WO2020207430A1 - 订单派发方法、装置、电子设备及计算机可读存储介质 - Google Patents

订单派发方法、装置、电子设备及计算机可读存储介质 Download PDF

Info

Publication number
WO2020207430A1
WO2020207430A1 PCT/CN2020/083947 CN2020083947W WO2020207430A1 WO 2020207430 A1 WO2020207430 A1 WO 2020207430A1 CN 2020083947 W CN2020083947 W CN 2020083947W WO 2020207430 A1 WO2020207430 A1 WO 2020207430A1
Authority
WO
WIPO (PCT)
Prior art keywords
historical
order
service provider
dispatch
service
Prior art date
Application number
PCT/CN2020/083947
Other languages
English (en)
French (fr)
Inventor
秦志伟
焦岩
黎敏讷
王晨曦
汪军
吴国斌
叶杰平
宫志晨
杨耀东
Original Assignee
北京嘀嘀无限科技发展有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京嘀嘀无限科技发展有限公司 filed Critical 北京嘀嘀无限科技发展有限公司
Publication of WO2020207430A1 publication Critical patent/WO2020207430A1/zh
Priority to US17/450,458 priority Critical patent/US20220027822A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0633—Lists, e.g. purchase orders, compilation or processing
    • G06Q30/0635—Processing of requisition or of purchase orders
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06315—Needs-based resource requirements planning or analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0639—Item locations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/40—Business processes related to the transportation industry

Definitions

  • This application relates to the field of data processing, and in particular to an order dispatch method, device, electronic equipment, and computer-readable storage medium.
  • Greedy algorithms generally dispatch orders based on the distance between the driver and the passenger, giving priority Orders are dispatched to the nearest driver, or ordered according to the value of the order, and the highest value order is given priority to the driver within the dispatch range.
  • the greedy algorithm only the optimal order in the current order queue (such as the closest order or the order with the highest value) is concerned, and other orders in the order queue cannot be taken into account, which leads to some service providers in the distribution process.
  • the response rate is relatively low.
  • the purpose of the present application is to provide an order dispatching method, device, electronic equipment, and computer-readable storage medium to solve the problem of low response rate of service providers to orders in the prior art.
  • an order dispatch method which includes:
  • a dispatch order is determined for the service provider, and the dispatch order maximizes the total amount of actual resources of the service provider and estimated resources of subsequent orders.
  • the attribute information includes location information and time information of the service provider, and the order information includes at least service start location information, service end location information, and current order estimated resources.
  • the determining the dispatch order for the service provider according to each of the association degrees includes:
  • the order with the greatest degree of relevance is taken as the dispatch order of the service provider.
  • it also includes:
  • the first historical attribute information of the historical service provider corresponding to the first historical order, the first historical association degree corresponding to the first historical order, the historical order characteristics of the first historical order, and the historical service provision The first historical average action of the party is input to the first action value network to obtain the first estimated resource of the first historical order, wherein the first historical average action is the historical service provider’s The supply-demand relationship between the historical service provider and the historical order at the service end position of the historical order;
  • it also includes:
  • the second historical order is an associated order of the historical service provider at the end position of the first historical order service
  • it also includes:
  • the parameters of the second action value network are updated based on the weighted processing result.
  • the supply and demand relationship is the ratio of the number of historical service providers to the number of historical orders.
  • the first historical order is determined based on a selection result obtained by inputting the degree of relevance of each first historical associated order associated with the historical service provider into a Boltzmann selector.
  • the associated orders are all orders within the dispatch range of the location of the service provider.
  • the actual resource is obtained by weighting the actual resource due to the service provider, the demand potential and penalty of the service provider at the service end position of the dispatch order.
  • an order dispatching device which includes:
  • the obtaining module is used to obtain the attribute information of the service provider and the order information of all associated orders received by the service provider;
  • a processing module configured to input the attribute information and the order information of all the associated orders into the order distribution strategy network to obtain the degree of association between the service provider and each of the associated orders;
  • the dispatch module is configured to determine a dispatch order for the service provider according to all the obtained association degrees, and the dispatch order maximizes the total amount of actual resources of the service provider and estimated resources of subsequent orders.
  • the attribute information includes location information and time information of the service provider, and the order information includes at least service start location information, service end location information, and current order estimated resources.
  • the dispatch module is specifically configured to:
  • the order with the greatest degree of relevance is taken as the dispatch order of the service provider.
  • an adjustment module configured to:
  • the first historical attribute information of the historical service provider corresponding to the first historical order, the first historical association degree corresponding to the first historical order, the historical order characteristics of the first historical order, and the historical service provision The first historical average action of the party is input to the first action value network to obtain the first estimated resource of the first historical order, wherein the first historical average action is the historical service provider’s The supply and demand relationship between the historical service provider and the historical order at the end of the service of the historical order;
  • the adjustment module is further used for:
  • the second historical order is an associated order of the historical service provider at the end position of the first historical order service
  • the adjustment module is further used for:
  • the parameters of the second action value network are updated based on the weighted processing result.
  • the supply and demand relationship is the ratio of the number of historical service providers to the number of historical orders.
  • the first historical order is determined based on a selection result obtained by inputting the degree of relevance of each first historical associated order associated with the historical service provider into a Boltzmann selector.
  • the associated orders are all orders within the dispatch range of the location of the service provider.
  • the actual resource is obtained by weighting the actual resource due of the service provider, the demand potential and penalty of the service provider at the service end position of the dispatch order.
  • an embodiment of the present application provides an electronic device, including: a processor, a storage medium, and a bus.
  • the storage medium stores machine-readable instructions executable by the processor.
  • the processor and the storage medium communicate through a bus, and the processor executes the machine-readable instructions to perform the steps of the method described in the first aspect when executed.
  • an embodiment of the present application provides a computer-readable storage medium with a computer program stored on the computer-readable storage medium, and the computer program executes the steps of the method described in the first aspect when the computer program is run by a processor.
  • the order dispatching method, device, electronic equipment, and computer-readable storage medium input the acquired attribute information of the service provider and the order information of all associated orders received by the service provider into the order dispatching strategy network , Obtain the degree of relevance between the service provider and each associated order, and then determine the dispatch order for the service provider based on the obtained degree of relevance.
  • the order dispatch strategy network distributes the order for the service provider to maximize the current and future resources of the service provider. , Which can improve the response rate of service providers to orders and reduce the delay time of order response.
  • Figure 1 shows a first flow chart of an order dispatch method provided by an embodiment of the present application
  • FIG. 2 shows a second flow chart of an order dispatching method provided by an embodiment of the present application
  • FIG. 3 shows a schematic diagram of an order dispatching environment provided by an embodiment of the present application
  • FIG. 4 shows a third process flow chart of an order dispatch method provided by an embodiment of the present application
  • FIG. 5 shows a fourth flow chart of an order dispatching method provided by an embodiment of the present application
  • FIG. 6 shows a schematic structural diagram of an order dispatch device provided by an embodiment of the present application.
  • FIG. 7 shows a schematic structural diagram of an electronic device provided by an embodiment of the present application.
  • the car platform At present, with the growth of travel demand, the car platform generates a large number of orders at all times.
  • the number of drivers is usually less than the number of orders, that is, the supply of service providers is less than the order. Demand. This imbalance between supply and demand prevents some passengers’ orders from being allocated to the corresponding drivers, leading to longer waiting times for passengers and reducing passenger experience. At the same time, some drivers may not be dispatched to bring greater potential. Orders for resources reduce the response rate of drivers to orders, which in turn leads to a decline in the experience of both passengers and drivers.
  • the vehicle platform In order to improve the response rate of the service provider to the order, the vehicle platform considers the best match between the driver and the order when dispatching the order.
  • multiple drivers When dispatching the order, multiple drivers usually have order-receiving interactions, such as location The range of dispatching orders between the approaching drivers overlaps. This interaction potentially affects the driver’s response rate to orders and the supply-demand relationship in local areas.
  • the interaction between drivers it can be realized that the number of drivers is certain. However, in the case of a large number of drivers, the response rate of the drivers to the order decreases and the response time is prolonged. Therefore, an order dispatch method is urgently needed to solve the low response rate of the service provider to the order. problem.
  • the embodiments of the present application may serve a car-using platform, which is used to provide users with corresponding services according to the received travel service request of the client.
  • the car-using platform may include multiple taxi-hailing systems, such as taxi-hailing systems, express taxi-hailing systems, dedicated taxi-hailing systems, and downwind taxi-hailing systems.
  • the attribute information of the service provider and the order information of all the associated orders received by the service provider are entered into the order dispatch strategy network to obtain the degree of association between the service provider and each associated order, and then based on the obtained association Degree for the service provider to determine the dispatch of orders.
  • the order distribution method of this application is not easily affected by the excessive number of drivers, and is suitable for scenarios where the number of drivers and orders changes over time, and has better robustness and real-time performance. The technical solution of the present application will be described in detail below.
  • the embodiment of the present application provides an order dispatching method, which is applied to a car-using platform server, as shown in FIG. 1, which specifically includes the following:
  • S101 Obtain attribute information of a service provider and order information of all associated orders received by the service provider;
  • the service provider is generally the driver.
  • the service provider in this application refers to the driver in the car platform that can provide real-time services to the passengers.
  • the service provider can receive real-time orders broadcast by the car platform; the attributes of the service provider
  • the information generally includes the location information and time information of the service provider.
  • the location information is generally positioning information obtained through the Global Satellite Positioning System (GPS), and the time information is generally the time of the service provider’s location, for example, the driver December 20, 2018 It is located at Q Street at 16:00 on the day, and the location information is the location information of Q Street (40 degrees 54 minutes 20 seconds north latitude, 116 degrees 23 minutes 30 seconds east longitude), and the time information is 16:00.
  • GPS Global Satellite Positioning System
  • the location information may include positioning information obtained through the Beidou system, the global satellite navigation system (GLONASS), or the Galileo satellite navigation system.
  • the location information may also include location information obtained through WiFi positioning technology, geomagnetic positioning technology, base station positioning technology, and the like.
  • the attribute information of the service provider may also include, but is not limited to, any combination of one or more of the service provider's car model, available seats, and driver information.
  • Associated orders are all orders within the dispatch range at the location of the service provider.
  • the dispatch range is generally a preset range, which can be set according to actual conditions.
  • the dispatch range can be centered on the driver’s location.
  • Service start location information represents the service start location in the current order
  • service end location information represents the current The service end position in the order
  • the estimated resource of the current order represents the estimated value of the current order.
  • the associated order may also be all orders that meet the requirements within the dispatch range where the service provider is located. For example, passengers can attach requirements to the order (such as a specific model, number of seats, etc.) when placing an order.
  • the order that meets the requirements is that the attribute information of the service provider can meet the additional requirements of the passengers.
  • the vehicle platform server uses the vehicle platform server to obtain all orders (or all orders that meet the requirements) in the dispatch range of the service provider’s location information.
  • S102 Input the attribute information and all the order information into the order distribution strategy network to obtain the degree of association between the service provider and each of the associated orders.
  • the order distribution strategy network may generally be a Perceptron (Perceptron) neural network, for example, a Multilayer Perceptron (MLP) neural network.
  • the order dispatch strategy network can estimate the uncertainty of the service provider’s subsequent orders by observing the state of the service provider’s environment, that is, the order within the dispatch range of the service provider’s location, so that the service provider
  • the provider’s total amount of actual resources for dispatching orders and estimated resources for subsequent orders is the largest, while the total amount of actual resources for dispatching orders and estimated resources for subsequent orders depends on the discount factor. The higher the discount factor, the greater the impact on subsequent orders.
  • the higher the degree of consideration of estimated resources that is, the more the total amount of actual resources for dispatching orders and the estimated resources for subsequent orders, the total amount of actual resources output through the order dispatch strategy network and estimated resources for subsequent orders
  • resources can be items, values, etc.
  • the degree of relevance represents the degree of matching between the service provider and the associated order.
  • the degree of relevance can be a score, and the greater the degree of relevance, It characterizes that the higher the degree of matching between the associated order and the service provider, it means that the service provider has a higher response rate to orders with a high degree of association.
  • obtain the location information and time information of the service provider and obtain the service start location information, service end location information and estimated resources of each related order in the dispatch range of the service provider’s location.
  • For each associated order enter the service start location information, service end location information, and estimated resources of the associated order, as well as the location information and time information of the service provider into the order distribution strategy network, and get the service provider and the associated order The degree of relevance between.
  • the service provider is driver A
  • the location of the service provider at 8:00 am is S (location information)
  • the attribute information of driver A includes location S and 8:00
  • driver A is within the dispatch range at location S
  • the order of order T1 is T1, T2
  • the order information of order T1 is the service start position S11, the service end position S12
  • the estimated resource is M1
  • the order information of order T2 is the service start position S21, the service end position S22, and the estimated resource
  • the order information is input into the order distribution strategy network, and the correlation degree R2 between driver A and order T2 is obtained.
  • S103 Determine a dispatch order for the service provider according to each degree of association, where the dispatch order maximizes the actual resources of the service provider and the estimated resources of subsequent orders.
  • the actual resources represent the resources actually obtained by the service provider after completing the order
  • the follow-up order is the order after the service provider completes the dispatch order
  • the estimated resource is the estimated resource that the service provider can obtain after completing the follow-up order.
  • resources can be goods, values, etc.
  • the order with the greatest degree of relevance is taken as the dispatch order of the service provider.
  • the degree of relevance is sorted in descending order, and it will be ranked first (that is, the degree of relevance is the largest) The order as the dispatch order of the service provider. In some embodiments, if the top two or more orders have the same degree of relevance, an order can be randomly selected from them.
  • the service provider is driver A
  • the dispatch range of driver A includes four orders, namely T1, T2, T3, and T4.
  • the correlation between driver A and order T1 is 0.8
  • the correlation between driver A and order T2 Is 0.9
  • the correlation degree between driver A and order T3 is 0.6
  • the correlation degree between driver A and order T4 is 0.5
  • the maximum correlation degree is 0.9
  • the order corresponding to 0.9 is T1. Therefore, order T1 is sent to as a dispatch order Driver A.
  • the current order data may be used to adjust the parameters of the order dispatch strategy network during the order dispatch process.
  • the historical order data in the vehicle usage platform may also be used to adjust the parameters of the order distribution strategy network. The specific adjustment method can be determined according to the actual application situation, which is not limited in this application.
  • the estimated value of the action value network and the matching degree output in the order distribution strategy network are generally adjusted by gradient descent.
  • Parameters In the process of adjusting the parameters in the order distribution strategy network, in order to improve the accuracy of the order distribution strategy network, further adjust the parameters of the action value network, which are described in detail below.
  • the method When adjusting the parameters in the order distribution strategy network, referring to Figure 2, the method also includes the following steps:
  • the first historical order is an order in the dispatch range where the historical service provider is located.
  • the location of the historical service provider is the same as or close to the location of the service provider (for example, the distance is less than 100 meters).
  • the time when the historical service provider is at the location is the same as the time information of the service provider (for example, both are 16:00) or similar (for example, the time difference is less than 10 minutes).
  • the first historical order is determined based on the selection result obtained by inputting the correlation degree of each first historical associated order associated with the historical service provider into the Boltzmann selector.
  • the characteristic historical service provider and each first historical association order obtained by the order distribution strategy network are obtained The matching probability of the order.
  • the greater the matching probability the higher the matching degree between the first historical association order and the historical service provider.
  • a sampling is performed according to the distribution output by the Boltzmann selector (e.g., extracting the highest matching probability ) Obtain the corresponding first historical associated order as the first historical order.
  • the first historical associated orders are all orders in the dispatch range where the historical service provider is located.
  • o i ) is the probability of the j-th first historical association order of the i-th historical service provider
  • ⁇ i (o i ,a i,j ) is the i-th historical service degree of association between the provider and its associated j-th order first history
  • scale factor beta] is typically a decimal between 0 and 1
  • o i is the service starting position and time of the i-th historical service provider at the first historical order
  • a i,j is the j-th first historical associated order of the i-th historical service provider at the service starting position
  • a i,m is the m-th associated order of the i-th historical service provider at the service start position.
  • the first historical related orders of driver A1 are T01, T02, T03
  • the correlation between driver A1 and T01 orders is R1
  • the correlation between driver A1 and T02 orders is R2
  • the correlation between driver A1 and T02 orders is R2
  • the correlation degree of is R3, and R1, R2, and R3 are respectively input to the Boltzmann selector
  • the first historical correlation order of driver A1 is between T01 and the matching probability G1
  • the historical correlation order of driver A1 is between T02
  • the matching probability G2 of the driver A1 is the matching probability G3 between T03. If G1 is the maximum matching probability, the historical correlation order T01 is the first historical order of the driver A1.
  • S202 Combine the first historical attribute information of the historical service provider corresponding to the first historical order, the first historical relevance degree corresponding to the first historical order, the historical order characteristics of the first historical order, and the historical The first historical average action of the service provider is input to the first action value network to obtain the first estimated resource of the first historical order, where the first historical average action is that the historical service provider is in the The supply and demand relationship between the historical service provider at the service end position of the first historical order and the historical order;
  • the first historical attribute information is the location information and time information when the historical service provider receives the first historical order
  • the first historical correlation is the relationship between the historical service provider and the first historical order output by the order dispatching strategy network.
  • the historical order characteristics of the first historical order are the service start position information and the service end position information in the first historical order
  • the first historical average action represents the supply and demand relationship at the location of the historical service provider
  • the first historical average The action may be the ratio of the number of historical service providers in the neighborhood of the historical service provider to the number of all historical orders in the dispatch range when the historical service provider is at the service end position of the first historical order.
  • the neighborhood is within a preset range of the location of the service provider, and the preset range may be greater than or equal to the order dispatch range.
  • the radius of the neighborhood is twice the radius of the dispatch range.
  • the first action value network is used to estimate the value of the historical service provider at the service start position corresponding to the first historical order.
  • the first action value network may be a Perceptron neural network, for example, a multilayer perceptron (Multilayer Perceptron, MLP) neural network.
  • the location information and time information of the location of the historical service provider corresponding to the first historical order, the first historical correlation degree between the first historical order and the historical service provider, and the service of the first historical order are input into the first action value network to obtain the first estimated resource of the first historical order.
  • the first historical order is T0
  • the order feature corresponding to the first historical order is that the service start position is S01
  • the service end position is S02
  • the historical service provider is driver A1
  • the relationship between driver A1 and the first historical order The degree is R0
  • the location of driver A1 at 8:00 am is S0 (GPS information)
  • the first historical attribute information of driver A1 includes locations S0 and 8:00
  • driver A1’s neighborhood includes N1
  • the dispatch range of driver A1 includes M1 orders
  • the first historical average action of driver A1 is N1/M1
  • the order characteristics and the first historical average action of the historical service provider are input to the first action value network, and the first estimated resource of the first historical order T0 received by the driver A1 service is obtained.
  • S203 Adjust parameters of the order distribution strategy network according to the first estimated resource and the first historical correlation.
  • a small-batch gradient descent algorithm is used to perform gradient descent iterative processing on the first estimated resource and the first historical correlation, and adjust the parameters of the order dispatch strategy network.
  • the gradient between the first estimated resource and the first historical relevance is calculated by the following formula:
  • Is the gradient of the output result of the order dispatching strategy network of the i-th historical service provider Is the gradient of the first historical relevance corresponding to the first historical order of the i-th historical service provider, Is the gradient of the first estimated resource corresponding to the first historical order of the i-th historical service provider, a i is the associated order of the i-th historical service provider at the service start position of the first historical order, o i represents The service starting position and time of the i-th historical service provider in the first historical order, It is the first historical average action of the i-th historical service provider.
  • the method further includes the following steps:
  • S401 Acquire a second historical order, where the second historical order is an associated order of the historical service provider at the end position of the first historical order service;
  • the second historical order is all orders in the dispatch range of the historical service provider at the service end position of the first historical order.
  • S402 Input the second historical attribute information of the historical service provider, the second historical relevance degree, the historical order characteristics of the second historical dispatch order, and the second historical average action of the historical service provider into a second action value
  • the network obtains the second estimated resource of the second historical order, where the second historical average action is the historical service provider and the historical service provider at the service end position of the second historical dispatch order Supply and demand of historical orders;
  • the second historical attribute information is the location information and time information of the historical service provider at the service end position of the first historical order
  • the second historical relevance is the historical service provider and each second historical service provider output by the order distribution strategy network.
  • the degree of correlation between historical orders, the second historical dispatch order is the order that the historical service provider ends in service; the historical order feature of the second historical dispatch order is the service start position information and service end position information in the second historical dispatch order
  • the second historical average action represents the supply-demand relationship at the location of the historical service provider.
  • the second historical average action can be the historical service provider’s history in the neighborhood of the historical service provider when the second historical dispatch order ends.
  • the second action value network is used to estimate the possible resources of the historical service provider at the service start position corresponding to the second historical dispatch order;
  • the two-action value network may be a Perceptron (Perceptron) neural network, for example, a Multilayer Perceptron (MLP) neural network.
  • Perceptron Perceptron
  • MLP Multilayer Perceptron
  • the location information and time information of the historical service provider s location, the second historical relevance degree of the second historical order and the historical service provider, the service starting location information of the second historical dispatch order and the service
  • the ending position information and the supply and demand relationship of the historical service provider at the service ending position of the second dispatch order are input into the second action value network to obtain the second estimated resource of the second historical order.
  • the second historical dispatch order is T00
  • the corresponding order feature of the second historical dispatch order is the service start position is S001
  • the service end position is S002
  • the historical service provider is driver A1
  • the second in the dispatch range of driver A1 Historical orders include T001, T002 and T003.
  • the correlation between driver A1 and the second historical dispatch order is R0
  • the second correlation between driver A1 and the second historical order T001 is R11
  • the second degree of correlation between T002 is R12
  • the second degree of correlation between driver A1 and the second historical order T003 is R13
  • the position of driver A1 at 9:00 am is S00 (GPS information)
  • the second degree of driver A1 Historical attribute information includes locations S00 and 9:00.
  • driver A1’s neighborhood includes N2 historical service providers
  • driver A1’s dispatch range includes M2 orders
  • the action is N2/M2.
  • the driver A1’ For each second historical order, the driver A1’s second historical attribute information, the second historical relevance, the order characteristics of the second historical dispatch order, and the second historical average of the historical service provider
  • the action is input to the second action value network, and each second estimated resource corresponding to each second historical order T001, second historical order T002, and second historical order T003 served by the driver A1 is obtained.
  • S403 Adjust a parameter of the first action value network according to the second estimated resource and the first estimated resource.
  • weighted calculation is performed on the second estimated resource of each second historical order to obtain a weighted average, and the above-mentioned weighted average, the actual resource of the first historical order, and the first estimated resource are input into the loss function, Make the loss function minimum adjust the parameters of the first action value network.
  • the weighted calculation of the second estimated resource of each second historical order may be the average value of the sum of the second estimated resources.
  • the actual resources of the first historical order are the actual resources of the first historical order, the demand potential of the first historical order at the end of the service, and the weighted value of the penalty for serving the first historical order, that is, the actual cost is calculated separately Obtain the sum of the product of the resource, demand potential, penalty and the corresponding weight, and use this sum as the actual resource of the first historical order.
  • the weights of actual deserved resources, demand potential, and punishment can be set according to the actual situation.
  • the weight of actual deserved resources is generally set to 1, and the weight of demand potential can be set to 1, 3, 5, 10, 20, etc.
  • the weight of the penalty can be set to 3, 5, 8, etc.
  • the actual due resource of the first historical order is the actual value of the historical service provider (for example, the actual income earned by the historical service provider after completing the first historical order).
  • the demand potential of the first historical order at the end of the service is, historical
  • the difference between the number of orders of the service provider in the dispatch range of the service end position of the first historical order and the number of historical service providers in the neighborhood, the order timeout penalty is based on the historical service provider and the first history
  • the order is determined by the distance between the service start positions.
  • the first historical order is T0
  • the actual revenue (actual resources) of the T0 order is 50
  • the order characteristics corresponding to the first historical order are that the service start position is S01 and the service end position is S02.
  • the historical service provider is driver A1
  • the distance between driver A1 and S01 is 1.5 kilometers. At this time, it is determined that the order overtime penalty for driver A1 is -1.5.
  • L( ⁇ i ) is the loss function value of the i-th historical service provider; r i is the actual resource of the first historical order of the i-th historical service provider; Is the average value of estimated resources of each second historical order associated with the i-th historical service provider; Is the estimated resource of the first historical order of the i-th historical service provider; ⁇ is the discount factor, which is generally a decimal between 0 and 1; o i represents the value of the i-th historical service provider’s first historical order Service start location and time, a i represents the first historical order of the i-th historical service provider, r i represents the actual resource of the first historical order of the i-th historical service provider, and o′ i represents the i-th history The service end position and time of the service provider in the first historical order; Is the first historical average action of the i-th historical service provider; Is the estimated resource of the second historical order of the i-th historical service provider, and ⁇ i (a
  • the sum of the actual resources of the first historical order of the historical service provider and the average value of the estimated resources of the associated orders after the first historical order is taken as the first historical order
  • the learning goal of the estimated resource that is, the estimated resource of the first historical order is infinitely close to the learning goal, when the estimated resource of the first historical order is infinitely close to the learning goal (for example, the estimated resource of the first historical order When the difference with the learning target is less than the preset threshold), the determined parameter is the parameter of the first action value network. In this way, the estimated resources obtained by the first action value network can be made more accurate.
  • the estimated resources estimated by the second action value network need to be used.
  • S501 Acquire parameters of the first action value network and parameters of the second action value network
  • the parameters of the first action value network and the parameters of the second action value network are acquired at the current moment, and the number of parameters of the first action value network is the same as the number of parameters of the second action value network.
  • the parameters of the first action value network will be adjusted in real time, and the parameters of the second action value network can be adjusted after the parameters of the first action value network are adjusted for a preset number of times. In this way, It is possible to improve the accuracy of the value estimated by the first action value network without increasing the processing volume.
  • S502 Perform weighting processing on the parameters of the first action value network and the parameters of the second action value network;
  • the weight of the first action value network is greater than the weight of the second action value network, and the weight of the first action value network and the second action value
  • the sum of the weights of the value network is 1.
  • the weight of the first action value network is set to 0.9
  • the weight of the second action value network is set to 0.1. In this way, the parameters of the second action value network will not decrease too much.
  • each parameter in the first action value network calculates the product of the parameter and the corresponding weight, use the product as the first value of the parameter, and calculate for each parameter in the second action value network
  • the product of the parameter and the corresponding weight use the product as the second value of the parameter, calculate the sum of each first value and the corresponding second value, and update the parameters in the second action value network according to the calculated sums .
  • the first action value network and the second action value network contain three parameters, the weight of the first action value network is 0.9, the weight of the second action value network is 0.1, and the parameters in the first action value network They are ⁇ 1, ⁇ 2, ⁇ 3.
  • the parameters of the second action value network are ⁇ 1, ⁇ 2, and ⁇ 3.
  • the product of the parameter and the weight in the first action value network are 0.9* ⁇ 1, 0.9* ⁇ 2, 0.9* ⁇ 3, and the second action
  • the product of the parameter and the weight in the value network are 0.1* ⁇ 1, 0.1* ⁇ 2, 0.1* ⁇ 3, respectively.
  • the parameter ⁇ 1 in the second action value network is updated to 0.9* ⁇ 1+0.1* ⁇ 1, and the second action value network is The parameter ⁇ 2 is updated to 0.9* ⁇ 2+0.1* ⁇ 2, and the parameter ⁇ 3 in the second action value network is updated to 0.9* ⁇ 3+0.1* ⁇ 3.
  • the associated information of the current order of each service provider is recorded, and the associated information of the current order includes the service start position of the current order.
  • the historical attribute information of the service provider, the correlation degree and order characteristics of the current order, the average action at the service end position of the current order, and the association information of the associated order at the service start position of the next order of each service provider includes the historical attribute information of the service provider, the association degree of the associated order of the next order, the order characteristics of the next order, and the service end position of the next order.
  • Average action taking the current order and the next order of each service provider as an order pair.
  • the associated information of the associated order of the location is input into the second action value network to obtain the estimated resource of each associated order.
  • the estimated resources of the current order use the small-batch gradient descent algorithm to reduce the gradient between the estimated resources of the first action value network and the matching degree of the order distribution strategy network, with the purpose of adjusting the parameters of the order distribution strategy network.
  • the parameters of the second action value network can be used in the first action. After adjusting the parameters of the value network for 100 times, adjust the parameters of the second action value network once.
  • the parameters of the first action value network and the parameters of the second action value network are weighted, and the parameters of the second action value network are updated based on the processing result.
  • the parameters of the order dispatching strategy network are different. In order to enable the car-using platform to dispatch more orders with a high response rate to the service provider as much as possible, from the historical order data, the completed orders in multiple dispatch cycles are selected. ,
  • the dispatch period can be a preset number of days, for example, the dispatch period is 1 day, 2 days, 7 days, etc.
  • the order dispatch strategy network Enter the attribute information of each service provider in the dispatch cycle and the order information of the associated order into the order dispatch strategy network to obtain dispatch orders from each service provider, estimate the estimated resources of each dispatch order, and judge each dispatch order Whether the estimated resources of all service providers in the cycle converge, that is to say, determine whether the sum of the estimated resources of all service providers in the dispatch cycle no longer increases, and determine whether the estimated resources of all service providers in the current dispatch cycle
  • the current order distribution strategy network is determined to be the final order distribution strategy network, and the order distribution strategy network distributes Orders will get a higher response rate and will improve the passenger experience.
  • the order dispatching method, device, electronic equipment, and computer-readable storage medium input the acquired attribute information of the service provider and the order information of all associated orders received by the service provider into the order dispatching strategy network , Obtain the degree of relevance between the service provider and each associated order, and then determine the dispatch order for the service provider based on the obtained degree of relevance.
  • the order dispatch strategy network dispatches the order to the service provider to maximize the current and future resources of the service provider, and This method is not easily affected by the excessive number of service providers in the order dispatching environment, and is suitable for dispatching scenarios where the service provider and the number of orders change over time. It has better robustness and real-time performance and is dispatched through orders. Orders distributed by the strategic network, on the one hand, increase the response rate of the service provider to the order and reduce the delay in order response due to order imbalance; on the other hand, it improves the experience of the service requester.
  • the embodiment of the present application provides an order dispatching device 60, as shown in FIG. 6, including:
  • the obtaining module 61 is configured to obtain attribute information of a service provider and order information of all associated orders received by the service provider;
  • the processing module 62 is configured to input the attribute information and the order information of all the associated orders into the order distribution strategy network to obtain the degree of association between the service provider and each of the associated orders;
  • the dispatch module 63 is configured to determine a dispatch order for the service provider according to all the obtained association degrees, and the dispatch order maximizes the total amount of actual resources of the service provider and estimated resources of subsequent orders.
  • the attribute information includes location information and time information of the service provider, and the order information includes at least service start location information, service end location information, and current order estimated resources.
  • the dispatch module 63 is specifically configured to:
  • the order with the greatest degree of relevance is taken as the dispatch order of the service provider.
  • an adjustment module 64 configured to:
  • the first historical attribute information of the historical service provider corresponding to the first historical order, the first historical association degree corresponding to the first historical order, the historical order characteristics of the first historical order, and the historical service provision The first historical average action of the party is input to the first action value network to obtain the first estimated resource of the first historical order, wherein the first historical average action is the historical service provider’s The supply-demand relationship between the historical service provider and the historical order at the service end position of the historical order;
  • the adjustment module 64 is further configured to:
  • the second historical order is an associated order of the historical service provider at the end position of the first historical order service
  • the adjustment module 64 is further configured to:
  • the parameters of the second action value network are updated based on the weighted processing result.
  • the supply and demand relationship is the ratio of the number of historical service providers to the number of historical orders.
  • the first historical order is determined based on a selection result obtained by inputting the relevance of each first historical associated order associated with the historical service provider into the Boltzmann selector.
  • the associated orders are all orders within the dispatch range of the location of the service provider.
  • the actual resource is obtained by weighting the actual resource due to the service provider, the demand potential of the service provider at the service end position of the dispatch order, and the penalty.
  • the embodiment of the present application also provides an electronic device 700.
  • the electronic device 700 may be a general-purpose computer or a special-purpose computer, both of which can be used to implement the order dispatch method of the present application. Although only one computer is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.
  • the electronic device 700 may include a network port 701 connected to a network, one or more processors 702 for executing program instructions, a communication bus 703, and different forms of storage media 704, such as magnetic disks, ROMs , Or RAM, or any combination thereof.
  • the computer platform may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. According to these program instructions, the method of this application can be realized.
  • the electronic device 700 also includes an input/output (Input/Output, I/O) interface 705 between the computer and other input and output devices (such as a keyboard and a display screen).
  • I/O input/output
  • step A and step B may also be executed by two different processors or be executed separately in one processor.
  • the first processor performs step A and the second processor performs step B, or the first processor and the second processor perform steps A and B together.
  • the processor 702 executes the following program instructions stored in the storage medium 704:
  • a dispatch order is determined for the service provider, and the dispatch order maximizes the total amount of actual resources of the service provider and estimated resources of subsequent orders.
  • the above attribute information includes location information and time information of the service provider, and the order information includes at least service start location information, service end location information, and current order estimated resources.
  • the program instructions executed by the processor 702 are specifically configured to determine the dispatch order for the service provider according to the respective degrees of association, including:
  • the order with the greatest degree of relevance is taken as the dispatch order of the service provider.
  • program instructions executed by the processor 702 are specifically used to:
  • the first historical attribute information of the historical service provider corresponding to the first historical order, the first historical association degree corresponding to the first historical order, the historical order characteristics of the first historical order, and the historical service provision The first historical average action of the party is input to the first action value network to obtain the first estimated resource of the first historical order, wherein the first historical average action is the historical service provider’s The supply-demand relationship between the historical service provider and the historical order at the service end position of the historical order;
  • program instructions executed by the processor 702 are specifically used to:
  • the second historical order is an associated order of the historical service provider at the end position of the first historical order service
  • program instructions executed by the processor 702 are specifically used to:
  • the parameters of the second action value network are updated based on the weighted processing result.
  • the above-mentioned supply and demand relationship is the ratio of the number of historical service providers to the number of historical orders.
  • the aforementioned first historical order is determined based on the selection result obtained by inputting the correlation degree of each first historical associated order associated with the historical service provider into the Boltzmann selector.
  • the above-mentioned related orders are all orders within the dispatch range of the location of the service provider.
  • the foregoing actual resources are obtained by weighted processing of the actual resources due to the service provider, the demand potential of the service provider at the service end position of the dispatch order, and the penalty.
  • an embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program executes the above Steps of the order dispatch method.
  • the computer-readable storage medium can be a general storage medium, such as a removable disk, a hard disk, etc., and when the computer program on the storage medium is run, it can execute the above-mentioned order dispatch method, thereby solving the problem of dispatching orders in the prior art. The question of balance.
  • the embodiments of the present application also provide a computer program product, which includes a computer-readable storage medium storing program code.
  • the instructions included in the program code can be used to execute the steps of the above order dispatching method. For specific implementation, see The above method embodiments will not be repeated here.
  • the order dispatching method, device, electronic equipment, and computer-readable storage medium input the acquired attribute information of the service provider and the order information of all associated orders received by the service provider into the order dispatching strategy network , Obtain the degree of relevance between the service provider and each associated order, and then determine the dispatch order for the service provider based on the obtained degree of relevance.
  • the order dispatch strategy network dispatches the order to the service provider to maximize the current and future resources of the service provider, and This method is not easily affected by the excessive number of service providers in the order dispatching environment, and is suitable for dispatching scenarios where the service provider and the number of orders change over time. It has better robustness and real-time performance and is dispatched through orders. Orders distributed by the strategic network, on the one hand, increase the response rate of the service provider to the order and reduce the delay in order response due to order imbalance; on the other hand, it improves the experience of the service requester.
  • the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • each unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the function is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a nonvolatile computer readable storage medium executable by a processor.
  • a computer device which may be a personal computer, a server, or a network device, etc.
  • the aforementioned storage media include: U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Tourism & Hospitality (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Game Theory and Decision Science (AREA)
  • Educational Administration (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种订单派发方法、装置、电子设备及计算机可读存储介质，其中，该方法包括：获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息（S101）；将所述属性信息和所有所述订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度（S102）；根据各所述关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源最大（S103）。上述方法能够提高服务提供方对订单的响应率。

Description

订单派发方法、装置、电子设备及计算机可读存储介质
交叉引用
本申请要求2019年4月9日提交的中国申请201910281576.2的优先权，全部内容通过引用并入本文。
技术领域
本申请涉及数据处理领域，具体而言，涉及一种订单派发方法、装置、电子设备及计算机可读存储介质。
背景技术
随着汽车电子技术的持续快速发展，乘坐出租车出行和预约乘坐私家车出行等出行方式得到了长足发展，在人们日常生活出行中起到了不可替代的作用，为广大人民的日常生活、交通出行带来了极大方便。
随着社会的进一步发展，传统的出租车已经不能满足人们出行的需求，为了满足用户的需求，目前市面上出现了网络预约车，方便用户通过用车软件预定符合自己行程的车辆。
随着提供服务的出租车和私家车数量的增多，已有的网约车平台在派单时通常通过贪心算法实现派单，贪心算法一般是按照司机和乘客之间的距离派单，优先把订单派给距离最近的司机，或者按照订单的价值排序，优先将价值最高的订单派给派单范围内的司机。但是，通过贪心算法派单时，只关注当前订单队列中的最优订单(如距离最近的订单或价值最高的订单)，无法考虑订单队列中的其它订单，在分配过程中导致部分服务提供方的响应率比较低。
发明内容
有鉴于此，本申请的目的在于提供一种订单派发方法、装置、电子设备及计算机可读存储介质，以解决现有技术中服务提供方对订单响应率低的问题。
第一方面，本申请实施例提供了一种订单派发方法，该方法包括：
获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息；
将所述属性信息和所述所有关联订单的订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度；
根据得到的所有关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源的总量最多。
可选地，所述属性信息包括所述服务提供方的位置信息和时间信息，所述订单信息至少包括服务起始位置信息、服务结束位置信息和当前订单预估资源。
可选地，所述根据各所述关联度，为所述服务提供方确定派发订单，包括：
将所述关联度最大的订单作为所述服务提供方的派发订单。
可选地，还包括：
获取第一历史订单；
将所述第一历史订单对应的历史服务提供方的第一历史属性信息、所述第一历史订单对应的第一历史关联度、所述第一历史订单的历史订单特征和所述历史服务提供方的第一历史平均动作输入到第一动作值网络，得到所述第一历史订单的第一预估资源，其中，所述第一历史平均动作为所述历史服务提供方在所述第一历史订单的服务结束位置的历史服务提供方与历史订单的供求关系；
根据所述第一预估资源和所述第一历史关联度，调整所述订单派发策略网络的参数。
可选地，还包括：
获取第二历史订单，所述第二历史订单为所述第一历史订单服务结束位置处的所述历史服务提供方的关联订单；
将所述历史服务提供方的第二历史属性信息、第二历史关联度、第二历史派发订单的历史订单特征和所述历史服务提供方的第二历史平均动作输入到第二动作值网络，得到所述第二历史订单的第二预估资源，其中，所述第二历史平均动作为所述历史服务提供方在所述第二历史派发订单的服务结束位置的历史服务提供方与历史订单的供求关系；
根据所述第二预估资源和所述第一预估资源，调整所述第一动作值网络的参数。
可选地，还包括：
获取所述第一动作值网络的参数和所述第二动作值网络的参数；
对所述第一动作值网络的参数和所述第二动作值网络的参数进行加权处理；
基于加权处理结果更新所述第二动作值网络的参数。
可选地，所述供求关系为历史服务提供方的数量与历史订单的数量的比值。
可选地，所述第一历史订单是基于将与所述历史服务提供方关联的各第一历史关联订单的关联度输入到玻尔兹曼选择器得到的选择结果确定的。
可选地，所述关联订单为所述服务提供方所处位置的派单范围内的所有订单。
可选地，所述实际资源为对所述服务提供方的实际应得资源、所述服 务提供方在所述派发订单的服务结束位置的需求潜力和惩罚进行加权处理得到的。
第二方面，本申请实施例提供了一种订单派发装置，该装置包括：
获取模块，用于获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息；
处理模块，用于将所述属性信息和所述所有关联订单的订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度；
派发模块，用于根据得到的所有关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源的总量最多。
可选地，所述属性信息包括所述服务提供方的位置信息和时间信息，所述订单信息至少包括服务起始位置信息、服务结束位置信息和当前订单预估资源。
可选地，所述派发模块具体用于：
将所述关联度最大的订单作为所述服务提供方的派发订单。
可选地，还包括：调整模块，所述调整模块用于：
获取第一历史订单；
将所述第一历史订单对应的历史服务提供方的第一历史属性信息、所述第一历史订单对应的第一历史关联度、所述第一历史订单的历史订单特征和所述历史服务提供方的第一历史平均动作输入到第一动作值网络，得到所述第一历史订单的第一预估资源，其中，所述第一历史平均动作为所述历史服务提供方在所述第一历史订单的服务结束位置的历史服务提供方与历史订单之间的供求关系；
根据所述第一预估资源和所述第一历史关联度，调整所述订单派发策略网络的参数。
可选地，所述调整模块还用于：
获取第二历史订单，所述第二历史订单为所述第一历史订单服务结束位置处的所述历史服务提供方的关联订单；
将所述历史服务提供方的第二历史属性信息、第二历史关联度、第二历史派发订单的历史订单特征和所述历史服务提供方的第二历史平均动作输入到第二动作值网络，得到所述第二历史订单的第二预估资源，其中，所述第二历史平均动作为所述历史服务提供方在所述第二历史派发订单的服务结束位置的历史服务提供方与历史订单的供求关系；
根据所述第二预估资源和所述第一预估资源，调整所述第一动作值网络的参数。
可选地，所述调整模块还用于：
获取所述第一动作值网络的参数和所述第二动作值网络的参数；
对所述第一动作值网络的参数和所述第二动作值网络的参数进行加权处理；
基于加权处理结果更新所述第二动作值网络的参数。
可选地，所述供求关系为历史服务提供方的数量与历史订单的数量的比值。
可选地，所述第一历史订单是基于将与所述历史服务提供方关联的各第一历史关联订单的关联度输入到玻尔兹曼选择器得到的选择结果确定的。
可选地，所述关联订单为所述服务提供方所处位置的派单范围内的所有订单。
可选地，所述实际资源为对所述服务提供方的实际应得资源、所述服务提供方在所述派发订单的服务结束位置的需求潜力和惩罚进行加权处理得到的。
第三方面，本申请实施例提供了一种电子设备，包括：处理器、存储介质和总线，所述存储介质存储有所述处理器可执行的机器可读指令，当电子设备运行时，所述处理器与所述存储介质之间通过总线通信，所述处理器执行所述机器可读指令，以执行时执行如第一方面所述的方法的步骤。
第四方面，本申请实施例提供了一种计算机可读存储介质，该计算机可读存储介质上存储有计算机程序，该计算机程序被处理器运行时执行如第一方面所述方法的步骤。
本申请实施例提供的订单派发方法、装置、电子设备及计算机可读存储介质，通过将获取的服务提供方的属性信息和服务提供方接收到的所有关联订单的订单信息输入到订单派发策略网络，得到服务提供方和各关联订单的关联度，进而基于得到的关联度为服务提供方确定派发订单，订单派发策略网络为服务提供方派发的订单使得服务提供方当前和未来的资源最多，这样，可以提高服务提供方对订单的响应率，减少订单响应延迟时长。
为使本申请的上述目的、特征和优点能更明显易懂，下文特举较佳实施例，并配合所附附图，作详细说明如下。
附图说明
为了更清楚地说明本申请实施例的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，应当理解，以下附图仅示出了本申请的某些实施例，因此不应被看作是对范围的限定，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他相关的附图。
图1示出了本申请实施例所提供的一种订单派发方法的第一种流程流 程图；
图2示出了本申请实施例所提供的一种订单派发方法的第二种流程流程图；
图3示出了本申请实施例所提供的一种服务提供方所处派单环境的示意图；
图4示出了本申请实施例所提供的一种订单派发方法的第三种流程流程图；
图5示出了本申请实施例所提供的一种订单派发方法的第四种流程流程图；
图6示出了本申请实施例所提供的一种订单派发装置的结构示意图；
图7示出了本申请实施例所提供的一种电子设备的结构示意图。
具体实施方式
为使本申请实施例的目的、技术方案和优点更加清楚，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，应当理解，本申请中附图仅起到说明和描述的目的，并不用于限定本申请的保护范围。另外，应当理解，示意性的附图并未按实物比例绘制。本申请中使用的流程图示出了根据本申请的一些实施例实现的操作。应该理解，流程图的操作可以不按顺序实现，没有逻辑的上下文关系的步骤可以反转顺序或者同时实施。此外，本领域技术人员在本申请内容的指引下，可以向流程图添加一个或多个其他操作，也可以从流程图中移除一个或多个操作。
另外，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。通常在此处附图中描述和示出的本申请实施例的组件可以以各种不同的配置来布置和设计。因此，以下对在附图中提供的本申请的实施例的 详细描述并非旨在限制要求保护的本申请的范围，而是仅仅表示本申请的选定实施例。基于本申请的实施例，本领域技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本申请保护的范围。
目前，随着出行需求的增长，用车平台每时每刻都会产生大量的订单，然而在用车平台中，司机的数量通常是小于订单的数量，也就是说，服务提供方的供给小于订单的需求。这种供给需求之间的不平衡使得一些乘客的订单无法分配给相应的司机，进而导致乘客等待时间较长，降低乘客的体验，同时，部分司机也可能没有被派到能带来较大潜在资源的订单，降低司机对订单的响应率，进而导致乘客和司机双方的体验均下降。
综合考虑用车平台、乘客、司机的需求，实现司机和乘客的智能匹配具有重要意义。为了提高服务提供方对订单的响应率，用车平台在派单时，考虑司机和订单之间的最佳匹配，在派单时，多个司机之间通常存在接单动作交互，如，位置接近的司机之间派单范围重叠，这种交互潜在地影响着司机对订单的响应率以及局部地区的供求关系，考虑到司机之间的交互，在司机数量一定的情况下，能够实现司机之间的相互协作，然而，在司机数量较多的情况下，导致司机对订单的响应率下降、响应时间延长，所以，亟需一种派单方法以解决服务提供方对订单的响应率低的问题。
为了使得本领域技术人员能够使用本申请内容，结合特定应用场景“智能派单”，给出以下实施方式。对于本领域技术人员来说，在不脱离本申请的精神和范围的情况下，可以将这里定义的一般原理应用于出行场景。虽然本申请主要围绕出行场景进行描述，但是应该理解，这仅是一个示例性实施例。
本申请实施例可以服务于用车平台，该用车平台用于根据接收的客户端的出行服务请求为用户提供相应的服务。用车平台可以包括多个打车系统，如包括出租车打车系统、快车打车系统、专车打车系统、顺风车打车系统等。
本申请实施例通过将获取服务提供方的属性信息和服务提供方接收到的所有关联订单的订单信息输入到订单派发策略网络，得到服务提供方和各关联订单的关联度，进而基于得到的关联度为服务提供方确定派发订单。本申请的订单派发方法不容易受司机数量过多的影响，适用于司机和订单数量随时间变化而变化的场景，具有更好的鲁棒性和实时性。以下针对本申请的技术方案进行详细说明。
本申请实施例提供了一种订单派发方法，应用于用车平台服务器，如图1所示，具体包括以下：
S101，获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息；
服务提供方一般为司机方，本申请中的服务提供方指用车平台中能够实时为乘客方提供服务的司机，如，服务提供方能够实时接收用车平台广播的订单；服务提供方的属性信息一般包括服务提供方的位置信息和时间信息，位置信息一般为通过全球卫星定位系统(GPS)获得的定位信息，时间信息一般为服务提供方所在位置的时间，如，司机2018年12月20日16:00位于Q街道，位置信息为Q街道的定位信息(北纬40度54分20秒，东经116度23分30秒)，时间信息为16:00。在一些实施例中，位置信息可以包括通过北斗系统、全球卫星导航系统(GLONASS)或伽利略卫星导航系统获得的定位信息。在一些实施例中，位置信息还可以包括通过WiFi定位技术、地磁定位技术、基站定位技术等方式获得的定位信息。在一些实施例中，服务提供方的属性信息还可以包括但不限于服务提供方的车型、空余座位、司机信息等一种或多种的任意组合。
关联订单为所述服务提供方所处位置的派单范围内的所有订单，派单范围一般为预先设置的范围，可以根据实际情况设置，例如，派单范围可以为司机所处位置为中心、半径2公里的圆形范围；订单信息至少包括服务起始位置信息、服务结束位置信息和当前订单预估资源，服务起始位置 信息表征当前订单中的服务起始位置，服务结束位置信息表征当前订单中的服务结束位置，当前订单预估资源表征预估的当前订单的价值。在一些实施例中，关联订单也可以为所述服务提供方所处位置的派单范围内的所有符合要求的订单。例如，乘客在发出订单时可以对订单附加要求(如特定车型、座位数等)。符合要求的订单即服务提供方的属性信息能够满足乘客附加要求的订单。
在具体实施中，用车平台服务器在获取到服务提供方的位置和时间后，用车平台服务器获取服务提供方所处位置的派单范围中的所有订单(或所有符合要求的订单)的订单信息。
S102，将所述属性信息和所有所述订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度。
这里，订单派发策略网络一般可以为感知器(Perceptron)神经网络，例如，多层感知器(Multilayer Perceptron，MLP)神经网络。订单派发策略网络可以通过观察服务提供方所处环境的状态，也就是，服务提供方所处位置的派单范围中的订单，从而对服务提供方的后续订单的不确定性进行估计，使得服务提供方的派发订单的实际资源和后续订单的预估资源的总量最多，而派发订单的实际资源和后续订单的预估资源的总量取决于折扣因子，折扣因子越高，对后续订单的预估资源的考虑程度越高，也就是说，派发订单的实际资源和后续订单的预估资源的总量也越多，通过订单派发策略网络输出的实际资源和后续订单的预估资源的总量越多的派发订单对应的关联度越大，其中，资源可以为物品、价值等；关联度表征服务提供方与关联订单之间的匹配程度，该关联度可以为打分，关联度越大，表征关联订单与服务提供方之间的匹配程度越高，也就意味着，服务提供方对关联度高的订单的响应率高。
在具体实施过程中，获取服务提供方的位置信息、时间信息，以及获取服务提供方所处位置的派单范围中各关联订单的服务起始位置信息、服 务结束位置信息以及预估资源，针对每个关联订单，将该关联订单的服务起始位置信息、服务结束位置信息以及预估资源，以及服务提供方的位置信息和时间信息输入到订单派发策略网络，得到服务提供方与该关联订单之间的关联度。
例如，服务提供方为司机A，服务提供方上午8:00所处位置为S(定位信息)，司机A的属性信息包括位置S和8:00，司机A在位置S处的派单范围内的订单有T1、T2，订单T1的订单信息为服务起始位置S11、服务结束位置S12、预估资源为M1，订单T2的订单信息为服务起始位置S21、服务结束位置S22、预估资源为M2，将司机A的位置信息、时间信息和订单T1的上述订单信息输入到订单派发策略网络，得到司机A和订单T1的关联度R1，将司机A的位置信息、时间信息和订单T2的订单信息输入到订单派发策略网络，得到司机A和订单T2的关联度R2。
S103，根据各所述关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源最多。
这里，实际资源表征服务提供方完成订单后实际得到的资源，后续订单为服务提供方完成派发订单之后的订单，预估资源为预估的服务提供方完成后续订单能够得到的资源。其中，资源可以为物品、价值等。
在所述根据各所述关联度，为所述服务提供方确定派发订单时，具体包括以下步骤：
将所述关联度最大的订单作为所述服务提供方的派发订单。
在具体实施中，在得到服务提供方与派单范围内各关联订单之间的关联度后，将关联度按照由大到小的顺序进行排序，将排在首位(也就是，关联度最大)的订单作为服务提供方的派发订单。在一些实施例中，若排在前两位或多位的订单的关联度相同，则可以从中随机选择某订单。
例如，服务提供方为司机A，司机A的派单范围内包括四个订单，分 别为T1、T2、T3、T4，司机A与订单T1的关联度为0.8，司机A与订单T2的关联度为0.9，司机A与订单T3的关联度为0.6，司机A与订单T4的关联度为0.5，最大的关联度为0.9，0.9对应的订单为T1，因此，将订单T1作为派单订单发送给司机A。
为了提高订单派发策略网络的准确度，需要调整订单派发策略网络中的参数。在一些实施例中，可以在订单派发过程中使用当前订单数据调整订单派发策略网络的参数。在一些实施例中，也可以使用用车平台中的历史订单数据对订单派发策略网络的参数进行调整。具体调整方式可以根据实际应用情况确定，本申请对此不予限制。
以下针对订单派发策略网络中的参数调整过程进行说明，在具体实施过程中，一般通过对动作值网络的预估值和订单派发策略网络中输出的匹配度进行梯度下降调整订单派发策略网络中的参数，在调整订单派发策略网络中的参数过程中，为了提高订单派发策略网络的准确度，进一步调整动作值网络的参数，以下详细叙述。
在调整订单派发策略网络中的参数时，参考图2，该方法还包括以下步骤：
S201，获取第一历史订单；
这里，第一历史订单为历史服务提供方所处位置的派单范围中的订单。在一些实施例中，历史服务提供方所处位置与服务提供方的位置相同或相近(如距离小于100米)。在一些实施例中，历史服务提供方处于所处位置的时间与服务提供方的时间信息相同(如均为16：00)或相近(如时间差小于10分钟)。
第一历史订单是基于将与所述历史服务提供方关联的各第一历史关联订单的关联度输入到玻尔兹曼(Boltzmann)选择器得到的选择结果确定的。
在具体实施中，通过将订单派发策略网络得到的历史服务提供方与各 第一历史关联订单之间的关联度输入到玻尔兹曼选择器，得到表征历史服务提供方与各第一历史关联订单的匹配概率，匹配概率越大，表征第一历史关联订单与历史服务提供方之间的匹配度越高，依照玻尔兹曼选择器输出的分布进行一次抽样(如，抽取匹配概率最大的)得到对应的第一历史关联订单作为第一历史订单。其中，第一历史关联订单为历史服务提供方所处位置的派单范围中的所有订单。
玻尔兹曼选择器对应的公式如下：
Figure PCTCN2020083947-appb-000001
其中，j＝1,…,M i
其中，π i(a i,j|o i)为第i个历史服务提供方的第j个第一历史关联订单的概率，μ i(o i,a i,j)为第i个历史服务提供方与其第j个第一历史关联订单之间的关联度，β是尺度因子，一般为0到1之间的小数，M i为第i个历史服务提供方的所有第一历史关联订单，o i为第i个历史服务提供方在第一历史订单的服务起始位置和时间，a i,j为第i个历史服务提供方在服务起始位置的第j个第一历史关联订单，a i,m为第i个历史服务提供方在服务起始位置的第m个关联订单。
例如，司机A1的第一历史关联订单为T01、T02、T03，司机A1与T01订单之间的关联度为R1，司机A1与T02订单之间的关联度为R2，司机A1与T02订单之间的关联度为R3，分别将R1、R2、R3输入到玻尔兹曼选择器，得到司机A1的第一历史关联订单为T01之间的匹配概率G1、司机A1的历史关联订单为T02之间的匹配概率G2、司机A1的第一历史关联订单为T03之间的匹配概率G3，若G1为最大匹配概率，则历史关联订单T01为司机A1的第一历史订单。
S202，将所述第一历史订单对应的历史服务提供方的第一历史属性信 息、所述第一历史订单对应的第一历史关联度、所述第一历史订单的历史订单特征和所述历史服务提供方的第一历史平均动作输入到第一动作值网络，得到所述第一历史订单的第一预估资源，其中，所述第一历史平均动作为所述历史服务提供方在所述第一历史订单的服务结束位置的历史服务提供方与历史订单之间的供求关系；
这里，第一历史属性信息为历史服务提供方接收第一历史订单时所处位置信息和时间信息，第一历史关联度为订单派发策略网络输出的历史服务提供方与第一历史订单之间的关联度，第一历史订单的历史订单特征为第一历史订单中的服务起始位置信息和服务结束位置信息，第一历史平均动作表征历史服务提供方所处位置的供求关系，第一历史平均动作可以为历史服务提供方在第一历史订单的服务结束位置时，历史服务提供方邻域中的历史服务提供方的数量与派单范围中所有历史订单的数目的比值。其中，邻域为服务提供方所处位置的预设范围内，该预设范围可以大于或等于派单范围。优选地，在邻域和派单范围为圆形时，邻域的半径是派单范围的半径的两倍，以单独一个服务提供方的派单环境为例，派单环境中服务提供方的邻域以及派单范围可以参考图3。第一动作值网络用于预估历史服务提供方位于第一历史订单对应的服务起始位置的价值，第一动作值网络可以为感知器(Perceptron)神经网络，例如，多层感知器(Multilayer Perceptron，MLP)神经网络。
在具体实施过程中，将第一历史订单对应的历史服务提供方所处位置的位置信息和时间信息、第一历史订单与历史服务提供方的第一历史关联度、第一历史订单的服务起始位置信息和服务结束位置信息、以及历史服务提供方在第一历史订单的服务结束位置处的供求关系输入到第一动作值网络，得到第一历史订单的第一预估资源。
例如，第一历史订单为T0，第一历史订单对应的订单特征为服务起始位置为S01、服务结束位置为S02，历史服务提供方为司机A1，司机A1 与第一历史订单之间的关联度为R0，司机A1上午8:00所处位置为S0(GPS信息)，司机A1的第一历史属性信息包括位置S0和8:00，司机A1位于S02位置时，司机A1的邻域内包括N1个历史服务提供方，司机A1的派单范围内包括M1个订单，司机A1的第一历史平均动作为N1/M1，将上述第一历史属性信息、第一历史关联度、第一历史订单的订单特征、以及历史服务提供方的第一历史平均动作输入到第一动作值网络，得到司机A1服务接收到第一历史订单T0的第一预估资源。
S203，根据所述第一预估资源和所述第一历史关联度，调整所述订单派发策略网络的参数。
在具体实施中，采用小批量梯度下降算法对所述第一预估资源和第一历史关联度进行梯度下降迭代处理，调整订单派发策略网络的参数。
通过以下公式计算第一预估资源和第一历史关联度之间的梯度：
Figure PCTCN2020083947-appb-000002
其中，
Figure PCTCN2020083947-appb-000003
为第i个历史服务提供方的订单派发策略网络的输出结果的梯度，
Figure PCTCN2020083947-appb-000004
为第i个历史服务提供方的第一历史订单对应的第一历史关联度的梯度，
Figure PCTCN2020083947-appb-000005
为第i个历史服务提供方的第一历史订单对应的第一预估资源的梯度，a i为第i个历史服务提供方在第一历史订单的服务起始位置的关联订单，o i表征第i个历史服务提供方在第一历史订单的服务起始位置和时间，
Figure PCTCN2020083947-appb-000006
为第i个历史服务提供方的第一历史平均动作。
在调整订单派发策略网络的参数时，需要对第一动作值网络的预估资源和订单配发策略网络的输出结果进行梯度下降处理，第一动作值网络得到的预估资源的准确度直接影响调整的订单派发策略网络参数的准确度，提高第一动作值网络的预估准确度可以提高订单派发策略网络的准确度。
在调整订单派发策略网络的参数过程中，如图4所示，该方法还包括以下步骤：
S401，获取第二历史订单，所述第二历史订单为所述第一历史订单服务结束位置处的所述历史服务提供方的关联订单；
这里，第二历史订单为所述第一历史订单的服务结束位置处的历史服务提供方的派单范围中的所有订单。
S402，将所述历史服务提供方的第二历史属性信息、第二历史关联度、第二历史派发订单的历史订单特征和所述历史服务提供方的第二历史平均动作输入到第二动作值网络，得到所述第二历史订单的第二预估资源，其中，所述第二历史平均动作为所述历史服务提供方在所述第二历史派发订单的服务结束位置的历史服务提供方与历史订单的供求关系；
这里，第二历史属性信息为历史服务提供方在第一历史订单的服务结束位置所处的位置信息和时间信息，第二历史关联度为订单派发策略网络输出的历史服务提供方与各第二历史订单之间的关联度，第二历史派发订单为历史服务提供方服务结束的订单；第二历史派发订单的历史订单特征为第二历史派发订单中的服务起始位置信息和服务结束位置信息，第二历史平均动作表征历史服务提供方所处位置的供求关系，第二历史平均动作可以为历史服务提供方在第二历史派发订单的服务结束位置时，历史服务提供方邻域中的历史服务提供方的数量与派单范围中所有历史订单的数目的比值，第二动作值网络用于预估历史服务提供方位于第二历史派发订单对应的服务起始位置的可能得到的资源；第二动作值网络可以为感知器(Perceptron)神经网络，例如，多层感知器(Multilayer Perceptron，MLP)神经网络。
在具体实施过程中，将历史服务提供方所处位置的位置信息和时间信息、第二历史订单与历史服务提供方的第二历史关联度、第二历史派发订单的服务起始位置信息和服务结束位置信息、以及历史服务提供方在第二派发订单的服务结束位置处的供求关系输入到第二动作值网络，得到第二历史订单的第二预估资源。
例如，第二历史派发订单为T00，第二历史派发订单对应的订单特征为服务起始位置为S001、服务结束位置为S002，历史服务提供方为司机A1，司机A1派单范围中的第二历史订单包括T001、T002和T003，司机A1与第二历史派发订单之间的关联度为R0，司机A1与第二历史订单T001之间的第二关联度为R11，司机A1与第二历史订单T002之间的第二关联度为R12，司机A1与第二历史订单T003之间的第二关联度为R13，司机A1上午9:00所处位置为S00(GPS信息)，司机A1的第二历史属性信息包括位置S00和9:00，司机A1位于S002位置时，司机A1的邻域内包括N2个历史服务提供方，司机A1的派单范围内包括M2个订单，司机A1的第一历史平均动作为N2/M2，针对每个第二历史订单，将该司机A1的第二历史属性信息、第二历史关联度、第二历史派发订单的订单特征、以及历史服务提供方的第二历史平均动作输入到第二动作值网络，得到司机A1服务各第二历史订单T001、第二历史订单T002、第二历史订单T003对应的各第二预估资源。
S403，根据所述第二预估资源和所述第一预估资源，调整所述第一动作值网络的参数。
在具体实施中，对各第二历史订单的第二预估资源进行加权计算，得到加权平均值，将上述加权平均值、第一历史订单的实际资源以及第一预估资源输入到损失函数，使得损失函数最小调整第一动作值网络的参数。其中，对各第二历史订单的第二预估资源进行加权计算可以为各第二预估资源的和值的平均值。
第一历史订单的实际资源为第一历史订单的实际应得资源、第一历史订单在服务结束位置的需求潜力、以及服务第一历史订单的惩罚的加权值，也就是说，分别计算实际应得资源、需求潜力和惩罚与相应权重的乘积的和值，将该和值作为第一历史订单的实际资源。其中，实际应得资源、需求潜力和惩罚的权重可以根据实际情况设定，例如，实际应得资源的权重 一般设置为1，需求潜力的权重可以设置为1、3、5、10、20等，惩罚的权重可以设置为3、5、8等。
第一历史订单的实际应得资源为历史服务提供方的实际价值(例如，历史服务提供方完成第一历史订单后实际所得的收入)，第一历史订单在服务结束位置的需求潜力为，历史服务提供方在第一历史订单的服务结束位置的派单范围中的订单的数量与邻域中的历史服务提供方的数量的差值，接单超时惩罚为基于历史服务提供方与第一历史订单的服务起始位置之间的距离确定的。
例如，延续步骤S302中的示例，第一历史订单为T0，T0订单的实际收益(实际资源)为50，第一历史订单对应的订单特征为服务起始位置为S01、服务结束位置为S02，历史服务提供方为司机A1，司机A1与S01之间的距离为1.5公里，此时确定司机A1的接单超时惩罚为-1.5，司机A1位于S02位置时，邻域内服务提供方的数量为5，派单范围内订单数量为7，司机A1的需求潜力为7-5＝2，实际回报资源的权重为1，需求潜力的权重为1，接单超时惩罚的权重为10，司机A1的第一历史订单的实际资源为50+2-15＝37。
在调整第一动作值网络参数时，一般通过调整学习目标与第一历史订单的预估资源之间的损失实现的，损失函数的公式如下：
Figure PCTCN2020083947-appb-000007
其中，
Figure PCTCN2020083947-appb-000008
其中，L(φ i)为第i个历史服务提供方的损失函数值；r i为第i个历史服务提供方的第一历史订单的实际资源；
Figure PCTCN2020083947-appb-000009
为第i个历史服务提供方关 联的各第二历史订单的预估资源的平均值；
Figure PCTCN2020083947-appb-000010
为第i个历史服务提供方的第一历史订单的预估资源；γ是折扣因子，一般为0到1之间的小数；o i表征第i个历史服务提供方在其第一历史订单的服务起始位置和时间，a i表征第i个历史服务提供方的第一历史订单，r i表征第i个历史服务提供方的第一历史订单的实际资源，o′ i表征第i个历史服务提供方在第一历史订单的服务结束位置和时间；
Figure PCTCN2020083947-appb-000011
为第i个历史服务提供方的第一历史平均动作；
Figure PCTCN2020083947-appb-000012
为第i个历史服务提供方的第二历史订单的预估资源，π i(a′ i|o′ i)为玻尔兹曼选择器输出的第i个历史服务提供方的第二历史订单的概率。
在调整第一动作值网络的参数时，将历史服务提供方的第一历史订单的实际资源和第一历史订单之后的各关联订单的预估资源的平均值的和值作为第一历史订单的预估资源的学习目标，也就是说，使得第一历史订单的预估资源无限接近学习目标，当第一历史订单的预估资源与学习目标无限接近(例如，第一历史订单的预估资源与学习目标的差值小于预设阈值)时，确定的参数为第一动作值网络的参数。这样，可以使得第一动作值网络得到的预估资源更加准确。
由于在调整第一动作值网络的参数时，需要使用第二动作值网络估计的预估资源，第二动作值网络的预估准确度越高，那么，调整的第一动作值网络的准确度越高，因此，在对第一动作值网络的参数进行调整时，同时也会调整第二动作值网络的参数，以下详细介绍第二动作值网络的参数的调整过程。
在调整第二动作值网络的参数时，参考图5，包括以下步骤：
S501，获取所述第一动作值网络的参数和所述第二动作值网络的参数；
这里，在更新第二动作值网络参数时获取当前时刻第一动作值网络的参数以及第二动作值网络的参数，第一动作值网络的参数的数目与第二动 作值网络的参数的数目相同。
为了提高订单派发策略网络的准确度，第一动作值网络的参数会实时进行调整，而第二动作值网络的参数可以在第一动作值网络的参数调整预设次数后再进行调整，这样，可以在不增加处理量的前提下，提高第一动作值网络预估的价值的准确度。
S502，对所述第一动作值网络的参数和所述第二动作值网络的参数进行加权处理；
S503，基于加权处理结果更新所述第二动作值网络的参数。
预先设置第一动作值网络的参数的权重和第二动作值网络的参数的权重，第一动作值网络的权重大于第二动作值网络的权重，且第一动作值网络的权重与第二动作值网络的权重的和值为1，例如，第一动作值网络的权重设置为0.9，第二动作值网络的权重设置为0.1。这样，使得第二动作值网络的参数不会降低太多。
在具体实施中，针对第一动作值网络中的每个参数，计算该参数与相应权重的乘积，将该乘积作为该参数的第一值，针对第二动作值网络中的每个参数，计算该参数与相应权重的乘积，将该乘积作为该参数的第二值，分别计算各第一值与相应第二值的和值，根据计算的各和值，更新第二动作值网络中的参数。
例如，第一动作值网络和第二动作值网络中包含的参数均为3个，第一动作值网络的权重为0.9，第二动作值网络的权重为0.1，第一动作值网络中的参数分别为α1、α2、α3，第二动作值网络的参数为γ1、γ2、γ3，第一动作值网络中参数与权重的乘积分别为0.9*α1、0.9*α2、0.9*α3，第二动作值网络中参数与权重的乘积分别为0.1*γ1、0.1*γ2、0.1*γ3，将第二动作值网络中的参数γ1更新为0.9*α1+0.1*γ1，将第二动作值网络中的参数γ2更新为0.9*α2+0.1*γ2，将第二动作值网络中的参数γ3 更新为0.9*α3+0.1*γ3。
在一种实施方式中，在用车平台中的各服务提供方服务完的历史订单数据后，记录各服务提供方的当前订单的关联信息，当前订单的关联信息包括当前订单的服务开始位置时的服务提供方的历史属性信息、当前订单的关联度和订单特征、当前订单的服务结束位置处的平均动作，以及相应各服务提供方的下一个订单的服务起始位置的关联订单的关联信息，下一个订单的服务起始位置的关联订单的关联信息包括服务提供方的历史属性信息、下一个订单的关联订单的关联度，下一个订单的订单特征、下一个订单的服务结束位置处的平均动作，将每个服务提供方的当前订单和下一个订单作为一个订单对。
从获取的各历史订单数据中，选择部分订单对，将订单对中当前订单的关联信息输入到第一动作值网络，得到当前订单的预估资源，将订单对中下一个订单的服务起始位置的关联订单的关联信息输入到第二动作值网络得到各关联订单的预估资源。
分别计算各关联订单的预估资源的平均值，将当前订单的实际资源和上述平均值作为第一动作值网络的学习目标，使得第一动作值网络的预估资源与学习目标之间的差最小调整第一动作值网络的参数。
在调整完第一动作值网络的参数后，从获取的各历史订单数据中，选择另外一部分订单对，将选择的该部分订单对中的当前订单的关联信息输入到第一动作值网络，得到当前订单的预估资源，利用小批量梯度下降算法来降低第一动作值网络的预估资源和订单派发策略网络的匹配度之间的梯度，目的在于调整订单派发策略网络的参数。
事实上每次调整订单派发策略网络的参数时，均会调整第一动作值网络的参数，为了减少订单派发网络的参数调整过程的数据处理量，第二动作值网络的参数可以在第一动作值网络的参数调整过如100次之后，调整 一次第二动作值网络的参数，在调整第二动作值网络参数时，获取第100次调整后的第一动作值网络的参数以及当前第二动作值网络中的参数，对第一动作值网络的参数和第二动作值网络中的参数进行加权处理，基于处理结果，更新第二动作值网络的参数。
每调整完一次订单派发策略网络的参数，可以将该订单派发策略网络应用于用车平台进行派单，事实上，在订单派发策略网络参数调整过程中，会得到大量的订单派发策略网络，不同的订单派发策略网络的参数不同，为了使得用车平台尽可能为服务提供方派发更多的、响应率高的订单，从历史订单数据中，选取多个派单周期中的已完成订单，其中，派单周期可以为预设天数，例如，派单周期为1天、2天、7天等。
将派单周期中的各服务提供方的属性信息和关联订单的订单信息输入到订单派发策略网络，得到各服务提供方的派发订单，预估各派发订单的预估资源，判断每个派单周期中所有服务提供方的预估资源是否收敛，也就是说，判断派单周期中所有服务提供方的预估资源的总和是否不再增加，在确定当前派单周期中所有服务提供方的预估资源收敛后，也就是，派单周期中所有服务提供方的预估资源的总和不再增加，则确定当前的订单派发策略网络为最终确定的订单派发策略网络，该订单派发策略网络派发的订单会得到较高的响应率，也会提高乘客的体验。
本申请实施例提供的订单派发方法、装置、电子设备及计算机可读存储介质，通过将获取的服务提供方的属性信息和服务提供方接收到的所有关联订单的订单信息输入到订单派发策略网络，得到服务提供方和各关联订单的关联度，进而基于得到的关联度为服务提供方确定派发订单，订单派发策略网络为服务提供方派发的订单使得服务提供方当前和未来的资源最多，且该方法不容易受派单环境中服务提供方数量过大的影响，适用于服务提供方和订单数量随时间变化而变化的派单场景，具有更好的鲁棒性和实时性，通过订单派发策略网络派发的订单，一方面，提高了服务提供 方对订单的响应率，减少由于订单不平衡带来的订单响应延迟时长，另一方面，提高了服务请求方的体验。
本申请实施例提供了一种订单派发装置60，如图6所示，包括：
获取模块61，用于获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息；
处理模块62，用于将所述属性信息和所述所有关联订单的订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度；
派发模块63，用于根据得到的所有关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源的总量最多。
在一种实施方式中，所述属性信息包括所述服务提供方的位置信息和时间信息，所述订单信息至少包括服务起始位置信息、服务结束位置信息和当前订单预估资源。
在一种实施方式中，所述派发模块63具体用于：
将所述关联度最大的订单作为所述服务提供方的派发订单。
在一种实施方式中，还包括：调整模块64，所述调整模块64用于：
获取第一历史订单；
将所述第一历史订单对应的历史服务提供方的第一历史属性信息、所述第一历史订单对应的第一历史关联度、所述第一历史订单的历史订单特征和所述历史服务提供方的第一历史平均动作输入到第一动作值网络，得到所述第一历史订单的第一预估资源，其中，所述第一历史平均动作为所述历史服务提供方在所述第一历史订单的服务结束位置的历史服务提供方与历史订单的供求关系；
根据所述第一预估资源和所述第一历史关联度，调整所述订单派发策略网络的参数。
在一种实施方式中，所述调整模块64还用于：
获取第二历史订单，所述第二历史订单为所述第一历史订单服务结束位置处的所述历史服务提供方的关联订单；
将所述历史服务提供方的第二历史属性信息、第二历史关联度、第二历史派发订单的历史订单特征和所述历史服务提供方的第二历史平均动作输入到第二动作值网络，得到所述第二历史订单的第二预估资源，其中，所述第二历史平均动作为所述历史服务提供方在所述第二历史派发订单的服务结束位置的历史服务提供方与历史订单的供求关系；
根据所述第二预估资源和所述第一预估资源，调整所述第一动作值网络的参数。
在一种实施方式中，所述调整模块64还用于：
获取所述第一动作值网络的参数和所述第二动作值网络的参数；
对所述第一动作值网络的参数和所述第二动作值网络的参数进行加权处理；
基于加权处理结果更新所述第二动作值网络的参数。
在一种实施方式中，所述供求关系为历史服务提供方的数量与历史订单的数量的比值。
在一种实施方式中，所述第一历史订单是基于将与所述历史服务提供方关联的各第一历史关联订单的关联度输入到玻尔兹曼选择器得到的选择结果确定的。
在一种实施方式中，所述关联订单为所述服务提供方所处位置的派单范围内的所有订单。
在一种实施方式中，所述实际资源为对所述服务提供方的实际应得资源、所述服务提供方在所述派发订单的服务结束位置的需求潜力和惩罚进行加权处理得到的。
本申请实施例还提供了一种电子设备700，电子设备700可以是通用计算机或特殊用途的计算机，两者都可以用于实现本申请的订单派发方法。本申请尽管仅示出了一个计算机，但是为了方便起见，可以在多个类似平台上以分布式方式实现本申请描述的功能，以均衡处理负载。
如图7所示，电子设备700可以包括连接到网络的网络端口701、用于执行程序指令的一个或多个处理器702、通信总线703、和不同形式的存储介质704，例如，磁盘、ROM、或RAM，或其任意组合。示例性地，计算机平台还可以包括存储在ROM、RAM、或其他类型的非暂时性存储介质、或其任意组合中的程序指令。根据这些程序指令可以实现本申请的方法。电子设备700还包括计算机与其他输入输出设备(例如键盘、显示屏)之间的输入/输出(Input/Output，I/O)接口705。
为了便于说明，在电子设备700中仅描述了一个处理器。然而，应当注意，本申请中的电子设备700还可以包括多个处理器，因此本申请中描述的一个处理器执行的步骤也可以由多个处理器联合执行或单独执行。例如，若电子设备700的处理器执行步骤A和步骤B，则应该理解，步骤A和步骤B也可以由两个不同的处理器共同执行或者在一个处理器中单独执行。例如，第一处理器执行步骤A，第二处理器执行步骤B，或者第一处理器和第二处理器共同执行步骤A和B。
下面以一个处理器为例，处理器702执行存储介质704中存储的如下程序指令：
获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息；
将所述属性信息和所述所有关联订单的订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度；
根据得到的所有关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源的总量最多。
上述属性信息包括所述服务提供方的位置信息和时间信息，所述订单信息至少包括服务起始位置信息、服务结束位置信息和当前订单预估资源。
在一种实施方式中，处理器702执行的程序指令具体用于根据各所述关联度，为所述服务提供方确定派发订单，包括：
将所述关联度最大的订单作为所述服务提供方的派发订单。
在一种实施方式中，处理器702执行的程序指令具体还用于：
获取第一历史订单；
将所述第一历史订单对应的历史服务提供方的第一历史属性信息、所述第一历史订单对应的第一历史关联度、所述第一历史订单的历史订单特征和所述历史服务提供方的第一历史平均动作输入到第一动作值网络，得到所述第一历史订单的第一预估资源，其中，所述第一历史平均动作为所述历史服务提供方在所述第一历史订单的服务结束位置的历史服务提供方与历史订单的供求关系；
根据所述第一预估资源和所述第一历史关联度，调整所述订单派发策略网络的参数。
在一种实施方式中，处理器702执行的程序指令具体还用于：
获取第二历史订单，所述第二历史订单为所述第一历史订单服务结束位置处的所述历史服务提供方的关联订单；
将所述历史服务提供方的第二历史属性信息、第二历史关联度、第二历史派发订单的历史订单特征和所述历史服务提供方的第二历史平均动作输入到第二动作值网络，得到所述第二历史订单的第二预估资源，其中， 所述第二历史平均动作为所述历史服务提供方在所述第二历史派发订单的服务结束位置的历史服务提供方与历史订单的供求关系；
根据所述第二预估资源和所述第一预估资源，调整所述第一动作值网络的参数。
在一种实施方式中，处理器702执行的程序指令具体还用于：
获取所述第一动作值网络的参数和所述第二动作值网络的参数；
对所述第一动作值网络的参数和所述第二动作值网络的参数进行加权处理；
基于加权处理结果更新所述第二动作值网络的参数。
上述供求关系为历史服务提供方的数量与历史订单的数量的比值。
上述第一历史订单是基于将与所述历史服务提供方关联的各第一历史关联订单的关联度输入到玻尔兹曼选择器得到的选择结果确定的。
上述关联订单为所述服务提供方所处位置的派单范围内的所有订单。
上述实际资源为对所述服务提供方的实际应得资源、所述服务提供方在所述派发订单的服务结束位置的需求潜力和惩罚进行加权处理得到的。
对应于图1至图5中的订单派发方法，本申请实施例还提供了一种计算机可读存储介质，该计算机可读存储介质上存储有计算机程序，该计算机程序被处理器运行时执行上述订单派发方法的步骤。
具体地，该计算机可读存储介质能够为通用的存储介质，如移动磁盘、硬盘等，该存储介质上的计算机程序被运行时，能够执行上述订单派发方法，从而解决现有技术中派单不平衡的问题。
基于相同的技术构思，本申请实施例还提供了一种计算机程序产品，包括存储了程序代码的计算机可读存储介质，程序代码包括的指令可用于执行上述订单派发方法的步骤，具体实现可参见上述方法实施例，在此不 再赘述。
本申请实施例提供的订单派发方法、装置、电子设备及计算机可读存储介质，通过将获取的服务提供方的属性信息和服务提供方接收到的所有关联订单的订单信息输入到订单派发策略网络，得到服务提供方和各关联订单的关联度，进而基于得到的关联度为服务提供方确定派发订单，订单派发策略网络为服务提供方派发的订单使得服务提供方当前和未来的资源最多，且该方法不容易受派单环境中服务提供方数量过大的影响，适用于服务提供方和订单数量随时间变化而变化的派单场景，具有更好的鲁棒性和实时性，通过订单派发策略网络派发的订单，一方面，提高了服务提供方对订单的响应率，减少由于订单不平衡带来的订单响应延迟时长，另一方面，提高了服务请求方的体验。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统和装置的具体工作过程，可以参考方法实施例中的对应过程，本申请中不再赘述。在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。以上所描述的装置实施例仅仅是示意性的，例如，所述模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，又例如，多个模块或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些通信接口，装置或模块的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元 中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个处理器可执行的非易失的计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
以上仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以权利要求的保护范围为准。

Claims (22)

  1. 一种订单派发方法，其特征在于，所述方法包括：
    获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息；
    将所述属性信息和所述所有关联订单的订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度；
    根据得到的所有关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源的总量最多。
  2. 如权利要求1所述的方法，其特征在于，所述属性信息包括所述服务提供方的位置信息和时间信息，所述订单信息至少包括服务起始位置信息、服务结束位置信息和当前订单预估资源。
  3. 如权利要求1所述的方法，其特征在于，所述根据各所述关联度，为所述服务提供方确定派发订单，包括：
    将所述关联度最大的订单作为所述服务提供方的派发订单。
  4. 如权利要求1所述的方法，其特征在于，还包括：
    获取第一历史订单；
    将所述第一历史订单对应的历史服务提供方的第一历史属性信息、所述第一历史订单对应的第一历史关联度、所述第一历史订单的历史订单特征和所述历史服务提供方的第一历史平均动作输入到第一动作值网络，得到所述第一历史订单的第一预估资源，其中，所述第一历史平均动作为所述历史服务提供方在所述第一历史订单的服务结束位置的历史服务提供方与历史订单的供求关系；
    根据所述第一预估资源和所述第一历史关联度，调整所述订单派发策略网络的参数。
  5. 如权利要求4所述的方法，其特征在于，还包括：
    获取第二历史订单，所述第二历史订单为所述第一历史订单服务结束位置处的所述历史服务提供方的关联订单；
    将所述历史服务提供方的第二历史属性信息、第二历史关联度、第二历史派发订单的历史订单特征和所述历史服务提供方的第二历史平均动作输入到第二动作值网络，得到所述第二历史订单的第二预估资源，其中，所述第二历史平均动作为所述历史服务提供方在所述第二历史派发订单的服务结束位置的历史服务提供方与历史订单的供求关系；
    根据所述第二预估资源和所述第一预估资源，调整所述第一动作值网络的参数。
  6. 如权利要求5所述的方法，其特征在于，还包括：
    获取所述第一动作值网络的参数和所述第二动作值网络的参数；
    对所述第一动作值网络的参数和所述第二动作值网络的参数进行加权处理；
    基于加权处理结果更新所述第二动作值网络的参数。
  7. 如权利要求5所述的方法，其特征在于，所述供求关系为历史服务提供方的数量与历史订单的数量的比值。
  8. 如权利要求4所述的方法，其特征在于，所述第一历史订单是基于将与所述历史服务提供方关联的各第一历史关联订单的关联度输入到玻尔兹曼选择器得到的选择结果确定的。
  9. 如权利要求1所述的方法，其特征在于，所述关联订单为所述服务提供方所处位置的派单范围内的所有订单。
  10. 如权利要求1所述的方法，其特征在于，所述实际资源为对所述服务提供方的实际应得资源、所述服务提供方在所述派发订单的服务结束位置的需求潜力和惩罚进行加权处理得到的。
  11. 一种订单派发装置，其特征在于，该装置包括：
    获取模块，用于获取服务提供方的属性信息和所述服务提供方接收到的所有关联订单的订单信息；
    处理模块，用于将所述属性信息和所述所有关联订单的订单信息输入到订单派发策略网络，得到所述服务提供方与各所述关联订单之间的关联度；
    派发模块，用于根据得到的所有关联度，为所述服务提供方确定派发订单，所述派发订单使得所述服务提供方的实际资源和后续订单的预估资源的总量最多。
  12. 如权利要求11所述的装置，其特征在于，所述属性信息包括所述服务提供方的位置信息和时间信息，所述订单信息至少包括服务起始位置信息、服务结束位置信息和当前订单预估资源。
  13. 如权利要求11所述的装置，其特征在于，所述派发模块具体用于：
    将所述关联度最大的订单作为所述服务提供方的派发订单。
  14. 如权利要求11所述的装置，其特征在于，还包括：调整模块，所述调整模块用于：
    获取第一历史订单；
    将所述第一历史订单对应的历史服务提供方的第一历史属性信息、所述第一历史订单对应的第一历史关联度、所述第一历史订单的历史订单特征和所述历史服务提供方的第一历史平均动作输入到第一动作值网络，得到所述第一历史订单的第一预估资源，其中，所述第一历史平均动作为所述历史服务提供方在所述第一历史订单的服务结束位置的历史服务提供方与历史订单之间的供求关系；
    根据所述第一预估资源和所述第一历史关联度，调整所述订单派发策略网络的参数。
  15. 如权利要求14所述的装置，其特征在于，所述调整模块还用于：
    获取第二历史订单，所述第二历史订单为所述第一历史订单服务结束位置处的所述历史服务提供方的关联订单；
    将所述历史服务提供方的第二历史属性信息、第二历史关联度、第二历史派发订单的历史订单特征和所述历史服务提供方的第二历史平均动作输入到第二动作值网络，得到所述第二历史订单的第二预估资源，其中，所述第二历史平均动作为所述历史服务提供方在所述第二历史派发订单的服务结束位置的历史服务提供方与历史订单的供求关系；
    根据所述第二预估资源和所述第一预估资源，调整所述第一动作值网络的参数。
  16. 如权利要求15所述的装置，其特征在于，所述调整模块还用于：
    获取所述第一动作值网络的参数和所述第二动作值网络的参数；
    对所述第一动作值网络的参数和所述第二动作值网络的参数进行加权处理；
    基于加权处理结果更新所述第二动作值网络的参数。
  17. 如权利要求16所述的装置，其特征在于，所述供求关系为历史服务提供方的数量与历史订单的数量的比值。
  18. 如权利要求14所述的装置，其特征在于，所述第一历史订单是基于将与所述历史服务提供方关联的各第一历史关联订单的关联度输入到玻尔兹曼选择器得到的选择结果确定的。
  19. 如权利要求11所述的装置，其特征在于，所述关联订单为所述服务提供方所处位置的派单范围内的所有订单。
  20. 如权利要求11所述的装置，其特征在于，所述实际资源为对所述服务提供方的实际应得资源、所述服务提供方在所述派发订单的服务结束位置的需求潜力和惩罚进行加权处理得到的。
  21. 一种电子设备，其特征在于，包括：处理器、存储介质和总线，所述存储介质存储有所述处理器可执行的机器可读指令，当电子设备运行时，所述处理器与所述存储介质之间通过总线通信，所述处理器执行所述机器可读指令，以执行时执行如权利要求1至10任一项所述的方法的步骤。
  22. 一种计算机可读存储介质，其特征在于，该计算机可读存储介质上存储有计算机程序，该计算机程序被处理器运行时执行如权利要求1至10任一项所述的方法的步骤。
PCT/CN2020/083947 2019-04-09 2020-04-09 订单派发方法、装置、电子设备及计算机可读存储介质 WO2020207430A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/450,458 US20220027822A1 (en) 2019-04-09 2021-10-09 Method, apparatus, electronic device, and computer-readable storage medium for distributing orders

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910281576.2 2019-04-09
CN201910281576.2A CN111798283A (zh) 2019-04-09 2019-04-09 订单派发方法、装置、电子设备及计算机可读存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/450,458 Continuation US20220027822A1 (en) 2019-04-09 2021-10-09 Method, apparatus, electronic device, and computer-readable storage medium for distributing orders

Publications (1)

Publication Number Publication Date
WO2020207430A1 true WO2020207430A1 (zh) 2020-10-15

Family

ID=72752146

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/083947 WO2020207430A1 (zh) 2019-04-09 2020-04-09 订单派发方法、装置、电子设备及计算机可读存储介质

Country Status (3)

Country Link
US (1) US20220027822A1 (zh)
CN (1) CN111798283A (zh)
WO (1) WO2020207430A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114066105A (zh) * 2022-01-11 2022-02-18 浙江口碑网络技术有限公司 运单配送超时预估模型的训练方法，存储介质和电子设备

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116757893B (zh) * 2023-04-10 2024-03-05 南京勤力信息科技有限公司 跨区域企业政务服务智能匹配方法及互联网平台系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090143965A1 (en) * 2007-12-03 2009-06-04 National Taiwan University Vehicle dispatch system
CN107767206A (zh) * 2016-08-23 2018-03-06 北京嘀嘀无限科技发展有限公司 订单分配方法及装置
CN108805411A (zh) * 2018-05-18 2018-11-13 北京嘀嘀无限科技发展有限公司 网约车订单分配方法、装置、服务器、终端和可读存储介质
CN109146348A (zh) * 2017-06-27 2019-01-04 阿里巴巴集团控股有限公司 一种物流数据处理方法及装置
CN109635972A (zh) * 2018-11-23 2019-04-16 杭州优行科技有限公司 网约车派单方法与装置

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10657581B2 (en) * 2015-02-02 2020-05-19 Beijing Didi Infinity Technology And Development Co., Ltd. Methods and systems for order processing
US20190370922A1 (en) * 2016-10-27 2019-12-05 University Of Southern California Price-aware real-time auction-based ride-sharing system
CN109284881A (zh) * 2017-07-20 2019-01-29 北京嘀嘀无限科技发展有限公司 订单分配方法、装置、计算机可读存储介质及电子设备
CN108596428B (zh) * 2018-03-13 2021-08-03 新智数字科技有限公司 一种订单派发的方法、装置以及设备
US10552773B1 (en) * 2018-09-07 2020-02-04 Lyft, Inc. Efficiency of a transportation matching system using geocoded provider models
US11796330B2 (en) * 2020-02-11 2023-10-24 Delphi Technologies Ip Limited System and method for providing value recommendations to ride-hailing drivers

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090143965A1 (en) * 2007-12-03 2009-06-04 National Taiwan University Vehicle dispatch system
CN107767206A (zh) * 2016-08-23 2018-03-06 北京嘀嘀无限科技发展有限公司 订单分配方法及装置
CN109146348A (zh) * 2017-06-27 2019-01-04 阿里巴巴集团控股有限公司 一种物流数据处理方法及装置
CN108805411A (zh) * 2018-05-18 2018-11-13 北京嘀嘀无限科技发展有限公司 网约车订单分配方法、装置、服务器、终端和可读存储介质
CN109635972A (zh) * 2018-11-23 2019-04-16 杭州优行科技有限公司 网约车派单方法与装置

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114066105A (zh) * 2022-01-11 2022-02-18 浙江口碑网络技术有限公司 运单配送超时预估模型的训练方法，存储介质和电子设备

Also Published As

Publication number Publication date
CN111798283A (zh) 2020-10-20
US20220027822A1 (en) 2022-01-27

Similar Documents

Publication Publication Date Title
US11416795B2 (en) Systems and methods for vehicle resource management
US11162803B2 (en) Providing alternative routing options to a rider of a transportation management system
US20190236742A1 (en) Method for requesting a ride service in a ride service system
JP2020515951A (ja) オンデマンドサービスのための乗り物を割り当てるシステム及び方法
US9746332B2 (en) Method and system for scheduling vehicles along routes in a transportation system
WO2017106256A1 (en) Systems and methods for adjusting ride-sharing schedules and routes
JP2019506640A (ja) 推奨サービス位置を推奨するシステム及び方法
JP2018533778A (ja) 共有可能な注文を割り当てるためのシステムおよび方法
US10593005B2 (en) Dynamic forecasting for forward reservation of cab
US20180314998A1 (en) Resource Allocation in a Network System
US9217647B2 (en) Guidebook transit routing
CN110832561B (zh) 用于确定和推荐车辆上车地点的系统和方法
US20210117874A1 (en) System for dispatching a driver
WO2019104078A1 (en) Dynamic route recommendation and progress monitoring for service providers
CN112236787A (zh) 用于生成个性化目的地推荐的系统和方法
WO2020207430A1 (zh) 订单派发方法、装置、电子设备及计算机可读存储介质
JP7078357B2 (ja) 配信装置、配信方法および配信プログラム
WO2018146622A1 (en) Dynamic selection of geo-based service options in a network system
CN111954891B (zh) 一种跨业务的共享汽车资源复用方法
US20170178085A1 (en) Method, apparatus, and system for managing reservations
JP2002024659A (ja) 配車予約システム
US11948165B2 (en) Apparatus for service acquisition
WO2020228607A1 (en) Method and system for multi-modal transportation
WO2020207418A1 (zh) 服务请求价格的计算方法、装置及系统
WO2016054700A1 (en) Online booking system

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20787558

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20787558

Country of ref document: EP

Kind code of ref document: A1