Crowdsourcing rider intelligent order sending method and system based on space-time collaboration

Abstract

The invention provides a crowdsourcing rider intelligent order sending method and system based on space-time collaboration, and the method comprises the steps: calculating a service capability score of each rider in a future time period and a specific geographic position, and a time aggregation value of the service capability score; aggregating the service capability scores of the plurality of riders in the target area, and constructing an area serviceability tensor of the target area; obtaining an order demand density tensor of the current target area, and constructing a scheduling priority tensor and a scheduling priority area set of each target area; obtaining a current order, constructing a scheduling score for each candidate rider, and evaluating the matching priority of the current candidate rider for executing the current order; and sorting the scheduling scores of all the candidate riders, selecting the rider with the highest score, and pushing the current order to the candidate rider client with the highest score through an order distribution system to complete task distribution. According to the method, the key problems of supply and demand mismatching, outbreak lagging, unstable supply, uncontrollable distribution and the like in a crowdsourcing scene are systematically solved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent order dispatching for crowdsourced riders, and particularly relates to a method and system for intelligent order dispatching for crowdsourced riders based on spatiotemporal collaboration. Background Technology

[0002] Urban same-city delivery is entering a period of peak growth, and crowdsourced riders have become a key force in ensuring timely delivery. Unlike fixed-staff riders, crowdsourced riders have a high degree of freedom in their online time, activity area, and order preferences. Platforms find it difficult to accurately regulate their behavior, which often leads to problems where traditional order dispatch strategies based on the shortest distance or shortest estimated arrival time seem to match orders reasonably but are difficult to implement. For example, orders may be rejected after being pushed, canceled after being accepted, or arrival times may be amplified by traffic congestion.

[0003] Existing methods typically simplify supply and demand modeling to static thermodynamics or short-term time-series forecasts, neglecting the hierarchical relationship from individual to regional levels. On the one hand, service availability at the individual level is not simply a matter of "online or offline" at a single moment, but rather a comprehensive capability influenced by historical frequency, response speed, and fulfillment stability. On the other hand, regional supply is not a simple summation of individual capabilities at a particular moment, but is affected by factors such as rider mobility, temporary aggregation, and spillover from border areas, exhibiting strong spatial continuity and time variability. More importantly, order surges are often triggered by a short-term acceleration in demand, rather than the absolute level of existing demand. If only the supply-demand ratio is used for ranking, scheduling often lags behind.

[0004] In addition, some areas have a seemingly sufficient supply, but it is supported by a few highly capable riders. Once these individuals move away or go offline, the capacity of the area will collapse instantly. Traditional models do not adequately depict this instability.

[0005] Ultimately, when the system moves from a regional perspective back to "person-order" matching, relying solely on distance or a single score fails to address two fundamental contradictions: first, high-stress areas need faster relief; second, the historical uncertainty of individual riders' responses needs effective constraint. This results in a disconnect between scheduling and actual delivery, a mismatch between resource allocation and order timeliness targets, and limited overall fulfillment efficiency. The industry urgently needs a systematic approach that starts with individual capability modeling, aggregates regional serviceability, prioritizes scheduling based on supply-demand discrepancies, and ultimately achieves controllable order dispatch within high-priority areas. This approach must balance spatiotemporal continuity, possess sensitivity and resilience to sudden growth and unstable supply, and be deployable using existing platform data and computing power. Summary of the Invention

[0006] The purpose of this invention is to design a crowdsourced rider intelligent order dispatch method and system based on spatiotemporal collaboration to solve the above-mentioned technical problems.

[0007] To achieve the above objectives, a first aspect of the present invention provides a crowdsourced rider intelligent order dispatch method based on spatiotemporal collaboration, the method comprising:

[0008] For each rider, based on at least one of their historical attendance frequency, order response time, and fulfillment stability, calculate their service capability score for future time periods and specific geographical locations, as well as their time aggregation value;

[0009] The service capability scores of multiple riders in the target area are aggregated and combined with the time aggregation value to construct the regional serviceability tensor of the target area.

[0010] Obtain the order demand density tensor of the current target region, and combine it with the regional serviceability tensor to construct the scheduling priority tensor of each target region and the set of scheduling priority regions;

[0011] The current order is obtained, and candidate riders are selected based on the set of priority scheduling regions, the scheduling priority tensor, and the service capability score of each rider. A scheduling score is constructed for each candidate rider, and the scheduling score evaluates the matching priority of the current candidate rider to execute the current order. The scheduling scores of all candidate riders are sorted, the rider with the highest score is selected, and the current order is pushed to the client of the candidate rider with the highest score through the order dispatch system to complete the task allocation.

[0012] Furthermore, the service capability score is obtained by weighted summation of the historical attendance frequency, order response time, and fulfillment stability.

[0013] Furthermore, the time aggregation value is used to measure the rider's overall dispatching ability level over a certain period of time.

[0014] Furthermore, the service capability scores of multiple riders in the target area are aggregated, and combined with the aggregated time value, a regional serviceability tensor for the target area is constructed, specifically as follows:

[0015] Obtain known urban grid division information from the platform;

[0016] Based on the known urban grid division information, the set of riders who are currently determined to be likely to provide services in the target area is obtained. The service capability score is combined with the scheduling elasticity factor to calculate the regional serviceability value in the target area in the future.

[0017] A neighborhood capability diffusion operation based on the neighborhood grid set centered on the target region is performed on the region serviceability value in the future time target region to generate a convolutionally smoothed region serviceability tensor.

[0018] Furthermore, the scheduling elasticity factor is calculated based on the service capability score and its time aggregation value, and is used to dynamically adjust the contribution of riders with high volatility or low stability to the regional capability.

[0019] Furthermore, the step of obtaining the order demand density tensor of the current target region, and combining it with the regional serviceability tensor to construct the scheduling priority tensor of each region and the set of scheduling priority regions, specifically involves:

[0020] Obtain a city map of the known target area to obtain the order demand density tensor;

[0021] By introducing an order surge inertia term and a service fluctuation correction term, and combining the order demand density tensor and the regional serviceability tensor, a structural priority function is constructed to generate a scheduling priority value for the target region at the current time, which is used to evaluate whether the target region needs to be prioritized for scheduling.

[0022] In actual operation, the platform refreshes the scheduling priority value every 5 minutes and selects the appropriate priority after sorting. Value is located in front % of the grid is used as the set of priority scheduling regions.

[0023] Furthermore, the order surge inertia term is calculated based on the order demand density tensor and is used to reflect the growth trend of orders within a short time window in the target area; the service fluctuation correction term is calculated based on the variance of rider service capability scores and is used to reflect the uncertainty of service capability in the area.

[0024] Furthermore, the obtained current order is used to filter candidate riders based on the set of scheduling priority regions, the scheduling priority tensor, and the service capability score of each rider, and a scheduling score is constructed for each candidate rider, specifically as follows:

[0025] Obtain the scheduling priority tensor and the set of scheduling priority regions, as well as the location and order time of all currently undispatched orders, the current location of all active riders, the path duration estimate returned by the map API service, and the historical order dispatch response behavior of each rider;

[0026] For all currently undispatched orders in the aforementioned priority dispatch area set, a candidate rider set is selected;

[0027] For each candidate rider in the candidate rider set, a scheduling score function is constructed to generate a scheduling score.

[0028] Furthermore, the criteria for selecting the candidate rider set must simultaneously meet the following conditions: the rider can arrive within the response time set by the platform, the rider has the minimum service capacity in the area at that time, and the current order is ranked high in the scheduling priority. The area;

[0029] The scheduling scoring function includes service capability scoring, regional scheduling priority tensor, path duration estimation returned by map API service, and generation of regional supply and demand pressure term and rider behavior confidence suppression term;

[0030] The regional supply and demand pressure term represents the deviation between the current regional order density and service capacity, and is used to proactively adjust matching priorities by scheduling local supply and demand tensions; the rider behavior confidence inhibition term is used to penalize individuals with large fluctuations in response behavior.

[0031] A second aspect of the present invention provides a crowdsourced rider intelligent dispatch system based on spatiotemporal collaboration, the system comprising:

[0032] The rider capability field modeling unit is used to calculate, for each rider, a service capability score and its time aggregation value for a future time period and a specific geographical location based on at least one of their historical attendance frequency, order response time and fulfillment stability.

[0033] The regional serviceability aggregation unit is used to aggregate the service capability scores of multiple riders in the target area and combine them with the time aggregation value to construct the regional serviceability tensor of the target area.

[0034] The scheduling priority determination unit is used to obtain the order demand density tensor of the current target area, and combine it with the regional serviceability tensor to construct the scheduling priority tensor of each target area and the set of scheduling priority areas.

[0035] The rider-order matching unit is used to obtain the current order, filter candidate riders based on the scheduling priority area set, scheduling priority tensor, and service capability score of each rider, and construct a scheduling score for each candidate rider. The scheduling score evaluates the matching priority of the current candidate rider to execute the current order; sort the scheduling scores of all candidate riders, select the one with the highest score, and push the current order to the client of the candidate rider with the highest score through the order dispatch system to complete the task allocation.

[0036] The beneficial technical effects of the present invention are at least as follows:

[0037] This invention constructs a complete bottom-up technical chain to address the spatiotemporal collaborative order dispatching challenges of crowdsourced riders. The invention's key features are the organic coupling of four levels: First, at the individual level, it proposes capability field modeling oriented towards dispatch feasibility. This model uses rider historical frequency of appearance, response latency, and fulfillment stability as core elements to form a spatiotemporally continuous capability representation, rather than relying on single-time intentions or temporary online status as the basis for judgment, thus improving the interpretability and reliability of individual dispatchability from the source. Second, at the regional level, it proposes a method for constructing a serviceability tensor. By aggregating individual capabilities, adding an elastic inhibition factor, and performing neighborhood diffusion, regional supply can reflect true intensity while possessing spatial continuity and robustness against sporadic supply, providing a stable basis for subsequent regional decision-making. The system first defines the available supply; then, at the strategy level, it proposes a scheduling priority function for explosive scenarios and unstable supply, using the demand-to-supply ratio as the basis, while introducing a demand surge trend term and a supply fluctuation correction term. This allows the system to identify areas of accelerated growth early and downweight fragile supply supported by a few individuals, thus aligning regional ranking with actual scheduling risks. Finally, at the execution level, it proposes an individual matching mechanism within high-priority regions, incorporating individual capabilities, regional priority, path cost, and local supply and demand pressure into the same scoring framework, and superimposing a behavioral confidence suppression term based on historical order acceptance, cancellation, and no-response statistics. This ensures that the final delivery simultaneously meets the dual goals of alleviating high-pressure areas and ensuring controllable fulfillment. These four layers of technology are interconnected in terms of data and computation, based on the platform's existing order and trajectory logs, location, and path services. They do not rely on external sensitive data or highly complex training processes, and can operate stably within the real-time scheduling cycle. This achieves closed-loop optimization from individuals to regions and back to individuals, systematically solving key problems such as supply-demand mismatch, explosive delays, unstable supply, and uncontrollable delivery in crowdsourcing scenarios. Attached Figure Description

[0038] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0039] Figure 1 This is a flowchart of the crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration of the present invention. Detailed Implementation

[0040] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0041] In one or more embodiments, such as Figure 1 As shown, a crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration is disclosed, the method including the following:

[0042] S1. For each rider, calculate their service capability score for future time periods and specific geographical locations, and its time aggregation value, based on at least one of their historical attendance frequency, order response time, and fulfillment stability.

[0043] Specifically, this step aims to construct the distribution of service capabilities of crowdsourced riders at a specific geographical location within a future timeframe, forming a capability scoring tensor with temporal and spatial components. This tensor characterizes the platform's rider dispatchability under uncertain supply conditions. This capability is not an abstract prediction of willingness, but rather integrates collectable indicators such as historical attendance records, response behavior, and fulfillment stability, reflecting the platform's comprehensive judgment from an operational perspective on whether a rider is suitable for undertaking tasks in a particular area. The final output capability field will serve as direct input for subsequent regional service aggregation modeling. Required data includes: a sequence of order acceptance records for each rider on the platform over the past 30 days, with data fields including order timestamp, task location, completion status, response time, etc., and each record containing a rider identifier; and the rider's real-time location. Uploaded via the client's GPS module, with an update cycle of 30 seconds; current system time. Dispatch is centrally managed by the dispatch center. The city's delivery area is discretely divided into fixed grids. Each coordinate is assigned to the nearest grid number for discretization.

[0044] The construction of the ability rating tensor depends on the following formula:

[0045]

[0046] in, Indicates the first One rider in the future In the grid Service capability score; The normalization function ensures that the score range is normalized to... Within the range, improve the comparability of riders across different riders; This indicates the frequency of orders received by the rider in this region and time period in history, derived from the projection results of their 30-day order-receiving trajectory, smoothed using a Gaussian kernel. This indicates the rider's historical average response time, expressed in order acceptance response time. The data comes from platform logs and has been normalized. This indicates the rider's historical fulfillment rate in this area and time period, derived from the calculation results of task completion markers and timeliness conditions; The weight parameters configured for the system are used to adjust the importance of each behavioral feature. Typical values ​​are 0.4, 0.3, and 0.3. These are set by business operations personnel and do not participate in model training.

[0047] For example, if a rider has 15 order-taking records in a business district between 11:30 and 11:35 AM, with an average response time of 4 seconds and a fulfillment rate of 95%, these values, normalized to 0.88, 0.92, and 0.95 respectively, are weighted as follows: , , The score is then calculated as follows:

[0048]

[0049] In addition, for the convenience of subsequent service aggregation modeling calculations, a global total capacity needs to be output to measure the overall dispatching capacity level of riders in a certain period of time. The expression is as follows:

[0050]

[0051] in The time aggregation value represents the rider's... In time The total value of city-level service capabilities at any given moment serves as the input basis for subsequent regional capability accumulation.

[0052] S2. Aggregate the service capability scores of multiple riders in the target area, and combine them with the time aggregation value to construct the regional serviceability tensor of the target area.

[0053] Specifically, the input data for this step comes from the previous step: the service capability score tensor for each rider in the future spatiotemporal region. (Based on weighted normalized statistics of historical attendance records, response time, and contract completion rate) and its total capability value across the entire city. (Used to characterize the overall activity level of riders). In addition, the platform is aware of urban grid division information. .

[0054] To avoid the high volatility caused by simply aggregating the abilities of all riders using traditional methods, this step introduces a "regional serviceability" calculation formula, combining the aggregation of individual abilities with a regional dispatch elasticity factor:

[0055]

[0056] in, Indicates future time Grid The serviceability value of the area; Indicates time Determined to be possible A group of riders providing services; It is the scheduling elasticity coefficient set by the platform, used to control the degree of influence of highly volatile riders on the regional ability value; This is a tiny offset to prevent the denominator from becoming zero. The innovation of this formula lies in the fact that it not only superimposes each rider's ability value for that area and time period, but also... It also introduced a "scheduling elasticity factor" item. This is used to dynamically adjust the contribution of riders with high volatility or low stability to regional capabilities. For example, when a rider's current regional capability deviates too much from their global capability (meaning they only appear occasionally or are unpredictable in that region), their weight in the total regional capability value will automatically decrease, thus ensuring that the regional capability tensor is more stable and more usable for scheduling decisions.

[0057] Furthermore, to further enhance the spatial continuity and practical usability of the region tensor, this step also... Based on this, a "neighborhood capability diffusion" operation is introduced to smoothly diffuse the capability value of each grid according to the supply status of neighboring grids, thereby simulating the accessibility of riders actually moving between areas:

[0058]

[0059] in, It is the region serviceability tensor after convolution and smoothing, which is used for subsequent supply and demand contrast calculation; Therefore The set of neighborhood grids centered on the platform (the platform defaults to this). window); It is a preset two-dimensional Gaussian kernel matrix, representing the influence weight of each neighboring grid in space on that location; This represents the original serviceability value of the region at the neighboring grid. The mapping representing the serviceability tensor of the original region. Indicates the horizontal index of the grid. Indicates the relative horizontal offset index. Indicates the grid vertical index, Indicates the relative vertical offset index. Indicates a time index.

[0060] The innovation of this convolutional smoothing lies in the fact that it doesn't simply perform spatial averaging, but instead uses a Gaussian kernel to simulate the real reachability probability of riders spreading from the central grid to the surrounding areas, balancing the continuity of supply capacity within the area with the buffering effect of edge transitions. For example, when rider density suddenly drops at the edge of a business district, the capacity value of the area after convolution will not show an unnatural cliff, but will form a smooth transition, which is closer to the real dispatching scenario.

[0061] For example, during the 11:30-11:35 AM time slot, there are 10 riders in business district A with ability values ​​of 0.85, 0.79, 0.77, etc. However, two of these riders had city-wide ability values ​​of only 0.15 and 0.18 over the past hour, indicating that they were only in this area by chance. Under traditional overlay methods, the regional ability of business district A would be overestimated. In this patented formula, the scheduling elasticity factor automatically reduces the contribution of these two riders, resulting in a more accurate final assessment. More realistic and reliable. Then, neighborhood convolution pairs are used. After smoothing, the capacity value in the boundary area naturally decreases, forming a pattern that more closely resembles the actual supply structure. .

[0062] This step outputs two variables: the first is... The first is the original regional service capacity tensor after adjustment by the scheduling elasticity factor; the second is... That is, a smoother and more realistic regional serviceability tensor after spatial diffusion.

[0063] S3. Obtain the order demand density tensor of the current target area, and combine it with the regional serviceability tensor to construct the scheduling priority tensor of each target area and the set of scheduling priority areas.

[0064] Specifically, the core objective of this step is to construct a "scheduling priority tensor" with decision-making guidance value within the city's dynamic supply and demand structure. This tensor serves as the core input for the patent system before dispatching tasks, ultimately achieving a quantifiable ranking of task scheduling urgency across different regions. This step directly relies on two types of regional data output from the previous stage: one is the service capacity tensor. First, it already considers neighborhood diffusion and smoothed distribution of serviceable capacity; second, it considers the order demand density tensor. These two tensors record the order concentration within a unit grid per unit time. They characterize the city's operational status from the supply and demand sides, respectively, and form the basis for establishing the regional scheduling priority function. Furthermore, to more accurately capture sudden order surges, local service imbalances, and structural redundancy issues in crowdsourcing scenarios, this step designs two innovative scheduling intervention factors: an order surge inertia term and a service fluctuation correction term. These two factors combined constitute the innovative core of the entire priority function.

[0065] Wherein, the order demand density tensor The calculation is as follows:

[0066] The city map is divided into a regular grid system (e.g., 100m x 100m per grid). Each grid has unique coordinates (x, y). Simultaneously, time is divided into continuous windows (e.g., every 5 minutes is a time period t). The system backtracks historical order data over a period of time (e.g., the past 30 days) to calculate the average number of historical orders occurring within the "same time period" and "same grid." Within the current scheduling cycle (e.g., the current 5 minutes), the system continuously calculates the number of newly generated, unassigned orders in each grid, using this as the order demand density tensor. The order demand density tensor Typically, it involves combining historical baseline forecasts with real-time order statistics to reflect demand "that is happening right now," rather than purely historical patterns.

[0067] Furthermore, in actual urban delivery processes, common dispatch systems only use the supply-demand ratio. As a scheduling and prioritization indicator, where D represents real-time demand intensity and S represents supply capacity intensity, although logically simple, it easily overlooks two key issues: firstly, the time lag in order demand changes (e.g., overlooking a region's rapid order growth before backlog); and secondly, the instability of service capacity (e.g., a region with high service capacity but composed of unstable riders, lacking sustained support). Therefore, in this step, the patented solution proposes for the first time the following structural priority function:

[0068]

[0069] in Represents city grid In time The scheduling priority value is used to assess whether the region needs to be prioritized for scheduling. It must be a very small positive number to prevent division by zero; and These are the adjustment coefficients for the two structural items, which the platform can set based on historical scheduling experience, such as... , .

[0070] This is the order surge inertia term, which depicts the growth trend of orders within a short time window in a region. Its construction method is as follows:

[0071]

[0072] in The short-term prediction window length (e.g., 5 minutes) set for the platform is essentially a first-order approximate derivative, describing the order growth rate per unit time. Cut off the negative growth portion to prevent mis-scheduling behavior caused by the scheduling system responding to a decline in orders. Indicates time The required grid density within the area.

[0073] The significance of this design lies in the fact that in crowdsourcing scenarios, order surges are highly unpredictable. If the platform only determines priority based on the current order volume, scheduling delays are very likely to occur. By adding... The system can detect "areas with rapidly increasing orders" in advance and prioritize rider allocation.

[0074] The service fluctuation correction term, used to reflect the uncertainty of service capacity in the region, is defined as the variance of the ability values ​​of each active rider in the region:

[0075]

[0076] in Is currently in The region contains a set of riders with recorded ability values, which can be constructed in the first step. Here, 'i' represents the rider index. This term is used as a deduction in the formula, and its innovative significance lies in the fact that riders exhibit significant instability in the crowdsourcing model. Even if the total service capacity of a certain area is high, it may be contributed by a very small number of riders, forming a "concentrated dependency" problem, leading to insufficient confidence in scheduling that area by the system. Higher variance indicates a more concentrated and volatile service capacity, and the system should lower the scheduling priority for that area. For example, if the service capacity of a certain area... However, if all the contributions are made by a single high-ability rider, while the remaining riders have an ability of 0, then... Extremely large, leading to This effectively suppresses the risk of the system misjudging service stability due to high overall volume. Overall, this priority function has triple scheduling intervention capabilities: basic supply-demand ratio determination, order surge trend detection, and service instability penalty. In actual operation, the platform refreshes the priority function every 5 minutes. And select after sorting Value is located in front % of the grid is designated as a high-priority order dispatch area. It is used to limit the rider dispatch range, strengthen the centralized task push, and improve the rider response rate.

[0077] This step outputs two variables: That is, a scheduling priority tensor with structure item modifications; This refers to the priority scheduling area set, which is used for the next step of matching orders with riders.

[0078] S4. For the current order obtained, candidate riders are selected based on the set of priority scheduling regions, the scheduling priority tensor, and the service capability score of each rider. A scheduling score is constructed for each candidate rider. The scheduling score evaluates the matching priority of the current candidate rider in executing the current order. The scheduling scores of all candidate riders are sorted, the rider with the highest score is selected, and the current order is pushed to the client of the candidate rider with the highest score through the order dispatch system to complete the task allocation.

[0079] Specifically, this step is the core stage in the entire patent dispatching system for completing the final "person-order" matching decision. Its significance lies in integrating the macro-dispatch indicators (i.e., regional supply capacity) constructed in the first three steps. Order density Regional scheduling priority This process transforms into specific micro-level dispatch actions. This stage must balance the overall dispatch from the platform's perspective (e.g., alleviating structural imbalances in supply and demand) with the feasibility from the individual rider's perspective (e.g., path cost, response preferences, and order rejection tendencies). Therefore, this step proposes a "multi-factor coupled order dispatch scoring function" for crowdsourcing scenarios with high degrees of freedom. This function not only integrates variables such as space, time, service capacity, path cost, and regional priority, but also introduces two innovative control factors with real-world business backgrounds: a regional supply and demand pressure weighting term and an individual behavior confidence inhibition term, which constitute the final rider dispatch score.

[0080] This step directly uses the scheduling priority region set output in the previous step. and regional scheduling priority tensor At the same time, combined with the rider capability tensor in the first step In addition, in actual operation, the platform also relies on the following real-time system data:

[0081] All currently undelivered orders Location With order time It is uploaded by the user's order system;

[0082] Current location of all active riders The location is automatically uploaded by the App's GPS positioning module every 30 seconds;

[0083] Path duration estimate returned by map API service This is obtained by calling the real-time path interface of a module such as the AppGPS location module;

[0084] The platform's order-taking behavior database contains statistics on each rider's historical order dispatch response behavior (order acceptance rate, cancellation rate, and no response rate).

[0085] First, the platform only targets those who fall into the category. orders Initiate fine-grained scheduling and matching to filter the set of candidate riders. The judgment condition is that all three of the following conditions are met simultaneously:

[0086] The rider can arrive within the response time set by the platform, for example... minute;

[0087] The rider has the minimum service capacity in this area at this time (e.g.) );

[0088] Region: Orders ranked at the top of the scheduling priority list The area.

[0089] Then for each candidate rider Construct a scheduling scoring function, defined as follows:

[0090]

[0091] in, This represents extremely small positive numbers, used to prevent the denominator from being zero and to maintain computational stability. represents the weight coefficient of the behavior confidence suppression term, and j represents the order index; This represents the confidence suppression term for rider behavior;

[0092] The scoring function Used to assess riders Execute orders The higher the matching priority, the stronger the willingness to assign orders, and the more likely the system is to assign orders to that rider.

[0093] in: It is the capability tensor value constructed in the first step, representing the rider's service adaptability at that spatiotemporal location; It is the regional scheduling priority score in the third step, which reflects the current urgency of supply and demand in the region; It is the estimated time for the shortest path from the rider's current location to the order's location at the current moment; It is a regional supply and demand pressure term, defined as the deviation between the current regional order density and service capacity:

[0094]

[0095] in, It represents extremely small positive numbers, used to prevent the denominator from being zero and to maintain computational stability;

[0096] Its meaning is: if the order density of a certain area Significantly higher than service capacity If the pressure value is greater than zero, it indicates that the region is in dire need of supply and requires incentives to improve the score; conversely, if the supply is sufficient, the pressure value is negative, which naturally weakens the score. The pressure adjustment coefficient (e.g., 0.2~0.5) can be adjusted on the platform. This factor combines the output variables of the first two steps and is the core innovation of the patent in "actively adjusting the matching priority of local supply and demand tension".

[0097] This is a rider behavior confidence suppression term, designed to punish individuals with large fluctuations in response behavior, specifically defined as:

[0098]

[0099] in , , These are the rider's order acceptance rate, cancellation rate, and no-response rate over the past 72 hours, derived from the platform's order dispatch logs; Weighting parameters for behavioral risk (e.g., set to) The innovation lies in incorporating the rider's historical response performance into the matching scoring function in a structured manner, rather than using it as an auxiliary filter, only applying it when... Orders are only allowed to be dispatched when the level is high enough, thus the platform’s ability to control performance risks is demonstrated by the fine-grained penalty for unstable behavior through the subtraction of items in the chain.

[0100] Ultimately, the platform selected riders based on all candidate riders' scores. Sort the results and select the highest scorer. And through the order dispatch system, the orders are distributed. Push to rider The client completes task assignment.

[0101] For example: An order is located in business district A, currently... , , Rider A in the area Estimated path time Minutes, behavioral indicators are , , ,but , ,set up , :

[0102]

[0103] The rider with the highest score among multiple candidate riders will be the one assigned the order.

[0104] This step outputs two variables: That is, the order dispatch results, and The score matrix is ​​used for platform scheduling log recording and subsequent model iteration optimization.

[0105] In one or more embodiments, a crowdsourced rider intelligent dispatch system based on spatiotemporal collaboration is disclosed, the system comprising:

[0106] The rider capability field modeling unit is used to calculate, for each rider, a service capability score and its time aggregation value for a future time period and a specific geographical location based on at least one of their historical attendance frequency, order response time and fulfillment stability.

[0107] The regional serviceability aggregation unit is used to aggregate the service capability scores of multiple riders in the target area and combine them with the time aggregation value to construct the regional serviceability tensor of the target area.

[0108] The scheduling priority determination unit is used to obtain the order demand density tensor of the current target area, and combine it with the regional serviceability tensor to construct the scheduling priority tensor of each target area and the set of scheduling priority areas.

[0109] The rider-order matching unit is used to obtain the current order, filter candidate riders based on the scheduling priority area set, scheduling priority tensor, and service capability score of each rider, and construct a scheduling score for each candidate rider. The scheduling score evaluates the matching priority of the current candidate rider to execute the current order; sort the scheduling scores of all candidate riders, select the one with the highest score, and push the current order to the client of the candidate rider with the highest score through the order dispatch system to complete the task allocation.

[0110] It is worth noting that the specific workflow of the crowdsourced rider intelligent dispatch system based on spatiotemporal collaboration provided in this embodiment of the invention is the same as that of the crowdsourced rider intelligent dispatch method based on spatiotemporal collaboration described in the above embodiments, and will not be repeated here.

[0111] This invention also provides a crowdsourced rider intelligent dispatching device based on spatiotemporal collaboration, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the above embodiments of the crowdsourced rider intelligent dispatching method based on spatiotemporal collaboration, for example... Figure 1 The steps S1 to S4 described above; or, when the processor executes the computer program, it implements the functions of each module in the above system embodiments.

[0112] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the spatiotemporal collaborative crowdsourced rider intelligent dispatching device.

[0113] The spatiotemporal collaborative crowdsourcing rider intelligent dispatching device can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the spatiotemporal collaborative crowdsourcing rider intelligent dispatching device may also include input / output devices, network access devices, buses, etc.

[0114] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the spatiotemporal collaborative crowdsourcing rider intelligent dispatching device, connecting all parts of the device via various interfaces and lines.

[0115] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the spatiotemporal collaborative crowdsourcing rider intelligent dispatching device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the operation of the air conditioner controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart memory card (SmartMedaCard, SMC), secure digital card (SecureDagatal, SD), flash memory card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0116] The modules integrated into the spatiotemporal collaborative crowdsourced rider intelligent dispatching device, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0117] 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 program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0118] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A crowdsourced rider intelligent order dispatch method based on spatiotemporal collaboration, characterized in that, The method includes: For each rider, based on at least one of their historical attendance frequency, order response time, and fulfillment stability, calculate their service capability score for future time periods and specific geographical locations, as well as their time aggregation value; The service capability scores of multiple riders in the target area are aggregated and combined with the time aggregation value to construct the regional serviceability tensor of the target area. Obtain the order demand density tensor of the current target region, and combine it with the regional serviceability tensor to construct the scheduling priority tensor of each target region and the set of scheduling priority regions; The current order is obtained, and candidate riders are selected based on the set of priority scheduling regions, the scheduling priority tensor, and the service capability score of each rider. A scheduling score is constructed for each candidate rider, and the scheduling score evaluates the matching priority of the current candidate rider to execute the current order. The scheduling scores of all candidate riders are sorted, the rider with the highest score is selected, and the current order is pushed to the client of the candidate rider with the highest score through the order dispatch system to complete the task allocation.

2. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 1, characterized in that, The service capability score is obtained by weighted summation of the historical attendance frequency, order response time, and fulfillment stability.

3. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 1, characterized in that, The time aggregation value is used to measure the rider's overall dispatching ability level over a certain period of time.

4. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 1, characterized in that, The service capability scores of multiple riders in the target area are aggregated, and combined with the aggregated time value, a regional serviceability tensor for the target area is constructed, specifically as follows: Obtain known urban grid division information from the platform; Based on the known urban grid division information, the set of riders who are currently determined to be likely to provide services in the target area is obtained. The service capability score is combined with the scheduling elasticity factor to calculate the regional serviceability value in the target area in the future. A neighborhood capability diffusion operation based on the neighborhood grid set centered on the target region is performed on the region serviceability value in the future time target region to generate a convolutionally smoothed region serviceability tensor.

5. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 4, characterized in that, The scheduling elasticity factor is calculated based on the service capability score and its time aggregation value, and is used to dynamically adjust the contribution of riders with high volatility or low stability to the regional capability.

6. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 1, characterized in that, The process of obtaining the order demand density tensor of the current target region, combining it with the regional serviceability tensor, and constructing the scheduling priority tensor of each region and the set of scheduling priority regions is as follows: Obtain a city map of the known target area to obtain the order demand density tensor; By introducing an order surge inertia term and a service fluctuation correction term, and combining the order demand density tensor and the regional serviceability tensor, a structural priority function is constructed to generate a scheduling priority value for the target region at the current time, which is used to evaluate whether the target region needs to be prioritized for scheduling. In actual operation, the platform refreshes the scheduling priority value every 5 minutes and selects the appropriate priority after sorting. Value is located in front % of the grid is used as the set of priority scheduling regions.

7. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 6, characterized in that, The order surge inertia term is calculated based on the order demand density tensor and is used to reflect the growth trend of orders in the target area within a short time window; the service fluctuation correction term is calculated based on the variance of rider service capability scores and is used to reflect the uncertainty of service capability in the area.

8. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 1, characterized in that, The obtained current order is used to filter candidate riders based on the set of scheduling priority regions, the scheduling priority tensor, and the service capability score of each rider, and a scheduling score is constructed for each candidate rider, specifically as follows: Obtain the scheduling priority tensor and the set of scheduling priority regions, as well as the location and order time of all currently undispatched orders, the current location of all active riders, the path duration estimate returned by the map API service, and the historical order dispatch response behavior of each rider; For all currently undispatched orders in the aforementioned priority dispatch area set, a candidate rider set is selected; For each candidate rider in the candidate rider set, a scheduling score function is constructed to generate a scheduling score.

9. The crowdsourced rider intelligent order dispatching method based on spatiotemporal collaboration according to claim 8, characterized in that, The criteria for selecting candidate riders must simultaneously meet the following conditions: the rider can arrive within the response time set by the platform, the rider has the minimum service capacity in the area at that time, and the current order is among the top in the scheduling priority ranking. The area; The scheduling scoring function includes service capability scoring, regional scheduling priority tensor, path duration estimation returned by map API service, and generation of regional supply and demand pressure term and rider behavior confidence suppression term; The regional supply and demand pressure term represents the deviation between the current regional order density and service capacity, and is used to proactively adjust matching priorities by scheduling local supply and demand tensions; the rider behavior confidence inhibition term is used to penalize individuals with large fluctuations in response behavior.

10. A crowdsourced rider intelligent order dispatch system based on spatiotemporal collaboration, characterized in that: The system includes: The rider capability field modeling unit is used to calculate, for each rider, a service capability score and its time aggregation value for a future time period and a specific geographical location based on at least one of their historical attendance frequency, order response time and fulfillment stability. The regional serviceability aggregation unit is used to aggregate the service capability scores of multiple riders in the target area and combine them with the time aggregation value to construct the regional serviceability tensor of the target area. The scheduling priority determination unit is used to obtain the order demand density tensor of the current target area, and combine it with the regional serviceability tensor to construct the scheduling priority tensor of each target area and the set of scheduling priority areas. The rider-order matching unit is used to obtain the current order, filter candidate riders based on the scheduling priority area set, scheduling priority tensor, and service capability score of each rider, and construct a scheduling score for each candidate rider. The scheduling score evaluates the matching priority of the current candidate rider to execute the current order; sort the scheduling scores of all candidate riders, select the one with the highest score, and push the current order to the client of the candidate rider with the highest score through the order dispatch system to complete the task allocation.

Images