A real-time bus idle driving scheduling method and system

Abstract

The present invention relates to a real-time bus idle run scheduling method and system. The scheduling method comprises the following steps: acquiring positioning data, preprocessing the positioning data, inputting the preprocessed positioning data into a trained idle run decision model to calculate real-time idle run decisions, and issuing real-time idle run decision instructions. The idle run decision model is established based on a gradient descent decision tree. The preprocessing comprises the following steps: performing map matching on the positioning data and a route map; dividing the positioning data of each vehicle after map matching into shifts; identifying the shift type for each divided shift; matching the identified shifts with the scheduled shifts; and extracting idle run features from the matched shifts. Compared with the existing technology, the present invention has the advantages of ensuring the normal execution of the planned shifts after external disturbances and highly real-time scheduling decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of urban traffic operation management, and in particular to a method and system for real-time bus idle running scheduling. Background Art

[0002] Public transportation systems are crucial for meeting both commuter and temporary travel needs in urban transportation. Whether buses arrive at their stops on time and according to schedule is a key criterion for evaluating the quality of public transportation operations. However, traffic congestion is often caused by real-time traffic conditions, and random factors such as vehicle breakdowns and driver illness can affect the planned operation of public transportation systems. Therefore, scientific operational management is essential to maintain optimal public transportation operations.

[0003] like Figure 1 As shown in the figure, an idle run occurs when a bus, after arriving at a terminal, forgoes the next scheduled service, choosing a faster route to reach the opposite terminal. Real-time idle run scheduling is one of many management techniques that ensures the smooth execution of the overall scheduled service by abandoning some scheduled services.

[0004] The invention patent with application number CN201810770055.9 discloses a bus dispatching system and dispatching method, which increases the number of buses by collecting passenger flow data.

[0005] The invention patent application number CN201710542534.0 discloses an intelligent bus dispatching method, which assigns empty buses to designated stations when the real-time passenger flow is greater than the capacity of three buses. This method also requires passenger flow data.

[0006] Existing technologies for bus idle run scheduling primarily determine whether to dispatch additional buses by comparing passenger flow data with remaining bus capacity. However, it takes time for an empty bus to reach its designated location, during which actual passenger flow may fluctuate, but the impact of these fluctuations is not considered. Furthermore, existing technologies focus on meeting real-time passenger flow demands, making it difficult to ensure the proper execution of subsequent scheduled bus schedules when external disturbances impact bus operations.

[0007] Currently, there is no real-time bus scheduling method for real-time idle run scheduling. Existing real-time idle run scheduling is mostly made by dispatchers on the spur of the moment, and their decisions rely on personal experience, and the quality of the decisions is affected by the dispatchers' work experience. Summary of the Invention

[0008] The purpose of the present invention is to provide a method and system for real-time bus idle run scheduling in order to overcome the above-mentioned defects in the prior art.

[0009] The purpose of the present invention can be achieved by the following technical solutions:

[0010] As one aspect of the present invention, a method for real-time bus empty run scheduling is provided, the method comprising the following steps: acquiring positioning data, preprocessing the positioning data, inputting the preprocessed positioning data into a trained empty run decision model to calculate a real-time empty run decision, and issuing a real-time empty run decision instruction;

[0011] The idle driving decision model is established based on the gradient descent decision tree.

[0012] The pretreatment comprises the following steps:

[0013] Map matching of positioning data with route maps;

[0014] After map matching, the positioning data of each vehicle is divided into shifts;

[0015] Identify the shift type for each divided shift;

[0016] Match the shifts after type identification with the planned shifts;

[0017] Extract the empty trip features from the shifts after shift matching.

[0018] As a preferred technical solution, the map matching includes the following steps:

[0019] Calculate the distance from each positioning data point to each section on the route. According to the rule of minimum distance from point to segment, map each bus positioning data point one by one to the operating line segment. According to the time sequence of the positioning data points, calculate the cumulative distance from the starting station to each data point on the operating line.

[0020] As a preferred technical solution, the specific steps of shift division include:

[0021] For each vehicle’s positioning data, the longest continuous trajectory point sequence in each direction is divided into a shift according to the vehicle’s up and down directions;

[0022] In each shift, the first point whose cumulative distance is less than the set distance threshold is taken and recorded as the first data point;

[0023] If the trajectory points between the starting point and the first data point of the shift meet the set quantity conditions, then the trajectory point sequence with the first data point as the starting point and the end point of the original shift as the end point is intercepted from the original shift data as the first new shift;

[0024] If the number of trajectory points between the starting point and the first data point of the shift is less than the set number, the shift will be remap-matched on the reverse operating route. After remap matching, the first point with a cumulative distance less than the set distance threshold is taken as the second data point, and the trajectory point sequence starting from the starting point of the original shift and ending at the second data point is intercepted from the original shift data to form the second new shift.

[0025] Use the first new shift or the second new shift to replace the original shift.

[0026] As a preferred technical solution, the shift type is identified as including an empty shift and a normal shift, and the shift type identification specifically includes the following steps:

[0027] Eliminate invalid shifts caused by data anomalies;

[0028] Determine the departure and arrival times of each bus;

[0029] The frequency distribution of the cumulative distance of each shift is made, and each shift is divided into two types of identification: empty and normal according to the frequency distribution characteristic value of empty shifts in the historical empty decision data;

[0030] The invalid shifts to be eliminated are determined by determining whether the difference between the 90% quantile and the 10% quantile of the cumulative distance of each shift data point is greater than a set threshold. If so, the shift is eliminated.

[0031] The departure time is the timestamp corresponding to the last track point whose cumulative distance is less than the set threshold minus the running time, where the running time is the cumulative distance of the last track point divided by the speed;

[0032] The arrival time is the timestamp corresponding to the last track point whose cumulative distance is within a deviation of plus or minus 10% of the operating line length;

[0033] The characteristic values ​​of the frequency distribution of empty flights include the median values ​​of the first and last groups and the median value of the middle group.

[0034] As a preferred technical solution, the shift matching step specifically includes:

[0035] The divided shifts are traversed in the order of departure time, and the departure time, arrival time and shift type information of each divided shift are added to the corresponding planned shift, so that the divided shifts and the planned shifts match one by one.

[0036] As a preferred technical solution, the idle running characteristics include rest time, estimated available rest time for the next shift, distance to the preceding vehicle, and distance to the following vehicle.

[0037] As the preferred technical solution,

[0038] The available rest time is the scheduled departure time minus the actual arrival time of the previous trip;

[0039] The estimated available rest time for the next shift is the next shift's scheduled departure time - max (actual arrival time of the previous shift + minimum rest time, planned departure time) + one-way running time);

[0040] The minimum rest time is set according to the operation requirements of the bus company, and the one-way running time is set according to the operation plan of the bus company.

[0041] As a preferred technical solution, the idle driving decision model training process specifically includes the following steps:

[0042] A dataset for model training and testing is established using the historical shift dataset. Abnormal data in the dataset is removed and the dataset is divided into a training set and a test set. The training set is preprocessed, and the results obtained from the training set preprocessing are used as input and the shift type as output to train the gradient descent decision tree model. The model is tested using the test set, and when the test results meet the set standards, the empty run decision model is output.

[0043] As another aspect of the present invention, a system for implementing the above-mentioned method for real-time bus idle scheduling is provided, the system comprising a vehicle, a cloud database, a station computing device, and a data processing engine;

[0044] The vehicle is provided with a positioning device, which transmits vehicle positioning data to a cloud database in real time;

[0045] The cloud database acquires and stores vehicle positioning data and idle driving decision data in real time;

[0046] The data processing engine calls the data in the cloud database, pre-processes the data, and then trains the idle driving decision model;

[0047] When a vehicle arrives at a site, the site computing device requests real-time vehicle positioning data from the cloud database. After pre-processing the positioning data, the data is input into the empty run decision model to calculate the real-time empty run decision result, and the empty run decision instruction is sent to the corresponding vehicle.

[0048] As a preferred technical solution, the site computing device updates the idle driving decision model according to a certain period.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1) The present invention provides a method based on real-time bus idle run scheduling. When an unplanned situation occurs, idle run scheduling is performed, and some shifts are abandoned to ensure the normal execution of the overall planned shifts. It can ensure the normal execution of the planned shifts after external disturbances affect the bus operation status.

[0051] 2) The present invention provides an empty run scheduling decision model constructed based on a gradient boosting decision tree according to historical scheduling data. It does not require real-time passenger flow data, overcomes the lag of scheduling decisions based on real-time passenger flow data in the existing technology, and ensures the normal operation of planned shifts.

[0052] 3) The present invention provides a method that uses machine learning to replace manual scheduling based on historical scheduling data, which improves the stability of idle run decisions. With data as the support, it avoids the influence of the dispatcher's work experience on the idle run decision, and gives the idle run decision results in real time and efficiently, thereby improving the operational effect after the idle run decision.

[0053] 4) The present invention provides a method for extracting bus schedules from operational data. Although the present invention uses a real-time idle bus scheduling example to explain the technology, the processing method is also applicable to other scheduling methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a schematic diagram of a normal route and an empty route;

[0055] Figure 2 Schematic diagram of the real-time empty-run decision system of the present invention;

[0056] Figure 3 This is a data preprocessing flow chart of the present invention;

[0057] Figure 4 This is a flow chart of the real-time idle driving decision model training of the present invention. DETAILED DESCRIPTION

[0058] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0059] As one embodiment of the present invention, a real-time bus idle run scheduling method is provided. Based on bus satellite positioning data and idle run decision data, idle run decision features are extracted. A gradient boosting decision tree (GBDT) is used to establish an idle run decision model for automated, real-time idle run decision making. This data-driven method mitigates the influence of dispatcher experience on idle run decisions, while providing efficient, real-time results of idle run decisions, thereby improving operational effectiveness after idle run decisions are made.

[0060] like Figure 2 As shown, as one aspect of this embodiment, a schematic diagram of the structure of a real-time idle running scheduling system for buses. First, the bus needs to be equipped with a satellite positioning signal transmitting device, which transmits satellite positioning data to the cloud database through the network in real time. In addition, the cloud database also stores real-time idle running decision data. The data in the cloud database is processed by a unified data processing engine, including data preprocessing and training of the idle running decision model. In order to meet the real-time response requirements, computing equipment is equipped at the site to calculate the real-time idle running decision results. When the vehicle arrives at the site, the site computing equipment requests real-time data from the cloud, and then inputs the obtained data into the idle running decision model to calculate the real-time idle running decision results, and finally sends the corresponding idle running instructions to the vehicle.

[0061] As another aspect of this embodiment, a method for real-time bus empty run scheduling is provided, the method comprising the steps of: acquiring positioning data, preprocessing the positioning data, inputting the preprocessed positioning data into a trained empty run decision model to calculate a real-time empty run decision, and issuing a real-time empty run decision instruction;

[0062] Figure 3 The real-time idle driving decision data preprocessing flow chart provided in an embodiment of the present invention specifically includes the following steps:

[0063] S1: Matching satellite positioning data to route map data, i.e. map matching;

[0064] The distance from each location data point to each route segment is calculated. Following the minimum point-to-segment distance rule, each bus location data point is mapped one by one to a segment point on the operating route. Next, the cumulative distance from the departure station to each data point on the operating route is calculated, following the chronological order of the location data points.

[0065] S2: Divide the positioning data after map matching into multiple segments, so that each segment corresponds to an operating shift, that is, shift division.

[0066] For each bus, the longest sequence of continuous trajectory points in the same direction is divided into a data segment, i.e., a shift, based on the uplink and downlink directions in the bus positioning data. The data is sorted in ascending time order, and for each shift, the first point whose cumulative distance is less than a distance threshold (e.g., 50 meters) is taken as the first data point. If the trajectory points from the start of the shift to this point meet the requirements (e.g., greater than or equal to 10 data points at a 1s sampling interval), the last first data point with a distance less than the distance threshold (e.g., 50 meters) is taken, and the first new shift is obtained from the first data point to the end of the original shift.

[0067] If the number of trajectory points between the start and this point is too small (e.g., less than 10 data points at a 1s sampling interval), map matching is performed again on the reverse route. After rematching, the first point that is less than a distance threshold (e.g., 50 meters) is selected as the second data point. A second new shift is generated from the start of the original shift to the second data point. The first and second new shifts replace the original shift.

[0068] S3: After the shifts are divided, the type of each shift is identified, including empty shifts and normal shifts, i.e., shift type identification.

[0069] Eliminate invalid shifts caused by data anomalies. That is, for each shift, if the difference between the 90th and 10th percentiles of the cumulative distance in the data points is greater than 1 km, then eliminate the shift. Afterwards, traverse each shift and select the last trajectory point with a cumulative distance less than a threshold (e.g., 150m). The corresponding timestamp is the cumulative distance of the point divided by the speed (e.g., 20km / h). The speed is set according to the operating conditions of different shifts, which is the departure time of the shift. Select the last trajectory point with a cumulative distance within a deviation of plus or minus 10% of the operating line length. The corresponding timestamp is the arrival time of the shift. Finally, perform a frequency distribution on the cumulative distance of the shift. Based on the empty route data in the historical empty route decision data, determine the characteristics of the frequency distribution of empty shifts (e.g., the median of the first and last groups and the median of the middle group, etc.). Based on this characteristic, identify the type of each satellite positioning shift as either empty or normal.

[0070] S4: Match the shifts obtained in the previous step with the planned shifts.

[0071] The satellite positioning shifts are traversed in the order of departure time, and the departure time, arrival time and shift type of each shift are added to the planned shifts, so that the satellite positioning shifts and the planned shifts are matched one by one.

[0072] S5: Extract features related to the idle driving decision.

[0073] Calculate the available rest time (i.e., planned departure - actual arrival of the previous bus), the estimated available rest time for the next bus (i.e., planned departure of the next bus - max (actual arrival of the previous bus + minimum rest time, planned departure) + one-way running time), the distance to the preceding bus, and the distance to the following bus, for a total of four features. The minimum rest time is determined according to the operating requirements of the bus company and is generally 5 minutes. Features can be increased or decreased based on actual operating conditions. The one-way running time is set according to the bus company's schedule operation plan.

[0074] like Figure 4 As shown in FIG, the real-time idle driving decision model training flow chart of the present invention specifically includes the following steps:

[0075] Establish an empty driving decision model based on the gradient descent decision tree model:

[0076] A dataset was established for training and testing the gradient descent decision tree model. After removing outliers, the historical shift dataset was divided into a training dataset and a test dataset using an 80:20 split along the time dimension. The four previously extracted features were used as input, and the shift type as output. Once acceptable results (e.g., precision and accuracy both greater than 80%) were obtained on both the training and test data, the final no-run decision model was determined. Steps 1 through 5 above all used a data processing engine to preprocess the data in the cloud database and establish the no-run decision model.

[0077] Finally, the real-time application of the present invention is real-time idle driving decision-making:

[0078] The station's computing device requests real-time operational data from a cloud database, extracts local empty-run features, and feeds them into a local empty-run decision model. It then outputs the empty-run decision results and issues empty-run decision instructions to the corresponding buses. In practice, the empty-run decision model needs to be updated periodically (e.g., weekly), with the specific period determined by actual operations.

[0079] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A real-time bus idle running scheduling method, characterized in that: The scheduling method comprises the following steps: acquiring positioning data, preprocessing the positioning data, inputting the preprocessed positioning data into a trained empty run decision model to calculate a real-time empty run decision, and issuing a real-time empty run decision instruction; The idle driving decision model is established based on the gradient descent decision tree. The pretreatment comprises the following steps: Map matching of positioning data with route maps; After map matching, the positioning data of each vehicle is divided into shifts. The specific steps of the shift division include: For each vehicle’s positioning data, the longest continuous trajectory point sequence in each direction is divided into a shift according to the vehicle’s up and down directions; In each shift, the first point whose cumulative distance is less than the set distance threshold is taken and recorded as the first data point; If the trajectory points between the starting point and the first data point of the shift meet the set quantity conditions, then the trajectory point sequence with the first data point as the starting point and the end point of the original shift as the end point is intercepted from the original shift data as the first new shift; If the number of trajectory points between the starting point and the first data point of the shift is less than the set number, the shift will be remap-matched on the reverse operating route. After remap matching, the first point with a cumulative distance less than the set distance threshold is taken as the second data point, and the trajectory point sequence starting from the starting point of the original shift and ending at the second data point is intercepted from the original shift data to form the second new shift. Use the first new flight or the second new flight to replace the original flight; The shift type is identified for each divided shift. The shift types include empty shifts and normal shifts. The shift type identification specifically includes the following steps: Eliminate invalid shifts caused by data anomalies; Determine the departure and arrival times of each bus; The frequency distribution of the cumulative distance of each shift is calculated, and the type of each shift is identified based on the frequency distribution characteristic value of the empty shifts in the historical empty shift decision data, including empty and normal types; The invalid shifts to be eliminated are determined by determining whether the difference between the 90% quantile and the 10% quantile of the cumulative distance of each shift data point is greater than a set threshold. If so, the shift is eliminated. The departure time is the timestamp corresponding to the last track point whose cumulative distance is less than the set threshold minus the running time, where the running time is the cumulative distance of the last track point divided by the speed; The arrival time is the timestamp corresponding to the last track point whose cumulative distance is within a deviation of plus or minus 10% of the operating line length; The characteristic values ​​of the frequency distribution of empty flights include the median of the first and last groups and the median of the middle group; Match the shifts after type identification with the planned shifts; Extracting empty run features from the matched shifts, the empty run features including rest time, estimated available rest time for the next shift, distance to the preceding vehicle, and distance to the following vehicle; The rest time is the scheduled departure time minus the actual arrival time of the previous bus; The estimated available rest time for the next shift is (the next shift's scheduled departure time – max (actual arrival time of the previous shift + minimum rest time, planned departure time) + one-way running time); The minimum rest time is set according to the operation requirements of the bus company, and the one-way running time is set according to the operation plan of the bus company.

2. A method for real-time bus idle running scheduling according to claim 1, characterized in that: The map matching comprises the following steps: Calculate the distance from each positioning data point to each section on the route. According to the rule of minimum distance from point to segment, map each bus positioning data point one by one to the operating line segment. According to the time sequence of the positioning data points, calculate the cumulative distance from the starting station to each data point on the operating line.

3. A method for real-time bus idle running scheduling according to claim 1, characterized in that: The shift matching step specifically includes: The divided shifts are traversed in the order of departure time, and the departure time, arrival time and shift type information of each divided shift are added to the corresponding planned shift, so that the divided shifts and the planned shifts match one by one.

4. A method for real-time bus idle running scheduling according to claim 1, characterized in that: The training process of the idle driving decision model specifically includes the following steps: A dataset for model training and testing is established using the historical shift dataset. Abnormal data in the dataset is removed and the dataset is divided into a training set and a test set. The training set is preprocessed, and the results obtained from the training set preprocessing are used as input and the shift type as output to train the gradient descent decision tree model. The model is tested using the test set, and when the test results meet the set standards, the empty run decision model is output.

5. A system for implementing the real-time bus idle running scheduling method according to any one of claims 1 to 4, characterized in that: The system includes a vehicle, a cloud database, a site computing device, and a data processing engine; The vehicle is provided with a positioning device, which transmits vehicle positioning data to a cloud database in real time; The cloud database acquires and stores vehicle positioning data and idle driving decision data in real time; The data processing engine calls the data in the cloud database, pre-processes the data, and then trains the idle driving decision model; When a vehicle arrives at a site, the site computing device requests real-time vehicle positioning data from the cloud database. After pre-processing the positioning data, the data is input into the empty run decision model to calculate the real-time empty run decision result, and the empty run decision instruction is sent to the corresponding vehicle.

6. The system according to claim 5, characterized in that The station computing device updates the idle driving decision model at a certain period.

Images