A bus driving scheduling method based on bus shelter passenger flow data

Abstract

The application provides a bus driving scheduling method based on bus shelter passenger flow data, and relates to the technical field of bus operation optimization. The passenger density accumulation index is constructed by fusing the current congestion degree, the arrival rate deviation degree and the continuous accumulation degree, and is coupled and quantized with the bus shelter effective bearing capacity and the historical stable operation benchmark, so as to realize the fine perception of the passenger flow accumulation trend of each bus shelter. In this way, the density mutation interval with step change from a low-density interval to a high-density interval on the line can be accurately identified, and the defect of insufficient spatial targeting caused by the traditional scheduling method taking the line overall or section passenger flow as the control basis can be overcome.

Description

Technical Field

[0001] This invention relates to the field of public transport operation optimization technology, and more specifically, to a public transport dispatching method based on passenger flow data at bus shelters. Background Technology

[0002] As a core component of the urban passenger transport system, the operational scheduling efficiency of urban public transport directly impacts the public's travel experience and the overall operational efficiency of the urban transportation system. Public transport scheduling refers to the decision-making process of formulating vehicle departure times, routes, and stop strategies based on the spatiotemporal distribution characteristics of passenger flow, vehicle resource allocation, and road operating conditions. Early public transport scheduling relied primarily on fixed timetables and manual experience; once departure intervals and vehicle configurations were determined, they remained fixed for extended periods, making it difficult to effectively address dynamic factors such as traffic congestion and passenger flow fluctuations. With the large-scale deployment of IoT sensing technologies such as GPS, automated vehicle positioning, intelligent ticketing systems, and video passenger flow detection, the granularity and real-time nature of public transport operation data collection have significantly improved. Against this backdrop, public transport scheduling research has gradually evolved from static timetable optimization to dynamic responsive scheduling. Existing dynamic scheduling methods can be broadly categorized into two types: one adjusts departure frequency or intervals based on short-term passenger flow forecasts, utilizing machine learning models to predict future passenger flow and optimize capacity allocation accordingly; the other addresses the uneven spatial and temporal distribution of passenger flow by employing a combination of full-journey and express trains or interval operation modes, identifying high-passenger-flow sections and configuring differentiated services to improve capacity utilization efficiency. Furthermore, some studies have introduced multi-agent collaborative frameworks or deep reinforcement learning algorithms to attempt adaptive scheduling decisions in more complex operational scenarios.

[0003] However, the aforementioned existing technologies still have several unresolved technical shortcomings. First, most existing dynamic scheduling methods rely on overall or cross-sectional passenger flow as the basis for regulation, lacking the ability to spatially refine the perception of passenger flow at each waiting shelter along the route. When there are significant differences in passenger concentration at different waiting shelters on the same route, it is difficult to identify the "density abrupt change interval" where a step change occurs from a low-density area to a high-density area, resulting in insufficient targeting of scheduling interventions. Second, existing methods typically only consider the remaining passenger capacity of vehicles in capacity demand assessment, failing to incorporate the actual release capacity of preceding vehicles at the target station into the extrapolation. Even if a vehicle passes through a high-density waiting shelter, if its remaining capacity has been occupied by upstream passengers or the stop time is insufficient to complete full boarding, there is a significant discrepancy between the vehicle's "nominal capacity" and "actual release capacity," causing the capacity gap to be systematically underestimated. Third, existing dynamic scheduling research largely focuses on optimizing departure intervals, with less attention paid to the fine-grained control of vehicle stop durations. Furthermore, it lacks a systematic approach that uses vehicle kinematic safety constraints (such as braking distance, communication delay, and positioning error) as pre-verification conditions for scheduling instructions—leading to potential "train collisions" or safety risks during actual execution. To address these issues, this invention provides a bus scheduling method based on bus shelter passenger flow data. By constructing a passenger density accumulation index that integrates current congestion levels, arrival rate deviations, and continuous accumulation, it identifies density abrupt change intervals on the route and generates hierarchical scheduling instructions based on the release capacity of preceding vehicles and vehicle kinematic safety constraints. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a bus dispatching method based on bus shelter passenger flow data. This method, to a certain extent, solves the problem that bus departure intervals and speeds cannot dynamically respond to real-time changes and cumulative trends in bus shelter passenger density, resulting in excessively long passenger dwell times in high-density bus shelters and idle capacity in low-density bus shelters.

[0005] According to one aspect of the present invention, a bus dispatching method based on bus shelter passenger flow data is provided, comprising: The system acquires real-time passenger flow data and historical stable operation benchmarks for each bus shelter, and couples the current level of congestion, the degree of deviation in arrival rate, and the degree of continuous accumulation to generate a passenger density accumulation index. The passenger density cumulative index is sorted along the line stations, and density abrupt change intervals are determined by combining the distance between stations and historical spatial gradient benchmarks to generate a density gradient distribution map. Based on the density gradient distribution map, real-time vehicle location, remaining capacity, and capacity released by preceding vehicles, the unreleased demand of vehicles arriving at the density abrupt change interval is estimated, and a spatiotemporal correlation table of capacity demand is generated. Based on the spatiotemporal correlation table of the capacity demand, the passenger dwell time is prospectively assessed, and instructions to depart early or skip stops are generated for the queues to be departed, and instructions to shorten stops are generated for the queues in transit, thus obtaining a bus dispatching scheme.

[0006] Furthermore, the generation of the passenger density cumulative index includes: The effective capacity of a bus shelter is determined based on its currently available waiting area and allowable standing density. Based on the arrival time of the vehicle at the waiting pavilion, the opening time of the door, the closing time of the door, and the departure time of the vehicle, the decrease in the number of passengers during the period from the opening to the closing of the vehicle is marked as the vehicle service release amount, and the increase in the number of passengers after the vehicle leaves the station is included in the natural passenger arrival amount. The current passenger arrival rate is determined based on the natural passenger arrival volume, and the passenger density cumulative index is calculated by combining the baseline passenger arrival rate and baseline passenger number in the historical stable operation benchmark.

[0007] Furthermore, the passenger density cumulative index is determined according to the following formula:

[0008] in, Let \(i\) be the cumulative passenger density index of the \th (i\)th bus shelter. For the current number of passengers, In order to effectively accommodate the number of people, For the current passenger arrival rate, Based on the passenger arrival rate, To reach the rate residual correction, Based on the number of passengers, The length of the sliding time window. , and As weight.

[0009] Furthermore, the determination of the weights includes: The current congestion characteristics, arrival rate deviation characteristics, and continuous accumulation characteristics in the historical scheduling samples are used as input features, and the actual retention intensity after the corresponding scheduling cycle is used as the calibration target. With the objective of minimizing the deviation between the predicted retention intensity and the actual retention intensity, , and Perform constraint calibration.

[0010] Furthermore, determining the intensity of the density abrupt change includes: The cumulative passenger density index of each bus shelter is smoothed with edge preservation to obtain the cumulative passenger density index after edge preservation. The difference in passenger density cumulative index between adjacent bus shelters is scaled according to the distance between stations; The scaling results are compared with the scaling gradient center values ​​in the historical stable operation benchmark, and the density mutation intensity is obtained based on the gradient discrepancy and the spatial trend confidence coefficient. The intensity of the density abrupt change is determined according to the following formula:

[0011] in, The intensity of density abrupt change between adjacent bus shelters. and The cumulative passenger density index after edge protection smoothing. This represents the actual distance between stations. This is the reference distance between stations on the line. For the scaled gradient center value, For gradient discrete quantities, This is the gradient residual correction amount. This represents the spatial trend confidence coefficient.

[0012] Furthermore, the generation of the spatiotemporal correlation table of vehicle capacity demand includes: Based on the real-time location of buses, route topology, travel speed, and estimated stop time at intermediate bus shelters, the estimated arrival time of each bus shelter within the density change interval is determined. The short-term predicted passenger arrival rate is determined based on the current passenger arrival rate and the baseline passenger arrival rate in the historical stable operation baseline. Based on the current number of passengers at the bus shelter within the density mutation interval, the short-term predicted passenger arrival rate, and the additional capacity intervention boundary, determine the estimated remaining time when additional capacity intervention is required in the density mutation interval. The expected release capacity of the preceding vehicles is determined based on their expected arrival order, remaining capacity, and station service capacity. The unreleased passenger demand before the vehicle arrives, the remaining vehicle capacity, the intensity of density mutation intervals, and the estimated remaining time are written into the spatiotemporal correlation table of vehicle capacity demand.

[0013] Furthermore, the generation of the instruction to shorten the stop time includes: Determine the intermediate bus shelters between the current location of vehicles en route and the density abrupt change zone; When the intermediate waiting shelter is in a low-density stable state, and the boarding and alighting flow in the door area has been completed, there are no passengers staying in the door safety area, no new passengers approaching the vehicle on the platform, and no passengers requesting to alight at this station, the intermediate waiting shelter is identified as a waiting shelter with shortened stops. Before issuing an instruction to shorten the stop time, the minimum safe distance is determined based on the vehicle's speed, braking deceleration, communication delay, positioning refresh delay, platform berth length, positioning error, and road operation disturbance. The instruction to shorten the stop time can only be executed if the expected actual distance between the vehicle and the vehicle in front is not less than the minimum safe distance and the next waiting platform berth is not occupied by the vehicle in front.

[0014] According to another aspect of the present invention, a bus dispatching system based on bus shelter passenger flow data is provided, comprising: The passenger flow status perception module is used to obtain real-time passenger flow data and historical stable operation benchmarks for each bus shelter, and couple the current congestion level, the degree of deviation of arrival rate and the degree of continuous accumulation to generate the passenger density accumulation index for each bus shelter. The density gradient analysis module is used to arrange the passenger density cumulative index along the line stations in sequence, combine the distance between adjacent waiting shelters and the historical spatial gradient benchmark to determine the density change interval, and generate a waiting shelter density gradient distribution map. The capacity demand projection module is used to project the unreleased passenger demand when each bus arrives at the density change interval based on the bus shelter density gradient distribution map, real-time location of buses, remaining capacity and capacity released by preceding vehicles, and generate a spatiotemporal correlation table of vehicle capacity demand. The scheduling decision and instruction generation module is used to prospectively assess the passenger dwell time in the waiting area within the density change interval based on the spatiotemporal correlation table of vehicle capacity demand, and generate early departure instructions or skip-stop instructions for the queue to be departed, and generate instructions to shorten the stop time for the queue in transit, and output the bus dispatching plan.

[0015] Compared with existing technologies, the bus dispatching method based on bus shelter passenger flow data provided by this invention constructs a passenger density accumulation index that integrates the current level of congestion, the degree of deviation in arrival rate, and the degree of continuous accumulation. This index is then coupled with the effective carrying capacity of bus shelters and historical stable operating benchmarks for dimensionless quantification, thereby achieving a refined perception of the passenger flow accumulation trend at each bus shelter. This allows for the accurate identification of density abrupt change zones along the route where a step shift occurs from low-density to high-density areas, overcoming the spatial targeting deficiencies of traditional dispatching methods that rely on overall or cross-sectional passenger flow for control.

[0016] Furthermore, this invention introduces a coupled extrapolation of the actual release of capacity by preceding vehicles and the intervention time of additional capacity, combined with the quantitative correlation between the estimated arrival time of vehicles in density abrupt change intervals and the unreleased demand, to generate a spatiotemporal correlation table of vehicle capacity demand that includes capacity coverage pressure, spatial abrupt change intensity, and time matching relationship. This avoids the systematic bias of equating the remaining passenger capacity of vehicles with the actual releaseable capacity, making the estimation of capacity gaps more accurate. Simultaneously, this invention adopts a hierarchical scheduling architecture, generating advance departure instructions and / or skip-stop driving instructions for queues to be dispatched, and generating instructions to shorten stop times for queues en route. The shortened stop instructions include kinematic safety checks on braking distance, communication delay, positioning error, and platform berth occupancy, enabling precise allocation of capacity resources to high-density accumulation intervals without causing train collisions or safety accidents. Thus, this invention realizes the transformation of bus dispatching from "passive response" to "proactive prediction", effectively shortens the dwell time of passengers in high-density bus shelters, improves the overall capacity utilization efficiency and service balance of the line, and provides a safety guarantee for the physical executability of bus dispatching instructions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a system block diagram of a bus dispatching method based on bus shelter passenger flow data according to an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram illustrating the density abrupt change interval identification of the bus dispatching method based on bus shelter passenger flow data according to an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram illustrating the spatiotemporal correlation of capacity demand in a bus dispatching method based on bus shelter passenger flow data according to an embodiment of the present invention. Detailed Implementation

[0020] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0021] Figure 1 This is a system block diagram of a bus dispatching method based on bus shelter passenger flow data according to an embodiment of the present invention. Figure 1As shown, the bus dispatching method based on bus shelter passenger flow data includes: S1. The bus dispatch control system first reads real-time passenger count data, historical stable operation benchmark data, vehicle arrival service event data, and effective carrying capacity status data of each bus shelter on the same bus route from the passenger flow detection device at the bus shelter, the historical passenger flow database of the route, the vehicle arrival service event interface, and the bus shelter status database. Then, based on the route's direction of operation, it establishes a bus shelter sequence according to the station order and aligns the current passenger count, passenger arrival rate within the past time window, vehicle service release process, and historical stable operation benchmark with time. After the above data is prepared, the bus dispatch control system uses the effective carrying capacity of the bus shelter as a capacity reference and the historical stable operation benchmark as a dynamic reference to perform dimensionless coupling calculations on the current congestion level, the degree of deviation in arrival rate, and the degree of continuous accumulation, generating the passenger density accumulation index corresponding to each bus shelter in the current dispatch cycle.

[0022] Specifically, the bus dispatch control system records the bus shelters arranged sequentially along the same route in the direction of vehicle travel as the 1st. Each bus shelter will record the current dispatch time as... The system first reads the... Real-time passenger count at each bus shelter during the current dispatch cycle The real-time passenger count, measured in persons, can be obtained using video passenger flow recognition equipment, infrared counting equipment, platform area target detection equipment, or multi-source fusion counting equipment. Subsequently, historical stable operating benchmarks matching the current date type, current time period, current weather conditions, route direction, and surrounding activity status are read. These historical stable operating benchmarks include at least the benchmark passenger arrival rate. Baseline passenger numbers And baseline spatial gradient data. The date type may include weekdays, rest days, holidays, or days of major events, and the surrounding activity status may include school dismissal, business district activities, sports events, or temporary traffic control status.

[0023] Furthermore, the historical stable operation benchmark does not use single-day passenger flow records, but is obtained through robust statistical analysis of passenger flow data from the same time zone across multiple comparable historical operating days. Specifically, abnormal operating days can be removed from the historical sample, including days with widespread vehicle malfunctions, days with severe road congestion, days with malfunctioning passenger flow detection equipment, and days with sudden large-scale events. In the remaining stable sample, the system uses the median, quantile envelope center value, or weighted robust mean to determine the stability. and When a line is newly opened, undergoing initial route rerouting, or lacks sufficient historical samples, the system enters a cold start process: It first generates a temporary historical stable operating baseline using the line's designed passenger flow forecast, passenger flow samples from adjacent similar lines during the same time period, land use attributes around stations, and operational data from similar stations across the entire network. Once the line has accumulated enough valid operational samples to cover multiple weekdays and non-weekdays, the system gradually replaces the temporary baseline with stable samples from the line itself. Therefore, S1 can still execute during the early operational phase when historical data is lacking, and will subsequently be automatically corrected based on the actual operational data of the line.

[0024] Subsequently, the system retrieves vehicle arrival service event data, which includes the vehicle's arrival time at the bus shelter, door opening time, door closing time, departure time, and passenger pick-up and drop-off records at the bus shelter. The system then uses past time windows. The system analyzes changes in passenger numbers at the bus shelter and distinguishes between natural passenger arrival and vehicle service release processes based on vehicle arrival events: if the number of passengers decreases between the opening and closing of the vehicle doors, this decrease is marked as vehicle boarding release and not considered as natural passenger loss; if the number of passengers increases again after the vehicle departs, this increase is counted as newly arrived passengers at the bus shelter. The system calculates the current passenger arrival rate based on the number of newly arrived passengers within the window and the window duration. The unit is people per minute. The system is determined based on the route scheduling cycle, the average vehicle arrival interval, and the historical passenger flow fluctuation cycle of the bus shelter. When passenger flow changes rapidly during morning and evening peak hours, the system uses a shorter window to enhance response, while when passenger flow is smooth, it uses a longer window to enhance noise resistance.

[0025] Furthermore, the system determines the first The effective capacity of each bus shelter The unit is people. The effective passenger capacity is obtained by multiplying the currently available waiting area of ​​the bus shelter by the allowable standing density. The currently available waiting area is determined based on the design area of ​​the bus shelter, the passable area of ​​the platform, the status of temporary enclosures, the occupancy status of entrances and exits, the occupancy status of shared bicycles or non-motorized vehicles, and the platform availability information recorded by the operation management system. When the bus operation system cannot obtain the enclosure or occupancy status in real time, the system uses the static effective area of ​​the bus shelter as the base value and corrects it by combining the upper boundary of the number of passengers that can be served when there is no continuous congestion in the most recent stable sample. Through this processing, bus shelters with different areas, different platform shapes, and different temporary occupancy statuses can be compared under their own carrying capacity scale.

[0026] Once the real-time passenger count, current passenger arrival rate, historical stable operating baseline, and effective carrying capacity are all determined, the system calculates the passenger density cumulative index. In one implementation, the passenger density cumulative index can be determined according to the following formula:

[0027] in, For the first The cumulative passenger density index of each bus shelter at the current dispatch time is a dimensionless quantity. This represents the current real-time number of passengers, expressed in persons. To effectively accommodate a certain number of people, the unit is one person; The current passenger arrival rate, in persons per minute; The baseline passenger arrival rate is the historical stable operating benchmark, in person / minute. The arrival rate residual correction, measured in people per minute, is determined by the minimum change in passenger numbers that the passenger flow detection device can reliably identify within a scheduling cycle, and is used to avoid... When the value approaches zero, the ratio becomes abnormally amplified. The baseline number of passengers in historically stable operations, expressed in persons; Indicates when For positive time take ,when Take 0 when it is negative or zero; , and These are dimensionless weights, corresponding to the current level of congestion, the degree of positive deviation of arrival rate, and the degree of continuous accumulation, respectively, with the sum of the three being 1.

[0028] In the above formula, This indicates the occupancy status of the current number of people waiting for the bus relative to the effective carrying capacity of the bus shelter; This indicates the positive deviation of the current passenger arrival rate from the historical stable operating baseline; This represents the cumulative area where the number of passengers has consistently exceeded the historical stable operating baseline within a past time window, with both the numerator and denominator measured in person-minutes. The engineering objective is to maintain exponential smoothness when passenger flow at bus shelters fluctuates only slightly, to quickly reflect the cumulative state when the current number of people, arrival rate deviation and continuous accumulation increase simultaneously, and to suppress the excessive impact of a single false detection or short-term spike on the scheduling state.

[0029] Furthermore, to ensure that the weight calibration process can be performed by those skilled in the art, the system can employ a constrained least squares calibration method. , and Specifically, the system combines the three dimensionless features corresponding to each scheduling cycle in the historical scheduling samples into a vector. ,in Indicates the current congestion characteristics. Indicates the rate deviation characteristic, This represents a continuous cumulative characteristic; the actual observed passenger dwell time or normalized dwell intensity after this scheduling cycle is denoted as... The system determines the weights by minimizing the following objective function:

[0030] and make And each weight is not less than 0. Among them, Number the historical samples; The normalized dwell intensity is obtained from actual operational records. It is a dimensionless quantity and can be obtained by dividing the actual waiting time by the normal service time scale in the historical stable operation benchmark of the line. It is a dimensionless smoothing coefficient used to prevent weights from changing drastically due to a small number of outlier samples; This represents the weight vector from the previous calibration period. The objective function can be solved using projective gradient descent, least squares with equality constraints, or quadratic programming. During online operation, the system can update the weights daily or weekly using the latest stable samples. When there are insufficient samples for newly opened lines, the system first uses the weights of similar lines or the weights of similar stations across the entire network as initial weights, and then gradually replaces them after obtaining stable samples for the current line. This calibration method ensures that the weights have a clear data basis, rather than being directly specified by human experience.

[0031] For example, a bus shelter currently has an effective capacity of 80 people, with 64 passengers currently present. The passenger arrival rate over the past 10 minutes was 5 people / minute, while the baseline passenger arrival rate in the historical stable operation benchmark was 2 people / minute. The residual correction for the arrival rate is 0.2 people / minute, and the integral of exceeding the baseline passenger number over the past 10 minutes is 320 people / minute. After the aforementioned least squares calibration, the weights for the current congestion term, rate deviation term, and continuous accumulation term are 0.40, 0.35, and 0.25, respectively. The system calculates the current congestion term to be 0.80, the rate deviation term to be approximately 1.36, and the continuous accumulation term to be 0.40, resulting in a weighted average of approximately 0.90. After exponential compression, the passenger density accumulation index is approximately 0.59. This example demonstrates that even if the current number of waiting passengers has not yet exceeded the effective capacity of the bus shelter, as long as the passenger arrival rate significantly increases relative to the historical stable operation benchmark and has formed a continuous accumulation, the system can treat it as a potential congestion state in subsequent steps. The above example is only for illustrating the calculation process and does not limit the scope of protection.

[0032] S2, the bus dispatch control system first receives the cumulative passenger density index of each bus shelter output by S1, and generates a passenger flow status sequence of the bus shelters according to the stop order of the bus route and the direction of vehicle travel; then it reads the actual route distance between adjacent bus shelters, the spatial gradient benchmark in the historical stable operation benchmark, and the data reliability of each bus shelter; after the passenger flow status sequence and spatial reference data of the bus shelters are prepared, the system performs scaled gradient analysis on the changes in the cumulative passenger density index between adjacent bus shelters, identifies the bus shelter intervals in the route that rapidly transition from low cumulative state to high cumulative state, and generates a bus shelter density gradient distribution map. Figure 2 As shown.

[0033] Specifically, the system treats each bus shelter as a node in the route topology, and the route connections between adjacent bus shelters as edges between nodes. For bidirectional bus routes, the system establishes separate passenger flow sequences for the upstream and downstream directions. For circular bus routes, the system determines the bus shelter arrangement order based on the current vehicle service direction and the route starting point defined by the dispatch system. For situations where the same physical station has upstream and downstream platforms or multiple separate berths, the system treats different waiting areas as independent bus shelter nodes. Each bus shelter node records the bus shelter number, route mileage coordinates, cumulative passenger density index, historical stable operation benchmark, data reliability, and most recent vehicle service event.

[0034] Furthermore, the system reads the spatial gradient data of adjacent bus shelters from historical stable operation benchmarks. This spatial gradient data includes the scaled gradient center values ​​of the same adjacent bus shelter edge under comparable date types, comparable time zones, and comparable route operating directions. and gradient discretization For newly opened routes or routes with insufficient historical samples, the system adopts a cold start spatial benchmark: first, temporary benchmarks are generated based on similar routes, similar station distances, similar station functional attributes, and the spatial distribution of passenger flow during the same time period across the entire network. and The data is then updated using the data from this line after accumulating stable operating samples. In this way, even without historical gradients for this line, S2 can still identify spatial abrupt changes, and the subsequent identification results will converge to the spatial structure of the line itself as the actual operating data is collected.

[0035] Subsequently, the system performs edge-preserving smoothing on the passenger flow sequence at bus shelters. This edge-preserving smoothing is used to suppress single-point detection noise while retaining true spatial abrupt changes. The system traverses the bus shelter sequence, forming a local window for each bus shelter and its one or two adjacent bus shelters. It reads the cumulative passenger density index, inter-station distance, and data reliability of each bus shelter within the local window. If the index difference between adjacent bus shelters within the local window is close to the normal spatial difference in historical stable operating benchmarks, the system allows for a small smoothing of that local window. If the index difference between adjacent bus shelters continues to increase in the same direction over multiple consecutive scheduling cycles, the system reduces the smoothing amplitude, preserving the edge change. Through this processing, the system avoids identifying instantaneous false detections of a single bus shelter as density abrupt changes, while also avoiding smoothing out the boundaries of truly high-density areas.

[0036] After edge smoothing is completed, the system calculates the density abrupt change intensity between adjacent bus shelters. In one embodiment, the density abrupt change intensity can be determined according to the following formula:

[0037] in, For the first The bus shelter and the first The intensity of density abrupt changes between individual bus shelters is a dimensionless quantity. and These are the first two lines after edge-preserving smoothing. The bus shelter and the first The cumulative passenger density index of each bus shelter; For the first The bus shelter and the first The actual distance between bus shelters along the bus route is expressed in meters. This is the reference station distance for the line, in meters. The median of all station distances on the line can be used. This is a dimensionless scale correction quantity; The central value of the scaled gradient of adjacent bus shelters in the historical stable operation benchmark is a dimensionless quantity. The gradient discrete quantity of adjacent waiting shelters in the historical stable operation benchmark is a dimensionless quantity that can be determined by the median absolute deviation or quantile width. The gradient residual correction is determined by the passenger detection resolution, the effective capacity of the waiting shelter, and the distance between reference stations. The spatial trend reliability coefficient ranges from 0 to 1 and is determined by the data reliability of two adjacent bus shelters, the continuity of gradient direction in the most recent scheduling cycles, and whether a concentrated passenger release event has just occurred.

[0038] The above formula serves to interpret the cumulative passenger density index difference between adjacent bus shelters within the context of the route's spatial scale and historical stable operating benchmarks. If the distance between two bus shelters is relatively long, the spatial abrupt change corresponding to the same index difference is weaker; conversely, if the distance is short, the same index difference may represent a stronger local passenger flow abrupt change. The system further compares the current scaled gradient with the central value of the historical stable spatial gradient at the same location. Compare and use historical gradient discretization. Standardization is performed to ensure that the identification results reflect "the degree of deviation from the normal spatial structure at that location along the line." Therefore, the intensity of this density abrupt change is not simply a comparison of the number of people at two stations, nor is it based on a uniform artificial boundary across the entire line. Instead, it is determined jointly by the current passenger flow status, station scale, historical stable operating benchmarks, and data reliability.

[0039] Subsequently, the system determines density mutation intervals based on the local prominence of density mutation intensity in the spatial sequence of the route, the gradient direction, and the continuous deviation of the bus shelter index from the historical stable operating benchmark. The system traverses each adjacent bus shelter edge along the vehicle's direction of travel: if the density mutation intensity of an edge forms a prominence point within its local adjacent range, and the passenger density accumulation index of the downstream bus shelter is higher than the stable envelope formed by the historical stable operating benchmark within a continuous scheduling cycle, then the system marks this edge as an entry edge of a high-density accumulation interval; if an edge shows a decline from a high-accumulation state to a stable state, and the passenger density accumulation index of its downstream bus shelter returns to the vicinity of the stable envelope formed by the historical stable operating benchmark, then the system marks this edge as a departure edge of a high-density accumulation interval. Once both entry and departure edges are determined, the system marks the set of bus shelters between them as a density mutation interval. If multiple bus shelters experience local fluctuations, but the intermediate bus shelters do not return to the vicinity of the historical stable operating benchmark, the system merges them into the same density mutation interval; if there are consecutive stable bus shelters between two high-density areas, the system splits them into different density mutation intervals.

[0040] After all adjacent bus shelter edges have been traversed, the system creates a data structure for the bus shelter density gradient distribution map. This data structure can be organized as a route topology map, a two-dimensional table, or key-value pairs. Each bus shelter node corresponds to a node entry in the graph, and each adjacent bus shelter edge corresponds to an edge entry in the graph. Node entries include at least the bus shelter number, route mileage coordinates, passenger density cumulative index, historical stable operation benchmark, data reliability, whether it belongs to a density abrupt change interval, and the interval number. Edge entries include at least the adjacent bus shelter numbers, distance between stations, scaled gradient, density abrupt change intensity, gradient direction, entry or exit edge identifier, and abrupt change reliability. After writing, the system performs an integrity check on the bus shelter density gradient distribution map to confirm that each bus shelter node and each adjacent edge has the spatial location, passenger flow status, and abrupt change attributes required for subsequent capacity projection.

[0041] For example, the cumulative passenger density indices of five consecutive bus shelters on a certain route, after edge-protection smoothing, are 0.21, 0.24, 0.68, 0.73, and 0.31, respectively, with the distances between adjacent shelters close to the reference distances between stations on the route. The system calculates the current scaled gradients of adjacent edges to be approximately 0.03, 0.44, 0.05, and -0.42. If the gradient center values ​​of the second and fourth edges in the historical stable operating baseline are close to 0.02 and the gradient dispersion is small, and the cumulative passenger density indices of the third and fourth shelters are consistently higher than the stable envelope formed by the historical stable operating baseline over multiple scheduling cycles, then the system marks the second edge as an entering edge, the fourth edge as a leaving edge, and the area between the third and fourth shelters as a density abrupt change interval, and writes this interval and its abrupt change intensity into the shelter density gradient distribution map. This example is only for illustrating the processing procedure and does not limit the number of bus shelters, index values, or distances between stations on the route.

[0042] S3, the bus dispatch control system first receives the bus shelter density gradient distribution map generated by S2, and reads the spatial location of each density abrupt change interval, the starting and ending bus shelters of the interval, the set of bus shelters within the interval, and the interval abrupt change intensity; then it receives the real-time location, speed, current passenger load, remaining capacity, vehicle queue status, and the relationship between preceding and following vehicles of each bus; after the density abrupt change interval information and vehicle operation information are prepared, the system couples and extrapolates the arrival time of each candidate vehicle in the density abrupt change interval, the continued accumulation of passengers before the vehicle's arrival, the time for additional capacity intervention in the density abrupt change interval, and the available capacity of the preceding vehicle, generating a spatiotemporal correlation table of vehicle capacity demand. For example... Figure 3 As shown.

[0043] Specifically, the system reads the vehicle number, direction of travel, real-time location coordinates, route mileage coordinates, current speed, rated passenger capacity, current passenger capacity, door status, whether it is currently stopped, whether it is at the starting station waiting to depart, whether it can access the route from the parking lot or temporary dispatch point, and its spatiotemporal relationship with the preceding and following vehicles. The system projects the real-time location coordinates of the vehicles onto the bus route topology to obtain the mileage position of the vehicles in the route, and determines the positional relationship of the vehicles relative to each density change interval. If a vehicle has already crossed the target density change interval and will not return to serve that interval within the current dispatch cycle, the system does not include the vehicle in the candidate service vehicle set for that interval; if the vehicle is located upstream of the target interval, inside the interval, at the starting station waiting to depart, or at a dispatchable access point, the system includes the vehicle in the candidate service vehicle set.

[0044] Subsequently, the system calculates the estimated time for each candidate vehicle to reach each bus shelter within the density mutation interval. Instead of using a simple extrapolation method of dividing the distance from the current location to the target bus shelter by the current vehicle speed, the system calculates the travel time segment by segment along the route topology from the vehicle's current location to the target bus shelter. For each segment, the system reads the current traffic speed, the travel speed of the same time period in historical stable operating benchmarks, road speed limits, signal control status, congestion status, and vehicle acceleration / deceleration capabilities to form the feasible travel speed for that segment. For intermediate bus shelters, the system reads the cumulative passenger density index, the estimated number of passengers getting on and off, the number of passengers requesting to alight from the vehicle, and the door passage capacity to calculate the estimated stop time. The system then sequentially sums the travel time of each segment and the intermediate stop time to obtain the estimated time for the [previous] bus shelter. The car arrived at the Estimated time for each bus shelter The unit is minutes.

[0045] Furthermore, the system calculates the continued accumulation of passengers within the density abrupt change interval before vehicle arrival. To avoid short-term passenger arrival rate prediction remaining merely a qualitative description, the system employs an adaptive fusion prediction method jointly determined by the current arrival rate, historical stable operating benchmarks, and rate fluctuation stability. In one implementation, the first... The short-term predicted passenger arrival rate of a bus shelter can be determined using the following formula:

[0046] in, For short-term prediction of passenger arrival rate, the unit is people / minute; The current passenger arrival rate, in persons per minute; The baseline passenger arrival rate is the historical stable operating benchmark, in person / minute. These are dimensionless adaptive fusion coefficients, ranging from 0 to 1. It can be determined by the degree of continuous deviation of the current arrival rate from the historical stable operating benchmark and the fluctuation stability over the most recent scheduling cycles. Specifically, when Compared to When there is a continuous increase and the direction of change in adjacent cycles is consistent. Increase the value to make the prediction result closer to the current arrival rate; when When it only manifests as a single peak or repeated changes in direction between adjacent periods The system reduces [the number of parameters] to bring the prediction results closer to the historical stable operating benchmark; when the reliability of the bus shelter data decreases, the system also reduces [the number of parameters]. This method enhances the constraint effect of historical stable operating benchmarks and the status of adjacent bus shelters on the prediction results. Both terms in the prediction formula are in people per minute, and the output after linear fusion remains in people per minute, maintaining consistent dimensions. Furthermore, the fusion coefficient is determined by the data distribution and trends.

[0047] Furthermore, based on the effective carrying capacity of the bus shelters within the density mutation interval and historical stable operation benchmarks, the system calculates the estimated remaining time for the interval to evolve from its current state to a state requiring additional transportation capacity intervention. The requirement for additional transportation capacity is not determined by subjective human judgment, but is jointly determined by the carrying capacity of the waiting shelter, the upper limit of historical stable operation, and the constraints of platform safety operation. Specifically, the system is for the first... Each bus shelter defines the boundaries for additional transport capacity intervention. The unit is a person; the stated It can be by Confirmed, among which The dimensionless load-bearing capacity correction coefficient can be determined based on the maximum service capacity quantile, platform safety management requirements, changes in available platform area, and passenger queue overflow records in the historical stable operation samples of the bus shelter, and is updated according to the effective load-bearing status of the bus shelter. In one embodiment, the estimated remaining time can be determined according to the following formula:

[0048] in, For the first The estimated remaining time, in minutes, for each density mutation interval requiring additional capacity intervention; For the first A collection of bus shelters within a density abrupt change interval; For the first The additional capacity of each bus shelter is limited to a single person. For the first The current number of passengers at each bus shelter, in persons; For short-term prediction of passenger arrival rate, the unit is people / minute; This is the residual correction for arrival rate, expressed in people per minute, used to avoid denominator instability when the predicted arrival rate approaches zero. If the current number of passengers at a bus shelter has reached or exceeded the additional capacity intervention boundary, then the interval... A value of 0 indicates that the bus shelter needs to be immediately included in the capacity scheduling; if multiple bus shelters have not reached the boundary, the system takes the time required for the first bus shelter to reach the boundary as the interval. The numerator in the above formula is in people, the denominator is in people / minute, and the result is in minutes, which can be used as a quantitative input for subsequent time matching items.

[0049] On the other hand, the system calculates the available capacity of preceding vehicles according to their expected arrival times. For the same density abrupt change interval, the system will estimate the capacity available for vehicles arriving earlier than the first vehicle. The arriving vehicle is designated as the preceding vehicle, and its remaining capacity, door access capacity, estimated stop duration, number of passengers waiting at the target bus shelter, and service demand at subsequent stops are read for each vehicle. Based on the estimated number of passengers waiting at the bus shelter when the preceding vehicle arrives, the remaining capacity of the preceding vehicle, and the actual number of passengers that can be boarded, the system calculates the number of passengers waiting for the next vehicle. The number of passengers released by the preceding vehicle before the arrival of the vehicle. The unit is a person. The aforementioned It is not directly equal to the remaining capacity of the preceding vehicles, but is constrained by the available capacity of the preceding vehicles, the actual number of people waiting at the platform, the efficiency of passenger boarding and alighting, and the stopping time, so that the system can identify the state of "vehicles arriving but insufficient capacity release".

[0050] Once the vehicle arrival time, the continued accumulation of passengers, the time of additional capacity intervention, and the release volume of preceding vehicles are all determined, the system calculates the capacity demand correlation between the vehicles and the density abrupt change interval. In one implementation, the capacity demand correlation can be determined according to the following formula:

[0051] in, For the first The bus and the first The correlation between capacity demand between density abrupt intervals is a dimensionless quantity. For the first A collection of bus shelters within a density abrupt change interval; For the first The current number of passengers at each bus shelter, in persons; For short-term prediction of passenger arrival rate, the unit is people / minute; For the first The car arrived at the The estimated time for each bus shelter is in minutes. For the first The number of passengers expected to be released by the preceding vehicle before the arrival of the vehicle, in persons; For the first The current remaining passenger capacity of the vehicle, in persons; The remaining correction amount for the vehicle's remaining capacity, in person, is determined by the stable recognition resolution of the onboard passenger flow detection device. For the first The mutation intensity statistics of each density mutation interval are dimensionless and can be obtained from the robust average values ​​of mutation intensity of the interval entry edge, exit edge, and internal edge. For the first The estimated time, in minutes, for a vehicle to reach the location represented by the density abrupt change interval; For the first The estimated remaining time, in minutes, for each density mutation interval requiring additional capacity intervention; The time scale for normal vehicle service in this section is based on historical stable operation benchmarks, and the unit is minutes. The length of the scheduling cycle is in minutes.

[0052] In the above formula, the numerator of the fractional part represents the first... The estimated passenger demand, expressed in persons, represents the number of passengers a vehicle cannot release before it reaches its target segment; the denominator represents the vehicle's remaining passenger capacity, also expressed in persons. Therefore, this ratio is dimensionless. Multiplied by The aim is to incorporate the degree of spatial density abruptness into the vehicle matching process, so that intervals with more pronounced spatial abruptness and a higher probability of local congestion have a stronger correlation in vehicle scheduling. The exponential term represents the degree of matching between vehicle arrival time and the time when additional capacity demand occurs in the interval, where... The denominator represents the absolute time deviation between the arrival time of the vehicle and the time of increased demand. The time scale is used. Through this formula, the system unifies passenger accumulation, insufficient release of preceding vehicles, remaining vehicle capacity, spatial mutation intensity, and arrival time into the same correlation degree, thereby avoiding scheduling based solely on fixed departure intervals or the principle of closest vehicle distance.

[0053] Subsequently, the system creates a data structure for a spatiotemporal association table of vehicle capacity demand. This association table can be organized as a two-dimensional table, a set of key-value pairs, or a scheduling database record; each record corresponds to a service association between a candidate vehicle and a density mutation interval. The system writes the vehicle number, vehicle queue status, vehicle current location, current speed, remaining capacity, target density mutation interval number, interval start and end waiting shelters, interval mutation intensity, estimated time for the vehicle to arrive at the first waiting shelter in the interval, estimated time to arrive at the last waiting shelter in the interval, estimated number of new passengers in the interval before the vehicle's arrival, estimated release volume of preceding vehicles, estimated remaining time for additional capacity intervention, unreleased demand, capacity demand association degree, and data reliability into the association table. If the same vehicle can serve multiple density mutation intervals sequentially along the travel direction, the system generates multiple association records for that vehicle and retains the sequential service relationship between each record.

[0054] For example, a density abrupt change interval includes the third and fourth bus shelters, with current passenger numbers of 64 and 58 respectively, and short-term predicted passenger arrival rates of 4 passengers / minute and 3 passengers / minute respectively. A vehicle is expected to arrive at the third bus shelter in 8 minutes and the fourth bus shelter in 9 minutes, with a current remaining capacity of 42 passengers. Before this vehicle arrives, the preceding vehicle is expected to release 30 passengers at the third bus shelter and 20 passengers at the fourth bus shelter. The system calculates the unreleased demand before the vehicle's arrival. That is, approximately 131 people. If the additional capacity intervention boundary for the third waiting shelter is 90 people, then the estimated remaining time for the third waiting shelter to reach that boundary is... The estimated time to reach the fourth bus shelter is approximately 6.19 minutes; if the additional capacity intervention threshold for the fourth bus shelter is 82 people, then the estimated remaining time for the fourth bus shelter to reach that threshold is... The interval is approximately 7.50 minutes. The system takes the smaller of the two values ​​as the interval. If the statistical value of the mutation intensity in this interval is 0.8, the time it takes for the vehicle to arrive at the representative location in the interval is 8.5 minutes, and the sum of the normal service timescale and the scheduling cycle length is 12 minutes, then the system writes the vehicle's capacity coverage pressure, spatial mutation intensity, additional capacity intervention time, and time matching relationship into the vehicle capacity demand spatiotemporal correlation table. This example is only for illustrating the processing procedure and does not limit vehicle capacity, the number of waiting shelters, intervention boundaries, or specific time values.

[0055] S4. The bus dispatch control system first receives the spatiotemporal correlation table of vehicle capacity demand generated by S3, and performs a forward-looking assessment of the future passenger dwell time at each bus shelter within the density mutation interval based on the correlation table. Then, it divides the vehicles on the route into a queue waiting to depart and a queue in transit according to their operating status. After the vehicle queues are divided, the system generates an early departure instruction and a skip-stop driving instruction for the queue waiting to depart, and generates a shortened stop time instruction for the queue in transit. Under the condition that the safe operation boundaries of the front and rear vehicles and the service constraints of the upstream bus shelters are satisfied, the system outputs a bus dispatch plan.

[0056] Specifically, the system first establishes a future waiting population evolution sequence for each density mutation interval. The system reads the current number of passengers, short-term predicted passenger arrival rate, historical stable operating baseline, additional capacity intervention boundary, expected vehicle arrival time, and preceding vehicle release capacity for each waiting area within that interval. It then arranges vehicles capable of serving that interval into a vehicle service sequence according to their expected arrival time. Subsequently, starting from the current scheduling time, the system continuously extrapolates waiting population changes in scheduling cycles: within each scheduling cycle, the system first adds new passengers based on the short-term predicted passenger arrival rate, then deducts the number of releaseable passengers based on the remaining capacity of expected arriving vehicles, door capacity, and stop service strategy within that cycle, obtaining the expected waiting population for each waiting area at the end of that cycle. The system determines the passenger dwell time corresponding to that density mutation interval based on the time required for the expected waiting population to fall back to the stable envelope formed by the historical stable operating baseline, or the time required for currently waiting passengers to be released by the vehicle service sequence.

[0057] The system then divides vehicles into a waiting queue and an in-transit queue. The waiting queue includes vehicles located at the originating station, parking lot, temporary dispatch point, or other accessible location on the route, whose vehicle status, driver scheduling, energy status, and route access status all meet operational requirements. The in-transit queue includes vehicles that have already entered the route and can continue to serve the bus shelters in the current direction of travel. For in-transit vehicles, the system further determines their position relative to the target density abrupt change interval: if the vehicle is upstream of the interval, it can arrive at the interval earlier by shortening the stop time at low-density bus shelters or optimizing the service rhythm; if the vehicle is within the interval, it prioritizes ensuring sufficient passenger boarding at the bus shelters within the interval; if the vehicle is downstream of the interval and cannot return to serve the interval within the current dispatch cycle, it will not participate in the direct dispatch of that interval.

[0058] For vehicles in the departure queue, the system calculates the arrival time of the vehicle in the target density mutation zone under two scenarios: departure as originally planned and departure ahead of schedule. It also calculates the difference in passenger dwell time in the target zone under these two scenarios. If early departure allows the vehicle to arrive in the target zone shortly before or after reaching the additional capacity intervention boundary, and the headway between the vehicle and the preceding and following vehicles meets the safe interval formed by the road operating boundary, platform berth capacity, vehicle braking capacity, and signal control conditions, the system identifies this vehicle as a candidate for early departure. Subsequently, the system sorts all early departure candidate vehicles according to the reduction in dwell time, remaining capacity, arrival time matching degree, and disturbance to subsequent departure plans, selecting the vehicle that best matches the demand of the target zone to generate an early departure instruction. The early departure instruction includes the vehicle number, departure time, target density mutation zone, set of waiting shelters requiring priority service, suggested driving speed range, safety interval verification results for preceding and following vehicles, and the effective time of the instruction.

[0059] When generating a skip-stop driving instruction, the system starts from the departure position of the vehicle to be dispatched and sequentially traverses the upstream bus shelters along the vehicle's direction of travel within the target density mutation interval. For each upstream bus shelter, the system reads the bus shelter's cumulative passenger density index, historical stable operating baseline, current number of waiting passengers, additional capacity intervention boundary, in-vehicle alighting requests, station attributes, estimated arrival time of subsequent vehicles, and remaining capacity of subsequent vehicles. If the bus shelter is within the stable envelope formed by the historical stable operating baseline, the current number of waiting passengers can be accommodated by subsequent vehicles within the normal service time scale, there are no alighting requests at this station, and the bus shelter is not a transfer hub, school / hospital, traffic control support station, or a mandatory stop marked by the operating system, the system marks the bus shelter as a skip-stop bus shelter. If the cumulative passenger density index of the bus shelter continuously deviates from the historical stable operating baseline, the current number of waiting passengers approaches or reaches the additional capacity intervention boundary, the remaining capacity of subsequent vehicles is insufficient, there are alighting requests on the vehicle, or the station has mandatory service attributes, the system marks the bus shelter as a mandatory stop bus shelter. Once all upstream bus shelters have been traversed, the system generates a skip-stop driving instruction based on the set of skippable bus shelters. This instruction includes the vehicle number, the number of the bus shelter that can be skipped, the corresponding alternative service vehicle, the estimated arrival time of the alternative service vehicle, the valid skip-stop route, and the conditions for canceling the instruction. If, before a vehicle approaches a skippable bus shelter, the passenger flow at that shelter deviates continuously, the current number of waiting passengers reaches the additional capacity intervention boundary, a passenger requests to disembark, or the alternative service vehicle is delayed, the system cancels the skip-stop arrangement for that bus shelter.

[0060] For vehicles in the queue, the system first determines the route distance from the vehicle's current location to the target density mutation interval and the set of intermediate waiting shelters. Then, it sequentially reads the current number of passengers, cumulative passenger density index, estimated number of passengers boarding and alighting, passenger alighting requests, platform passenger approach to vehicle status, remaining vehicle capacity, and whether the waiting shelter belongs to the density mutation interval. If an intermediate waiting shelter is in a low-density stable state, and passenger flow detection at the door area shows that boarding and alighting have been completed, no passengers are lingering in the door safety area, no new passengers are approaching the vehicle on the platform, and no passengers are requesting to alight at that station, the system marks the waiting shelter as a shelter with a shortened stop. If a waiting shelter belongs to the density mutation interval, or if there are still boarding and alighting passengers, passenger alighting requests, or platform passenger approach to vehicle status at the waiting shelter, the system marks the waiting shelter as a shelter maintaining normal stopping. After traversing all intermediate waiting shelters, the system generates candidate instructions for shortening the stop time for waiting shelters with shortened stops.

[0061] Before issuing a candidate command to shorten the stop duration, the system performs a safety check on the spatiotemporal position of the preceding vehicle. The system reads the current vehicle's speed, the speed of the preceding vehicle, the distance between the two vehicles on the route mileage coordinates, the estimated arrival time of both vehicles at the next bus shelter, vehicle braking performance, positioning error, communication delay, road speed limit, number of bus stop spaces, and traffic signal status, and based on this, forms a safe operating boundary for the current road segment. In one implementation, the minimum safe distance within the safe operating boundary can be determined according to the following formula:

[0062] in, Let $\{j\}$ be the minimum safe distance that the $j\th vehicle needs to maintain relative to the vehicle in front, in meters. This is the current speed of the vehicle, in meters per second. This is the safe braking deceleration that the vehicle can achieve under the current road conditions, expressed in meters per second squared. It can be obtained by correcting for vehicle braking performance calibration data, road surface wetness, and vehicle load conditions. The scheduling communication delay, measured in seconds, is determined by the round-trip time of the most recent communication between the vehicle terminal and the scheduling platform. The location refresh delay, measured in seconds, is determined by the sampling period of the vehicle positioning device. The length occupied by the next waiting shelter platform berth or the safe length for vehicles to park, in meters; The distance for compensation of positioning error is in meters and is estimated from the calibration error of the positioning equipment or the real-time positioning accuracy. The distance for compensating for road traffic disturbances is measured in meters and can be determined by fluctuations in current road congestion, changes in signal queue length, or changes in road speed limits.

[0063] The system records the actual distance between this vehicle and the vehicle in front on the route mileage coordinates as: If this train shortens its stops, the estimated stop at the next section of road and the next bus shelter will be... Still not less than If the next bus shelter platform berth is not occupied by the preceding vehicle, the system will confirm the candidate command to shorten the stop duration as an executable command. If the preceding vehicle has not yet left the next bus shelter, the road ahead is congested causing a reduction in vehicle distance, the positioning error is too large, or there are insufficient platform berths, the system will not issue a command to shorten the stop duration, but will maintain the original stop service or generate a waiting command. This safety check uses vehicle kinematics, communication status, positioning error, and platform berth conditions to jointly constrain the service rhythm adjustment of vehicles en route, avoiding treating "shortening the stop" as a simple experience-based acceleration.

[0064] Subsequently, the system resolves conflicts between early departure instructions, skip-stop instructions, and shortened stop duration instructions. If the same vehicle simultaneously fulfills multiple scheduling actions, the system determines the final instruction combination based on the reduction in passenger dwell time in the target density mutation interval, the utilization rate of the vehicle's remaining capacity, the impact of upstream waiting shelter services, the interval between preceding and following vehicles, and the safe operating boundary. If the simultaneous early departure of two vehicles would lead to excessive vehicle density at the origin, the system prioritizes the vehicle with a better arrival time in the target interval, higher remaining capacity, and less disruption to subsequent departure plans, with the other vehicle serving as a backup or entering the line according to the updated departure interval. If a skip-stop instruction would cause a new passenger accumulation trend at the upstream waiting shelter in subsequent scheduling cycles, the system cancels the skip-stop arrangement and replaces it by adjusting the departure of subsequent vehicles or shortening the stop duration at low-density waiting shelters.

[0065] Once all dispatch instructions are finalized, the system creates a data structure for the bus dispatch plan. This plan includes vehicle number, vehicle queue type, instruction category, execution start time, target density mutation interval, affected bus shelter set, suggested driving speed range, permitted skip-stop bus shelters, mandatory stop bus shelters, alternative service vehicles, expected dwell time changes, preceding vehicle's spatiotemporal position safety verification result, instruction validity time, and instruction cancellation conditions. The system sends the bus dispatch plan to the bus dispatch platform, onboard terminal, or driver notification system. In the next dispatch cycle, it continues to receive new bus shelter passenger numbers, vehicle locations, and vehicle service events, re-executing S1 to S4, ensuring that bus departure intervals, stop strategies, and on-route operation rhythm are continuously updated in real-time according to the passenger density at the bus shelters and their cumulative trends.

[0066] For example, a certain bus route experiences a density abrupt change between the third and fourth waiting shelters. The spatiotemporal correlation table of vehicle capacity demand shows that the nearest en route vehicle will take 8 minutes to reach this section, and there is insufficient remaining capacity. A vehicle waiting to depart from the originating station can immediately enter the route and arrive at the section in 6 minutes. The cumulative passenger density indices of the first and second waiting shelters are both within the stable envelope formed by historical stable operating benchmarks. Subsequent vehicles are expected to arrive within the normal service timescale with sufficient remaining capacity, and there are no requests to disembark from the waiting vehicle. Based on this, the system generates an early departure instruction for the waiting vehicle, includes the first and second waiting shelters in the skip-stop passage set, and records the corresponding alternative service vehicle and cancellation conditions. For another vehicle en route located upstream of the target section, the system detects that the vehicle preceding it has already left the next bus shelter. Both vehicles meet the safe operating boundaries formed by braking capacity, positioning error, communication delay, and platform berth capacity at the next road segment and the next platform. Furthermore, the passenger pick-up and drop-off process at the low-density bus shelter where the en route vehicle is currently stopped has been completed. Therefore, a command to shorten the stop duration is generated. This example is only used to illustrate the generation process of the bus dispatching scheme and does not limit the route structure, number of vehicles, number of bus shelters, or specific time values.

[0067] In summary, the bus dispatching method based on passenger flow data at bus shelters, as described in this invention, is explained. It constructs a passenger density accumulation index that integrates current congestion levels, arrival rate deviations, and continuous accumulation levels. This index is then coupled with the effective carrying capacity of bus shelters and historical stable operating benchmarks for dimensionless quantification, thereby achieving a refined perception of passenger flow accumulation trends at each bus shelter. This allows for accurate identification of density abrupt change zones along the route, where a step shift occurs from low-density to high-density areas. This overcomes the shortcomings of traditional dispatching methods that rely on overall or cross-sectional passenger flow for control, resulting in insufficient spatial targeting.

[0068] Here, those skilled in the art will understand that the specific operations of each step in the bus dispatching system based on bus shelter passenger flow data have been referenced above. Figures 1 to 3 The bus dispatching method based on bus shelter passenger flow data has been described in detail, and therefore, its repeated description will be omitted.

[0069] In summary, the bus dispatching system based on passenger flow data at bus shelters, as described in this invention, is explained. It constructs a passenger density accumulation index that integrates current congestion levels, arrival rate deviations, and continuous accumulation levels. This index is then coupled with the effective carrying capacity of bus shelters and historical stable operating benchmarks for dimensionless quantification, thereby achieving refined perception of passenger flow accumulation trends at each bus shelter. This allows for accurate identification of density abrupt changes along the route, transitioning from low-density to high-density zones, overcoming the spatial targeting deficiencies of traditional dispatching methods that rely on overall or cross-sectional passenger flow for control.

Claims

1. A bus dispatching method based on bus shelter passenger flow data, characterized in that, include: The system acquires real-time passenger flow data and historical stable operation benchmarks for each bus shelter, and couples the current level of congestion, the degree of deviation in arrival rate, and the degree of continuous accumulation to generate a passenger density accumulation index. The passenger density cumulative index is sorted along the line stations, and density abrupt change intervals are determined by combining the distance between stations and historical spatial gradient benchmarks to generate a density gradient distribution map. Based on the density gradient distribution map, real-time vehicle location, remaining capacity, and capacity released by preceding vehicles, the unreleased demand of vehicles arriving at the density abrupt change interval is estimated, and a spatiotemporal correlation table of capacity demand is generated. Based on the spatiotemporal correlation table of the capacity demand, the passenger dwell time is prospectively assessed, and instructions to depart early or skip stops are generated for the queues to be departed, and instructions to shorten stops are generated for the queues in transit, thus obtaining a bus dispatching scheme.

2. The bus dispatching method based on bus shelter passenger flow data according to claim 1, characterized in that, The generation of the passenger density cumulative index includes: The effective capacity of a bus shelter is determined based on its currently available waiting area and allowable standing density. Based on the arrival time of the vehicle at the waiting pavilion, the opening time of the door, the closing time of the door, and the departure time of the vehicle, the decrease in the number of passengers during the period from the opening to the closing of the vehicle is marked as the vehicle service release amount, and the increase in the number of passengers after the vehicle leaves the station is included in the natural passenger arrival amount. The current passenger arrival rate is determined based on the natural passenger arrival volume, and the passenger density cumulative index is calculated by combining the baseline passenger arrival rate and baseline passenger number in the historical stable operation benchmark.

3. The bus dispatching method based on bus shelter passenger flow data according to claim 2, characterized in that, The passenger density cumulative index is determined according to the following formula: ； in, Let \(i\) be the cumulative passenger density index of the \th (i\)th bus shelter. For the current number of passengers, In order to effectively accommodate the number of people, For the current passenger arrival rate, Based on the passenger arrival rate, To reach the rate residual correction, Based on the number of passengers, The length of the sliding time window. , and For weights.

4. The bus dispatching method based on bus shelter passenger flow data according to claim 3, characterized in that, The determination of the weights includes: The current congestion characteristics, arrival rate deviation characteristics, and continuous accumulation characteristics in the historical scheduling samples are used as input features, and the actual retention intensity after the corresponding scheduling cycle is used as the calibration target. With the objective of minimizing the deviation between the predicted retention intensity and the actual retention intensity, , and Perform constraint calibration.

5. The bus dispatching method based on bus shelter passenger flow data according to claim 4, characterized in that, The determination of the intensity of the density abrupt change includes: The cumulative passenger density index of each bus shelter is smoothed with edge preservation to obtain the cumulative passenger density index after edge preservation. The difference in passenger density cumulative index between adjacent bus shelters is scaled according to the distance between stations; The scaling results are compared with the scaling gradient center values ​​in the historical stable operation benchmark, and the density mutation intensity is obtained based on the gradient discrepancy and the spatial trend confidence coefficient. The intensity of the density abrupt change is determined according to the following formula: ； in, The intensity of density abrupt change between adjacent bus shelters. and The cumulative passenger density index after edge protection smoothing. This represents the actual distance between stations. This is the reference distance between stations on the line. For the scaled gradient center value, For gradient discrete quantities, This is the gradient residual correction amount. This represents the spatial trend confidence coefficient.

6. The bus dispatching method based on bus shelter passenger flow data according to claim 5, characterized in that, The generation of the spatiotemporal correlation table of vehicle capacity demand includes: Based on the real-time location of buses, route topology, travel speed, and estimated stop time at intermediate bus shelters, the estimated arrival time of each bus shelter within the density change interval is determined. The short-term predicted passenger arrival rate is determined based on the current passenger arrival rate and the baseline passenger arrival rate in the historical stable operation baseline. Based on the current number of passengers at the bus shelter within the density mutation interval, the short-term predicted passenger arrival rate, and the additional capacity intervention boundary, determine the estimated remaining time when additional capacity intervention is required in the density mutation interval. The expected release capacity of the preceding vehicles is determined based on their expected arrival order, remaining capacity, and station service capacity. The unreleased passenger demand before the vehicle arrives, the remaining vehicle capacity, the intensity of density mutation intervals, and the estimated remaining time are written into the spatiotemporal correlation table of vehicle capacity demand.

7. The bus dispatching method based on bus shelter passenger flow data according to claim 6, characterized in that, The generation of the instruction to shorten the stop time includes: Determine the intermediate bus shelters between the current location of vehicles en route and the density abrupt change zone; When the intermediate waiting shelter is in a low-density stable state, and the boarding and alighting flow in the door area has been completed, there are no passengers staying in the door safety area, no new passengers approaching the vehicle on the platform, and no passengers requesting to alight at this station, the intermediate waiting shelter is identified as a waiting shelter with shortened stops. Before issuing an instruction to shorten the stop time, the minimum safe distance is determined based on the vehicle's speed, braking deceleration, communication delay, positioning refresh delay, platform berth length, positioning error, and road operation disturbance. The instruction to shorten the stop time can only be executed if the expected actual distance between the vehicle and the vehicle in front is not less than the minimum safe distance and the next waiting shelter platform berth is not occupied by the vehicle in front.

8. A bus dispatching system based on bus shelter passenger flow data, characterized in that, include: The passenger flow status perception module is used to obtain real-time passenger flow data and historical stable operation benchmarks for each bus shelter, and couple the current congestion level, the degree of deviation of arrival rate and the degree of continuous accumulation to generate the passenger density accumulation index for each bus shelter. The density gradient analysis module is used to arrange the passenger density cumulative index along the line stations in sequence, combine the distance between adjacent waiting shelters and the historical spatial gradient benchmark to determine the density change interval, and generate a waiting shelter density gradient distribution map. The capacity demand projection module is used to project the unreleased passenger demand when each bus arrives at the density change interval based on the bus shelter density gradient distribution map, real-time location of buses, remaining capacity and capacity released by preceding vehicles, and generate a spatiotemporal correlation table of vehicle capacity demand. The scheduling decision and instruction generation module is used to prospectively assess the passenger dwell time in the waiting area within the density change interval based on the spatiotemporal correlation table of vehicle capacity demand, and generate early departure instructions or skip-stop instructions for the queue to be departed, and generate instructions to shorten the stop time for the queue in transit, and output the bus dispatching plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the bus dispatching method based on bus shelter passenger flow data as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the bus dispatching method based on bus shelter passenger flow data as described in any one of claims 1 to 7.

Images