A chain line design method for a reservation type bus gathering and distributing station

Abstract

The application discloses a chain line design method for a reservation type bus gathering and distributing station, and belongs to the technical field of public traffic scheduling. The method first maps a reservation request into a station pair demand, constructs a reservation service feasible directed chain graph through multi-dimensional mutual feedback chain edge weights and comprehensive costs, generates a main chain skeleton, and expands stations based on an insertion gain. A rolling time domain mechanism is used to divide a locked section and an unlocked section, and new reservation demands are adaptively classified. Finally, chain edge weights are dynamically updated according to operation feedback. The application can efficiently absorb new demands on the premise of ensuring stable confirmed services, improve line operation adaptability, and is suitable for dynamic reservation bus scenes such as commuting and connection.

Description

Technical Field

[0001] This invention belongs to the field of urban public transportation operation scheduling and demand-responsive bus route planning technology. Specifically, it relates to a method for designing a chain of reserved bus hub stations, which is applicable to bus service scenarios where reservation demands are continuously and dynamically updated, such as peak commuting, park shuttle, medical shuttle, and event evacuation. Background Technology

[0002] Reservation-based public transport is a public transport service model that organizes and operates based on passengers' pre-submitted travel needs. It combines the advantages of traditional public transport, such as large capacity and low cost, with the flexibility of customized services, and is widely used in scenarios such as commuter connections and travel for special groups.

[0003] Existing reservation-based bus route designs typically involve sequentially collecting demand, selecting stops, generating routes, and dispatching vehicles, which works well in scenarios with small and stable demand. However, in scenarios with dense and continuously increasing demand, such as peak commuting and large-scale event evacuations, existing technologies have the following drawbacks:

[0004] Firstly, every new booking request requires a new global route calculation, which is computationally intensive and can lead to frequent changes in stations and arrival times that have already been confirmed with passengers, reducing service credibility and passenger experience.

[0005] Secondly, the evaluation of station connectivity only takes distance and travel time as the core criteria, without considering the demand-carrying capacity between stations, the continuity of time windows, the vehicle capacity turnover efficiency, the continuity of downstream services, and the chain-like feedback relationship of historical passengers. As a result, the generated route operation adaptability is poor, and problems such as long-term vehicle capacity occupation and arrival time deviating from passenger expectations are likely to occur.

[0006] Third, the dynamic scheduling lacks a tiered adaptation mechanism, making it impossible to absorb new demands in a tiered manner while ensuring the stability of confirmed reservation services. This results in either excessive disruption to existing lines or rejection of too many new demands, making it difficult to balance service coverage and operational stability. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a chain-like route design method for pre-booked bus hubs and distribution stations. This method solves problems such as unclear route generation rules, lack of calculable basis for station connection relationships, difficulty in incorporating passenger reservation feedback into route design, susceptibility to global route reconstruction during dynamic scheduling, and insufficient stability of confirmed reservation services. This method is applicable to bus service scenarios where passengers submit reservation requests in advance, the operating system confirms services before departure, and new reservation requests may still be received during vehicle operation.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A chain-like route design method for reservation-based public transport hubs is disclosed. This method takes a reservation request set, a candidate station set, a vehicle set, and road travel time data as input. The reservation request set describes the passenger's desired origin and destination, reservation service time window, number of passengers, maximum acceptable walking time, maximum acceptable detour rate, and current confirmation status. The candidate station set describes the station location, available stopping directions, service radius, and stop service time. The vehicle set describes the vehicle capacity, current location, current service chain, and dispatchability status. The road travel time data describes the travel time, distance, and traffic restrictions between candidate stations during different time periods.

[0010] During the reservation request standardization phase, the system represents any reservation request as a data item containing the passenger's desired origin, desired destination, reservation service time window, number of passengers, walking restrictions, detour restrictions, and reservation service status, which can be represented as:

[0011]

[0012] In the formula, and These represent the passenger's expected origin and expected destination, respectively. and These represent the lower and upper limits of the appointment service time window, respectively. Indicates the number of passengers. Indicates the maximum acceptable walking time. This indicates the maximum acceptable detour rate. This indicates the reservation service status. The reservation service status includes: not matched, matched, confirmed, boarded, completed, cancelled, and reservation not fulfilled. Confirmed requests are subject to the locking rules, while unconfirmed requests can participate in partial adjustments within the unlocked segment.

[0013] During the candidate station mapping phase, the system filters the set of candidate boarding stations based on passengers' acceptable walking time and station stopping direction. and candidate drop-off stations Candidate stations should simultaneously meet the walking time threshold and vehicle stopping direction conditions. If the candidate boarding station set or candidate alighting station set is empty, the reservation request will not be included in this round of chain route generation; if multiple candidate station pairs exist, the default station pair and alternative station pairs will be retained for replacement later if time windows, capacity, or detour constraints are not met.

[0014] For candidate site pairs with the same reservation request, the system determines the default site pair based on the principle of minimizing the service cost of the site pair, and retains a preset number of alternative site pairs in ascending order of service cost. The service cost of a site pair can be expressed as:

[0015]

[0016] In the formula, This indicates the walking time for passengers from their desired origin to the candidate boarding station. This indicates the road travel time between stations within a given time period. This indicates the walking time from the candidate drop-off point to the desired destination. Indicates expected waiting penalty, Indicates the acceptance rate of historical appointments. to This is a service cost weighting coefficient. The smaller the cost, the more suitable the corresponding site pair is as the default pick-up and drop-off combination for this request.

[0017] After generating demand pairs between stations, the system calculates passenger volume, passenger drop-off volume, unserved reservation demand, available service time windows, and unidirectional demand intensity for each candidate station according to service hours, and treats the candidate stations as chain-like route nodes. Directed connections between candidate stations are called candidate chain edges. Whether a candidate chain edge is retained is determined by reservation demand acceptance, time window alignment, passenger drop-off and pick-up service balance, downstream continuous service, and historical reservation feedback.

[0018] Candidate Sites To candidate sites During the period The weight of the feed-in chain edge is used to represent the contribution of the chain edge to the reservation-based continuous bus service, and can be expressed as:

[0019]

[0020]

[0021] In the formula, To normalize the site's demand intensity, To ensure seamless connection between appointment service time windows, To balance passenger pick-up and drop-off services, For downstream continuous service, Provide feedback on passenger acceptance; to This is a non-negative weighting coefficient. This weight is used to unify the scale of reservation demand, the availability of time windows, vehicle capacity turnover, the service capacity of subsequent stations, and historical passenger feedback into the same evaluation chain.

[0022] During the chain-side cost calculation phase, the system combines road travel time, road network distance, stop service time, published reservation route disturbance penalties, turnaround or detour penalties, and feedback chain-side weights to obtain the comprehensive chain-side cost, which can be expressed as:

[0023]

[0024] In the formula, Indicates the road travel time between stations. Indicates the distance between stations on the road network. Indicates the estimated downtime of the station. This indicates the penalty for any disturbance caused by connecting this link to the already booked route. This indicates a penalty for turning back or taking a detour. Indicates the weight of the feed-in chain edge. to This represents the cost weighting coefficient. The higher the weight of the feed-in chain edge, the lower the overall cost of the chain edge; the greater the running time, downtime, or line disturbance, the higher the overall cost of the chain edge.

[0025] During the feasible directed chain graph construction phase, the system will set up candidate sites. As a set of nodes, candidate chain edges that satisfy the chain-like advancement direction and the feasibility conditions of reservation services will be added to the edge set. This forms a feasible directed chain graph for the appointment service:

[0026]

[0027] The edge set can be represented as:

[0028]

[0029] The feasibility conditions for the reservation service include passenger reservation time windows, maximum detour rate, vehicle capacity, and station stopping directions.

[0030] During the route generation phase, the system determines the starting and ending anchor points based on the net distribution intensity of stations, the main direction of demand, and station accessibility. The minimum total cost path for each link from the starting anchor point to the ending anchor point is calculated to obtain the main chain skeleton:

[0031]

[0032] Once the main chain skeleton is determined, the system calculates the insertion gain between candidate sites and adjacent sites in the main chain for reservation requests that have not yet been served.

[0033] Site to be inserted Insert adjacent sites within the chain and The insertion gain between them can be expressed as:

[0034]

[0035] In the formula, This indicates the number of new service reservations that can be added through the site to be inserted. This represents the change in the mutual feedback weights of the chain edges before and after the insertion. Indicates the newly added running time. This indicates the new stop time. This indicates a restriction on the reservation service or a confirmed penalty for disrupting the reservation. to This is the insertion gain weighting coefficient. When the insertion gain is greater than the preset threshold and does not violate the feasibility conditions of the reservation service, the system will add the site to be inserted to the corresponding chain segment and synchronously update the arrival time, departure time, number of passengers in the vehicle, and request binding relationship; otherwise, it will try the next candidate site pair or generate a sub-chain.

[0036] During the rolling scheduling phase, the system divides the chained route into locked and unlocked segments based on vehicle location, reservation confirmation time, and passenger notification status. Locked segments include the sequence of stations where vehicles have arrived, are about to arrive, or have confirmed with passengers; the station order and service hours are generally not changed. Unlocked segments include the sequence of stations that have not yet been served and have not yet been finally confirmed, and are used to absorb new reservation requests. For new reservation requests, the system selects a scheduling method in the following order: direct absorption within the chain, partial insertion into the unlocked segment, branch-link connection, and reconstruction of the unlocked tail chain.

[0037] During the operational feedback update phase, the system updates the weights of the mutual feedback links based on the actual passenger acceptance rate, service completion rate, delay penalties, and reservation non-fulfillment rate after operation, which can be represented as:

[0038]

[0039] In the formula, For passenger acceptance rate, To improve service completion rate, In order to delay punishment, To determine the booking non-fulfillment rate, To update the step size, to The weighting coefficients are used for feedback. If a chain edge experiences a long-term history of passenger rejections, service delays, or unfulfilled reservations, its subsequent weight will decrease; if a chain edge operates stably and has a high passenger acceptance rate, its subsequent weight will increase.

[0040] Through the above processing, this invention incorporates reservation request mapping, station-to-service cost, feedback chain edge weights, chain edge comprehensive cost, feasible directed chain graph for reservation services, main chain and sub-chain generation, rolling adjustment of locked and unlocked segments, and operational feedback updates into the same route design process. Compared with methods that generate routes solely based on shortest path or ordinary vehicle path models, this method can process new reservation requests while maintaining the stability of confirmed reservation services, and allows subsequent route design to gradually reflect passenger reservation behavior and road operation status within the target service area.

[0041] The present invention has the following advantages and beneficial effects:

[0042] 1. Ensure service stability and improve scheduling efficiency: By adopting a rolling time domain and locking segment mechanism, the station sequence of confirmed services is locked, and new requests are processed only in the unlocked segment. This avoids frequent global recalculation, which reduces the amount of computation, ensures the service expectations of confirmed passengers, and improves the passenger experience.

[0043] 2. Multi-dimensional feedback evaluation with strong route adaptability: The system constructs a feedback chain weight that includes demand acceptance, time window connection, capacity balance, downstream services, and passenger feedback. This breaks through the limitation of evaluating station connections only based on distance / time, making the routes more in line with the characteristics of reservation bus ride-sharing operations and improving vehicle turnover efficiency and service punctuality.

[0044] 3. Tiered adaptation to new demands, balancing coverage and stability: A four-level adaptation strategy is set up for new reservation requests, prioritizing the absorption of demands in the way with the least disturbance, and only performing partial reconstruction when necessary, so as to maximize service coverage while ensuring the overall stability of the line.

[0045] 4. Feedback-based iterative optimization and continuous adaptation to operations: By dynamically updating the weights of the chain edges through actual operational data, the route design can autonomously adapt to regional passenger flow patterns, road conditions and passenger behavior characteristics, thereby continuously improving operational effectiveness. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the overall process of the method described in this invention;

[0047] Figure 2 A schematic diagram illustrating the composition of the weights of the feed-through chain edges;

[0048] Figure 3 A schematic diagram of the main chain, sub-chains, and connecting chains;

[0049] Figure 4 This is a schematic diagram illustrating the updates of locked and unlocked segments in the rolling time domain;

[0050] Figure 5 A flowchart illustrating the process of hierarchical adaptation for new requests. Detailed Implementation

[0051] The present invention will be further described below with reference to the embodiments. It should be noted that these are merely examples and descriptions of the inventive concept. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in the claims, they should all be considered to fall within the protection scope of the present invention.

[0052] This invention takes as its implementation premise a candidate site set with attributes such as location coordinates, road lateral orientation, permitted stopping direction, service radius, expected stop service time and available service period. It can be provided by site selection methods such as bus stops, temporary stops, park gates, hospital outpatient entrances, school gates, rail station entrances and exits, and main entrances and exits of communities.

[0053] Before the system goes live, four types of data need to be prepared. The first type is reservation request data, including fields such as passenger origin, destination, reservation submission time, expected departure or arrival time window, reservation confirmation deadline, number of passengers, maximum acceptable walking time, maximum acceptable detour rate, whether ridesharing is accepted, whether station replacement is allowed, and current confirmation status. The second type is candidate station data, including station number, latitude and longitude or projected coordinates, road lateral direction, types of vehicles that can stop, whether temporary stops are allowed, estimated stop time, and historical passenger acceptance rate. The third type is vehicle data, including vehicle number, seating capacity, current location, current passenger load, current service chain, whether it is dispatchable, whether it can be modularly disassembled, and vehicle operating costs. The fourth type is road operation data, including the operation time matrix between candidate stations, road network distance matrix, congestion status, and rules prohibiting turning or stopping.

[0054] The "chain-type passenger flow route" referred to in this invention refers to a route structure in which vehicles depart from one or more upstream passenger pick-up points, sequentially serve several intermediate stations along the main demand direction, and arrive at one or more downstream evacuation points. A chain-type route can be a single main chain or composed of a main chain and several sub-chains. The main chain handles the main passenger flow direction, while the sub-chains are used to absorb localized new demand or provide connections to the main chain.

[0055] Let V be the set of candidate stations, R be the set of reservation requests, and B be the set of vehicles. The service period is divided into several rolling windows, and a set of chained route plans is generated or updated within each window. If the operator prioritizes route stability, the rolling window can be appropriately increased; if the operator prioritizes response speed, the rolling window can be appropriately shortened.

[0056]

[0057] in, Represents the set of candidate sites. The total number of candidate sites; This represents a set of reservation requests. This represents the total number of reservation requests. Indicates a collection of vehicles. The number of vehicles that can be dispatched; This indicates a service period or a rolling calculation window. The element numbers in the above set are only used for system calculations and are not limited to the specific format of actual site numbers, order numbers, or vehicle numbers.

[0058] Step S1: Standardize appointment requests;

[0059] The system first converts all appointment requests into a unified data structure:

[0060]

[0061] in, and These are the passenger's expected origin and destination, respectively. and The lower and upper limits of the service time window, For the number of passengers, For the maximum acceptable walking time, To achieve the maximum acceptable detour rate, The status is "request".

[0062] The reservation service status is used for rolling time-domain updates. Unmatched requests can participate in the whole-chain design; matched but unconfirmed requests can be adjusted within the unlocked segment; confirmed requests generally do not change their boarding station, alighting station, and estimated service time unless the passenger reconfirms; boarded requests only adjust the downstream evacuation segment that has not yet arrived; completed, canceled, and unfulfilled reservation requests no longer participate in the generation of the current route, but are used as reservation feedback data to update the weights of the mutual feedback chain edges.

[0063] The road operation time matrix is ​​updated in time periods, denoted as T. If a station pair is unreachable due to road lateral movement, no-stopping rules, or road connectivity, then the station pair is set as unconnectable and will not participate in candidate chain edge generation. The system also determines the chain advancement direction based on road direction, main origin-destination (OD) direction, and operation rules to minimize the need for repeated back-and-forth turns during route generation.

[0064] Vehicle status is uniformly represented as:

[0065]

[0066] in, The vehicle's current location. For vehicle capacity, The current number of people in the vehicle. For the current service chain of the vehicle, Vehicle status includes at least the following: idle, waiting to depart, en route, fully loaded, unschedulable, splittable, and returned to the depot.

[0067] Step S2: Appointment Request - Candidate Site Mapping;

[0068] For each appointment request The system first searches for candidate boarding and alighting stations that meet the walking time requirements.

[0069] The set of candidate boarding stations is as follows:

[0070]

[0071] The set of candidate alighting stations is as follows:

[0072]

[0073] in, Indicates a reservation request The corresponding set of candidate boarding stations, Indicates a reservation request The corresponding set of candidate drop-off stations; This indicates that passengers are from the expected origin. The time to walk to candidate boarding station i This indicates that the passenger walks from the candidate drop-off point j to the desired destination. Time; or This indicates that the station is allowed to stop under the current vehicle direction and road lateral conditions. If the value is 0, it means that the station cannot be used as a candidate pick-up or drop-off station for this request.

[0074] If the candidate boarding station set or the candidate alighting station set is empty, it means that the request cannot be directly served under the current candidate station set. In this case, the passenger can be prompted to expand their acceptable walking range, or the request can be marked as requiring temporary station service, or it can be postponed to the next scrolling window for manual processing by the operator or an external station module. Since the focus of this invention is not on the generation of temporary stations, such requests do not affect the main process of chain route design.

[0075] If multiple candidate site pairs exist, the system determines the default site pair based on the principle of minimizing the service cost of the site pair. Specifically, the system first calculates the service cost of each candidate site pair:

[0076]

[0077] in, This represents the station pair service cost generated when request r selects candidate boarding station i and candidate alighting station j. For time period The road travel time for a vehicle to travel from station i to station j; To account for potential waiting penalties that may result from adopting this station; For passengers' station pairings in historical orders Acceptance rate; to This is a service cost weighting coefficient, which can be set based on passengers' sensitivity to walking time, waiting time, travel time, and station acceptance. The smaller this value, the more suitable the station is as the default pick-up and drop-off station combination for request r.

[0078] The system selects the site pair with the lowest service cost as the default site pair, while retaining several alternative site pairs with lower costs. This process allows the system to continue trying alternative site pairs if inserting the default site pair into the main chain would cause the time window or reservation service stability to be compromised, thus reducing the workload of manually reassigning sites.

[0079] After mapping is completed, the system calculates the demand intensity, passenger volume, passenger alighting volume, and unserved demand for each time period. For requests where both boarding and alighting occur within the same small area, they are not simply merged into a single point. Instead, multiple candidate site pairs are retained, and subsequent feedback chain edge calculations determine which connection method is more suitable.

[0080] Step S3: Calculation of mutual feedback chain edge weights and overall cost;

[0081] The weighting of feed-back links is what distinguishes this solution from ordinary shortest path or ordinary vehicle routing problems. Ordinary route design primarily relies on the distance or time between two stations, while this solution also considers whether the connection is beneficial to subsequent services. A link from station i to station j is more suitable if it can handle more direct demand, has a smooth time window connection, releases vehicle capacity at station j, continues to serve downstream demand from station j, and has historically had good passenger acceptance.

[0082] For the site To the station During the period The weight of the feed-in chain edge is defined as:

[0083]

[0084] in, To normalize the site's demand intensity, To ensure seamless service time window connectivity, To balance passenger boarding and alighting capacity, For downstream continuous service, Provide feedback on passenger acceptance. to The weight coefficients of each component in the mutual feedback chain edge weights are all non-negative numbers and satisfy the following conditions: The weights of each item can be adjusted according to the service scenario.

[0085] Demand intensity for normalized sites can be calculated using the following formula:

[0086]

[0087] in, This represents the demand for direct site pairs from candidate site i to candidate site j within a given time period. This represents the maximum demand among all candidate site pairs within the same time period; A very small positive number is set to prevent the denominator from being zero. After normalization, The value ranges from 0 to 1. The larger the value, the stronger the direct demand between station i and station j. This indicator is used to indicate whether there is sufficient direct passenger flow between station i and station j.

[0088] Service time window continuity is calculated using the following formula:

[0089]

[0090] in, This represents the set of serviceable request time windows related to site i within the time period τ. This represents the set of serviceable request time windows related to site j within the time period τ; This represents the connectable time window obtained by shifting the service time window of station j forward according to the running time from station i to station j; Indicates the length of the time window or the effective duration of the time window set; To prevent extremely small positive numbers with a denominator of zero, the larger this index, the easier it is for vehicles to meet the time window requirements of both preceding and following requests when continuing from station i to station j. A smaller value would be preferable if arriving too early would require waiting, or if arriving too late would cause delays.

[0091] Passenger capacity balance is calculated using the following formula:

[0092]

[0093] in, This represents the expected passenger volume at station i within the time period τ. This represents the expected number of passengers disembarking at station j within the time period τ. To prevent extremely small positive numbers with a denominator of zero. When and The closer the stations are, the more efficient the vehicle capacity turnover is, meaning that after passengers board at station i, the vehicle can release its capacity near station j. The higher the value, the lower the capacity balance. Conversely, if passengers only board but do not disembark, or if the difference between boarding and disembarking is too large, the capacity balance decreases. This indicator reflects the turnover of vehicle capacity on the chain. If there is strong demand for boarding at station i and strong demand for disembarking at station j, then the vehicle can release capacity after moving from i to j, making room to continue serving downstream stations.

[0094] Downstream continuity of service is calculated using the following formula:

[0095]

[0096] in, This represents other candidate downstream sites besides site j; This represents the normalized demand intensity from station j to station k; This is an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise. This represents the running time from station j to station k within the time period τ; This represents the maximum permissible inter-station travel time threshold. This formula measures whether, after a vehicle arrives at station j, there are still a number of downstream stations that can continue to serve within a reasonable timeframe, and whether there are multiple stations after station j that are reachable within a reasonable timeframe and have high demand. If there are multiple stations after station j that are reachable within a reasonable timeframe and have high demand, this value is large, indicating that station j is suitable as a relay hub in the chain.

[0097] Passenger feedback acceptance It can be calculated using recent historical data on a rolling basis:

[0098]

[0099] In the formula, To provide passengers with the number of trips that include that station, Number of times passengers accept This is the smoothing factor. The purpose of setting a smoothing factor is to prevent sudden fluctuations in weight due to new sites or a small number of samples. The higher the value, the higher the passenger acceptance of the station pair's history, and the more suitable the station pair is to be selected in subsequent chain routes.

[0100] After obtaining the weights of the mutual feedback chain edges, the overall cost of building the chain edges in the system is as follows:

[0101]

[0102] The first five items represent running time, network distance, stop time, line disturbance, and turnaround / detour costs, respectively, while the last item represents feedback benefits. The stronger the feedback effect of a link edge, the lower its overall cost, and the easier it is to be selected as a line during the main chain generation and insertion process.

[0103] The weight of feed-in links is not simply a metric for passenger traffic. A station may have high direct demand, but if its time windows are not aligned, its vehicle capacity is occupied for extended periods, or its historical acceptance rate is low, its weight will not be high. Conversely, a station with moderate direct demand but capable of releasing capacity, connecting multiple downstream stations, and having a high passenger acceptance rate can also become a priority link.

[0104] Step S4: Generate a feasible directed chain graph for the appointment service;

[0105] A feasible directed chain graph for the reservation service is used to narrow down the search scope for route generation. The system treats candidate stations as graph nodes and candidate connections that meet the feasibility conditions of the reservation service as directed edges. Candidate chain edges are added to the edge set only if the following conditions are met:

[0106]

[0107] in, Let represent the set of candidate chain edges in the feasible directed chain graph during time period τ. This indicates that the travel time between the two stations does not exceed the maximum allowed inter-station travel time. Indicates site Located at the station In the direction of chain propulsion, This indicates that the capacity is not violated. This indicates that the time window is not violated. This indicates that passengers taking a detour is not a violation. This indicates that the station's stopping direction is permitted.

[0108] The direction of chain-like advancement can be determined according to the service scenario. In the commuting scenario, the system determines the advancement direction based on the main OD directions of the morning and evening peak hours respectively; in the medical treatment scenario, the advancement direction is usually from the residential area or rail station to the hospital; in the event evacuation scenario, the advancement direction is usually from the event venue to multiple evacuation areas; in the park connection scenario, the advancement direction can be from the rail station or residential area to the park entrance and exit.

[0109] To reduce invalid searches, the system can also set candidate edge filtering rules. For example, only the top K outgoing edges with the lowest overall chain edge cost can be retained for each site; edges with mutual feedback chain edges whose weights are below a threshold and have no direct booking demand can be deleted; edges with obvious loops or detours can be penalized more severely and deleted directly if necessary. This controls the search scale while retaining candidate connections with practical operational significance.

[0110] After the feasible directed chain graph is generated, the system outputs the starting station, ending station, running time, comprehensive chain edge cost, and mutual feedback chain edge weight for each candidate chain edge. This information is directly used for main chain generation, site insertion, and real-time scheduling.

[0111] Step S5: Generation of chain circuits;

[0112] The main chain skeleton is the initial framework of the chained route, used to determine the basic service direction and main demand coverage of the route. The system first determines the starting anchor point and the ending anchor point, and then generates the minimum overall cost path from the starting anchor point to the ending anchor point on the feasible directed chain graph.

[0113] Calculate the net distribution intensity for each candidate site:

[0114]

[0115] in, This indicates the net aggregation intensity of candidate site i within the time period τ; This indicates the station's expected passenger volume; This indicates the expected number of passengers disembarking at the station. Larger stations indicate stronger demand for passengers to board than to disembark, making them suitable as upstream customer gathering nodes. Smaller stations or those with high passenger drop-off demand are suitable as downstream evacuation nodes. The system then selects the starting and ending anchor points based on the main direction of demand and station accessibility.

[0116] When there are multiple candidate start or end anchor points, the system can use an anchor point scoring function:

[0117]

[0118]

[0119] in, This indicates the degree of consistency between the site and the main direction of demand. This indicates a risk of station congestion or stopping. This indicates the additional cost for vehicles to enter the station. and The system scores the starting and ending anchor points separately, and prioritizes stations with higher scores.

[0120] After determining the anchor points, the system solves for the minimum total cost path on the feasible directed chain graph to obtain the main chain skeleton:

[0121]

[0122] in, As the starting point and anchor point, The endpoint anchor point. The main chain skeleton should meet the following conditions: no duplicate sites, feasible chain edges, no capacity violation, no time window violation, and no maximum detour rate violation.

[0123] If no feasible path exists between the starting anchor point and the ending anchor point, the system can adopt the following approaches: reduce the demand coverage target of a single chain; enable a branch chain structure; add the next scheduled vehicle; or convert some requests to alternative station pairs and recalculate. Through the above processing, an operable solution can still be formed even under tight constraints.

[0124] Even after the main chain skeleton is generated, there may still be unserved requests. The system further absorbs these unserved requests by inserting intermediate sites. (Sites to be inserted) Insert adjacent stations in the chain and The insertion gain between them is:

[0125]

[0126] In the above formula, the first term represents the unserved demand that can be added after insertion; the second term represents the change in the mutual feedback revenue of the chain edges after insertion; the third term represents the added runtime; the fourth term represents the added downtime; and the fifth term represents the constraint violation penalty. If the penalty term is infinite, it means that the insertion violates a hard constraint and cannot be executed. If the value is greater than the preset insertion threshold and all hard constraints are met, then site k is suitable for insertion into the chain segment.

[0127] The site insertion process is performed as follows: First, establish a set of unserved sites, including sites capable of serving at least one unserved request. Second, iterate through each unserved site and each adjacent site pair in the main chain, calculating the insertion gain. Third, select the candidate solution with the largest insertion gain. Fourth, perform constraint checks, including vehicle capacity, request time window, maximum detour rate, site stopping direction, and published route disturbances. Fifth, if the constraints are satisfied, insert the site and update the chain; if the constraints are not satisfied, discard the insertion solution and try the next candidate solution. Sixth, repeat the above process until no feasible insertion solution with a gain greater than the threshold exists.

[0128] After insertion, the system binds the reservation request to the chained route. The conditions for a request to be bound to a route are: there is a boarding station i and an alighting station j in the route, and i is before j; the arrival time of the vehicle at i meets the boarding time window of the request; the arrival time of the vehicle at j meets the alighting time requirement; and the passenger's time on the vehicle does not exceed the allowed value.

[0129] A binding request can be represented as:

[0130]

[0131] in, Indicates whether request r is bound to chain line L, when When it indicates that request r is served by line L, This indicates that request r is not served by that line. Indicates site On the line Located in the middle of the station Before, This means that when request r selects station i to board and station j to alight, and is served by line L, it can simultaneously satisfy time window, capacity, detour rate and stop constraints; This indicates other situations where the above conditions are not met.

[0132] Arrival times are recursively calculated for chain-type railway lines:

[0133]

[0134] And perform vehicle passenger capacity estimation:

[0135]

[0136] in, This represents the estimated time when the vehicle will arrive at the m-th station in the chain. This indicates the estimated time when the vehicle will arrive at the (m+1)th station; and These represent the m-th and (m+1)-th stations in the chained route, respectively. This indicates the dwell time of the vehicle at the m-th station; This indicates that the vehicle departed from the station during the specified time period. Arrive at the station The runtime; This indicates the number of people inside the vehicle after it leaves station m. and They represent the sites respectively. If the number of passengers boarding and alighting exceeds the capacity at any time during the expected passenger load period, it indicates that the line capacity is insufficient and some insertions need to be cancelled or a sub-chain needs to be generated to handle part of the demand.

[0137] To control passenger detours, the system calculates the requests. On the line The ratio of actual on-vehicle time to direct station travel time:

[0138]

[0139] in, This indicates the actual detour factor for request r on line L; This indicates the on-train time from the boarding station to the alighting station when request r is served by line L. This indicates that the request to the site corresponding to r is... The direct runtime; To prevent extremely small positive numbers with a denominator of zero, if the ratio of actual on-board time to direct station pair travel time exceeds the passenger-allowed detour threshold, the request cannot be bound to that line, or other alternative station pairs need to be tried.

[0140] After the above process, the system outputs one or more operable chain routes. Each route includes at least a route number, station stop sequence, estimated arrival time, estimated departure time, vehicle number, service request list, passenger notification content, and adjustable chain segment range.

[0141] When all demands cannot be met by a single main chain, the system generates a split chain. A split chain is not simply an addition of vehicles; rather, it is determined based on the main chain's capacity pressure, detour pressure, and the breakpoint of the feed-in edge weights. If the insertion gain after a certain position on the main chain remains negative, or if the feed-in edge weight of a certain chain edge is significantly lower than that of its adjacent chain edges, it indicates that this position may be suitable as a split point.

[0142] The segmentation score can be defined as:

[0143]

[0144] in, Indicates the capacity pressure of the chain segment. This indicates the pressure of delays. Indicates bypass pressure, This represents the mutual feedback weight between adjacent chain edges. The larger the value, the more suitable it is as a chain splitting point.

[0145] Subchains fall into two categories. The first category is independent short subchains, used to serve local needs that are similar in direction to the main chain but would cause significant detours if directly inserted into the main chain. The second category is main chain-connected short subchains, used to transport passengers from scattered upstream stations to main chain connection stations, or to transport main chain passengers from connection stations to multiple downstream destinations.

[0146] The connection between the branch chain and the main chain should meet the connection time constraint:

[0147]

[0148] in, Main link to the connecting station Time, To reach the connecting station via the branch chain Time, This is the maximum allowed waiting time for connection.

[0149] For modular buses, if vehicles support splitting or merging, subchains can be formed by splitting vehicle groups. In this case, it is necessary to simultaneously verify the remaining vehicle capacity of the main chain and the vehicle capacity of the subchain after the split to avoid affecting the confirmed passengers on the main chain after splitting vehicles for new requests.

[0150] Step S6: Rolling time-domain scheduling adaptation;

[0151] After departure, reservation requests may continue to arrive, travel times may change, and some passengers may have already received reservation confirmation information such as boarding station, estimated boarding time, or estimated arrival time. If the entire route were reconstructed for each change, passenger information would change frequently. Therefore, this invention employs a rolling time domain mechanism. The rolling time domain length is [length missing], and the unlocked segment is updated every minute.

[0152] For each operating line The system divides it into locked segments. and unlocked segments :

[0153]

[0154] in This indicates that the sequence of stations is spliced ​​according to the original route. Locked segments include stations that the vehicle has already arrived at, is about to arrive at, or has been confirmed with the passenger and is currently en route to. Unlocked segments include station sequences that have not yet been served and have not yet been finalized.

[0155] The locking rules can be determined based on the advance confirmation time. If the vehicle is expected to... If a passenger arrives at a station within a specified timeframe, or if a passenger at that station has already received a confirmation notification, then that station and all stations preceding it enter a locked segment. Locked segments generally do not change the station order, cancel stations, or significantly alter arrival times; unlocked segments can be inserted, replaced, split, and restructured.

[0156] The segment locking mechanism protects the travel arrangements of passengers with confirmed reservations and prevents the system from overturning existing routes due to a small number of new reservation requests. For operators, driver-side navigation is more stable; for passengers, pick-up points and estimated arrival times are more reliable.

[0157] When a new request enters the system, the system does not directly recalculate globally. Instead, it processes the request in the following order: direct absorption within the chain, local insertion of unlocked chain segments, reconnection of sub-chains, and reconstruction of the unlocked tail chain.

[0158] First, direct absorption within the chain. If the boarding and alighting stations of a new request already appear in the unlocked segment of a certain line, and the boarding station is before the alighting station, and the vehicle capacity and time window are both satisfied, then the line station order remains unchanged, and only the service request list, the number of passengers boarding and alighting at the corresponding station, and the stop time are updated. This method causes minimal disturbance to existing lines.

[0159] Second, partial insertion of unlocked segments. If one or more candidate stations in a new request can be inserted into an unlocked segment of a line, the insertion position is calculated according to the insertion gain function. The system only changes the unlocked segment, not the locked segment. If a new request requires the addition of both a boarding and alighting station, the paired insertion positions of the two stations are calculated simultaneously, ensuring that the boarding station precedes the alighting station.

[0160] Third, branch chain connection. If a partial insertion would cause a significant detour in the main chain or affect confirmed passengers, but the number of new requests is large or has already formed a localized concentrated demand, the system generates a short branch chain. The short branch chain can be executed by an idle vehicle or by a detachable modular vehicle.

[0161] Fourth, unlocked tail chain reconstruction. If the number of new requests is large and has changed the main demand structure of the unlocked segment, the system only re-executes S3 to S6 for the unlocked tail chain, without changing the locked segment. The inputs for tail chain reconstruction include the end station of the locked segment, existing requests for the unlocked segment, new requests, remaining vehicle capacity, and current road running time.

[0162] To select a specific scheduling scheme, the system calculates a scheduling suitability score:

[0163]

[0164] in, To increase the number of users that can be served, This is the amount by which the weight of the feed-in chain edge is increased. To compensate existing passengers for the delays, To add a detour, To mitigate the disruption to already published lines, For additional vehicle or trainset costs.

[0165] The system selects the highest-rated option that meets the hard constraints of the reservation service. If none of the options meet the reservation service requirements, the system can return suggestions for the next departure time, alternative stops, or other modes of transportation to the passenger.

[0166] Step S7: Update operational feedback weights;

[0167] After the actual operation ends, the system updates the weights of the feedback chain edges based on passenger behavior and vehicle operation data. This feedback mechanism allows the route design to be gradually revised based on operational results, rather than relying solely on initial parameters.

[0168] For the site To the station For the chain edge, the feedback update formula is:

[0169]

[0170] Among them, represents passenger acceptance rate, service completion rate, delay penalty, and reservation non-fulfillment rate. If a chain edge experiences a long-term history of passenger rejections or reservation non-fulfillment, its weight decreases; if a chain edge operates stably and has a high passenger acceptance rate, its weight increases.

[0171] The update step size should not be too large. If it is too large, a single anomaly may cause excessive fluctuations in the line design; if it is too small, the system will not be able to adapt quickly enough to real changes. Generally, it can be set according to the operating cycle. For example, a smaller update step size is used for daily commuting, and a larger update step size is used for temporary events and evacuations.

[0172] Feedback updates can also be maintained separately for different passenger groups and service scenarios. For example, hospital patients typically focus more on walking distance, commuters on arrival time, and students on study tours on safe stops and group arrivals. The system can assign different weighting coefficients to different service scenarios and maintain them separately during feedback updates.

[0173] Example

[0174] Taking the morning rush hour commuter service of an industrial park as an example, the service area includes three residential areas A, B, and C, and two park entrances P and Q. The system receives reservation requests from 6:30 to 7:30, with passengers mainly traveling from residential areas to the park. Candidate stops include existing bus stops at the entrances of residential communities, subway station exits, and park entrances / exits.

[0175] The system first maps each reservation request to a candidate station pair with a walking time of no more than 8 minutes. Statistical results show that stations A, B, and C mainly serve the upstream passenger gathering function, while stations P and Q mainly serve the downstream evacuation function. The time windows between stations A and B, and between B and C, are well connected. There is strong direct demand between station C and P. Station P can still serve a small number of passengers heading to Q, therefore, the relevant link edges have high weights.

[0176] The system selects A as the starting anchor point and P as the ending anchor point, generating a main chain ABCP in the feasible directed chain graph. It then finds that station D still has 12 unserved passengers. If D is inserted between B and C, adding 4 minutes of travel time, it can serve 12 more passengers without causing delays for passengers at stations A and B. Therefore, the insertion gain is positive, and the route is updated to ABDCP.

[0177] After departure, segments A, B, and D gradually enter the locked segments. If a new request from D to P is added at 7:45, since both D and P are already in the unlocked segment or subsequent service segment, the system directly absorbs the request without changing the station order. If a new group of requests from X to Q is added, directly inserting it into the main chain would cause existing passengers to experience delays exceeding the threshold. In this case, the system schedules an idle vehicle to generate a short sub-chain XQ, which is then connected to the main chain at point Q.

[0178] After the operation ends, if the ABDCP route has a high passenger acceptance rate and low delays, the weight of the relevant chain edge will be increased; if the XQ short chain has a high non-fulfillment rate, the weight of the corresponding chain edge will be decreased. When generating routes the next day, the system will prioritize the chain edge with more stable operation.

[0179] Let's take a hospital as an example to illustrate the specific implementation method. Parking is difficult around a hospital, and elderly patients and those returning for follow-up appointments book public transportation to get there. Passengers originate from various communities, with their destinations concentrated at the hospital's outpatient and inpatient entrances. In this scenario, passengers are sensitive to walking distance and on-time arrival, so the bus routes should minimize detours.

[0180] The system increases the weight of walking time when mapping candidate stations, and increases the weight of time window connection and passenger feedback acceptance in the mutual feedback link weight. For hospital outpatient entrances, although the demand is high, if the road is congested or parking is difficult, the stopping and disturbance penalties in the link edge comprehensive cost will increase, and vehicles may choose the inpatient department side gate or hospital shuttle point as the drop-off station.

[0181] When a new patient requests entry, the system first determines whether it can be absorbed by an existing station within the chain. If a new stop is needed, the system only inserts it within an unlocked segment and verifies the arrival time of confirmed patients. If the insertion would affect the appointment time of confirmed patients, the system provides the new requesting passenger with the next available train or an alternative drop-off station.

[0182] In a modular bus scenario, a single bus can be split into multiple modules while meeting safety requirements. A mainline bus departs from the rail station to an industrial park, and a new request for a bus to go to the park's side gate arises en route. If the entire bus detours to the side gate, it will affect the on-time arrival of most passengers on the mainline; if the new request is rejected outright, it will reduce service coverage.

[0183] In this invention, after determining the capacity and time window of the unlocked segment of the main chain, the system can choose to split a module at the connection point to form a short sub-chain. The split verification includes: the remaining capacity of the main chain after splitting is not less than the number of confirmed passengers; the capacity of the short sub-chain meets the number of new requests; and the splitting and merging time does not affect the service of the locked segment of the main chain. If the verification passes, the short sub-chain serves side-door requests, and the main chain continues to run in its original direction.

[0184] This implementation shows that dynamic demand does not necessarily require triggering a full-line recalculation, but can be absorbed locally through a branch chain structure, thereby reducing the impact on passengers on the main chain.

[0185] When a sudden traffic jam causes a significant increase in the running time of a certain chain edge, the system first determines whether the affected chain edge is in a locked segment. If it is in an unlocked segment, the overall cost of that chain edge is increased, and the unlocked tail chain is re-inserted or reconstructed; if it is in a locked segment, the station order remains unchanged, only the estimated arrival time is updated, and a delay notification is sent to the passenger.

[0186] When a vehicle experiences a temporary malfunction, the system releases unserved requests from the chain containing the malfunctioning vehicle and prioritizes finding an unlocked chain in the same direction to absorb them. If the chain cannot absorb the requests, an alternative sub-chain is generated or the next shift is scheduled. For confirmed requests that are about to be served, their service time window is protected first.

[0187] When a passenger's reservation is not fulfilled, the system reduces the actual passenger volume at the corresponding station and updates the vehicle passenger capacity projection. If the reservation non-fulfillment occurs during a locked period, subsequent stations are not immediately cancelled; if the same station experiences a consistently high reservation non-fulfillment rate, the corresponding link edge weight is reduced in the feedback update.

[0188] This invention can also be implemented as a reservation-based public transport chain-based route design system. This system includes a data receiving module, a reservation request mapping module, a mutual feedback chain edge calculation module, a reservation service feasible directed chain graph generation module, a chain-based route generation module, a rolling scheduling adaptation module, a passenger notification module, a driver navigation module, and a feedback update module.

[0189] The data receiving module receives data from the reservation platform, vehicle terminal, and road operation monitoring system; the reservation request mapping module converts passenger reservation requests into station-to-station demand; the feedback chain edge calculation module calculates the weights and overall costs of the feedback chain edges; the reservation service feasible directed chain graph generation module generates a feasible directed chain graph for the reservation service; the chain route generation module generates the main chain, branch chains, and connecting chains; the rolling scheduling adaptation module handles new reservation requests; the passenger notification module pushes boarding stations, estimated arrival times, reservation confirmations, and change information to passengers; the driver navigation module outputs the station sequence and navigation route to the driver; and the feedback update module updates the chain edge weights based on operational results.

[0190] The system can be deployed on public transport company dispatch platforms, urban transportation cloud platforms, or pre-booked public transport operation service platforms. Small-scale scenarios can be deployed using a single server; multi-route, large-area scenarios can employ regional parallel computing. Because this invention controls the search range through a feasible directed chain graph and an unlocked segment mechanism, it is suitable for online dispatch scenarios.

[0191] This invention outputs not only simple geometric routes, but also structured data that can be directly used by the operating system. Each route output includes at least the route number, service period, vehicle number, station stop sequence, estimated arrival time at each station, estimated departure time at each station, passenger pick-up request list at each station, passenger drop-off request list at each station, locked segment range, unlocked segment range, list of insertable backup stations, passenger notification text, and driver execution instructions.

[0192] The passenger's information includes recommended pick-up point, walking navigation, estimated pick-up time, estimated arrival time, license plate or vehicle number, whether ridesharing is possible, and estimated confirmation time if the route is not yet locked. The driver's information includes next stop, estimated arrival time, number of passengers boarding and alighting, whether temporary adjustments are allowed, and instructions for splitting or connecting routes.

[0193] To ensure a smooth experience for passengers using the pre-booked bus service, the route information and locking rules released by the system remain consistent. Unlocked segments can be displayed as the planned route; once locked, a reservation confirmation message is sent to the passenger. This maintains scheduling flexibility while preventing passengers from developing distrust due to frequent information changes.

[0194] In ordinary commuting scenarios, the weights of direct demand intensity and time window connection can be set relatively high; in hospital medical treatment scenarios, the weights of passenger feedback acceptance and walking time cost can be increased; in large-scale event dispersal scenarios, the weights of downstream continuous service can be increased, making it easier for the route to form multi-directional evacuation chains.

[0195] The rolling time range length can be set according to the service scenario. During peak commuting hours, it can be set to 15 to 30 minutes to reduce frequent route fluctuations; for community micro-circulation, it can be set to 5 to 15 minutes to improve response speed; and for large-scale event dispersal, it can be dynamically set according to batch evacuation plans.

[0196] Insertion gain threshold controls the aggressiveness of site insertion. A lower threshold makes the system more receptive to new demand, but may increase line length; a higher threshold makes the line more stable, but some demand may be delayed or served by a branch chain. Operators can adjust this threshold based on service commitments and vehicle resources.

[0197] This invention differs from route design based solely on the shortest path. While the shortest path method primarily optimizes vehicle travel distance or time, this invention simultaneously considers demand delivery, time window connection, capacity release, downstream continuous service, and passenger feedback through mutual feedback chain edge weights, making station connections more in line with the ride-sharing operation characteristics of reservation-based public transport.

[0198] This invention also differs from the conventional vehicle routing problem. The conventional vehicle routing problem typically treats all customer points as nodes that must be visited, while in this invention, candidate stations can be inserted, replaced, chained, or delayed. Passenger requests are bound to a chained route through both boarding and alighting stations, making it more suitable for bus stop-based services.

[0199] This invention differs from simple real-time scheduling methods. Simple real-time scheduling typically re-optimizes vehicle routes after a new request arrives, which can easily cause changes to confirmed passenger routes. This invention, through reservation confirmation, locked segment protection, and local adjustment of unlocked segments, maintains the stability of existing reservation services as much as possible while responding to new reservation requests.

[0200] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A chain-like route design method for reservation-based bus hubs, characterized in that, include: S1. Receive the set of reservation requests, the set of candidate stations, the set of vehicles, and the road operation time data within the target service area; standardize the reservation requests and form a reservation request data table. S2. Map each reservation request to one or more candidate boarding and alighting stations to form reservation station pair requirements with default station pairs and alternative station pairs, and statistically analyze the reservation demand characteristics for each service period. S3. For any two candidate sites, calculate the mutual feedback chain edge weight based on reservation demand acceptance, service time window connection and historical reservation feedback, and construct the chain edge comprehensive cost by combining road running time, station time and published route disturbance penalty to form a candidate chain connection relationship. S4. Based on the chain-like advancement direction and the feasibility conditions of the reservation service, select candidate chain edges that can be connected and generate a feasible directed chain graph for the reservation service. S5. Determine the starting anchor point and the ending anchor point in the feasible directed chain graph, generate the main chain skeleton, and insert the intermediate stations corresponding to the unserved reservation requests into the main chain or generate sub-chains based on the insertion gain of the reservation requests, so as to obtain an initial chain route scheme containing the station stopping sequence, the expected arrival time, the reservation service request list and the adjustable chain segment range. S6. Adopting a rolling time-domain mechanism, the chain line is divided into locked segments and unlocked segments according to the reservation confirmation status. For new reservation requests, the chain direct absorption, unlocked segment local insertion, branch link reconnection and unlocked tail chain reconstruction are performed in stages. S7. Update the weights of the feedback chain edges based on passenger acceptance, reservation fulfillment, service delays, and actual vehicle arrival times after actual operation, and optimize subsequent route design.

2. The method according to claim 1, characterized in that, In step S1, the reservation request is represented as: in, and These represent the passenger's expected origin and expected destination, respectively. and These represent the lower and upper limits of the appointment service time window, respectively. Indicates the number of passengers. Indicates the maximum acceptable walking time. This indicates the maximum acceptable detour rate. The status of the reservation service is indicated, including unmatched, matched, confirmed, boarded, completed, canceled, or reservation not fulfilled.

3. The method according to claim 1, characterized in that, In step S2, for each reservation request Filter the set of candidate boarding stations and candidate drop-off stations The selection criteria are: in, For the candidate site set, For passengers from their desired origin to the candidate boarding station walking time, Candidate drop-off stations Walking time to the passenger's expected destination This indicates whether stopping at the corresponding station is permitted in the direction of vehicle travel; When multiple candidate site pairs exist for the same reservation request, the default site pair is determined according to the principle of minimizing the service cost of the site pair, and a preset number of alternative site pairs are retained in ascending order of cost; the service cost of the site pair is: in, For time period Internal candidate sites To candidate sites Road travel time, The station will impose a pre-existing waiting penalty. For historical passengers to the station Acceptance rate to These are the weighting coefficients.

4. The method according to claim 1, characterized in that, In step S3, candidate sites To candidate sites During the period mutual feedback chain edge weights Calculate using the following formula: in, To normalize the site's demand intensity, To ensure seamless service time window connectivity, To balance passenger boarding and alighting capacity, Candidate sites Downstream continuity of service For passenger feedback on acceptance, to The weights are non-negative and satisfy the following conditions: ; The service time window connectivity Calculate using the following formula: in, For the site During the period The time window for merging service requests within the scope. To deduct sites To the station Sites after runtime Connectable service time windows To prevent constants with a denominator of zero; The balance of passenger boarding and alighting capacity Calculate using the following formula: in, Candidate sites During the period The expected number of passengers inside, Candidate sites During the period The projected number of passengers disembarking within the area; the downstream continuous service level. Calculate using the following formula: in, For indicator functions, This is the maximum inter-station operating time threshold; Chain edge comprehensive cost Calculate using the following formula: in, For the site To the station Road network distance, For the site Estimated service stop time, To compensate for the disturbance caused to the published lines by connecting this chain edge, Penalties for turning back or taking a detour The weight of the chain edge. to Cost weighting.

5. The method according to claim 1, characterized in that, In step S4, the conditions for adding a candidate chain edge to the feasible directed chain graph of the reservation service include: the candidate site is located in the chain-driven direction; the comprehensive cost of the chain edge or the weight of the mutually fed chain edge meets the preset screening conditions; the reservation service feasibility conditions are met after adding the chain edge; and the candidate site meets the feasibility of vehicle parking in the corresponding direction.

6. The method according to claim 1, characterized in that, In step S5, the starting anchor point and the ending anchor point are jointly determined based on the net distribution intensity of the station, the main direction of demand, and the accessibility of the candidate stations. The net distribution intensity of the station is: when Greater than the first threshold and the site When located upstream of the main demand direction, it is used as a candidate starting point anchor point; when Less than the second threshold or Greater than the third threshold and the site When it is located downstream of the main demand direction, it is used as a candidate endpoint anchor point; The main chain skeleton is obtained by solving for the minimum comprehensive edge cost path from the starting anchor point to the ending anchor point on the feasible directed chain graph of the reservation service. The main chain skeleton is represented as follows: in, As the starting point and anchor point, As the final anchor point, to As an intermediate distribution station, the main chain skeleton satisfies the following conditions: no duplicate stations within the chain, all chain edges belong to the feasible directed chain graph of reservation services, and it does not violate the reservation service constraints. In step S5, the site to be inserted Insert adjacent sites within the chain and Insertion gain between Calculate using the following formula: in, This represents the number of unserved appointments that can be added to the service list at the site. Penalties for violations of reservation service constraints caused by inserting a site. to This is the insertion gain weight.

7. The method according to claim 6, characterized in that, Site insertion is performed according to the following rules: traverse the candidate sites corresponding to unserved reservation requests and all adjacent site pairs in the main chain, and calculate the insertion gain; select the site and position with the largest insertion gain; if the insertion gain is greater than the insertion threshold and the reservation service feasibility condition is met, then insert the request and update the arrival time, number of passengers, service request list, and chain edge status; if the reservation service feasibility condition is not met, then the insertion scheme is listed as infeasible and the next candidate insertion scheme is tried.

8. The method according to claim 1, characterized in that, In step S5, the arrival time and passenger capacity of the chain-type line are calculated recursively using the following formula: in, The time when the vehicle arrives at the i-th station within the chain. For the first Service downtime at each station For the vehicle to leave the The number of passengers in the vehicle after each station; if the reservation service constraints are not met, the corresponding insertion scheme or chain segment will be rolled back.

9. The method according to claim 1, characterized in that, In step S6, the locked segment is a station sequence where the vehicle has arrived, is about to arrive, the passenger has received a confirmation notification, or the system has set an unchangeable station sequence; the unlocked segment is a station sequence where the vehicle has not yet served and the passenger has not yet received a final confirmation; the system only performs new request insertion, station replacement, sub-chain generation, or tail chain reconstruction within the unlocked segment. New reservation requests will be scheduled and adapted according to the following priority: first, determine whether they can be directly bound within the existing line's unlocked section; If direct binding is not possible, determine whether partial insertion can be performed in the unlocked chain segment through alternative sites; if partial insertion does not meet stability requirements, generate a branch link; when the scale of new requests reaches the reconstruction threshold, only the unlocked tail chain is reconstructed. Add request scheduling adaptation score Calculate using the following formula: in, To increase the number of users that can be served, This is the amount by which the weight of the feed-in chain edge is increased. To compensate existing passengers for the delays, To increase the amount of traffic by taking detours, To mitigate the disruption to already published lines, For additional vehicle or train formation costs, the system selects vehicles that meet hard constraints and The best strategy.

10. The method according to claim 1, characterized in that, In step S7, the weights of the mutual feedback chain edges are updated based on actual operational feedback: in, For passenger acceptance rate, To improve service completion rate, In order to delay punishment, To determine the booking non-fulfillment rate, To update the step size, to This refers to the feedback weighting coefficient.

Images