A Public Transport Automatic Scheduling Method with Multi-Model Mutual Learning

Abstract

The present invention discloses a public transportation automatic scheduling method for multi-model mutual learning, which relates to the technical field of public transportation. It obtains the real-time arrival, departure, and stop times of each vehicle on each route at each stop for each trip, as well as the historical arrival, departure, and stop times; obtains the real-time passenger flow and historical passenger flow of the stop; based on the collected data, it analyzes in real time various scheduling events occurring in actual operation, including vehicle failure events, large passenger flow events, and vehicle delay events; generates corresponding scheduling plans for each scheduling event; and the dispatcher makes a final scheduling decision according to each scheduling event and the corresponding scheduling plan. By establishing analysis models and processing models for different scheduling events, integrating the characteristics and results of different models, and dynamically selecting the most suitable model according to real-time scheduling requirements, the present invention realizes simulation prediction and scheduling decision, and finally realizes the intelligent and automatic scheduling of bus vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of public transportation, and in particular to a public transportation automatic scheduling method for multi-model mutual learning. Background Art

[0002] Currently, intelligent scheduling systems have been introduced in the bus systems of many cities. By real-time monitoring information such as vehicle positions and passenger flows, the scheduling plans are optimized. However, most of the current scheduling products still require a large amount of manual operations by dispatchers, with limited automation and unable to fully exploit the potential of intelligent technologies. Summary of the Invention

[0003] In order to overcome the defects in the above-mentioned prior art, the present invention provides a public transportation automatic scheduling method for multi-model mutual learning, which establishes analysis models and processing models for different scheduling events, combines the characteristics and results of different models, dynamically selects the most suitable model according to real-time scheduling requirements, realizes simulation prediction and scheduling decision-making, and finally realizes the intelligent and automatic scheduling of bus vehicles.

[0004] To achieve the above object, the present invention adopts the following technical solutions, including:

[0005] A public transportation automatic scheduling method for multi-model mutual learning, including:

[0006] Data collection: Obtain the real-time arrival, departure, and stop times of each vehicle on each line at each stop for each trip, as well as the historical arrival, departure, and stop times; obtain the real-time passenger flow and historical passenger flow of the stop.

[0007] Scheduling event analysis: Based on the collected data, analyze in real time each scheduling event occurring in actual operation, including vehicle failure events, large passenger flow events, and vehicle delay events.

[0008] Scheduling event processing: Generate corresponding scheduling plans for each scheduling event.

[0009] Preferably, the analysis of vehicle failure events is as follows:

[0010] If the vehicle speed drops by more than a set range and the vehicle position does not change for a certain period of time, it is determined as a vehicle failure event; otherwise, it is not determined as a vehicle failure event.

[0011] Analyze the impact brought by the vehicle failure event: Take the average total passenger flow from the vehicle failure position to the terminal at the same historical time as the passenger flow P0 affected by the vehicle failure event; the delay of the next vehicle's driving time caused by the vehicle failure event is T g =P0×λ; where λ is the average boarding and alighting time of each passenger.

[0012] Preferably, the analysis of large passenger flow events is as follows:

[0013] Predict the passenger flow at each station within a future period of time. If the predicted passenger flow P1 within the future period of time is greater than k1 times the historical passenger flow P2 at the same time period, it is determined as a large passenger flow event; otherwise, it is not determined as a large passenger flow event. Among them, 1 < k1 < 1.1;

[0014] Analyze the impact brought by the large passenger flow event: The passenger flow affected by the large passenger flow event is P3 = P1 - P2; The delay of the next vehicle's running time due to the large passenger flow is T k = P3 × λ; where λ is the average boarding and alighting time of each passenger;

[0015] Classify the level of the large passenger flow situation: The greater the passenger flow P3 affected by the large passenger flow event, the higher the level, indicating that the large passenger flow situation is more urgent.

[0016] Preferably, based on the historical passenger flow of the station, combined with the real-time passenger flow of the station and the passenger flow affected by vehicle failure events, predict the passenger flow at the station within a future period of time.

[0017] Preferably, the analysis of vehicle delay events is as follows:

[0018] Calculate the predicted running time T of the vehicle from the current position to the terminal station p as:

[0019] ;

[0020] Among them, the station closest to the current position of the vehicle is the mth station, 1 ≤ m ≤ n, and there are n stations in total; T m is the running time of the vehicle from the current position to the closest station, that is, the mth station; T i,i+1 is the running time of the vehicle from the ith station to the (i + 1)th station; D i is the stopping time of the vehicle at the ith station; T g is the delay of the previous vehicle failure event on the running time of this vehicle; T k is the delay of the large passenger flow event on the running time of this vehicle; α and β are the weights of vehicle failure events and large passenger flow events respectively;

[0021] According to the predicted running time T of the vehicle from the current position to the terminal station p , obtain the predicted arrival time of the vehicle;

[0022] Compare the predicted arrival time with the planned arrival time in the vehicle operation plan. If the predicted arrival time is later than the planned arrival time, it is determined as a vehicle delay event; otherwise, it is not determined as a vehicle delay event;

[0023] Classify the level of vehicle delays: The number of remaining stops from the current position of the delayed vehicle to the terminal is n - m + 1, and the proportion of the remaining stops in the total number of stops is (n - m + 1) / n; the larger the proportion of the remaining stops in the total number of stops, the higher the level, indicating that the vehicle delay situation is more urgent.

[0024] Preferably, according to the historical arrival, departure, and stop times, predict the travel time between each pair of stops and the stop time at each stop.

[0025] Preferably, generate an optimal scheduling plan based on the analysis results of each scheduling event, as follows:

[0026] For vehicle failure events, based on the status of the remaining vehicles on the same route as the faulty vehicle and the original operation plan, adjust the departure times of the remaining vehicles with the goal of minimizing the number of adjusted trips to generate a new operation plan;

[0027] For large passenger flow events, generate an additional vehicle scheduling plan based on the location of large passenger flow stations and the predicted passenger flow in the next period to meet the actual passenger flow demand;

[0028] For vehicle delay events, generate a skipping or detouring scheduling plan based on the actual operation situation and the predicted arrival time to meet the vehicle operation plan.

[0029] Preferably, present the scheduling plans of each scheduling event in a card form on the scheduling interface and display the urgency of each scheduling event.

[0030] Preferably, it also includes scheduling decision-making: The dispatcher makes a final scheduling decision based on each scheduling event and the corresponding scheduling plan, that is, the dispatcher confirms whether to adopt the scheduling plan or adjusts and then adopts the scheduling plan.

[0031] The present invention also provides a multi-model mutual learning public transportation automated scheduling system applicable to the above-mentioned multi-model mutual learning public transportation automated scheduling method. The system includes: a data acquisition module, a scheduling event analysis module, a scheduling event processing module, and a scheduling decision-making module;

[0032] The data acquisition module obtains the following data: Obtain the real-time arrival, departure, and stop times of each vehicle on each route for each trip at each stop and the historical arrival, departure, and stop times; obtain the real-time passenger flow and historical passenger flow of the stops;

[0033] The dispatching event analysis module includes a vehicle failure analysis model, a large passenger flow prediction model, and a vehicle delay prediction model, which are respectively used to analyze in real time vehicle failure events, large passenger flow events, and vehicle delay events occurring in actual operations; the vehicle failure analysis model, the large passenger flow prediction model, and the vehicle delay prediction model respectively input the analysis results into the dispatching event processing module for generating corresponding dispatching plans; among them, the large passenger flow prediction model combines the analysis results of the vehicle failure analysis model to analyze large passenger flow events; the vehicle delay prediction model combines the analysis results of the vehicle failure analysis model and the large passenger flow prediction model to analyze vehicle delay events.

[0034] The dispatching event processing module includes a vehicle failure processing model, a large passenger flow processing model, and a vehicle delay processing model, which are used to generate dispatching plans for different dispatching events.

[0035] The dispatching decision-making module is used for the dispatcher to view each dispatching event and the corresponding dispatching plan and make a final dispatching decision.

[0036] The advantages of the present invention are as follows:

[0037] (1) By using advanced information technology, big data analysis and other means, the present invention establishes analysis models and processing models for different dispatching events, combines the characteristics and results of different models, dynamically selects the most suitable model according to real-time dispatching requirements, realizes simulation prediction and dispatching decision-making, and finally realizes the intelligent and automatic dispatching of bus vehicles.

[0038] (2) A data closed-loop is formed among the vehicle failure analysis, large passenger flow prediction and vehicle delay prediction of the present invention (such as the vehicle failure analysis result is used for large passenger flow prediction and vehicle delay prediction), realizing cross-validation and correction in a dynamic environment, data feedback between multiple models, improving the accuracy of complex event correlation analysis, and enhancing the overall analysis accuracy.

[0039] (3) For vehicle failure events, the present invention not only calculates the current delay, but also predicts the chain reaction through historical passenger flow data at the same location, avoiding systemic paralysis caused by local problems.

[0040] (4) The present invention introduces weight coefficients of vehicle failure events and large passenger flow events in vehicle delay analysis, optimizes weight allocation through historical data learning, and enhances the adaptability of the model to complex scenarios.

[0041] (5) The present invention classifies the urgency of large passenger flow events and vehicle delay events, combines quantitative indicators such as passenger flow increment and delay time, provides a priority basis for subsequent dispatching strategies, and is conducive to the accurate allocation of resources according to the urgency.

[0042] (6) For vehicle failure events, the present invention generates a new operation plan with the goal of minimizing the number of adjusted train trips, reducing operation costs, and minimizing scheduling disruptions. For large passenger flows, an additional train strategy is adopted, and for delays, a skip-stop or detour strategy is adopted to balance the passenger waiting time and the overall line efficiency.

[0043] (7)After the present invention generates a scheduling plan, it displays each scheduling event and the urgency of the event in the form of cards on the scheduling interface, allowing manual adjustment methods, taking into account both algorithm efficiency and dispatcher experience.

[0044] (8)The public transportation automatic scheduling method of the present invention can improve operation efficiency, reduce operation costs, and enhance the passenger experience.

[0045] (9)The public transportation automatic scheduling system of the present invention adopts a modular design (data collection, analysis, processing, decision-making), has strong scalability, supports the rapid access of new event types, and the model mutual learning framework reserves an interface for integrating more AI algorithms in the future.

[0046] (10)Through multi-model collaborative analysis, hierarchical response mechanisms, and human-machine collaborative decision-making, the present invention constructs an efficient and low-cost public transportation automatic and intelligent scheduling method and system, providing an innovative solution to the problems of real-time and complexity in urban public transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of a public transportation automatic scheduling system with multi-model mutual learning.

[0048] Figure 2 It is a flowchart of a public transportation automatic scheduling method with multi-model mutual learning. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0050] Embodiment 1

[0051] As shown by Figure 1 A public transportation automatic scheduling system with multi-model mutual learning includes: a data collection module, a scheduling event analysis module, a scheduling event processing module, and a scheduling decision module.

[0052] The data acquisition module obtains the following data: obtain the operation plan of the line vehicles; obtain the real-time arrival, departure, and stop times of each vehicle on each line for each trip at each station, as well as the arrival, departure, and stop times within a certain period of historical time; obtain the real-time passenger flow (waiting passengers) and historical passenger flow of the station; at the same time, collect data on uncertain factors such as weather, sudden accidents, and route changes.

[0053] The dispatching event analysis module is used to monitor and analyze various dispatching events occurring in actual operations in real time, including vehicle failures (or vehicle accidents), large passenger flows, and operation plan change events caused by vehicle delays, simply referred to as vehicle failure events, large passenger flow events, and vehicle delay events.

[0054] Multiple models are built in the dispatching event analysis module, which are respectively used to analyze each dispatching event, and input the analysis results of each dispatching event into the dispatching event processing module to generate corresponding dispatching plans.

[0055] Specifically, the dispatching event analysis module includes a vehicle failure analysis model, a large passenger flow prediction model, and a vehicle delay prediction model.

[0056] The vehicle failure analysis model determines whether a vehicle has a failure or an accident through certain rules, so as to judge whether the vehicle can continue to operate. If the vehicle speed suddenly drops and stays stagnant for a long time, that is, the speed drop amplitude exceeds the set range and the vehicle position does not change for a certain period of time, it is determined that the vehicle has a failure or an accident. Specifically, the on-vehicle device uploads the real-time vehicle speed every 10 seconds. During the non-station entry and exit period, the vehicle driving speed suddenly drops significantly, for example, from 30 km / h suddenly drops to less than 5 km / h and finally drops to 0 km / h.

[0057] The vehicle failure analysis model takes factors such as the location, time, and current weather of the vehicle failure as inputs, and determines and outputs the influence results of the vehicle failure event, including the passenger flow and road traffic conditions affected by the vehicle failure event; among them, the road traffic conditions affected by the event specifically refer to the delay of the driving time of the next vehicle caused by the event.

[0058] Take the average total passenger flow from the vehicle failure location to the terminal station at the same factor and the same time in history as the passenger flow P0 affected by the vehicle failure, and calculate the delay of the driving time of the next vehicle caused by the vehicle failure as T g = P0×λ; where λ is the average boarding and alighting time of each passenger, which is determined by the historical stop time and the number of boarding and alighting passengers.

[0059] For example, at 7:35 during the peak period, vehicle V on the upward route A breaks down at stop i. If the average total passenger flow from stop i to the terminal on the upward route under the same historical factors and at the same time is P0, then the passenger flow affected by this vehicle breakdown event is P0, and the delay in the travel time of the next vehicle due to this vehicle breakdown event is T g = P0 / λ.

[0060] The vehicle breakdown analysis model takes the impact results of vehicle breakdown events as an input variable for subsequent vehicle delay prediction and large passenger flow prediction, and inputs them into the vehicle delay prediction model and the large passenger flow prediction model respectively; at the same time, it also inputs the impact results of vehicle breakdown events into the dispatching event processing module to generate a dispatching plan, and then the dispatcher confirms the events and the plan

[0061] For each stop, the large passenger flow prediction model, based on the historical passenger flow data of the stop, historical weather, working days or holidays, seasons and other factors, combines the number of passengers waiting (real-time passenger flow at the stop) collected in real time at the stop, and the passenger flow affected by vehicle breakdown events, etc., and uses a general passenger flow prediction algorithm to predict the passenger flow within a future period (30 minutes). The predicted passenger flow P1 within the future period is compared with the passenger flow P2 in the same historical period under the same factors. If the predicted passenger flow P1 is greater than 1.1 times the passenger flow P2 in the same historical period under the same factors, it is determined as a large passenger flow event, and the large passenger flow situation is classified

[0062] The classification of the large passenger flow situation is as follows:

[0063] If the predicted passenger flow P1 is greater than 1.1 times and less than or equal to 1.2 times the passenger flow P2 in the same historical period under the same factors, the large passenger flow is determined to be of grade one; if the predicted passenger flow P1 is greater than 1.2 times and less than or equal to 1.5 times the passenger flow P2 in the same historical period under the same factors, the large passenger flow is determined to be of grade two; if the predicted passenger flow P1 is greater than 1.5 times the passenger flow P2 in the same historical period under the same factors, the large passenger flow is determined to be of grade three; the higher the grade, the more urgent the large passenger flow situation

[0064] The large passenger flow prediction model takes the large passenger flow location (stop), time, current weather, predicted passenger flow and grade as inputs, determines and outputs the impact results of large passenger flow events, including the passenger flow and road traffic conditions affected by large passenger flow events

[0065] If the large passenger flow is of grade one, the passenger flow affected by the large passenger flow event is P3 = P1 - P2, and the delay in the travel time of the next vehicle due to the large passenger flow event is T k = P3×λ; where λ is the average boarding and alighting time of each passenger, which is determined by the historical stop time and the number of boarding and alighting passengers

[0066] If the level of the large passenger flow is level two or level three, the passenger flow affected by the large passenger flow event is P3 = P1 - P2. However, at this time, the delay T of the driving time of the next vehicle caused by the large passenger flow event k needs to be determined according to the processing result after the dispatching decision. It is necessary to return according to the dispatching strategy selected by the dispatcher and then determine the delay of the driving time of the next vehicle.

[0067] For example, at 7:15 during the peak period, a large passenger flow occurs at station j on the up line of line A, and the passenger flow affected by the large passenger flow event is P3, and the level of the large passenger flow is level one. Then the delay of the driving time of the next vehicle caused by the large passenger flow event is T k = P3 × λ.

[0068] The large passenger flow prediction model takes the influence result of the large passenger flow event as an input variable for the subsequent vehicle delay prediction and inputs it into the vehicle delay prediction model; at the same time, it also inputs the influence result of the large passenger flow event into the dispatching event processing module to generate a dispatching plan, and then the dispatcher confirms the event and the plan.

[0069] The vehicle delay prediction model is used to predict the driving time of the vehicle from any position after the start of operation to the terminal station. Among them, the driving time of the vehicle from any position after the start of operation to the terminal station consists of the stop time at each intermediate station, the driving time of the vehicle from the current position to its nearest station, the driving time between each intermediate station, plus the influence variables output by other models.

[0070] The vehicle delay prediction model calculates the predicted driving time T of the vehicle from the current position to the terminal station p as:

[0071] ;

[0072] Among them, the nearest station to the current position of the vehicle is the mth station, 1 ≤ m ≤ n, there are a total of n stations, and the terminal station is the nth station; T m is the driving time of the vehicle from the current position to the nearest station, that is, the mth station; T i,i+1 is the driving time of the vehicle from the ith station to the i + 1th station; D i is the stop time of the vehicle at the ith station; T g is the delay of the driving time of this vehicle caused by the vehicle failure of the previous vehicle; T g is the delay of the driving time of this vehicle caused by the large passenger flow, T g and T kVariables input into the vehicle fault analysis model and the large passenger flow prediction model respectively; α and β are the weights of the vehicle fault event and the large passenger flow event respectively, and these weights are dynamically adjusted according to the stability and accuracy of the output results of other historical models after the processing of the scheduling decision module; during the peak period, the result of the vehicle fault event has a greater impact on the prediction of vehicle delay, so the weight of the vehicle fault event will be increased during the peak period; during the off-peak period, the large passenger flow event has a greater impact on the prediction of vehicle delay, so the weight of the large passenger flow event will be increased during the off-peak period.

[0073] In this embodiment, the vehicle delay prediction model classifies the inbound and outbound data based on historical GPS data, weather, weekdays or holidays, seasons and other factors. By combining the above influencing factors and with a time granularity of 15 minutes, the driving time data between each pair of stations and the docking time data of each station within each time period under different historical factors can be obtained, including the mean and standard deviation. Based on the weather, weekdays or holidays, seasons and other conditions on the prediction day, the data under the same historical factors is selected as the prediction basis, and a general time series prediction algorithm is used to predict the driving time between each pair of stations and the docking time of each station within each time period under different factors.

[0074] According to the above formula, the predicted arrival time of the vehicle can be obtained based on the predicted driving time of the vehicle from the current position to the terminal station.

[0075] The vehicle delay prediction model compares the predicted arrival time of the vehicle with the planned arrival time in the operation plan. If the predicted arrival time is later than the planned arrival time, it is determined that the vehicle is delayed, and the vehicle delay situation is classified according to the current position of the delayed vehicle.

[0076] The classification of the vehicle delay situation is as follows:

[0077] The number of remaining stations from the current position of the delayed vehicle to the terminal station is n - m + 1; the proportion of the number of remaining stations to the total number of stations is (n - m + 1) / n. If the proportion (n - m + 1) / n of the number of remaining stations to the total number of stations is greater than 40%, the vehicle delay level is level one; if the proportion (n - m + 1) / n of the number of remaining stations to the total number of stations is greater than 20% and less than or equal to 40%, the vehicle delay level is level two; if the proportion (n - m + 1) / n of the number of remaining stations to the total number of stations is less than or equal to 20%, the vehicle delay level is level three; the higher the level, the more urgent the vehicle delay situation.

[0078] The vehicle delay prediction model inputs the predicted arrival time of the delayed vehicle and the corresponding situation level into the scheduling event processing module to generate a scheduling plan.

[0079] The scheduling event processing module has different processing models (based on a general machine learning model) for different scheduling events, including vehicle failure handling models, large passenger flow handling models, and vehicle delay handling models, which are used to generate scheduling plans for different scheduling events.

[0080] For vehicle failure events, since the vehicle cannot continue to operate according to the plan due to failures or accidents during actual operation, it is necessary to adjust the original operation plan in real time to meet subsequent operation requirements. Specifically, according to the status of the remaining vehicles on the same line as the faulty vehicle, including the status of the currently operating vehicles and the situation of the vehicles that can operate subsequently, the departure times of the remaining vehicles are adjusted based on the original operation plan. With the goal of minimizing the number of adjusted trips, factors such as the headway and turnover time in different peak periods are comprehensively considered, and a new operation plan with reduced trips is generated from a global perspective.

[0081] For large passenger flow events, according to the location of large passenger flow stations and the predicted passenger flow in the next period of time, a scheduling plan for adding vehicles is generated to meet the actual passenger flow demand.

[0082] For vehicle delay events, according to the actual operation situation and the predicted arrival time, a skipping-stop or detour scheduling plan is generated to meet the vehicle operation plan.

[0083] Among them, skipping-stop means that in the public transportation system, the stops at some stations are temporarily cancelled to speed up the operation speed or deal with special situations. Detour means that in traffic route planning, in order to avoid congestion, accidents or other obstacles, an alternative route is selected to reach the destination. Adding vehicles means that in the public transportation system, additional vehicles or trips are added to meet the passenger demand or relieve the congestion during peak hours.

[0084] In this embodiment, the scheduling event processing module establishes multiple different processing models according to specific scheduling requirements, and can dynamically select the most suitable model for processing according to real-time scheduling requirements and resource status, and generate a scheduling plan corresponding to the scheduling event.

[0085] The scheduling decision-making module centrally presents each scheduling event and the corresponding scheduling plan in the form of cards on the scheduling interface, and uses different colors to display the urgency levels of different scheduling events, which is used for the dispatcher to view each scheduling event and the corresponding scheduling plan and make the final scheduling decision.

[0086] For automated scheduling, the dispatcher mainly plays a supervisory role. Since the system has a high degree of automation, the dispatcher only needs to pay attention to what scheduling events exist and the generated scheduling plans. After the dispatcher enters each card, they can observe the results of the scheduling plan, and decide whether to adopt the scheduling plan according to the actual situation, or adjust the generated scheduling plan and then adopt it, so as to form the final scheduling decision.

[0087] Example 2

[0088] As Figure 2 shown, based on the multi-model mutual learning public transportation automatic scheduling system provided in the above-mentioned Embodiment 1, the specific process of the public transportation automatic scheduling method is as follows:

[0089] S1. Data collection: Obtain the real-time arrival, departure, and stop times of each vehicle on each route at each stop for each trip, as well as the historical arrival, departure, and stop times; obtain the real-time passenger flow and historical passenger flow of the stop; at the same time, collect data on uncertain factors such as weather data, sudden accidents, and route diversions.

[0090] S2. Scheduling event analysis: Based on the collected data, analyze in real time each scheduling event that occurs in actual operation, including vehicle failure events, large passenger flow events, and vehicle delay events;

[0091] S3. Scheduling event handling: Generate corresponding scheduling plans for each scheduling event;

[0092] S4. Scheduling decision: The dispatcher makes a final scheduling decision according to each scheduling event and the corresponding scheduling plan, that is, the dispatcher confirms whether to adopt the scheduling plan or adopts it after adjusting the scheduling plan.

[0093] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A public transportation automated scheduling method for multi-model mutual learning, characterized in that, Including: Data collection: Obtain the real-time arrival, departure, and stop times of each vehicle on each route at each station for each trip, as well as the historical arrival, departure, and stop times; Obtain the real-time and historical passenger flows of the stations; Dispatch event analysis: Based on the collected data, analyze in real time each dispatch event that occurs during actual operation, including vehicle failure events, large passenger flow events, and vehicle delay events; Dispatch event handling: Generate corresponding dispatch plans for each dispatch event; Take the average total passenger flow from the vehicle failure location to the terminal at the same historical time as the passenger flow P0 affected by the vehicle failure event; The delay of the fault event to the driving time of the next vehicle is T g = P0×λ; where λ is the average boarding and alighting time of each passenger According to the historical passenger flow of the station, combined with the real-time passenger flow of the station and the passenger flow affected by vehicle failure events, predict the passenger flow P1 of the station within a future period of time; if the predicted passenger flow P1 within a future period of time is greater than k1 times the historical passenger flow P2 in the same period, it is determined as a large passenger flow event; otherwise, it is not determined as a large passenger flow event; among them, 1 < k1; the passenger flow affected by the large passenger flow event is P3 = P1 - P2; the delay of the driving time of the next vehicle caused by the large passenger flow event is T k = P3 × λ; where P2 is the historical passenger flow in the same period; According to the predicted driving time T of the vehicle from the current position to the final stop p , the predicted arrival time of the vehicle is obtained; calculate the predicted driving time T of the vehicle from the current position to the final stop p as follows: ； Among them, the nearest station to the current position of the vehicle is the m-th station, where 1 ≤ m ≤ n and there are n stations in total; T m is the driving time of the vehicle from the current position to the nearest station, i.e., the m-th station; T i,i+1 is the driving time of the vehicle from the i-th station to the (i + 1)-th station; D i is the stopping time of the vehicle at the i-th station; T g is the delay of the driving time of this vehicle caused by the vehicle failure event of the previous vehicle; T k is the delay of the driving time of this vehicle caused by the large passenger flow event; α and β are the weights of the vehicle failure event and the large passenger flow event respectively. During the peak period, the weight α of the vehicle failure event in the peak period is increased; during the off-peak period, the weight β of the large passenger flow event in the off-peak period is increased.

2. The public transportation automatic scheduling method with multi-model mutual learning according to claim 1, characterized in that, The analysis of vehicle failure events is specifically as follows: If the vehicle speed drops by more than the set range and the vehicle position does not change for a certain period of time, it is determined as a vehicle failure event; otherwise, it is not determined as a vehicle failure event.

3. A public transportation automated scheduling method with multi-model mutual learning according to claim 1, characterized in that, The analysis of large passenger flow events is specifically as follows: Classify the large passenger flow situation: The larger the passenger flow P3 affected by the large passenger flow event, the higher the level, indicating that the large passenger flow situation is more urgent.

4. A public transportation automatic scheduling method with multi-model mutual learning according to claim 1, characterized in that, The analysis of vehicle delay events is specifically as follows: Compare the predicted arrival time with the planned arrival time in the vehicle operation plan. If the predicted arrival time is later than the planned arrival time, it is determined as a vehicle delay event; otherwise, it is not determined as a vehicle delay event; Classify the vehicle delay situation: The remaining number of stations from the current position of the delayed vehicle to the terminal is n - m + 1, and the proportion of the remaining number of stations to the total number of stations is (n - m + 1) / n; the larger the proportion of the remaining number of stations to the total number of stations, the higher the level, indicating that the vehicle delay situation is more urgent.

5. A public transportation automated scheduling method for multi-model mutual learning according to claim 4, characterized in that Based on the historical arrival, departure, and stop times, predict the travel time between each station and the stop time at each station.

6. A public transportation automated scheduling method for multi-model mutual learning according to claim 1, characterized in that According to the analysis results of each dispatch event, generate the optimal dispatch plan, specifically as follows: For vehicle failure events, based on the status of the other vehicles on the same route as the faulty vehicle and the original operation plan, adjust the departure times of the other vehicles with the goal of minimizing the number of adjusted trips to generate a new operation plan; For large passenger flow events, generate a dispatch plan for adding vehicles according to the location of the large passenger flow stations and the predicted passenger flow in the next period of time to meet the actual passenger flow demand; For vehicle delay events, generate a dispatch plan for skipping stations or taking a detour according to the actual operation situation and the predicted arrival time to meet the vehicle operation plan.

7. A public transportation automated scheduling method with multi-model mutual learning according to claim 1, characterized in that Present the dispatch plans of each dispatch event in a card form on the dispatch interface and display the urgency of each dispatch event.

8. A public transportation automated scheduling method for multi-model mutual learning according to claim 1, characterized in that, It also includes dispatch decision-making: The dispatcher makes a final dispatch decision based on each dispatch event and the corresponding dispatch plan, that is, the dispatcher confirms whether to adopt the dispatch plan or adjust the dispatch plan and then adopt it.

9. A public transportation automatic scheduling system for multi-model mutual learning, characterized in that, Applicable to a public transportation automated dispatch method of multi-model mutual learning described in any one of the above claims 1-8, the system includes: a data collection module, a dispatch event analysis module, a dispatch event handling module, and a dispatch decision-making module; The data collection module obtains the following data: The real-time arrival, departure, and stop times of each vehicle on each route at each station for each trip, as well as the historical arrival, departure, and stop times; the real-time and historical passenger flows of the stations; The dispatching event analysis module includes a vehicle fault analysis model, a large passenger flow prediction model, and a vehicle delay prediction model, which are respectively used to analyze in real time vehicle fault events, large passenger flow events, and vehicle delay events occurring in actual operations; the vehicle fault analysis model, the large passenger flow prediction model, and the vehicle delay prediction model respectively input the analysis results into the dispatching event processing module for generating corresponding dispatching plans; among them, the large passenger flow prediction model combines the analysis results of the vehicle fault analysis model to analyze large passenger flow events; the vehicle delay prediction model combines the analysis results of the vehicle fault analysis model and the analysis results of the large passenger flow prediction model to analyze vehicle delay events. The dispatching event processing module includes a vehicle fault processing model, a large passenger flow processing model, and a vehicle delay processing model, which are used to generate dispatching plans for different dispatching events. The dispatching decision-making module is used for dispatchers to view each dispatching event and the corresponding dispatching plan and make a final dispatching decision.

Images