A Mobile Crowd Sensing Incentive Method Based on Optimal Information Freshness

Abstract

This invention discloses a mobile crowdsourcing sensing incentive method based on optimal information freshness, used to incentivize mobile users to actively participate in crowdsourcing sensing tasks within budget constraints. The method includes the following steps: Modeling is performed based on the budget constraints and cost pricing in real-world mobile crowdsourcing sensing tasks, establishing an optimization equation with optimal information freshness as the objective. A maximum-greedy strategy based on a reverse auction model is used to select a suitable candidate set of mobile users. Based on the obtained candidate set, a sampling time allocation scheme is proposed under two sampling methods. Finally, using the user candidate set and the sampling time allocation scheme as basic conditions, an incentive compensation mechanism is designed that satisfies realism, budget feasibility, and individual rationality. The method proposed in this invention can incentivize mobile users to participate in sensing tasks and achieve optimal information freshness under different sampling modes, thereby ensuring the efficient completion of crowdsourcing sensing tasks.

Description

Technical Field

[0001] This invention belongs to the field of incentive mechanism design and optimization in mobile crowd sensing, specifically involving an incentive method for mobile crowd sensing based on optimal information freshness. Background Technology

[0002] With the proliferation of mobile devices and their sensing capabilities, a new and promising sensing paradigm—mobile crowdsensing (MCS)—has recently garnered significant attention in both research and industry.

[0003] Roughly speaking, MCS (Multi-Sensing Components) refers to a group of mobile devices that perform large-scale sensing tasks through their onboard sensors. In this context, many attractive MCS applications, such as indoor positioning, environmental monitoring, road and traffic prediction, and healthcare, can be accomplished with very high quality and low service costs. However, mobile users are generally reluctant to provide services to others. This is not only because performing mobile crowdsourcing sensing tasks inevitably incurs costs for users, such as energy consumption, but also because sharing sensing data may expose user privacy, which is considered a privacy threat. Therefore, promoting user participation is particularly important for ensuring the success of MCS applications.

[0004] To incentivize mobile user participation in mobile crowd sensing (MCS), designing incentive mechanisms within MCS has been a focus of much research (NDLane, E. Miluzzo, H. Lu, D. Peebles, T. Choudhury and ATCampbell, “A survey of mobilephone sensing,” IEEE Communications Magazine, vol. 48, no. 9, pp. 140–150, Sept. 2010). Auctions have been widely used in mobile crowd sensing due to their ability to resolve conflicts of interest between buyers and sellers. Typically, performing mobile crowd sensing tasks involves resource / data transactions between task requesters and mobile users. By viewing the resource transaction system as an ecosystem, task requesters and mobile users can be modeled as buyers and sellers of resources, respectively. In this context, auctions aim to properly handle the conflict of interest between users and task requesters, as well as the internal competition among them. Typically, a designed auction should be able to appropriately allocate tasks to users and determine the amount of reward given to them. Research objectives involved include optimizing service quality, social welfare, or reducing the cost of mobile crowd sensing.

[0005] While existing solutions are effective at incentivizing users to share information with others, using reverse auction methods to model mobile crowdsensing tasks and optimizing for task completion (M. Xiao, B. An, J. Wang, G. Gao, S. Zhang and J. Wu, “CMAB-based reverse auction for unknown worker recruitment in mobile crowdsensing,” IEEE Transactions on Mobile Computing, doi:10.1109 / TMC.2021.3059346, 2021.), few have considered the freshness of the information provided by users. However, data freshness is particularly important. For example, in a traffic monitoring or prediction system, traffic conditions change over time, and only the most up-to-date information can support some intelligent applications, such as driving assistant applications. Therefore, it is necessary to consider the issues arising from incentive mechanism design in scenarios that take information freshness into account. Summary of the Invention

[0006] The main objective of this invention is to overcome the aforementioned shortcomings in the prior art and propose a design scheme for an incentive mechanism based on optimal information freshness. Modeling of mobile crowdsourcing sensing tasks is also included.

[0007] The present invention is achieved by at least one of the following technical solutions.

[0008] A mobile crowd-sensing incentive method based on optimal information freshness includes the following steps:

[0009] S1. Construct a many-to-one random sampling model and a many-to-many pre-sampling model based on the actual mobile crowd sensing scenario.

[0010] S2. Introduce information freshness and construct an optimization equation with the goal of minimizing information freshness for different sampling models;

[0011] S3. The incentive mechanism design problem of mobile crowd sensing task is constructed into a reverse auction model;

[0012] S4. Design a maximum-greedy algorithm based on the reverse auction model to solve the task allocation problem and obtain the winning candidate set in the task scenario;

[0013] S5. Based on the winning candidate set, establish an incentive mechanism model that is realistic and budget-feasible.

[0014] Further, step S1 includes:

[0015] Setting up a mobile crowd sensing task scenario: Define the mobile crowd sensing task publisher and set the set of interest points for sensing sampling R = {r1, r2, ..., r}. n}, Mobile user candidate set Within the budget B constraint, and within the time interval Within, where T represents the maximum sampling time, the optimal sampling time t is selected, and the best mobile user is allocated. Go to the specified sampling point of interest r i Collect information and data, and provide corresponding financial compensation to mobile users. i ;

[0016] Constructing a many-to-one random sampling model: For the many-to-one model, the sampling interest point is simplified to a single point; each mobile user s i At most l can be generated i One sample; l i Let x represent the maximum number of times a user can sample, used to quantify the user's sampling capability; i,t Indicates whether to select mobile user s i Samples are collected in time slot t;

[0017] Construct a many-to-many presampling model: derive s based on the user's movement trajectory. i The number of sampling points of interest that can be covered in each time slot, i.e., when s i At the point of interest r in time slot t j When within the perceptual scale, it is the point of interest r. j Generate data; use Indicates user s i At time slot t, for point of interest r j Perform data sampling.

[0018] Furthermore, step S2 specifically includes:

[0019] Let Γ(t) represent the AoI of the data at time slot t:

[0020]

[0021] In the formula, t k This represents the set of mobile users s that the data requester wants to recruit, specifically the set of mobile users S. i The time slot for generating the k-th update, i.e. Indicates whether to select mobile user s i In time slot t k Sample collection was carried out. This represents the set of time slots updated for all data at a single sampling point. t0 = 0, let X represent the vector of sampling decisions, i.e., X = <xi,t |s i ∈S,1≤t≤T>,x i,t Indicates whether to select mobile user s i Samples are collected in time slot t; time slot t is defined as the elapsed time since the latest data sample was generated.

[0022] The data update method for information freshness is as follows:

[0023]

[0024] Define the buyer utility optimization equation under the random sampling model, using U w (S) is represented as:

[0025]

[0026] Where α represents the parameter balancing the importance of average AoI and data quality, and Γ(S,X) represents the average AoI of the set of mobile users S assigned sampling tasks over time T, given the update decision vector function.

[0027]

[0028] Q(S,X) represents the overall data quality obtained by all users, let q i For user s i Data quality of the sampled data:

[0029]

[0030] Furthermore, the constraints of the buyer utility optimization equation under the random sampling model are: S represents the set of mobile users that the data requester wants to recruit. B indicates that the total payments provided to mobile users must not exceed their budget. l i s i It has a maximum number of sample collections; x i,t Indicates whether to select mobile user s i Samples are collected in time slot t.

[0031] Furthermore, under the presampling model, buyer utility U p The optimization equation for (S) is expressed as:

[0032]

[0033] in AoI represents the total data sample population of the set of points of interest R sampled within time T. Representing the point of interest r j In time slot t, at AoI, then:

[0034]

[0035]

[0036] Where n represents the maximum number of points of interest, Q p (S) represents the overall data quality that all candidate mobile users can provide:

[0037]

[0038] q i,j s i Can provide points of interest r j The data quality and related constraints are as follows: S represents the set of mobile users that the data requester wants to recruit. B indicates that the total payments provided to mobile users must not exceed their budget. l i s i It has a maximum number of sample collections; x i,t Indicates whether to select mobile user s i Samples are collected in time slot t.

[0039] Furthermore, the issuance, allocation, and compensation of mobile crowdsourcing sensing tasks are modeled as a reverse auction model, with mobile users as sellers and task issuers as buyers. The reverse auction process of the model consists of the following five steps: the buyer issues a data request to the seller, including the sampling area and requirements; interested sellers submit their offers to the buyer to compete for the task; based on the offers received from the sellers, the buyer selects the recruited sellers within the budget limit; the recruited sellers sample the data according to the requirements and upload the data to the buyer; the buyer provides compensation to the sellers according to the agreement between the two parties.

[0040] Further, step S4 includes:

[0041] The objective is to solve the buyer utility optimization equation U under the random sampling model. w Minimizing (S) is transformed into maximizing the submodular function f(S):

[0042]

[0043] because This indicates that the set S of mobile users that the data requester wants to recruit is an empty set. This represents the buyer's utility value when the recruited set of mobile users is empty. Therefore:

[0044]

[0045] Use S g This represents the candidate solution obtained using the greedy algorithm. Initially, let...

[0046] from Select the best candidate s i ,in:

[0047]

[0048] Δf(s i )=f(S g \{s i})-f(S g )

[0049] Δf(s i ) represents maximizing the marginal benefit of the submodular function f(S), and s i Budgetary feasibility conditions must be met:

[0050]

[0051] s i Add to candidate set S g In the middle, S g =S g ∪{s i};

[0052] Loop from Select the best candidate s i , making s i Meet the budget feasibility requirements, and s i Add to candidate set S g The process continues until all sellers join S g Or, it violates the conditions of budgetary feasibility;

[0053] Sellers who achieve maximum AoI reduction benefits and quality while keeping submission costs within the initial budget.

[0054]

[0055] f({s i}) indicates that when the candidate set is {s i The value of the submodular function at time} is obtained by randomly selecting from S. g and A solution is randomly selected from the options.

[0056] Further, step S5 includes:

[0057] For the sellers selected in S, if This means that only the seller with the largest reduction in AoI can win the bid. In this case, let... Indicates the seller The compensation that can be obtained is the budget allocated to the seller.

[0058] If S = S g S is determined as follows: g The winning seller i The compensation received;

[0059] From candidate bids Delete s i , This means excluding s. i The candidate bids, i.e.

[0060] Using the maximum-greedy algorithm from Seller S will be reselected from the candidate sellers. -i ;

[0061] s i With the winning set S -i Compare each user in the list and calculate s i and s i″ The marginal contribution, of which s″ i ∈S -i ;

[0062]

[0063]

[0064] in Let S be the candidate set. _i The value of the submodular function of the first i′-1 mobile user sets, Let S be the candidate set. _i The set of the first i′-1 mobile users and {s i The value of the submodular function obtained by the union of} Let S be the candidate set. -i The set of the first i′-1 mobile users and {s i″ The value of the submodular function obtained by the union of}, Δf′(s) i ) represents s i The marginal contribution, Δf′(s i″ ) represents s i″ marginal contribution S represents -i The first i'-1 sellers in the list receive the following two payments, respectively using and express:

[0065]

[0066]

[0067] Where c i″ Indicates mobile user s i″ Sampling cost, Indicates in s i″ Previously selected s i , The total compensation paid is guaranteed to be less than the budget B,s i The bid is no greater than s i That is, comparing sellers s i″ Winning in the position;

[0068] S -i All users and s i After comparison, take The maximum value is taken as s i Critical payment p i .

[0069] Furthermore, the average AoI of the interest points in the optimization equation is as follows:

[0070] Let L represent the maximum number of samples that the set of mobile users S that the data requester wants to recruit can generate:

[0071]

[0072] l i Indicates mobile user s i The number of samples that can be generated;

[0073] The entire update time Within this context, using mobile users in S, time is divided into L+1 update intervals:

[0074]

[0075] in:

[0076]

[0077] Where σk represents the update interval between the (k-1)th update and the kth update; T represents the maximum sampling time;

[0078] Γ(S, X) can be rewritten as:

[0079]

[0080] Let X represent the average AoI of a set of mobile users S assigned sampling tasks within time T, given an update decision vector X. This represents the average AoI when the update interval is σ;

[0081] The AoI optimization problem is restated as follows:

[0082]

[0083] stσ k ≥0, for all k

[0084] σ1+σ2+…+σ L+1 =T

[0085] σ L+1 Let L represent the update interval between the Lth update and time T, where L is the maximum number of samples that the data requester can generate from the set of mobile users S to be recruited.

[0086] Furthermore, the optimization equation Γ(S,X) is as follows:

[0087] Using the KKT conditions to solve the optimal solution to the AoI optimization problem, we get:

[0088]

[0089] σ * Indicates that it can make The time interval with the smallest value is the optimal time interval;

[0090] The final formula for calculating the updated decision vector function Γ(S,X) is expressed as follows:

[0091]

[0092] The optimal time interval σ represents * The average AoI under the following conditions.

[0093] Compared with the prior art, the present invention has the following beneficial effects:

[0094] (1) This invention optimizes the average AoI provided by the user and proposes an efficient MCS incentive mechanism for perceiving freshness. By introducing AoI to measure and define the freshness of data information, it balances data quality and data timeliness, which is more in line with the needs of actual mobile crowd intelligence perception tasks.

[0095] (2) In order to capture the conflict of interest between users, this invention defines two data sampling models and constructs an auction model, in which mobile users act as data sellers and data requesters act as data buyers. The proposed freshness incentive mechanism, in addition to considering how to select suitable winning sellers and determine their rewards, also completes the complex task of arranging data sampling for sellers;

[0096] (3) The auction proposed in this invention achieves several ideal properties, such as individual rationality, budget balance, authenticity and computational efficiency, which are practical and feasible. Attached Figure Description

[0097] Figure 1 This is a flowchart illustrating a design method for a mobile swarm intelligence sensing incentive mechanism based on a reverse auction model, according to an embodiment of the present invention.

[0098] Figure 2 This invention provides an embodiment of the trend of the update of the Point of Interest (AoI) over time.

[0099] Figure 3 This is a structural diagram of the mobile crowd-sensing task auction model provided in an embodiment of the present invention. Detailed Implementation

[0100] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0101] This example uses driving trajectory and outdoor temperature data recorded by sensors on taxis in Rome, Italy, from 2015 to 2016. This data includes records from 300 sensors. This example performs a mobile crowdsourcing sensing task, selecting the area with the most trajectory coverage as the mobile crowdsourcing sensing point of interest (POI) based on the taxi's trajectory. The sensing range for mobile users is set as a circular area with a sensing radius of r. When the POI is within the sensing range, information can be sampled and the sampled data uploaded. For a random sampling model, s... i The sampling capability, i.e., the maximum number of updates l i This represents the number of times each taxi travels around the area. For the presampled model, s i The sampling capability is exactly the time the taxi travels within the sensing area. The sampling cost for mobile users / sellers is randomly set within the range of (0, 10).

[0102] like Figure 1As shown, the mobile crowd-sensing incentive method based on the reverse auction model includes the following steps:

[0103] Based on actual mobile crowd sensing scenarios, we construct both a many-to-one random sampling model and a many-to-many pre-sampling model:

[0104] Since this example scenario involves a many-to-one mobile crowdsourcing sensing scenario, a many-to-one random sampling model is constructed. The set of sensing sampling interest points is R = {r1}, and the set of mobile user candidates is... Within the budget limit of B=3000, and within the specified time interval Choose an appropriate sampling time t to allocate the optimal mobile users. Go to the specified sampling point of interest r i Collect information and data, and provide corresponding financial compensation to mobile users. i Based on the geographic data uploaded by taxis, the value of each mobile user's s is calculated. i The maximum number of samples that can be generated for sampling points of interest in a movement trajectory is l i l i This represents the maximum number of times a user can sample, used to quantify a user's sampling capability. Let x... i,t Indicates whether to select mobile user s i Samples are collected in time slot t.

[0105] Considering both information freshness and data quality, we introduce AoI to measure information freshness and Q to measure the quality of sampled data, establishing an easily solvable optimization model. The specific steps are as follows:

[0106] Use t k This represents the time slot when a mobile user in S generates the k-th update, i.e. use This represents the set of time slots updated for all data at a single sampling point. The initial time slot data is the latest data, therefore t0 = 0. Let X represent the sampling decision vector, i.e., X =<x i,t |s i ∈S, 1≤t≤T>.

[0107] Data update methods for information freshness, such as Figure 2 As shown:

[0108]

[0109] In the current embodiment, the objective function to be optimized is:

[0110]

[0111] Where Γ(S,X) represents the average AoI of the set of mobile users S assigned sampling tasks over time T, given the update decision vector X:

[0112]

[0113] Q(S,X) represents the overall data quality obtained by all users, let q i For user s i Data quality of the sampled data:

[0114]

[0115] As a preferred embodiment, the maximum budget B is set to 3000. Since the mobile crowd sensing task has a maximum budget constraint, while considering the optimization objective function, a budget feasibility constraint also needs to be set, that is, the total compensation paid should not exceed the budget.

[0116]

[0117] In addition, since the sampling capabilities of mobile users are heterogeneous and limited, it is also necessary to consider that the total number of samples for each candidate mobile user cannot exceed its maximum number of samples:

[0118]

[0119] The computation of the average AoI integral is complex and difficult to apply in practice. Therefore, the time slots are intervalized to relax the objective function.

[0120] calculate The maximum number of samples L that the winning set S can generate during the period:

[0121]

[0122] The entire update time Within this timeframe, using the moving users in the winning set S, the time can be divided into L+1 update intervals. Using σ... k This represents the update interval between the (k-1)th update and the kth update.

[0123]

[0124] in:

[0125]

[0126] σ k Substituting the objective function into the equation, the objective function can be rewritten as follows:

[0127]

[0128] The AoI optimization problem can be restated as follows:

[0129]

[0130] stσ k ≥0, for all k

[0131] σ1+σ2+…+σ L+1 =T

[0132] Using the KKT conditions to find the optimal solution to the above problem, we can obtain:

[0133]

[0134] Therefore, for a many-to-one random sampling model, the average AoI value is optimal when the sampling time intervals are equal. Thus, the final formula for calculating Γ(S,X) can be expressed as:

[0135]

[0136] The optimization problem is concretized into a reverse auction model, which involves determining the candidate set of winners and the compensation for the winners.

[0137] As another preferred embodiment, taxis providing data in Rome are considered sellers, and mobile crowd sensing task providers are considered buyers. The flowchart of the auction model is as follows: Figure 3 As shown, the specific auction steps are as follows:

[0138] The entire reverse auction process consists of the following five steps: the buyer initiates and publishes a data request to the seller, including the sampling area and requirements; interested sellers submit their bids to the buyer to compete for the task; based on the bids received from the sellers, the buyer selects the recruited sellers within a budget limit according to certain criteria; the recruited sellers sample the data according to the requirements and upload the data to the buyer; the buyer provides compensation to the sellers according to the agreement between the two parties.

[0139] Next, based on auction mechanism design theory, and under the premise of ensuring budget feasibility, individual rationality, authenticity, and computational efficiency, we design an algorithm for selecting the winning candidate set:

[0140] Minimize the objective function U w (S) can be transformed into maximizing the submodular function f(S), which can be obtained by subtracting the objective function from the value of the initial state of any chosen taxi:

[0141]

[0142] because so:

[0143]

[0144] First, the candidate set S is obtained using a greedy algorithm that maximizes marginal benefit. g The specific steps are as follows:

[0145] Initialization command Sort all candidates in the candidate set according to their unit marginal benefit value:

[0146]

[0147] The candidate with the highest marginal benefit per unit is selected as the winner of the current round. i , determine s i Does the budget feasibility condition meet?

[0148]

[0149] If the conditions are not met, the algorithm terminates.

[0150] If satisfied, s i Add to candidate set S g In the middle, S g =S g ∪{s i}. And repeat the above operation to sort the remaining 299 taxis in the candidate set according to their unit marginal benefit:

[0151]

[0152] The candidate with the highest marginal benefit per unit is selected as the winner of the current round. i , determine s i Does it meet the budget feasibility condition? And so on. Finally, the winning candidate set S selected by the greedy algorithm is obtained. g .

[0153] Since there exists an extreme case where a single candidate can satisfy all sampling requirements, in order to avoid violating the monotonicity of the algorithm, it is also necessary to calculate the candidate with the maximum benefit from all candidate sets.

[0154]

[0155] To preserve the monotonicity of the algorithm, finally, based on probabilities of 0.6 and 0.4 respectively, S is randomly selected from... g and A solution is randomly selected from the options.

[0156] Based on the winning candidate set, design an incentive mechanism that is realistic and budget-feasible. The specific implementation steps are as follows:

[0157] like This means that only the seller with the largest reduction in AoI (Amount of Income) can win the bid. In this case, let... Budget B will be directly allocated

[0158] If S = S g S is determined as follows: g The winning seller i The compensation received.

[0159] First, extract the candidate set S. g The first winner, s1, is selected. The set of s1 candidates is then determined. When deleted, the remaining 299 candidates were used Indicates. (The s) i With the winning set S -i Each user in the "s" i Comparison, Calculate the first winners s1 and s” of the current round respectively. i Marginal benefits:

[0160]

[0161]

[0162] Based on the calculated marginal benefit, two boundary values ​​can be calculated:

[0163]

[0164]

[0165] Each time s1 and candidate s” i The smaller of the two boundary values ​​obtained after comparison is retained. That is, After comparing with all users, take... The maximum value is taken as the critical payment p1 of s1.

[0166] Next, repeat the entire process, taking the second winner from the winning set, s2, and... The critical payout p_2 of s2 is obtained by comparing all users in the winning set. This continues until the payouts for all winners in the winning set are determined.

[0167] This invention proposes a mobile crowd-based intelligent perception incentive method based on optimal information freshness by studying the computational offloading optimization problem in mobile edge computing. This method satisfies the requirements of authenticity, individual rationality, computational efficiency feasibility, and budget feasibility.

[0168] Authenticity: The auction must be able to induce mobile users to submit genuine bids throughout the entire auction process. In other words, the auction must guarantee that no seller can increase the auction's utility by submitting bids that differ from actual costs. That is, for any mobile user s... i In terms of its actual cost quote c i Its utility is always no less than that of an unrealistic cost quote c′ i utility.

[0169] Individual rationality: In the actual auction process, the individual rationality of all participating mobile users means that the auction mechanism must guarantee that each winning seller receives a non-negative net benefit. In other words, it must receive a reward no less than its actual cost bid.

[0170] Budgetary feasibility: For the data recruiter, i.e. the buyer, the total cost paid to the mobile users they recruit, i.e. the seller, should not exceed their budget.

[0171] Computational efficiency and feasibility: The proposed incentive mechanism operates in polynomial time, exhibiting high computational efficiency.

[0172] It also considers complex issues such as information freshness and data quality, as well as the heterogeneity among mobile users. An incentive mechanism scheme was designed that satisfies budget feasibility, authenticity, individual rationality, and computational efficiency, and can be used for practical mobile crowdsourcing sensing tasks.

[0173] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A mobile crowd-sensing incentive method based on optimal information freshness, characterized in that, Includes the following steps: S1. Construct a many-to-one random sampling model and a many-to-many pre-sampling model based on the actual mobile crowd sensing scenario. S2. Introducing information freshness, for different sampling models, constructing an optimization equation with the objective of minimizing information freshness; Step S2 specifically includes: Let Γ(t) represent the AoI of the data at time slot t: In the formula, t k This represents the set of mobile users s that the data requester wants to recruit, specifically the set of mobile users S. i The time slot for generating the k-th update, i.e. Indicates whether to select mobile user s i In time slot t k Sample collection was carried out. This represents the set of time slots updated for all data at a single sampling point. t0 = 0, let X represent the vector of sampling decisions, i.e., X = <x i,t |s i ∈S,1≤t≤T>,x i,t Indicates whether to select mobile user s i Samples are collected in time slot t; time slot t is defined as the elapsed time since the latest data sample was generated. The data update method for information freshness is as follows: Define the buyer utility optimization equation under the random sampling model, using U w (S) is represented as: Where α represents the parameter balancing the importance of average AoI and data quality, and Γ(S,X) represents the average AoI of the set of mobile users S assigned sampling tasks over time T, given the update decision vector function. Q(S,X) represents the overall data quality obtained by all users, let q i For user s i Data quality of the sampled data: S3. The incentive mechanism design problem of mobile crowd sensing task is constructed into a reverse auction model; S4. Design a maximum-greedy algorithm based on the reverse auction model to solve the task allocation problem and obtain the winning candidate set in the task scenario; S5. Based on the winning candidate set, establish an incentive mechanism model that is realistic and budget-feasible.

2. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 1, characterized in that, Step S1 includes: Setting up a mobile crowd sensing task scenario: Define the mobile crowd sensing task publisher and set the set of interest points for sensing sampling R = {r1, r2, ..., r}. n }, Mobile user candidate set Within the budget B constraint, and within the time interval Within, where T represents the maximum sampling time, the optimal sampling time t is selected, and the best mobile user is allocated. Go to the specified sampling point of interest r i Collect information and data, and provide corresponding financial compensation to mobile users. i ; Constructing a many-to-one random sampling model: For the many-to-one model, the sampling interest point is simplified to a single point; each mobile user s i At most l can be generated i One sample; l i Let x represent the maximum number of times a user can sample, used to quantify the user's sampling capability; i,t Indicates whether to select mobile user s i Samples are collected in time slot t; Construct a many-to-many presampling model: derive s based on the user's movement trajectory. i The number of sampling points of interest that can be covered in each time slot, i.e., when s i At the point of interest r in time slot t j When within the perceptual scale, it is the point of interest r. j Generate data; use Indicates user s i At time slot t, for point of interest r j Perform data sampling.

3. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 1, for a many-to-many presampling model, is characterized in that: The constraints of the buyer utility optimization equation under the random sampling model are: S represents the set of mobile users that the data requester wants to recruit. B indicates that the total payments provided to mobile users must not exceed their budget. l i s i It has a maximum number of sample collections; x i,t Indicates whether to select mobile user s i Samples are collected in time slot t.

4. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 1, for a many-to-many presampling model, is characterized in that: Under the presampling model, buyer utility U p The optimization equation for (S) is expressed as: in AoI represents the total data sample population of the set of points of interest R sampled within time T. Representing the point of interest r j In time slot t, at AoI, then: Where n represents the maximum number of points of interest, Q p (S) represents the overall data quality that all candidate mobile users can provide: q i,j s i Can provide points of interest r j The data quality and related constraints are as follows: S represents the set of mobile users that the data requester wants to recruit. B indicates that the total payments provided to mobile users must not exceed their budget. l i s i It has a maximum number of sample collections; x i,t Indicates whether to select mobile user s i Samples are collected in time slot t.

5. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 1, characterized in that, The issuance, allocation, and compensation of mobile crowdsourcing sensing tasks are modeled as a reverse auction model, with mobile users as sellers and task issuers as buyers. The reverse auction process of the model consists of the following five steps: the buyer issues a data request to the seller, including the sampling area and requirements; interested sellers submit their offers to the buyer to compete for the task; based on the offers received from the sellers, the buyer selects from the recruited sellers within the budget limit; the recruited sellers sample the data according to the requirements and upload the data to the buyer; the buyer provides compensation to the sellers according to the agreement between the two parties.

6. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 1, characterized in that, Step S4 includes: The objective is to solve the buyer utility optimization equation U under the random sampling model. w Minimizing (S) is transformed into maximizing the submodular function f(S): because This indicates that the set S of mobile users that the data requester wants to recruit is an empty set. This represents the buyer's utility value when the recruited set of mobile users is empty. Therefore: Use S g This represents the candidate solution obtained using the greedy algorithm. Initially, let... from Select the best candidate s i ,in: Δf(s i )=f(S g \{s i })-f(S g ) Δf(s i ) represents maximizing the marginal benefit of the submodular function f(S), and s i Budgetary feasibility conditions must be met: s i Add to candidate set S g In the middle, S g =S g ∪{s i }; Loop from Select the best candidate s i , making s i Meet the budget feasibility conditions, and s i Add to candidate set S g The process continues until all sellers join S g Or, it violates the conditions of budgetary feasibility; Sellers who achieve maximum AoI reduction benefits and quality while keeping submission costs within the initial budget. f({s i }) indicates that when the candidate set is {s i The value of the submodular function at time} is obtained by randomly selecting from S. g and A solution is randomly selected from the options.

7. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 4, characterized in that, Step S5 includes: For the sellers selected in S, if This means that only the seller with the largest reduction in AoI can win the bid. In this case, let... Indicates the seller The compensation that can be obtained is the budget allocated to the seller. If S = S g S is determined as follows: g The winning seller i The compensation received; From candidate bids Delete s i , This means excluding s. i The candidate bids, i.e. Using the maximum-greedy algorithm from Seller S will be reselected from the candidate sellers. -i ; s i With the winning set S -i Compare each user in the list and calculate s i and s i″ The marginal contribution, of which s” i ∈S -i ; in Let S be the candidate set. -i The value of the submodular function of the first i′-1 mobile user sets, Let S be the candidate set. -i The set of the first i′-1 mobile users and {s i The value of the submodular function obtained by the union of} Let S be the candidate set. -i The set of the first i′-1 mobile users and {s i″ The value of the submodular function obtained by the union of}, Δf′(s) i ) represents s i The marginal contribution, Δf′(s i″ ) represents s i″ marginal contribution S represents -i The first i'-1 sellers in the list receive the following two payments, respectively using and express: Where c i″ Indicates mobile user s i″ Sampling cost, Indicates in s i″ Previous selection Guarantee that the total compensation paid is less than the budget B,s i The bid is no greater than s i That is, comparing sellers s i″ Winning in the position; S -i All users and s i After comparison, take The maximum value is taken as s i Critical payment p i .

8. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 1, characterized in that, The average AoI of the interest points in the optimization equation is as follows: Let L represent the maximum number of samples that the set of mobile users S that the data requester wants to recruit can generate: l i Indicates mobile user s i The number of samples that can be generated; The entire update time Within this context, using mobile users in S, time is divided into L+1 update intervals: in: Where σ k Γ(S,X) represents the update interval between the (k-1)th update and the kth update; T represents the maximum sampling time; Γ(S,X) is rewritten as: Let X represent the average AoI of a set of mobile users S assigned sampling tasks within time T, given an update decision vector X. This represents the average AoI when the update interval is σ; The AoI optimization problem is restated as follows: σ L+1 Let L represent the update interval between the Lth update and time T, where L is the maximum number of samples that the data requester can generate from the set of mobile users S to be recruited.

9. The mobile swarm intelligence sensing incentive method based on optimal information freshness according to claim 8, characterized in that, The optimization equation Γ(S,X) is as follows: Using the KKT conditions to solve the optimal solution to the AoI optimization problem, we get: σ * Indicates that it can make The time interval with the smallest value is the optimal time interval; The final formula for calculating the updated decision vector function Γ(S,X) is expressed as follows: The optimal time interval σ represents * The average AoI under the following conditions.

Images