing multi-scale prediction models, this method can proactively identify and mitigate foreseeable transportation risks such as seasonality and cyclicality, achieving a shift from passive response to proactive planning. The method generates multiple forward-looking strategy solutions covering short-term, medium-term, and long-term scenarios, and rigorously evaluates their resilience through simulations. This significantly improves the long-term stability, strategic adaptability, and overall risk resistance of cargo sourcing recommendations in complex and volatile environments, effectively ensuring the continuous and efficient completion of transportation tasks.

[0060] Example 3: In Example 3, by constructing a spatiotemporal path reliability map, a dynamic and reliable forward-looking driving path is planned for a single vehicle. This allows for real-time monitoring and replanning of path risks, significantly improving the resilience of single-vehicle task execution. However, this method focuses on the internal optimization of a single vehicle. When multiple vehicles simultaneously or sequentially execute their independently planned optimal reliable paths, implicit conflicts may arise due to competition for the capacity of the same road segment at similar times. This causes the predicted path reliability to fail under the superposition of actual road network traffic flow, leading to new congestion or delays. In other words, individual optimality does not necessarily lead to global optimality. To achieve efficient and equitable utilization of road network spatiotemporal resources in complex multi-vehicle operation scenarios and fundamentally avoid multi-vehicle path conflicts, a mechanism needs to be introduced to coordinate the occupation of critical road resources by different vehicles.

[0061] In some embodiments, step S33, which involves a dual resilience assessment of the primary recommended solution and the set of alternative solutions, further includes: S321, based on the enhanced prediction stitching map, constructs a spatiotemporal path reliability map covering the next few hours on the current driving paths of vehicles involved in the main recommended scheme and the backup scheme, and calculates the road travel time.

[0062] Specifically, based on the generated primary and backup plans, a list of all node access sequences within each plan is extracted, with the starting point of the node list being the vehicle's real-time location. Combining the vehicle's current speed, load status, and the loading / unloading time requirements for each node, the earliest and latest estimated arrival times for each node in the plan are calculated. All transport segments connecting nodes, such as the route from the current location to the first loading point and subsequent routes between nodes, are uniformly defined as the core path set requiring precise evaluation.

[0063] For each specific road segment within the actual road network covered by the core path set, a parallel computing process is initiated. For each road segment, historical average travel time data for the same period, current real-time vehicle speed at the minute level, detailed weather forecast information for the next few hours, and known road construction or traffic control plans are collected. This multi-source data is input into a pre-trained time series prediction model, which automatically analyzes patterns and outputs the estimated travel time for that road segment at multiple consecutive time points over the next few hours, while also providing the possible error range for the estimated value at each time point.

[0064] For any possible route between any two consecutive nodes in the plan, based on the vehicle's planned departure time from the starting point, the estimated travel time for each road segment along the route at the vehicle's expected entry time is sequentially queried, and these times are accumulated to obtain the total estimated travel time for the route. The estimated error ranges of all road segments on the route are aggregated, and a comprehensive path reliability index between 0 and 1 is calculated using a specific mathematical aggregation algorithm. This index directly reflects the degree of certainty in completing the journey within the specified time.

[0065] Based on the above calculations, a structured mapping data table is generated for all node pairs and their candidate routes involved in the schemes, namely, the local spatiotemporal path reliability map required for this decision. This map records the total estimated time and comprehensive reliability index of each candidate route. According to a preset strategy (such as most reliable priority, i.e., selecting the path with the highest reliability index, or shortest time priority), an optimal path is selected from the map for the movement between each node segment. The total time of this selected path will be assigned the standard movement time attribute as the directed edge connecting the two nodes in the enhanced prediction stitched graph. The reliability index of this path will be converted into a risk correction value, specifically: Risk Correction Value = (1 - Reliability Index) The adjustment coefficient is used to dynamically adjust the comprehensive risk weight coefficient of the corresponding edge in the enhanced prediction splicing graph, thereby quantifying the traffic uncertainty of actual roads in the risk assessment of the cargo network.

[0066] S322, when enhancing the stitching scheme of the prediction stitching map, dynamically binds the moving segments between each node to the spatiotemporal path reliability map, and immediately marks the risk of moving segments with road travel time below the threshold or insufficient reliability.

[0067] Specifically, a reliable path is matched and bound for each mobile segment. For each mobile segment from the preceding node to the subsequent node in the scheme, based on the constructed spatiotemporal path reliability map, all feasible physical paths connecting the two nodes are found and obtained, along with their estimated travel time and reliability index. According to the preset strategy objective, a current optimal path is selected from the candidate paths. The estimated total travel time and path reliability index of this selected path are assigned as the estimated transportation time and travel reliability of the current mobile segment, respectively.

[0068] Assess the timeliness and reliability risks of the mobile segment. Calculate the available time budget for the mobile segment, i.e., the time window from the completion of the preceding node task to the necessity of starting the subsequent node task. Multiply this time budget by a preset buffer coefficient to obtain the maximum allowable travel time threshold for the mobile segment. Compare the estimated transport time bound in the previous step with this threshold. If the estimated time exceeds the threshold, the mobile segment is determined to have a high risk of timeout. Set a minimum reliability standard and compare the bound path reliability index with this standard. If the reliability index is lower than the standard, the mobile segment is determined to have a high risk of travel uncertainty.

[0069] High-risk movement segments are identified and their attributes are updated in real time. For any movement segment determined to be high-risk, a mark is made on the corresponding directed edge in the enhanced prediction stitching graph. This mark records the specific type and severity level of the risk. The risk level is calculated by weighting the timeout ratio ((estimated time - threshold) / threshold) and the degree to which reliability falls below the standard ((standard value - reliability index) / standard value). This risk level will exist as an independent attribute alongside the original risk coefficient.

[0070] S323 monitors the progress and route risk status of vehicles during execution. When a change in the reliability of the bound route is detected, it actively triggers local route replanning and splices the nodes in the main recommended route that need route replanning with the local routes in the backup route.

[0071] Specifically, during the execution of the recommended route, the vehicle's real-time location and speed are continuously tracked, and the latest predicted travel time and reliability index of the current and next road segments are dynamically queried based on the current travel time. The estimated arrival time is estimated in real time based on the distance between the vehicle's current location and the target node, as well as its current speed.

[0072] Two core risk indicators are continuously assessed to determine whether intervention is necessary. First, the latest reliability index of the current or next route segment is checked to see if it falls below a dynamic threshold. This threshold is defined as: Dynamic Threshold = Base Threshold + Urgency Coefficient × (1 - Remaining Time Budget Proportion), where the remaining time budget proportion = (Maximum Allowable Travel Time Threshold for the Route - Time Already Used) / Maximum Travel Time Threshold. Therefore, as the remaining time decreases, this threshold automatically increases, and the reliability requirements become increasingly stringent. Second, the real-time estimated arrival time of the vehicle is compared with the originally predicted arrival time based on the bound route to determine if the deviation exceeds a preset allowable range. When either of these conditions is triggered, it is immediately determined that local route replanning needs to be initiated for the current or affected next travel segment.

[0073] Once replanning is triggered, the new route planning starts at the vehicle's latest location (if it has left the starting point) or the original planned starting point, and ends at the originally planned next node. A new route search is then performed in the latest spatiotemporal path reliability map. This search aims to find a new, comprehensively optimal route based on the latest traffic forecasts. Its evaluation criteria consider not only predicted travel time but also path reliability and the urgency of remaining time due to current delays, forming a new composite optimization objective.

[0074] Once a new candidate route is determined, the system automatically retrieves all available alternative route segments from a pre-generated backup plan library. These segments must contain the originally planned next node and their pre-defined local transport segments must be compatible with the vehicle's current cargo status and the new route in both time and space. The matching degree between the backup segments and the newly planned route is quantitatively evaluated, taking into account factors such as the time window fit, the reliability of the new route, and the similarity between the new route and the pre-defined routes of the backup segments. The backup transport segment with the highest matching degree is selected, and the newly planned route is seamlessly and logically combined with this segment to generate an adjusted transport plan that extends from the vehicle's current location to multiple subsequent nodes. This revised plan, the new navigation route, and the updated timeline are then pushed to the driver, and the global plan record in the backend is updated simultaneously.

[0075] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By constructing a spatiotemporal path reliability map covering several hours into the future, and deeply integrating road travel time and risk prediction into cargo splicing decisions, high-delay-risk road sections can be proactively avoided during the plan generation phase. Furthermore, the optimal reliable path is assigned to each segment of the journey based on dynamic road conditions. During transportation execution, real-time monitoring of path status changes triggers local replanning, and intelligent splicing with backup plan segments enables dynamic and seamless adjustments to the transportation plan. This significantly improves the on-time completion rate, overall resilience, and adaptability to complex road conditions.

[0076] Example 4: In Example 3, by fusing real-time path reliability maps, a forward-looking plan was developed for each vehicle, and reliable paths were dynamically bound, significantly improving the resilience and adaptability of single-vehicle transportation to road conditions. However, this method is essentially still an independent optimization for a single vehicle. When multiple drivers simultaneously or sequentially choose the same or overlapping optimal reliable paths, the system cannot anticipate and coordinate this potential resource competition. This can lead to multiple vehicles competing for the same road segment resources in the same time and space, potentially causing new local congestion. Consequently, the reliability of the predicted path is attenuated in actual execution due to traffic flow overlap, thus weakening the overall effectiveness of the method.

[0077] In some embodiments, step S322, which involves immediately marking the movement segments with road travel times below a threshold or insufficient reliability, further includes: 2A discretizes the road segments and timelines for vehicle traffic into reservable spatiotemporal resource blocks and initializes their capacity and global state table.

[0078] Specifically, core road networks requiring refined resource management are selected, such as highways and major arterial roads. Based on the physical separation of lanes and traffic direction, each independently passable road is divided into a series of continuous road segment units along its centerline at fixed lengths. Each road segment unit is assigned a unique geographical identifier, and its static attributes such as the number of lanes, design speed, gradient, and key facilities along the route are recorded.

[0079] Starting from the current moment, the timeline of a future planning period is divided into continuous and non-overlapping time slices of fixed duration. Each time slice is assigned a globally unique temporal index to identify its absolute start and end time window. Combining a road segment unit with a time slice defines a spatiotemporal resource block. Based on the physical attributes of the road segment unit and the traffic flow theory model, the maximum number of vehicles that can safely pass through the unit within the corresponding time slice duration is calculated as the theoretical capacity of the resource block. Then, combining road grade and historical operational data, its initial available reservation capacity is calculated using a reduction factor.

[0080] Create a global resource status table with the spatiotemporal resource block identifier as the key. Initialize the theoretical capacity, initial available reservation capacity, current number of reservations, and reservation queue for each resource block entry in the table. Simultaneously, import published long-term road construction or permanent traffic restriction information, directly marking the status of the corresponding resource blocks during the affected period as unavailable and setting their available capacity to zero.

[0081] 2B: After generating a cargo splicing scheme and determining the specific path for the cargo source, the resource requirements of the path are analyzed, mapped into a spatiotemporal resource block sequence, and a structured reservation request is generated.

[0082] Specifically, the process involves obtaining the established freight transportation plan and its complete directed routes. Based on a pre-established geospatial index of road segment units, this continuous geometric path is mapped into a sequence of discrete road segment units connected end-to-end. For each road segment unit in this sequence, the planned entry and exit times of vehicles in each unit are recursively calculated based on the planned departure time of the plan, vehicle performance parameters, and the estimated traffic speed of each road segment, thereby determining its planned occupancy time window.

[0083] The planned occupancy time window for each road segment unit is compared with the globally discretized time slice sequence. Based on preset occupancy rules, specific time slice sets that are covered or partially covered by this time window and require reservation are identified. A unique spatiotemporal resource block identifier is generated for each road segment unit and time slice combination requiring reservation. Through this process, the entire physical path is transformed into an ordered sequence of spatiotemporal resource block demands. Finally, a structured reservation request is generated based on this sequence. This request includes a unique identifier for the vehicle or task making the reservation, an associated freight order identifier, the spatiotemporal resource block demand sequence, and a priority score calculated from multiple dimensions, including the urgency of the freight and the credit rating of the entity.

[0084] 2C: Upon receiving a reservation request, immediately query the global resource status table to check the resource block capacity, and perform real-time monitoring and elastic adjustment of the reservation execution process.

[0085] Specifically, upon receiving a structured reservation request, the system immediately queries the global resource status table to verify the current remaining capacity of each spatiotemporal resource block in the request sequence. If the remaining capacity of all resource blocks meets the current requirement, the reservation is approved, the corresponding resource is locked in the status table (i.e., its current reserved quantity is increased, and the reservation information is added to its reservation queue).

[0086] If any resource block is found to be undercapacitated, a conflict resolution process is triggered. The process attempts the following strategies in sequence: First, route fine-tuning is performed, searching for an alternative path with sufficient resources and minimal incremental travel costs among the current request's candidate paths. If this fails, priority negotiation is initiated, comparing the priority scores of the current request with those of existing reservations in the target resource block's reservation queue. If the current request has a higher priority, attempts are made to replan the path for lower-priority reservations in the queue to free up resources; otherwise, a new path is planned for the current request. If this still fails, a time offset is suggested for the current request, adjusting the planned departure time to avoid resource peaks. Upon successful conflict resolution, the reservation is approved and the resource status is updated.

[0087] After a reservation takes effect, execution is monitored by comparing the vehicle's real-time trajectory with the spatiotemporal range of the reserved resource block. If the vehicle's progress deviates significantly from the plan, such as arriving at the corresponding road segment of the resource block too early or too late, all subsequent unused resource blocks for that vehicle are automatically released, and other tasks that have reserved adjacent spatiotemporal resource blocks are notified. If a resource block becomes unavailable due to a sudden road event, an emergency replanning is initiated for all tasks occupying that resource block, allocating alternative resources and adjusting their subsequent reservations accordingly. All changes to resource status are updated to the global status table in real time, and the released capacity can be immediately used to respond to new requests or optimize existing reservations.

[0088] The technical solutions described in the embodiments of this application above have at least the following technical effects or advantages: By introducing spatiotemporal resource blocks and a collaborative reservation mechanism, this method achieves refined management of road capacity in the spatiotemporal dimensions, enabling global optimization of multi-vehicle collaboration. When generating freight recommendation schemes, this method not only plans reliable routes for individual vehicles but also proactively coordinates the occupation of key spatiotemporal resources by multiple vehicles through the reservation mechanism, fundamentally avoiding hidden congestion and delays caused by route conflicts. During transportation execution, the system monitors and dynamically adjusts the reservation status in real time. When vehicles deviate from the plan or resource status changes abruptly, it can quickly trigger resource renegotiation and reallocation. This significantly improves the overall efficiency, resource utilization, and robustness of multi-vehicle collaborative transportation in complex road networks.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. 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.

Claims

1. A method for recommending cargo sources to drivers, characterized in that, include: S1 collects real-time data streams of cargo sources, transportation capacity, and environment. The data streams are then filtered and standardized through events, and output in a unified format with timestamps and cargo source credibility scores. When the cargo source credibility score is higher than the threshold, it is considered as valid cargo source information. S2, construct and update a dynamic cargo source mosaic based on the processed data stream, generate nodes for each cargo source, calculate the spatiotemporal connectivity between nodes to establish feasible edges, and evaluate the interference risk coefficient for each edge; S3. Based on the dynamic cargo splicing map, a set of cargo splicing schemes is generated through an optimization algorithm. The cargo splicing scheme group includes a main recommended scheme and multiple backup schemes that are pre-constructed after analyzing the vulnerabilities in the main recommended scheme. Anti-interference simulation evaluation is performed on all schemes in the group. S4 monitors the real-time processed data stream to identify interference events and automatically selects the most suitable solution from the backup solution library based on the event type and the current transportation status of the transport vehicles. S5 collects data on the results of the scheme implementation and dynamically adjusts and updates the risk coefficient in the mosaic diagram through feedback learning.

2. The method for recommending cargo to drivers as described in claim 1, characterized in that, The process of constructing and updating a dynamic cargo source mosaic based on the processed data stream specifically includes: creating a corresponding graph node for each valid cargo source; calculating the spatial proximity and temporal continuity between any two nodes based on their spatiotemporal attributes; establishing a directed edge representing the feasibility of transportation connection between a set of nodes if and only if a set of nodes simultaneously satisfies both the spatial proximity and temporal continuity conditions; evaluating a quantified interference risk coefficient for each directed edge by comprehensively considering the historical interruption frequency of the associated path and the current environmental risk information; continuously monitoring changes in the data stream and incrementally and dynamically updating the cargo source mosaic, including creating, updating, or deleting nodes based on newly added, changed, or invalid cargo source information, as well as recalculating associated edges and reassessing risk coefficients.

3. The method for recommending cargo to drivers as described in claim 1, characterized in that, The process of analyzing vulnerabilities in the primary recommended path and pre-constructing multiple backup paths includes: starting from the vehicle's current position based on a mosaic map, selecting subsequent nodes iteratively according to a multi-objective evaluation function to generate a primary recommended path; performing vulnerability analysis on the primary recommended path to identify links in its sequence with high interference risk or strict time window constraints as vulnerabilities; for each identified vulnerability, applying time elasticity and spatial redundancy strategies to generate local or global alternative paths to pre-construct multiple backup paths; and performing Monte Carlo simulations on all generated paths to quantitatively evaluate the task completion rate and cost change rate of each path by simulating random interference events, and calculating the stability score of the path accordingly.

4. The method for recommending cargo to drivers as described in claim 1, characterized in that, The automatic selection of the most suitable solution from the backup solution library specifically includes: real-time monitoring of standardized data streams to identify interference events and classifying them; immediately initiating an assessment of the current transport vehicle status upon identification of an interference event to obtain its location, load, and task progress information; based on the type of interference event and the assessment results of the current transport vehicle status, selecting all logically feasible candidate solutions from the backup solution library; performing multi-dimensional matching degree scoring on each feasible candidate solution, and selecting the solution with the highest comprehensive matching degree as the final execution solution based on the scoring results.

5. The method for recommending cargo to drivers as described in claim 1, characterized in that, Step S3, which involves performing anti-interference simulation evaluation on all schemes within the group, further includes: S31, by accessing historical data streams to extract spatiotemporal pattern factors of interference events, and constructing an enhanced predictive mosaic that integrates historical patterns and short-term trends in real time; S32, based on the enhanced predictive stitching map, generates a forward-looking cargo stitching scheme that covers multiple time windows and has different strategy orientations; S33 combines historical scenario backtesting with multi-perturbation Monte Carlo simulation to conduct a dual resilience assessment of the primary recommended scheme and the set of backup schemes.

6. The method for recommending cargo to drivers as described in claim 5, characterized in that, The stitching scheme includes: loading an enhanced prediction stitched image and obtaining the real-time status of the vehicle; setting a planning period containing multiple consecutive time windows; starting from the vehicle's current position, within the first time window, performing path searches using multiple strategies in parallel based on the enhanced prediction stitched image to generate initial path segments corresponding to different strategies; using the endpoint status of the planned path in the previous time window as the virtual starting point of the next time window, iteratively performing multi-strategy path searches to generate a complete strategy scheme covering the planning period; and outputting scheme summary information for each generated complete strategy scheme.

7. The method for recommending cargo to drivers as described in claim 6, characterized in that, The enhanced predictive mosaic includes: combining real-time vehicle coordinates, current date and time, weather codes, and calendar marker parameters to form a structured query request; based on this request, extracting complete historical records of the same period and region from historical data storage to form a highly correlated dataset; sending the dataset to a time-series analysis service for real-time scanning and calculation, statistically analyzing the probability and average duration of delays on each road segment during the same period in history, comparing recent traffic flow with the historical average to calculate the regional trend percentage, and analyzing historical event chains to identify the probability of influence of related locations, thereby generating a spatiotemporal pattern factor with confidence; obtaining a basic cargo source mosaic, dynamically injecting the spatiotemporal pattern factor through a graph traversal program, adjusting the risk weight of corresponding edges in the graph and the priority of fine-tuning points according to the factor type, and adding virtual connection edges between nodes where chain reactions are predicted, outputting the enhanced predictive mosaic.

8. The method for recommending cargo to drivers as described in claim 5, characterized in that, The dual resilience assessment of the primary recommended solution and the set of backup solutions includes: S321, Based on the enhanced prediction stitching map, a spatiotemporal path reliability map covering the next few hours is constructed on the current driving path of the vehicle involved in the main recommended plan and the backup plan, and the road travel time is calculated. S322, when enhancing the stitching scheme of the prediction stitching map, dynamically binds the moving segments between each node to the spatiotemporal path reliability map, and immediately marks the risk of moving segments with road travel time below the threshold or insufficient reliability. S323 monitors the progress and route risk status of vehicles during execution. When a change in the reliability of the bound route is detected, it actively triggers local route replanning and splices the nodes in the main recommended route that need route replanning with the local routes in the backup route.

9. The method for recommending cargo to drivers as described in claim 8, characterized in that, The step of splicing the nodes in the main recommended plan that require route replanning with the local paths in the backup plan includes: after triggering local path replanning, planning a comprehensive optimal new path from the current planning starting point to the originally planned next node based on the latest path prediction information; automatically retrieving all backup plan segments from the pre-generated backup plan library that contain the originally planned next node and whose preset local transportation segments are spatiotemporally compatible with the current vehicle loading status and the new path; quantitatively evaluating the matching degree of these backup segments, with evaluation factors including time window fit, reliability of the new path, and similarity to the preset route of the backup segments; selecting the backup transportation segment with the highest matching degree, and seamlessly logically splicing the newly planned path with this segment to generate a revised transportation plan that extends from the vehicle's current position to multiple subsequent nodes.

10. The method for recommending cargo to drivers as described in claim 9, characterized in that, The method of instantly marking the risk of moving segments with road passage time below the threshold or insufficient reliability includes: discretizing the road segments and time axis of vehicle passage into reservable spatiotemporal resource blocks, and initializing their capacity and global state table; after generating a cargo splicing scheme for the cargo and determining the specific path, parsing the resource requirements of the path, mapping it into a spatiotemporal resource block sequence and generating a structured reservation request; upon receiving the reservation request, immediately querying the global resource state table to check the resource block capacity, and performing real-time monitoring and flexible adjustment of the reservation execution process.

Images