Intelligent matching and load balancing dispatching system for short-distance orders of cruising taxis

Abstract

The application discloses a short-distance order intelligent matching and load balancing scheduling system for cruising taxis, and particularly relates to the technical field of intelligent traffic management, and the system comprises a topology evidence module, a queue gating module, a batch matching module and a closed-loop processing module; wherein the topology evidence module is used for unified access, time benchmark alignment and site topology modeling of multi-source data of key sites and surrounding roads to generate a sequence evidence chain of vehicles; the queue gating module is used for reconstructing real-time sequences of vehicle queue areas, forming a queue sequence list, and generating a sequence evidence credibility and a load gating quantity of each vehicle; the batch matching module is used for completing short-distance order identification, release batch generation, sequence layer range determination, vehicle and order matching and passenger pickup window allocation within a rolling scheduling window; and the closed-loop processing module is used for executing issuing, process monitoring, abnormality processing, rollback adjustment and audit solidification.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic management technology, and more specifically, to a system for intelligent matching and load balancing of short-distance taxi orders. Background Technology

[0002] Taxis still undertake a large number of on-demand transportation tasks in key areas such as train stations, bus stations, hospital entrances, and commercial district entrances. Especially in scenarios where short-distance orders are concentrated, stations typically adopt a one-way management method from the waiting area to the passenger area. On-site, it is required that vehicles enter in an orderly manner while balancing traffic flow efficiency and passenger dispersal efficiency. Existing taxi dispatching methods mostly rely on vehicle positioning, fence entry and exit status, order information, or human experience as the main basis, which can complete basic vehicle allocation and order matching. However, they still lack sufficient consideration for the constraints of queuing order within key stations, the priority rules for short-distance vehicles, and the linkage between changes in the load of surrounding roads. Therefore, during peak hours, temporary traffic control, or when traffic is dense and slow, inconsistencies can easily arise between dispatching results and on-site order requirements.

[0003] Existing technologies have the following shortcomings: In the process of dispatching short-distance orders at key stations, some solutions typically sort waiting vehicles based on the number of vehicles within the electronic fence, vehicle arrival time, or the order of vehicle terminal trajectories, and then implement short-distance priority release or order dispatch accordingly. This approach is feasible in general scenarios, but in actual operation, waiting areas and vehicle storage areas are often mixed with private vehicles, ride-hailing vehicles, or vehicles temporarily dropping off passengers. Furthermore, factors such as bus obstruction, nighttime backlighting, rain glare, damaged license plates, low-speed movement, and vehicle terminal outages can affect the system, leading to gaps, breaks, or misorders in the data from checkpoint recognition, video events, and vehicle trajectories. In such cases, if the system still treats vehicles located within the fence or roughly their arrival order as queuing facts, it is difficult to reliably reconstruct the true queue order. This can easily lead to misjudging vehicles that cut in line as those at the front of the queue, causing the short-distance priority mechanism to evolve into cross-order release, resulting in delayed orderly vehicles, increased on-site disputes, overflow of waiting areas, and deterioration of road order around the station. Existing schemes generally lack a mechanism for recording the basis for release, the order of priority, and the version of the rules. It is difficult to explain afterward why a certain vehicle was given priority or why it was assigned to a specific pick-up window, which is not conducive to the traffic police department's on-site management and review of key stations.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a smart matching and load balancing scheduling system for short-distance taxi orders to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The intelligent matching and load balancing scheduling system for short-distance taxi orders includes a topology evidence module, a queue gating module, a batch matching module, and a closed-loop processing module.

[0008] The topology evidence module is used to uniformly access, standardize field definitions, verify integrity and align time bases for multi-source data within the scope of key stations and surrounding roads, and generate vehicle sequence evidence chains by combining station topology and traffic organization version.

[0009] The queue gating module is used to reconstruct the real-time queue position of the vehicle holding area based on the sequence evidence chain, the result of the missing evidence segment marking, the site topology and the traffic organization version, form a queue position list, and generate the confidence level of the sequence evidence and the load gating level for each vehicle.

[0010] The batch matching module is used to complete short-distance order identification, release batch generation, order layer range determination, order and vehicle matching within the order layer, passenger boarding window allocation, and load balancing scheduling inside and outside the station within the scrolling scheduling window, based on the queue order list, the credibility of each vehicle's order evidence and its threshold judgment result, the load control quantity, and the station-related order data.

[0011] The closed-loop processing module is used to issue and execute, monitor, handle anomalies, adjust and audit information on release batches, sequence layer ranges, short-distance orders and vehicle matching results, target passenger pick-up windows, entry time limits and rebalancing guidance information in areas outside the station, and based on the execution feedback, correct the credibility of the sequence evidence and the load gating amount of each vehicle in the subsequent rolling dispatch window.

[0012] In a preferred embodiment, the data accessed by the topological evidence module includes at least the location data and vehicle operation status data reported by the taxi vehicle terminal, the passenger status data and business status data returned by the meter or operating terminal, the license plate passing events generated by the station entrance and exit checkpoint or video structured recognition, the entry and exit records generated by the station gate, the traffic signal release status and temporary traffic organization information at the traffic police side intersection, and the traffic speed monitoring data or queue length monitoring data of the road sections around the station.

[0013] In a preferred embodiment, the topology evidence module establishes a unified time reference, records the arrival time and original timestamp of data packets from each source, and retains the alignment residual after time alignment; the topology evidence module also performs solidified modeling of site spatial elements to form a site topology, which includes at least a parking area electronic fence, a passenger pick-up area electronic fence, a set of passenger pick-up windows and a set of key sections, and binds it to the traffic organization version so that the sequence determination corresponds to the currently effective on-site organization rules.

[0014] In a preferred embodiment, the topological evidence module associates multi-source events with the vehicle identifier as the primary key to generate a vehicle sequence evidence chain. The sequence evidence chain includes at least the events of entering the parking area, reaching the stop line or gate, entering the boarding area, entering the boarding window, and leaving the boarding area. For events missing due to obstruction, recognition failure, license plate damage, or terminal disconnection, the topological evidence module retains the gaps and marks them as evidence missing segments, while recording the start and end times and the missing type of the evidence missing segments.

[0015] In a preferred embodiment, the queue gating module uses key cross-section events as anchor points and station topology as constraints to organize the sequence evidence chain for each vehicle, extracting events such as entering the holding area, reaching the stop line or gate, entering the boarding area, and entering the boarding window, and corresponding the extracted events with the key cross-section locations. For vehicles with stop line or gate passing events, the time of occurrence of the corresponding cross-section event is preferentially used as the sorting anchor point. For vehicles lacking cross-section anchor points, a conservative sequence interval is given without violating the single-lane no-overtaking constraint and the minimum headway constraint, and a missing evidence mark is retained. The queue gating module forms a queue sequence list accordingly.

[0016] In a preferred embodiment, the confidence level of vehicle sequence evidence is used to characterize the reliability of vehicle sequence evidence in the current queuing session, and is determined based on the completeness rate of key section events, topology consistency deviation and time alignment residual; the queue gating module divides vehicles into a set of vehicles that meet the threshold condition and a set of vehicles that do not meet the threshold condition according to the threshold determination result of the confidence level of vehicle sequence evidence, so as to serve as the basis for short-distance priority access control.

[0017] In a preferred embodiment, the load threshold is used to characterize the overall pressure status of the current station and its surrounding roads, and is determined based on the reliable release capacity, the intensity of the short-distance supply and demand gap in the region, and the intensity of queue overflow; the load threshold is used to uniformly adjust the release batch size, entry time limit, sequence layer range, passenger boarding window allocation method, and the intensity of rebalancing guidance in the area outside the station.

[0018] In a preferred embodiment, the batch matching module marks orders entering the rolling scheduling window as short-distance orders. The short-distance marking is completed according to the short-distance order judgment rules bound to the station. The short-distance order judgment rules generate short-distance identifiers based at least on whether the pick-up point is located within the electronic fence of the pick-up area, whether the mileage or estimated duration exceeds the short-distance threshold, and whether the cost of returning the vehicle to the depot empty after the short-distance is completed is lower than a preset threshold. The short-distance identifier and the rule version number are written into the order record.

[0019] In a preferred embodiment, the batch matching module first generates a release batch and determines the sequence layer range within each rolling scheduling window. The sequence layer range is selected from the queue sequence list in consecutive order, and the sequence not crossing is used as a hard constraint. Within the sequence layer, the batch matching module performs admission control on the candidate vehicle set based on the threshold judgment result of the credibility of the sequence evidence for each vehicle, so that short-distance priority is only implemented within the vehicle set that meets the threshold condition. The batch matching module also assigns a target boarding window and entry direction to each released vehicle according to the boarding window occupancy status and capacity, and after completing the sequence layer matching within the station, it implements the rebalancing guidance of the area outside the station based on the load gating quantity.

[0020] In a preferred embodiment, the closed-loop processing module issues departure batches, sequence layer ranges, matching results, target boarding windows, entry time limits, and rebalancing guidance information for areas outside the station to both the station management terminal and the vehicle terminal, and continuously monitors the execution process under a unified time benchmark. The closed-loop processing module identifies events such as timeouts, unauthorized entry, cross-sequence layer insertion, entry into non-designated boarding windows, and window congestion anomalies through execution event correlation, and performs rematching, re-entry, window adjustment, subsequent batch convergence, or guidance intensity adjustment according to the anomaly type. The closed-loop processing module also audits and solidifies the end-to-end results of each departure and matching, and the audit content includes at least the departure batch identifier, sequence layer range, vehicle and order identifier, target boarding window, key section event summary, execution status chain node time, and anomaly handling information.

[0021] The technical effects and advantages of this invention are as follows:

[0022] This invention transforms the traditional method of short-distance order scheduling at key stations—which relies on fence locations, rough arrival order, or manual experience—into an ordered scheduling method based on sequence evidence chains, key section anchor points, and station topology constraints. This is achieved by introducing a topological evidence module, a queue gating module, a batch matching module, and a closed-loop processing module. For events such as occlusion, recognition failure, license plate damage, or terminal disconnection, this invention does not simply replace them with estimated results. Instead, it preserves the missing evidence segments and performs conservative sequence reconstruction by combining single-lane non-overtaking constraints and minimum headway constraints. Simultaneously, it uses the credibility of each vehicle's sequence evidence to gating short-distance priority access. This allows for more reliable maintenance of queuing order from the waiting area to the boarding area under complex conditions, reducing misjudgments of queue jumping, cross-sequence releases, and related on-site disputes, thereby improving the feasibility and fairness of the short-distance order priority mechanism.

[0023] This invention further unifies the adjustment of batch size, entry time limit, sequence layer range, passenger boarding window allocation, and external area rebalancing guidance within the station by using load gating. Combined with short-distance order identification, continuous sequence layer matching, and execution process monitoring within the rolling dispatch window, it achieves coordinated load balancing dispatching both inside and outside the station. During the execution phase, it can also identify, adjust, and audit abnormalities such as late arrivals, unauthorized entry, cross-sequence layer insertion, entry into non-designated passenger boarding windows, and window congestion, ensuring traceability for every release, matching, and anomaly handling. Therefore, this invention not only improves the turnover efficiency of short-distance orders and passenger boarding organization efficiency at key stations but also suppresses the adverse effects of queue overflow on surrounding road traffic order, enhancing the traffic police department's ability to supervise, review, and refine the dispatching process at key stations. Attached Figure Description

[0024] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0025] Figure 1 This is a schematic diagram of the intelligent matching and load balancing scheduling system for short-distance taxi orders of the present invention;

[0026] Figure 2 A schematic diagram of the topology and key sections of key stations;

[0027] Figure 3 A schematic diagram illustrating the generation of the sequence evidence chain and the reconstruction of the queue sequence.

[0028] Figure 4 This is a flowchart of batch matching and closed-loop processing under the rolling scheduling window. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Example 1: The intelligent matching and load balancing scheduling system for short-distance taxi orders of the present invention, such as... Figure 1 As shown, it includes the following modules:

[0031] Module 1: Topological Evidence Module; In this module, the traffic police brigade's dispatch platform uses key stations and surrounding roads as dispatch areas, accessing the data required for dispatching short-distance taxi orders. This includes at least the location and vehicle operation status information reported by the taxi's onboard terminal, passenger status and business status information reported by the meter or operating terminal, license plate passing events generated by station entrance / exit checkpoints or video structured recognition, station gate entry / exit information, traffic signal release status and temporary traffic organization information at traffic police side intersections, and traffic speed or queue length monitoring information for roads surrounding the station. The dispatch platform first performs field standardization and integrity checks on all source data. For fields such as vehicle identification, station identification, event type, occurrence time, and spatial location, consistent parsing can be performed under the same data model. For missing fields or outliers, the dispatch platform does not discard them, but adds missing and outlier markers to the original records to prevent evidence loss during subsequent tracing.

[0032] During the access phase, the scheduling platform establishes a unified time base, using GPS time synchronization or an NTP server to synchronize its own system time as the unified time base. For each type of data source, the platform records the arrival time of data packets. And extract the original timestamp The average network latency for this type of data source is calculated using a sliding window. For example, the sliding window contains the 100 most recent events: ; Calibrated event occurrence time The calculation formula is For each event record, the scheduling platform also stores an alignment residual. This is used to represent the deviation between the original timestamp of the event and the calibration time under a unified timeline. Reserved. This can be used in subsequent reliability and anomaly analysis to distinguish between changes in actual behavior and timing misalignments caused by reporting delay jitter; when When data consistently exceeds a threshold within a short period, the scheduling platform identifies it as an abnormal time segment and reduces the priority of data within that segment in subsequent decisions. The threshold can be set according to the 95th percentile of historical network latency, such as 3 seconds. Maintenance personnel can adjust this based on on-site communication conditions.

[0033] After unifying the time base, the scheduling platform solidifies and models the spatial elements of the stations to form the station topology, such as... Figure 2As shown, the station topology includes at least an electronic fence for the holding area, an electronic fence for the pick-up area, a set of pick-up windows, and a set of key sections. The pick-up window set is represented as a dockable line segment window with capacity attributes. The set of key sections includes at least entrance gates, stop lines, and diversion points to reflect the queuing process. The dispatch platform also fixes channel direction constraints and single-lane overtaking constraints to regulate the legal travel direction and queuing order of vehicles between the holding area and the pick-up area. For temporary traffic organization information from the traffic police side, such as entrance closures, temporary pick-up point relocations, prohibited U-turns, or one-way traffic zones, the dispatch platform saves these information in the form of a traffic organization version bound to the station topology. This ensures that the order determination within the same time period corresponds to the currently valid organization rules, avoiding the interpretation of on-site behavior using outdated rules.

[0034] Based on the aforementioned data and topology, the scheduling platform uses vehicle identifiers as the primary key to correlate multi-source events, generating a sequence evidence chain for vehicles, such as... Figure 3 As shown, the sequence evidence chain describes the entire process of a vehicle entering the holding area, advancing along the queue to the key section, entering the boarding area, and reaching the specific boarding window. During the association process, the dispatch platform prioritizes using license plate passing events and gate entry / exit records as section-level anchor points, with vehicle trajectory as supplementary evidence. Following a unified timeline, the events of the same vehicle entering the holding area, reaching the stop line or gate, entering the boarding area, entering the boarding window, and leaving the boarding area are sequentially written into the vehicle's sequence evidence chain. At each event node, the event source type, station topology version, traffic organization version, and corresponding information are recorded. Information. For events missing due to occlusion, recognition failure, license plate damage, or terminal disconnection, the scheduling platform does not directly replace the missing facts with interpolation. Instead, it retains the gap and marks it as a missing evidence segment. At the same time, it records the start and end times and the missing type of the missing segment so that it can be conservatively repaired by cross-section anchoring and topological constraints in the subsequent sequence reconstruction stage, and provides interpretable evidence for audit review.

[0035] Module Two: Queue Gating Module; In this module, the dispatching platform reconstructs the real-time queue position of the vehicle holding area using the sequence evidence chain set, missing evidence segment markers, and station topology and traffic organization version inputs provided in Module One. Based on this, it outputs a gating quantity suitable for subsequent release and dispatch. Station scenarios often exhibit characteristics such as low-speed movement, discontinuous occlusion recognition, and vehicle-mounted reporting delays and jitter. Therefore, the dispatching platform's sequence reconstruction is not limited to a coarse reconstruction of the order of vehicle appearance within the fence, but rather focuses on key cross-sectional events and uses station topology as constraints to reconstruct the order of vehicle advancement within the same time period, thus aligning with on-site queuing rules and avoiding the risks of queue-jumping disputes and erroneous releases.

[0036] Specifically, the dispatching platform first organizes the sequence evidence chain for each vehicle on a unified timeline, extracting cross-sectional events directly related to queue progression, including events of entering the holding area, reaching the stop line or gate, entering the boarding area, and entering the boarding window, and mapping these events one-to-one with the key cross-sectional locations in the station topology. For vehicles with stop line or gate passage events, the dispatching platform prioritizes the occurrence time of that cross-sectional event as the sorting anchor point. When multiple vehicles have similar times in the same cross-sectional event, the dispatching platform combines upstream cross-sectional events and channel direction constraints to disambiguate, ensuring the sorting result conforms to the unidirectional progression from the holding area to the channel and then to the boarding area. For vehicles lacking cross-sectional anchor points, the dispatching platform does not directly assign a definitive sequence position. Instead, it uses the missing evidence segment as the boundary, providing a conservative sequence position interval without violating the single-lane overtaking constraint and the minimum headway constraint, and retains a missing evidence marker for the vehicle in the queue sequence position list for more robust access control during subsequent release. Using the above method, the dispatching platform generates a real-time queue position list for stations. The queue position list includes at least the vehicle identifier, position number or position range, corresponding station topology version and traffic organization version, and a marker for missing evidence segments.

[0037] While generating the queue order list, the scheduling platform further calculates the credibility of the order evidence for each vehicle. This is used to characterize the reliability of the vehicle's sequence number evidence in the current session, thereby distinguishing available and unavailable sequence numbers at the data level. The reliability of the sequence number evidence... The value range is [0,1]. A larger value indicates more complete evidence, better compliance with site topology constraints, and more stable time alignment. The calculation formula is: ; in This means cropping the result to the [0,1] interval; Right now ; The function is defined as: ; in .

[0038] , , For the weighting coefficients, satisfying The meanings and calculation methods of each parameter are as follows.

[0039] Completeness of critical section events The scheduling platform has a pre-set set of key section events. ={Entering the waiting area, reaching the stop line or gate, entering the boarding area, entering the boarding window}, the trigger conditions for each type of event are:

[0040] Vehicle entry event: Triggered when a vehicle first enters the electronic fence of the parking area, which can be generated by the vehicle's GPS positioning or geomagnetic sensor;

[0041] Events that occur when a vehicle reaches the stop line or gate: triggered when a vehicle passes through the station entrance gate, either by a gate lifting signal or by a license plate recognition camera capturing the image.

[0042] Passenger pick-up area entry event: Triggered when a vehicle enters the electronic fence of the passenger pick-up area, which can be generated by vehicle positioning or geomagnetic sensor;

[0043] Passenger boarding window event: Triggered when a vehicle stops at a designated passenger boarding window, either by the window's geomagnetic sensor or video recognition.

[0044] Vehicle statistics Number of critical section events successfully associated in the current queuing session ,but: ,in , The value ranges from 0 to 4.

[0045] Topology consistency deviation The dispatching platform performs a topology rule check on each vehicle in the current queue at a fixed interval, such as every 5 seconds. The check includes the following:

[0046] Single-lane overtaking constraint: Sorted by the time of occurrence of events at critical sections, if the event of the following vehicle occurs earlier than that of the preceding vehicle, it is considered a violation;

[0047] Channel direction constraints: The direction of vehicle travel must be consistent with the direction specified by the station topology, for example, the entrance can only be entered and the exit can only be exited;

[0048] Section sequence constraint: Events must occur in the following order: entering the holding area, reaching the stop line, entering the boarding area, and entering the boarding window. Any disorder in this sequence will be considered a violation. For vehicles The total number of times the queued session is checked. Each check is considered a violation if any of the three rules mentioned above are not met. To accumulate the number of violations, then: ; indicates a violation of proportion, function and These are operations to find the minimum and maximum values, respectively.

[0049] Time alignment residual index For vehicles For all associated events in the current queuing session, collect the alignment residual for each event. The absolute values, take the median of these absolute values. To resist the impact of extreme values, and with a preset tolerance limit. Normalize: ; The method for determining this is based on the device's reporting cycle setting. For example, if the reporting cycle of the vehicle terminal is 1 second, then take... The maximum allowable alignment deviation is 3 seconds.

[0050] Weighting coefficient , , Determination: Fixed empirical values ​​are adopted, preset based on site equipment reliability and data quality. Considering that the completeness of critical section events is the most important, followed by topology consistency, and then time alignment quality, the following settings are established. , , This weight can be adjusted periodically by system maintenance personnel based on actual operational results.

[0051] The scheduling platform received Set threshold after This is used to categorize vehicles into a trustworthy set and an untrustworthy set. A fixed threshold is used, for example, by setting... That is, only Only vehicles meeting the minimum requirements will be included in the short-distance priority queue. This threshold can be adjusted based on station operational experience.

[0052] After completing the vehicle-level reliability calculation, the scheduling platform uses a rolling scheduling window. Further, a gating quantity for load balancing inside and outside the site is formed. Scrolling dispatch window It can be set according to the station's business cycle, for example, 30 seconds or 60 seconds. The value range is [0,1]. A larger value indicates higher pressure on the current station and surrounding roads or a more significant short-distance traffic gap in the area, requiring a more conservative approach to traffic release and stronger balanced guidance. Its calculation formula is: ;

[0053] in , , Let be the weighting coefficient, satisfying The meanings and calculation methods of each parameter are as follows.

[0054] Trustworthy release capability In the scrolling schedule window Internally, count the total number of queues. That is, the number of vehicles in the queue sequence list that are in the holding area and have not yet been released, and the number of vehicles that meet the following conditions. Number of vehicles ,but: ;

[0055] Intensity of regional short-distance supply and demand gap The dispatching platform displays the dispatching unit to which the site belongs in the window. Short-distance demand within The moving average method is used for forecasting. The average number of actual short-distance order arrivals over the past three windows of equal length is taken, with the window length set to 60 seconds. Then: ;in This represents the actual number of orders placed in the first i windows.

[0056] Effective and executable capacity In the window The system displays the number of available vehicles that can be dispatched to a station. Filtering criteria include: vehicle status (available, online, no assigned orders), and reachability assessment: the estimated travel time from the vehicle's current location to the station entrance is calculated using a route planning API; if this time is less than 5 minutes, the vehicle is considered reachable. The dispatch platform displays this information in the window. The system scans all eligible vehicles in real time, deduplicating and counting them to obtain the result. .

[0057] The intensity of the supply-demand gap is calculated as follows: ;

[0058] Queue overflow strength Overflow length The length is obtained through video detection. Using cameras at the station entrance, an image recognition algorithm detects the number of vehicles queuing, and multiplies this number by the average vehicle length (e.g., 4.5 meters per vehicle) to calculate the maximum allowable overflow length. The distance from the exit of the parking area to the nearest intersection stop line, such as 100 meters, or a threshold set by traffic police based on road capacity. Then: ;

[0059] Weighting coefficient , , Determination: A fixed empirical value is adopted. Considering that reliable clearance capacity reflects internal station order, supply-demand gap reflects external demand, and spillover intensity reflects road pressure, for example, a set... , , This indicates that the importance of the three factors decreases in that order. This weight can be adjusted periodically by system maintenance personnel based on actual operational performance.

[0060] Therefore, the output of module two is the queue order and the number of cars. Based on the threshold determination results, the rolling scheduling window W is used. , , and The outputs are applied to batch size, sequence level range, short-distance order matching admission set, and load balancing inside and outside the site, respectively, to improve the turnaround efficiency of short-distance orders and suppress queue overflow without affecting the queuing order of the site.

[0061] Module 3: Batch Matching Module; such as Figure 4 As shown, in this module, the scheduling platform uses the queue order list and the credibility of the order evidence for each vehicle output by module two. and its threshold determination results, load gating quantity under the rolling scheduling window W As input to the scheduling constraints, and simultaneously receiving order flows related to the stations as pending objects, the system completes short-distance order identification, release batch generation, order-vehicle matching within the sequence layer, and passenger boarding window allocation within the scrolling scheduling window. It also links in-station scheduling with out-of-station regional scheduling, ensuring that the rapid turnover of short-distance orders does not come at the expense of disrupting queuing order or causing spillover congestion. The scheduling platform maintains the current valid station topology and traffic organization version for each scrolling scheduling window, ensuring that the entrance direction, traffic restrictions, and available parking window status used for matching and release are consistent with the on-site organization by traffic police, thereby avoiding discrepancies where online assignment is feasible but on-site execution is not.

[0062] In order management, orders entering window W on the dispatch platform are marked as short-distance orders. This marking is based on short-distance order determination rules bound to the station. These rules consider at least the following conditions: whether the pick-up point is within the electronic fence of the pick-up area, whether the mileage or estimated duration exceeds the short-distance threshold, and whether the return empty-run cost after the short-distance trip is less than a certain threshold. The short-distance identifier and rule version number are then written to the order record. To prevent mis-dispatch due to duplication or changes in status, the order status is verified for consistency after marking as short-distance orders. This ensures the order is allocable and cannot be received from other channels. Verified short-distance orders are added to the current window's matching queue as the target set for subsequent ordering layers.

[0063] Regarding the organization of release and matching, the dispatching platform does not directly dispatch orders globally to all vehicles in the station's queues. Instead, it first generates release batches and determines the sequence level range within each rolling dispatch window W. The size and validity period of the release batches are determined by the load gating. Unified regulation: When When the level is high, the dispatching platform tends to reduce the size of the release batches and shorten the entry time limit, allowing vehicles to enter the pick-up area in batches within a controllable number, thereby reducing the probability of cutting in, overtaking, and congestion at the windows; when When the passenger flow is low and the boarding window is idle, the scheduling platform increases the batch size to improve throughput. The sequence layer is selected from the queue sequence list in a continuous sequence. It usually starts from the frontmost unreleased position to form a sequence layer, and the sequence layer is hard-constrained to ensure that the release and matching do not change the basic order of the queue on site. This mechanism avoids queue-jumping disputes caused by strategic priority.

[0064] When matching orders and vehicles within the sequence layer, the dispatch platform first determines based on... The threshold determination result is used to control the admission of the candidate vehicle set, ensuring that short-distance priority is only implemented within the vehicle set that meets the threshold condition. Specifically, the scheduling platform divides vehicles within the sequence layer into those that meet the threshold condition. The system prioritizes short-distance orders within the priority set and other vehicle sets, while maintaining the constraint of not crossing order positions. For vehicles that do not meet the threshold, the dispatching platform can restrict them to participating only in regular matching for non-short-distance orders, or require them to complete the key cross-section events in a subsequent rolling window before entering the priority set. This reduces the risk of disorder caused by vehicles with insufficient evidence being improperly released due to short-distance priority. To improve the executability of matching, the dispatching platform introduces arrival feasibility judgment in the matching calculation. Based on the vehicle's current location, traffic organization version, and road network traffic constraints, it estimates the estimated arrival time of the vehicle at the passenger pick-up area entrance and pick-up window. It also incorporates terminal online quality, recent confirmation timeouts, or order rejection behaviors as risk factors into the candidate screening, making the final output closer to the executable results that can be implemented on-site.

[0065] In passenger pick-up window allocation, the scheduling platform decomposes the pick-up area into multiple available pick-up window segments and stores the scroll window occupancy status and capacity of each window. Based on the occupancy status, it can statistically analyze events entering the pick-up window, events leaving the pick-up area, and vehicle parking duration to determine whether the window is idle or saturated. The matching output assigns a target pick-up window and entry direction to each vehicle to be released. Window allocation depends on… :when When the traffic volume is high, the dispatch platform prioritizes distributed allocation, causing multiple vehicles to converge in the same window within a short period of time, resulting in localized congestion; when... When the traffic flow is low, the window is ample, allowing the dispatching platform to implement more compact parallel allocation to improve passenger boarding efficiency. Simultaneously, the dispatching platform sets entry time limits and validity periods for each release and matching result, and outputs the station topology version and traffic organization version. This allows the station management terminal and vehicle terminals to execute using the same rule version, avoiding execution deviations caused by rule switching.

[0066] Regarding load balancing scheduling outside the site, after completing the order-level matching within the site, the scheduling platform utilizes the supply-demand gap intensity under window W. With spillover intensity The generation area rebalancing guides maintain a relative balance between the distribution of empty vehicles and short-distance demand around stations and adjacent dispatching units. The dispatching platform selects idle vehicles not currently in the current sequence layer that will not exacerbate station entrance conflicts as guidance targets, issues recommended waiting points and arrival time limits to dispatching units with larger gaps, and sets cooldown times to suppress frequent vehicle back-and-forth trips between adjacent units; when This indicates an increase in spillover and When traffic is high, the dispatching platform implements a flow control and absorption switching strategy around the station. Specifically, when the station's capacity allows, it prioritizes guiding empty vehicles from the periphery into the waiting area and incorporating them into the queue management system. When the station's pressure is too high, it reduces the intensity of guidance in the station's direction and instead distributes vehicles to adjacent dispatching units or backup pick-up points to prevent the station from becoming a single point of aggregation. By coordinating batch release, queue level matching, window allocation within the station with the rebalancing guidance outside the station within the same rolling window, the dispatching platform can improve the turnaround efficiency of short-distance orders while ensuring queuing order and suppressing the impact of queue overflow on the passage of other lanes.

[0067] Module Four: Closed-Loop Processing Module; In this module, the dispatch platform uses the release batches, sequence ranges, short-distance order and vehicle matching results, target passenger pick-up windows and entry time limits, and regional rebalancing guidance lists output from Module Three as the results to be executed. It then issues executable instructions to both the station management terminal and the vehicle terminal, continuously monitoring the instruction execution process under a unified time benchmark to ensure consistency between instructions, execution, and evidence. The station management terminal can be a station control terminal deployed by the traffic police brigade or a control unit connected to the station's gate, display screen, and voice broadcasting equipment. After receiving the release batch instructions from the dispatch platform, it controls the gate to release or the display screen to guide vehicles according to the vehicle identification, entry direction, and entry time limit within the batch, and transmits execution events such as gate lifting and vehicle passage back to the dispatch platform. After receiving the matching instructions, the vehicle terminal enters the execution state. The terminal interface displays the driver the matched order, recommended entrance, target passenger pick-up window, and validity period, and continuously uploads its location and status changes during the journey, enabling the dispatch platform to determine in real time whether the vehicle has entered the station channel and reached the designated window according to the instructions. For regional rebalancing guidance, the dispatch platform sends recommended waiting points, arrival time limits, and cooldown times to idle vehicle terminals and binds them to the traffic organization version, so that the vehicle guidance path is consistent with the traffic police's on-site restrictions such as no-entry and one-way traffic, thereby avoiding guiding vehicles into temporary control sections or forming new road-occupying gathering points around the stations.

[0068] To achieve verifiable monitoring of the execution process, the dispatch platform, based on the sequence evidence chain mechanism established in Module 1, continuously accesses events such as vehicle license plate passage at checkpoints or via video, gate entry and exit records, entry into the passenger pick-up area, entry into the passenger pick-up window, and order status feedback, and correlates these execution events under a unified timeline. The dispatch platform maintains an execution trajectory segment for each released vehicle, covering at least the entire process from the effective time of the release instruction to the vehicle entering the passenger pick-up area, driving into the target passenger pick-up window, and completing the pick-up confirmation. Simultaneously, it maintains an execution status chain for each short-distance order, including at least the following status nodes: dispatched, confirmed, arrived at the passenger pick-up area, arrived at the passenger pick-up window, passenger boarding, and order billing commencement. By cross-validating the execution trajectory segment with the execution status chain and the release batch with the sequence layer range, the dispatching platform can identify abnormal situations such as whether a vehicle has entered without authorization, whether it has crossed the sequence layer to insert, whether it has entered a non-designated boarding window, and whether it has failed to arrive within the validity period. The platform also solidifies the abnormalities and the corresponding evidence fragment index together for the traffic police to review.

[0069] When an anomalies are detected, the dispatch platform does not use a simple reassignment process. Instead, it performs interpretable rollback based on the anomaly type to suppress command storms and on-site chaos. For vehicles that fail to arrive on time, the dispatch platform removes the vehicle from the valid set of the current release batch and triggers a rematch for the corresponding short-distance order. Simultaneously, it reduces the vehicle's participation privileges in subsequent rolling dispatch windows, preventing it from participating in the short-distance priority set until stable execution is restored. For unauthorized entry or insertion across sequence levels, the dispatch platform compares the behavior with the corresponding key section events, traffic organization version, and station topology constraints. After confirming inconsistencies with the sequence level range, it outputs an anomaly alert to the station management terminal along with an evidence fragment index. Simultaneously, it sets temporary restrictions on the vehicle on the system side, such as requiring it to re-enter the waiting area or only allowing it to participate in normal matching in subsequent windows, thus bringing vehicle behavior back into a controllable order. For entry into non-designated boarding windows or situations where window congestion prevents parking, the dispatch platform combines the window occupancy status with the current... The system employs adaptive handling at different levels: when there is still room in the window and the deviation is acceptable, adjacent windows can be reallocated and window adjustment instructions issued within the same release validity period; when the window is close to saturation or overflow pressure is increasing, the system tends to converge the size of subsequent batches and increase the intensity of window distribution to avoid repeated occurrences of similar congestion. All of the above rollback actions are recorded in the audit log with the reasons for cancellation and alternative decisions, ensuring that every adjustment has a traceable basis.

[0070] In this embodiment, the scheduling platform uses execution feedback to update the calculation input for subsequent windows. Abnormal events are included in the violation statistics of topology rule checks, causing the vehicle's topology consistency deviation statistics to change with execution deviations. Simultaneously, the event sequence disorder and reporting delay changes caused by anomalies are included in the alignment residual statistics, ensuring that the distribution of time alignment residuals reflects the current network and device status. Based on this, the scheduling platform recalculates the vehicle's sequence evidence credibility in subsequent windows. Based on this, the set of vehicles that meet the threshold conditions is updated, and then the reliable release capacity is recalculated within the rolling scheduling window W. This is in turn related to the intensity of the short-distance supply and demand gap in the region. and queue overflow strength Update the load gating value as well. .in, The calculation formula remains as follows: ; In the formula This represents the number of vehicles within the queue management range within window W. For which it satisfies The number of vehicles; through this update, the batch size, sequence layer range, window allocation strategy and regional rebalancing guidance intensity of subsequent windows will be automatically adjusted according to the on-site execution effect, so that the system can still maintain stable operation after changes in occlusion, traffic flow and traffic organization, and will not be continuously amplified into misassignment and spillover due to a single anomaly.

[0071] Based on the traffic police brigade's monitoring and review needs, the dispatch platform audits and solidifies the entire chain result of each release and matching. The audit log includes at least: release batch identifier, sequence layer range, vehicle and order identifier, short-distance labeling rule version, station topology version and traffic organization version, target passenger pick-up window and entry time limit, key section event summary, execution status chain node time, and anomaly handling and cancellation reasons. The dispatch platform can query relevant audit logs by station, time period, vehicle, or order. The traffic police can query why the vehicle was released, why the window was assigned, why rollback and rematch were performed, and why the batch size was reduced or increased, and quickly verify the information by combining evidence fragments. Thus, Module 4 will fully close the loop from issuance, execution, anomaly handling, and parameter correction, transforming the short-distance order of taxis from intelligent matching and load-balanced dispatch under station order constraints into a supervised and traceable process.

[0072] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0073] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0074] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart matching and load balancing dispatching system for short-distance taxi orders, characterized in that: It includes a topology evidence module, a queue gating module, a batch matching module, and a closed-loop processing module; The topology evidence module is used to uniformly access, standardize field definitions, verify integrity and align time bases for multi-source data within the scope of key stations and surrounding roads, and generate vehicle sequence evidence chains by combining station topology and traffic organization version. The queue gating module is used to reconstruct the real-time queue position of the vehicle holding area based on the sequence evidence chain, the result of the missing evidence segment marking, the site topology and the traffic organization version, form a queue position list, and generate the confidence level of the sequence evidence and the load gating level for each vehicle. The batch matching module is used to complete short-distance order identification, release batch generation, order layer range determination, order and vehicle matching within the order layer, passenger boarding window allocation, and load balancing scheduling inside and outside the station within the scrolling scheduling window, based on the queue order list, the credibility of each vehicle's order evidence and its threshold judgment result, the load control quantity, and the station-related order data. The closed-loop processing module is used to issue and execute, monitor, handle anomalies, adjust and audit information on release batches, sequence layer ranges, short-distance orders and vehicle matching results, target passenger pick-up windows, entry time limits and rebalancing guidance information in areas outside the station, and based on the execution feedback, correct the credibility of the sequence evidence and the load gating amount of each vehicle in the subsequent rolling dispatch window.

2. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 1, characterized in that: The data accessed by the topology evidence module includes at least the location data and vehicle operation status data reported by the taxi vehicle terminal, the passenger status data and business status data returned by the meter or operating terminal, the license plate passing events generated by the station entrance and exit checkpoint or video structured recognition, the entry and exit records generated by the station gate, the traffic signal release status and temporary traffic organization information at the traffic police side intersection, and the traffic speed monitoring data or queue length monitoring data of the road sections around the station.

3. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 2, characterized in that: The topological evidence module establishes a unified time reference, records the arrival time and original timestamp of data packets from each source, and retains the alignment residual after time alignment. The topological evidence module also solidifies the spatial elements of the site to form a site topology. The site topology includes at least the electronic fence of the parking area, the electronic fence of the passenger pick-up area, the set of passenger pick-up windows, and the set of key sections, and is bound to the traffic organization version so that the order determination corresponds to the currently effective on-site organization rules.

4. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 3, characterized in that: The topology evidence module associates multi-source events with the vehicle identifier as the primary key to generate a sequence evidence chain for the vehicle. The sequence evidence chain includes at least the events of entering the parking area, reaching the stop line or gate, entering the boarding area, entering the boarding window, and leaving the boarding area. For events missing due to obstruction, recognition failure, license plate damage, or terminal disconnection, the topology evidence module retains the gaps and marks them as evidence missing segments, while recording the start and end times and the type of missing evidence segment.

5. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 4, characterized in that: The queue gating module uses key cross-section events as anchor points and station topology as constraints to organize the sequence evidence chain for each vehicle, extracting events such as entering the holding area, reaching the stop line or gate, entering the boarding area, and entering the boarding window, and corresponding the extracted events to the key cross-section locations. For vehicles with stop line or gate passing events, the time of occurrence of the corresponding cross-section event is used as the sorting anchor point. For vehicles lacking cross-section anchor points, a conservative sequence interval is given without violating the single-lane overtaking constraint and the minimum headway constraint, and a missing evidence mark is retained. The queue gating module then forms a queue sequence list based on this.

6. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 5, characterized in that: The confidence level of vehicle sequence evidence is used to characterize the reliability of vehicle sequence evidence in the current queuing session, and is determined based on the completeness rate of key section events, topology consistency deviation and time alignment residual. The queue gating module divides vehicles into a set of vehicles that meet the threshold conditions and a set of vehicles that do not meet the threshold conditions based on the threshold determination result of the confidence level of vehicle sequence evidence, so as to serve as the basis for short-distance priority access control.

7. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 5, characterized in that: The load threshold is used to characterize the overall pressure status of the current station and its surrounding roads, and is determined based on the reliable release capacity, the intensity of the short-distance supply and demand gap in the region, and the intensity of queue overflow. The load threshold is used to uniformly adjust the release batch size, entry time limit, sequence layer range, passenger boarding window allocation method, and the intensity of rebalancing guidance in the area outside the station.

8. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 6, characterized in that: The batch matching module marks orders entering the rolling dispatch window as short-distance orders. The short-distance marking is completed according to the short-distance order judgment rules bound to the station. The short-distance order judgment rules are based on at least whether the pick-up point is located within the electronic fence of the pick-up area, whether the mileage or estimated time exceeds the short-distance threshold, and whether the cost of the vehicle returning to the depot empty after the short-distance is completed is lower than the preset threshold. The short-distance mark and the rule version number are written into the order record.

9. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 8, characterized in that: The batch matching module first generates a release batch and determines the sequence layer range within each rolling scheduling window. The sequence layer range is selected from the queue sequence list in consecutive order, and the sequence does not cross boundaries as a hard constraint. Within the sequence layer, the batch matching module performs admission control on the candidate vehicle set based on the threshold judgment result of the credibility of the sequence evidence for each vehicle, so that short-distance priority is only implemented within the vehicle set that meets the threshold condition. The batch matching module also assigns a target boarding window and entry direction to each released vehicle according to the boarding window occupancy status and capacity, and after completing the sequence layer matching within the station, it implements the rebalancing guidance of the area outside the station based on the load gating quantity.

10. The intelligent matching and load balancing scheduling system for short-distance taxi orders according to claim 9, characterized in that: The closed-loop processing module issues departure batches, sequence layer ranges, matching results, target boarding windows, entry time limits, and rebalancing guidance information for areas outside the station to both the station management terminal and the vehicle terminal, and continuously monitors the execution process under a unified time benchmark. The closed-loop processing module identifies events such as timeouts, unauthorized entry, cross-sequence layer insertion, entry into non-designated boarding windows, and window congestion anomalies through execution event correlation, and performs rematching, re-entry, window adjustment, subsequent batch convergence, or guidance intensity adjustment according to the anomaly type. The closed-loop processing module also audits and solidifies the end-to-end results of each departure and matching, and the audit content includes at least the departure batch identifier, sequence layer range, vehicle and order identifier, target boarding window, key section event summary, execution status chain node time, and anomaly handling information.

Images