A large-scale intelligent dispatching method and system for ride-hailing services

Abstract

This invention relates to the field of order management technology, specifically disclosing a large-scale intelligent dispatching method and system for ride-hailing services. The method includes: acquiring multimodal dispatching data for orders; obtaining a dispatching dataset based on the multimodal dispatching data, and acquiring dynamic pick-up timeliness and regional supply-demand tension based on the dispatching dataset; acquiring driver status data near the order location, and generating a candidate order-driver set based on the driver status data, dynamic pick-up timeliness, and regional supply-demand tension; obtaining a screening and matching score for each driver based on the candidate order-driver set; and generating an output dispatching result based on the screening and matching score. This invention significantly improves the accuracy and speed of large-scale ride-hailing dispatching through the above-described processing steps from data acquisition, feature extraction, candidate screening, score calculation to dispatching output.

Description

Technical Field

[0001] This invention relates to the field of order management technology, and in particular to a large-scale intelligent dispatching method and system for ride-hailing services. Background Technology

[0002] Ride-hailing platforms use intelligent dispatch systems to allocate passenger orders to suitable available drivers in real time, thereby improving matching efficiency and user experience. Existing large-scale ride-hailing dispatch methods typically include the following process: obtaining the order's origin location and the driver's real-time location based on GPS positioning; defining a fixed geographical radius (e.g., 3 kilometers) centered on the order's origin to filter candidate drivers; then calculating the matching score for each order-driver, with scoring factors typically including Euclidean distance, estimated pick-up time, driver rating, and order value; and finally, using a combinatorial optimization algorithm (such as the Hungarian algorithm, KM algorithm, or greedy strategy) for global optimization to maximize overall utility and allocate orders to the optimal driver.

[0003] Existing order dispatching methods typically employ a linear weighted summation model when calculating order-driver matching scores. This model multiplies a few static factors, such as pick-up distance, order value, and driver rating, by fixed weights and then sums them. This simplistic model has the following fundamental limitations: pick-up time estimation relies solely on Euclidean distance or historical average speed, failing to dynamically reflect the non-linear impact of real-time traffic events and environmental traffic density on travel time. This results in estimation errors of several times the normal rate in congestion or sudden accident scenarios. Furthermore, it lacks a quantitative perception of passenger urgency, treating all orders as homogeneous demands, and failing to prioritize urgent orders. Simultaneously, it ignores the dynamic changes in drivers' willingness to accept orders, assuming all idle drivers have the same level of enthusiasm. This can lead to orders being assigned to less active drivers, resulting in delayed order acceptance or order rejection. Summary of the Invention

[0004] The purpose of this invention is to provide a large-scale intelligent dispatching method and system for ride-hailing services to solve the technical problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A large-scale intelligent dispatching method for ride-hailing services includes: Obtain multimodal data on order dispatch; The dispatch dataset is obtained based on the dispatch multimodal data, and the dynamic pick-up time and regional supply and demand tension are obtained based on the dispatch dataset. Obtain driver status data near the order location, and generate a candidate order-driver set based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension; The filtering and matching score for each driver is obtained based on the candidate order-driver set; The output dispatch result is generated based on the filtering and matching score.

[0006] Preferably, the step of obtaining the dispatch dataset based on the dispatch multimodal data includes: The location information of the order is obtained based on the dispatch multimodal data, wherein the location information includes the order's starting point location information and the order's expected destination location information; Based on the order origin location information and the order pre-arrival destination location information, obtain the real-time traffic events, environmentally perceived traffic density and static road network topology of the order route, and obtain the dynamic road condition impedance coefficient based on the real-time traffic events, environmentally perceived traffic density and static road network topology of the order route; Based on the order dispatch multimodal data, the order inflow rate, idle driver density, and environmentally perceived traffic density are obtained; based on the order inflow rate, idle driver density, and environmentally perceived traffic density, the regional supply and demand tension is obtained. The passenger refresh frequency and input dwell time are obtained based on the dispatch multimodal data, and the passenger urgency index is obtained based on the passenger refresh frequency and input dwell time. Based on the dispatch multimodal data, the real-time speed-to-speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate of nearby idle drivers are obtained, and the driver order acceptance vitality value is obtained based on the real-time speed-to-speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate. The dynamic road condition impedance coefficient, regional supply and demand tension, passenger urgency index, and driver order acceptance vitality value are combined to form the order dispatch dataset.

[0007] Preferably, the step of obtaining the dynamic pick-up time based on the dispatch dataset includes: Extract the dynamic road condition impedance coefficient, passenger urgency index, and driver order-accepting vitality value from the dispatch dataset; The real-time traffic status of the road segments around the order starting point location information and the isochronous travel distance in each direction of the order starting point location information are obtained based on the dynamic road condition impedance coefficient. Input the passenger urgency index and driver order-accepting vitality value into the dynamic pick-up timeliness coefficient calculation formula to obtain the dynamic pick-up timeliness coefficient; The dynamic pick-up timeliness is obtained based on the passenger urgency index, driver order-accepting vitality value, and dynamic pick-up timeliness coefficient.

[0008] Preferably, the step of obtaining the regional supply and demand tension based on the dispatch dataset includes: The regional supply and demand tension gradient is obtained based on the order inflow speed, idle driver density, and environmentally perceived traffic density. The regional supply and demand tension is obtained based on the regional supply and demand tension gradient.

[0009] Preferably, the step of generating a candidate order-driver set based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension includes: The dynamic pick-up reachability domain of the order is obtained based on the dynamic pick-up time and the real-time road condition impedance coefficient around the order's starting point. Obtain the real-time locations of multiple available drivers, spatially match the real-time location of each available driver with the dynamic pick-up reach domain, and generate an initial candidate driver set; The regional supply and demand matching degree of candidate drivers is obtained based on the regional supply and demand tension and the initial candidate driver set. Drivers whose regional supply and demand match rate is higher than a preset matching threshold are selected to form a priority candidate driver set; Based on the driver order-acceptance vitality value in the driver status data, drivers with vitality values ​​higher than the vitality admission threshold are selected from the priority candidate driver set and combined with the corresponding orders to generate a candidate order-driver set.

[0010] Preferably, the step of obtaining the filtering and matching score for each driver based on the candidate order-driver set includes: Order information is obtained from the candidate order-driver set, and the first supply-demand tension is obtained based on the passenger urgency index and order information corresponding to the order. Based on the candidate order-driver set, obtain the driver order-acceptance vitality value of the driver, and obtain the second supply and demand tension based on the driver order-acceptance vitality value; The order origin location information and driver real-time location are obtained based on the order information. The dynamic pick-up time of the order and driver is obtained based on the order origin location information, driver real-time location, and dynamic road condition impedance coefficient of each road segment. Based on the passenger urgency index and the driver order-acceptance vitality value, obtain the comprehensive adjustment coefficient of urgency-vitality for this order-driver; The order-driver matching score is obtained based on the dynamic pick-up time, urgency-vitality comprehensive adjustment coefficient, order base value, first supply-demand tension and second supply-demand tension.

[0011] Preferably, the step of generating the output dispatch result based on the filtering and matching score includes: Obtain the filtering and matching score of each order-driver in the candidate order-driver set, and generate a set of matching scores to be dispatched; Based on the matching score set of the orders to be dispatched, the candidate drivers corresponding to each order are sorted in descending order of score to obtain the driver matching priority for each order; Based on the driver matching priority, the driver ranked first for each order is selected as the pre-matched driver, and it is detected whether the same driver is selected as the pre-matched driver for multiple orders at the same time. If it does not exist, the order will be directly linked to the driver; If such a driver exists, the driver is assigned the highest-scoring order based on the filtering and matching scores of the multiple orders corresponding to that driver. The other orders are then released back to the candidate order-driver set for rematching until each order is assigned a unique driver, thus generating the final dispatch matching pair. The final dispatch matching pair is used to generate the output dispatch result, and the dispatch instruction is pushed to the corresponding driver terminal.

[0012] This invention also discloses a large-scale intelligent dispatching system for ride-hailing services, comprising: The data acquisition module is used to acquire multimodal data on order dispatch. The dispatch dataset acquisition module is used to acquire the dispatch dataset based on the dispatch multimodal data, and to acquire the dynamic pick-up timeliness and regional supply and demand tension based on the dispatch dataset. The order-driver set acquisition module is used to obtain driver status data near the order location and generate candidate order-driver sets based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension. The matching score acquisition module is used to obtain the filtering matching score of each driver based on the candidate order-driver set; The dispatch result generation module is used to generate output dispatch results based on the filtering and matching scores.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a large-scale intelligent dispatching method for ride-hailing services.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a large-scale intelligent dispatching method for ride-hailing services. The beneficial effects of this application are as follows: By collecting multi-source real-time data, this invention enables the dispatch system to perceive multi-dimensional information such as traffic events, environmental traffic flow, passenger behavior, and driver status. Based on this, deep fusion and feature extraction of multimodal data are performed to generate dynamic road condition impedance coefficients, regional supply and demand tension, passenger urgency index, and driver order-acceptance vitality values, thereby supporting the dispatch decision to accurately respond to real-time road conditions, supply and demand dynamics, and individual behavioral characteristics. Multi-level screening of candidate drivers based on dynamic pick-up timeliness and regional supply and demand tension not only effectively reduces the size of the candidate pool but also takes into account regional supply and demand balance and driver willingness to respond. Furthermore, by comprehensively calculating the matching score using dynamic pick-up time, urgency-vitality adjustment coefficient, and supply-demand difference penalty coefficient, a multi-objective refined evaluation of order value, pick-up efficiency, user needs, driver willingness, and regional balance is achieved, effectively reducing pick-up time prediction errors and the risk of supply-demand imbalance. Finally, a priority-based iterative conflict resolution mechanism is adopted, ensuring global matching coordination while avoiding complex optimization calculations, enabling the system to quickly output feasible dispatch results during peak periods. This invention significantly improves the accuracy and speed of large-scale ride-hailing order dispatch through the above-described processing steps, from data collection, feature extraction, candidate screening, score calculation to order dispatch output. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of a method flow according to an embodiment of this application.

[0016] Figure 2 This is a schematic diagram of the system structure according to an embodiment of this application.

[0017] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application.

[0018] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] like Figure 1 As shown, this application provides a large-scale intelligent dispatching method for ride-hailing services, including: S1. Obtain multimodal data of order dispatch; For example, ride-hailing platforms receive passenger-initiated order requests in real time and collect multimodal information related to order dispatch from multiple data sources. Order dispatch multimodal data is a multi-source heterogeneous data set encompassing order text information, spatiotemporal environmental information, user behavior information, and vehicle status information. Specifically, order information includes the passenger's input start and end points, estimated trip distance, and value; real-time traffic event data is obtained through map service provider APIs, including the location, type, impact range, and estimated duration of events such as accidents, road closures, construction, and temporary traffic control; environmental perception data is obtained through roadside equipment or third-party traffic information service providers, including real-time traffic density on various road segments, average waiting time at intersections, and average vehicle speed on road segments; passenger behavior data is collected through embedded points on the app, including the frequency of passenger refreshes on the order page, dwell time when inputting start and end points, and operation records before order cancellation; driver status data is collected through the driver's app and onboard sensors, including the driver's real-time location, empty / carrying status, real-time speed ratio to road segment speed limit, frequency of rapid acceleration and deceleration per unit time, and historical order acceptance rate based on historical data statistics.

[0021] S2. Obtain the dispatch dataset based on the dispatch multimodal data, and obtain the dynamic pick-up time and regional supply and demand tension based on the dispatch dataset; For example, the multimodal data collected in step S1 is fused to generate a structured dispatch dataset. The dispatch dataset is a core data set used in subsequent dispatch processes after cleaning, alignment, and feature extraction. It includes derived indicators such as dynamic road condition impedance coefficient, regional supply and demand tension, passenger urgency index, and driver order-accepting vitality value. The specific generation process is as follows: Based on the order's origin and destination locations, combined with real-time traffic events, environmentally perceived traffic density, and static road network topology (road grade, connectivity, speed limit, etc.), path analysis technology is used to construct the traffic impedance of each road segment at the current moment, obtaining the dynamic road condition impedance coefficient; based on the new order inflow rate per unit time, idle driver density, and environmentally perceived traffic density, the supply and demand tension gradient of each grid area is calculated using a supply and demand balance model and divided into high tension, medium tension, and low tension levels; based on passenger refresh frequency and input dwell time, combined with order context (such as whether it is a pre-booked order, whether it is in a special area such as an airport), a passenger urgency index is generated; based on the driver's real-time speed to speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate, a driver order-accepting vitality value is generated. After obtaining the dispatch dataset, key information is further extracted to obtain dynamic pick-up timeliness and regional supply and demand tension: Dynamic pick-up timeliness refers to the available pick-up range or time threshold centered on the order origin, adjusted based on real-time road conditions, passenger urgency, and driver vitality value. Specifically, it is obtained by generating an isochronous travel distance map by integrating dynamic road condition impedance coefficients and correcting the basic range by combining passenger urgency index and driver order-accepting vitality value; Regional supply and demand tension is directly obtained from the dispatch dataset and used for subsequent candidate driver screening and matching score calculation.

[0022] S3. Obtain driver status data near the order location, and generate a candidate order-driver set based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension; For example, after obtaining dynamic pick-up timeliness and regional supply-demand tension, candidate drivers are selected for each order based on real-time driver status data, generating a candidate order-driver set. First, based on dynamic pick-up timeliness and the dynamic road condition impedance coefficient around the order's origin, an isochronous diffusion algorithm is used to determine the dynamic pick-up reachability of the order, i.e., the spatial range that can actually be reached from the order's origin within a preset pick-up time threshold. This range is determined based on real-time road conditions, unlike traditional fixed-radius selection. Second, the real-time locations of all available drivers are obtained, and drivers located within the dynamic pick-up reachability are selected to form an initial candidate driver set. Then, based on regional supply-demand tension, the matching degree between the supply-demand tension of the driver's region and the order's region is calculated for each driver in the initial candidate set: if the driver comes from a low-tension area and the order is located in a high-tension area, the matching degree is high, indicating that the driver's arrival to pick up the order helps alleviate regional imbalance; drivers with matching degrees higher than a preset threshold are retained to form a priority candidate driver set. Finally, based on the driver order-acceptance vitality value in the driver status data, drivers with vitality values ​​higher than the vitality admission threshold are selected from the priority candidate set and combined with the corresponding orders to generate the final candidate order-driver set.

[0023] S4. Obtain the screening and matching score for each driver based on the candidate order-driver set; For example, for each pair in the candidate order-driver set, a screening and matching score is calculated by integrating multi-dimensional information. For each order-driver, the passenger urgency index, basic order value, and first supply-demand tension of the order's location are extracted from the dispatch dataset. Simultaneously, the driver's order-accepting vitality value and second supply-demand tension of the driver's location are extracted. Based on the order's origin and the driver's real-time location, combined with the dynamic road condition impedance coefficients corresponding to each road segment, the dynamic pick-up time is determined by segmented accumulation: the driving path is divided into segments, the length of each segment is divided by the product of the free-flow velocity and (1 - impedance coefficient), and the results of each segment are summed to obtain the estimated pick-up time. Based on the passenger urgency index and driver order-accepting vitality value, an urgency-vitality comprehensive adjustment coefficient is generated: the urgency index is multiplied by a preset urgency weight, and the vitality value is multiplied by a preset vitality value weight, and then summed. The screening and matching score for the order-driver is obtained by multiplying these three factors together.

[0024] S5. Generate the output dispatch result based on the filtering and matching score; For example, based on the screening and matching scores of all candidate order-driver pairs, the final dispatch result is generated through conflict resolution. First, the scores of all candidate pairs are collected to form a set of matching scores to be dispatched. For each order, its candidate drivers are sorted from highest to lowest score to obtain a driver matching priority. Pre-matching and conflict detection are performed: the driver ranked first for each order is selected as the pre-matching driver; if the same driver is selected as the pre-matching driver by multiple orders, the order with the highest score is retained and bound to that driver, while the remaining orders are released back to the candidate set and re-sorted and matched until all orders are bound to a unique driver, generating the final dispatch matching pair. Finally, based on the final dispatch matching pair, the dispatch instruction is pushed to the corresponding driver's terminal. This process uses priority-based iterative conflict resolution, ensuring global coordination of the matching while avoiding complex optimization calculations, enabling rapid output of dispatch results in large-scale concurrent scenarios.

[0025] As described in steps S1-S5 above, this application acquires multi-source real-time data, enabling the dispatch system to comprehensively perceive traffic events, environmental traffic flow, passenger behavior, and driver status, significantly improving the richness and accuracy of input information. Step S2 performs deep fusion and feature extraction on the multimodal data to generate dynamic road condition impedance coefficients, regional supply and demand tension, passenger urgency index, and driver order-acceptance vitality value, allowing dispatch decisions to accurately reflect real-time road conditions, supply and demand dynamics, and individual behavioral characteristics. Step S3 performs multi-level screening of candidate drivers based on dynamic pick-up timeliness and regional supply and demand tension, effectively reducing the size of candidate pairs. Simultaneously, it takes into account regional supply and demand balance and driver willingness to respond, significantly improving the quality of the candidate set and subsequent calculation efficiency; by calculating the matching score through step S4, which integrates dynamic pick-up time, urgency-activity adjustment coefficient and supply-demand difference penalty factor, it achieves a multi-objective refined evaluation of order value, pick-up efficiency, user needs, driver willingness and regional balance, significantly reducing the error in pick-up time prediction and the risk of supply-demand imbalance; by adopting a priority iterative conflict resolution mechanism in step S5, it avoids complex optimization calculations while ensuring global coordination of matching, enabling the system to output feasible order dispatch results during peak periods and increasing order dispatch accuracy.

[0026] In one embodiment, the step of obtaining the dispatch dataset based on the dispatch multimodal data includes: S201. Obtain the location information of the order based on the dispatch multimodal data, wherein the location information includes the order's starting point location information and the order's expected destination location information; For example, this invention extracts the order's location information from multimodal dispatch data by parsing the order request initiated by the passenger. The order's origin location information refers to the latitude and longitude coordinates or structured address of the pick-up point entered by the passenger, while the order's expected destination location information refers to the latitude and longitude coordinates or structured address of the destination entered by the passenger. In actual ride-hailing scenarios, passengers select points on the app map or manually input to determine the origin and destination. The platform instantly captures these coordinate data as a spatial reference for subsequent route analysis and traffic condition matching. By obtaining accurate origin and destination locations, a clear spatial range is provided for subsequent calculation of dynamic traffic condition impedance based on real-time traffic events and environmental perception data, ensuring that all subsequent analyses revolve around the actual driving path of the order, avoiding route misjudgment due to ambiguous locations.

[0027] S202. Based on the order origin location information and the order pre-arrival destination location information, obtain the real-time traffic events, environmentally perceived traffic density and static road network topology of the order route, and obtain the dynamic road condition impedance coefficient according to the real-time traffic events, environmentally perceived traffic density and static road network topology of the order route; For example, this invention uses the origin and destination of an order as constraints to determine the planned travel route of the order, and collects real-time traffic events, environmentally perceived traffic density, and static road network topology information from multi-source data along the route. Real-time traffic events refer to dynamic events such as accidents, road closures, construction, and temporary traffic control occurring on the route, obtained through map service provider interfaces. Each event includes location, type, scope of impact, and timeliness. Environmentally perceived traffic density refers to the real-time vehicle density on each segment of the route, obtained through roadside equipment or floating car data, reflecting the current level of congestion. Static road network topology refers to the inherent attributes of the road network, including road grade, number of lanes, connectivity, speed limits, and turning restrictions, which are pre-stored in the road network database. After acquiring the above data, this invention maps real-time traffic events to corresponding road segments, generates event impact weights based on event type and severity, compares environmentally perceived traffic density with road capacity to generate a density impact factor, and combines road grade and speed limit information from the static road network topology to generate basic traffic impedance. Subsequently, by fusing event impact, density impact, and static topology, a dynamic road condition impedance coefficient is generated for each road segment on the route. This coefficient quantifies the actual traffic difficulty of that road segment at the current moment; a higher value indicates greater traffic difficulty. This process integrates discrete real-time events with continuous traffic flow perception into a unified road segment impedance metric, enabling subsequent pick-up time estimation to dynamically reflect real-world road condition changes and significantly improving the accuracy of time prediction.

[0028] S203. Obtain the order inflow speed, idle driver density and environmentally perceived traffic density based on the order dispatch multimodal data, and obtain the regional supply and demand tension based on the order inflow speed, idle driver density and environmentally perceived traffic density; For example, this invention extracts order inflow rate, idle driver density, and environmentally sensed traffic density from multimodal dispatch data, using urban grids as units. Order inflow rate refers to the number of newly generated orders per unit time (e.g., per minute), reflecting the demand intensity of the area; idle driver density refers to the number of drivers in an empty state within a unit area who can accept orders, reflecting the supply capacity of the area; and environmentally sensed traffic density reflects the road load level of the area. After obtaining the above data, this invention constructs a supply-demand balance model, comparing the order inflow rate with the idle driver density to preliminarily determine the supply-demand gap; simultaneously, it introduces environmentally sensed traffic density as a congestion suppression factor, because even with sufficient supply, severe road congestion will reduce actual service capacity. Specifically, by dividing the order inflow rate by the product of the idle driver density and the congestion suppression factor, a regional supply-demand tension gradient is obtained, which quantifies the demand pressure borne per unit of supply capacity. Subsequently, the continuous gradient values ​​are discretized into levels, such as highly tense, moderately tense, and lowly tense, forming the regional supply-demand tension level.

[0029] S204. Obtain the passenger refresh frequency and input dwell time based on the dispatch multimodal data, and obtain the passenger urgency index based on the passenger refresh frequency and input dwell time; For example, passenger refresh frequency refers to how frequently a passenger performs a pull-to-refresh or re-search operation on the application's order page. It is typically measured in refreshes per unit of time (e.g., per minute). This metric reflects a passenger's impatience with the current lack of ride availability and their eagerness to receive an order as soon as possible. Input dwell time refers to the time interval between a passenger entering their origin or destination location and finally confirming and submitting the order. The length of this time can indirectly reflect the passenger's decisiveness or hesitation in decision-making. Generally, a shorter dwell time indicates that the passenger may have already predetermined their trip and is eager to place an order, while a longer dwell time may mean that the passenger is hesitant or repeatedly modifying their decision, indicating a relatively low sense of urgency. The passenger urgency index is a comprehensive quantitative value used to characterize a passenger's sensitivity to the speed of order dispatch and the urgency of their need at the current moment. This index ranges from 0 to 1, with a higher value indicating a more urgent need for a ride.

[0030] S205. Based on the dispatch multimodal data, obtain the real-time speed-to-speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate of nearby idle drivers, and obtain the driver order acceptance vitality value based on the real-time speed-to-speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate. For example, this invention extracts the real-time driving status and historical behavior data of each idle driver from the multimodal dispatch data. The real-time speed-to-speed limit ratio refers to the ratio of the driver's current speed to the legal speed limit of the road segment. If the ratio is close to or exceeds 1, it indicates that the driver is driving at a relatively high speed and may be actively accepting orders; if the ratio is much lower than 1, it may indicate that the driver is driving slowly or stopped, with a low willingness to accept orders. The frequency of sudden acceleration and deceleration refers to the number of times the driver accelerates or brakes suddenly per unit time. High-frequency sudden acceleration and deceleration usually indicates that the driver is in a congested area or has an aggressive driving style, but it may also reflect their frustration with road conditions, indirectly affecting their order-accepting response. The historical order acceptance rate is the ratio of the number of orders accepted to the number of orders dispatched by the driver over a past period, reflecting their long-term service willingness and reliability. After obtaining the above data, this invention normalizes the real-time speed-to-speed limit ratio as the base vitality value, uses the frequency of sudden acceleration and deceleration as an adjustment factor (too high a frequency may reduce the vitality value), and uses the historical order acceptance rate as a weighted correction for long-term trust, ultimately generating the driver's order-accepting vitality value.

[0031] S206. Combine the dynamic road condition impedance coefficient, regional supply and demand tension, passenger urgency index and driver order acceptance vitality value to form the order dispatch dataset; For example, this invention structurally encapsulates the dynamic road condition impedance coefficient, regional supply and demand tension, passenger urgency index, and driver order-accepting activity value generated in the aforementioned steps to form an order dispatch dataset. This dataset is centered on orders, with each order record associated with its origin, destination, dynamic road condition impedance coefficient (which can be refined into impedance for each segment of the path), regional supply and demand tension, and passenger urgency index. Simultaneously, each idle driver record is associated with its current location, driver order-accepting activity value, and regional supply and demand tension. Through this combination, the order dispatch dataset unifies multi-source heterogeneous raw information into decision features with clear business meaning, providing complete input for subsequent dynamic pick-up timeliness calculation, candidate driver screening, and matching score evaluation.

[0032] As described in steps S201-S206 above, this invention obtains precise order origin and destination locations in step S201, providing an accurate spatial reference for subsequent route analysis and road condition matching; in step S202, it integrates real-time traffic events, environmentally perceived traffic density, and static road network topology to generate a dynamic road condition impedance coefficient, enabling the pickup time estimate to be dynamically adjusted according to actual road conditions such as accidents and congestion, significantly reducing time prediction deviation; in step S203, it combines order inflow speed, idle driver density, and environmentally perceived traffic density to obtain regional supply and demand tension, enabling the system to identify supply and demand imbalances in each region caused by changes in the number of orders and drivers, providing regional situation reference for subsequent order dispatch decisions; and in step S204, it uses passenger refresh frequency and... By inputting the dwell time to obtain the passenger urgency index, the system can perceive the user's actual need for order dispatch speed, thus giving appropriate preference to urgent orders during the matching process. In step S205, the driver's order-accepting vitality value is obtained based on the driver's real-time speed-to-speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate, enabling the system to judge the driver's current order-accepting willingness and response status, avoiding assigning orders to drivers with low activity. In step S206, the dynamic road condition impedance coefficient, regional supply and demand tension, passenger urgency index, and driver order-accepting vitality value are combined to form an order dispatch dataset, enabling the subsequent order dispatch process to evaluate each order from four dimensions: road conditions, regional supply and demand, user needs, and driver status, providing multi-dimensional data support for improving matching accuracy.

[0033] In one embodiment, the step of obtaining the dynamic pick-up time based on the dispatch dataset includes: S207. Extract the dynamic road condition impedance coefficient, passenger urgency index and driver order acceptance vitality value from the dispatch dataset. For example, three key features are first extracted from the constructed order dispatch dataset: dynamic road condition resistance coefficient, passenger urgency index, and driver order acceptance vitality value. The dynamic road condition resistance coefficient is a quantitative value of the traffic difficulty generated for each road segment around the order origin or along the planned order route, recorded in the order dispatch dataset. This coefficient integrates real-time traffic events, environmentally perceived traffic density, and static road network topology information, reflecting the actual traffic resistance of each road segment at the current moment. The passenger urgency index is a quantitative value of the urgency of user demand, generated based on passenger refresh frequency and input dwell time, recorded in the order dispatch dataset. This value ranges from 0 to 1; a higher value indicates a stronger expectation from passengers for order dispatch speed. The driver order acceptance vitality value is a quantitative value of the driver's current willingness to accept orders and their responsiveness, generated based on the driver's real-time speed-to-speed-limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate, recorded in the order dispatch dataset. This value also ranges from 0 to 1; a higher value indicates a more proactive driver and a higher likelihood of quickly accepting orders.

[0034] S208. Obtain the real-time road segment traffic status around the order starting point location information and the isochronous travel distance in each direction of the order starting point location information based on the dynamic road condition impedance coefficient. For example, this invention divides the surrounding space into multiple directional sectors or grid units, centered on the order's starting point. For each road segment in each direction, the traffic status is evaluated using dynamic road condition impedance coefficients. Real-time road segment traffic status refers to the traffic efficiency of each road segment at the current moment, expressed as the level of impedance coefficients; higher impedance coefficients indicate greater difficulty in passage. Based on this, this invention employs an isochronous diffusion algorithm, starting from the order's starting point and gradually extending outwards along the road network, accumulating the travel time for each road segment. The accumulation method for travel time is: dividing the road segment length by the product of the free-flow velocity of that road segment and the dynamic road condition impedance coefficient of that road segment, yielding the time required to pass through that road segment. When the accumulated time reaches a preset pick-up time threshold (e.g., 5 minutes), diffusion stops. At this point, all covered spatial locations form a closed isochronous line, and the area enclosed by this isochronous line represents the achievable spatial range from the order's starting point within the threshold time. Furthermore, the furthest reachable distance in each direction is extracted to form the isochronous travel distance in each direction.

[0035] S209. Input the passenger urgency index and driver order-accepting vitality value into the dynamic pick-up timeliness coefficient calculation formula to obtain the dynamic pick-up timeliness coefficient. For example, this invention generates a dynamic adjustment coefficient for modifying the basic pick-up range by integrating the passenger urgency index and the driver's order-accepting vitality value. The dynamic pick-up timeliness coefficient is a scaling factor that adjusts the basic pick-up time or distance based on a combination of passenger urgency and driver willingness to respond. Preset urgency amplification coefficient and vitality value attenuation coefficient are obtained. and The value is obtained based on historical order data statistics: by analyzing the relationship between passenger urgency index and actual pick-up time in historical orders, the appropriate proportion for expanding the pick-up range for every 0.1 increase in urgency is determined, thereby determining... By analyzing the relationship between driver vitality value and order acceptance response rate, the appropriate reduction in the pick-up range for every 0.1 increase in vitality value was determined, thus establishing... The coefficient is obtained using the following formula: ; In the formula, This indicates the dynamic pick-up timeliness coefficient. This indicates the passenger urgency index. This indicates the driver's activity level in accepting orders. This indicates the preset urgency amplification factor. This indicates the preset vitality value decay coefficient; when the passenger urgency index is high, Increasing this means that the pick-up area can be appropriately expanded to meet demand more quickly; when a driver's order-taking activity level is high... Reducing the distance means that drivers are more responsive and the pick-up area can be appropriately narrowed to ensure pick-up efficiency. Through this adjustment, pick-up timeliness can adaptively respond to the dynamic characteristics of both supply and demand.

[0036] S210. Obtain dynamic pick-up timeliness based on the passenger urgency index, driver order-accepting vitality value, and dynamic pick-up timeliness coefficient. For example, this invention combines a basic pick-up time threshold (e.g., 5 minutes) with a dynamic pick-up efficiency coefficient to obtain the final dynamic pick-up efficiency. Dynamic pick-up efficiency refers to the actual available pick-up time threshold or spatial range used to screen candidate drivers starting from the order origin, after comprehensively considering real-time road conditions, passenger urgency, and driver activity levels. In specific implementation, the basic pick-up time threshold can be multiplied by the dynamic pick-up efficiency coefficient to obtain a corrected pick-up time threshold; then, based on this threshold and combined with the isochronous travel distances in each direction obtained in step S208, the reachable domain is redefined to form the available pick-up range corresponding to the dynamic pick-up efficiency. For example, if the basic threshold is 5 minutes, a high urgency index makes... Increasing it to 1.2 will correct the pick-up time to 6 minutes, and the corresponding reachable area will expand accordingly; if the vitality value is high... If it drops to 0.8, the pick-up time will be reduced to 4 minutes, and the reach will shrink.

[0037] As described in steps S207-S210 above, this invention transforms the multi-source features in the dispatch dataset into dynamic pick-up timeliness. This timeliness is based on real-time traffic conditions and incorporates a two-way adjustment of passenger urgency and driver activity levels. This allows the determination of the pick-up range to break through the limitations of the traditional fixed radius. It can appropriately expand the range to improve response speed when demand is urgent, and shrink the range to ensure pick-up efficiency when drivers are active, thereby significantly improving the matching quality of candidate drivers and the accuracy of dispatch.

[0038] In one embodiment, the step of obtaining the regional supply and demand tension based on the dispatch dataset includes: S211. Obtain the regional supply and demand tension gradient based on the order inflow speed, idle driver density, and environmentally perceived traffic density. For example, this invention extracts order inflow rate, idle driver density, and environmentally sensed traffic density from the order dispatch dataset, using city grids as units. Order inflow rate refers to the number of newly generated orders in the grid area per unit time (e.g., per minute), reflecting the real-time demand intensity; idle driver density refers to the ratio of the number of drivers in the grid area who are available to accept orders to the area of ​​the grid, reflecting the real-time supply capacity; environmentally sensed traffic density refers to the vehicle density of the main roads in the grid area obtained through roadside equipment or floating car data, usually expressed in standard vehicle equivalents per kilometer, used to characterize the current road congestion level. After obtaining the above three data points, this invention constructs a supply-demand tension gradient model to compare demand pressure with supply capacity, and introduces traffic density as a congestion suppression factor. The congestion suppression coefficient (…) is obtained. Based on the characteristics of urban roads, a value is typically set between 0.3 and 0.7 to reflect the degree to which congestion reduces service capacity. The specific value can be obtained by fitting the dispatch efficiency during historical congestion periods. Road capacity limit. , (As a preset constant, determined based on road grade and design capacity), the regional supply-demand tension gradient is obtained using the following formula: ; In the formula, Indicates the regional supply and demand tension gradient. Indicates the speed at which orders flood in. Indicates the density of idle drivers. Indicates the perceived traffic density in the environment. This indicates the maximum road capacity (preset value) for this area. The term in the denominator of the formula represents the congestion mitigation coefficient. This indicates the degree to which congestion reduces supply capacity: as traffic density approaches road capacity, this term tends to... This reduces the denominator and increases the gradient, indicating that congestion exacerbates the supply-demand imbalance. When traffic density is low, this term approaches 1, and the gradient is mainly determined by the number of orders and drivers. Through this integrated calculation, the regional supply-demand tension gradient comprehensively reflects the three factors of demand pressure, supply capacity, and road congestion, quantifying the current supply-demand imbalance in each grid area in the form of continuous numerical values, providing a quantitative basis for subsequent classification of tension levels.

[0039] S212. Obtain the regional supply and demand tension level based on the regional supply and demand tension gradient; For example, the present invention converts the continuous gradient values ​​obtained in step S211 into discrete, easily understood and applicable regional supply and demand tension levels. Regional supply and demand tension refers to a classification label used to intuitively identify the supply and demand status of each region, typically divided into three levels: highly tense, moderately tense, and low tense. More levels can be added as needed. In specific implementation, two thresholds are preset: a first threshold... Second threshold (and The gradient values ​​for each grid cell. Compare with a threshold: If If so, the area is judged to be highly strained, indicating that order demand far exceeds supply capacity, and may be accompanied by severe congestion; if If so, it is judged as moderately tight, with supply and demand basically balanced but under slight pressure; if If the situation is as described above, it is considered to be under-stressed, with relatively sufficient supply. This tiered processing transforms the abstract continuous gradient into regional status labels with clear business implications, enabling the dispatch system to quickly identify hotspot areas, areas with supply-demand imbalances, and areas with idle resources. This provides intuitive contextual information for subsequent candidate driver screening, dynamic adjustment of pick-up times, and matching score calculation. For example, in highly stressed areas, the system can appropriately expand the pick-up range to process orders more quickly, while guiding drivers from less stressed areas to more stressed areas; in moderately stressed areas, the system maintains its usual strategy; and in low-stress areas, the pick-up range can be appropriately reduced to improve driver utilization.

[0040] As described in steps S211-S212 above, this invention integrates the original order inflow speed, idle driver density, and environmentally perceived traffic density into a regional supply-demand tension gradient, and then transforms it into an intuitive regional supply-demand tension level through threshold division. This process enables the dispatch system to perceive the demand pressure, supply capacity, and road congestion status of each region in real time, and outputs it in the form of hierarchical tags, which facilitates rapid invocation by subsequent decision-making modules, thereby effectively guiding driver scheduling and optimizing dispatch strategies.

[0041] In one embodiment, the step of generating a candidate order-driver set based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension includes: S301. Obtain the dynamic pick-up reachability domain of the order based on the dynamic pick-up timeliness and the real-time road condition impedance coefficient around the order's starting point. For example, this invention combines dynamic pick-up timeliness with real-time traffic impedance coefficient to determine the spatial range that can actually be reached within a given time from the order origin. Dynamic pick-up timeliness refers to the pick-up time threshold obtained in step S21, after correction for passenger urgency and driver vitality; real-time traffic impedance coefficient refers to the quantitative value of traffic difficulty obtained by fusing traffic events, traffic density, and road topology for each road segment in the dispatch dataset. Dynamic pick-up reachability domain refers to the maximum spatial area that can be covered by diffusion along the road network in all directions within the dynamic pick-up timeliness, centered on the order origin, based on the real-time traffic impedance coefficient. In specific implementation, this invention uses an isochronous diffusion algorithm: the road network around the order origin is divided into several directions or grid units, and the travel time is accumulated segment by segment along the path from the origin (the length of each segment is divided by the product of the free flow velocity of that segment and the impedance coefficient), until the accumulated time reaches the dynamic pick-up timeliness threshold. The area formed by connecting all grid units whose accumulated time does not exceed the threshold is determined as the dynamic pick-up reachability domain. This process transforms the abstract time threshold into a spatial boundary based on real road conditions, ensuring that subsequent driver screening is strictly limited to the actual range of available drivers, thus avoiding time estimation errors caused by a fixed radius.

[0042] S302. Obtain the real-time locations of multiple idle drivers, and spatially match the real-time location of each idle driver with the dynamic pick-up reachable domain to generate an initial candidate driver set; For example, this invention extracts the real-time location coordinates of all available drivers from driver status data, ensuring that each available driver is in an unloaded state and can accept orders. The coordinates of each driver are then compared to the dynamic pick-up reachability domain determined in step S301 to determine spatial inclusion: if the driver's location falls within the reachability domain, they are included in the initial candidate driver set; otherwise, they are excluded. This spatial matching operation ensures that every driver in the initial candidate set can reach the order origin within the dynamic pick-up timeframe, thus guaranteeing the feasibility of pick-up times from the outset. Compared to traditional fixed-radius filtering, this step dynamically adjusts the filtering range based on real-time traffic conditions, enabling more accurate identification of truly available driver resources and preventing drivers who are actually unreachable due to congestion or accidents from entering the candidate set.

[0043] S303. Obtain the regional supply and demand matching degree of candidate drivers based on the regional supply and demand tension and the initial candidate driver set; For example, this invention introduces a regional supply and demand balance factor to evaluate the degree of supply and demand matching between each driver in the initial candidate set and the order. Regional supply and demand tension refers to the tension level (e.g., high tension, medium tension, low tension) of each grid region obtained in step S22. Regional supply and demand matching degree refers to the coordination between the tension level of the driver's region and the tension level of the order's region, used to measure whether assigning the driver to the order helps alleviate or exacerbate the regional supply and demand imbalance. Specifically, this invention presets a matching score for each combination of tension levels: for example, if the driver is from a low-tension region and the order is in a high-tension region, the matching degree is high because the transfer helps supplement capacity to high-demand regions; if the driver is from a high-tension region and the order is in a low-tension region, the matching degree is low because it may further exacerbate the supply shortage in the high-tension region; if both levels are the same, the matching degree is moderate. By querying a preset matching score table, a corresponding regional supply and demand matching degree value is generated for each candidate driver.

[0044] S304. Select drivers whose regional supply and demand matching degree is higher than the preset matching threshold to form a priority candidate driver set; For example, the present invention sets a regional supply-demand matching threshold to filter out drivers who are better in terms of supply-demand balance. The matching degree of each candidate driver obtained in step S303 is compared with this threshold, and only drivers with a matching degree higher than the threshold are retained, forming a priority candidate driver set. This threshold can be dynamically adjusted according to city, time period, or operational strategy. For example, the threshold can be appropriately lowered during peak periods to ensure transport capacity, and raised during off-peak periods to optimize the supply-demand structure.

[0045] S305. Based on the driver order-acceptance vitality value in the driver status data, select drivers whose vitality value is higher than the vitality admission threshold from the priority candidate driver set, and combine them with the corresponding orders to generate a candidate order-driver set. For example, this invention extracts the driver order-accepting vitality value for each candidate driver from driver status data. This value reflects the driver's current willingness to accept orders and their responsiveness (obtained by fusing real-time vehicle speed, frequency of rapid acceleration and deceleration, and historical order acceptance rate). A vitality threshold (e.g., 0.6) is set, and only drivers with vitality values ​​higher than this threshold are retained. These drivers are then associated with current orders to form the final candidate order-driver set. This step ensures that the drivers ultimately participating in the matching process are not only spatially and temporally accessible and have a good supply-demand match, but also possess a positive attitude towards accepting orders, thereby significantly improving the order dispatch success rate and driver response speed. If no driver in the priority candidate set meets the vitality threshold, the threshold can be appropriately lowered or the search scope expanded. However, in most cases, multi-level screening ensures a sufficient number of high-quality candidates.

[0046] As described in steps S301-S305 above, this invention first delineates the dynamic pick-up reachability domain based on real-time traffic conditions to ensure spatiotemporal feasibility; secondly, it filters based on regional supply and demand matching to consider overall supply and demand balance; and finally, it uses driver order-acceptance activity values ​​to ensure individual responsiveness. This process significantly improves the quality of candidate orders and drivers, reducing invalid calculations and increasing order dispatch accuracy and success rate. In one embodiment, the step of obtaining the filtering and matching score for each driver based on the candidate order-driver set includes: S401. Obtain order information from the candidate order-driver set, and obtain the first supply-demand tension based on the passenger urgency index and order information corresponding to the order. For example, this invention extracts the currently processed order information from the candidate order-driver set. This order information includes basic attributes such as the order's unique identifier, origin location, destination location, and basic order value. Simultaneously, it obtains the passenger urgency index corresponding to the order from the dispatch dataset. This index has been generated in previous steps based on passenger refresh frequency and input dwell time. The first supply-demand tension refers to the supply-demand tension level of the grid area where the order is located. This level is directly read from the regional supply-demand tension mapping table in step S22 based on the area to which the order's origin location belongs. The specific grid is located by the order's origin location, and then the corresponding tension label (e.g., highly tense, moderately tense, or low tense) is obtained based on the grid's supply-demand tension gradient threshold division result.

[0047] S402. Obtain the driver's order-acceptance vitality value according to the candidate order-driver set, and obtain the second supply and demand tension according to the driver's order-acceptance vitality value; For example, this invention extracts currently considered driver information from the candidate order-driver set, including the driver's unique identifier, real-time location, and driver order-acceptance activity value. The driver order-acceptance activity value has been generated in step S205 based on the real-time vehicle speed to speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate, and is stored in the dispatch dataset. The second supply-demand tension refers to the supply-demand tension level of the grid area where the driver is currently located, and is obtained in a similar way to the first supply-demand tension: by locating the driver's grid area through real-time location, the corresponding tension label is read from the regional supply-demand tension mapping table.

[0048] S403. Obtain the order origin location information and the driver's real-time location based on the order information, and obtain the dynamic pick-up time of the order-driver based on the order origin location information, the driver's real-time location, and the dynamic road condition impedance coefficient of each road segment along the route. For example, this invention extracts the order origin coordinates from order information and the driver's real-time location coordinates from driver status data. Based on the road network topology, it plans the optimal driving path from the driver's real-time location to the order origin and divides the path into a continuous sequence of road segments. For each road segment, it obtains the dynamic road condition impedance coefficient corresponding to that segment from the dispatch dataset. This coefficient integrates real-time traffic events, traffic density, and inherent road properties. The dynamic pick-up time refers to the total time required for the driver to travel along the path to the order origin, which is obtained by dividing the length of each road segment by the product of the free-flow velocity of that segment and a factor minus the impedance coefficient, and then summing the results for each segment. This process is expressed by the formula: ; In the formula, Indicates the dynamic pick-up time. Indicates the first Length of the path segment Indicates the free flow velocity. It represents the dynamic road condition impedance coefficient, and the estimated pick-up time can reflect traffic events and congestion in real time, significantly improving accuracy compared to traditional estimates based on distance or historical average speed.

[0049] S404. Based on the passenger urgency index and the driver order-acceptance vitality value, obtain the comprehensive adjustment coefficient of urgency-vitality for the order-driver; For example, the present invention uses the passenger urgency index corresponding to the order. Driver order-taking vitality value corresponding to the driver The factors are then integrated to generate a comprehensive adjustment coefficient, which characterizes the synergistic effect of this pairing on demand urgency and supply positivity. Urgency-Potential Comprehensive Adjustment Coefficient Obtained through linear combination: ; in, and The preset weighting coefficients (weighting coefficients) and Statistical analysis of historical successfully matched orders revealed that, using passenger urgency, driver activity level, and final transaction details as samples, regression analysis was employed to determine the contribution weights of urgency and activity level to the matching success rate, thereby determining... and The numerical values ​​(indicating urgency and responsiveness) reflect the contribution of urgency and responsiveness to the final matching score. A higher coefficient indicates a more urgent need from the passenger and a more proactive attitude from the driver, meaning the pairing should receive a higher matching priority under similar conditions. By introducing this adjustment factor, the dispatch system can appropriately favor combinations of high urgency and high responsiveness, thereby improving user experience and driver satisfaction.

[0050] S405. Based on the dynamic pick-up time, urgency-vitality comprehensive adjustment coefficient, order base value, first supply-demand tension and second supply-demand tension, obtain the order-driver screening and matching score; For example, the present invention integrates the above-mentioned information from various dimensions and generates the final order-driver matching score through a nonlinear fusion model to obtain the supply-demand difference penalty coefficient. This was determined by analyzing the relationship between supply and demand imbalances and order cancellation rates in historical orders. Specifically, the average order cancellation rate was calculated for different absolute values ​​of supply and demand imbalances, and a penalty strength was fitted (generally between 0.1 and 0.3). A matching score was then used for filtering. The calculation formula is: ; in, Indicates the basic value of the order. Indicates the dynamic pick-up time. Represents the smoothing coefficient. This represents the urgency-vitality combined adjustment coefficient. and These are the first and second levels of supply and demand tension, respectively. This represents the penalty coefficient for supply and demand disparities. The formula achieves multi-objective fusion through the product of three factors: the first factor... This reflects the order value per unit of time, i.e., pick-up efficiency; the second factor It amplifies the advantages of combining high urgency with high vitality. The third factor This approach penalizes matches where there is a significant difference in supply and demand levels between the driver and the order's region, in order to maintain regional supply and demand balance. Through this integration, the matching score can simultaneously consider efficiency, value, user needs, driver willingness, and overall supply and demand, avoiding the one-sidedness of a single indicator.

[0051] As described in steps S401-S405 above, this invention performs a multi-dimensional, refined evaluation of each pair in the candidate order-driver set. This evaluation process uses dynamic pick-up time to reflect real road conditions, urgency-activity adjustment to reflect individual behavioral characteristics, and supply-demand tension differences to maintain regional balance. Finally, a comprehensive score is generated through non-linear fusion. This mechanism enables dispatch decisions to simultaneously optimize pick-up efficiency, user experience, driver response, and global supply and demand, significantly improving the accuracy and rationality of dispatch. The introduction of a supply-demand difference penalty coefficient allows the system to apply corresponding penalties to cross-regional dispatch pairs based on the supply-demand tension differences between the order's region and the driver's region, effectively suppressing capacity mismatch caused by local supply-demand imbalances and maintaining the stability of the global supply-demand structure.

[0052] In one embodiment, the step of generating the output dispatch result based on the filtering and matching score includes: S501. Obtain the filtering and matching score of each order-driver in the candidate order-driver set, and generate a set of matching scores to be dispatched; For example, the present invention extracts the screening and matching score corresponding to each order and driver from the candidate order-driver set output in step S40. The screening and matching score refers to the comprehensive value calculated in the previous step based on the dynamic pick-up time, urgency-activity comprehensive adjustment coefficient, order basic value, first supply and demand tension, and second supply and demand tension, which is used to quantify the matching quality of the order-driver combination. All orders-drivers and their scores are summarized to form a matching score set to be dispatched. This set is indexed by orders, and each order is associated with one or more candidate drivers and their scores, providing complete data input for subsequent sorting and conflict resolution.

[0053] S502. Based on the matching score set of the orders to be dispatched, sort the candidate drivers corresponding to each order in descending order of scores to obtain the driver matching priority for each order. For example, this invention sorts all candidate drivers associated with each order in the set of matching scores to be dispatched, from highest to lowest, to form a driver matching priority order for that order. The driver matching priority order refers to the order in which candidate drivers are considered for the same order, with the driver with the highest score listed first, indicating the highest match between that driver and the order under the current evaluation dimension. This sorting operation gives each order a clear sequence of selection preferences, providing a basis for subsequent pre-matching and conflict resolution. Through sorting, the dispatching system can quickly locate the optimal choice for each order in subsequent steps, and can also make choices based on scores when conflicts occur.

[0054] S503. According to the driver matching priority, select the driver ranked first for each order as the pre-matching driver, and detect whether the same driver is selected as the pre-matching driver by multiple orders at the same time. If it does not exist, the order will be directly linked to the driver; If such a driver exists, the driver is assigned the highest-scoring order based on the filtering and matching scores of the multiple orders corresponding to that driver. The other orders are then released back to the candidate order-driver set for rematching until each order is assigned a unique driver, thus generating the final dispatch matching pair. For example, the present invention first selects the driver ranked first as the pre-matching driver for each order based on the driver matching priority, forming a preliminary matching scheme. Then, the system iterates through all pre-matching drivers to detect if the same driver is selected for multiple orders simultaneously, i.e., a matching conflict occurs. If a conflict is detected, for all conflicting orders associated with that driver, the system compares the filtering matching scores of these orders with that driver, and binds the order with the highest score to that driver, forming a one-to-one final match; the remaining conflicting orders are released, i.e., their current pre-matching status is canceled, they return to the set of pending order matching scores, and the bound drivers are removed. Afterwards, the system repeats steps S502 to S503 for the released orders, i.e., reordering (at this time, the bound drivers have been removed from the candidate drivers) and performing pre-matching and conflict detection again, until all orders are successfully bound to a unique driver. This iterative process ensures that, globally, each driver serves at most one order, while prioritizing the highest-scoring pairing, achieving a fair and efficient matching result. Finally, once all orders are bound to a single driver, the final dispatch matching pairs are generated, which is the set of correspondences between each order and a single driver.

[0055] S504. Generate an output dispatch result based on the final dispatch matching pair, and push the dispatch instruction to the corresponding driver terminal. For example, this invention generates a structured output dispatch result based on the final dispatch matching pair. The output dispatch result includes detailed information for each order (such as order number, passenger information, origin and destination, etc.) and information of the assigned driver (such as driver ID, contact information, vehicle information, etc.). Subsequently, the system sends the dispatch instruction to the corresponding driver's terminal device (such as the driver's app) in real time via a message push service, notifying the driver to pick up the passenger. At the same time, the passenger will also receive a notification that the driver has accepted the order.

[0056] As described in steps S501-S504 above, this invention achieves global integration and conflict resolution of candidate matching results. This process, centered on filtering matching scores, employs sorting, pre-matching, conflict detection, and iterative re-matching to ensure both computational efficiency and matching fairness and global optimality. Ultimately, it outputs high-quality dispatch results, significantly improving the reliability of the dispatch system and user experience.

[0057] like Figure 2 As shown, the present invention also provides a large-scale intelligent dispatching system for ride-hailing services, comprising: The data acquisition module is used to acquire multimodal data on order dispatch. The dispatch dataset acquisition module is used to acquire the dispatch dataset based on the dispatch multimodal data, and to acquire the dynamic pick-up timeliness and regional supply and demand tension based on the dispatch dataset. The order-driver set acquisition module is used to obtain driver status data near the order location and generate candidate order-driver sets based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension. The matching score acquisition module is used to obtain the filtering matching score of each driver based on the candidate order-driver set; The dispatch result generation module is used to generate output dispatch results based on the filtering and matching scores.

[0058] like Figure 2 As shown, the present invention also provides a large-scale ride-hailing intelligent dispatch system, in which multiple modules are used to implement the steps of a large-scale ride-hailing intelligent dispatch method.

[0059] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a large-scale intelligent dispatching method for ride-hailing services.

[0060] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0061] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0062] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent results or equivalent process transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A large-scale intelligent dispatching method for ride-hailing services, characterized in that, include: Obtain multimodal data on order dispatch; The dispatch dataset is obtained based on the dispatch multimodal data, and the dynamic pick-up time and regional supply and demand tension are obtained based on the dispatch dataset. Obtain driver status data near the order location, and generate a candidate order-driver set based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension; The filtering and matching score for each driver is obtained based on the candidate order-driver set; The output dispatch result is generated based on the filtering and matching score.

2. The large-scale intelligent dispatching method for ride-hailing services according to claim 1, characterized in that, The step of obtaining the dispatch dataset based on the dispatch multimodal data includes: The location information of the order is obtained based on the dispatch multimodal data, wherein the location information includes the order's starting location information and the order's expected destination location information; Based on the order origin location information and the order pre-arrival destination location information, obtain the real-time traffic events, environmentally perceived traffic density and static road network topology of the order route, and obtain the dynamic road condition impedance coefficient based on the real-time traffic events, environmentally perceived traffic density and static road network topology of the order route; Based on the order dispatch multimodal data, the order inflow rate, idle driver density, and environmentally perceived traffic density are obtained; based on the order inflow rate, idle driver density, and environmentally perceived traffic density, the regional supply and demand tension is obtained. The passenger refresh frequency and input dwell time are obtained based on the dispatch multimodal data, and the passenger urgency index is obtained based on the passenger refresh frequency and input dwell time. Based on the dispatch multimodal data, the real-time speed-to-speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate of nearby idle drivers are obtained, and the driver order acceptance vitality value is obtained based on the real-time speed-to-speed limit ratio, frequency of rapid acceleration and deceleration, and historical order acceptance rate. The dynamic road condition impedance coefficient, regional supply and demand tension, passenger urgency index, and driver order acceptance vitality value are combined to form the order dispatch dataset.

3. The large-scale intelligent dispatching method for ride-hailing services according to claim 2, characterized in that, The step of obtaining the dynamic pick-up time based on the dispatch dataset includes: Extract the dynamic road condition impedance coefficient, passenger urgency index, and driver order-accepting vitality value from the dispatch dataset; The real-time traffic status of the road segments around the order starting point location information and the isochronous travel distance in each direction of the order starting point location information are obtained based on the dynamic road condition impedance coefficient. Input the passenger urgency index and driver order-accepting vitality value into the dynamic pick-up timeliness coefficient calculation formula to obtain the dynamic pick-up timeliness coefficient; The dynamic pick-up timeliness is obtained based on the passenger urgency index, driver order-accepting vitality value, and dynamic pick-up timeliness coefficient.

4. The large-scale intelligent dispatching method for ride-hailing services according to claim 2, characterized in that, The step of obtaining the regional supply and demand tension based on the dispatch dataset includes: The regional supply and demand tension gradient is obtained based on the order inflow speed, idle driver density, and environmentally perceived traffic density. The regional supply and demand tension is obtained based on the regional supply and demand tension gradient.

5. The large-scale intelligent dispatching method for ride-hailing services according to claim 1, characterized in that, The step of generating a candidate order-driver set based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension includes: The dynamic pick-up reachability domain of the order is obtained based on the dynamic pick-up time and the real-time road condition impedance coefficient around the order's starting point. Obtain the real-time locations of multiple available drivers, spatially match the real-time location of each available driver with the dynamic pick-up reach domain, and generate an initial candidate driver set; The regional supply and demand matching degree of candidate drivers is obtained based on the regional supply and demand tension and the initial candidate driver set. Drivers whose regional supply and demand match rate is higher than a preset matching threshold are selected to form a priority candidate driver set; Based on the driver order-acceptance vitality value in the driver status data, drivers with vitality values ​​higher than the vitality admission threshold are selected from the priority candidate driver set and combined with the corresponding orders to generate a candidate order-driver set.

6. The large-scale intelligent dispatching method for ride-hailing services according to claim 1, characterized in that, The step of obtaining the filtering and matching score for each driver based on the candidate order-driver set includes: Order information is obtained from the candidate order-driver set, and the first supply-demand tension is obtained based on the passenger urgency index and order information corresponding to the order. Based on the candidate order-driver set, obtain the corresponding driver order-acceptance vitality value, and based on the driver order-acceptance vitality value, obtain the second supply and demand tension. The order origin location information and driver real-time location are obtained based on the order information. The dynamic pick-up time of the order and driver is obtained based on the order origin location information, driver real-time location, and dynamic road condition impedance coefficient of each road segment. Based on the passenger urgency index and the driver order-acceptance vitality value, obtain the comprehensive adjustment coefficient of urgency-vitality for this order-driver; The order-driver matching score is obtained based on the dynamic pick-up time, urgency-vitality comprehensive adjustment coefficient, order base value, first supply-demand tension and second supply-demand tension.

7. The large-scale intelligent dispatching method for ride-hailing services according to claim 1, characterized in that, The step of generating the output dispatch result based on the filtering and matching score includes: Obtain the filtering and matching score of each order-driver in the candidate order-driver set, and generate a set of matching scores to be dispatched; Based on the matching score set of the orders to be dispatched, the candidate drivers corresponding to each order are sorted in descending order of score to obtain the driver matching priority for each order; Based on the driver matching priority, the driver ranked first for each order is selected as the pre-matched driver, and it is detected whether the same driver is selected as the pre-matched driver for multiple orders at the same time. If it does not exist, the order will be directly linked to the driver; If such a driver exists, the driver is assigned the highest-scoring order based on the filtering and matching scores of the multiple orders corresponding to that driver. The other orders are then released back to the candidate order-driver set for rematching until each order is assigned a unique driver, thus generating the final dispatch matching pair. The final dispatch matching pair is used to generate the output dispatch result, and the dispatch instruction is pushed to the corresponding driver terminal.

8. A large-scale intelligent dispatching system for ride-hailing services, characterized in that: include: The data acquisition module is used to acquire multimodal data on order dispatch. The dispatch dataset acquisition module is used to acquire the dispatch dataset based on the dispatch multimodal data, and to acquire the dynamic pick-up timeliness and regional supply and demand tension based on the dispatch dataset. The order-driver set acquisition module is used to obtain driver status data near the order location and generate candidate order-driver sets based on driver status data, dynamic pick-up timeliness, and regional supply and demand tension. The matching score acquisition module is used to obtain the filtering matching score of each driver based on the candidate order-driver set; The dispatch result generation module is used to generate output dispatch results based on the filtering and matching scores.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images