CN110444008B

CN110444008B - Vehicle scheduling method and device

Info

Publication number: CN110444008B
Application number: CN201910779734.7A
Authority: CN
Inventors: 杨旭光
Original assignee: Euler Information Services Co ltd
Current assignee: Euler Information Services Co ltd
Priority date: 2019-08-22
Filing date: 2019-08-22
Publication date: 2021-06-04
Anticipated expiration: 2039-08-22
Also published as: CN110444008A

Abstract

The invention provides a vehicle scheduling method and a vehicle scheduling device, which comprise the following steps: inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster; inputting historical data and hot spot areas into a preset prediction model to obtain prediction data; determining the vehicle dispatching quantity of the hot spot area at the target moment according to the first quantity, the prediction data and the third quantity of the current parked vehicles in the hot spot area; and at the target moment, vehicle scheduling is carried out on the hot spot areas according to the vehicle scheduling number. The invention can input the historical data into the clustering algorithm model and the prediction model, and accurately acquire the prediction data of the vehicle at the target moment. And when the target moment is reached, a vehicle dispatching plan is rapidly made according to the predicted data and the current data collected in real time. The accuracy and the efficiency of vehicle dispatching operation are improved, and meanwhile, the timeliness of a dispatching plan is improved, so that the dispatching plan is more in line with the requirements of actual business.

Description

Vehicle scheduling method and device

Technical Field

The invention relates to the technical field of computers, in particular to a vehicle scheduling method and device.

Background

Due to the fact that the problem of traffic jam in the large cities is gradually serious, some people choose not to buy private cars, and therefore, some enterprises begin to carry out services such as car sharing and car rental in a time-sharing mode. For enterprises, it is too high to place own shared automobiles in every corner of a city, so that how to accurately park limited vehicle resources in a hot spot area with large business demand becomes a key problem.

Most of the existing vehicle dispatching methods in the industry are manual dispatching. Firstly, an enterprise can arrange and send daily data to an operation scheduling person according to the business operation condition. The operation scheduling personnel can judge the vehicle supply and demand relationship of each region of the city according to the vehicle distribution data, the daily operation condition and the actual work experience of the operation scheduling personnel. And determining how many vehicles need to be called in the hot spot area in which time period, and how many vehicles need to be taken out in the cold area in which time period. And then the scheduled dispatching plan is issued to ground staff, the ground staff monitors the actual vehicle storage amount of each area at any time, and adjusts the actual vehicle storage amount of each area according to the dispatching plan.

However, manual scheduling methods have many problems. Firstly, in the method, the formulation of the dispatching plan depends heavily on the working experience of local dispatchers, and if the working experience of the dispatchers is insufficient, the vehicle demand cannot be estimated accurately, which results in lower precision of dispatching operation. Meanwhile, the manual scheduling method is time-consuming, labor-consuming and low in efficiency. In addition, because vehicle operation is a dynamic process, the manual scheduling method cannot determine the vehicle supply and demand relationship in real time, so that the timeliness of the manual scheduling method is poor.

Disclosure of Invention

In view of this, the present invention aims to provide a vehicle scheduling method and apparatus, so as to solve the problems of low accuracy, low efficiency and poor timeliness of scheduling operation in the vehicle scheduling process due to the adoption of a manual scheduling method in the prior art.

In a first aspect, an embodiment of the present invention provides a vehicle scheduling method, where the method includes:

inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster, wherein the historical data comprises vehicle historical position data, the target vehicle cluster comprises a hot spot region with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot region at the target moment;

inputting the historical data and the hot spot area into a preset prediction model to obtain prediction data, wherein the prediction data comprises a second number of vehicles to be parked corresponding to the hot spot area at the target moment;

determining the vehicle dispatching number of the hot spot area at the target moment according to the first number, the second number and the third number of the current parked vehicles in the hot spot area;

and at the target moment, vehicle scheduling is carried out on the hot spot area according to the vehicle scheduling number.

Further, the clustering algorithm model includes a first clustering algorithm model and a second clustering algorithm model, and the obtaining of the target vehicle cluster by inputting the historical data at the historical time into the preset clustering algorithm model includes:

inputting the historical data into the first clustering algorithm model to obtain an initial vehicle cluster, wherein the initial vehicle cluster comprises an initial area with a preset size, and at the target moment, the initial area comprises a corresponding relation between the vehicle to be moved out and the position to be moved out;

and inputting the initial vehicle cluster into the second clustering algorithm model to obtain the target vehicle cluster, wherein the number of the target vehicle clusters is greater than or equal to that of the initial vehicle clusters.

Further, the step of inputting the historical data and the hot spot region into a preset prediction model to obtain prediction data includes:

determining, in the predictive model, a fourth number of historical parked vehicles corresponding to the hot spot zone at the historical time according to the historical data and the hot spot zone;

and in the prediction model, obtaining the prediction data at the target moment according to the fourth quantity.

Further, the step of determining the vehicle dispatching number of the hot spot area at the target time according to the first number, the second number and the third number of the currently parked vehicles in the hot spot area includes:

subtracting the second quantity and the third quantity from the first quantity to obtain the vehicle dispatching quantity;

if the vehicle dispatching quantity is a positive value, determining the vehicle dispatching quantity as the vehicle dispatching quantity of the hot spot area at the target moment;

and if the vehicle dispatching quantity is a negative value, determining the absolute value of the vehicle dispatching quantity as the vehicle dispatching quantity of the hot spot area at the target moment.

Further, the vehicle scheduling for the hot spot area according to the vehicle scheduling number includes:

establishing a calling vehicle array corresponding to all the hot spot areas according to the vehicle calling quantity;

establishing a called vehicle array corresponding to all the hot spot areas according to the number of the called vehicles;

determining vehicle updating data corresponding to each hot spot area according to the called vehicle array and the called vehicle array, wherein the vehicle updating data comprises a first plan of calling out vehicles with a first target quantity from the hot spot areas to other hot spot areas, or a second plan of calling out vehicles with a second target quantity from the other hot spot areas to the hot spot areas;

and at the target moment, vehicle scheduling is carried out on the hot spot area according to the vehicle updating data.

In a second aspect, an embodiment of the present invention provides a vehicle dispatching device, where the device includes:

the system comprises a clustering algorithm module, a data processing module and a data processing module, wherein the clustering algorithm module is used for inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster, the historical data comprises vehicle historical position data, the target vehicle cluster comprises a hot spot region with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot region at the target moment;

the prediction model module is used for inputting the historical data and the hot spot area into a preset prediction model to obtain prediction data, and the prediction data comprises a second number of vehicles to be parked corresponding to the hot spot area at the target moment;

the determining module is used for determining the vehicle dispatching number of the hot spot area at the target moment according to the first number, the second number and the third number of the current parked vehicles in the hot spot area;

and the scheduling module is used for scheduling the vehicles in the hot spot area at the target moment according to the vehicle scheduling number.

Further, the clustering algorithm module includes:

the first clustering algorithm sub-module is used for inputting the historical data into the first clustering algorithm model to obtain an initial vehicle cluster, the initial vehicle cluster comprises an initial area with a preset size, and at the target moment, the initial area comprises the corresponding relation between the vehicle to be moved out and the position to be moved out;

and the second clustering algorithm submodule is used for inputting the initial vehicle clusters into the second clustering algorithm model to obtain the target vehicle clusters, and the number of the target vehicle clusters is greater than or equal to that of the initial vehicle clusters.

Further, the prediction model module comprises:

a first determining submodule, configured to determine, in the prediction model, a fourth number of historical parked vehicles corresponding to the hot spot area at the historical time according to the historical data and the hot spot area;

and the prediction model submodule is used for obtaining the prediction data at the target moment in the prediction model according to the fourth quantity.

Further, the determining module includes:

the calculation submodule is used for subtracting the second quantity and the third quantity from the first quantity to obtain the vehicle dispatching quantity;

the second determining submodule is used for determining the vehicle dispatching quantity as the vehicle dispatching quantity of the hot spot area at the target moment if the vehicle dispatching quantity is a positive value;

and the third determining submodule is used for determining the absolute value of the vehicle dispatching quantity as the vehicle dispatching quantity of the hot spot area at the target moment if the vehicle dispatching quantity is a negative value.

Further, the scheduling module includes:

the first array submodule is used for establishing calling vehicle arrays corresponding to all the hot spot areas according to the vehicle calling quantity;

the second array submodule is used for establishing called vehicle arrays corresponding to all the hot spot areas according to the number of the called vehicles;

the planning sub-module is used for determining vehicle updating data corresponding to each hot spot area according to the called vehicle array and the called vehicle array, wherein the vehicle updating data comprises a first plan of calling out vehicles with a first target quantity from the hot spot areas to other hot spot areas, or a second plan of calling out vehicles with a second target quantity from the other hot spot areas to the hot spot areas;

and the scheduling submodule is used for performing vehicle scheduling on the hot spot area at the target moment according to the vehicle updating data.

The embodiment of the invention provides a vehicle scheduling method and a vehicle scheduling device, which comprise the following steps: inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster, wherein the historical data comprises vehicle historical position data, the target vehicle cluster comprises a hot spot region with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot region at the target moment; inputting the historical data and the hot spot area into a preset prediction model to obtain prediction data, wherein the prediction data comprises a second number of vehicles to be parked corresponding to the hot spot area at the target moment; determining the vehicle dispatching quantity of the hot spot area at the target moment according to the first quantity, the second quantity and the third quantity of the current parked vehicles in the hot spot area; and at the target moment, vehicle scheduling is carried out on the hot spot areas according to the vehicle scheduling number. The invention can input the historical data into the clustering algorithm model and the prediction model, and accurately acquire the prediction data of the vehicle at the target moment. And when the target moment is reached, a vehicle dispatching plan is rapidly made according to the predicted data and the current data collected in real time. The accuracy and the efficiency of vehicle dispatching operation are improved, and meanwhile, the timeliness of a dispatching plan is improved, so that the dispatching plan is more in line with the requirements of actual business.

Drawings

FIG. 1 is a flowchart illustrating steps of a method for scheduling vehicles according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a target vehicle cluster generation process based on a K-means clustering algorithm according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of a target vehicle cluster generation process based on a K-means clustering algorithm according to an embodiment of the present invention;

FIG. 4 is a flowchart illustrating steps of another vehicle dispatching method according to an embodiment of the present invention;

FIG. 5 is a schematic illustration of a first planned vehicle dispatch in accordance with an embodiment of the present invention;

FIG. 6 is a schematic illustration of a second planned vehicle dispatch in accordance with an embodiment of the present invention;

fig. 7 is a block diagram of a vehicle dispatching device according to an embodiment of the invention.

Detailed Description

Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

Referring to fig. 1, a flowchart illustrating steps of a vehicle scheduling method according to an embodiment of the present invention is shown.

Step 101, inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster.

The historical data comprises vehicle historical position data, the target vehicle cluster comprises hot spot areas with preset sizes, and the first number of the vehicles to be moved out corresponding to the hot spot areas at the target moment.

In this step, when the first number of vehicles to be moved out at the hot spot area and the target time needs to be predicted, the server may input the historical data at the corresponding historical time into a preset clustering algorithm model to obtain a required target vehicle cluster.

The core of the clustering algorithm model is a clustering algorithm.

The server can input a data set needing data mining into a clustering algorithm model, and obtains a data cluster after data mining by calculating the similarity between data points in the data set, wherein the data cluster can comprise a classification region of similar data points and data points in the classification region.

In this step, the selection of the clustering algorithm model may include, but is not limited to, one or a combination of several of the following: a K-means (K-means) Clustering algorithm, a Density-Based Noise application space (DBSCAN) Clustering algorithm, a Mean-Shift (Mean-Shift) Clustering algorithm, a Gaussian mixture model-Based expectation-maximization Clustering algorithm, and a coacervation hierarchy Clustering algorithm.

Specifically, the historical data includes vehicle historical location data. When the user uses the shared automobile service, the client of the user generates and uploads an order to the server, wherein the order can comprise related information such as a vehicle number, a vehicle getting-on position, a vehicle getting-off position, vehicle getting-on starting time, vehicle getting-off ending time, an order generating position, order generating time and the like. The historical data is extracted from historical order information, wherein the vehicle historical location data may include a boarding location, a disembarking location, and an order generation location.

In this step, the server may input the getting-on position as vehicle historical position data to the preset clustering algorithm model, or may input the order generation position as vehicle historical position data to the preset clustering algorithm model.

Further, in the manual dispatching mode, the actual position of the vehicle, that is, the getting-on position of the user, is used as the historical position data of the vehicle by the dispatcher in the vehicle dispatching business. However, since the data of the boarding location is generated by the influence of the vehicle location, for example, when the user generates an order at intersection a, a vehicle is needed, but there is no vehicle around the user, and there is a vehicle at the nearest intersection B. Therefore, the user can only select to get on the vehicle at the nearest intersection B, and if this occurs, the server considers the intersection B as the position of the vehicle to be requested each time the vehicle prediction request is made, but actually, the intersection a is the true position of the vehicle requested by the user. Therefore, in order to avoid the error caused by the vehicle position selection and accurately predict the user requirement, in the embodiment of the present invention, it is preferable that the order generation position is input to the preset clustering algorithm model as the vehicle historical position data.

Specifically, the target vehicle cluster includes a hot spot area with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot area at the target time. The target vehicle cluster is a classification region that includes similar data points and a collection of data points within the classification region. The classification area of the similar data points is a hot spot area, the data points in the classification area are vehicles to be moved out corresponding to the hot spot area, and the number of the data points in the classification area is the first number of the vehicles to be moved out corresponding to the hot spot area.

Further, the vehicles to be removed refer to vehicles selected and used by the user, and each vehicle used by the user is recorded in the order generated by the client of the user, so that the number of the vehicles to be removed may be equivalent to the number of the order of the user.

For example, to predict XX cities, each hotspot zone on the day of 2019, 10 months, 1 and a first number of vehicles to be removed for each hotspot zone. The server can select XX city, all historical data generated on the day of 2018, 10 months and 1 day and input the historical data into a preset clustering algorithm model. Among them, the history data may be the vehicle number 00001 and the corresponding order generation location (longitude 110, latitude 109), the vehicle number 00002 and the corresponding order generation location (longitude 110.5, latitude 107), and so on. Through the clustering algorithm model, the server can obtain a target vehicle cluster containing prediction information such as the hot spot region 1 of the XX city, the hot spot region 1 of the day of 10 months and 1 of 2019, the corresponding first number (30 vehicles), the hot spot region 2, the corresponding first number (15 vehicles) and the like.

Referring to fig. 2 and fig. 3, schematic diagrams of a generation process of a target vehicle cluster based on a K-means clustering algorithm according to an embodiment of the present invention are shown. Fig. 2 is historical data before inputting into a K-means clustering algorithm model, wherein fig. 2 includes a plurality of feature points 10, and each feature point 10 represents an order generation position in the historical data. Fig. 3 shows a target vehicle cluster obtained through calculation by the K-means clustering algorithm model, where the K-means clustering algorithm model divides the historical data into 4 hot spot regions, which are a hot spot region 20, a hot spot region 30, a hot spot region 40, a hot spot region 50, and feature points 10 correspondingly included in each hot spot region.

Specifically, different types of target vehicle clusters are obtained by using different types of clustering algorithm models, and because the types of selectable clustering algorithm models are too many, schematic diagrams of the target vehicle clusters obtained when other clustering algorithm models are used are not repeated in the embodiment of the invention.

In the embodiment of the invention, when the first number of the vehicles to be moved at the hot spot area and the target moment needs to be predicted, the server can input the historical data at the corresponding historical moment into the preset clustering algorithm model to obtain the required target vehicle cluster. The influence of human factors on analysis and mining of historical data under a method of manually dispatching vehicles is avoided. Compared with the method that the first number of the vehicles to be moved at the moment of determining the hot spot areas and predicting the target is determined by the dispatcher, the method that the server acquires the hot spot areas and the first number is high in speed, high in efficiency and accurate in result. Therefore, the server obtains the prediction data by using the clustering algorithm model, so that a large amount of labor cost is saved, and the accuracy of vehicle scheduling operation is improved.

In addition, in the embodiment of the invention, the order generation position is used as the historical position data of the vehicle and is input to the preset clustering algorithm model, so that the influence of the vehicle distribution position on the historical data is avoided, and the accuracy of the prediction result of the clustering model is improved.

And 102, inputting the historical data and the hot spot area into a preset prediction model to obtain prediction data.

The predicted data comprise a second number of vehicles to be parked corresponding to the hot spot area at the target moment.

Specifically, in actual business, the number of vehicles in a region is dynamic, and vehicles move out at any time and also drive in at any time, so that the server also needs to accurately predict the number of vehicles to be parked in the hot spot region on the same day.

In step 101, the server has obtained a target cluster of vehicles that contains a hot spot zone. In this step, the server may determine, for the acquired hot spot area, how many vehicles are parked in the hot spot area by the user in a past certain stage by combining the get-off position in the history data. Further, the server may input historical data of the vehicles staying in the hot spot area by the user into a preset prediction model, and obtain the number of vehicles to be parked in the hot spot area at the target time.

In this step, the selection of the prediction model may include, but is not limited to, one or a combination of the following: a first order exponential smoothing model, a second order smoothing exponential model, a third order smoothing exponential model, and the like.

For example, in a case where the server has obtained history data of the hot spot area 1, the hot spot area 2 of the XX city, and all of the get-off positions of the XX city between 6 month 1 day in 2019 and 9 month 30 in 2019. The server can determine the number of vehicles staying in the hot spot areas by the users in the hot spot areas 1 and 2 every day between 1 and 1 month and 2019 and 6 and 1 month and 2019 according to the hot spot areas and the getting-off positions in the historical data. Further, the server may input the historical data of the number of vehicles staying in the hot spot area by the user into a preset prediction model, so as to obtain the number of vehicles to be parked staying in the hot spot area 1 by the user on the day of 10 months and 1 days of 2019 at the target moment, and the number of vehicles to be parked by the user staying in the hot spot area 2.

In the embodiment of the invention, the server can input the historical data and the hot spot area into the preset prediction model, and obtain the number of vehicles to be parked in the hot spot area on the prediction day. The influence caused by the user behavior is avoided by further predicting the vehicle change quantity at the target time, so that the server can further improve the vehicle scheduling accuracy at the target time.

Step 103, determining the vehicle dispatching number of the hot spot area at the target moment according to the first number, the second number and the third number of the current parked vehicles in the hot spot area.

In the embodiment of the present invention, step 101 and step 102 are both calculations performed based on historical data. Since the historical data is usually huge in practical application, the required calculation time is also long. Therefore, the implementation of

steps

101 and 102 may be performed before the target time, for example, one day or one week in advance, and the server performs the calculation online, and stores the data obtained in

steps

101 and 102 in the database. And when the target time is reached, the server calculates the vehicle dispatching quantity on line through the acquired real-time data and the data stored in advance in the database.

In this step, the first number and the second number may be stored in the database in advance, and the third number of currently parked vehicles in the hotspot area is data collected by the server in real time. Because the server finishes the step with large calculation amount in advance, the calculation amount of the vehicle dispatching amount of the hot spot area at the target moment is small according to the first amount, the second amount and the third amount, and the calculation link is simple. The server can completely obtain the vehicle dispatching quantity in real time under the condition of not bearing large workload.

Specifically, each vehicle in the service is in a state of being connected to the server for 24 hours, and vehicle state information, such as a vehicle number, a vehicle position, a vehicle order state, vehicle remaining energy, vehicle remaining warranty period, and the like, may be periodically transmitted through a network device built in the vehicle. The server can judge whether the vehicle is idle or not, whether the vehicle needs to be recalled or not, whether the vehicle can participate in scheduling or not and the like according to the vehicle state information sent by the vehicle.

Further, the server may determine, by using the vehicle state information and in combination with the hot spot area, a third number of vehicles currently parked in the hot spot area.

For example, when the target time is reached, the server first collects vehicle state information periodically transmitted by the vehicle, and the vehicle state information may include: vehicle number 00001, no order, remaining 80% gasoline, remaining 50% maintenance period located in XX street XX parking lot; vehicle number 00002, in the XX street XX parking lot, ordered, 50% gasoline remaining, 80% maintenance remaining, etc. Then the server determines that the current parked vehicle number 00001 in the hotspot area 1 is available and the current parked vehicle number 00002 in the hotspot area 2 is unavailable according to the vehicle state information. After the server sequentially processes all the vehicle state information, data that the third number of the current parked vehicles in the hot spot area 1 is 5, the third number of the current parked vehicles in the hot spot area 2 is 3 and the like can be obtained. Further, the server retrieves the first number 10 and the second number 3 of the hot spot areas 1, the first number 5 and the second number 4 of the hot spot areas 2, which have been previously saved in the database. The server can perform the corresponding calculation of the hotspot area 1 on line: 10-3-5 is 2, namely determining that the vehicle dispatching number of the hot spot region 1 at the target moment is 2; the hot spot region 2 corresponds to the calculation: 5-4-3 ═ 2, that is, it is determined that the scheduled number of vehicles in hotspot area 2 at the target time is-2.

In the embodiment of the invention, all the first quantity and the second quantity are obtained in advance and stored in the database, and the server only needs to collect the third quantity of the current parked vehicles in the hot spot area at the target moment, namely the vehicle dispatching quantity of the hot spot area at the target moment can be calculated through the first quantity, the second quantity and the third quantity. Because the server completes partial calculation in advance under the line, the calculation process of the vehicle dispatching quantity on the line is simple, and the calculation amount is small. By the steps, the real-time performance of the vehicle dispatching process can be further improved, and the vehicle dispatching efficiency is improved.

And 104, performing vehicle scheduling on the hot spot areas according to the vehicle scheduling quantity at the target moment.

In this step, the server has obtained the vehicle dispatch number corresponding to each hotspot zone. The server can match all the hot spot areas, pair the hot spot areas needing to be called into the vehicle with the hot spot areas needing to be called out of the vehicle, and make a specific scheduling plan. The server can issue the well-formulated vehicle dispatching plan to the ground staff nearest to the hot spot area where the vehicle needs to be called out, so that the ground staff can move the vehicle to the hot spot area where the vehicle needs to be called in time according to the issued vehicle dispatching plan.

Specifically, each ground service personnel can send own position and work information regularly through a customized mobile terminal, wherein the work information can comprise an idle state and a work state, and the server can receive the position and work information sent by the ground service personnel in real time and assign a vehicle scheduling plan to the idle ground service personnel according to the information.

For example, the server has obtained the vehicle scheduling number corresponding to each hotspot area, such as: the hot spot area 1 can call 10 vehicles, the hot spot area 2 can call 5 vehicles, the hot spot area 3 needs to call 8 vehicles, the hot spot area 4 needs to call 10 vehicles and the like. Furthermore, the server matches all the hot spot areas, and the hot spot areas with close distance can be preferentially matched according to the principle of close distance during matching. If the hot spot area 1 is closest to the hot spot area 3, 8 vehicles which can be called out from the hot spot area 1 are dispatched to the hot spot area 3, and a dispatching plan 001 is generated; and if the hot spot area 2 is closest to the hot spot area 4, 5 vehicles which can be called out by the hot spot area 2 are dispatched to the hot spot area 4, and if the hot spot area 1 is second closest to the hot spot area 4, 2 vehicles which can be called out by the hot spot area 1 are dispatched to the hot spot area 4 until the calling requirement of the hot spot area 4 is met, so that a dispatching plan 002 is generated. Further, the server may collect work information periodically sent by each ground staff, and the work information may include: number 001 of ground service, XX number on XX street, no task; ground service number 002, street XX located XX, tasking, etc. The server may issue the scheduling plan 001 to the 10 ground crew members closest to the hot spot area 1, and issue the scheduling plan 002 to the 5 ground crew members closest to the hot spot area 2 and the 2 ground crew members closest to the hot spot area 1. And after receiving the dispatching plan, the dispatcher moves the corresponding vehicle to the corresponding hot spot area within the specified time, and then the vehicle dispatching plan is completed.

In the embodiment of the invention, the vehicle dispatching plan issued by the server is formulated according to the real-time data, and the ground staff performing the dispatching plan is also selected according to the real-time data. The problem of poor timeliness caused by a manual scheduling mode is avoided, and the timeliness of the whole vehicle scheduling plan is further improved.

In summary, an embodiment of the present invention provides a vehicle scheduling method, including: inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster, wherein the historical data comprises vehicle historical position data, the target vehicle cluster comprises a hot spot region with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot region at the target moment; inputting the historical data and the hot spot area into a preset prediction model to obtain prediction data, wherein the prediction data comprises a second number of vehicles to be parked corresponding to the hot spot area at the target moment; determining the vehicle dispatching quantity of the hot spot area at the target moment according to the first quantity, the second quantity and the third quantity of the current parked vehicles in the hot spot area; and at the target moment, vehicle scheduling is carried out on the hot spot areas according to the vehicle scheduling number. The invention can input the historical data into the clustering algorithm model and the prediction model, and accurately acquire the prediction data of the vehicle at the target moment. And when the target moment is reached, a vehicle dispatching plan is rapidly made according to the predicted data and the current data collected in real time. The accuracy and the efficiency of vehicle dispatching operation are improved, and meanwhile, the timeliness of a dispatching plan is improved, so that the dispatching plan is more in line with the requirements of actual business.

Referring to fig. 4, a flowchart illustrating steps of another vehicle dispatching method according to an embodiment of the invention is shown.

Step 201, inputting the historical data into the first clustering algorithm model to obtain an initial vehicle cluster.

The initial vehicle cluster comprises an initial area with a preset size, and at the target moment, the corresponding relation between the vehicle to be moved out and the position to be moved out in the initial area is obtained.

In this step, the first clustering algorithm model may be a DBSCAN clustering algorithm. The DBSCAN clustering algorithm is a density-based clustering algorithm. It defines clusters as the largest set of densely connected points, can divide areas with a sufficiently high density into clusters, and can find arbitrarily shaped clusters in noisy spatial databases. The DBSCAN clustering algorithm includes two density parameters: the scan radius (eps) and the number of points (MinPts) that need to be included within the scan radius. The DBSCAN clustering algorithm can output a set (cluster) of high-density regions having correlation in the spatial data according to the input spatial data, and remove some isolated noise (outliers) having no correlation with each region.

Specifically, the historical data is input into the DBSCAN clustering algorithm, so that abnormal points in the historical data can be removed, and the initial vehicle cluster of the high-density area in the historical data can be obtained. The server may input the order generation position of the corresponding historical time as vehicle historical position data to the DBSCAN clustering algorithm. The longitude and latitude coordinates corresponding to each vehicle are position data in the DBSCAN clustering algorithm. For example, the order generation location of vehicle number 0001 (longitude 110, latitude 109), the order generation location of vehicle number 00002 (longitude 110.5, latitude 107), may be represented in the DBSCAN clustering algorithm as data point 0001(110, 109) and data point 0002(110.5, 107).

In this step, the initial vehicle cluster includes an initial region of a preset size, and at the target time, the initial region includes a correspondence between the vehicle to be moved out and the position to be moved out. The initial area is a high-density area with correlation obtained through a DBSCAN clustering algorithm, and the corresponding relation between the vehicle to be moved out and the position to be moved out in the initial area is a data point of the high-density area with correlation obtained through the DBSCAN clustering algorithm.

Specifically, the DBSCAN clustering algorithm needs to set an initial value of a parameter, an iteration termination condition, and an iteration step length in advance.

Setting initial values of the parameters. The initial values of the parameters are two, the scan radius eps and the number of points MinPts that must be included in the scan radius eps. The larger the eps, the larger the range of the obtained initial region; the larger MinPts is, the more correspondence between the vehicle to be moved out and the position to be moved out in the initial region is obtained. Therefore, when setting the initial values of the parameters, the server can set eps to a high value with priority and set MinPts to a low value with priority. With each subsequent iteration, decreasing eps and increasing MinPts, the initial region can be gradually narrowed and outliers removed. For example, the server may set the initial values eps of the parameters to 1 and MinPts to 3.

And setting parameter iteration step size. In order to output an accurate calculation result, multiple iterations of the input data are required. If both the eps and MinPts parameters are changed at the same time in each iteration, it will result in excessive calculation. Therefore, in a preferable case, the server may preferentially decrease eps and fix MinPts to rapidly increase the density of the initial region, fix eps and increase MinPts to gradually increase the density of the initial region when the number of removed abnormal points reaches a preset first proportional threshold, and end the iteration when the number of removed abnormal points reaches a preset iteration termination condition. For example, the server may set the iteration step size for eps to be 0.1 and the iteration step size for MinPts to be 1, i.e., eps decreases by 0.1 or MinPts increases by 1 per iteration. The server may set half of the iteration termination condition to the first proportional threshold.

And setting a parameter iteration termination condition. The number and range of the initial regions obtained by the DBSCAN clustering algorithm cannot be known or set in advance. Therefore, the server can set the proportion range of the removed abnormal point to all the data points as the iteration termination condition. For example, the server may set the range of outliers removed to be 5% -9%, i.e., when the outliers removed reach 5% -9% of the total data points, the iteration terminates and the server outputs the initial cluster of vehicles.

In the embodiment of the invention, the server can carry out preliminary filtering and sorting on the historical data through a DBSCAN clustering algorithm, remove abnormal points which may influence the prediction result, and obtain the corresponding relation between the initial area and the vehicle to be moved out and the position to be moved out of the initial area. This step can increase the accuracy of the whole clustering algorithm model by removing outliers that may affect the prediction results.

Step 202, inputting the initial vehicle cluster into the second clustering algorithm model to obtain the target vehicle cluster.

Wherein the number of target vehicle clusters is greater than or equal to the number of initial vehicle clusters.

In this step, the second clustering algorithm model may be a K-means clustering algorithm. The K-means clustering algorithm is a clustering analysis algorithm for iterative solution, and takes distance as a standard for measuring similarity between data objects, i.e. the smaller the distance between data objects is, the higher the similarity is, and the more likely they are in the same cluster. The method comprises the steps of randomly selecting K objects as initial clustering centers, then calculating the distance between each object and each seed clustering center, and allocating each object to the nearest clustering center. The cluster centers and the objects assigned to them represent a cluster. The cluster center of a cluster is recalculated for each sample assigned based on the objects existing in the cluster. This process will be repeated until some termination condition is met. The termination condition may be that no object is reassigned to a different cluster, or that no cluster center is changed again, or that the sum of squared errors is locally minimal.

Specifically, the server may further process the initial vehicle cluster by using a K-means clustering algorithm, so as to divide the initial vehicle cluster having a larger initial area into a plurality of target vehicle clusters having smaller hot spot areas, and the correlation between data points is higher.

Specifically, the K-means clustering algorithm needs to set a K value, that is, the number of clusters in advance. First, the server may set the total number of hot spots in a city or region, denoted as n, according to the predicted traffic demand of the city or region. The server may then note p as the ratio of the number of data points included in one initial cluster of vehicles to the number of data points included in all initial clusters of vehicles. Finally, the server can obtain the K value correspondingly set by each initial vehicle cluster, namely: and if the obtained value has decimal places, rounding off the decimal places to obtain integer digits as a final result. For example, n is set to 35, and the initial cluster 1 includes 30 dataPoint, when all initial clusters of vehicles currently have 100 data points, p₁Is 0.3, then K₁N × p 35 × 0.3 10.5, further rounded, K₁Then 10, and thus the K value for the initial cluster 1 of vehicles is 10.

Further, the K-means clustering algorithm needs to set a termination condition. The termination condition may be that no data point is reassigned to a different hot spot region, or that no center of the hot spot region is changed, or that the sum of squared errors of each hot spot region is locally minimized.

In the embodiment of the invention, the server can divide the initial vehicle cluster containing a larger initial area into one or more target vehicle clusters containing smaller hot spot areas through a K-means clustering algorithm, so as to obtain the target vehicle clusters with more quantity and higher correlation among data points in the clusters. The step can divide the initial vehicle cluster into the target vehicle clusters with smaller hot spot areas and higher similarity, and improves the accuracy of the vehicle dispatching plan.

It should be noted that, after the step 202 is completed, the operations of converting the format of the hot spot area and obtaining the first number according to the target vehicle cluster may also be performed.

In step 202, the server has obtained a target vehicle cluster through a K-means clustering algorithm model, wherein the data format of the hot spot areas in the target vehicle cluster is also latitude and longitude. In this step, the server may convert the latitude and longitude format of the hotspot area into specific address information, and determine a specific location of the hotspot area center in the actual geographic location. Meanwhile, the server may record a ratio of the number of data points included in each hot spot area to the total number of all data points of the city or the area as a first ratio, then multiply the first ratio by the total number of available service vehicles of the city or the area, and determine the result as the first number of vehicles to be moved out corresponding to each hot spot area. Wherein, if the first number has decimal place, the decimal place is cut off and only integral digit is reserved.

For example, if the hot spot region 1 includes 20 data points and all the hot spot regions have 100 data points, the first ratio f of the hot spot region 1 is 0.2. If the total number g of the urban available service vehicles is 120, the first number d is 24, where f × g is 0.2 × 120, and the first number d is 24.

In the embodiment of the invention, the latitude and longitude format of the hotspot area is converted into the specific address information, so that the use habit in the actual service application is better met. Meanwhile, since part of the outliers are removed in step 201, the total number of data points of the hot spot region and the actual number of vehicles are not necessarily equal. Therefore, the first number of the vehicles to be moved out corresponding to the hot spot area is determined according to the ratio of the data points of the hot spot area to the total data points, the first number can be acquired more accurately, and the accuracy of vehicle scheduling operation is improved.

Step 203, determining a fourth number of historical parked vehicles corresponding to the hot spot area at the historical time in the prediction model according to the historical data and the hot spot area.

In this step, the predictive model may be a seasonal exponential smoothing model. The seasonal exponential smoothing method is essentially a cubic exponential smoothing method, and a new parameter is added to represent the smoothed trend. The cubic exponential smoothing method can predict a time series containing both trends and seasonality. Since traffic has strong periodicity and seasonality, for example, the habits of users are basically repeated from monday to sunday, and thus, the historical data of each week is very similar. Meanwhile, the traveling habits of the user are different in summer and winter as a whole. Thus, the historical data of traffic is very consistent with the use of seasonal exponential smoothing.

In this step, the vehicle to be predicted is a vehicle that may be actively driven to the hot spot area by the user within the target time. Therefore, the vehicle history position data input into the prediction model is the get-off position. The server may determine a fourth number of the historical parked vehicles in the hot spot area within the corresponding historical time according to the get-off position of the historical data and the hot spot area of the target vehicle cluster.

Specifically, the server may respectively calculate, according to the get-off position of the historical data, whether the parking distance between the get-off position and the central position of the hot spot area in the prediction time period corresponding to each day in the historical time is smaller than a preset distance threshold. And if the parking distance of one getting-off position is smaller than a preset distance threshold, adding 1 to the fourth number of the historical parked vehicles in the hot spot area in the corresponding historical moment. And the server calculates all get-off positions in sequence to obtain the accumulated fourth number.

And 204, obtaining the prediction data at the target moment in the prediction model according to the fourth quantity.

In this step, the predictive model may be a seasonal exponential smoothing model. The server may input the obtained fourth quantity into a prediction model of a seasonal exponential smoothing method to obtain prediction data of the corresponding target time. Wherein, the calculation formula of the seasonal index smoothing method is as follows:

s_i＝α×(x_i-p_i-k)+(1-α)×(s_i-1+t_i-1)

t_i＝β×(s_i-s_i-1)+(1-β)×t_i-1

p_i＝γ×(x_i-s_i)+(1+γ)×p_i-k

F_i+h＝s_i+h×t_i+p_{i-k+(h mod k)}

wherein: s_i-smoothed values of phase i;

x_i-observed value of phase i;

p_i-periodic factors of stage i;

t_i-trend of stage i;

k is the cycle length;

i + h-predicted target time, wherein h is the interval time between the ith period and the target time;

i-serial number;

α -data smoothing factor, and { α |0 < α < 1 };

β -trend smoothing factor, and { β |0 < β < 1 };

γ -the seasonal change smoothing factor, and { β |0 < β < 1 };

F_i+h-expected value of i + h period;

mod — remainder operation.

Specifically, the seasonal index smoothing method requires s to be set in advance_i、p_i、t_iAnd the values of the three coefficients alpha, beta, gamma. Wherein, due to s_i、p_i、t_iThe initial value of (a) has little influence on the whole prediction model, and the initial values can be respectively as follows: initial value s₀＝x₀(ii) a Initial value p₀0; initial value t_i＝x₁-x₀. According to a plurality of tests on the prediction model, three coefficients alpha, beta and gamma are respectively taken as follows: α is 0.35, β is 0.05, and γ is 0.9.

In particular, x_iIs the observed value of the ith period, i.e. the fourth quantity obtained in step 203, and the serial number i, i.e. the sequence of all the fourth quantities arranged in sequence according to the actual date within a period of historical time. For example, if the historical time is from 1/2019 to 1/7/2019, x is₂₀The hot spot areas on the day of 20/1/2019 are in the corresponding fourth number. k is the cycle length, and as the traffic service has strong periodicity and seasonality, the living habits of the users are basically repeated from Monday to Sunday, and therefore k is 7. F_i+hThe predicted value is the expected value of the (i + h) th period, that is, the predicted data of the target time needing to be predicted, wherein h is the interval time between the (i) th period and the target time. For example, if the target time to be predicted is the 110 th period and the historical time data has a total of 100 periods, i is 100 and h is 10, i.e., F_i+hIs F₁₀₀₊₁₀。

In the embodiment of the invention, the historical data can be selected according to the actual service condition, if the service development of the predicted area is fast, the historical data should be selected to have a shorter period so as to adapt to the fast-developing service; if the service development of the prediction region is slow, the historical data should be selected to have a longer period number, so that the input data can be increased, and the prediction accuracy is improved. Therefore, by using the prediction model to obtain the prediction data, the accuracy and timeliness of the whole vehicle dispatching business can be improved.

It should be noted that, in the embodiment of the present invention, step 201 to step 204 are all calculations performed based on historical data. Since the historical data is usually huge in practical application, the required calculation time is also long. Therefore, the implementation of steps 201 to 204 may be performed before the target time, for example, one day or one week ahead, and the server performs the calculation online, and stores the data obtained in steps 201 to 204 in the database. And when the target time is reached, the server calculates the vehicle dispatching quantity on line through the acquired real-time data and the data stored in advance in the database.

And step 205, subtracting the second quantity and the third quantity from the first quantity to obtain the vehicle dispatching quantity.

In this step, the server may collect the status information of the vehicle in real time when the target time is reached. Every vehicle is in a state of being connected with the server for 24 hours in the business, and vehicle state information, such as a vehicle number, a vehicle position, a vehicle order state, vehicle remaining energy, vehicle remaining warranty period and the like, can be periodically sent through a network device built in the vehicle. The server can judge whether the vehicle is idle or not, whether the vehicle needs to be recalled or not, whether the vehicle can participate in scheduling or not and the like according to the vehicle state information sent by the vehicle.

In the step, the server calls the first quantity and the second quantity which are stored in the database in advance before, calculates the vehicle dispatching quantity on line by combining the collected third quantity, and determines the vehicle dispatching quantity of each hot spot area at the target moment. The server can subtract the number of the vehicles to be moved out corresponding to the hot spot area from the number of the vehicles to be parked corresponding to the hot spot area and the number of the vehicles currently parked in the hot spot area to obtain the vehicle dispatching number of each hot spot area. That is, the server subtracts the second number and the third number from the first number to obtain the vehicle dispatching number. Wherein, the calculation formula of the vehicle dispatching quantity is as follows:

S_I＝R_I-V_I-D_I

wherein: s_I-a vehicle dispatch number;

R_I-a first number;

V_I-a second number;

D_I-a third number;

i-the corresponding serial number of the hot spot area.

For example, a first number R of hot spot regions 1₁Is 20, the second number V ₁10, a third number D₁If the number is 2, the calculation formula is substituted to obtain the vehicle dispatching number S of the hot spot area 1₁The vehicle dispatching quantity S is recorded for 20-10-2-8₁Is 8; first number R of hot spot regions 2₂Is 5, a second number V₂Is 6, a third quantity D₂If the number is 3, the calculation formula is substituted to obtain the vehicle dispatching number S of the hot spot area 2₂The vehicle dispatching quantity S is recorded for 5-6-2 ═ 3₂Is-3.

In the embodiment of the invention, the server can calculate the clustering algorithm and the prediction model through the line to obtain the first quantity and the second quantity, and obtain the third quantity by collecting real-time vehicle data at the target moment. The server may then subtract the second number and the third number from the first number to obtain a vehicle dispatch number. Therefore, the server can implement the collected data through the accurate prediction of the clustering algorithm and the prediction model, and the accuracy and the real-time performance of the vehicle scheduling data are improved.

Step 206, if the vehicle dispatching quantity is a positive value, determining the vehicle dispatching quantity as the vehicle dispatching quantity of the hot spot area at the target moment.

In step 205, the server obtains the vehicle dispatching number of the hot spot area, wherein if the vehicle dispatching number is a positive value, it indicates that the total number of the vehicles to be parked and the current parked vehicles in the hot spot area at the target time cannot meet the user requirement, and a part of the vehicles need to be called. Therefore, the server may determine the vehicle scheduled number as the vehicle call-in number of the hot spot area at the target time in a case where the vehicle scheduled number is a positive value. For example, if the vehicle scheduling number of the hot spot area 1 is 8, the server may record the vehicle call-in number of the hot spot area 1 at the target time as 8.

Step 207, if the vehicle dispatching quantity is a negative value, determining an absolute value of the vehicle dispatching quantity as the vehicle dispatching quantity of the hot spot area at the target time.

In step 205, the server obtains the vehicle dispatching number of the hot spot area, wherein if the vehicle dispatching number is a negative value, it indicates that the total number of the vehicles to be parked and the current parked vehicles in the hot spot area at the target time has met the user requirement and some vehicles will be in an idle state. Therefore, the server may determine the absolute value of the vehicle dispatch number as the vehicle dispatch number of the hot spot area at the target time in the case that the vehicle dispatch number is a negative value. For example, if the number of vehicles scheduled for hot spot area 2 is-3, the server may record the number of vehicle callouts for hot spot area 2 at the target time as 3.

Further, if the vehicle dispatching quantity is zero, it indicates that the total quantity of the vehicles to be parked and the current parked vehicles in the hot spot area at the target moment just meets the user requirement, and no vehicle is free and does not need to be called. Therefore, the server may determine the hot spot region as not participating in the vehicle dispatching plan at the target time in the case where the vehicle dispatching number is zero. For example, if the vehicle dispatch number of hot spot zone 3 is 0, the server may determine that hot spot zone 3 does not participate in the vehicle dispatch plan at the target time.

And 208, establishing a calling vehicle array corresponding to all the hot spot areas according to the vehicle calling quantity.

In this step, the server may add the hot spot regions to the corresponding call-in vehicle array according to the vehicle call-in number corresponding to the hot spot region, and sequentially add all the hot spot regions having the vehicle call-in number to the call-in vehicle array. Meanwhile, after all the hot spot areas with the vehicle calling number are added to the calling vehicle array by the server, the hot spot areas in the calling vehicle array can be arranged from large to small according to the number of the respective calling vehicles.

For example, if the number of vehicle calls in the hot spot area 1 at the target time is recorded as 8, the number of vehicle calls in the hot spot area 4 at the target time is recorded as 2, and the number of vehicle calls in the hot spot area 5 at the target time is recorded as 5, the server may add the hot spot area 1, the hot spot area 4, and the hot spot area 5 to the call-in vehicle array a [ n ]. Wherein a 0 is hot spot region 1, a 1 is hot spot region 5, and a 2 is hot spot region 4.

And 209, establishing a called vehicle array corresponding to all the hot spot areas according to the vehicle calling quantity.

In this step, the server may add the hot spot regions to the corresponding called vehicle arrays according to the number of vehicle calls corresponding to the hot spot regions, and sequentially add all the hot spot regions having the number of vehicle calls to the called vehicle arrays. Meanwhile, after all the hot spot areas with the vehicle calling number are added to the called vehicle array by the server, the hot spot areas in the called vehicle array can be arranged from large to small according to the number of the respective called vehicles.

For example, if the number of vehicle calls in the hot spot area 2 at the target time is 3, the number of vehicle calls in the hot spot area 6 at the target time is 4, and the number of vehicle calls in the hot spot area 7 at the target time is 1, the server may add the hot spot area 1, the hot spot area 6, and the hot spot area 7 to the set of call-in vehicles s [ n ]. Wherein s 0 is a hot spot region 6, s 1 is a hot spot region 2, and s 2 is a hot spot region 7.

And step 210, determining vehicle updating data corresponding to each hot spot area according to the called vehicle array and the called vehicle array.

Wherein the vehicle update data comprises a first schedule of calling a first target number of vehicles from the hot spot area to another hot spot area, or a second schedule of calling a second target number of vehicles from the other hot spot area to the hot spot area.

In the step, the server determines vehicle updating data corresponding to each hot spot area by adopting a greedy algorithm. The greedy algorithm is an algorithm that adopts local optimization, that is, only the conditions of the current target object are considered and the next target object is processed after the conditions of the current target object are fully met.

Specifically, when the server formulates a second plan for calling out vehicles with a second target number from other hot spot areas to the hot spot areas, the server can preferentially match the hot spot areas with the first sequence in the current calling-in vehicle array, namely the hot spot areas with the maximum number of the calling-in vehicles in the calling-in vehicle array, and when the scheduling plan of the hot spot areas is formulated, the hot spot areas are moved out of the calling-in vehicle array. Meanwhile, if the number of the called vehicles in one hot spot area in the called vehicle array is matched, the hot spot area is moved out of the called vehicle array.

Further, when the server matches the called vehicle array, the server may select a hot spot area closest to the hot spot area corresponding to the called vehicle array from the called vehicle array according to a distance priority principle.

Preferably, when the server makes a first plan for calling the vehicles with the first target number from the hot spot area to other hot spot areas, the server may preferentially match the hot spot area with the first sequence in the current called vehicle array, that is, the hot spot area with the largest number of the called vehicles in the called vehicle array, and when the making of the scheduling plan of the hot spot area is completed, the hot spot area is moved out of the called vehicle array. Meanwhile, if the number of the calling vehicles in one hot spot area in the calling vehicle array is matched, the hot spot area is moved out of the calling vehicle array.

Further, when the server matches the called vehicle array, the server may select a hot spot area closest to the hot spot area corresponding to the called vehicle array from the called vehicle array according to a distance priority rule.

For example, the serial number of the hotspot area 1 in the calling vehicle array is 0, and the number of the calling vehicles is 15; the sequence number of the hot spot area 2 in the called vehicle array is 5, the number of the called vehicles is 10, and the hot spot area is closest to the hot spot area 1; the sequence number of the hot spot area 3 in the called vehicle array is 8, the number of the called vehicles is 8, and the hot spot area is the second closest to the hot spot area 1. The server can preferentially match the hot spot area 1 with the serial number of 0 in the calling vehicle array, and then select the hot spot area 2 closest to the hot spot area 1 from the calling vehicle array. At this time, the number of the called vehicles 10 in the hot spot area 2 does not satisfy the number of the called vehicles 15 in the hot spot area 1, and the server continues to select the hot spot area 3 which is the second closest to the hot spot area 1 from the called vehicle array. At this time, the number of the outgoing vehicles 10 of the hot spot area 2 and the number of the outgoing vehicles 8 of the hot spot area 3 together satisfy the number of the incoming vehicles 15 of the hot spot area 1, so that the server may add the hot spot area 1, the hot spot area 2, the hot spot area 3, the second target number 10 of the hot spot area 2, and the second target number 5 of the hot spot area 3 to the second plan, remove the hot spot area 1 from the incoming vehicle array, remove the hot spot area 2 from the outgoing vehicle array, and adjust the number of the outgoing vehicles of the hot spot area 3 to 3.

In this step, the server may collect data of the ground crew through a mobile terminal worn by the ground crew. The ground service personnel can regularly send own position and work information through the customized mobile terminal, wherein the work information can comprise an idle state and a work state, and the server can receive the position and work information sent by the ground service personnel in real time and assign a first plan or a second plan to the idle ground service personnel according to the information.

Further, after the server determines vehicle update data corresponding to the hot spot area, the server determines an idle ground crew member closest to the hot spot area of the called vehicle according to a distance priority principle, and brings the ground crew member into a first plan corresponding to the hot spot area of the vehicle to be called or a second plan corresponding to the hot spot area of the vehicle to be called.

Specifically, the first plan or the second plan includes a vehicle number to be scheduled, a ground crew number, a current vehicle position, a vehicle scheduling destination, and a scheduling time.

For example, referring to FIG. 5, a first planned vehicle dispatch diagram is shown in accordance with an embodiment of the present invention. Wherein fig. 5 includes a ground crew 60 number 011, a ground crew 90 number 015, a vehicle 81 number 0011, a vehicle 82 number 0010, a hot spot area 80, a hot spot area 100, and a hot spot area 110. The first plan may include: the ground crew 60 of number 011 would move the vehicle 81 of number 0011 from the hot spot area 80 to the hot spot area 110 before 8/1/09: 30 in 2019; the commuter 90 at number 015 should move the vehicle 82 at number 0010 from the hot spot area 80 to the hot spot area 100 before 8/1/10: 00 in 2019.

For example, referring to FIG. 6, a schematic diagram of a second planned vehicle dispatch is shown in accordance with an embodiment of the present invention. Fig. 6 includes a number 003 of the ground crew 140, a number 005 of the ground crew 150, a number 0001 of the vehicle 131, a number 0002 of the vehicle 161, a hot spot area 120, a hot spot area 130, and a hot spot area 160. The second plan may include: the commuter 140, number 003, should move the vehicle 131, number 0001, from the hot spot area 130 to the hot spot area 120 before 09:30, 8/1/2019; the commuter 150, number 005, should move the vehicle 161, number 0002, from the hot spot area 160 to the hot spot area 120 before 8/1/30 in 2019.

In the embodiment of the invention, the server preferentially matches the hot spot areas with the largest number of the called vehicles or preferentially matches the hot spot areas with the largest number of the called vehicles through a greedy algorithm, preferentially matches the hot spot areas with the closest distance to the hot spot areas, preferentially selects the ground crew members with the closest distance to the hot spot areas, and meets the area with the highest user demand to the maximum extent. Through the local optimal scheduling strategy, the calculated amount of the server for making the vehicle scheduling plan on line is reduced, and the service requirements of most users are met.

And step 211, performing vehicle scheduling on the hot spot area according to the vehicle updating data at the target moment.

In this step, the server may issue the first plan corresponding to the hot spot area where the vehicle needs to be called out or the second plan corresponding to the hot spot area where the vehicle needs to be called in, which is well formulated, to the corresponding ground service staff. And when the ground staff receives the first plan or the second plan, vehicle scheduling is carried out according to the plan content.

Further, after the server issues the first plan or the second plan, the vehicle scheduling plan can be updated in real time according to the information sent by the vehicle and the ground service personnel regularly, so that the accuracy and timeliness of the first plan or the second plan are guaranteed.

In summary, the vehicle scheduling method provided in the embodiment of the present invention includes: inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster, wherein the historical data comprises vehicle historical position data, the target vehicle cluster comprises a hot spot region with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot region at the target moment; inputting the historical data and the hot spot area into a preset prediction model to obtain prediction data, wherein the prediction data comprises a second number of vehicles to be parked corresponding to the hot spot area at the target moment; determining the vehicle dispatching quantity of the hot spot area at the target moment according to the first quantity, the second quantity and the third quantity of the current parked vehicles in the hot spot area; and at the target moment, vehicle scheduling is carried out on the hot spot areas according to the vehicle scheduling number.

According to the invention, the server can accurately acquire the vehicle dispatching requirement from the historical data through the first clustering algorithm model, the second clustering algorithm model and the prediction model, and data calculation with large calculation amount is completed in advance before the target moment. In addition, the server can acquire real-time data of the vehicle and the ground staff through the information uploaded by the vehicle and the ground staff regularly, and a scheduling plan is formulated quickly. The accuracy and the efficiency of the dispatching operation in the vehicle dispatching process are improved, the unstable factor of manually making a dispatching plan is avoided, and meanwhile, the timeliness of the dispatching plan is improved.

On the basis of the above embodiment, the embodiment of the invention also provides a vehicle dispatching device.

Referring to fig. 7, a block diagram of a vehicle dispatching device according to an embodiment of the present invention is shown, which may specifically include the following modules:

the clustering algorithm module 301 is configured to input historical data at a historical time into a preset clustering algorithm model to obtain a target vehicle cluster, where the historical data includes vehicle historical position data, the target vehicle cluster includes a hot spot region with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot region at the target time;

the prediction model module 302 is configured to input the historical data and the hot spot area into a preset prediction model to obtain prediction data, where the prediction data includes a second number of vehicles to be parked corresponding to the hot spot area at the target time;

a determining module 303, configured to determine, according to the first number, the second number, and a third number of currently parked vehicles in the hot spot area, a vehicle dispatching number of the hot spot area at the target time;

and the scheduling module 304 is configured to perform vehicle scheduling on the hot spot areas at the target time according to the vehicle scheduling number.

Optionally, the clustering algorithm module 301 includes:

a first clustering algorithm sub-module 3011, configured to input the historical data into the first clustering algorithm model to obtain an initial vehicle cluster, where the initial vehicle cluster includes an initial region with a preset size, and at the target time, the initial region includes a corresponding relationship between the vehicle to be moved out and a position to be moved out;

and the second clustering algorithm submodule 3012 is configured to input the initial vehicle cluster into the second clustering algorithm model to obtain the target vehicle clusters, where the number of the target vehicle clusters is greater than or equal to the number of the initial vehicle clusters.

Further, the prediction model module 302 includes:

a first determining submodule 3021, configured to determine, in the prediction model, a fourth number of historical parked vehicles corresponding to the hotspot region at the historical time according to the historical data and the hotspot region;

a prediction model submodule 3022, configured to obtain, in the prediction model, the prediction data at the target time according to the fourth quantity.

Further, the determining module 303 includes:

a calculation submodule 3031, configured to subtract the second quantity and the third quantity from the first quantity to obtain the vehicle scheduling quantity;

a second determining submodule 3032, configured to determine the vehicle scheduling number as the vehicle calling number of the hot spot area at the target time if the vehicle scheduling number is a positive value;

a third determining submodule 3033, configured to determine, if the vehicle scheduling number is a negative value, an absolute value of the vehicle scheduling number as the vehicle dispatch number of the hot spot area at the target time.

Further, the scheduling module 304 includes:

a first array submodule 3041, configured to establish, according to the vehicle call-in number, call-in vehicle arrays corresponding to all the hot spot areas;

a second array submodule 3042, configured to establish, according to the vehicle callout number, callout vehicle arrays corresponding to all the hot spot areas;

a plan submodule 3043, configured to determine, according to the called-in vehicle array and the called-out vehicle array, vehicle update data corresponding to each hot spot area, where the vehicle update data includes a first plan for calling out a first target number of vehicles from the hot spot area to another hot spot area, or a second plan for calling out a second target number of vehicles from the other hot spot area to the hot spot area;

the scheduling submodule 3044 is configured to perform vehicle scheduling on the hot spot area according to the vehicle update data at the target time.

In summary, an embodiment of the present invention provides a vehicle scheduling apparatus, including: inputting historical data at a historical moment into a preset clustering algorithm model to obtain a target vehicle cluster, wherein the historical data comprises vehicle historical position data, the target vehicle cluster comprises a hot spot region with a preset size, and a first number of vehicles to be moved out corresponding to the hot spot region at the target moment; inputting the historical data and the hot spot area into a preset prediction model to obtain prediction data, wherein the prediction data comprises a second number of vehicles to be parked corresponding to the hot spot area at the target moment; determining the vehicle dispatching quantity of the hot spot area at the target moment according to the first quantity, the second quantity and the third quantity of the current parked vehicles in the hot spot area; and at the target moment, vehicle scheduling is carried out on the hot spot areas according to the vehicle scheduling number. The invention can input the historical data into the clustering algorithm model and the prediction model, and accurately acquire the prediction data of the vehicle at the target moment. And when the target moment is reached, a vehicle dispatching plan is rapidly made according to the predicted data and the current data collected in real time. The accuracy and the efficiency of vehicle dispatching operation are improved, and meanwhile, the timeliness of a dispatching plan is improved, so that the dispatching plan is more in line with the requirements of actual business.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the method, the apparatus and the unit described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A vehicle scheduling method, characterized in that the method comprises:

at the target moment, vehicle scheduling is carried out on the hot spot areas according to the vehicle scheduling number;

the method for obtaining the target vehicle cluster comprises the following steps of:

inputting the historical data into the first clustering algorithm model to obtain an initial vehicle cluster, wherein the initial vehicle cluster comprises an initial area with a preset size, and at the target moment, the initial area comprises the corresponding relation between the vehicle to be moved out and the position to be moved out;

inputting the initial vehicle cluster into the second clustering algorithm model to obtain the target vehicle cluster, wherein the number of the target vehicle clusters is greater than or equal to that of the initial vehicle cluster;

wherein inputting the historical data into the first clustering algorithm model to obtain an initial cluster of vehicles comprises:

inputting the historical data into a DBSCAN clustering algorithm, and filtering abnormal points influencing a prediction result in the historical data to obtain the initial vehicle cluster;

wherein inputting the initial vehicle cluster into the second clustering algorithm model to obtain the target vehicle cluster comprises: inputting the initial vehicle cluster into a K-means clustering algorithm, and randomly selecting K objects in the initial vehicle cluster as clustering centers for clustering to obtain the target vehicle cluster;

at the target moment, according to the vehicle scheduling number, performing vehicle scheduling on the hot spot area comprises: inputting the vehicle dispatching quantity into a greedy algorithm to determine vehicle updating data, and carrying out vehicle dispatching on the hot spot area according to the vehicle updating data;

the prediction model comprises at least one of a primary exponential smoothing model, a secondary smoothing exponential smoothing model and a tertiary smoothing exponential smoothing model; the vehicle is a shared automobile.

2. The method of claim 1, wherein the step of inputting the historical data and the hot spot region into a preset prediction model to obtain prediction data comprises:

3. The method of claim 1, wherein the step of determining the scheduled number of vehicles for the hot spot zone at the target time based on the first number, the second number, and a third number of currently parked vehicles in the hot spot zone comprises:

4. The method of claim 3, wherein the vehicle scheduling the hotspot area according to the vehicle scheduling number comprises:

5. A vehicle dispatching device, comprising:

the scheduling module is used for scheduling vehicles in the hot spot area at the target moment according to the vehicle scheduling number;

wherein, the clustering algorithm module comprises:

a second clustering algorithm submodule, configured to input the initial vehicle cluster into the second clustering algorithm model to obtain the target vehicle clusters, where the number of the target vehicle clusters is greater than or equal to the number of the initial vehicle clusters;

the first clustering algorithm sub-module is specifically configured to input the historical data into a DBSCAN clustering algorithm, and filter abnormal points affecting a prediction result in the historical data to obtain the initial vehicle cluster;

the second clustering algorithm sub-module is specifically configured to input the initial vehicle cluster into a K-means clustering algorithm, and randomly select K objects from the initial vehicle cluster as clustering centers for clustering to obtain the target vehicle cluster;

the scheduling module is specifically configured to input the vehicle scheduling number into a greedy algorithm to determine vehicle update data, and perform vehicle scheduling on the hot spot area according to the vehicle update data;

6. The apparatus of claim 5, wherein the predictive model module comprises:

7. The apparatus of claim 5, wherein the determining module comprises:

8. The apparatus of claim 7, wherein the scheduling module comprises: