A dispatching method for a connected vehicle system

Abstract

The present invention discloses a scheduling method for a networked vehicle system, which belongs to the field of intelligent transportation. Specifically, the method comprises the following steps: first, for the current moment, obtaining in real time the number of service providers and service objects in a to-be-scheduled area and adjacent areas as supply and demand information; then, inputting the supply and demand information into a trained neural network, and outputting the value estimation of all pricing factors in each area; and using an ε-greedy strategy, selecting a unique pricing factor for each area; finally, inputting the number of service objects and the unique pricing factor in each area into an adjustment function, calculating the number of service objects that accept pricing in each area, and randomly selecting an equal number of service providers to provide services; at the same time, inputting the remaining service providers in each area that have not accepted pricing into a learned network flow scheduling model, and outputting and executing the scheduling strategy for each service provider; the present invention takes into account constraints in multiple aspects, such as scheduling cost, idle time, spatiotemporal pricing, and supply and demand balance, to maximize the benefits of the service provider.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation, and specifically is a dispatching method for a networked vehicle system. Background Art

[0002] Vehicle-to-Everything (V2X) is a new generation of communication technology that connects vehicles to things, where V stands for vehicle (service provider) and X stands for any object that interacts with the vehicle, including vehicles, people and networks. V2X organically links the elements of traffic participation together, which not only enables vehicles to obtain more sensory information than a single vehicle, but also helps to build a smart network system, which is of great significance to improving traffic efficiency, saving resources, reducing pollution, reducing the accident rate and improving traffic management.

[0003] However, a major obstacle to connected car systems is the imbalance between supply and demand in time and space; for example, when demand is higher than supply, service recipients endure longer waiting times, reducing user satisfaction; conversely, service providers drive around looking for orders, reducing their enthusiasm.

[0004] In order to solve the problem of supply and demand imbalance, existing technologies have proposed two solutions, namely dynamic pricing and vehicle scheduling. Dynamic pricing suppresses price-sensitive service objects and helps ensure high reliability services. Vehicle scheduling recommends areas with high potential income or high demand to service providers to improve system performance.

[0005] However, the existing technology does not consider the high correlation between optimized pricing and vehicle dispatching strategies, which will lead to poor performance of the service platform. If it relies solely on pricing, the service provider will tend to choose areas with high pricing and cannot fairly meet the needs of other areas. And vehicle dispatching that relies solely on demand cannot maximize the profit of the service provider because pricing will affect the demand distribution of the service objects. Summary of the invention

[0006] In view of the technical problem that the existing connected vehicle system does not jointly consider pricing and scheduling, resulting in an imbalance in supply and demand between the service parties, the present invention provides a scheduling method for a connected vehicle system, which simultaneously improves the order response rate of the connected vehicle system and the revenue of the service party, and reduces the waiting time of the service recipients.

[0007] The scheduling method of the networked vehicle system comprises the following steps:

[0008] Step 1: At the current moment, the number of service providers and service objects in the area to be scheduled and the adjacent areas is obtained in real time as supply and demand information;

[0009] Step 2: Use the samples to train the neural network, input the supply and demand information into the trained neural network, output the estimated value of all pricing factors in each region, and use the ε-greedy strategy to select a unique pricing factor for each region.

[0010] The steps to train a neural network are as follows:

[0011] First, obtain experience samples from the sample pool, input the experience samples of each region at the current historical moment into the estimated Q network, and output the value estimate of the corresponding regional pricing factor;

[0012] Then, the experience samples of each region at the next historical moment are input into the target Q network, and the value target value of the corresponding regional pricing factor is output;

[0013] Finally, for the estimated value and target value of the same area, based on minimizing the difference as the training goal, a loss function is constructed to update the parameters of the estimated Q network; and the target Q network is updated every fixed number of iterations; until the maximum number of iterations is reached, the estimated Q network is determined as the trained neural network model;

[0014] The loss function is as follows:

[0015]

[0016] Among them, θ is the parameter of the estimated Q network, θ - are the parameters of the target Q network, Refers to area Z i Supply and demand information at the current historical moment t, Refers to area Z i The pricing factor at the current historical moment t, γ is a hyperparameter that serves as the weight between immediate return and future value.

[0017] r t i Refers to area Z i The return value obtained by selecting the pricing factor at the current historical moment t is calculated as follows:

[0018]

[0019] in is the number of objects that the service provider can satisfy. It is a fixed price obtained by the order distance and travel time of each area. It is region Z i The number of service providers corresponding to the current historical moment t; It is region Z i The number of service objects corresponding to the current historical moment t; pently is a preset value.

[0020] Step 3: Input the number of service objects and the unique pricing factor in each area into the adjustment function, calculate the number of service objects that accept pricing in each area, and randomly select an equal number of service parties to provide services;

[0021] For region Z i , the number of service recipients receiving pricing The calculation formula is:

[0022]

[0023] in Indicates area Z i The number of service object requests, A regulation function indicating that the number of service recipients receiving pricing is negatively linearly correlated with the pricing factor.

[0024] Step 4: At the same time, the remaining service providers in each area that have not accepted the pricing are input into the learned network flow scheduling model, and the scheduling strategy for each service provider is output and executed;

[0025] The specific process is as follows:

[0026] Step 401: Input the supply and demand information of each region at each moment to calculate the supply and demand difference of each region. The supply and demand difference of all regions at that moment constitutes the supply and demand imbalance matrix

[0027] The formula for calculating the gap between supply and demand is:

[0028] Step 402, construct a node set and an edge set according to the supply-demand imbalance matrix to form a rescheduling network flow model corresponding to time t; the rescheduling network flow model is represented as a directed bipartite graph

[0029] Each region is regarded as a node, and the node set is represented according to the difference between supply and demand as follows: The service side is greater than the set of nodes of the service object, that is The service side is smaller than the set of nodes of the service object.

[0030] The edge set ε is composed of Nodes in The directed edges of the nodes in

[0031] Node Z in i arrive Node Z in j The weights of the directed edges are calculated as follows,

[0032]

[0033] Among them, the weight is node Z j The future value of node Z minus i The value of and the fuel cost; cost(Z i ,Z j ) is node Z i To Node Z j The fuel cost of directed edges; is the time t for node Z i The regional value of a single service provider within the region; e(Z i ,Z j ) is from node Z i Transition to Node Z j The moment required; is the time t+e(Z i ,Z j ) for node Z i The regional value of a single service provider within the region.

[0034] Step 403, based on the rescheduling network flow model, generate an objective function corresponding to time t that maximizes the total revenue of the service provider:

[0035]

[0036] in Represents the supply node Z i Dispatch to demand node Z j The number of idle servers; ε is the set of all regions;

[0037] At the same time, set the constraints corresponding to the objective function;

[0038] C1

[0039] C2

[0040] C3

[0041] C4

[0042] C5

[0043] The constraints include:

[0044] C1 ensures that the node Z i Dispatched to each demand node Z j The number of idle service providers does not exceed the supply node Z i The number of idle servers; represents time t area Zi The number of idle servers;

[0045] C2 ensures that all supply nodes Z i Dispatch to demand node Z j The number of idle service providers does not exceed the demand node Z j The remaining demand quantity at represents time t area Z i The number of remaining demand service providers;

[0046] C3 limits the scheduling space of idle servers;

[0047] C4 limits the relocation time from a supply node to a demand node to ensure that the demand node remains a demand node;

[0048] C5 ensures that the number of service parties is greater than or equal to 0;

[0049] Step 404, based on solving the objective function of linear programming, obtain the dispatch probability of the service provider corresponding to different areas at time t, and dispatch the service provider according to the probability;

[0050] At time t, it stays in area Z i The probability h(t,Z i ,Z i ) is calculated as:

[0051]

[0052] At time t, from area Z i Dispatch to region Z j The probability h(t,Z i ,Z j ) is calculated as:

[0053]

[0054] Step 405, updating the service provider scheduling strategy until all moments are completed, and obtaining the final vehicle scheduling strategy;

[0055] The update formula is:

[0056] h(t,Z i ,Z i )=(1-α)h'(t,Z i ,Z i )+αh(t,Z i ,Z i );

[0057] Among them, h'(t,Z i ,Z i) is the historical server scheduling probability strategy; α is a hyperparameter, which is used as the weight to measure the historical and current server scheduling strategies.

[0058] The advantages of the present invention are:

[0059] 1) A scheduling method for a connected vehicle system can be decomposed into multiple spatiotemporal pricing strategies for local areas according to the discretization of time and space domains. Distributed reinforcement learning is used to adaptively generate the optimal spatiotemporal pricing strategy, alleviating the complexity of the solution strategy.

[0060] 2) A scheduling method for a connected vehicle system is proposed. A network flow scheduling model is designed for joint spatiotemporal pricing and vehicle scheduling. The network flow method is used to maximize the service provider's revenue while meeting the needs of the service recipients.

[0061] 3) A scheduling method for a connected vehicle system aims to maximize the service provider's revenue. By considering multiple constraints such as scheduling costs, idle time, time-space pricing, and supply-demand balance, a vehicle scheduling strategy that maximizes the service provider's revenue is implemented, effectively improving the system's order response rate, the service provider's profit, and reducing the waiting time of the service recipient. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a flow chart of a scheduling method for a networked vehicle system of the present invention;

[0063] Figure 2 is a flow chart of the present invention for training a neural network model;

[0064] Figure 3 It is a schematic diagram of the internal structure of the neural network of the present invention;

[0065] Figure 4 is a flow chart of the present invention for learning a network flow scheduling model;

[0066] Figure 5 A structural diagram of an optimization device for a networked vehicle system is provided for an embodiment of the present invention;

[0067] Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The present invention will be further described below in conjunction with the embodiments and the accompanying drawings. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. After reading the present invention, modifications to various equivalent forms of the present invention made by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0069] The present invention proposes a scheduling method for a networked vehicle system, which specifically includes: obtaining the number of service providers and the number of service objects within each area in real time; inputting the real-time information into a trained neural network and outputting a pricing strategy for each area; inputting the number of service objects and the pricing strategy in each area into an adjustment function, calculating the number of service objects that accept pricing, and randomly selecting an equal number of service providers for service; inputting the remaining service providers in each area into a learned network flow scheduling model, outputting a scheduling strategy for the service providers and executing it; The present invention improves user benefits and order service levels

[0070] The scheduling method of a networked vehicle system, such as Figure 1 As shown, the following steps are included:

[0071] Step 1: At the current moment, the number of service providers and service objects in the area to be scheduled and the adjacent areas is obtained in real time as supply and demand information;

[0072] The supply and demand information includes: the number of service providers and service objects in each area and adjacent areas at the current moment;

[0073] Step 2: Use the samples to train the neural network, input the supply and demand information into the trained neural network, output the estimated value of all pricing factors in each region, and use the ε-greedy strategy to select a unique pricing factor for each region.

[0074] like Figure 2 As shown, the steps for training a neural network are as follows:

[0075] First, the supply and demand information at each historical moment, the pricing factor of the region, and the estimated value of the region are combined, that is, Store in a sample pool; perform sampling with replacement on the sample pool to obtain empirical samples, thereby breaking the association between data;

[0076] Get experience samples from the sample pool, input the experience samples of each region at the current historical moment into the estimated Q network, and output the estimated value of the corresponding regional pricing factor;

[0077] Then, the experience samples of each region at the next historical moment are input into the target Q network, and the value target value of the corresponding regional pricing factor is output;

[0078] Finally, for the estimated value and target value of the same area, based on minimizing the difference as the training goal, a loss function is constructed, and the loss function is minimized by stochastic gradient descent to update the parameters of the estimated Q network; and the target Q network is updated every fixed number of iterations; until the maximum number of iterations is reached, the estimated Q network is determined as the trained neural network model;

[0079] The loss function is as follows:

[0080]

[0081] Among them, θ is the parameter of the estimated Q network, θ - is the parameter of the target Q network. The estimated Q network and the target Q network are two neural networks with the same structure and initial weights, but the estimated Q network is updated every iteration, while the target Q network is updated every once in a while. The purpose is to prevent overfitting and help prevent the uncontrolled bias of the neural network bootstrapping from dominating numerically, thus causing the estimated Q value to diverge. If only one neural network is used, the target will also be updated every time it is updated, which can easily lead to parameter non-convergence.

[0082] Refers to area Z i Supply and demand information at the current historical moment t, Refers to area Z i The pricing factor at the current historical moment t, γ is a hyperparameter that serves as the weight between immediate return and future value.

[0083] r t i Refers to area Z i The return value obtained by selecting the pricing factor at the current historical moment t is calculated as follows:

[0084]

[0085] in is the number of objects that the service provider can satisfy. It is a fixed price obtained by the order distance and travel time of each area. It is region Z i The number of service providers corresponding to the current historical moment t; It is region Z i The number of service objects corresponding to the current historical moment t; pently is a preset value.

[0086] In order to balance the revenue of service providers who accept orders and the number of service providers who fail to accept orders, the reward objectives are twofold: (1) the total revenue of service providers who successfully accept orders in the area, and (2) the penalty for idle service providers.

[0087] Step 3: Input the number of service objects and the unique pricing factor in each area into the adjustment function, calculate the number of service objects that accept pricing in each area, and randomly select an equal number of service parties to provide services;

[0088] For region Z i , the number of service recipients receiving pricing The calculation formula is:

[0089]

[0090] in Indicates area Z i The number of service object requests, A regulation function indicating that the number of service recipients receiving pricing is negatively linearly correlated with the pricing factor.

[0091] Step 4: At the same time, the remaining service providers in each area that have not accepted the pricing are input into the learned network flow scheduling model, and the scheduling strategy for each service provider is output and executed;

[0092] The remaining service providers are service providers that have no service objects.

[0093] The specific process is as follows:

[0094] Step 401: Input the supply and demand information of each region at each moment to calculate the supply and demand difference of each region. The supply and demand difference of all regions at that moment constitutes the supply and demand imbalance matrix

[0095] The formula for calculating the gap between supply and demand is:

[0096] Step 402, construct a node set and an edge set according to the supply-demand imbalance matrix to form a rescheduling network flow model corresponding to time t; the rescheduling network flow model is represented as a directed bipartite graph

[0097] Each region is regarded as a node, and the node set is divided based on the supply-demand imbalance matrix. The node set is expressed as: The service side is greater than the set of nodes of the service object, that is The service side is smaller than the set of nodes of the service object.

[0098] The edge set ε is composed of Nodes in The directed edges of the nodes in represent feasible rescheduling; according to and and the spatiotemporal constraints determine the edge set ε; Node Z in i arrive Node Z in j The weights of the directed edges are calculated as follows,

[0099]

[0100] Among them, the weight is the destination node Z j The future value minus the departure node Z iThe value (regional value) and the fuel cost; cost (Z i ,Z j ) is node Z i To Node Z j The fuel cost of directed edges; is the time t for node Z i The regional value of a single service provider within the region (the estimated value of regional pricing / the number of drivers); e(Z i ,Z j ) is from node Z i Transition to Node Z j The moment required; is the time t+e(Z i ,Z j ) for node Z i The regional value of a single service provider within the region.

[0101] Step 403, based on the rescheduling network flow model, generate an objective function corresponding to time t that maximizes the total revenue of the service provider:

[0102]

[0103] in Represents the supply node Z i Dispatch to demand node Z j The number of idle servers; ε is the set of all regions;

[0104] At the same time, set the constraints corresponding to the objective function;

[0105] C1

[0106] C2

[0107] C3

[0108] C4

[0109] C5

[0110] The constraints include:

[0111] C1 ensures that the node Z i Dispatched to each demand node Z j The number of idle service providers does not exceed the supply node Z i The number of idle servers; represents time t area Z i The number of idle servers;

[0112] C2 ensures that all supply nodes Zi Dispatch to demand node Z j The number of idle service providers does not exceed the demand node Z j The remaining demand quantity at represents time t area Z i The number of remaining demand service providers;

[0113] C3 limits the scheduling space of idle servers;

[0114] C4 limits the relocation time from a supply node to a demand node to ensure that the demand node remains a demand node;

[0115] C5 ensures that the number of service parties is greater than or equal to 0;

[0116] Step 404, based on solving the objective function of linear programming, obtain the dispatch probability of the service provider corresponding to different areas at time t, and dispatch the service provider according to the probability;

[0117] At time t, it stays in area Z i The probability h(t,Z i ,Z i ) is calculated as:

[0118]

[0119] At time t, from area Z i Dispatch to region Z j The probability h(t,Z i ,Z j ) is calculated as:

[0120]

[0121] Step 405, updating the service provider scheduling strategy until all moments are completed, and obtaining the final vehicle scheduling strategy;

[0122] The update formula is:

[0123] h(t,Z i ,Z i )=(1-α)h'(t,Z i ,Z i )+αh(t,Z i ,Z i );

[0124] Among them, h'(t,Z i ,Z i ) is the historical server scheduling probability strategy; α is a hyperparameter, which is used as the weight to measure the historical and current server scheduling strategies.

[0125] For example, there are 10 service providers in area 1, and their destination areas are a, b, and c. The corresponding probabilities are 0.2, 0.3, and 0.5, respectively. Then the scheduling strategy is that 2 go to area a, 3 go to area b, and 5 go to area c.

[0126] The present invention provides an optimization device for a connected vehicle system, comprising:

[0127] The acquisition module obtains real-time supply and demand information within the area through V2X communication;

[0128] The pricing module inputs real-time information into the trained neural network and outputs the pricing strategy and value of each area;

[0129] Training module, used to train neural network;

[0130] Input the supply and demand information within a region into the trained neural network, and output the pricing factor and its value for this region; select the pricing factor with the largest regional value with a certain probability, otherwise, select a random pricing factor; until all regions are completed, output the pricing strategy and its value for each region.

[0131] The adjustment module inputs the number of service objects and the pricing strategy of each area into the adjustment function to calculate the number of service objects that accept the pricing; wherein the number of service objects that accept the pricing is negatively linearly correlated with the pricing factor;

[0132] The scheduling module inputs the remaining service providers in each area into the learned network flow scheduling model, outputs the scheduling strategy for the service providers and executes it;

[0133] The network flow module is used to construct the rescheduling network flow model corresponding to time slot t; specifically: first, the node set is divided based on the supply and demand imbalance matrix is the set of nodes where supply exceeds demand. is the set of nodes whose supply is less than demand; then, according to and And the space-time constraints determine the edge set ε; where ε is from Nodes in The nodes in the graph are composed of directed edges, which represent feasible rescheduling. The weight of each edge is calculated based on the fuel cost and regional value. Finally, based on the node set, edge set and edge weight, a relocation network flow model is constructed, which is represented as a directed bipartite graph.

[0134] The learning module is used to learn the network flow scheduling model; specifically, first, input the supply and demand information and regional value of each area in each time slot; calculate the degree of supply and demand imbalance (supply and demand difference) of each area in each time slot, and generate a supply and demand imbalance matrix; then, based on the supply and demand imbalance matrix and the regional value, construct a rescheduling network flow model corresponding to time slot t, generate an objective function that maximizes the total benefit of users corresponding to time slot t, and set the constraints corresponding to the objective function; finally, based on the linear programming solution of the objective function, obtain the vehicle scheduling strategy corresponding to time slot t; according to the calculation results, update the vehicle scheduling strategy; until all time slots are completed, the learned vehicle scheduling strategy is obtained.

[0135] The storage module stores the supply and demand information, regional pricing, and regional value estimates at each historical moment in the sample pool;

[0136] The sampling module performs sampling with replacement on the sample pool to obtain empirical samples;

[0137] When the present invention implements the optimization device of the networked vehicle system, the following electronic devices need to be applied, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus;

[0138] Memory, used to store computer programs;

[0139] The processor is used to implement any of the above-mentioned scheduling methods for the connected vehicle system when executing the computer program stored in the memory.

[0140] At the same time, it also includes a computer-readable storage medium for storing a computer program, which implements the above-mentioned scheduling method of the connected vehicle system when executed by a processor.

[0141] Example:

[0142] The scheduling method of the networked vehicle system is specifically as follows:

[0143] Step S11, obtaining supply and demand information within the area.

[0144] Divide the city into M non-overlapping hexagonal areas, that is, Z = {Z 1 ,...,Z M}; the area range of each area includes the current area and adjacent areas. Divide each day into T time slots, that is, T = {1,...,T}. Count the Z of each area in time slot t i The number of service providers and service objects is recorded as and As supply and demand information;

[0145] Step S12, input the supply and demand information into the trained neural network, and output the pricing strategy and value of each region, specifically:

[0146] Step 1: Input the supply and demand information within a region into the trained neural network, and output the pricing factor and its value of this region; where the pricing factor is a discrete multiplier factor in a given interval;

[0147] Step 2: Select the pricing factor with the largest regional value with a certain probability, otherwise, select a random pricing factor;

[0148] Step 3: Return to the previous step until all regions are completed, and output the pricing strategy and value of each region;

[0149] The internal structure of the neural network model is as follows Figure 3 As shown in the figure, it specifically includes the estimated Q network, the target Q network and the sample pool. In each training step, the empirical samples are extracted and input into the estimated Q network and the target Q network to update the parameters of the two networks. Until the two networks converge. In the actual step, the estimated Q network is used to estimate the corresponding regional value and determine the pricing factor. Finally, the obtained empirical samples are stored in the sample pool.

[0150] Step S13, generating the number of service objects that accept pricing according to the pricing factor of the calculated area, which is calculated as follows;

[0151]

[0152] in Indicates area Z i The number of service object requests, is the pricing factor, which is the product of the base price. It indicates that the number of service recipients accepting pricing is negatively linearly correlated with the pricing factor.

[0153] Step S14, input the remaining service providers in each area into the learned network flow scheduling model, output the scheduling strategy for the service providers and execute it;

[0154] The network flow scheduling model is learned through the following steps: Figure 4 As shown:

[0155] Step 1: Input the supply and demand information and regional value of each region in each time slot;

[0156] Step 2: Calculate the supply and demand imbalance matrix based on the supply and demand imbalance degree of each area in each time slot

[0157] Step 3: Based on the supply-demand imbalance matrix and the regional value, a rescheduling network flow model corresponding to time slot t is constructed;

[0158] Based on the node set, edge set and edge weight, a relocation network flow model is constructed, represented as a directed bipartite graph

[0159] Step 4: Generate an objective function that maximizes the total revenue of the user corresponding to time slot t according to the rescheduling network flow model;

[0160] Step 5: Set the constraints corresponding to the objective function;

[0161] C1

[0162] C2

[0163] C3

[0164] C4

[0165] C5

[0166] Step 6: Solve the objective function based on linear programming to obtain the vehicle scheduling strategy corresponding to time slot t

[0167] Step 7: Update the vehicle dispatching strategy based on the calculation results;

[0168] h(t,Z i ,Z i )=(1-α)h'(t,Z i ,Z i )+αh(t,Z i ,Z i )

[0169] Among them, h'(t,Z i ,Z i ) is the historical vehicle scheduling strategy;

[0170] Step 8, return to step 2 until all time slots are completed, and obtain the learned vehicle scheduling strategy;

[0171] A first structural diagram of an optimization device for a networked vehicle system provided by an embodiment of the present invention is as follows: Figure 5 As shown, it includes the following modules: acquisition module, pricing module, adjustment module and scheduling module;

[0172] The pricing module is specifically used to: input the supply and demand information within a region into the trained neural network, and output the pricing factor and its value of this region; the pricing factor is a multiplier within the range; select the pricing factor with the largest regional value with a certain probability, otherwise, select a random pricing factor; until all regions are completed, output the pricing strategy and its value of each region;

[0173] Optionally, the optimization device also includes a training module, a learning module, a storage module, a sampling module and a network flow module.

[0174] The embodiment of the present invention further provides an electronic device, such as Figure 6 As shown, it includes a processor 701, a communication interface 702, a memory 703 and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 communicate with each other through the communication bus 704.

[0175] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0176] The communication interface is used for communication between the above electronic device and other devices.

[0177] The memory includes a random access memory (RAM), a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0178] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0179] In another embodiment provided by the present invention, a computer-readable storage medium is also provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned connected vehicle travel time estimation methods are implemented.

[0180] In another embodiment provided by the present invention, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any of the scheduling methods for the connected vehicle system in the above embodiments.

[0181] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0182] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0183] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, electronic device, and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0184] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A dispatching method for a connected vehicle system, characterized in that: The specific steps are as follows: First, for the current moment, the number of service providers and service objects in the area to be scheduled and the adjacent areas is obtained in real time as supply and demand information; Then, the neural network is trained using the samples, and the supply and demand information is input into the trained neural network to output the estimated value of all pricing factors in each region; and the ε-greedy strategy is used to select the unique pricing factor for each region; Finally, the number of service objects and the unique pricing factor in each area are input into the adjustment function, the number of service objects that accept pricing in each area is calculated, and an equal number of service parties are randomly selected to provide services; For region Z i , the number of service recipients receiving pricing The calculation formula is: in Indicates area Z i The number of service object requests, The adjustment function indicates that the number of service recipients receiving pricing is negatively linearly correlated with the pricing factor; At the same time, the remaining service providers in each area who have not accepted the pricing are input into the learned network flow scheduling model, and the scheduling strategy for each service provider is output and executed.

2. A dispatching method for a connected vehicle system as claimed in claim 1, characterized in that: The steps of training the neural network are as follows: First, the supply and demand information at each historical moment and the pricing factors of the region are stored in the sample pool; the sample pool is sampled with replacement to obtain the experience sample, and the experience sample of each region at the current historical moment is input into the estimation Q network to output the value estimate of the corresponding regional pricing factor; Then, the experience samples of each region at the next historical moment are input into the target Q network, and the value target value of the corresponding regional pricing factor is output; Finally, for the estimated value and target value of the same area, a loss function is constructed based on minimizing the difference as the training goal. The loss function is minimized by stochastic gradient descent, and the parameters of the estimated Q network are updated. The target Q network is updated every fixed number of iterations until the maximum number of iterations is reached, and the estimated Q network is determined as the trained neural network model.

3. A dispatching method for a connected vehicle system as claimed in claim 2, characterized in that: The loss function is as follows: Among them, θ is the parameter of the estimated Q network, θ - are the parameters of the target Q network, Refers to area Z i Supply and demand information at the current historical moment t, Refers to area Z i The pricing factor at the current historical moment t, γ is a hyperparameter that serves as the weight of immediate return and future value; r t i Refers to area Z i The return value obtained by selecting the pricing factor at the current historical moment t is calculated as follows: in is the number of objects that the service provider can satisfy. It is a fixed price obtained by the order distance and travel time of each area. It is region Z i The number of service providers corresponding to the current historical moment t; It is region Z i The number of service objects corresponding to the current historical moment t; pently is a preset value.

4. The method for dispatching a connected vehicle system according to claim 1, characterized in that: The network flow scheduling model is learned, and the specific process is as follows: Step 401: Input the supply and demand information of each region at each moment to calculate the supply and demand difference of each region. The supply and demand difference of all regions at that moment constitutes the supply and demand imbalance matrix The formula for calculating the gap between supply and demand is: Step 402, constructing a node set and an edge set according to the supply-demand imbalance matrix to form a rescheduling network flow model corresponding to time t; The rescheduling network flow model is represented as a directed bipartite graph Each region is regarded as a node, and the node set is represented according to the difference between supply and demand as follows: The service side is greater than the set of nodes of the service object, that is The service side is smaller than the set of nodes of the service object. The edge set ε is composed of Nodes in The directed edges of the nodes in Step 403, based on the rescheduling network flow model, generate an objective function corresponding to time t that maximizes the total revenue of the service provider: in Represents the supply node Z i Dispatch to demand node Z j The number of idle servers; Represents node Z i To Node Z j The weights of the directed edges; At the same time, set the constraints corresponding to the objective function; C1 C2 C3 C4 C5 The constraints include: C1 ensures that the node Z i Dispatched to each demand node Z j The number of idle service providers does not exceed the supply node Z i The number of idle servers; represents time t area Z i The number of idle servers; C2 ensures that all supply nodes Z i Dispatch to demand node Z j The number of idle service providers does not exceed the demand node Z j The remaining demand quantity at represents time t area Z i The number of remaining demand service providers; C3 limits the scheduling space of idle servers; C4 limits the relocation time from a supply node to a demand node to ensure that the demand node remains a demand node; C5 ensures that the number of service parties is greater than or equal to 0; Step 404, based on solving the objective function of linear programming, obtain the dispatch probability of the service provider corresponding to different areas at time t, and dispatch the service provider according to the probability; At time t, it stays in area Z i The probability h(t,Z i ,Z i ) is calculated as: At time t, from area Z i Dispatch to region Z j The probability h(t,Z i ,Z j ) is calculated as: Step 405, update the service provider scheduling strategy until all moments are completed, and obtain the final vehicle scheduling strategy.

5. A dispatching method for a connected vehicle system as claimed in claim 4, characterized in that: The directed edge weight The calculation is as follows: Among them, the weight is node Z j The future value of node Z minus i The value of and the fuel cost; cost(Z i ,Z j ) is node Z i To Node Z j The fuel cost of directed edges; is the time t for node Z i The regional value of a single service provider within the region; e(Z i ,Z j ) is from node Z i Transition to Node Z j The moment required; is the time t+e(Z i ,Z j ) for node Z i The regional value of a single service provider within the region.

6. A method for dispatching a connected vehicle system according to claim 4, characterized in that: The update formula is: h(t,Z i ,WITH i )=(1-α)h'(t,Z i ,WITH i )+αh(t,Z i ,WITH i ); Among them, h'(t,Z i ,Z i ) is the historical server scheduling probability strategy; α is a hyperparameter, which is used as the weight to measure the historical and current server scheduling strategies.

Images