A data acquisition task allocation method based on the Internet of Things

Abstract

The present invention discloses a data collection task allocation method based on the Internet of Things. The problem to be solved by task allocation is to maximize the average benefit of the system through selection control and task allocation under the condition of satisfying the stability of the sensing task queue, and define the objective function; adopt the Lyapunov optimization theory for online task allocation, and perform task allocation according to the feasible solution. The online task allocation algorithm in the mobile crowd sensing of the present invention is effective and feasible, can ensure the stability of the task queue, and can increase the system benefit by adjusting the value of V.

Description

Technical Field

[0001] The present invention belongs to the field of information technology, and particularly relates to a method for allocating data collection tasks based on the Internet of Things. Background Art

[0002] With the rapid development of information technology, today's society has entered the big data era. The huge value of data has gradually been recognized by society, and the amount of data has shown an explosive growth. On the other hand, prominent contradictory problems such as incomplete and outdated data exist to varying degrees in many fields. Among them, the acquisition of end data in high-dynamic and strong situations is more difficult, but such data often has higher value, such as damage data that can reflect the battlefield situation in real time, real-time performance data of equipment, and troop distribution data in the battlefield environment. The Internet of Things, as an upgrade of traditional sensor networks, can integrate various types of sensors to collect various required data such as sound, light, heat, electricity, mechanics, chemistry, biology, and location in real time. Therefore, it is an important technical means to support end data collection.

[0003] Meanwhile, with the development of computer hardware technology and the miniaturization of sensors, more and more sensing resources are embedded in mobile devices, such as cameras, microphones, accelerometers, and electronic compasses. To make full use of such scattered mobile devices with data collection capabilities, the concept of Mobile Crowdsensing (MCS) has emerged. It combines the idea of crowdsourcing and mobile sensing to construct a novel sensing paradigm for the Internet of Things. Mobile Crowdsensing is centered around mobile user devices, which can provide sensing, computing, and communication capabilities. Through the recruitment and coordination of servers, they sense the surrounding environment and upload data to the server to support other applications. For example, in military operations, the rich sensing data collected through Mobile Crowdsensing can effectively carry out intelligence acquisition and collection, and reconnaissance and surveillance of combat targets.

[0004] Compared with traditional sensor networks, mobile crowd sensing systems have many new advantages: First, mobile crowd sensing systems can make full use of existing mobile devices to perform sensing tasks. The sensing resources are widely distributed, rich in variety, and highly scalable. There is no need to deploy sensors separately, which not only improves resource utilization but also saves the costs and time of deployment and maintenance. Second, the sensing capabilities of related devices are rich, and some also have functions such as sensing, computing, communication, and storage, which are not possessed by traditional sensors. Third, in terms of sensing data, in mobile crowd sensing systems, mobile intelligent devices are mostly held and used by mobile users. Different users have different understandings and preferences for sensing tasks, so the uploaded sensing data is rich in content, which is more convenient for different applications to reuse data. Mobile crowd sensing technology has received extensive attention in recent years and has broad application prospects in fields such as traffic management, health monitoring, and environmental protection, making environmental sensing and information collection more flexible, convenient, and efficient.

[0005] In mobile crowd sensing based on the Internet of Things, its process can be roughly summarized into four stages: task allocation, data sensing, data processing, and data uploading. Currently, related research mainly focuses on the above four stages, among which task allocation is the key link of mobile crowd sensing. After the sensing platform receives the sensing tasks issued by external applications, according to the task allocation scheme, it selects appropriate sensing participants to execute the tasks, and then collects and organizes the sensing data and sends it to the external applications to obtain benefits. The formulation of the task allocation scheme usually needs to consider various factors, including both the energy consumption, sensing cost, and privacy security of sensing participants, as well as factors such as data quality, task deadline, expenditure constraint, and fairness. When the sensing tasks are different, the task allocation schemes are different. For example, the allocation schemes for online tasks and offline tasks are different, and the allocation schemes for multi-type tasks and single-type tasks are different. The rationality of the task allocation scheme directly affects the success of sensing tasks. Therefore, formulating a reasonable task allocation scheme is a hot topic and key point in related research on mobile crowd sensing. In actual situations, the online task allocation problem is more common than the offline task allocation problem, and multi-type tasks are more common than single-type tasks. Summary of the Invention

[0006] In view of this, the present invention studies the online allocation problem of multi-type tasks oriented to system benefits. Based on the Lyapunov optimization algorithm, the present invention takes the system utility as the objective function and solves the problem, realizing the long-term online dynamic control of the task admission strategy and task allocation scheme.

[0007] A data collection task allocation method based on the Internet of Things disclosed by the present invention is applied to a sensing platform. After receiving a sensing task issued by an external application, the sensing platform selects a suitable sensing participant to execute the task according to the task allocation scheme, collects and organizes the sensing data, sends it to the external application, and obtains benefits. The problem to be solved by the task allocation is to maximize the average benefit of the system through selection control and task allocation under the condition of satisfying the stability of the sensing task queue. The objective function is as follows:

[0008]

[0009]

[0010] Q j (t + 1) = max[Q j (t) - r j (t), 0] + o j (t) (3)

[0011]

[0012] Among them, t is the time slot, is the expected value of U(t), U(τ) is the benefit of the sensing platform in time slot τ, β represents the benefit that the sensing platform can obtain by executing a unit number of tasks, and o j (t) is the number of sensing tasks of type j at time slot t, and Q j (t) is the backlog vector of the sensing task queue of type j in the platform at time slot t, and r j (t) is the number of tasks of task type j that can be executed at time slot t, and o j (t) is the number of sensing tasks of type j that enter the sensing platform after the platform selection decision at time slot t;

[0013] The Lyapunov optimization theory is adopted to solve the feasible solution of online task allocation, which is specifically as follows:

[0014] Define the Lyapunov drift function as:

[0015]

[0016] The drift-plus-penalty function is:

[0017]

[0018] Among them, V is a non-negative constant. By adjusting the value of V, the proportional relationship between stability and the utility function can be controlled;

[0019] At each time slot t, any feasible solution of the objective functions (1)-(4) satisfies the following inequality:

[0020]

[0021] Among them

[0022] Perform task allocation according to the feasible solution.

[0023] Furthermore, the types and quantities of sensing tasks arriving at the platform within each time slot are different. The quantity O j (t) of sensing tasks of type j arriving at the sensing platform at time slot t is an independently and identically distributed random variable in each time slot t. Not all arriving sensing tasks can enter the sensing platform to be executed. The sensing platform decides the tasks that can enter the system and be executed according to the task execution situation in the previous time slot, that is, makes a selection decision.

[0024] Furthermore, in order to achieve queue stability and utility maximization, it is necessary to minimize this upper bound in each time slot, and the following conditions need to be satisfied:

[0025]

[0026] Furthermore, for the objective functions (1)-(4), the queue Q(t) is stable, that is, all sensing tasks entering the sensing platform will be executed; when {Q j (t)}, j ∈ {1,..., M} all satisfy the objective functions (1)-(4), then the mobile crowd sensing system is stable.

[0027] Furthermore, Varies linearly with o j (t), so the optimal solution to the task allocation problem is:

[0028] When Q j (t) - Vβ < 0, o j (t) = O j (t);

[0029] Otherwise, o j (t) = 0;

[0030] To minimize the upper bound, maximize , and the specific method is as follows:

[0031] When Q j (t) + o j (t) ≥ N, x ij (t) = 1, i ∈ {1,..., N}, j ∈ {1,..., M}, that is, in order to achieve benefit maximization, all participants execute tasks;

[0032] When Q j (t) + o jWhen (t) < N, select (Q j (t) + o j (t)) participants to execute the tasks in the queue.

[0033] The beneficial effects of the present invention are as follows:

[0034] The online task allocation algorithm in the mobile crowd sensing of the present invention is effective and feasible, which can ensure the stability of the task queue and can increase the system benefit by adjusting the value of V. Description of the Drawings

[0035] Figure 1 The task allocation flowchart of the present invention;

[0036] Figure 2 The change of the system benefit U(t) when the value of V is different in the experiment of the present invention

[0037] Figure 3 The change of the task queue Q_avg(t) when the value of V is different in the experiment of the present invention;

[0038] Figure 4 The graph of the total benefit and the average benefit varying with the value of V in the experiment of the present invention. Detailed Embodiments

[0039] The present invention will be further described below with reference to the drawings, but the present invention is not limited in any way. Any transformation or replacement based on the teachings of the present invention falls within the protection scope of the present invention.

[0040] First, the basic definitions of the symbols involved in the present invention are given.

[0041]

[0042]

[0043] In actual situations, the data required for different application scenarios are not the same. Therefore, there are multiple types of sensing tasks in the MCS system studied in the present invention. The time is divided into discrete equal-length time slots t, t ∈ {0, 1, 2,...}. The types and quantities of the sensing tasks arriving at the platform within each time slot are different. O j (t) is an independent and identically distributed random variable in each time slot, Not all the arriving sensing tasks can enter the sensing platform for execution. The platform decides the tasks that can enter the system and be executed according to the task execution situation in the previous time slot, that is, the selection decision. 0 ≤ o j (t) ≤ O j (t).

[0044] In each time slot, the platform needs to allocate different types of tasks to the sensing participants, i.e., the allocation decision. x ij x ij (t) represents whether task j is allocated to participant i in time slot t, i.e., whether participant i executes task j. When x ij ij (t) = 1, participant i executes task j; when x ij ij (t) = 0, participant i does not execute task j. Within each time slot, the same participant can execute multiple types of tasks with a unit quantity. After the task allocation, the number of tasks j that can be executed in time slot t is denoted as r j (t):

[0045]

[0046] where 0 ≤ r j (t) ≤ N.

[0047] Let Q(t) = (Q1(t), Q1(t),..., Q M (t)) represent the backlog vector of different types of sensing request queues within the platform in time slot t, and its dynamic change follows the following formula:

[0048] Q j (t + 1) = max[Q j (t) - r j (t), 0] + o j (t) (2)

[0049] where Q j (0) = 0, j ∈ {1,..., M}. Q j (t) ≥ 0. It can be seen from

[19] that if Q j (t) satisfies:

[0050]

[0051] Then Q j (t) is stable, that is, all sensing tasks entering the sensing platform will be executed; when {Q j (t)}, j ∈ {1,..., M} all satisfy the above formula, the MCS system is said to be stable.

[0052] (2) System benefit

[0053] Let U(t) represent the benefit of the sensing platform in time slot t, and its magnitude depends on the number of sensing tasks entering the platform,

[0054]

[0055] where β represents the benefit that the sensing platform can obtain by executing a unit quantity of tasks, and it is a positive constant.

[0056] (3) Objective function

[0057] The problem to be solved by the present invention is to maximize the average benefit of the system through control selection and task allocation under the condition of satisfying the stability of the sensing task queue. Therefore, the objective function of this problem is designed as follows:

[0058]

[0059]

[0060] Q j (t + 1) = max[Q j (t) - r j (t), 0] + o j (t) (7)

[0061]

[0062] (4) Online task allocation strategy

[0063] To solve the above optimization problem, the present invention adopts the Lyapunov optimization theory, which greatly reduces the complexity of problem solving.

[0064] The Lyapunov optimization theory is a stochastic network optimization theory. After the research and development of Professor M.J. Neely and his team, this theory is currently relatively mature and has been applied in many fields. Different from other static optimization algorithms, the algorithm proposed based on the Lyapunov optimization theory is an online, dynamic, and adaptive algorithm. When the system state changes, no manual adjustment is required. This algorithm has a certain self-learning ability and can maintain asymptotic optimality. In addition, the algorithm based on the Lyapunov optimization theory is relatively easy to implement, requires less prior knowledge, and has a certain decoupling ability, which can greatly reduce the difficulty of algorithm design.

[0065] Define the Lyapunov optimization function at time slot t as:

[0066]

[0067] Define the Lyapunov drift function as:

[0068]

[0069] Add the utility function as a penalty function to the above drift function to obtain the drift-plus-penalty function:

[0070]

[0071] Among them, V is a non - negative constant. By adjusting the value of V, the proportional relationship between stability and the utility function can be controlled.

[0072] Theorem 1: At each time slot t, any feasible solution of the objective functions (5)-(8) satisfies the following inequality:

[0073]

[0074] Where

[0075] Proof: From It can be seen that:

[0076]

[0077] In addition:

[0078] (max[Q j (t)-r j (t), 0]) 2 ≤(Q j (t)-r j (t)) 2 (14)

[0079] Squaring both sides of equation (2) and according to the above - mentioned inequality, we can get:

[0080]

[0081] Therefore:

[0082]

[0083] Where is a positive constant.

[0084]

[0085] Taking the expectation of equation (13), summing over j ∈ {1,..., M}, and adding the right - hand side of equation (14), we get the following equation:

[0086]

[0087] Theorem 1 is proved.

[0088] From Theorem 1, an upper - bound of the drift - plus - penalty function can be obtained. In order to achieve queue stability and utility maximization, it is necessary to minimize this upper - bound at each time slot. Transforming the right - hand side of equation (15) gives:

[0089]

[0090] Next, the stability of the task queue Q(t) is proved.

[0091] Theorem 2: For the objective functions (5)-(8), the queue Q(t) is stable.

[0092] Proof: First, if the objective functions (5)-(8) have a feasible solution, then for any δ > 0, there exists a dynamic load balancing strategy that satisfies:

[0093] For any t,

[0094] E{U(t)} ≥ U opt -δ (20)

[0095] E{V c (t)} ≤ V max +δ (21)

[0096] E{o j (τ)} ≤ E{r j (t)} + δ (22)

[0097] where U opt represents the supremum of the system benefits among all feasible solutions.

[0098] From equation (11), we can obtain:

[0099]

[0100] When δ = 0, from formulas (19)(20)(21), we can get:

[0101] Δ v (Q(t)) ≤ B - VU opt (24)

[0102]

[0103] Thus, it can be known that the queue Q(t) is stable.

[0104] Theorem 2 is proved.

[0105] The benefit-optimal algorithm design based on Lyapunov optimization is shown in Table 1.

[0106]

[0107] Among them, varies linearly with o j (t), so its optimal solution is:

[0108] ① When Q j (t) - Vβ < 0, o j (t) = O j (t);

[0109] ②Otherwise, o j (t) = 0.

[0110] To minimize the upper bound, should be maximized:

[0111] ①When Q j (t) + o j (t) ≥ N, x ij (t) = 1, i ∈ {1,..., N}, j ∈ {1,..., M}, that is, to maximize the benefit, all participants perform tasks;

[0112] ②When Q j (t) + o j (t) < N, select (Q j (t) + o j (t)) participants to perform the tasks in the queue.

[0113] The second part:

[0114] Next, the performance of the method proposed in the present invention is experimentally evaluated. Five types of sensing tasks are set, and the number of different types of tasks arriving at the sensing platform in each time slot is generated by a Poisson distribution with different parameters λ, where λ1 = 10, λ2 = 15, λ3 = 15, λ4 = 25, λ5 = 15. It is assumed that there are 5 participants performing sensing tasks in the experiment. Set β = 1.

[0115] Experiment 1: Set V = 3000, V = 6000, V = 9000, V = 12000, V = 15000, V = 18000, V = 21000, V = 24000, V = 27000 respectively, and observe the change of the system benefit U(t). The experiment runs for 1000 time slots, that is, t ∈ {0, 1,..., 999}.

[0116] It can be seen from Figure 2 that when the value of V is different, the change of the system benefit U(t) is also different. As the value of V gradually increases, the mean value of U(t) also gradually increases, and the stable value within 1000 time slots also gradually approaches 80. The above phenomenon is caused by Equation (12). By adjusting the value of V, the proportional relationship between stability and system benefit can be controlled. When V is larger, the system benefit is larger, but the convergence speed of the task queue Q(t) is slower, and the time required for the queue to stabilize is longer; conversely, the system benefit is smaller, the convergence speed is faster, and the time required for the queue to stabilize is shorter.

[0117] To further observe the change of the task queue Q(t) and its stability, this paper proposes the average queue length index Q_avg(t) and conducts Experiment 2, where:

[0118] Experiment 2: Set V = 3000, V = 6000, V = 9000, V = 12000, V = 15000, V = 18000, V = 21000, V = 24000, V = 27000 respectively, and observe the change of system benefit. The experiment runs for 6000 time slots in total, that is, t ∈ {0, 1,..., 5999}.

[0119] From Figure 3 It can be seen that under different V values, the task queue Q_avg(t) finally tends to be stable, verifying the effectiveness and feasibility of the algorithm proposed in this paper. However, with different V values, the queue stabilization speed is different. The larger the V value, the longer the time required for the queue to stabilize. It can be seen that it is consistent with the conclusion obtained in Experiment 1.

[0120] In order to further study the influence degree of V value on system benefit, this paper proposes to refine the system benefit index and conducts Experiment 3.

[0121] The system benefit index is divided into the total benefit U_sum index and the average benefit U_avg index.

[0122] Total benefit U_sum:

[0123]

[0124] Average benefit U_avg:

[0125]

[0126] Experiment 3: Set V = 3000, V = 6000, V = 9000, V = 12000, V = 15000, V = 18000, V = 21000, V = 24000, V = 27000 respectively, and observe the change of system benefit. The experiment runs for 1000 time slots in total, that is, t ∈ {0, 1,..., 999}.

[0127] Figure 4 The left figure is the graph of the total benefit changing with the V value, and the right figure is the graph of the average benefit changing with the V value. From Figure 4 It can be seen that with the increase of the V value, both the total benefit U_sum and the average benefit U_avg gradually increase. Therefore, when performing task allocation in the mobile crowd sensing system, the system benefit can be improved by increasing the V value.

[0128] From the above three experiments, it can be seen that the online task allocation algorithm proposed in this paper for mobile crowd sensing is effective and feasible, which can ensure the stability of the task queue and can increase the system benefit by adjusting the V value.

[0129] The beneficial effects of the present invention are as follows:

[0130] The online task assignment algorithm in the mobile crowd sensing of the present invention is effective and feasible, which can ensure the stability of the task queue and increase the system benefit by adjusting the value of V.

[0131] As used herein, the term "preferred" is intended to be used as an example, illustration, or exemplification. Any aspect or design described as "preferred" in the present invention need not be construed as more advantageous than other aspects or designs. On the contrary, the use of the term "preferred" is intended to present concepts in a specific manner. As used in this application, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X uses A or B" means any one of the permutations is naturally included. That is, if X uses A; X uses B; or X uses both A and B, then "X uses A or B" is satisfied in any of the foregoing examples.

[0132] Moreover, although the present disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art based on a reading and understanding of this specification and the drawings. The present disclosure includes all such modifications and variations and is limited only by the scope of the appended claims. In particular, with respect to the various functions performed by the above-described components (e.g., elements, etc.), the terms used to describe such components are intended to correspond to any component that performs the specified function of the component (e.g., it is functionally equivalent), unless otherwise indicated, even if it is not structurally equivalent to the disclosed structure that performs the function in the exemplary implementations of the present disclosure shown herein. In addition, although a particular feature of the present disclosure has been disclosed with respect to only one of several implementations, such a feature may be combined with one or other features of other implementations as may be desired and advantageous for a given or particular application. Moreover, insofar as the terms "comprises," "has," "contains," or any variation thereof are used in a particular embodiment or claim, such terms are intended to include in a manner similar to the term "includes."

[0133] Each functional unit in the embodiments of the present invention may be integrated into a processing module, or each unit may exist physically alone, or multiple or more than multiple units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a magnetic disk, or an optical disc, etc. Each of the above-mentioned devices or systems may execute the storage method in the corresponding method embodiment.

[0134] In summary, the above embodiments are an implementation manner of the present invention. However, the implementation manner of the present invention is not limited by the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A data acquisition task allocation method based on the Internet of Things, which is applied to a perception platform. After receiving a perception task published by an external application, the perception platform selects appropriate perception participants to execute the task according to the task allocation scheme, collects and organizes the perception data, sends it to the external application, and obtains benefits. It is characterized in that, The problem to be solved by the task allocation is to maximize the average benefit of the system through selection control and task allocation under the condition of satisfying the stability of the sensing task queue. The objective function is as follows: Q j (t + 1) = max[Q j (t) - r j (t), 0] + o j (t) (3) where t is the time slot, is the expected value of U(t), U(τ) is the benefit of the sensing platform at time slot τ, β represents the benefit that the sensing platform can obtain by executing a unit number of tasks, o j (t) is the number of sensing tasks of type j at time slot t, Q j (t) is the backlog vector of sensing task queues of type j in the platform at time slot t, r j (t) is the number of tasks of type j that can be executed at time slot t, o j (t) is the number of sensing tasks of type j that enter the sensing platform after the platform selection decision at time slot t; The Lyapunov optimization theory is used to solve the feasible solution of the online task allocation, which is specifically as follows: Define the Lyapunov drift function as: The drift-plus-penalty function is: where V is a non-negative constant, and the proportional relationship between stability and the utility function can be controlled by adjusting the value of V; At each time slot t, any feasible solution of the objective functions (1)-(4) satisfies the following inequality: Among them Task allocation is performed according to the feasible solution.

2. The data acquisition task allocation method based on the Internet of Things according to claim 1, wherein The type and quantity of sensing tasks arriving at the platform in each time slot are different. The number of sensing tasks of type j arriving at the sensing platform at time slot t, O j (t) is an independent and identically distributed random variable in each time slot t. Not all arriving sensing tasks can enter the sensing platform for execution. The sensing platform decides which tasks can enter the system and be executed based on the task execution situation in the previous time slot, that is, the selection decision.

3. The data acquisition task allocation method based on the Internet of Things according to claim 1, characterized in that, To achieve queue stability and minimize the upper bound at each time slot, the following conditions are satisfied:

4. The data acquisition task allocation method based on the Internet of Things according to claim 1, wherein For the objective functions (1)-(4), the queue Q(t) is stable, that is, all sensing tasks entering the sensing platform are executed; when {Q j (t)}, j ∈ {1, …, M} all satisfy the objective functions (1)-(4), the mobile crowdsensing system is stable.

5. The data acquisition task allocation method based on the Internet of Things according to claim 1, wherein With o j (t) varies linearly, so the optimal solution to the task assignment problem is: When Q j (t) - Vβ < 0, o j (t) = O j (t); Otherwise, o j (t) = 0; To minimize the upper bound, maximize as follows: When Q j (t) + o j (t) ≥ N, then x ij (t) = 1, i ∈ {1, …, N}, j ∈ {1, …, M}, that is, in order to maximize the benefits, all participants execute the tasks; When Q j (t) + o j (t) < N, select (Q j (t) + o j (t)) participants to execute the tasks in the queue.

Images