A method and system for transfer scheduling considering subway service induced impulse demand

Abstract

This invention discloses a connection scheduling method and system that considers the pulse demand induced by subway train schedules, belonging to the field of intelligent transportation and urban public transportation technology. First, a four-layer progressive pulse demand expression framework is constructed, consisting of time slot mapping, pulse cluster generation, intensity estimation, and spatiotemporal coordination calculation, aggregating dispersed travel requests into three-dimensional pulse clusters. Then, a constrained mixed-integer linear programming model is established with the objectives of minimizing operating costs, unserved service losses, and transfer mismatch penalties. Subsequently, a Pulse-ALNS solution framework adapted to pulse demand is designed, solving for the optimal scheduling scheme through steps such as pulse perception disruption and Train-Batch batch repair. Finally, real-time monitoring of operational status and dynamic scheduling adjustment are achieved. This invention fully considers the pulse demand characteristics induced by subway train schedules, significantly improving the solution efficiency and transfer reliability of connection scheduling, enabling second-level scheduling decisions, and adapting to the industrialized operation needs of subway "last mile" connections.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation and urban public transportation service technology, specifically involving a connection scheduling method and system that considers the demand induced by subway train schedules. It is applicable to the "last mile" connection scenario of subway, realizes demand-responsive coordinated scheduling of buses and subways, and improves the efficiency of connection services, transfer reliability and operational economy. Background Technology

[0002] As a high-capacity backbone of urban transportation, subways play a vital role in ensuring efficient urban travel and improving the public transportation network. With the continuous expansion of urban rail transit, according to statistics from the Ministry of Transport, by the end of 2024, 54 cities in mainland my country had opened urban rail transit systems, with an operating mileage exceeding 10,000 kilometers and a daily passenger volume exceeding 80 million. However, traditional subway services suffer from a significant "last mile" limitation, failing to directly cover passengers' origin and destination, and thus failing to meet passengers' needs for convenient travel throughout their journey. Therefore, demand-responsive public transport (DRT) is widely used in urban public transportation systems as a supplement, leveraging its dynamic scheduling and on-demand service characteristics to compensate for the spatial limitations of subways and achieve efficient connections between subways and passengers' origin and destination.

[0003] Compared to traditional fixed-route bus connections, DRT scheduling in subway connection scenarios faces new challenges. First, most existing scheduling research and applications fail to consider the pulse demand characteristics induced by subway schedules—that is, the phenomenon of concentrated release of exit connection demand after train arrival and pre-departure connection demand accumulation before train departure, making it difficult to adapt to dynamic demand changes. Second, existing scheduling methods cannot accurately characterize this demand aggregation structure; the solution process does not utilize pulse demand correlation information, resulting in high computational overhead and slow response speed, making it difficult to meet large-scale real-time scheduling needs. More importantly, existing scheduling schemes do not deeply integrate subway schedules, leading to insufficient transfer reliability. Furthermore, the ALNS algorithm is not optimized for pulse demand, making it difficult to balance operational efficiency and service coverage, and it lacks a robust dynamic adjustment mechanism, resulting in insufficient robustness.

[0004] Finding a balance between scheduling efficiency and transfer reliability, and addressing various shortcomings caused by pulse demand through optimizing DRT scheduling design, so that DRT connections can not only achieve deep coordination between subway and bus under normal conditions, but also resist the impact of dynamic disturbances and ensure the efficiency and stability of connection services, are problems that need to be solved at present. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, this invention provides a connection scheduling method and system that considers the demand induced by subway train schedules. It aims to solve the problems of inaccurate demand characterization, low solution efficiency, poor transfer reliability, and insufficient robustness in existing subway connection scheduling, and achieve multi-objective optimization of operating costs, service coverage, and transfer reliability, providing an efficient and intelligent scheduling solution for urban subway connection services.

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

[0007] A connection scheduling method considering the demand for subway train induced pulses includes the following steps:

[0008] S1: Considering the characteristics of subway train schedule guidance, a four-layer progressive pulse demand expression framework is constructed, which includes time slot mapping, pulse cluster generation, intensity estimation, and spatiotemporal coordination calculation. This framework aggregates scattered individual travel requests into a three-dimensional pulse cluster consisting of station, time slot, and direction.

[0009] S2: With the goal of minimizing operating costs, unserved losses, and transfer mismatch penalties, a mixed integer linear programming model with time windows, capacity constraints, and pick-up / delivery pairing constraints is established by combining request allocation, path continuity, time feasibility, and capacity feasibility constraints.

[0010] S3: Based on the impulse demand structure and mixed integer linear programming model, the Pulse-ALNS solution framework is designed. The optimal connection scheduling scheme is solved through the steps of initial solution construction, impulse sensing destruction, Train-Batch repair, solution evaluation and acceptance, weight update and computation acceleration.

[0011] S4: Assign the optimal scheduling plan to the corresponding vehicles, monitor vehicle status, subway schedule changes and road congestion in real time, and dynamically adjust the scheduling plan to ensure service reliability.

[0012] As a further technical solution of the present invention: the impulse demand modeling in step S1 specifically includes:

[0013] S101: Time slot mapping, assuming the subway train frequency reference interval is... The target connection time for request r is Define the time slot mapping function: This maps individual requests over continuous time to discrete time slot indexes.

[0014] S102: Pulse cluster construction, based on the three-dimensional attributes of station m, time slot k, and direction d, constructs a pulse cluster:

[0015]

[0016] And define cluster strength:

[0017]

[0018] in For the request The number of passengers;

[0019] S103: Arrival rate characterization, defining the arrival rate function at the station-direction level:

[0020]

[0021] in For the first The pulse amplitude corresponding to each subway train number , The pulse center time, The pulse spread coefficient is... This is a background requirement.

[0022] S104: Robustness estimation, calculating the expected demand and variance within time slot k:

[0023]

[0024] Define robust pulse intensity estimation:

[0025]

[0026] in These are the discretized variance coefficients. This is the risk preference coefficient;

[0027] S105: Spatiotemporal coordination calculation, constructing cluster-level spatiotemporal coordination:

[0028]

[0029] in, For time synchronization of requests within the cluster, For the sake of space compactness of intra-cluster requests, , These are the weighting coefficients.

[0030] As a further technical solution of the present invention: the mixed integer linear programming model in step S2 specifically includes:

[0031] S201: The objective function is expressed as:

[0032] (1)

[0033] Where V represents the set of vehicles; A represents the set of road network arcs; and R represents the set of travel requests. The fixed start-up cost for vehicle v; Enable state variables for vehicle v; For arc The unit operating cost; For vehicle v to pass through arc State variables; The state variable for request r being served; To request the lateness duration of r; For late arrival penalty function;

[0034] S202: The constraint conditions are specifically expressed as follows:

[0035] (2)

[0036] (3)

[0037] (4)

[0038] (5)

[0039] (6)

[0040] (7)

[0041] (8)

[0042] (9)

[0043] (10)

[0044] (11)

[0045] (12)

[0046] (13)

[0047] (14)

[0048] (15)

[0049] Where R represents the set of travel requests; The state variable for requesting r to be served by vehicle v; , These represent the pick-up and drop-off points for request r, respectively; 0 indicates a parking lot node. , These are the start and end times of the service time window for node i, respectively. The service start time for vehicle v at node i; The service time for node i; For arc Travel time; M is a sufficiently large constant; The maximum acceptable travel time for request r; The passenger load when vehicle v leaves node i; The rated capacity of vehicle v; , These represent the times when vehicle v departs from and returns to the parking lot, respectively. This refers to the maximum operating time for a single shift of a vehicle.

[0050] As a further technical solution of the present invention: step S3 further includes:

[0051] S301: Initial solution construction adopts a two-layer solution representation method of vehicle path sequence + request status index. The path layer records the access node order of each vehicle, and the status layer records the vehicle allocation, service time, late status and pulse cluster to which each request belongs.

[0052] S302: Pulse-sensing disruption, defining cluster removal priority score:

[0053]

[0054] in, , , These are the weighting coefficients. To mitigate the risk of in-cluster request default, To determine the cost of bypassing the cluster edge, three types of destructive operators are employed: whole-cluster removal operator, local sampling removal operator, and high-penalty request removal operator.

[0055] S303: Train-Batch Fix, Define Batch Insert Increment:

[0056]

[0057] in For the request Insert vehicle Individual incremental costs, For bulk discount factors, Conflict penalties are imposed; a two-tier rollback and repair mechanism is employed to ensure the feasibility of repairs.

[0058] S304: Solution Evaluation and Acceptance. Simulated annealing is used as the acceptance criterion. Let the objective difference between the candidate solution and the current solution be... The probability of acceptance is:

[0059]

[0060] in The temperature parameter at each iteration time;

[0061] S305: Weight Update, introducing an adaptive weight update mechanism. Let the set of operators be Ω, and the operators... The weight is Perform weight update: Operator selection probability ;

[0062] S306: Computational acceleration, employing an integrated acceleration module to reduce the cost per unit iteration evaluation, including constructing a lower bound for cost based on the triangle inequality, adopting a version number-driven lazy re-evaluation mechanism, and employing a selective repair strategy.

[0063] As a further technical solution of the present invention: step S4 further includes:

[0064] S401: Add request processing, set the request update cycle to 30 seconds, when a new travel request arrives, only re-insert the request in vehicles with unlocked tasks;

[0065] S402: Handling subway schedule changes and setting transfer safety margins Every 2-5 minutes, the system receives real-time train schedule updates from the subway operation system and recalculates the transfer availability for each request.

[0066]

[0067] in This refers to the arrival times of subway trains. If the estimated arrival time of the vehicle at the drop-off point is... < If so, the subway train schedules will be rematched and the vehicle dispatching plan will be adjusted.

[0068] S403: Road congestion management, real-time acquisition of road segment congestion status through road network data interface, and real-time monitoring of the remaining time for vehicles to reach their drop-off points: ,like This will trigger the emergency response plan.

[0069] As a further technical solution of the present invention: in step S105, time synchronization... Spatial compactness is calculated by the standard deviation of the connection times of all requesting targets within the cluster. It is calculated by the average pairwise distance of all visitor points within the cluster.

[0070] As a further technical solution of the present invention: the lateness penalty function It is a piecewise linear function, and the penalty value varies with the duration of lateness. The increase is in a step-like manner.

[0071] As a further technical solution of the present invention: the initial solution construction in step S301 first involves pulse grouping the requests according to the three-dimensional attributes of station-time slot-direction, then sorting them according to cluster strength, time urgency, and spatiotemporal coordination, and constructing the initial path using the minimum incremental insertion strategy.

[0072] A connection scheduling system considering the demand for subway train induced pulses, used to implement the above method, includes:

[0073] The pulse demand modeling module is used to construct a four-layer progressive pulse demand expression framework, which aggregates scattered individual travel requests into a three-dimensional pulse cluster of station-time slot-direction;

[0074] The optimization modeling module is used to build a constrained mixed-integer linear programming model with the goal of minimizing operating costs, unserved losses, and transfer mismatch penalties.

[0075] The intelligent solution module has a built-in Pulse-ALNS solution framework for solving optimal connection scheduling schemes.

[0076] The dynamic scheduling module is used to allocate scheduling plans to corresponding vehicles, monitor the operational status in real time, and dynamically adjust the scheduling plans.

[0077] The human-computer interaction module communicates with each of the above modules and is used for parameter configuration, data display, log query and fault alarm.

[0078] As a further technical solution of the present invention, it also includes a data docking module, which is used to interact with the subway operation system, road network data platform, vehicle terminal and factory MES system to obtain real-time subway schedule dynamics, road congestion status and vehicle operation data.

[0079] This technology proposes a connection scheduling method that considers the demand for subway train induced pulses, which has the following advantages and beneficial effects:

[0080] 1. Accurately depicting the characteristics of subway pulse demand and solving the problem of disconnect between demand modeling and actual scenarios: This invention constructs a four-layer progressive pulse demand expression framework of "time slot mapping - pulse cluster generation - intensity estimation - spatiotemporal coordination calculation", which explicitly depicts the concentrated arrival characteristics of demand under subway train guidance. It aggregates scattered individual requests into three-dimensional pulse clusters of station-time slot-direction, fully explores the spatiotemporal coordination value of requests in the same cluster, and fundamentally solves the defects of existing technologies that treat travel requests as independent individuals and do not consider the characteristics of subway train guidance.

[0081] 2. Design a Pulse-ALNS solution framework adapted to pulse requirements, significantly improving solution efficiency and scheduling quality: This invention introduces the pulse cluster structure into the main search loop of the ALNS algorithm, designs a pulse-aware destruction operator and a Train-Batch batch repair mechanism, and makes full use of the collaborative value of requests within the same cluster. Compared with the traditional ALNS algorithm, the solution efficiency is improved by more than 60%, and second-level scheduling decisions can be achieved in large-scale request scenarios, fully meeting the real-time requirements of actual operation. At the same time, through three types of computation acceleration strategies, the computational overhead per unit iteration is further reduced, adapting to large-scale industrial scheduling scenarios.

[0082] 3. Deep integration of subway schedules significantly improves transfer reliability and system robustness: This invention embeds subway schedules and transfer reliability penalties into the optimization model and solution evaluation process. Combined with a robust dynamic adjustment mechanism, it can effectively cope with dynamic disturbances such as changes in subway schedules, road congestion, and new requests, preventing passengers from missing their target subway trains or waiting for long periods of time, and improving the on-time transfer rate to over 99%. At the same time, through robust impulse intensity estimation, it effectively resists the risk of service interruption caused by demand fluctuations, and the overall operating efficiency (OEE) of the system is ≥98%.

[0083] 4. Achieves multi-objective comprehensive optimization and possesses strong practical application value: This invention constructs a comprehensive evaluation system encompassing four dimensions: cost, service, reliability, and computational efficiency. It takes into account operating costs, service coverage, transfer reliability, and computational efficiency, achieving multi-objective collaborative optimization. It can be directly embedded into the existing subway connection operation system without large-scale transformation, has a short investment return cycle, and can provide scientific scheduling decision support for urban subway connection services, promoting the intelligent and efficient development of subway connection services. Attached Figure Description

[0084] Figure 1 The overall flowchart of the connection scheduling method considering the demand for subway train induced pulses provided by the present invention;

[0085] Figure 2 This is a flowchart of S1 in a connection scheduling method that considers the demand for subway train induced pulses according to the present invention;

[0086] Figure 3 This is a flowchart of S2 in a connection scheduling method that considers the demand for subway train induced pulses according to the present invention;

[0087] Figure 4 This is a flowchart of S3 in a connection scheduling method that considers the demand for subway train induced pulses according to the present invention.

[0088] Figure 5 This is a flowchart of step S4 in a connection scheduling method that considers the demand for subway train induced pulses according to the present invention. Detailed Implementation

[0089] 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.

[0090] like Figure 1-5 As shown, the present invention proposes a connection scheduling method that considers the demand for subway train induced pulses. The implementation process includes the following steps:

[0091] S1: Considering the characteristics of subway train schedule guidance, a four-layer progressive pulse demand expression framework is constructed, which includes "time slot mapping - pulse cluster generation - intensity estimation - spatiotemporal coordination calculation", and aggregates scattered individual travel requests into a three-dimensional pulse cluster of station-time slot-direction.

[0092] To accurately characterize the concentrated arrival characteristics of demand induced by subway schedules, this invention proposes a four-layer progressive pulse demand expression framework. The core idea of ​​this framework is to aggregate dispersed individual travel requests according to three dimensions: station, time slot, and direction, forming a three-dimensional pulse cluster with spatiotemporal collaborative value. For example... Figure 2 As shown, the framework includes the following five steps:

[0093] Time slot mapping: Let the reference interval of subway trains be H, and the target connection time of request r be tᵣ. Define the time slot mapping function. This algorithm maps individual requests over continuous time to discrete time slot indices, enabling it to identify highly relevant travel requests within the same subway train schedule window and providing time-dimensional support for request aggregation.

[0094] Pulse cluster construction: Based on the three-dimensional attributes of station m, time slot k, and direction d, a pulse cluster is constructed. and define cluster strength ,in For the number of people requesting r, cluster strength is used to characterize the scale of cluster demand at a specific site, time slot, and direction, intuitively reflecting the intensity of pulse demand.

[0095] Arrival rate characterization: Defining the arrival rate function at the station-direction level ,in For the first The pulse amplitude corresponding to each subway train number The pulse center time corresponds to the arrival time of the subway train. The pulse diffusion coefficient is used to characterize the temporal diffusion range of demand. The background demand term represents regular travel demand not induced by subway schedules. This function can be fitted using historical card swipe data, arrival records, or reservation request data, providing a solid data foundation for the impulse demand structure.

[0096] Robustness estimation: Calculate the expected demand within time slot k With variance Where ξ is the discretization variance coefficient, set according to the actual demand fluctuation, and a robust pulse intensity estimation is defined. Where ζ is the risk preference coefficient. The larger ζ is, the stronger the conservative reservation strategy for potential peak demand, which can effectively improve the robustness of the scheduling scheme under high demand window and avoid service interruption due to demand fluctuations.

[0097] Spatiotemporal Coordination Calculation: Constructing Cluster-Level Spatiotemporal Coordination ,in Time synchronization is calculated using the standard deviation of the target connection times of all requests within the cluster; the smaller the standard deviation, the higher the time synchronization. Spatial compactness is calculated by the average pairwise distance of all pick-up points within the cluster. The smaller the average distance, the higher the spatial compactness. α and β are weighting coefficients that reflect the importance of time synchronization and spatial compactness in collaborative services, respectively, and can be adjusted according to actual operational needs.

[0098] S2: With the goal of minimizing operating costs, unserved losses, and transfer mismatch penalties, a mixed-integer linear programming model with time windows, capacity constraints, and pick-up / delivery pairing constraints is established by combining request allocation, path continuity, time feasibility, and capacity feasibility constraints.

[0099] Objective function: ,in For operating costs, For service loss, Penalty for mismatch during transfer. For the weighting coefficients to satisfy It can be flexibly adjusted according to the actual situation such as city operation strategy and time period demand to achieve multi-objective optimization.

[0100] The objective function is specifically expressed as:

[0101] (1)

[0102] The objective function (1) represents minimizing the total operating cost, which consists of three parts: the first part is the operating cost, including the fixed vehicle activation cost fᵥ and the arc operation cost cᵢⱼ, used to measure the economics of the scheduling scheme; the second part is the unservice loss, i.e., the number of passengers qᵣ corresponding to the unserved requests, used to measure the service coverage; the third part is the transfer mismatch penalty, i.e., the penalty function P corresponding to the lateness time τᵣ of request r. late (τᵣ) is used to measure the reliability of transfers. The delay time is defined as the difference between the time a passenger arrives at the drop-off point and the departure time of the target subway train. The larger the difference, the higher the risk of the passenger missing the target train and the greater the penalty cost.

[0103] The constraints are specifically expressed as follows:

[0104] (2)

[0105] (3)

[0106] (4)

[0107] (5)

[0108] (6)

[0109] (7)

[0110] (8)

[0111] (9)

[0112] (10)

[0113] (11)

[0114] (12)

[0115] (13)

[0116] (14)

[0117] (15)

[0118] Where R represents the set of travel requests; V represents the set of vehicles; and A represents the set of road network arcs. The fixed start-up cost for vehicle v; Enable state variables for vehicle v; For arc The unit operating cost; For vehicle v to pass through arc State variables; The number of people requesting r; The state variable for request r being served; To request the lateness duration of r; For late arrival penalty function; The state variable for requesting r to be served by vehicle v; , These represent the pick-up and drop-off points for request r, respectively; 0 indicates a parking lot node. These are the start and end times of the service time window for node i, respectively. The service start time for vehicle v at node i; The service time for node i; For arc Travel time; M is a sufficiently large constant; The maximum acceptable travel time for request r; The passenger load when vehicle v leaves node i; The rated capacity of vehicle v; , These represent the times when vehicle v departs from and returns to the parking lot, respectively. This refers to the maximum operating time for a single shift of the vehicle. For the weighting coefficients to satisfy .

[0119] Constraint (2) indicates the consistency between the service requested and the vehicle allocation, ensuring that each request can only be served by one vehicle; Constraints (3) and (4) indicate that the pick-up point and drop-off point must be accessed by the same vehicle at the same time, ensuring the integrity of the shuttle service; Constraint (5) indicates that the vehicle departs from the parking lot and eventually returns to the parking lot, forming a closed path, which conforms to the actual operation specifications; Constraint (6) indicates the balance of vehicle inflow and outflow at intermediate nodes, ensuring the continuity of the path and avoiding vehicle stagnation or path breakage at the nodes; Constraint (7) indicates the time recursion relationship, ensuring that the vehicle accesses between nodes. The time order; constraint (8) indicates that the vehicle completes the node service within the specified time to meet the passenger's time requirements; constraint (9) indicates that the vehicle visits the pick-up point first and then the corresponding drop-off point, which conforms to the logical order of the connection service; constraint (10) avoids excessive passenger travel time and improves the travel experience; constraints (11) and (12) clarify the definition of the late time of request r, and provide a basis for the calculation of transfer mismatch penalty; constraint (13) ensures the continuity of passenger capacity calculation; constraint (14) avoids vehicle overloading and ensures operational safety; constraint (15) conforms to the actual operating time requirements.

[0120] S3: Based on the impulse demand structure and mixed integer linear programming model, the Pulse-ALNS solution framework is designed. The optimal connection scheduling scheme is solved through the steps of initial solution construction, impulse sensing destruction, train-batch repair, solution evaluation and acceptance, weight update and computation acceleration.

[0121] This framework introduces a pulse sensing mechanism on the basis of the traditional ALNS algorithm, taking the pulse cluster structure as the core operation object of the search process, making full use of the collaborative value of requests within the same cluster, and improving the solution efficiency and solution quality.

[0122] Initial solution construction: A two-layer solution representation method of "vehicle path sequence + request status index" is adopted. The path layer records the access node order of each vehicle, and the status layer records the vehicle allocation, service time, lateness status, and pulse cluster to which each request belongs. A hierarchical heuristic strategy is adopted, first grouping requests into pulses according to the three-dimensional attributes of station-time slot-direction, and then grouping them according to cluster strength. Time urgency and time-space coordination The initial path is constructed step by step using a minimum incremental insertion strategy. Requests that cannot be directly inserted are temporarily stored in the unserved pool and then periodically retried for insertion to ensure the feasibility and rationality of the initial solution.

[0123] Impulse-sensing destruction: Defining cluster removal priority scores ,in The weighting coefficients respectively reflect the importance of spatiotemporal coordination, default risk, and detour cost. To mitigate the risk of default or delay in requests within the cluster, This represents the marginal detour cost introduced by the cluster in the current solution. Three types of disruptive operators are employed: a whole-cluster removal operator identifies high-risk clusters based on priority scores and removes them entirely; a local sampling removal operator samples and removes requests within a specific site and time slot; and a high-penalty request removal operator prioritizes removing the set of requests with the highest lateness penalty contribution, while retaining a small number of random requests for removal to improve search diversity and prevent premature convergence. A Train-Batch repair mechanism is proposed, prioritizing highly collaborative requests from the same site, time slot, and direction as a batch for overall insertion to improve repair efficiency. The batch insertion increment is defined. ,in Let γ be the individual incremental cost of inserting vehicle v into the path of request r, and let γ be the batch discount factor, preferably 0.8, to reflect the collaborative advantage of batch insertion requests within the same cluster. Conflict penalties are used to penalize batches that generate constraint conflicts after insertion. A two-level rollback mechanism is adopted: the first-level rollback performs partial whole-cluster insertions based on the lateness risk or time urgency of requests within the cluster; the second-level rollback switches to request-by-request insertion to ensure feasibility under complex constraints.

[0124] Solution Evaluation and Acceptance: The simulated annealing (SA) acceptance criterion is used to balance the global search capability and local convergence capability of the algorithm. Let the objective difference between the candidate solution and the current solution be... The probability of acceptance is ,in For temperature parameters (λ is the cooling coefficient) decreasing. Simultaneously, an adaptive weight update mechanism is introduced. Let the set of operators be Ω, and the operators... The weight is After the evaluation window ends, scores are given based on the operator's contribution to the improvement of the global optimum, the improvement of the current solution, and the feasibility recovery. Perform weight update (ρ is the update coefficient), operator selection probability This enables the algorithm to adaptively adjust its search strategy.

[0125] Computational acceleration: An integrated acceleration module is used to reduce the cost of unit iteration evaluation and ensure real-time scheduling requirements. Specifically, it includes three strategies: 1. Based on the triangle inequality, a lower bound for cost is constructed to prune invalid insertion positions and reduce invalid calculations; 2. A version number-driven lazy re-evaluation mechanism is used to perform local recalculation only on path segments affected by disturbances to avoid the redundant overhead of global recalculation; 3. A selective repair strategy is used to prioritize the processing of removal requests in this round and perform full repair after a fixed period to improve repair efficiency.

[0126] S4: Assign the optimal scheduling plan to the corresponding vehicles, monitor vehicle status, subway schedule changes and road congestion in real time, and dynamically adjust the scheduling plan to ensure service reliability.

[0127] New request handling: The request update cycle is set to 30 seconds. When a new travel request arrives, the request is only re-inserted into vehicles with unlocked tasks (unlocked tasks refer to tasks that are more than 10 minutes away from the next pick-up point). The locked task paths and times are not adjusted to ensure the stability of the original scheduling scheme. The average response time for new requests is ≤5 seconds.

[0128] Handling subway schedule changes: Setting a safety margin for transfers The adjustment period is typically 2-5 minutes, depending on the size of the subway station. It receives real-time updates on train schedules from the subway operation system. If a delay or arrival exceeding one minute is detected, the transfer availability for each request is recalculated. ( This refers to the arrival times of subway trains. (The estimated time the vehicle will arrive at the drop-off point), if The system will then re-match the nearest subway train and adjust the speed and stopping order of the corresponding trains to ensure that passengers can transfer smoothly.

[0129] Road congestion management: Real-time acquisition of road congestion status via road network data interface, and real-time monitoring of remaining time for vehicles to reach their drop-off points. ,like The emergency response plan includes: dispatching the nearest backup vehicle to make high-priority requests for connecting and diverting passengers; pushing subsequent subway train information and transfer suggestions to affected passengers; planning emergency priority passage routes for connecting vehicles; coordinating regional traffic control to reduce traffic delays; and ensuring that the average response time of the emergency response plan is ≤1 minute.

[0130] The above is an exemplary description of the invention. Obviously, the specific implementation of the invention is not limited to the above-described manner. Any non-substantial improvement made using the inventive concept and technical solution of the invention, or the direct application of the inventive concept and technical solution to other situations without modification, is within the protection scope of the invention.

Claims

1. A connection scheduling method considering the demand for subway train induced pulses, characterized in that, Includes the following steps: S1: Considering the characteristics of subway train schedule guidance, a four-layer progressive pulse demand expression framework is constructed, which includes time slot mapping, pulse cluster generation, intensity estimation, and spatiotemporal coordination calculation. This framework aggregates scattered individual travel requests into a three-dimensional pulse cluster consisting of station, time slot, and direction. S2: With the goal of minimizing operating costs, unserved losses, and transfer mismatch penalties, a mixed integer linear programming model with time windows, capacity constraints, and pick-up / delivery pairing constraints is established by combining request allocation, path continuity, time feasibility, and capacity feasibility constraints. S3: Based on the impulse demand structure and mixed integer linear programming model, the Pulse-ALNS solution framework is designed. The optimal connection scheduling scheme is solved through the steps of initial solution construction, impulse sensing destruction, Train-Batch repair, solution evaluation and acceptance, weight update and computation acceleration. S4: Assign the optimal scheduling plan to the corresponding vehicles, monitor vehicle status, subway schedule changes and road congestion in real time, and dynamically adjust the scheduling plan to ensure service reliability.

2. The large-span current-carrying acceleration test apparatus for evaluating the durability of the connection point according to claim 1, characterized in that, The pulse demand modeling in step S1 specifically includes: S101: Time slot mapping, assuming the subway train frequency reference interval is... The target connection time for request r is Define the time slot mapping function: This maps individual requests over continuous time to discrete time slot indexes. S102: Pulse cluster construction, based on the three-dimensional attributes of station m, time slot k, and direction d, constructs a pulse cluster: And define cluster strength: in For the request The number of passengers; S103: Arrival rate characterization, defining the arrival rate function at the station-direction level: in For the first The pulse amplitude corresponding to each subway train number , The pulse center time, The pulse spread coefficient is... Background requirements; S104: Robustness estimation, calculating the expected demand and variance within time slot k: Define robust pulse intensity estimation: in These are the discretized variance coefficients. This is the risk preference coefficient; S105: Spatiotemporal coordination calculation, constructing cluster-level spatiotemporal coordination: in, For time synchronization of requests within the cluster, For the sake of space compactness of intra-cluster requests, , These are the weighting coefficients.

3. The large-span current-carrying acceleration test apparatus for evaluating the durability of the connection point according to claim 1, characterized in that, The mixed-integer linear programming model mentioned in step S2 specifically includes: S201: The objective function is expressed as: （1） Where V represents the set of vehicles; A represents the set of road network arcs; and R represents the set of travel requests. The fixed start-up cost for vehicle v; Enable state variables for vehicle v; For arc The unit operating cost; For vehicle v to pass through arc State variables; The state variable for request r being served; To request the lateness duration of r; For late arrival penalty function; S202: The constraint conditions are specifically expressed as follows: （2） （3） （4） （5） （6） （7） （8） （9） （10） （11） （12） （13） （14） （15） Where R represents the set of travel requests; The state variable for requesting r to be served by vehicle v; , These represent the pick-up and drop-off points for request r, respectively; 0 indicates a parking lot node. , These are the start and end times of the service time window for node i, respectively. The service start time for vehicle v at node i; The service time for node i; For arc Travel time; M is a sufficiently large constant; The maximum acceptable travel time for request r; The passenger load when vehicle v leaves node i; The rated capacity of vehicle v; , These represent the times when vehicle v departs from and returns to the parking lot, respectively. This refers to the maximum operating time for a single shift of a vehicle.

4. The large-span current-carrying acceleration test apparatus for evaluating the durability of the connection point according to claim 1, characterized in that, Step S3 further includes: S301: Initial solution construction adopts a two-layer solution representation method of vehicle path sequence + request status index. The path layer records the access node order of each vehicle, and the status layer records the vehicle allocation, service time, late status and pulse cluster to which each request belongs. S302: Pulse-sensing disruption, defining cluster removal priority score: in, , , These are the weighting coefficients. To mitigate the risk of in-cluster request default, To determine the cost of bypassing the cluster edge, three types of destructive operators are employed: whole-cluster removal operator, local sampling removal operator, and high-penalty request removal operator. S303: Train-Batch Fix, Define Batch Insert Increment: in For the request Insert vehicle Individual incremental costs, For bulk discount factors, Conflict penalties are imposed; a two-tier rollback and repair mechanism is employed to ensure the feasibility of repairs. S304: Solution Evaluation and Acceptance. Simulated annealing is used as the acceptance criterion. Let the objective difference between the candidate solution and the current solution be... The probability of acceptance is: in The temperature parameter at each iteration time; S305: Weight Update, introducing an adaptive weight update mechanism. Let the set of operators be Ω, and the operators... The weight is Perform weight update: Operator selection probability ; S306: Computational acceleration, employing an integrated acceleration module to reduce the cost per unit iteration evaluation, including constructing a lower bound for cost based on the triangle inequality, adopting a version number-driven lazy re-evaluation mechanism, and employing a selective repair strategy.

5. The large-span current-carrying acceleration test apparatus for evaluating the durability of the connection point according to claim 1, characterized in that, Step S4 further includes: S401: Add request processing, set the request update cycle to 30 seconds, when a new travel request arrives, only re-insert the request in vehicles with unlocked tasks; S402: Handling subway schedule changes and setting transfer safety margins Every 2-5 minutes, the system receives real-time train schedule updates from the subway operation system and recalculates the transfer availability for each request. in This refers to the arrival times of subway trains. If the estimated arrival time of the vehicle at the drop-off point is... < If so, the subway train schedules will be rematched and the vehicle dispatching plan will be adjusted. S403: Road congestion management, real-time acquisition of road segment congestion status through road network data interface, and real-time monitoring of the remaining time for vehicles to reach their drop-off points: ,like This will trigger the emergency response plan.

6. The large-span current-carrying acceleration test apparatus for evaluating the durability of the connection point according to claim 1, characterized in that, In step S105, time synchronization Spatial compactness is calculated by the standard deviation of the connection times of all requesting targets within the cluster. It is calculated by the average pairwise distance of all visitor points within the cluster.

7. The large-span current-carrying acceleration test apparatus for evaluating the durability of the connection point according to claim 1, characterized in that, The lateness penalty function It is a piecewise linear function, and the penalty value varies with the duration of lateness. The increase is in a step-like manner.

8. The large-span current-carrying acceleration test apparatus for evaluating the durability of the connection point according to claim 1, characterized in that, The initial solution construction in step S301 involves first grouping requests into pulse groups based on the three-dimensional attributes of station-time slot-direction, then sorting them by cluster strength, time urgency, and spatiotemporal coordination, and constructing the initial path using a minimum incremental insertion strategy.

9. A connection scheduling system considering the demand for subway train induced pulses, used to implement the connection scheduling method according to any one of claims 1-8, characterized in that, include: The pulse demand modeling module is used to construct a four-layer progressive pulse demand expression framework, which aggregates scattered individual travel requests into a three-dimensional pulse cluster of station-time slot-direction; The optimization modeling module is used to build a constrained mixed-integer linear programming model with the goal of minimizing operating costs, unserved losses, and transfer mismatch penalties. The intelligent solution module has a built-in Pulse-ALNS solution framework for solving optimal connection scheduling schemes. The dynamic scheduling module is used to allocate scheduling plans to corresponding vehicles, monitor the operational status in real time, and dynamically adjust the scheduling plans. The human-computer interaction module communicates with each of the above modules and is used for parameter configuration, data display, log query and fault alarm.

10. The system according to claim 9, characterized in that, It also includes a data interface module, which is used to interact with the subway operation system, road network data platform, vehicle terminal and factory MES system to obtain real-time subway schedule dynamics, road congestion status and vehicle operation data.

Images