Digital twin resource allocation method based on UAV network flow model

Abstract

The present invention discloses a digital twin resource allocation method based on a drone network flow model. The network flow model is constructed according to the information of the digital twin system. When allocating resources, the IoT device with the largest marginal utility value in the remaining device set is selected as the candidate IoT device. Then, it is judged whether the current resource capacity of the digital twin cloud platform can meet the resource requirements of the candidate IoT device. Based on the network flow model, it is judged whether the bandwidth requirements of the current winner set and the candidate IoT device can be met. If both can be met, the candidate IoT device is taken as the winner, and so on until the maximum marginal utility value is less than or equal to 0 or all devices are allocated. Finally, the payment price of the corresponding user of the winning IoT device is determined according to the resource allocation plan. The present invention combines the network flow model with the reverse auction mechanism to solve the resource allocation and pricing problems in digital twin services.

Description

Technical Field

[0001] The present invention belongs to the technical field of cloud computing, and more specifically, relates to a digital twin resource allocation method based on an unmanned aerial vehicle network flow model. Background Art

[0002] Digital Twin (DT) is an effective way to connect physical space and digital space. It has a wide range of applications, such as constructing digital cities in the metaverse, optimizing the performance of the Internet of Vehicles, and assisting medical diagnosis and decision-making. In DT applications, there is a huge amount of data that needs to be processed and transmitted in real time, which poses new challenges to computing power and communication capabilities. DT is generally divided into four layers: physical layer, communication layer, virtual layer, and application layer. The devices in the physical layer transmit the operation status and data processing requirements to the virtual layer through the communication layer. The DT service builds a mirror of the real world in the virtual layer and provides it to the application layer. The application layer can analyze and feedback the real world by analyzing and processing the data mirror. For example, in the construction of DT digital cities, data from a large number of IoT devices need to be collected, processed, and transmitted in real time for monitoring and simulating the real world. In DT Internet of Vehicles, the analysis of current vehicle information can assist in making decisions on vehicle task unloading or resource allocation. In medical DT services, patient twins can be established to collect time-varying information of patients, so that doctors can predict changes in their condition in advance to prevent the disease from worsening. However, in DT services, large-scale data generated by different physical devices need to be transmitted to the virtual layer for DT modeling and updating, which increases the burden on communication links and data quality, thus affecting the efficiency of DT services.

[0003] In the design of traditional edge computing or cloud computing mechanisms, the connection between the end-edge or edge-cloud is usually represented as an edge, and deployment constraints are used to indicate which base stations or servers the device can be connected to. It is also stipulated that the user's data can only be transmitted through one of the edges. The advantage of this method is that the resource allocation problem can be converted into an integer programming problem for solution. The disadvantage is that it cannot reflect the actual network situation. In the drone base station networking mode, due to the distance between the device and the drone base station, the number of drone base station channels, and the data throughput of the drone base station, a single drone cannot meet the user's large data transmission needs (because the channel bandwidth is very small due to the distance factor). Therefore, the user's data may be transmitted through multiple drone base stations and finally gathered to the DT service provider.

[0004] Figure 1 It is a schematic diagram of the digital twin service mechanism model based on the drone base station network process. Figure 1In the figure, (1) indicates that the IoT device collects data and pre-processes it; (2) indicates that after the device evaluates the value of the data, it transmits the resource requirements, data value, bandwidth requirements, bidding, etc. to the DT service provider for decision-making; (3) indicates that the DT service provider determines the winner through the auction mechanism and obtains the network traffic plan; (4) indicates that the winning device transmits data through the UAV-BS network and finally processes it on the DTC; (5) indicates that the DT service provider pays the corresponding amount to the winning bidder.

[0005] Unlike traditional cloud computing or edge computing, in DT applications, heterogeneous data from different devices will be received. For example, vehicle driving information and air quality information will be transmitted to the digital city DT service at the same time, so the validity and value of the data have become important indicators for measuring the quality of DT services. How to select the most valuable users has become an important challenge in building a DT model. DT operators are willing to pay for collecting high-quality data, and data owners will also give a bid for the data they provide. This is essentially a problem of selecting high-quality users and pricing the value of data. The auction mechanism can allocate and price the resources used according to the supply and demand relationship in the market, and is an effective means of determining the value of commodities. In DT, service providers first need to evaluate the value of the data submitted by users, and then bid based on the user's bid information to select the winning user. For example, the paper "Lotfi Ismail, Dusit Niyato, Sumei Sun, DongIn Kim, Melike Erol-Kantarci, and Chunyan Miao, "Semantic Information Market For The Metaverse An Auction Based Approach," ArXiv, April 2022." uses the semantic information evaluation function of data to evaluate the data submitted by users, and then uses the reverse auction mechanism to select the winners and finally determine the fees paid to them. In the reverse auction model, the DT provider is the buyer of the data, selects the user, purchases the user's data, and pays the winning user. However, users may lie about information in order to increase their profits, so the design of the auction mechanism needs to meet credibility and individual rationality.

[0006] Another feature of DT services is that the geographical locations of data providers are dispersed. In traditional edge computing, it is always assumed that there are fixed base stations or roadside units (RSUs) to receive user data, but this is not realistic in DT services. For example, data from IoT devices in places with harsh natural environments need to be collected, or data from vehicles in places with poor communication conditions need to be collected. In these cases, there may be no fixed data transmission equipment, but instead a dynamic data transmission method is used, such as UAV-BS networking. Dynamic networking of UAV-BS can effectively overcome the problem of geographical dispersion of users, and is also a research hotspot in current network models. In the dynamic networking mode of UAV-BS, user data is first transmitted to the UAV-BS, and then transmitted to the DT service provider through the UAV. The biggest challenge of using UAV networking is that the power consumption of the UAV-BS is very small, and it cannot have a large transmit and receive power like an ordinary base station, as well as a data forwarding rate. These factors need to be considered when building a network model. Summary of the invention

[0007] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a digital twin resource allocation method based on the drone network flow model, which combines the network flow model with the reverse auction mechanism to solve the resource allocation and pricing problems in digital twin services.

[0008] In order to achieve the above-mentioned purpose of the invention, the digital twin resource allocation method based on the drone network flow model of the present invention comprises the following steps:

[0009] S1: Record the collection of IoT devices in the digital twin system Drone base station collection

[0010] The data value v of each IoT device i is calculated using the following formula: i :

[0011]

[0012] Among them, i represents the number of objects contained in the data collected by IoT device i, Indicates the deviation between the moving direction of object j and the preset reference direction, represents the moving speed of object j, Represents the area of ​​object j in the captured video image, j = 1, 2, ..., o j ;ω j represents the weather state of the observed object j, ω j is a real number greater than 1, and when the weather conditions are worse, its value is larger, and when the weather conditions are better, its value is closer to 1; ιij Represents the relative distance between object j and IoT device i. The smaller the value, the closer the distance between object j and IoT device i is, and vice versa.

[0013] The bandwidth requirement of IoT device i is bw i , the user corresponding to the IoT device i determines the resource demand s according to its actual situation i =(s i1 ,s i2 ,…,s iR ), s ir represents the demand of IoT device i for the rth resource, r = 1, 2, ..., R, and the bid of the user corresponding to IoT device i for its resource demand is b i , IoT device i reports parameter θ to the digital twin cloud platform i =(b i ,bw i ,s i ,v i );

[0014] Note the resource capacity of the digital twin cloud platform represents the capacity of the rth resource owned by the digital twin cloud platform;

[0015] S2: The digital twin cloud platform uses the following method to construct the network flow model of the digital twin system:

[0016] Set up 2 nodes for each drone base station, namely the drone base station ingress node and the drone base station egress node, and set up 2 nodes for the digital twin cloud platform, namely the digital twin cloud platform ingress node and the digital twin cloud platform egress node;

[0017] Set the starting point and the end point, together with the IoT device, the drone base station ingress node, the drone base station egress node, the digital twin cloud platform ingress node, and the digital twin cloud platform egress node as nodes to construct a directed graph G as the network flow model of the digital twin system, where:

[0018] The starting point is connected to each IoT device separately, and the capacity of the corresponding edge is the bandwidth demand bw of IoT device i i , the traffic is fixedly set to the bandwidth requirement bw of IoT device i i ;

[0019] Each IoT device is connected to each drone base station ingress node, and the capacity of the corresponding edge is the channel capacity BW between IoT device i and drone base station m im ;

[0020] Connect each drone base station ingress node m to the corresponding drone base station egress node m, and the capacity of the corresponding edge is the data throughput limit THP of the drone base station m. m ;

[0021] Connect each drone base station egress node m to the digital twin cloud platform ingress node, and the capacity of the corresponding edge is the channel capacity BW between the drone base station m and the digital twin cloud platform m,cloud ;

[0022] Connect the digital twin cloud platform ingress node to the digital twin cloud platform egress node. The capacity of the corresponding edge is the data throughput limit THP of the digital twin cloud platform. cloud ;

[0023] Connect the digital twin cloud platform egress node to the endpoint, and the capacity of the corresponding edge is a preset absolute maximum value;

[0024] Except for the edge between the starting point and each IoT device, the flow of the edge in the network flow model is the transmission rate between two nodes;

[0025] S3: Resource allocation based on the network flow model of the digital twin system. The specific method is as follows:

[0026] S3.1: Initialize the winner set of IoT devices Loser Collection Remaining equipment collection Marginal utility value of each IoT device Total Utility Maximum flow identifier result = 0;

[0027] S3.2: The marginal utility value of each IoT device i in the remaining device set u is calculated using the following formula:

[0028]

[0029] Then select the IoT device with the largest marginal utility value as the candidate IoT device i * ;

[0030] S3.3: Determine whether the IoT device i * The marginal utility value of If yes, go to step S3.4, otherwise the resource allocation ends;

[0031] S3.4: Determine whether the current resource capacity of the digital twin cloud platform can meet the needs of IoT devices * resource requirements, that is, to determine whether each resource meets r∈R, if satisfied, proceed to step S3.5, otherwise proceed to step S3.7;

[0032] S3.5: In the network flow model constructed in step S2, Delete all IoT devices except , solve the maximum flow of the obtained network flow model, and remember the maximum flow solution is The maximum flow identifier result is used to indicate whether there is a feasible flow solution. If so, the maximum flow identifier result = 1, otherwise, the maximum flow identifier result = 0;

[0033] S3.6: Determine whether the maximum flow identifier result = 0, if yes, proceed to step S3.7, otherwise proceed to step S3.8;

[0034] S3.7: Connect IoT devices * Add the loser set ε and delete it from the remaining device set u, and go to step S3.10;

[0035] S3.8: Connect IoT devices * Join the winners And delete it from the remaining device set u, and go to step S3.9;

[0036] S3.9: Update the resource capacity of the digital twin cloud platform using the following formula

[0037]

[0038] S3.10: Determine whether there are any remaining device sets If yes, the resource allocation ends, otherwise, return to step S3.2.

[0039] S4: According to the resource allocation plan obtained in step S3, the payment price of each winning IoT device corresponding to the user is solved.

[0040] The digital twin resource allocation method of the present invention is based on the network flow model of the unmanned aerial vehicle. The network flow model is constructed according to the information of the digital twin system. When allocating resources, the Internet of Things device with the largest marginal utility value in the remaining device set is screened out as the alternative Internet of Things device, and then it is judged whether the current resource capacity of the digital twin cloud platform can meet the resource requirements of the alternative Internet of Things device. Based on the network flow model, it is judged whether the bandwidth requirements of the current winner set and the alternative Internet of Things device can be met. If both can be met, the alternative Internet of Things device is selected as the winner, and so on until the maximum marginal utility value is less than or equal to 0 or all devices are allocated. Finally, the payment price of the user corresponding to the winning Internet of Things device is determined according to the resource allocation plan.

[0041] The present invention accurately models the network part that is ignored in traditional mechanism design, transforms the multi-UAV base station networking problem in digital twin services into a mixed integer programming model with resource and network flow constraints, and combines it with the reverse auction mechanism to solve the resource allocation and pricing problems in digital twin services. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the digital twin service mechanism model based on the drone base station network process;

[0043] Figure 2 It is a structural diagram of the digital twin model based on network flow in the present invention;

[0044] Figure 3 is an example diagram of the digital twin system in this embodiment;

[0045] Figure 4 It is a schematic diagram of the network traffic model of the improved digital twin system;

[0046] Figure 5 is a flow chart of a digital twin resource allocation method based on a UAV network traffic model in the present invention;

[0047] Figure 6 is a flow chart of resource allocation in the present invention;

[0048] Figure 7 It is a flowchart of the pricing algorithm based on the dichotomy method in this embodiment. DETAILED DESCRIPTION

[0049] The specific implementation of the present invention is described below in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0050] In order to better illustrate the technical solution of the present invention, the digital twin model based on network flow constructed by the present invention is first briefly described.

[0051] The network flow-based digital twin model designed by the present invention involves a three-layer structure: digital twin cloud, drone base station network layer, and user equipment. Figure 2 It is a structural diagram of the digital twin model based on network flow in the present invention. Figure 2As shown in the figure, 1. After collecting environmental data, the user device extracts data information and evaluates the value of the collected data according to the value function; 2. The user proposes his own bid based on the data value, the data resource requirements and bandwidth requirements to be processed, and transmits the above information to the digital twin cloud platform DTC (Digital Twins Cloud) for decision-making. 3. The digital twin cloud platform selects the winner user and payment price plan based on the information submitted by the user and the communication status of the drone base station, and plans the user data transmission network flow model. 4. The user uploads data according to the allocation results and the network flow model.

[0052] Because drone base stations are used to transmit data, the bandwidth requirements of some IoT devices may exceed the channel capacity, so the devices will use network flow technology to transmit data through different base stations. At the same time, wireless transmission is also used between the base station and the digital twin cloud platform, which is also limited by the base station bandwidth and throughput. Therefore, a feasible network flow data transmission solution should be designed for the winning user to ensure that there are no communication problems such as data loss during the transmission process. This is also an important difference between drone base station networking and traditional cloud computing data transmission. The main technologies in the digital twin model based on network flow of the present invention are explained below.

[0053] IoT device model

[0054] Assume that there are N IoT devices in the system, use the collection Define, each device Real-world environmental data will be collected After that, IoT device i will first pre-process the raw data (such as encoding) and then transmit it to the digital twin cloud platform. Therefore, the data size requested to be transmitted by IoT device i is the size of the pre-processed data, which is defined as d i The IoT device requests the corresponding CPU instruction cycle resources from the digital twin cloud platform to process the data transmitted by the device. The number of CPU instruction cycle resources is related to d i Depending on the data type, the number of CPU instruction cycles s requested by IoT device i i1 It can be written as

[0055] s i1 =d i τ i (1)

[0056] Among them, τ i It is the number of CPU instruction cycles required to process 1 unit of data corresponding to the user of IoT device i. This parameter is fixed. Similarly, the memory required for processing data is defined as:

[0057] s i2 =di λ i (2)

[0058] λ i It is the amount of memory required by the digital twin cloud platform to process 1 unit of data transmitted by the user corresponding to the IoT device i. Generally speaking, τ i , i It is related to the type of user data and the task to be completed. The device has a time limit for transmitting data. i This means that the IoT device needs to transmit data to the digital twin cloud platform within this time limit. Therefore, the bandwidth bw requested by IoT device i i Must meet:

[0059]

[0060] Therefore, in order to convert the preprocessed data d i The device will request resources from the digital twin cloud platform using vector s i =(s i1 ,s i2 ). At the same time, in order to transmit data to the digital twin cloud platform within the specified time, the transmission bandwidth requested by IoT device i is bw i For example, to transmit a 720P, 30Fps H.264-encoded video to DTC for processing, a 3.3GHz CPU, 2Mbps bandwidth, and 1GB memory may be required. The value of user data v can be calculated based on the value function built into the device. i (See formula (15)). At the same time, the bid of the user corresponding to IoT device i for the proposed data is b i Therefore, the parameter θ reported by IoT device i to the digital twin cloud platform is i =(b i ,bw i ,s i ,v i ).

[0061] Drone base station model

[0062] In traditional cloud computing and edge computing models, the communication connection between the device and the server is assumed to be in a state of 0 or 1 based on deployment constraints, and the user's data can only be transmitted from one connection. This is inconsistent with the actual situation. In the case of drone base station networking, a single drone may not be able to meet the user's large data transmission needs (because the channel bandwidth is very small due to distance factors), so the user's data may be transmitted through multiple drone base stations. Assume that IoT device i can be connected to multiple drone base stations, but the channel capacity of the connection is limited by the distance between the device and the drone base station, the transmission power of the drone base station antenna, and the signal-to-noise ratio and other parameters. Assume that there are M drone base stations in the system, and use the set Definition, define the channel capacity BW between IoT device i and drone base station m im as follows:

[0063]

[0064] According to Shannon's theorem, B im is the bandwidth between IoT device i and drone base station m, S im is the average signal power between IoT device i and drone base station m, N i is the average noise power between IoT device i and drone base station m. In order to avoid data loss caused by excessive data received by the base station, the total network flow received by the drone base station must be less than or equal to the maximum throughput (data forwarding rate) of the base station. m At the same time, the network flow finally transmitted between IoT device i and drone base station m must be less than or equal to the channel capacity BW of the channel im .

[0065] In order to ensure that multiple devices do not transmit data on the same channel when the IoT device transmits data to the drone base station, it is necessary to frequency-division multiplex the channel frequency of the drone base station. Therefore, the drone base station m has a limited number of channels CH m This means that the number of IoT devices that the drone base station m receives data from must be less than or equal to the number of its own channels CH m .

[0066] At the same time, wireless data transmission is also carried out between the drone base station and the digital twin cloud platform, which is limited by factors such as channel capacity and signal power, channel bandwidth, and distance. The channel capacity BW between the drone base station m and DTC m,cloud The definition is as follows:

[0067]

[0068] Among them, B m,cloudis the bandwidth for communication between DTC and base station m. The bandwidth frequency interval is allocated to prevent data from being mixed when the base station transmits data. S m,cloud is the average signal power between the drone base station and DTC, N m,cloud is the average noise power between the unmanned base station m and the DTC.

[0069] In summary, there are M drone base stations in the digital twin system of the present invention, and the set Indicates that the base station The throughput constraint is limited to THP m , and the number of channels per base station m is constrained to CH m , the channel capacity constraint between drone base station m and IoT device i is BW im , the channel capacity constraint between the drone base station m and the digital twin cloud platform is BW m,cloud .

[0070] Digital twin cloud model

[0071] In the digital twin cloud model, the digital twin cloud platform needs to receive data transmitted by the drone base station. However, in order to ensure that the digital twin cloud platform can receive and process data in a timely manner, the total data flow forwarded by the drone base station to the digital twin cloud platform must be less than or equal to the throughput THP of the digital twin cloud platform. cloud , ensuring that such constraints will not cause data loss due to untimely data reception. The throughput of the digital twin cloud platform THP cloud It is determined by its own equipment (such as the data receiving rate of the core switch and the processing capacity of the digital twin cloud platform). At the same time, it is assumed that DTC can provide two resources such as CPU instruction cycles. and memory However, the total amount of these resources is limited, so it is also necessary to ensure that the total resource demand of IoT devices in the winner set is less than or equal to the resource capacity of the digital twin cloud platform.

[0072] ·Calculation of digital twin service fees and data value

[0073] To realize the resource allocation of DT services, it is necessary to describe the data collection cost of IoT devices, the cost of using drone base stations, the data transmission cost, the computing cost of the digital twin cloud platform, and the data value.

[0074] 1) Cost of IoT devices: The cost of IoT devices includes the cost of collecting environmental data, the cost of data preprocessing, and the cost of communication. It can be written as

[0075]

[0076] Among them, col represents the stage of IoT devices collecting environmental data, is the size of the original data collected by IoT device i, α i It is the unit cost of collecting raw data by IoT device i.

[0077] After collecting environmental data, IoT devices need to pre-process the data, such as grayscale processing, compression, feature extraction, or semantic information extraction. It can be written as:

[0078]

[0079] Among them, pre represents the data preprocessing stage of IoT devices, γ i is the unit cost of data preprocessing of IoT device i. Data computation cost of IoT device i is the cost of collecting environmental data and the cost of preprocessing the data The sum can be written as:

[0080]

[0081] Where comp represents the data calculation stage of IoT devices. After the IoT device preprocesses the data, the data size will be reduced accordingly. The data size d after the IoT device i preprocesses the data i It can be written as:

[0082]

[0083] Among them, λ i It is the data compression rate of the preprocessed data of IoT device i. The data transmitted by IoT devices through drone base stations are preprocessed data. For example, IoT devices use H264 algorithm to compress the collected video data, use grayscale processing algorithm to preprocess the collected environmental data, or use FCN and Seg-Net algorithms to preprocess environmental data and extract semantic data information such as object status, object quantity, weather conditions, etc. from environmental data.

[0084] The pre-processed data needs to be transmitted to the drone base station. The communication cost of IoT device i transmitting data to the drone base station is It can be written as:

[0085]

[0086] Among them, comm represents the communication stage of IoT devices, β i is the communication cost per unit time of IoT device i, bw iis the bandwidth requested by IoT device i. The total service cost of IoT device i is the data computation cost and communication costs The sum can be written as:

[0087]

[0088] 2) Service cost of drone base station: The service cost of drone as base station includes the construction cost of drone base station and the operation cost of base station. cons It can be written as:

[0089]

[0090] Among them, cons represents the stage of building a drone base station. is the construction cost of the drone base station m (such as take-off and landing costs, and energy consumption costs during movement).

[0091] After the drone base station enters the designated area, it will stabilize and start receiving and transmitting data. At this time, the operating cost is c oper It can be written as:

[0092]

[0093] Among them, oper represents the operation stage of the drone base station, t m is the time that the drone base station m stays in the air after it is built, δ m is the unit time operation cost of the drone base station m (including hovering energy consumption and communication energy consumption). The service cost of the drone base station c uav is the construction cost c cons and operating costs c oper The sum can be written as:

[0094]

[0095] c) IoT device data value: The multivariate information contained in the data collected by IoT devices determines the value of the data. Data with high value is more conducive to building digital twin services and is also an important basis for resource allocation in the system. In this invention, it is assumed that the data collected by IoT device i has a lot of semantic information, and the generated data value v i and the number of objects collected o i , weather conditions ω, object status κ, and also the relative distance ι. Data value v i It can be written as

[0096]

[0097] Among them, iRepresents the number of objects (such as pedestrians, vehicles, and trees) contained in the data collected by IoT device i. Indicates the deviation between the moving direction of object j and the preset reference direction. represents the moving speed of object j. Represents the area size of object j in the captured video image. j represents the weather state of the observed object j, ω j It is a real number greater than 1, and when the weather conditions are worse (such as rain, strong wind), its value is larger, and when the weather conditions are better (such as sunny days), its value is closer to 1. ι ij Represents the relative distance between object j and IoT device i. The smaller the value, the closer the distance between object j and IoT device i is, and vice versa. The closer the distance, the greater the data value brought by the object. This is because the closer the distance, the clearer the information such as the state of the object observed by the IoT device, which is more conducive to building a more realistic DT virtual environment.

[0098] 4) Data value obtained by the digital twin cloud platform: The data value obtained by the digital twin cloud platform is the winner set The sum of the data value of IoT devices in the digital twin cloud platform. The definition is as follows:

[0099]

[0100] in, Represents the winner set among IoT devices during resource allocation.

[0101] 5) Utility maximization problem of digital twin system based on network flow

[0102] In the model framework designed by the present invention, there are N sellers (IoT device users), M drone base stations responsible for data transmission and one buyer (digital twin cloud platform). The digital twin cloud platform provider is the data buyer, who selects users, purchases their data, and pays the winning user. Therefore, the digital twin cloud platform needs to select the winning user set in a reverse auction manner.

[0103] The utility of IoT devices i is the amount p paid by the digital twin cloud platform to the user corresponding to IoT device i i and the user's bid b i If user i does not win, the utility is 0, which is defined as follows:

[0104]

[0105] In the case of user i's true report b i and Equal. Among them, x i is a binary decision variable, and when its value is 1, it means that IoT device i wins (x i =1), otherwise x i =0.

[0106] The utility of digital twin cloud platform DTC It is the data value of the IoT devices in the winner set minus the amount paid by the digital twin cloud platform to the IoT devices minus the service cost of the drone base station, and is defined as follows:

[0107]

[0108] The social welfare of the entire model system, i.e., the total utility It is the sum of the utility of IoT devices and the utility of the digital twin cloud platform, and is defined as follows:

[0109]

[0110] The present invention transforms the communication problem between the digital twin cloud platform and the IoT devices into a network flow model (such as Figure 2 As shown in (19), it is used as a constraint condition of formula (19). First, all entities in the system (including IoT devices, drone base stations, and DTC) are represented as a set Knowable Second, define Represents the set of connections from IoT devices to drone base stations, such as the connection from IoT device i to drone base station m. Representation, definition Represents the set of connections from the drone base station to the DTC, such as the connection from the drone base station m to the DTC Indicates that each connection in the network has a channel bandwidth constraint, such as BW im or BW m,cloud At the same time, the drone base station and the digital twin cloud platform also have throughput constraints THP m and THP cloud The entire system communication model can be represented as a directed graph Define the decision variable x i Indicates whether IoT device i is assigned, defines the decision variable, represents the network flow decision variable, f im 、f m,cloud They represent the transmission rates from IoT devices to drone base stations and from drone base stations to DTC nodes respectively.

[0111] In order to express that the sum of the number of valid flows for any UAV base station cannot exceed the number of its channels, the present invention introduces the sign function auxiliary representation, which is defined as follows:

[0112]

[0113] The goal of the present invention is to solve the system utility maximization problem under the constraints of resources and communication conditions. According to the above conditions, the problem can be transformed into a mixed integer linear programming problem, which is defined as follows:

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124] x i ∈{0,1}(21j)

[0125] f im ≥0,f m,cloud ≥0(21k)

[0126] Among them, formula (21) represents the maximization of system utility, which is the data value of the IoT devices in the winner set minus the bid of the IoT devices in the winner set minus the service cost of the unmanned base station. The above mixed integer linear programming can enable the model system designed by the present invention to obtain the maximum social welfare. (21a) indicates that the resource demand of the IoT devices in the winner set is less than or equal to the resource capacity provided by the digital twin cloud platform. (21b) indicates that the network flow between IoT device i and the drone base station is less than the channel bandwidth between the two nodes. (21c) indicates that the throughput of the drone base station is greater than or equal to the sum of the network flows accepted by the base station. (21d) indicates that the network flow between the drone base station and the digital twin cloud platform is less than the channel bandwidth between the two. (21e) indicates that the sum of the flows transmitted by all drone base stations to the digital twin cloud platform is less than or equal to the throughput of the cloud. (21f) indicates that the data throughput of the digital twin cloud platform is greater than or equal to the sum of the requested bandwidths of the IoT devices in the winner set. (21g) indicates that for any drone base station, the size of its input network flow is equal to the size of its output network flow. (21h) For any IoT device i, the sum of the network flows it sends should be equal to its bandwidth requirement. (21i) indicates that the number of network diversions received by the drone base station must be less than or equal to the number of its channels. (21j) indicates that the decision variable of the user corresponding to the IoT device is an integer, and (21k) indicates that any network flow is greater than or equal to 0, which also shows that the model is a mixed integer programming model.

[0127] Figure 3 This is an example diagram of the digital twin system in this embodiment. Figure 3 As shown in the figure, there are 3 devices and 2 drone base stations in the digital twin system. The information submitted by IoT device 1 is θ1=(4,7,(8,5),15), which means that the bid of the user corresponding to IoT device 1 for its own data is 4, the bandwidth requirement is 7, the CPU and memory requirements are (8,5), and the data value is 15. At the same time, the channel capacity from the user corresponding to IoT device 1 to drone base station 1 and drone base station 2 is 6 and 4 respectively. The throughput of drone base station 1 is 10, and the channel capacity from drone base station 1 to DTC is 12. The data throughput of the digital twin cloud platform is 20, and the CPU and memory resource capacities are 15 and 10 respectively. The rest of the information is similar. It is worth noting that the channel capacity between IoT device 3 and drone base station 2 is 0, which means that device 3 cannot transmit data through drone base station 2.

[0128] Formula (21) is an ideal problem model. In theory, it can be solved by designing an optimal digital twin reverse auction mechanism based on the VGG mechanism, which is defined as follows:

[0129]

[0130] Among them, p i is the amount paid by the digital twin cloud platform to the user corresponding to IoT device i. It is the winner set of IoT devices when the social welfare of the system is maximized. The winner set is The utility achieved by the system. It is the winner set when the utility of the system is maximized without including IoT device i. The winner set is It is easy to know that

[0131] However, in this mechanism, solving the optimal allocation solution in the optimal resource allocation and pricing mechanism is NP-hard and cannot be solved in polynomial time. Therefore, the present invention designs a real reverse auction mechanism to solve the utility maximization of digital twin service providers and the resource allocation and pricing problems of digital twin cloud platforms.

[0132] After research, it is found that in practice, the earlier a device becomes a winner, the greater the value of its data should be. This is because in the initial stage of building a digital twin system, there is no data redundancy, and the data transmitted by the device has greater utility for the construction of the digital twin system. Defined as:

[0133]

[0134] in is the current set of winners.

[0135] because:

[0136]

[0137]

[0138] x i According to the collection Determine the binary decision variables, when Time i =1, otherwise x i =0;x i′ According to the collection Determine the binary decision variables, when Time i′ =1, otherwise x i′ = 0. Comparison set and collection It can be seen that only the binary decision variables of IoT device i are different, so the marginal utility value The calculation formula can be simplified as:

[0139]

[0140] That is to say, the resource allocation algorithm proposed in the present invention is based on the principle of monotone submodule and uses the marginal utility value as the basis for winner decision.

[0141] In addition, in order to determine whether the transmission capacity of the drone base station can meet the bandwidth requirements of the winner set when searching for the winner in the IoT device, the present invention also aims at Figure 2 The model shown is improved by adding the starting node and the target node, and converting each node with throughput constraints (such as the UAV base station and the digital twin cloud platform) into two nodes, such as the UAV base station in node (UAV-BS-in) and the UAV base station out node (UAV-BS-out), which are connected by an edge with bandwidth limited by THP, unifying the throughput constraint of the node with the bandwidth capacity constraint of the edge. Accordingly, the nodes and edges in the graph G0 are updated to obtain the graph G. Figure 4 It is a schematic diagram of the network traffic model of the improved digital twin system.

[0142] Based on the above analysis, the present invention proposes a digital twin resource allocation method based on the UAV network traffic model. Figure 5 is a flow chart of the digital twin resource allocation method based on the drone network traffic model in the present invention. Figure 5 As shown, the specific steps of the digital twin resource allocation method based on the drone network traffic model in the present invention include:

[0143] S501: Obtain digital twin system information:

[0144] Remember the IoT device collection in the digital twin system Drone base station collection

[0145] The data value v of each IoT device i is calculated using the following formula: i :

[0146]

[0147] Among them, i represents the number of objects contained in the data collected by IoT device i, Indicates the deviation between the moving direction of object j and the preset reference direction, represents the moving speed of object j, Represents the area of ​​object j in the captured video image, j = 1, 2, ..., o j ;ω j represents the weather state of the observed object j, ω jis a real number greater than 1, and when the weather conditions are worse, its value is larger, and when the weather conditions are better, its value is closer to 1; ι ij Represents the relative distance between object j and IoT device i. The smaller the value, the closer the distance between object j and IoT device i is, and vice versa.

[0148] The bandwidth requirement of IoT device i is bw i , the user corresponding to the IoT device i determines the resource demand s according to its actual situation i =(s i1 ,s i2 ,…,s iR ), s ir represents the demand of IoT device i for the rth resource, r = 1, 2, ..., R, and the bid of the user corresponding to IoT device i for its resource demand is b i , IoT device i reports parameter θ to the digital twin cloud platform i =(b i ,bw i ,s i ,v i );

[0149] Note the resource capacity of the digital twin cloud platform represents the capacity of the rth resource owned by the digital twin cloud platform;

[0150] S502: Constructing the network flow model of the digital twin system:

[0151] The digital twin cloud platform uses the following method to construct the network flow model of the digital twin system:

[0152] Two nodes are set for each drone base station, namely the drone base station ingress node and the drone base station egress node, and two nodes are set for the digital twin cloud platform, namely the digital twin cloud platform ingress node and the digital twin cloud platform egress node.

[0153] Set the starting point and the end point, together with the IoT device, the drone base station ingress node, the drone base station egress node, the digital twin cloud platform ingress node, and the digital twin cloud platform egress node as nodes to construct a directed graph G as the network flow model of the digital twin system, where:

[0154] The starting point is connected to each IoT device separately, and the capacity of the corresponding edge is the bandwidth demand bw of IoT device i i , the traffic is fixedly set to the bandwidth requirement bw of IoT device i i ;

[0155] Each IoT device is connected to each drone base station ingress node, and the capacity of the corresponding edge is the channel capacity BW between IoT device i and drone base station m im ;

[0156] Connect each drone base station ingress node m to the corresponding drone base station egress node m, and the capacity of the corresponding edge is the data throughput limit THP of the drone base station m. m ;

[0157] Connect each drone base station egress node m to the digital twin cloud platform ingress node, and the capacity of the corresponding edge is the channel capacity BW between the drone base station m and the digital twin cloud platform m,cloud ;

[0158] Connect the digital twin cloud platform ingress node to the digital twin cloud platform egress node. The capacity of the corresponding edge is the data throughput limit THP of the digital twin cloud platform. cloud ;

[0159] Connect the digital twin cloud platform outgoing node to the end point, and the capacity of the corresponding edge is a preset absolute maximum value.

[0160] Except for the edge between the starting point and each IoT device, the flow of the edge in the network flow model is the transmission rate between two nodes.

[0161] S503: Resource allocation:

[0162] Next, resources are allocated based on the network flow model of the digital twin system. The basic idea is to determine the IoT device i with the largest marginal utility value next time based on the current winner set, and merge the winner set and IoT device i into a candidate set. Then verify whether the resources of the digital twin cloud platform can meet the resource requirements of the devices in the candidate set, and use the maximum flow algorithm to verify whether the network flow exists. If the above conditions are met, the device is added to the winner set, otherwise it is added to the loser set. The resource allocation process will go through multiple rounds of iterations until the maximum marginal utility value is less than or equal to 0 or all devices have been processed (all added to the winner set or loser set). Figure 6 It is a flow chart of resource allocation in the present invention.

[0163] like Figure 6 As shown, the specific steps of resource allocation of the present invention include:

[0164] S601: Initialization data:

[0165] Initialize the winner set of IoT devices Loser Collection Remaining equipment collection Marginal utility value of each IoT device Total Utility The maximum flow identifier result=0.

[0166] S602: Screening of candidate IoT devices:

[0167] The marginal utility value of each IoT device i in the remaining device set u is calculated using the following formula:

[0168]

[0169] Then select the IoT device with the largest marginal utility value as the candidate IoT device i * .

[0170] S603: Determine whether the IoT device i * The marginal utility value of If yes, proceed to step S604, otherwise the resource allocation ends.

[0171] S604: Determine whether the current resource capacity of the digital twin cloud platform can meet the needs of IoT devices. * resource requirements, that is, to determine whether each resource meets r∈R, if satisfied, proceed to step S605, otherwise proceed to step S607.

[0172] S605: Solve the maximum flow of the network:

[0173] In the network flow model constructed in step S502, Delete all IoT devices except , solve the maximum flow of the obtained network flow model, and remember the maximum flow solution is The maximum flow identifier result is used to indicate whether there is a feasible flow solution. If so, the maximum flow identifier result=1, otherwise, the maximum flow identifier result=0.

[0174] In this embodiment, the Edmonds-Karp algorithm is used to solve the maximum flow solution. The Edmonds-Karp algorithm is a basic algorithm for solving the maximum flow of the network, and its specific process is not repeated here. Since in the present invention, the flow of the edge in the network flow model is the transmission rate between two nodes, when the maximum flow flag is 1, it indicates that the current winner set and IoT devices * The bandwidth requirements of IoT devices can be met. * You can be the winner, otherwise you cannot be the winner.

[0175] S606: Determine whether the maximum flow identifier result = 0, if yes, proceed to step S607, otherwise proceed to step S608.

[0176] S607: Add the IoT device to the loser set:

[0177] IoT devices * Add the loser set ε and delete it from the remaining device set u, and go to step S610.

[0178] S608: Adding the IoT device to the winner set:

[0179] IoT devices * Join the winners And delete it from the remaining device set u and go to step S609.

[0180] S609: Update the resource capacity of the digital twin cloud platform:

[0181] Use the following formula to update the resource capacity of the digital twin cloud platform:

[0182]

[0183] S610: Determine whether there are any remaining device sets If yes, the resource allocation ends, otherwise the process returns to step S602.

[0184] S504: Determine the payment price:

[0185] According to the resource allocation scheme obtained in step S503, the payment price of each user corresponding to the winning IoT device is solved.

[0186] The payment price calculation is to calculate the fee that each user needs to pay based on the resource allocation result. The optimal payment algorithm can be designed based on the VCG (Vickrey-Clark-Groves) mechanism. The core of the VCG mechanism is that the price that the user corresponding to the IoT device i finally needs to pay has nothing to do with his own bid, which eliminates the user's attempt to use false quotations to make improper profits. The VCG mechanism has been proven to meet the credibility under the premise of optimal allocation, but its calculation complexity is high and the calculation time is unpredictable, so it cannot be used in practice. Therefore, this embodiment proposes a pricing algorithm based on dichotomy to calculate the payment price of each user corresponding to the IoT device.

[0187] Figure 7 is a flowchart of the pricing algorithm based on the dichotomy method in this embodiment. Figure 7 As shown, the specific steps of the pricing algorithm based on the dichotomy method in the present invention include:

[0188] S701: Initialize payment price:

[0189] For the winners set IoT devices in Initialize the lower limit of the payment price for the corresponding user Paying a price cap Bid Cache Set the loop exit value σ, in this embodiment σ=0.1,

[0190] S702: Order to pay the price cap

[0191] S703: Update bid

[0192] S704: Reallocate resources:

[0193] Other parameters remain unchanged, based on IoT devices Current bid price Re-adopt the method in step S503 to allocate resources and obtain the winner set

[0194] S705: Determine whether it is an IoT device If yes, return to step S702, otherwise go to step S706.

[0195] S706: Resume bidding

[0196] S707: Order to pay the price floor

[0197] S708: Update bid

[0198] S709: Determine whether If yes, go to step S710, otherwise go to step S715.

[0199] S710: reallocate resources:

[0200] Other parameters remain unchanged, based on IoT devices Current bid price Re-adopt the method in step S503 to allocate resources and obtain the winner set

[0201] S711: Determine whether it is an IoT device If yes, return to step S707, otherwise go to step S712.

[0202] S712: Order to pay the price cap

[0203] S713: Update bid

[0204] S714: Reallocate resources:

[0205] Other parameters remain unchanged, based on IoT devices Current bid price Re-adopt the method in step S503 to allocate resources and obtain the winner set Return to step S711.

[0206] S715: Determine the payment price:

[0207] The current bid As a digital twin cloud platform for IoT devices The price paid.

[0208] According to the above method, when determining the payment price, the present embodiment first searches for the IoT device Become the loser's payment price ceiling, and then in the IoT device If the winner is still the winner, the lower limit of the payment price and the bid are updated; if the bid is changed and the loser is changed, the upper limit of the payment price and the bid are updated. When the upper and lower limits of the payment price are the same, the payment price is obtained.

[0209] Example

[0210] In order to better illustrate the implementation process of the present invention, Figure 3 The digital twin system shown in FIG. 1 is an example of a resource allocation process in this embodiment.

[0211]

[0212] Table 1

[0213] As shown in Table 1, in the resource allocation process, we first need to calculate the marginal utility values ​​of all IoT devices. Taking IoT device 1 as an example, when the winner set is empty, its marginal utility value is The calculation process of the marginal utility value of IoT devices 2 and 3 is similar. Therefore, IoT device 1 is selected as the candidate IoT device in the first round. Next, it needs to be verified whether it can become the winner: At this time, the total remaining resources (15,10) of the digital twin cloud platform can meet the resource request (8,5) of IoT device 1. The bandwidth requirement of IoT device 1 is bw1=7. After calling the maximum flow algorithm, the traffic f diverted to drone base station 1 11 =6, the traffic diverted to drone base station 2 f 12 =1. That is, IoT device 1 can transmit data to the digital twin cloud platform, so IoT device 1 joins the winner set. Then the resource capacity of the digital twin cloud platform is updated to (7,5).

[0214] In the second round, IoT device 2, which has the largest marginal utility value at this time, is selected as the candidate IoT device. The remaining resources of DTC (7,5) cannot meet the resource requirements of IoT device 2 (4,6), so IoT device 2 is added to the loser set.

[0215] In the third round, the only remaining IoT device 3 is selected as the candidate IoT device. The total remaining resources of DTC (7, 5) can satisfy the resource request (7, 5) of IoT device 3. The bandwidths requested by IoT devices 1 and 3 are bw1=7 and bw3=6 respectively, of which the bandwidth diverted from IoT device 1 to base station 1f is 11 =4, diverted to base station 2f 12 =3, traffic is diverted from IoT device 3 to base station 1f 31 = 6. IoT devices 1 and 3 can transmit data to DTC, so device 3 is also included in the winner set.

[0216] According to the above process, the IoT devices in the winner set are device 1 and device 3. Using the binary pricing algorithm in this embodiment, the service provider of the digital twin cloud platform pays the users corresponding to IoT devices 1, 2, and 3 with amounts p1=5, p2=0, and p3=13, respectively.

[0217] Although the above describes the illustrative specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations using the concept of the present invention are protected.

Claims

1. A digital twin resource allocation method based on drone network flow model, characterized in that: The following steps are involved: S1: Record the collection of IoT devices in the digital twin system Drone base station collection The data value v of each IoT device i is calculated using the following formula: i : Among them, i represents the number of objects contained in the data collected by IoT device i, Indicates the deviation between the moving direction of object j and the preset reference direction, represents the moving speed of object j, Represents the area of ​​object j in the captured video image, j = 1, 2, ..., o j ;ω j represents the weather state of the observed object j, ω j is a real number greater than 1, and when the weather conditions are worse, its value is larger, and when the weather conditions are better, its value is closer to 1; ι ij Represents the relative distance between object j and IoT device i. The smaller the value, the closer the distance between object j and IoT device i is, and vice versa. The bandwidth requirement of IoT device i is bw i , the user corresponding to the IoT device i determines the resource demand s according to its actual situation i =(s i1 ,s i2 ,…,s iR ), s ir represents the demand of IoT device i for the rth resource, r = 1, 2, ..., R, and the bid of the user corresponding to IoT device i for its resource demand is b i , IoT device i reports parameter θ to the digital twin cloud platform i =(b i ,bw i ,s i ,v i ); Note the resource capacity of the digital twin cloud platform represents the capacity of the rth resource owned by the digital twin cloud platform; S2: The digital twin cloud platform uses the following method to construct the network flow model of the digital twin system: Set up 2 nodes for each drone base station, namely the drone base station ingress node and the drone base station egress node, and set up 2 nodes for the digital twin cloud platform, namely the digital twin cloud platform ingress node and the digital twin cloud platform egress node; Set the starting point and the end point, together with the IoT device, the drone base station ingress node, the drone base station egress node, the digital twin cloud platform ingress node, and the digital twin cloud platform egress node as nodes to construct a directed graph G as the network flow model of the digital twin system, where: The starting point is connected to each IoT device separately, and the capacity of the corresponding edge is the bandwidth demand bw of IoT device i i , the traffic is fixedly set to the bandwidth requirement bw of IoT device i i ; Each IoT device is connected to each drone base station ingress node, and the capacity of the corresponding edge is the channel capacity BW between IoT device i and drone base station m im ; Connect the ingress node of each drone base station m to the egress node of the corresponding drone base station m. The capacity of the corresponding edge is the data throughput limit THP of the drone base station m. m ; Connect the outgoing node of each drone base station m to the ingoing node of the digital twin cloud platform. The capacity of the corresponding edge is the channel capacity BW between the drone base station m and the digital twin cloud platform. m,cloud ; Connect the digital twin cloud platform ingress node to the digital twin cloud platform egress node. The capacity of the corresponding edge is the data throughput limit THP of the digital twin cloud platform. cloud ; Connect the digital twin cloud platform egress node to the endpoint, and the capacity of the corresponding edge is a preset absolute maximum value; Except for the edge between the starting point and each IoT device, the flow of the edge in the network flow model is the transmission rate between two nodes; S3: Resource allocation based on the network flow model of the digital twin system. The specific method is as follows: S3.1: Initialize the winner set of IoT devices Loser Collection Remaining equipment collection Marginal utility value of each IoT device Total Utility Maximum flow identifier result = 0; S3.2: The marginal utility value of each IoT device i in the remaining device set u is calculated using the following formula: Then select the IoT device with the largest marginal utility value as the candidate IoT device i * ; S3.3: Determine whether the IoT device i * The marginal utility value of If yes, go to step S3.4, otherwise the resource allocation ends; S3.4: Determine whether the current resource capacity of the digital twin cloud platform can meet the needs of IoT devices * resource requirements, that is, to determine whether each resource meets r∈R, if satisfied, proceed to step S3.5, otherwise proceed to step S3.7; S3.5: In the network flow model constructed in step S2, Delete all IoT devices except , solve the maximum flow of the obtained network flow model, and remember the maximum flow solution is The maximum flow identifier result is used to indicate whether there is a feasible flow solution. If so, the maximum flow identifier result = 1, otherwise, the maximum flow identifier result = 0; S3.6: Determine whether the maximum flow identifier result = 0, if yes, proceed to step S3.7, otherwise proceed to step S3.8; S3.7: Connect IoT devices * Add the loser set ε and delete it from the remaining device set u, and go to step S3.10; S3.8: Connect IoT devices * Join the winners And delete it from the remaining device set u, and go to step S3.9; S3.9: Update the resource capacity of the digital twin cloud platform using the following formula S3.10: Determine whether there are any remaining device sets If yes, the resource allocation ends, otherwise return to step S3.2; S4: According to the resource allocation scheme obtained in step S3, the payment price of each winning IoT device corresponding to the user is solved. The specific method is: S4.1: For the winner set IoT devices in Initialize the lower limit of the payment price for the corresponding user Paying a price cap Bid Cache Set loop exit value σ; S4.2: Order to pay a price cap S4.3: Update Bid S4.4: Other parameters remain unchanged, based on IoT devices Current bid price Re-adopt the method in step S3 to allocate resources and obtain the winner set S4.5: Determine whether it is an IoT device If yes, return to step S4.2, otherwise go to step S4.6; S4.6: Resume bidding S4.7: Order to pay a price floor S4.8: Update Bid S4.9: Determine whether If yes, go to step S4.10, otherwise go to step S4.15; S4.10: Other parameters remain unchanged, based on IoT devices Current bid price Re-adopt the method in step S3 to allocate resources and obtain the winner set S4.11: Determine whether it is an IoT device If yes, return to step S4.7, otherwise go to step S4.12; S4.12: Order to pay a price cap S4.13: Update Bid S4.14: Reallocate resources: Other parameters remain unchanged, based on IoT devices Current bid price Re-adopt the method in step S3 to allocate resources and obtain the winner set Return to step S4.11; S4.15: Current bid As a digital twin cloud platform for IoT devices The price paid.

2. The digital twin resource allocation method according to claim 1, characterized in that: In step S3.5, the Edmonds-Karp algorithm is used to solve the maximum flow solution.

Images