Logistics fleet intelligent scheduling method and system based on Beidou high-precision positioning

Abstract

The embodiment of the invention provides a logistics fleet intelligent scheduling method based on Beidou high-precision positioning. The logistics fleet intelligent scheduling method comprises the steps of obtaining continuous high-precision track point data of logistics vehicles; recognizing a congested road section and a potential passing time delay area based on the continuous high-precision track point clustering analysis result; predicting the traffic efficiency of each road section in different time periods based on a historical traffic time delay model, and constructing a dynamic traffic capacity map; and according to the dynamic traffic capacity map and the vehicle real-time position information, carrying out dynamic redistribution regulation and control on a distribution time period window. According to the scheme of the embodiment of the invention, dynamic redistribution regulation and control can be carried out on the distribution period window so as to solve the problem of low time fulfillment rate of the end node.

Description

Technical Field

[0001] This application relates to the field of logistics technology, specifically to a method and system for intelligent dispatching of logistics fleets based on BeiDou high-precision positioning. Background Technology

[0002] In the logistics and distribution sector, intelligent scheduling methods are crucial for improving the time-fulfillment rate of last-mile delivery nodes. However, existing technologies typically rely on static route planning or simple trajectory monitoring based on ordinary GPS positioning, lacking deep clustering analysis of continuous high-precision vehicle trajectory data. This results in the inability to identify congested road sections and potential traffic delay areas in real time and accurately. At the same time, because historical traffic delay models fail to dynamically integrate with real-time traffic conditions, the accuracy of predicting the traffic efficiency of each road segment is insufficient. Consequently, delivery time windows are often fixed and rigid, unable to be dynamically reallocated and adjusted according to actual road conditions. This leads to vehicle scheduling delays, route optimization failures, and ultimately, low time-fulfillment rates at last-mile delivery nodes. Summary of the Invention

[0003] In view of this, the present disclosure provides an intelligent scheduling method for logistics fleets based on BeiDou high-precision positioning, which at least partially solves the problems existing in the prior art.

[0004] This application discloses an intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning, comprising: Acquire continuous, high-precision trajectory point data of logistics vehicles; Based on the results of the continuous high-precision trajectory point clustering analysis, congested road sections and potential traffic delay areas are identified. Based on historical traffic delay models, the traffic efficiency of each road segment at different time periods is predicted and a dynamic traffic capacity map is constructed. The delivery time window is dynamically reallocated and controlled based on the dynamic traffic capacity map and real-time vehicle location information.

[0005] In one specific implementation, the step of identifying congested road sections and potential travel delay areas based on the continuous high-precision trajectory point clustering analysis results further includes: Obtain the GPS coordinate sequence of logistics vehicles and extract the timestamp and latitude / longitude triplet; The timestamps are divided according to a specific time granularity to form a trajectory segment dataset; The trajectory points of each segment are clustered based on the K-means algorithm to obtain multiple dense point groups; If the number of cluster points exceeds the set threshold K0 and the average moving speed is lower than V0, it is judged as a congested road segment.

[0006] In one specific implementation, the step of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Collect traffic records for a specific road section at different times of the day; Calculate the average travel time Ta within each time period; The current traffic efficiency of this road segment is calculated using the following formula: E = 1 / (Ta × α + β), where α is the time variation weight and β is the basic delay compensation coefficient. Prioritize nodes based on traffic efficiency and update graph node information.

[0007] In one specific implementation, the dynamic reallocation and control of the delivery time window based on the dynamic capacity map and real-time vehicle location information further includes: Collect the current location coordinates and remaining distance D to the destination of the logistics vehicle; Compare the vehicle's location with available paths in the dynamic map; The following formula is used to determine whether a route needs to be replanned: if D < δ × Ct, then the route is reassigned; where δ is the safety distance coefficient and Ct is the current road segment's capacity. If rescheduling is required, update the scheduled delivery time window for that vehicle in the task dispatch system.

[0008] In one specific implementation, the dynamic reallocation and control of the delivery time window based on the dynamic capacity map and real-time vehicle location information further includes: The vehicle's latitude and longitude coordinates are obtained in real time through the Beidou receiver, and the current vehicle speed V is calculated. Compare the real-time load status S of the road network node where the vehicle is currently located; Use the following formula to determine whether the current road conditions require an adjustment to the delivery rhythm: If S > σ × Ta, then adjust the delivery window time ΔT = ΔT_max × (1 - S / Ta); where σ is the warning threshold and Ta is the average travel time during that period; If the adjustment conditions are met, the delivery scheduling optimization module will be triggered.

[0009] In one specific implementation, the step of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Collect a dataset M of hourly traffic records for a specific road during a particular season; The data is processed by weighted moving average to smooth out outlier data points, and the hourly average latency Th is calculated. A geospatial factor Fg is introduced to characterize the impact of different road topologies on traffic efficiency, and a dynamic map model G=Th×Fg is constructed. The model-generated data is fused with real-time traffic data to form a traffic capacity map with time-varying characteristics.

[0010] In one specific implementation, the step of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Extract multidimensional variables from historical big data: weather conditions W (0=sunny / 1=rainy), peak period H (0=off-peak / 1=peak), and emergencies E (0=none / 1=some). A linear regression model P=ω1×W+ω2×H+ω3×E+ω0 is established to estimate the average delay probability P during this period. Priority markers in dynamic maps are constructed using P-values ​​combined with real-time GPS data. If P > threshold, mark the route as a risky path and avoid selecting it during task assignment.

[0011] In one specific implementation, the step of identifying congested road sections and potential travel delay areas based on the continuous high-precision trajectory point clustering analysis results further includes: After preprocessing the trajectory data, key event points are extracted, and a road segment dwell time matrix R is generated. The Markov chain method was used to evaluate the dwell behavior transfer patterns of different vehicles on this road segment; Use the following formula to determine if there is a potential passage delay: Pz=(∑(R_i≥tm)) / N, where tm is the tolerance time and N is the number of samples; If Pz > T2, then the road segment is determined to have potential travel delay.

[0012] In one specific implementation, the dynamic reallocation and control of the delivery time window based on the dynamic capacity map and real-time vehicle location information further includes: Extract the real-time traffic density Df and average travel time Ta of the target road segment from the dynamic map; The comprehensive impact value of the road segment is calculated using the following formula: I = λ × Df + (1 - λ) × Ta, where λ is the traffic priority factor. If I exceeds the set threshold It, the emergency dispatch procedure will be initiated to dynamically reconstruct the time window. The adjusted delivery time window must meet the transportation timeliness SLA.

[0013] In one specific implementation, the step of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Collect daily traffic data for the same road segment over the past three months, normalize the data on a daily basis, and store it in the database; The normalized time series is used to predict the expected traffic efficiency E_forecast for the next day. Construct a weight function w(t)=exp((tt_start)^2 / (2σ²)) to represent the effect of time deviation on the weight; The predicted values ​​are multiplied by the weights and then stored and updated as weight parameters in the dynamic graph.

[0014] In one specific implementation, the dynamic reallocation and control of the delivery time window based on the dynamic capacity map and real-time vehicle location information further includes: Obtain the real-time load index η_i of all key nodes on the current path from the graph; The total load score L_total is calculated based on η of all nodes, where L_total = ∑(η_i × w_i), and w_i is the node weight. Set a maximum load limit η_max for a path. If L_total > η_max, then start searching for alternative paths. If a suitable route is found, the time window is reassigned; otherwise, the original plan is maintained and the terminal station is notified through the notification system.

[0015] This disclosure provides an intelligent scheduling method for logistics fleets based on BeiDou high-precision positioning, including: acquiring continuous high-precision trajectory point data of logistics vehicles; identifying congested road sections and potential traffic delay areas based on the clustering analysis results of the continuous high-precision trajectory points; predicting the traffic efficiency of each road section at different time periods based on a historical traffic delay model and constructing a dynamic traffic capacity map; and dynamically reallocating and regulating the delivery time window according to the dynamic traffic capacity map and the real-time location information of the vehicles. The solution provided by this disclosure can address the problem of low time fulfillment rates at the last mile by dynamically reallocating and regulating the delivery time window based on the clustering analysis results of continuous high-precision trajectory points and a historical traffic delay model. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the exemplary embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this disclosure and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning of the present invention; Figure 2 This is a step in one embodiment of the present invention to dynamically reallocate and regulate the delivery time window based on the dynamic traffic capacity map and the real-time vehicle location information; Figure 3 This is a block diagram of the intelligent dispatching system for logistics fleets based on BeiDou high-precision positioning, which is based on the present invention. Detailed Implementation

[0018] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0019] The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning of the present invention will be described in detail below with reference to the accompanying drawings. Those skilled in the art will understand that the following description is intended to clearly illustrate the technical solution of the present invention, and not to limit the scope of protection of the present invention.

[0020] The core of this invention lies in its ability to intelligently and dynamically reallocate logistics delivery time windows by integrating high-precision real-time BeiDou trajectory data, dynamic congestion identification based on trajectory clustering, and traffic delay prediction combining historical and real-time data, thereby systematically improving the on-time delivery rate at the last mile. Figure 1 As shown, the method specifically includes: Step S101: Obtain continuous high-precision trajectory point data of logistics vehicles.

[0021] Specifically, a BeiDou high-precision positioning terminal (such as an intelligent vehicle terminal supporting BeiDou-3 system B1C and B2a dual-frequency signals) deployed on each logistics vehicle continuously collects the vehicle's location information at a frequency of once per second (i.e., a sampling rate of 1Hz). Each trajectory point data packet contains at least the following fields: a unique vehicle identifier (VehicleID), a UTC timestamp accurate to milliseconds, latitude and longitude coordinates based on the Earth-fixed coordinate system (longitude and latitude, with accuracy down to the centimeter level), instantaneous speed, and heading angle. This encrypted data is transmitted in real-time or near real-time (latency less than 2 seconds) to the data receiving and parsing server of the cloud-based intelligent dispatch platform via the vehicle's built-in 4G / 5G communication module. After verifying, cleaning (such as removing obvious drift points), and formatting the raw data, the server stores it in a spatiotemporal trajectory database in chronological order, forming a continuous high-precision trajectory point sequence for each vehicle's journey.

[0022] Step S102: Identify congested road sections and potential traffic delay areas based on the continuous high-precision trajectory point clustering analysis results.

[0023] The cloud-based dispatch platform's data processing engine retrieves trajectory points from the trajectory database for all vehicles within a specified area (e.g., a city's delivery range) within the most recent time window (e.g., the past 15 minutes). First, the trajectory points are segmented by vehicle and time, for example, using a 1-minute time granularity, with each vehicle's 1-minute trajectory point set as an analysis unit. Then, an improved DBSCAN density clustering algorithm or the K-means algorithm is applied to spatially cluster the trajectory points of all vehicles within the same geographical area during the same time period. During clustering, the algorithm uses geographical location (latitude and longitude) as the primary feature and speed as an auxiliary weight. The clustering result generates multiple point clusters.

[0024] The system can preset judgment rules: if the number of trajectory points contained in a point cluster exceeds the threshold K0 (e.g., 50 points), and the average moving speed of all trajectory points in the point cluster is lower than the threshold V0 (e.g., 10 km / h), then the continuous road segment covered by the point cluster is judged as a "real-time congested road segment". Simultaneously, it can analyze the evolution trend of point clusters. For example, if the density of point clusters on a certain road segment continuously increases in a short period of time but the speed has not yet dropped below the threshold, then the road segment is marked as a "potential traffic delay area". The system updates the GIS information (such as road segment ID, start-end nodes, severity level) of the identified congested road segments and potential delay areas to the dynamic traffic layer in real time.

[0025] Step S103: Based on the historical traffic delay model, predict the traffic efficiency of each road segment in different time periods and construct a dynamic traffic capacity map.

[0026] Specifically, a historical traffic delay database can be maintained, storing historical traffic records for each road segment over a long period (e.g., the past 6 months) at different time intervals (e.g., 15-minute intervals), including average travel time, median time, and speed standard deviation. Based on the real-time identification in step S102, the model layer performs the following operations: For a target road segment, firstly, based on the current date type (weekday, weekend, holiday) and time point (e.g., Wednesday morning 10:15), historical traffic delay data under similar conditions is matched from the historical database as a baseline. Then, a multivariate correction factor is introduced. For example, by accessing the meteorological API to obtain the current weather conditions (W), combined with the time period attribute (H) and the sudden events (E) inferred from social media or traffic management department data, a real-time delay probability (P) or delay correction coefficient (α, β) is calculated using a pre-trained linear regression model. Finally, the traffic efficiency E is predicted, for example, using the following method: E = baseline efficiency / (1 + P) or a similar function. Based on the city's road network map, the system dynamically assigns the predicted traffic efficiency (E), real-time congestion status, and comprehensive delay risk level of each road segment as attributes to each edge (road segment) and node (intersection) in the map, thereby constructing a "dynamic traffic capacity map" that changes over time (from tens of minutes to several hours) and in real-time conditions. This map not only reflects the current situation but also predicts the time-varying trend of the road network's traffic capacity.

[0027] Step S104: Dynamically reallocate and regulate the delivery time window based on the dynamic traffic capacity map and real-time vehicle location information.

[0028] Specifically, the scheduling engine maintains a task queue for each vehicle performing a delivery mission, which includes a pre-defined sequence of delivery nodes and corresponding committed time windows (e.g., "Customer A: 14:00-15:00"). Figure 2 As shown, the engine executes the following loop every second (or as needed): Status awareness: Obtain the latest high-precision position, speed, and heading of vehicles, and accurately map them onto specific road segments in the dynamic traffic capacity map.

[0029] Future path simulation: Based on the vehicle's current task queue, combined with the current traffic efficiency attributes and prediction attributes of road segments in the map, the time-dependent shortest path algorithm is used to recalculate the estimated arrival time (ETA) of the vehicle to each subsequent delivery node.

[0030] Performance Risk Assessment: The recalculated ETA is compared with the task commitment time window. If the ETA is later than the end of the window, or even if it is still within the window but the remaining margin (safe time) is less than the threshold δ, the delivery task is determined to be high-risk.

[0031] Dynamic reallocation decision-making and execution: Once a high risk is identified, the system immediately initiates dynamic reallocation. This is not merely simple route replanning, but rather the reallocation of delivery time windows, specifically including: Inter-vehicle task rebalancing: The system searches the global fleet for other vehicles that are currently ahead of the high-risk vehicle and have a lighter load or a better route, and attempts to transfer the high-risk delivery task (or part of the cargo in the task) to the vehicle in real time, and updates the task time windows of the two vehicles.

[0032] Time window negotiation and update: If the task cannot be transferred, the system can automatically calculate a new and more feasible estimated delivery time window based on the SLA rules (such as adjusting 14:00-15:00 to 14:30-15:30), and push the change notification simultaneously through the driver's APP and customer service system to manage customer expectations in advance.

[0033] Micro-path and rhythm adjustment: Based on the real-time load of the specific intersection where the vehicle is located, dynamically suggest that the vehicle adjust its waiting or passage strategy, and even fine-tune the order of adjacent delivery tasks to optimize the overall rhythm.

[0034] All redistribution decisions are instantly fed back to the navigation terminals of all relevant vehicles in the fleet and the monitoring screen of the dispatch center, forming a closed-loop intelligent control system of "perception-prediction-decision-execution-re-perception".

[0035] Traditional scheduling methods rely on static route planning before departure, which cannot cope with dynamic changes along the way. This invention creatively constructs a core data model called "dynamic capacity map" and continuously compares real-time vehicle status with it, enabling scheduling decisions to be based on predictions of future road network conditions, achieving truly dynamic and intelligent control throughout the entire process.

[0036] In a specific embodiment, the identification of congested road sections and potential travel delay areas based on the continuous high-precision trajectory point clustering analysis results specifically includes: Obtain the GPS coordinate sequence of logistics vehicles and extract the timestamp and latitude / longitude triplet; The timestamps are divided into segments with a 1-minute granularity to form a trajectory segment dataset; The trajectory points of each segment are clustered based on the K-means algorithm to obtain multiple dense point groups; If the number of cluster points exceeds the set threshold K0 and the average moving speed is lower than V0, it is judged as a congested road segment.

[0037] Specifically, first, extract the original trajectory data reported by all logistics vehicles in the target area within the most recent analysis period (e.g., the past 10 minutes) from the spatio-temporal trajectory database. Each piece of data is a quadruple containing a timestamp (accurate to milliseconds), longitude, latitude, and instantaneous speed. The data preprocessing module divides the trajectory points of all vehicles into time windows with a fixed time granularity of 1 minute, forming individual trajectory segment data sets (for example, grouping all trajectory points between 10:00:00 and 10:01:00 into one data set).

[0038] Subsequently, the spatial analysis engine independently performs clustering analysis on each data set within a 1-minute time window. The classic K-means clustering algorithm is used, with longitude and latitude as the feature vectors. Before the algorithm execution, the number of clusters K is preset through the elbow method or based on historical experience. After clustering, each cluster is labeled as a "cluster of dense trajectory points". The system sets two key judgment thresholds: the quantity threshold K0 (for example, set to 30) and the speed threshold V0 (for example, set to 15 km / h). The system traverses all clusters, calculates the number of points N in each cluster and the average moving speed V_avg of all trajectory points within it. The judgment logic is: if a certain cluster simultaneously meets the conditions N≥K0 and V_avg≤V0, the geographical area covered by this cluster (usually mapped to a continuous section of road) is identified by the system as the "current congested section". At the same time, the system marks the cluster area where V_avg is close to but slightly higher than V0 (such as V0<V_avg≤1.2×V0) and the points are dense as a "potential traffic delay area" and gives early warning attention. The recognition results will be associated with the specific road number or section ID and attached with a congestion intensity label (classified according to the proportion of points exceeding the threshold), and pushed to the downstream dynamic map construction module.

[0039] By introducing time window division and clustering analysis, the massive and scattered high-precision trajectory points are transformed into "congestion event" objects with clear spatial and state semantics, realizing the effective extraction of traffic state knowledge from the original data. Compared with the traditional simple threshold alarm based on the speed of a single vehicle or fixed detectors, this method uses the spatio-temporal aggregation characteristics of group vehicles for comprehensive judgment, significantly improving the accuracy and anti-interference ability of congestion recognition (for example, it can effectively filter out misjudgments caused by the temporary parking of individual vehicles). At the same time, by setting the double-threshold (K0 and V0) conditions, the "congestion" state can be more accurately defined, avoiding the over-identification of slight slowdowns, and providing reliable and high-confidence real-time road condition inputs for subsequent scheduling decisions.

[0040] In a specific embodiment, predicting the traffic efficiency of each section at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map includes: Collect the traffic records of a certain section at different times of each day; Calculate the average travel time Ta within each time period; The current traffic efficiency of this road segment is calculated using the following formula: E = 1 / (Ta × α + β), where α is the time variation weight and β is the basic delay compensation coefficient. Prioritize nodes based on traffic efficiency and update graph node information.

[0041] Specifically, firstly, the historical database module collects a sample of travel time records for this road segment over the past 90 days, 24 hours a day, at 15-minute intervals (calculated by back-calculating from a large number of historical vehicle trajectories). For the time period "Wednesday 10:00-10:15 AM", all historical Wednesday travel times during this time period are statistically analyzed, and their average value is calculated as the historical baseline travel time Ta (e.g., 500 seconds).

[0042] Then, the model calculation module applies the preset traffic efficiency calculation formula E=1 / (Ta×α+β). Here, α is the time-varying weighting coefficient (e.g., α=1.05), used to fine-tune historical benchmark values ​​to reflect long-term trend changes; β is the basic delay compensation coefficient (e.g., β=30), used to smooth out minor fluctuations in the calculation and ensure the numerical stability of the formula. After substituting the values, the theoretical traffic efficiency E of the road segment in the current time period is calculated. To incorporate real-time performance, the system dynamically discounts the E value based on whether the road segment is currently congested or has potential delays (e.g., if congestion is identified, the E value is multiplied by a penalty factor of 0.6).

[0043] Finally, the map construction module uses the city's digital road network map as a base map and creates a data structure for each road segment node. This structure not only contains static road level and length attributes, but also dynamically stores the currently calculated traffic efficiency (E value), historical traffic capacity (Ta value), and real-time status labels. These dynamic attributes of all road segments together constitute a "dynamic traffic capacity map" covering the entire region. The map is visualized in the dispatch center using color depth (e.g., dark red represents extremely low efficiency, green represents high efficiency) and numerical heatmaps, and serves as a queryable digital map for the route planning algorithm.

[0044] In this invention, the dynamic reallocation and control of the delivery time window based on the dynamic capacity map and real-time vehicle location information includes: Collect the current location coordinates and remaining distance D to the destination of the logistics vehicle; Compare the vehicle's location with available paths in the dynamic map; The following formula is used to determine whether a route needs to be replanned: if D < δ × Ct, then the route is reassigned; where δ is the safety distance coefficient and Ct is the current road segment's capacity. If rescheduling is required, update the scheduled delivery time window for that vehicle in the task dispatch system.

[0045] Specifically, the system acquires the vehicle's BeiDou high-precision positioning coordinates every second and calculates the remaining distance D (in meters) from the current location to the destination along the currently planned path based on the next destination in its delivery task list (such as "Customer C"). Simultaneously, the system queries a dynamic traffic capacity map to obtain the real-time passable capacity Ct of the road segment the vehicle is currently on and the key road segment it is about to enter (this capacity is calculated by comprehensively considering the road segment's traffic efficiency E, the number of lanes, and real-time traffic flow; it is a dynamic value reflecting the road segment's throughput capacity at the current moment).

[0046] The system incorporates a safety-based distance coefficient δ (e.g., δ=1.5). The reallocation trigger logic executes as follows: Calculate the ratio of D to δ×Ct. If the condition D<δ×Ct is met, it means that the vehicle's remaining travel distance before reaching its destination is less than a "safe buffer distance" dynamically determined by the road network capacity ahead. This indicates that, based on the current path and predicted road conditions ahead, the vehicle's safety margin for on-time delivery is insufficient, posing a high risk of delay.

[0047] Once the conditions are triggered, the system immediately initiates the "dynamic reallocation" procedure. In this invention, it's not simply about replanning a shortest path, but rather a task-level reorganization with the core objective of "ensuring delivery within the time window." The scheduling algorithm searches for a better solution in the global fleet and road network map, and may take the following measures: 1) planning an alternative route for the current vehicle that, while potentially slightly longer, offers higher efficiency and reliability; or 2) finding another vehicle in the fleet with a better location and sufficient load capacity, and dynamically transferring "Customer C's" delivery task along with the time window to it. After the decision is made, the system automatically updates the task timelines of the vehicle and / or relevant replacement vehicles in the central task dispatch system, and simultaneously distributes the new estimated arrival time and adjusted delivery time window through the vehicle terminal and customer notification system.

[0048] This invention creatively proposes a triggering mechanism based on the comparison between the remaining distance D and the dynamic safety buffer distance δ×Ct, so that the redistribution decision is no longer a simple judgment based on a fixed distance or fixed time, but is closely linked to the real-time traffic capacity of the road network ahead, realizing personalized and precise risk warning.

[0049] In this invention, the step of dynamically reallocating and regulating the delivery time window based on the dynamic traffic capacity map and real-time vehicle location information further includes: The vehicle's latitude and longitude coordinates are obtained in real time through the Beidou receiver, and the current vehicle speed V is calculated. Specifically, the vehicle-mounted BeiDou receiver collects and reports the vehicle's precise latitude and longitude coordinates in real time at a frequency of once per second. The speed calculation module in the dispatch platform calculates the current vehicle speed V=ΔS / ΔtV (unit: km / h) in real time based on the spherical distance difference ΔS and time difference Δt between two consecutive reporting points.

[0050] Compare the real-time load status S of the road network node where the vehicle is currently located; Specifically, the platform maps vehicle coordinates onto a dynamic traffic capacity map in real time, accurately locating the current road network node (such as a specific intersection or road segment connection point). The system reads the real-time load status value S from the node's attributes. This value is a normalized comprehensive index (range 0-1), calculated by comparing the number of vehicles currently passing through the node, the average speed, and historical traffic flow during the same period. A higher S value indicates a more congested node.

[0051] Use the following formula to determine whether the current road conditions require an adjustment to the delivery rhythm: If S > σ × Ta, then adjust the delivery window time ΔT = ΔT_max × (1S / Ta); where σ is the warning threshold and Ta is the average travel time during that period; Specifically, the system performs delivery rhythm adjustment judgments. It first queries the historical average travel time (Tavg) of the current vehicle's location on the road segment during the current time period (e.g., Friday evening rush hour) from the historical travel delay model. The system presets a warning threshold σ (e.g., σ=1.3). The adjustment judgment logic is as follows: if the real-time load state S satisfies the condition S>σ×Tavg, it is considered that the current node's load is significantly worse than the historical normal, which will soon have a substantial impact on the vehicle's travel rhythm, and adjustment must be triggered.

[0052] The adjustment amount is calculated using the formula ΔT = ΔTmax × (1 − S / Tavg). Here, ΔTmax is the maximum allowable adjustment window for the delivery task (e.g., 30 minutes). This formula means that the higher the deterioration of the real-time load S relative to the historical average Tavg, the larger the adjustment window ΔT needs to be.

[0053] If the adjustment conditions are met, the delivery scheduling optimization module will be triggered.

[0054] If the above conditions are met, the system first assesses the current task queue: it may instruct vehicles to appropriately extend their operating dwell time after arriving at the next delivery point (e.g., from the standard 5 minutes to 8 minutes) to avoid more congested sections of road later; or, under the premise of safety and compliance, it may suggest to the driver via the vehicle terminal to fine-tune the following distance and speed on the current road segment to pass through congested areas more smoothly. All these micro-rhythm adjustments are intended to partially absorb the time delay caused by high load at nodes during transportation, thereby protecting the integrity of the final delivery time window.

[0055] This solution combines the abstract node load state S in the dynamic graph with the specific vehicle's speed, historical average travel time Tavg, and maximum adjustable range ΔTmax. Before the full impact of congestion becomes apparent, it proactively fine-tunes the vehicle's driving and operation rhythm through a precisely calculated time window adjustment amount ΔT. This significantly improves the precision and foresight of dispatch instructions, enabling a smoother absorption of the impact of sudden local congestion, avoiding frequent and significant route changes, and improving both driving experience and fuel economy while ensuring contract fulfillment.

[0056] The following describes the steps of the present invention for predicting the traffic efficiency of each road segment at different time periods based on a historical traffic delay model and constructing a dynamic traffic capacity map. Specifically, this includes: Collect a dataset M of hourly traffic records for a specific road during a particular season; The data is processed by weighted moving average to smooth out outlier data points, and the hourly average latency Th is calculated. A geospatial factor Fg is introduced to characterize the impact of different road topologies on traffic efficiency, and a dynamic map model G=Th×Fg is constructed. The model-generated data is fused with real-time traffic data to form a traffic capacity map with time-varying characteristics.

[0057] First, the data acquisition engine extracts hourly traffic records for all weekdays from the historical database of the target road over the past two specific seasons (e.g., spring and autumn), forming a dataset MM. Each record contains: hour identifier (e.g., 7:00-8:00), actual travel time, and date stamp.

[0058] Next, data smoothing is performed. Since the original data may contain outliers caused by accidents or extreme weather, the system uses a weighted moving average method for smoothing. For the travel time series {ti} in the i-th hour, its smoothed value is... From the formula The calculation yields (weights are configurable). Based on the smoothed sequence, the hourly average historical latency Th for this road is calculated for each hourly interval.

[0059] Then, the geospatial factor Fgeo is introduced. This factor is used to quantify the global impact of inherent road topological features on traffic efficiency. Fgeo is calculated by analyzing digitized road alignment data, and its value comprehensively considers: 1) curvature (mean radius of curvature); 2) node density (number of intersections or ramps per unit length); and 3) frequency of lane number changes. For example, a straight expressway with few intersections and a stable number of lanes may have an Fgeo of 0.9 (indicating a positive impact on traffic); while a side road with many curves, multiple intersections, and frequent lane merging may have an Fgeo of 1.2 (indicating a negative amplification effect on traffic).

[0060] Finally, a dynamic traffic capacity map model is constructed. The system multiplies the hourly predicted travel time of the road segment by the geospatial factor, i.e., G = Th × Fg. The resulting G value is the weighted predicted travel time of the road segment in the map. This model generates a predicted value for each road segment at different times of the day, integrating historical temporal patterns and static spatial attributes. This predicted value is then fused with real-time traffic event data (such as accidents and construction) to ultimately generate a time-varying capacity map that profoundly reflects the spatiotemporal heterogeneity of the road network.

[0061] This solution introduces the geospatial factor Fgeo, enabling the constructed dynamic map to possess physical interpretability and scenario universality. Traditional models rely solely on historical time statistics, failing to distinguish whether traffic flow patterns are caused by high traffic volume or road complexity within the same historical timeframe. This solution, through the Fgeo factor, enhances the model's ability to generalize and infer new traffic flow patterns, which is crucial for planning routes in new areas or assessing the impact of sudden traffic surges on specific roads.

[0062] In one specific embodiment, predicting the traffic efficiency of each road segment at different time periods based on a historical traffic delay model and constructing a dynamic traffic capacity map further includes: Extract multidimensional variables from historical big data: weather conditions (W), peak hours (H), and emergencies (E); A linear regression model P=ω1×W+ω2×H+ω3×E+ω0 is established to estimate the average delay probability P during this period. Priority markers in dynamic maps are constructed using P-values ​​combined with real-time GPS data. If P > threshold, mark the route as a risky path and avoid selecting it during task assignment.

[0063] First, the feature engineering module extracts three key categorical variables affecting travel delay from the historical big data warehouse and quantifies them: Weather condition W: Sunny = 0, Rain / Snow / Fog = 1.

[0064] Peak hours H: Off-peak hours = 0, morning / evening peak hours = 1.

[0065] Emergency E: None = 0, Existing (Accident, Construction, Control) = 1.

[0066] At the same time, extract the sample labels (1 = delay occurred, 0 = no delay occurred) for road sections that actually experienced significant delays (such as travel time exceeding 20% ​​of the historical baseline) within the corresponding historical time period.

[0067] Next, a multiple linear regression model is trained using a large amount of historical sample data (e.g., all records for this road segment over the past year). The model has the form P = ω1 × W + ω2 × H + ω3 × E + ω0. Here, ω0 is the intercept term, and ω1, ω2, and ω3 are the regression coefficients of the three variables: weather, peak hours, and events, respectively, obtained through least squares fitting. The P-value output by the model is an estimated probability index of delay. For example, a sample taken on a sunny, off-peak day with no events might output P = 0.1; while a sample taken on a rainy, peak day with accidents might output P = 1.8.

[0068] Then, this model is applied to the real-time updating of dynamic maps. The system connects to the meteorological API, timers (to determine peak hours), and traffic event API in real time to obtain the W, H, and E values ​​corresponding to the current target road segment. These values ​​are then substituted into the regression model to calculate the delay probability index P at the current moment.

[0069] Finally, the map risk labeling: The system presets a risk threshold (e.g., 0.7). If P > threshold, the dynamic map will mark the road segment as a "high-risk path" in its graphical and data attributes, and display it on the scheduling interface with a prominent visual style (such as flashing or dark red). When assigning tasks or planning routes, the path search algorithm will impose a very high penalty weight on such marked road segments, actively avoiding or prioritizing detours, unless there is no other choice.

[0070] This solution transforms seemingly vague external risk factors into a unified and calculable delay risk index P, thereby achieving proactive risk-avoidance scheduling.

[0071] In another embodiment, identifying congested road sections and potential travel delay areas based on the continuous high-precision trajectory point clustering analysis results further includes: After preprocessing the trajectory data, key event points are extracted, and a road segment dwell time matrix R is generated. The Markov chain method was used to evaluate the dwell behavior transfer patterns of different vehicles on this road segment; Use the following formula to determine if there is a potential passage delay: Pz=(∑(R_i≥tm)) / N, where tm is the tolerance time and N is the number of samples; If Pz > T2, then the road segment is determined to have potential travel delay.

[0072] First, the continuous high-precision trajectory point data acquired in step S101 is preprocessed. Preprocessing includes coordinate correction (mapping the original coordinates to the actual road centerline), data smoothing (removing positioning jitter), and key event point extraction. The system defines a "key event point" as a trajectory point where the vehicle speed is below a threshold (e.g., 3 km / h) and persists for more than a preset time (e.g., 30 seconds). This point records non-driving behaviors caused by the vehicle waiting at intersections, loading / unloading, or abnormal congestion. The system marks the location (segment ID) and dwell time of all key event points on the trajectory of each vehicle.

[0073] Based on the above data, the system constructs a "road segment-dwell time matrix" R. The rows of the matrix represent different logistics vehicles or trips, and the columns represent various road segments in the road network. The matrix element R_{ij} represents the total dwell time of the i-th vehicle on the j-th road segment. If a vehicle does not dwell on a road segment, the corresponding element is 0.

[0074] Subsequently, the Markov chain method was used to model and analyze the patterns of vehicle dwell time transitions. Each road segment was considered a "state," and the behavior of a vehicle moving (or dwelling) from one road segment to the next was considered a "state transition." The system analyzed the historical trajectory data of all vehicles, statistically calculating the probability of moving from road segment A to road segment B, paying particular attention to the conditional probability P(Next|S_stay) of moving from a road segment where a dwell time has occurred (state S_stay) to the next road segment (state Next). Through this model, certain "chain reaction" patterns can be identified. For example, during the morning rush hour, once a vehicle dwells at "entry ramp A," the probability of it dwelling on "downstream main road B" will significantly increase to P1 (e.g., 0.85).

[0075] To quantify the potential delay risk, the system uses the following formula to calculate the "potential delay probability" Pz for road segment j: Pz_j = (∑I(R_{ij}≥tm)) / N. Where I() is an indicator function, taking a value of 1 when the dwell time R_{ij} of the i-th vehicle on road segment j exceeds the preset tolerance time tm (e.g., 120 seconds), and 0 otherwise; N is the total number of vehicles passing through road segment j in the analysis sample. This probability reflects the frequency of "over-tolerance dwell" on this road segment.

[0076] Finally, the system sets a judgment threshold T2 (e.g., 0.4). If Pz_j of a certain road segment is greater than T2, then based on the high chain transition probability shown in its Markov chain analysis, the road segment is judged as a high-risk road segment with "potential traffic delay". Even if the current instantaneous speed does not meet the standard, an early warning should be issued in the map.

[0077] This solution combines micro-level dwell behavior analysis (matrix R) with macro-level traffic flow pattern mining (Markov chain) to achieve a deeper and more forward-looking identification of potential traffic delays. Using the Markov chain model, the system can capture causal relationships and transmission effects within traffic flow, thereby predicting the spread of congestion. This allows the scheduling system to anticipate impending delays several kilometers or road segments in advance, providing longer warning times and more comprehensive decision-making basis for dynamic reallocation.

[0078] In a specific embodiment, the dynamic reallocation and control of the delivery time window based on the dynamic capacity map and real-time vehicle location information further includes: Extract the real-time traffic density Df and average travel time Ta of the target road segment from the dynamic map; The comprehensive impact value of the road segment is calculated using the following formula: I = λ × Df + (1 - λ) × Ta, where λ is the traffic priority factor. If I exceeds the set threshold It, the emergency dispatch procedure will be initiated to dynamically reconstruct the time window. The adjusted delivery time window must meet the transportation timeliness SLA.

[0079] Specifically, when the dispatch engine detects that a vehicle is heading to a target delivery point, it first extracts real-time data of all key nodes (such as major intersections, bridge entrances, tunnel exits, etc.) on the vehicle's planned route from the dynamic capacity map. For each node i, the map provides its real-time traffic density Df_i (unit: vehicles / km, estimated by the density of surrounding trajectory points) and current average travel time Ta_i (calculated based on recent vehicle data passing through that node).

[0080] Next, the system calculates a comprehensive impact value I_path for the path. This value is not a simple summation, but rather an aggregation of indicators from all key nodes along the path. First, the comprehensive impact value I_i for each node i is calculated as I_i = λ × Df_i + (1-λ) × Ta_i. Here, λ is a preset traffic priority factor (0 < λ < 1). For example, if λ is set to 0.7, it means that when evaluating the impact of a node, real-time traffic density (reflecting the scale of congestion) accounts for 70% of the weight, and average travel time (reflecting the time cost of congestion) accounts for 30%. This design allows for adjustments based on scheduling strategy preferences (such as focusing more on easing large-scale congestion points or focusing more on shortening single-point travel time).

[0081] Then, calculate the total score for the entire route: L_total=Σ(η_i×w_i). Here, η_i is the normalized load index of node i (which can be converted from I_i), and w_i is the weight of the node in the route, which can be assigned according to the topological importance of the node in the road network (such as whether it is a hub) or the estimated impact on delivery time.

[0082] The system presets a maximum load limit η_max for each path. This limit is an empirical value representing the total degree of road condition deterioration a path can withstand to ensure on-time delivery. The emergency trigger condition is: if the calculated L_total > η_max, the overall traffic conditions of the path are deemed unacceptable, and continuing along the original path will highly likely lead to severe delays. In this case, the system immediately initiates the emergency dispatch procedure.

[0083] Specifically, the system first interrupts the original timeline for the vehicle. Then, starting from the current vehicle location and ending at the original delivery point, it urgently searches the dynamic graph for alternative routes with a load score L_total lower than η_max. During the search, the algorithm actively avoids nodes marked as high-risk or extremely overloaded. If a suitable route is found, the system quickly recalculates the estimated arrival time for the new route and dynamically reconstructs (i.e., adjusts) the time window for the delivery task accordingly, while simultaneously synchronizing the adjustment information to the driver and customer. If no feasible alternative route is found within a reasonable timeframe (e.g., all detour options are severely time-consuming), the system maintains the original plan but sends delay warnings to the terminal station and customer via a notification system to allow for advance planning.

[0084] This solution upgrades the process from single-point traffic condition assessment to overall route health evaluation, avoiding over-scheduling caused by temporary anomalies at individual nodes and improving the overall rationality of decision-making. Furthermore, by introducing a traffic priority factor λ and node weights w_i, the route evaluation model becomes highly configurable and adaptable, allowing for flexible adjustments based on different cities, vehicle types, and even delivery tasks with varying priorities.

[0085] In a specific embodiment, the step of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Collect daily traffic data for the same road segment over the past three months, normalize the data on a daily basis, and store it in the database; The normalized time series is used to predict the expected traffic efficiency E_forecast for the next day. Construct a weight function w(t)=exp((t-t_start)^2 / (2σ²)) to represent the effect of time deviation on the weight; The predicted values ​​are multiplied by the weights and then stored and updated as weight parameters in the dynamic graph.

[0086] First, the data preparation phase. The system selects the target road segment and collects complete daily traffic efficiency time-series data for that segment from the historical database for the past three months (e.g., January 1st to March 31st). This data is typically structured as a time series of travel times or speeds at fixed time intervals (e.g., 15 minutes). Then, the raw data is normalized to eliminate the scaling effects caused by differences in absolute traffic volume between different days. For example, the daily series data is divided by its daily average to convert it into a ratio series reflecting relative traffic efficiency. The processed series is stored in a dedicated time-series database.

[0087] Secondly, model training and prediction. For a specific time period (e.g., 2:00 PM - 2:15 PM) on a future day (e.g., April 1st), the system constructs a time series using normalized efficiency values ​​from 2:00 PM to 2:15 PM on the corresponding days of the week (i.e., all Tuesdays) over the past three months. This series is then modeled using an ARIMA model. The ARIMA model captures the inherent periodicity, trend, and randomness of the series by analyzing its autocorrelation (AR), difference stationarity (I), and moving average (MA). After model training, inputting the latest series data allows for the prediction of the expected relative efficiency E_forecast for that time period on the future day (April 1st). (This is a normalized prediction ratio; for example, 1.1 indicates an expected efficiency 10% higher than the average efficiency for that time period on weekdays.)

[0088] Then, a time weighting function is introduced to handle the uncertainty of the prediction. The closer the predicted value is to the current time, the higher its reliability. The system defines a Gaussian weighting function w(t) = exp(-(t-t_start)^2 / (2σ²)). Where t is the current time, t_start is the start time of the prediction period, and σ is a parameter that controls the rate of weight decay (e.g., set to 1 hour). When t coincides with t_start (i.e., at the beginning of the prediction period), the weight w(t) = 1 (maximum); as time deviates (earlier or later), the weight decays according to a Gaussian curve.

[0089] Finally, the dynamic capacity map generates and updates the data. When storing the predicted efficiency value for a road segment during the 2:00-2:15 PM time period, the dynamic capacity map does not directly use E_forecast. Instead, it multiplies it by the real-time calculated time weight to obtain the final weight parameter: E_final = E_forecast × w(t). This E_final value is stored in the map as the "base prediction value" for that road segment during that time period. As the actual time t progresses, w(t) changes continuously. Even if E_forecast remains unchanged, E_final will automatically adjust, smoothly reducing the impact of the predicted value until it is completely replaced by real-time data. The map update module periodically (e.g., every 5 minutes) recalculates the E_final values ​​for all road segments for multiple future time periods to ensure the timeliness of the prediction information.

[0090] In another embodiment, the step of dynamically reallocating and adjusting the delivery time window based on the dynamic capacity map and real-time vehicle location information further includes: Obtain the real-time load index η_i of all key nodes on the current path from the graph; The total load score L_total is calculated based on η of all nodes, where L_total = ∑(η_i × w_i), and w_i is the node weight. Set a maximum load limit η_max for a path. If L_total > η_max, then start searching for alternative paths. If a suitable route is found, the time window is reassigned; otherwise, the original plan is maintained and the terminal station is notified through the notification system.

[0091] Specifically, when a vehicle needs to determine the feasibility of its current route, the system obtains the real-time load index η_i of all N key nodes (K_1, K_2, ..., K_N) on its planned route P from the dynamic traffic capacity map. This index is a normalized value that integrates information such as current traffic flow, traffic speed, and queue length, and ranges from 0 (completely unobstructed) to 1 (completely saturated).

[0092] The system does not treat all nodes equally; instead, it assigns a node weight w_i to each node. The value of w_i is based on the node's topological importance and its impact on the current delivery task. For example, a necessary bridge or tunnel entrance with no alternative route has a higher w_i (e.g., 0.3), while a small intersection with multiple detour options has a lower w_i (e.g., 0.05). The sum of all weights is 1.

[0093] Based on this, the total load score for the path is calculated as: L_total = Σ(η_i × w_i). This score reflects the weighted congestion level of the entire path after considering the importance of nodes.

[0094] The system presets a maximum load limit η_max for each type of delivery task based on its service availability level (SLA). For example, η_max is set to 0.6 for "2-hour delivery" tasks and 0.8 for "next-day delivery" tasks.

[0095] The core decision-making logic is as follows: If the calculated L_total ≤ η_max, the path is considered acceptable, and the original plan is executed. If L_total > η_max, an alternative path search is immediately initiated. This search is global, considering not only the current vehicle but also assessing whether adjustments will affect other vehicles in the fleet. The scheduling algorithm will attempt to find another path, P_alternative, with a total load score lower than η_max and the least impact on the overall fleet scheduling.

[0096] If P_alternative is successfully found, the system will perform a "reassignment of time window". This means that a new ETA will be calculated for the current vehicle, and the order and time window of subsequent delivery points may be fine-tuned, with all changes automatically synchronized.

[0097] If a global search fails to find any alternative path that satisfies L_total≤η_max without causing unacceptable impact on other high-priority tasks, the system enters a degraded processing mode: maintaining the original plan and notifying the end station through the notification system.

[0098] In addition, such as Figure 3 As shown, the present invention also provides a logistics fleet intelligent dispatching system 300 based on BeiDou high-precision positioning, comprising: The data acquisition module 301 is used to acquire continuous high-precision trajectory point data of logistics vehicles; The identification module 302 identifies congested road sections and potential traffic delay areas based on the continuous high-precision trajectory point clustering analysis results; Prediction module 303 predicts the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructs a dynamic traffic capacity map; The allocation module 304 dynamically reallocates and controls the delivery time window based on the dynamic traffic capacity map and the real-time vehicle location information.

[0099] The functions and implementation methods of each module of the system have been described above with reference to the attached diagrams, and will not be repeated here.

[0100] The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning of the present invention includes: acquiring continuous high-precision trajectory point data of logistics vehicles through the BeiDou high-precision positioning system, which can accurately reflect the real-time location and driving status of vehicles on the road; then, performing cluster analysis on these trajectory point data to identify road segments with similar driving characteristics, and further identifying congested road segments and potential traffic delay areas, thereby conducting a refined assessment of road traffic conditions; subsequently, combining historical traffic delay models, using big data analysis technology to predict the traffic efficiency of each road segment at different time periods, and constructing a dynamic traffic capacity map, which not only covers current real-time road condition information, but also includes historical experience data and time-dimensional influencing factors; based on the constructed dynamic traffic capacity map and the real-time location information of each vehicle, the system can dynamically reallocate and regulate the delivery time window, that is, flexibly adjusting the arrival time window of each vehicle based on considering the existing traffic conditions, expected travel time, and order urgency, to ensure timely delivery of goods and reduce the risk of delays. This technology effectively solves the problem of low time-delivery rates at the last mile. By combining high-precision positioning data with historical delay models, it improves the intelligence and flexibility of transportation scheduling, and significantly enhances the timeliness of logistics services and customer satisfaction.

[0101] The exemplary systems and methods of the present invention have been specifically shown and described with reference to the above embodiments, which are merely examples of the best mode for implementing the systems and methods. Those skilled in the art will understand that various changes can be made to the embodiments of the systems and methods described herein without departing from the spirit and scope of the invention as defined in the appended claims when implementing the systems and / or methods.

Claims

1. A method for intelligent scheduling of logistics fleets based on BeiDou high-precision positioning, characterized in that, include: Acquire continuous, high-precision trajectory point data of logistics vehicles; Based on the results of the continuous high-precision trajectory point clustering analysis, congested road sections and potential traffic delay areas are identified. Based on historical traffic delay models, the traffic efficiency of each road segment at different time periods is predicted and a dynamic traffic capacity map is constructed. The delivery time window is dynamically reallocated and controlled based on the dynamic traffic capacity map and real-time vehicle location information.

2. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 1, characterized in that, The identification of congested road sections and potential traffic delay areas based on the continuous high-precision trajectory point clustering analysis results further includes: Obtain the GPS coordinate sequence of logistics vehicles and extract the timestamp and latitude / longitude triplet; The timestamps are divided according to a specific time granularity to form a trajectory segment dataset; The trajectory points of each segment are clustered based on the K-means algorithm to obtain multiple dense point groups; If the number of cluster points exceeds the set threshold K0 and the average moving speed is lower than V0, it is judged as a congested road segment.

3. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 1, characterized in that, The method of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Collect traffic records for a specific road section at different times of the day; Calculate the average travel time Ta within each time period; The current traffic efficiency of this road segment is calculated using the following formula: E = 1 / (Ta × α + β), where α is the time variation weight and β is the basic delay compensation coefficient. Prioritize nodes based on traffic efficiency and update graph node information.

4. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 1, characterized in that, The dynamic reallocation and control of delivery time windows based on the dynamic capacity map and real-time vehicle location information further includes: Collect the current location coordinates and remaining distance D to the destination of the logistics vehicle; Compare the vehicle's location with available paths in the dynamic map; The following formula is used to determine whether a route needs to be replanned: if D < δ × Ct, then the route is reassigned; where δ is the safety distance coefficient and Ct is the current road segment's capacity. If rescheduling is required, update the scheduled delivery time window for that vehicle in the task dispatch system.

5. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 1, characterized in that, The dynamic reallocation and control of delivery time windows based on the dynamic capacity map and real-time vehicle location information further includes: The vehicle's latitude and longitude coordinates are obtained in real time through the Beidou receiver, and the current vehicle speed V is calculated. Compare the real-time load status S of the road network node where the vehicle is currently located; Use the following formula to determine whether the current road conditions require an adjustment to the delivery rhythm: If S > σ × Ta, then adjust the delivery window time ΔT = ΔT_max × (1 - S / Ta); where σ is the warning threshold and Ta is the average travel time during that period; If the adjustment conditions are met, the delivery scheduling optimization module will be triggered.

6. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 5, characterized in that, The method of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Collect a dataset M of hourly traffic records for a specific road during a particular season; The data is processed by weighted moving average to smooth out outlier data points, and the hourly average latency Th is calculated. A geospatial factor Fg is introduced to characterize the impact of different road topologies on traffic efficiency, and a dynamic map model G=Th×Fg is constructed. The model-generated data is fused with real-time traffic data to form a traffic capacity map with time-varying characteristics.

7. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 1, characterized in that, The method of predicting the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructing a dynamic traffic capacity map further includes: Extract multidimensional variables from historical big data: weather conditions (W), peak hours (H), and emergencies (E); A linear regression model P=ω1×W+ω2×H+ω3×E+ω0 is established to estimate the average delay probability P during this period. Priority markers in dynamic maps are constructed using P-values ​​combined with real-time GPS data. If P > threshold, mark the route as a risky path and avoid selecting it during task assignment.

8. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 7, characterized in that, The identification of congested road sections and potential traffic delay areas based on the continuous high-precision trajectory point clustering analysis results further includes: After preprocessing the trajectory data, key event points are extracted, and a road segment dwell time matrix R is generated. The Markov chain method was used to evaluate the dwell behavior transfer patterns of different vehicles on this road segment; Use the following formula to determine if there is a potential passage delay: Pz=(∑(R_i≥tm)) / N, where tm is the tolerance time and N is the number of samples; If Pz > T2, then the road segment is determined to have potential travel delay.

9. The intelligent dispatching method for logistics fleets based on BeiDou high-precision positioning according to claim 8, characterized in that, The dynamic reallocation and control of delivery time windows based on the dynamic capacity map and real-time vehicle location information further includes: Extract the real-time traffic density Df and average travel time Ta of the target road segment from the dynamic map; The comprehensive impact value of the road segment is calculated using the following formula: I = λ × Df + (1 - λ) × Ta, where λ is the traffic priority factor. If I exceeds the set threshold It, the emergency dispatch procedure will be initiated to dynamically reconstruct the time window. The adjusted delivery time window must meet the transportation timeliness SLA.

10. A smart dispatching system for logistics fleets based on BeiDou high-precision positioning, characterized in that, include: The data acquisition module is used to acquire continuous high-precision trajectory point data of logistics vehicles; The identification module identifies congested road sections and potential traffic delay areas based on the continuous high-precision trajectory point clustering analysis results; The prediction module predicts the traffic efficiency of each road segment at different time periods based on the historical traffic delay model and constructs a dynamic traffic capacity map. The allocation module dynamically reallocates and controls the delivery time window based on the dynamic traffic capacity map and the real-time vehicle location information.

Images