Distributed resource auction method for mobile group computing

Abstract

The invention belongs to the technical field of Internet of Things. The invention provides a distributed resource auction method for mobile group computing. According to the embodiment of the invention, an ICV network system model supporting a mobile group calculation service is constructed, road infrastructure-assisted mobile group calculation is realized, and a resource auction problem is expressed as a non-cooperative game of incomplete information among participants, so that a user utility maximization problem model is constructed; and solving the user utility maximization problem model based on sine disturbance in combination with reputation and distance factors so as to obtain an optimal transaction price and an optimal transaction quantity. According to the method, the resource auction problem that in a high-dynamic and distributed environment, self-benefited participants are motivated to faithfully disclose information, so that the utility of users is maximized is solved.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the technical field of Internet of Things, and in particular to a distributed resource auction method for mobile group computing. Background Art

[0002] Intelligent connected vehicle (ICV) networks, by combining sensing technologies, information processing, and communications with artificial intelligence (AI), promise to significantly improve driving and road safety. To support computationally intensive applications such as augmented reality, autonomous driving, and natural language processing, ICV networks require robust computing power to effectively assist drivers and passengers. However, with the rapid adoption of these applications, local processing of these tasks within vehicles presents significant challenges. Mobile edge computing (MEC) has emerged as an effective solution to meet the low-latency requirements of real-time services like autonomous driving. MEC offloads computing tasks to edge servers closer to the vehicle, providing more efficient network services. While MEC effectively reduces latency, it also presents challenges in practical applications. First, increasing edge server density increases installation costs, and base station coverage is difficult to determine. Second, due to the varying resource requirements of mobile users, some users may face resource shortages while others may have idle computing resources. This imbalance impacts system efficiency and service quality.

[0003] To address these challenges, mobile crowdsourcing (MCS) has been proposed as an emerging distributed computing model that fully leverages the sensing, communication, and computing capabilities of widely available mobile devices. Building on the concept of MCS, mobile crowd computing (MCC) has been proposed, leveraging the idle computing resources of mobile devices to compute latency-sensitive tasks. Through MCC, end users can outsource computing tasks to nearby mobile devices, which then complete the computation and feedback the results via wireless communication links. Mobile devices are often surrounded by other mobile devices, enabling them to collaboratively process computationally intensive tasks. Supported by a work-sharing framework, these devices can communicate with each other and collaborate on computations. Therefore, in mobile crowd computing environments, using private mobile devices as resource providers has become a viable development direction. Compared to base station deployment, some private devices offer greater flexibility and can effectively overcome the limitations of traditional approaches. Driven by this trend, sharing resources among mobile vehicles (MVs) can be considered to achieve local optimization and resource allocation, thereby improving system efficiency and reliability.

[0004] Specifically, consider integrating mobile devices with idle computing resources, such as phones and vehicles, with road infrastructure to form a ubiquitous network. Road infrastructure can serve as auxiliary nodes for edge computing, helping mobile devices achieve broader communication coverage. When computing resources are limited, mobile devices can offload computationally intensive tasks to nearby devices with idle computing resources, temporarily expanding their resources and improving computing efficiency and resource utilization.

[0005] However, effectively allocating computing tasks and resources across multiple mobile devices is a critical issue. Existing resource allocation methods, most of which rely on centralized control systems, face computational bottlenecks and network overload. Furthermore, these methods typically assume static resource requirements and are unable to cope with dynamic changes in tasks and resources.

[0006] Therefore, it is necessary to improve one or more problems existing in the above-mentioned related technical solutions.

[0007] It should be noted that this section is intended to provide background or context for the technical solutions of the present disclosure stated in the claims. The description herein is not admitted to be prior art by virtue of being included in this section. Summary of the Invention

[0008] The purpose of the embodiments of the present disclosure is to provide a distributed resource auction method for mobile group computing, thereby overcoming one or more problems caused by the limitations and defects of related technologies, at least to a certain extent.

[0009] According to an embodiment of the present disclosure, a distributed resource auction method for mobile group computing is provided, the method comprising: Step S1, constructing an ICV network system model supporting mobile group computing services; wherein the ICV network system model includes a network model, a task model, and a distributed auction model; The network model includes multiple task owner vehicles and multiple task vehicles that can share computing resources. The task owner vehicle is the buyer. , the task can share computing resources for the vehicle seller , the buyer and seller are connected to the RSU through V2I communication; In the task model, the seller The computing resources provided are , among which is the amount of computing resources that the seller can provide, is the maximum acceptable energy consumption, is the spatial coordinate of the seller; the buyer The computing resources required are ;in Input data size, is the amount of computing resources required for the task, is the maximum acceptable delay, is the spatial location coordinate of the buyer; In the distributed auction model, participating users are divided into two groups: the seller group , Buyer Group At the beginning of a trading round, the seller Submit Supply Vector ,in, Represents the seller's initial asking price, which is the lowest price the seller is willing to accept. Indicates the amount of idle computing resources; buyer Submit a demand vector ,in, Represents the buyer's initial bid, that is, the highest price the buyer is willing to pay. Indicates the amount of computing resources required to perform the task; Based on the quadratic utility function, the buyer's utility function and the seller's utility function are obtained; Step S2: Based on the ICV network system model, a distributed resource auction process is designed to construct a user utility maximization problem model; In step S3, the user utility maximization problem model is solved based on sinusoidal perturbations by combining reputation and distance factors to obtain the optimal transaction price and optimal transaction quantity to complete the distributed resource auction.

[0010] Furthermore, the distributed resource auction process is as follows: At the beginning of a trading round, the buyer and seller submit their demand vectors and the supply vector , and submit their respective bidding information, and determine the market type based on the surplus or deficit of total resources; market types include buyer's market and seller's market; If it is a buyer's market, calculate the first reputation-distance price of each seller based on the seller's bidding information, and sort the seller's first reputation-distance price; Calculate the buyer's reputation score based on the seller's historical actual transaction amount and historical agreed transaction amount in the bid information; Calculate the distance score between each buyer and each seller based on the location information in the buyer's and seller's bidding information; Calculate the seller's reputation-distance score based on the seller's reputation score and distance score; According to the initial asking price in the seller's bidding information and the seller's reputation-distance score, the first reputation-distance price of each seller is obtained, and the seller's first reputation-distance price is sorted by price; For the seller ranked first, calculate the first reputation-distance price of all corresponding buyers, and sort the first reputation-distance prices of all buyers to determine the sorting of transaction quantity.

[0011] Furthermore, if it is a seller's market, the second reputation-distance price of each buyer is calculated based on the buyer's bidding information, and the second reputation-distance prices of the buyers are sorted by price; Calculate the buyer's reputation score based on the historical actual transaction amount and historical agreed transaction amount in the buyer's bidding information; Calculate the distance score between each buyer and each seller based on the location information in the buyer's and seller's bidding information; Calculate the buyer's reputation-distance score based on the buyer's reputation score and distance score; According to the initial bid in the buyer's bidding information and the buyer's reputation-distance score, the second reputation-distance price of each buyer is obtained, and the second reputation-distance prices of the buyers are sorted by price; For the second-ranked buyer, calculate the second reputation-distance prices of all corresponding sellers, and sort the second reputation-distance prices of all sellers to determine the ranking of transaction quantity.

[0012] Furthermore, in a seller's market, the seller's reputation score is expressed as:

[0013] in, is the time discount factor, and , is the seller’s historical actual transaction amount, is the seller’s historical actual transaction amount, is the fault tolerance factor, The seller's historical reputation score; The distance score between each buyer and each seller is expressed as:

[0014]

[0015] in, For the seller and the buyer The Euclidean distance between and are the maximum and minimum distances among all distances respectively; The buyer's reputation-distance score is expressed as:

[0016] in, is the weight coefficient; The expression of the seller's first reputation-distance price is:

[0017] in, The buyer's initial bid.

[0018] Furthermore, the seller's utility function is expressed as:

[0019] in, Indicates the seller i Select a strategy When the strategy set of other participants is For the seller i and the buyer j The transaction price between For the seller i and the buyer j The number of transactions between is the seller’s QoS coefficient, is the seller's network and communication cost coefficient, is the seller’s initial asking price; The buyer's utility function is expressed as:

[0020] in, Indicates the buyer i Select a strategy When the strategy set of other participants is is the buyer’s QoS coefficient, is the buyer's network and communication cost coefficient, The buyer's initial bid.

[0021] Furthermore, by combining reputation and distance factors, the user utility maximization problem model is solved based on sinusoidal perturbations to obtain the optimal transaction price and optimal transaction quantity, thereby completing the steps of the distributed resource auction, including: In a seller's market, the number of transactions is determined based on the seller's available capacity and the buyer's task requirements; Based on the number of transactions and the order of the transaction numbers, the seller's utility function is sinusoidally perturbed; Perform Taylor expansion and gradient estimation on the utility function after sinusoidal perturbation to obtain the estimated gradient; The seller updates the strategy price based on the estimated gradient to bring it close to the Nash equilibrium point to obtain the final transaction price.

[0022] Furthermore, in a seller's market, the utility function after sinusoidal disturbance is:

[0023] in, For users i The strategy price at time t, is the disturbance amplitude, is the disturbance frequency, is the phase, is the first final transaction price; The Taylor expansion of the utility function after sinusoidal disturbance is:

[0024] The estimated gradient is:

[0025] in, is the utility function For user i's own policy variables gradient; The user adjusts their actions according to the estimated gradient to maximize utility:

[0026] in, is the updated strategy price, is the learning rate, which is used to control the amplitude of the policy update.

[0027] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects: In the disclosed embodiments, the distributed resource auction method for mobile group computing described above constructs, on the one hand, an ICV network system model supporting mobile group computing services, including road infrastructure-assisted mobile group computing. The resource auction problem is formulated as a non-cooperative game with incomplete information between participants to construct a user utility maximization model. This user utility maximization model is solved based on sinusoidal perturbations, incorporating reputation and distance factors, to obtain the optimal transaction price and optimal transaction quantity. Furthermore, this method addresses the resource auction problem of maximizing user utility by incentivizing self-interested participants to disclose information truthfully in a highly dynamic and distributed environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0029] Figure 1 A diagram showing the steps of a distributed resource auction method for mobile group computing in an exemplary embodiment of the present disclosure; Figure 2 A scene diagram of the present application in an exemplary embodiment of the present disclosure is shown; Figure 3 A flow chart showing a distributed auction resource method in an exemplary embodiment of the present disclosure; Figure 4 A framework diagram illustrating a distributed resource auction method in an exemplary embodiment of the present disclosure is shown; Figure 5 A schematic diagram illustrating a distributed resource auction method in an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0030] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0031] In addition, the accompanying drawings are merely schematic illustrations of embodiments of the present disclosure and are not necessarily drawn to scale. Like reference numerals in the figures represent like or similar parts, and thus repeated descriptions thereof will be omitted. Some of the blocks shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically separate entities.

[0032] This example embodiment provides a distributed resource auction method for mobile group computing. Figure 1 As shown in , the distributed resource auction method for mobile group computing may include: Step S1, constructing an ICV network system model supporting mobile group computing services; wherein the ICV network system model includes a network model, a task model, and a distributed auction model; The network model includes multiple task owner vehicles and multiple task vehicles that can share computing resources. The task owner vehicle is the buyer. , the task can share computing resources for the vehicle seller , the buyer and seller are connected to the RSU through V2I communication; In the task model, the seller The computing resources provided are , among which is the amount of computing resources that the seller can provide, is the maximum acceptable energy consumption, is the spatial coordinate of the seller; the buyer The computing resources required are ;in Input data size, is the amount of computing resources required for the task, is the maximum acceptable delay, is the spatial location coordinate of the buyer; In the distributed auction model, participating users are divided into two groups: the seller group , Buyer Group At the beginning of a trading round, the seller Submit Supply Vector ,in, Represents the seller's initial asking price, which is the lowest price the seller is willing to accept. Indicates the amount of idle computing resources; buyer Submit a demand vector ,in, Represents the buyer's initial bid, that is, the highest price the buyer is willing to pay. Indicates the amount of computing resources required to perform the task; Based on the quadratic utility function, the buyer's utility function and the seller's utility function are obtained; Step S2: Based on the ICV network system model, a distributed resource auction process is designed to construct a user utility maximization problem model; In step S3, the user utility maximization problem model is solved based on sinusoidal perturbations by combining reputation and distance factors to obtain the optimal transaction price and optimal transaction quantity to complete the distributed resource auction.

[0033] This distributed resource auction method for mobile swarm computing constructs an ICV network system model supporting mobile swarm computing services, featuring road infrastructure-assisted mobile swarm computing. The resource auction problem is formulated as a non-cooperative game with incomplete information between participants, modeling the user utility maximization problem. This user utility maximization model is solved based on sinusoidal perturbations, incorporating reputation and distance factors, to obtain the optimal transaction price and quantity. Furthermore, this method addresses the resource auction problem of maximizing user utility by incentivizing self-interested participants to disclose information truthfully in a highly dynamic and distributed environment.

[0034] Below, we will refer to Figures 1 to 5 Each step of the above-mentioned distributed resource auction method for mobile group computing in this example implementation is described in more detail.

[0035] In step S101, an ICV network system model supporting mobile group computing services is constructed; wherein the ICV network system model includes a network model, a task model and a distributed auction model.

[0036] Specifically, the network model: like Figure 2 As shown, the scenario diagram of the present application. To consider an ICV (Intelligently Connected Vehicle) network that supports mobile group computing services, all vehicles are equipped with wireless transceivers and can communicate with road infrastructure (RSU, BS, ES, etc.) for V2I (Vehicle to road Infrastructures). Assume that the vehicle's on-board computing resources are limited, and it is hoped to offload its computing-intensive tasks to surrounding vehicles with idle computing resources. Therefore, the road infrastructure will assist the vehicle in task allocation. For the convenience of description, the road infrastructure is referred to as RSU. One benefit of RSU-assisted communication is that the coverage of RSU is much larger than that of vehicle-to-vehicle (V2V) communication, so all vehicles within the coverage of RSU can be mobilized to share computing resources. Its function is limited to assisting communication and collecting a small amount of global control information or data, and does not participate in the specific calculation of the task. The task owner is called the buyer ( ), these vehicles cannot provide services for themselves or others, and vehicles that can share computing resources are called sellers ( ), the role of the vehicle is switchable, depending on its available capacity and mission requirements.

[0037] Mission Model: The V2I2V resource sharing task service process mainly includes two parts: task data transmission and task calculation. The computing resources provided are represented as ,in is the amount of computing resources (i.e., number of CPU cycles) that the user can provide, is the maximum acceptable energy consumption, is the spatial coordinate of the seller's vehicle; the buyer The required computing resources are expressed as ,in Input data size (unit: bits), is the amount of computing resources required for the task, is the maximum acceptable delay, are the spatial position coordinates of the buyer’s vehicle. The computing resources can be shared not only with the most recently associated RSU , can also be distributed to a wider distance through transmission between RSUs .

[0038] set up is a binary transaction decision variable. If the task of user j is offloaded to user i in a transaction round, ,otherwise In order to facilitate marking, the vehicle connection diagram is introduced , where S is the vertex set of all moving vehicles, and is an edge set. If user i can establish a feasible V2I2V link with user j, Then, in a transaction round, there are the following task matching constraints:

[0039] Since the data rates in communication are not equal, the transmission time delay is considered. For simplicity, it is assumed that users are allocated using orthogonal sub-channels (OFDMA). According to the Shannon formula, the achievable transmission rate of a user can be expressed as ,in, is the transmission power of user j, is the bandwidth of user j, is the Gaussian noise power, is the average channel gain. Since signal transmission between RSUs is optical fiber-based and very fast, it can be ignored. Note that the output size is always negligible compared to the input size, so the output latency is neglected here. Also, the energy consumption of the RSU is negligible.

[0040] At the beginning of a trading round, participating users are divided into two groups: the seller group , Buyer Group . To get data from Send to At this point, RSU forwarding is required, which means it will go through three steps: V2I, I2I, and I2V. The delay and energy consumption during this period are as follows: For resource buyers, the task Total service delay The following constraints should be met:

[0041] in, Sending delay , Receiving delay , calculate the delay , Assigned to computing tasks CPU frequency.

[0042] In order to ensure QoS requirements, the maximum energy consumption of idle devices needs to be considered. For resource sellers, the energy consumption constraint of resource sharing yes:

[0043] in, Receiving energy consumption , the energy cost is , is the energy cost per CPU cycle calculated at the current processor frequency.

[0044] Distributed auction model: The interaction between vehicles can be modeled as a double auction with the following participants: 1. (Seller): When a mobile vehicle acts as a seller, it provides its computing resources to facilitate other vehicles to complete computing tasks and is rewarded for doing so. Submit Supply Vector ,in, Represents the seller's initial asking price, which is the lowest price the seller is willing to accept. Indicates the amount of idle computing resources.

[0045] 2. (Buyer): When a mobile vehicle acts as a buyer, it pays a fee to purchase computing resources to facilitate the completion of its computing tasks. In order to participate in the auction, the buyer Submit a demand vector ,in, Represents the buyer's initial bid, which is also the highest price the buyer is willing to pay. Indicates the amount of computing resources required to execute the task.

[0046] 3. Auctioneer: In a buyer's market (i.e., a market with oversupply), sellers submit bids to compete for buyer task requirements. Each buyer acts as an auctioneer, making their own auction decisions locally. Due to intense competition among sellers, buyers are able to acquire resources at favorable prices, and the final transaction price is determined by the competition among sellers. Sellers' markets follow similar rules.

[0047] In auction theory, utility is an economic indicator that measures the effectiveness of a system. Due to the diminishing marginal utility hypothesis, that is, when consumers obtain more and more of a certain commodity, the increase in utility brought by an additional unit of the commodity is decreasing, so the utility of the participants is not linearly increasing or decreasing, but a quadratic function of the price. In addition, the quadratic utility function form is not only convenient for mathematical derivation, but also ensures the convexity of the optimization problem, thereby ensuring that the algorithm can effectively converge to the optimal solution. In large-scale resource auction scenarios, the use of quadratic functions can reduce computational complexity and ensure the stability and efficiency of the utility maximization process. Consider a game with a general quadratic utility function. Specifically, assume that the user's strategy is any option from the strategy set S, and the seller i has a strategy The utility of , buyer j in strategy The utility of , define the utility function:

[0048]

[0049] in, represents the strategy chosen by seller i When the strategy set of other participants is and are the transaction price and transaction quantity between seller i and buyer j respectively; is a coefficient related to QoS, is a coefficient related to network and communication costs; the same applies to buyer j.

[0050] In steps S2 and S3, based on the ICV network system model, a distributed resource auction process is designed to construct a user utility maximization problem model; combined with reputation and distance factors, the user utility maximization problem model is solved based on sinusoidal perturbations to obtain the optimal transaction price and optimal transaction quantity to complete the distributed resource auction.

[0051] Specifically, (1) Submitting a bid: When the submission window opens, users register to the market as sellers or buyers and submit their bids. The submitted bids contain the following information: user ID, reservation price, supply / demand quantity, and resource exchange time. The information can be displayed as a tuple: the seller is , the buyer is . Determine the current market type (buyer's market / seller's market) based on the total resource surplus or deficit.

[0052] (2) Bid sorting: 1) Obtain the auctioneer’s reputation and distance score. 2) Combine these two scores into a reputation-distance index. 3) Obtain a reputation-distance-based price (i.e., sorting price) based on the strategy price and the reputation-distance index. 4) Use the sorting price to sort the orders.

[0053] (3) Iterative Update: Buyers and sellers have the opportunity to adjust their strategies based on this information and any other available information and resubmit new bids to maximize their own utility. Through multiple iterations, the strategies of the participants are gradually adjusted and the ranking prices are continuously updated. This process will be repeated until no buyers or sellers are willing to adjust their bids, and eventually converge to a Nash equilibrium point.

[0054] (4) Winner Determination: A Nash equilibrium search algorithm based on reputation-distance is proposed. The final transaction price and final transaction quantity are determined in the convergence state. In a buyer's market (i.e., a market with oversupply), sellers face fierce competition and buyers' needs are met, which means that buyers can purchase resources at their preferred price. Therefore, in this case, the final transaction price is determined by the competition among sellers, and the seller's market also follows similar rules.

[0055] (5) Execution Record: Users trade resources based on transaction results and provide feedback on transaction performance to the platform to update their reputation scores. Sellers and buyers who expect to be matched must fulfill their commitments to the contract, otherwise their reputation levels will deteriorate, putting them at a disadvantage in the next auction.

[0056] For the sake of convenience, the following is explained from the perspective of a buyer's market. The same applies to a seller's market.

[0057] The interaction between resource buyers and sellers, where bids are repeatedly revised, can be described as a dynamic, non-cooperative game under incomplete information. During the considered time period, both the computing service request and the seller's computing server are active. The process of solving a Nash equilibrium is essentially a game. By analyzing the strategies and utilities of the players, the optimal response strategy for each player is sought, ultimately determining a stable strategy combination. The objective function is as follows:

[0058] This objective function quantifies the utility of each participant under different strategy combinations and, through iterative calculation, finds the Nash equilibrium point. This equilibrium point represents the point where, given information and strategy space, all participants have no intention of unilaterally changing their strategies, thus achieving the optimal allocation between resource buyers and sellers.

[0059] When a Nash equilibrium exists, the following conditions must be met:

[0060] in, is the optimal strategy for user i.

[0061] Once a Nash equilibrium is reached, the optimal trading parameters (i.e., clearing quantity and price) are established and the market closes.

[0062]

[0063]

[0064] Before conducting the final search, it is necessary to build a reputation-distance index model. This process has four parts: 1. Calculate a reputation score to evaluate the user’s current and past trading performance:

[0065] in, is a time discount factor that measures the importance of real-time behavior and historical behavior. is the actual transaction amount, It is the agreed transaction amount.

[0066] 2. Calculate the distance score, which is used to evaluate the communication link quality and signal transmission delay between vehicles, and is obtained by calculating the straight-line distance and The Euclidean distance between And normalize:

[0067]

[0068] 3. Calculate reputation-distance score:

[0069] 4. Price sorting: When user i is a buyer, that is, a seller's market, the buying sorting price is:

[0070] When user i is a seller, that is, a buyer's market, the selling sorting price is:

[0071] Furthermore, based on the sinusoidal perturbation, the user utility maximization problem model is solved, and the optimal transaction price and optimal transaction quantity are obtained as follows: (1) Sinusoidal disturbance Each user adds a small sinusoidal perturbation to their actions, allowing the system to explore new policy space.

[0072]

[0073]

[0074] in, is the strategic price of user i at time t, is the disturbance amplitude, is the perturbation frequency, is the phase. At this time, the user's utility function is:

[0075] (2) Taylor expansion Assuming the disturbance amplitude is small, so the utility function Perform Taylor expansion:

[0076] because is a periodic signal, and its integrated average value is 0, so:

[0077] (3) Gradient estimation To estimate the gradient of the utility function, players use the changes in utility values over a period of time to calculate. Assuming the perturbation period is T, the gradient estimation formula is:

[0078] Substituting Taylor expansion into the gradient estimation formula, we get:

[0079] Split the above integral into two parts:

[0080] because is a constant, which is The correlation integral of is zero:

[0081] The rest is:

[0082] because The integral average of is 1 / 2, and we get: .

[0083] (4) Action update The user adjusts their actions according to the estimated gradient to maximize utility:

[0084] In this way, the user's actions gradually approach the Nash equilibrium point.

[0085] Before moving on to the termination phase, let's briefly review the extremum search mechanism used by each player. Each player's game utility is a quadratic function of the price of the strategy they choose. Somewhere in the middle of this utility function lies an optimal strategy that maximizes utility. Players can start to the left or right of this optimal point, and sinusoidal perturbations become crucial here. By applying these perturbations, players can observe changes in their returns from the market.

[0086] For a player starting to the left of the optimal point, this sinusoidal perturbation will generate a corresponding sinusoidal change in utility with the same phase as the original perturbation. Therefore, when multiplied together, they generate a positive correction that, when added to the strategy from the previous iteration, causes the process to converge toward the optimal point. Conversely, if the player starts to the right of the optimal point, a negative correction will be generated, which will be added to the strategy from the previous iteration. Once the process converges to the optimal point, it will remain within a narrow band around that point, as very small negative corrections are generated.

[0087] like Figure 3 , which is a flow chart of a distributed auction resource method; like Figure 4 As shown, it is a framework diagram of the distributed resource auction method; like Figure 5 The figure shows the principle diagram of the distributed resource auction method.

[0088] This distributed resource auction method for mobile swarm computing constructs an ICV network system model supporting mobile swarm computing services, featuring road infrastructure-assisted mobile swarm computing. The resource auction problem is formulated as a non-cooperative game with incomplete information between participants, modeling the user utility maximization problem. This user utility maximization model is solved based on sinusoidal perturbations, incorporating reputation and distance factors, to obtain the optimal transaction price and quantity. Furthermore, this method addresses the resource auction problem of maximizing user utility by incentivizing self-interested participants to disclose information truthfully in a highly dynamic and distributed environment.

[0089] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly indicate the number of the technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0090] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification.

[0091] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the appended claims.

Claims

1. A distributed resource auction method for mobile group computing, characterized in that: The method includes: Step S1, constructing an ICV network system model supporting mobile group computing services; wherein the ICV network system model includes a network model, a task model, and a distributed auction model; The network model includes multiple task owner vehicles and multiple task vehicles that can share computing resources. The task owner vehicle is the buyer. , the task can share computing resources for the vehicle seller , the buyer and seller are connected to the RSU through V2I communication; In the task model, the seller The computing resources provided are , among which is the amount of computing resources that the seller can provide, is the maximum acceptable energy consumption, is the spatial coordinate of the seller; the buyer The computing resources required are ;in Input data size, is the amount of computing resources required for the task, is the maximum acceptable delay, is the spatial location coordinate of the buyer; In the distributed auction model, participating users are divided into two groups: the seller group , Buyer Group At the beginning of a trading round, the seller Submit Supply Vector ,in, Represents the seller's initial asking price, which is the lowest price the seller is willing to accept. Indicates the amount of idle computing resources; buyer Submit a demand vector ,in, Represents the buyer's initial bid, that is, the highest price the buyer is willing to pay. Indicates the amount of computing resources required to perform the task; Based on the quadratic utility function, the buyer's utility function and the seller's utility function are obtained; Step S2: Based on the ICV network system model, a distributed resource auction process is designed to construct a user utility maximization problem model; In step S3, the user utility maximization problem model is solved based on sinusoidal perturbations by combining reputation and distance factors to obtain the optimal transaction price and optimal transaction quantity to complete the distributed resource auction.

2. The distributed resource auction method for mobile group computing according to claim 1, characterized in that: The process of distributed resource auction is as follows: At the beginning of a trading round, the buyer and seller submit their demand vectors and the supply vector , and submit their respective bidding information, and determine the market type based on the surplus or deficit of total resources; market types include buyer's market and seller's market; If it is a buyer's market, calculate the first reputation-distance price of each seller based on the seller's bidding information, and sort the seller's first reputation-distance price; Calculate the buyer's reputation score based on the seller's historical actual transaction amount and historical agreed transaction amount in the bid information; Calculate the distance score between each buyer and each seller based on the location information in the buyer's and seller's bidding information; Calculate the seller's reputation-distance score based on the seller's reputation score and distance score; According to the initial asking price in the seller's bidding information and the seller's reputation-distance score, the first reputation-distance price of each seller is obtained, and the seller's first reputation-distance price is sorted by price; For the seller ranked first, calculate the first reputation-distance price of all corresponding buyers, and sort the first reputation-distance prices of all buyers to determine the sorting of transaction quantity.

3. The distributed resource auction method for mobile group computing according to claim 2, characterized in that: If it is a seller's market, calculate the second reputation-distance price of each buyer based on the buyer's bidding information, and sort the buyers' second reputation-distance prices; Calculate the buyer's reputation score based on the historical actual transaction amount and historical agreed transaction amount in the buyer's bidding information; Calculate the distance score between each buyer and each seller based on the location information in the buyer's and seller's bidding information; Calculate the buyer's reputation-distance score based on the buyer's reputation score and distance score; According to the initial bid in the buyer's bidding information and the buyer's reputation-distance score, the second reputation-distance price of each buyer is obtained, and the second reputation-distance prices of the buyers are sorted by price; For the second-ranked buyer, calculate the second reputation-distance prices of all corresponding sellers, and sort the second reputation-distance prices of all sellers to determine the ranking of transaction quantity.

4. The distributed resource auction method for mobile group computing according to claim 3, characterized in that: In a seller's market, the seller's reputation score is expressed as: in, is the time discount factor, and , is the seller’s historical actual transaction amount, is the seller’s historical actual transaction amount, is the fault tolerance factor, The seller's historical reputation score; The distance score between each buyer and each seller is expressed as: in, For the seller and the buyer The Euclidean distance between and are the maximum and minimum distances among all distances respectively; The buyer's reputation-distance score is expressed as: in, is the weight coefficient; The expression of the seller's first reputation-distance price is: in, The buyer's initial bid.

5. The distributed resource auction method for mobile group computing according to claim 3, characterized in that: The seller's utility function is expressed as: in, Indicates the seller i Select a strategy When the strategy set of other participants is For the seller i and the buyer j The transaction price between For the seller i and the buyer j The number of transactions between is the seller’s QoS coefficient, is the seller's network and communication cost coefficient, is the seller’s initial asking price; The buyer's utility function is expressed as: in, Indicates the buyer i Select a strategy When the strategy set of other participants is is the buyer’s QoS coefficient, is the buyer's network and communication cost coefficient, The buyer's initial bid.

6. The distributed resource auction method for mobile group computing according to claim 5, characterized in that: Combined with reputation and distance factors, the user utility maximization problem model is solved based on sinusoidal perturbations to obtain the optimal transaction price and optimal transaction quantity to complete the distributed resource auction process, including: In a seller's market, the number of transactions is determined based on the seller's available capacity and the buyer's task requirements; Based on the number of transactions and the order of the transaction numbers, the seller's utility function is sinusoidally perturbed; Perform Taylor expansion and gradient estimation on the utility function after sinusoidal perturbation to obtain the estimated gradient; The seller updates the strategy price based on the estimated gradient to bring it close to the Nash equilibrium point to obtain the final transaction price.

7. The distributed resource auction method for mobile group computing according to claim 6, characterized in that: In a seller's market, the utility function after sinusoidal disturbance is: in, For users i The strategy price at time t, is the disturbance amplitude, is the disturbance frequency, is the phase, is the first final transaction price; The Taylor expansion of the utility function after sinusoidal disturbance is: The estimated gradient is: in, is the utility function For user i's own policy variables gradient; The user adjusts their actions according to the estimated gradient to maximize utility: in, is the updated strategy price, is the learning rate, which is used to control the amplitude of the policy update.

Images