Sampling rate control and transfer calculation resource allocation joint optimization method based on Lyapunov

Abstract

The invention discloses a Lyapunov-based sampling rate control and transfer calculation resource allocation joint optimization method. The method comprises the following steps of S1, collecting an edge server set; s2, calculating the calculation task load of the edge server in the time slot; s3, calculating energy consumption generated during a sensing data calculation task; s4, obtaining a three-dimensional joint optimization formula oriented to the agricultural sensor sampling rate, the edge server preprocessing mode, the edge-cloud communication and the computing resources; s5, defining a function and optimizing the function; based on the scene characteristic and resource coupling relation of the agricultural Internet of Things, three core problems of dynamic regulation and control of the sampling rate of a sensing layer, collaborative allocation of resources of a transfer layer and precise adaptation of full-link service are focused, an integrated resource optimization framework and an efficient algorithm are constructed, and collaborative improvement of data acquisition precision, resource utilization efficiency and service response speed is realized.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology for agricultural Internet of Things (IoT), specifically to a joint optimization method for sampling rate control and computing resource allocation based on Lyapunov. Background Technology

[0002] In the process of integrating digital agriculture with rural revitalization, agricultural IoT is a core technology for solving pain points such as "data silos" in traditional agriculture, and is driving the transformation of production from "experience-driven" to "data-driven". Traditional centralized architecture, due to direct data connection to the cloud, exposes shortcomings such as large bandwidth consumption, high latency, insufficient computing power, and low service accuracy in large-scale scenarios. The "edge-cloud" collaborative architecture formed by the integration of edge computing and agricultural IoT has become a solution.

[0003] In this architecture, edge servers handle near-end real-time tasks, while the cloud center is responsible for all-domain non-real-time tasks, supporting the integration of "sensing-computing-service". These three layers are tightly coupled in terms of resources: the sensing layer needs to balance data collection accuracy and cost, the computing layer needs to dynamically allocate resources, and the service layer needs to match personalized production needs. The level of collaboration determines the system's performance.

[0004] Currently, resource optimization shortcomings still exist across the entire value chain: the sampling strategy at the perception layer is rigid, resource allocation at the computation layer is static, and there is a lack of closed-loop collaboration across the entire chain. These problems restrict system efficiency, increase operation and maintenance costs, and affect the implementation of precision agriculture, becoming key bottlenecks for the large-scale development of smart agriculture. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a joint optimization method for sampling rate control and computational resource allocation based on Lyapunov.

[0006] The technical solution of this invention to solve the above problems is: a joint optimization method for sampling rate control and computational resource allocation based on Lyapunov, comprising the following steps:

[0007] Step S1: Collect the edge server set: The server control system collects the set of all edge servers. ,in, This is the first traffic entry. This is the second traffic stream. This is the third traffic entry. Refers to the k-th traffic item;

[0008] Step S2, compute edge server In the time slot Computational workload: ;

[0009] in This represents the number of CPU cycles required to compute each bit of agricultural sensor data. The computational resource requirements for the preprocessing of agricultural sensing data;

[0010] Step S3: Calculate the energy consumption generated during the sensor data calculation task. ;

[0011] in, For effective switched capacitors in cloud computing centers, This indicates the CPU computing speed of the cloud computing center (unit: cycles / second). For time slots Assigned to edge server The proportion of cloud computing resources, For cloud computing centers in time slots Processing from edge servers The computation latency corresponding to agricultural sensor data computation tasks;

[0012] Step S4: Derive the three-dimensional joint optimization formula for agricultural sensor sampling rate, edge server preprocessing mode, and edge-cloud communication and computing resources;

[0013] Step S5: Define and optimize the function: ;

[0014] in, Indicates time slot Inner The deviation between the service latency of an edge server's agricultural sensor data computing task and the long-term latency requirements of agricultural applications is initially set to... ; This is a precision-loss queue.

[0015] Furthermore, the three-dimensional joint optimization formula in step S4 is:

[0016]

[0017] (a)

[0018] (b)

[0019] (c)

[0020] (d)

[0021] (e)

[0022] (f)

[0023] in, Represents the total number of system cycles. Represents the dimensions of decision variables. The set of decision variables for the i-th sensor at time step t is used to determine... The value of ; This refers to the energy consumption of edge server i within time slot t. It is the bandwidth allocation ratio of edge server i. The cloud computing resource allocation ratio of edge server i, and the adaptive decision variable for agricultural sensor data sampling rate. Selection variables for edge preprocessing mode All are discrete binary variables. and These represent the long-term latency and accuracy requirements of edge server application i, respectively. and Describe the edge servers respectively The latency and accuracy performance;

[0024] In step S4, equation (a) represents the bandwidth allocation constraint; equation (b) represents the cloud computing resource allocation constraint; equation (c) represents the preprocessing mode selection constraint; equation (d) represents the sampling rate decision constraint; and constraints (e) and (f) represent the edge servers for specific agricultural application needs (such as crop soil moisture monitoring and pest and disease early warning). The corresponding long-term average service latency constraints and processing accuracy constraints, among which and Threshold requirements related to agricultural application scenarios.

[0025] Furthermore, Description of edge server The latency performance, i.e. ,in This refers to the time consumed by local computing tasks such as data preprocessing and AI inference on the edge server i, including queuing time and actual execution time.

[0026] Size is Agricultural sensor data computing tasks are performed from edge servers. The transmission latency to the cloud computing center can be expressed as: ,in This represents the size of the agricultural sensor data computation task. Representing time slots Inner edge server Transmission speed between the cloud computing center and the edge server; The energy consumption for transmitting computational task data can be expressed as: ,in For edge servers The preset transmission power, Computing agricultural sensor data from edge servers Transmission latency to the cloud computing center; cloud computing center in time slots Processing from edge servers When performing agricultural sensor data computation tasks, the corresponding computation latency can be expressed as: ,in, This indicates the CPU computing speed of the cloud computing center (unit: cycles / second). The number of CPU cycles required for the cloud center to complete the computation of each bit of agricultural sensor data. For time slots Assigned to edge server The proportion of cloud computing resources, this parameter must meet the following constraints: ,and .

[0027] Furthermore, in step 5,

[0028] The Lyapunov drift penalty function can be expressed as: , ;

[0029] in, The comprehensive optimization objective for each time step is composed of a weighted average of "system stability cost" and "instant performance cost"; For the condition of Lyapunov drift, It is a non-negative constant used to adjust the priority between "queue stability" and "instant performance cost". Let be the conditional expectation operator, representing the state of the system at the current time t. Under the premise of eliminating random factors, the mathematical expectation of the expression within the parentheses is calculated.

[0030] Let a feasible solution to the optimization problem be: ,in Let the set of all feasible solutions be defined; let the time slot be defined. Using solution The probability is The problem can then be transformed into the following equivalent form:

[0031] in, regarded as probability The corresponding weights, in physical terms, represent the adoption of feasible solutions. The global objective function value of the agricultural IoT system is obtained; the problem can be further transformed using the logarithmic-supplier approximation method; firstly, a convex logarithmic-supplier function is introduced. Optimize the objective function for agricultural sensor data tasks An approximation is made, and its definition is as follows: ;in, For positive integers, satisfy the following relationship: ;

[0032] Among them, the upper bound of the approximate performance gap of agricultural Internet of Things systems is Logarithmic-Sum-Exponential Functions for Agricultural IoT Systems The value of is equivalent to the optimal solution to the following optimization problem: ;

[0033] In other words, the problem can be transformed into: ;

[0034] The probability distribution of the optimal solution can then be obtained. Its expression is as follows: .

[0035] The present invention has the following beneficial effects:

[0036] This invention provides a joint optimization method for sampling rate control and computing resource allocation based on Lyapunov. It focuses on the scenario characteristics and resource coupling relationship of agricultural Internet of Things, and addresses three core issues: dynamic regulation of sampling rate at the perception layer, collaborative allocation of resources at the computing layer, and precise adaptation of services across the entire chain. It constructs an integrated resource optimization framework and efficient algorithm to achieve a synergistic improvement in data acquisition accuracy, resource utilization efficiency, and service response speed. Detailed Implementation

[0037] The embodiments of the present invention are described in detail below, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The following embodiments are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0038] A joint optimization method for sampling rate control and computational resource allocation based on Lyapunov includes the following steps:

[0039] Step S1: Collect the edge server set: The server control system collects all edge servers and treats them as a single server set. ,in, This is the first traffic entry. This is the second traffic stream. This is the third traffic entry. Refers to the k-th traffic item;

[0040] Step S2, compute edge server In the time slot Computational workload: ;

[0041] in This represents the number of CPU cycles required to compute each bit of agricultural sensor data. The computational resource requirements for the preprocessing of agricultural sensing data.

[0042] Step S3: Calculate the energy consumption generated during the sensor data calculation task. ;

[0043] in, For effective switched capacitors in cloud computing centers, This indicates the CPU computing speed of the cloud computing center (unit: cycles / second). For time slots Assigned to edge server The proportion of cloud computing resources, For cloud computing centers in time slots Processing from edge servers The computational latency for agricultural sensor data computation tasks is shown. Denoising preprocessing of agricultural sensor data can typically reduce the amount of data required for computational tasks.

[0044] Step S4: Derive the three-dimensional joint optimization formula for agricultural sensor sampling rate, edge server preprocessing mode, and edge-cloud communication and computing resources:

[0045]

[0046] (a)

[0047] (b)

[0048] (c)

[0049] (d)

[0050] (e)

[0051] (f)

[0052] in, Represents the total number of system cycles. Represents the dimensions of decision variables. The set of decision variables for the i-th sensor at time step t is used to determine... The value of . This refers to the energy consumption of edge server i within time slot t. It is the bandwidth allocation ratio of edge server i. The cloud computing resource allocation ratio of edge server i, and the adaptive decision variable for agricultural sensor data sampling rate. Selection variables for edge preprocessing mode All are discrete binary variables. and These represent the long-term latency and accuracy requirements of edge server application i, respectively. and Describe the edge servers respectively The problem requires decision-making within each time slot, aiming to minimize the long-term global energy consumption of the agricultural IoT network under dynamic changes (i.e., the randomness of sensor data arrival and field channel state). The challenges of solving this problem lie mainly in the following two aspects: 1) In practical agricultural IoT systems, the statistical characteristics of field time-varying channel state (affected by terrain and weather) and sensor data arrival rate are often difficult or even impossible to obtain. This is determined by the massive historical information storage and processing pressure caused by the dispersed deployment of nodes in farmland and complex environmental interference; 2) As the number of farmland edge servers and agricultural sensing devices increases, the action space and state space of the problem expand rapidly, leading to an exponential increase in computational complexity, making traditional dynamic programming methods no longer applicable.

[0053] In step S4, equation (a) represents the bandwidth allocation constraint; equation (b) represents the cloud computing resource allocation constraint; equation (c) represents the preprocessing mode selection constraint; and equation (d) represents the sampling rate decision constraint. Constraints (e) and (f) represent the constraints for each edge server designed for specific agricultural applications (such as crop moisture monitoring and pest and disease early warning). The corresponding long-term average service latency constraints and processing accuracy constraints, among which and These are threshold requirements relevant to agricultural applications. It should be noted that... Description of edge server The latency performance, i.e. ,in This refers to the time consumed by local computing tasks such as data preprocessing and AI inference on the edge server i, including queuing time and actual execution time.

[0054] Size is Agricultural sensor data computing tasks are performed from edge servers. The transmission latency to the cloud computing center can be expressed as: ,in This represents the size of the agricultural sensor data computation task. Representing time slots Inner edge server Transmission speed between the cloud computing center and the edge server; The energy consumption for transmitting computational task data can be expressed as: ,in For edge servers The preset transmission power, Computing agricultural sensor data from edge servers Transmission latency to the cloud computing center; cloud computing center in time slots Processing from edge servers When performing agricultural sensor data computation tasks, the corresponding computation latency can be expressed as: ,in, This indicates the CPU computing speed of the cloud computing center (unit: cycles / second). The number of CPU cycles required for the cloud center to complete the computation of each bit of agricultural sensor data. For time slots Assigned to edge server The proportion of cloud computing resources, this parameter must meet the following constraints: ,and .

[0055] Step S5: Define and optimize the function: ;

[0056] in, Indicates time slot Inner The deviation between the service latency of an edge server's agricultural sensor data computing task and the long-term latency requirements of agricultural applications is initially set to... . This is a precision-loss queue.

[0057] The Lyapunov drift penalty function can be expressed as: , .in, The overall optimization objective for each time step is composed of a weighted average of "system stability cost" and "instant performance cost". For the condition of Lyapunov drift, It is a non-negative constant used to adjust the priority between "queue stability" and "instant performance cost". Let be the conditional expectation operator, representing the state of the system at the current time t. Under the premise of eliminating random factors, the mathematical expectation of the expression within the parentheses is calculated.

[0058] The upper bound of the drift penalty function given by Theorem 1 is minimized within each time slot. This strategy can maintain the processing accuracy and service latency of agricultural sensor data at the expected levels (such as centimeter-level accuracy for soil moisture monitoring and minute-level latency for pest and disease early warning) while minimizing the total energy consumption of farmland edge servers and cloud centers.

[0059] In step S5, the objective function of the optimization problem corresponds to the drift penalty function. The right side, in each time slot The optimal decisions for agricultural sensor sampling rate adaptation, edge preprocessing mode selection, and cloud-edge resource management can be obtained. Subsequently, the accuracy loss queue and latency excess queue are updated to provide state support for optimization in the next time slot. To solve the problem, a Markov approximation method is introduced to solve the problem in polynomial time, and the performance gap of the obtained approximate solution has a verifiable theoretical bound. This method treats the agricultural IoT network configuration scheme (including the agricultural sensor sampling rate adaptation strategy, the edge node preprocessing mode selection scheme, and the cloud center agricultural data computation and bandwidth allocation strategy) as the state of a time-reversible continuous-time Markov chain, and it can be proven that this Markov chain converges to a stationary state after a finite number of iterations.

[0060] Describe a feasible solution to the optimization problem (i.e., the objective function for minimizing agricultural sensor data tasks). The solution is ,in Let be the set of all feasible solutions. Let time slots be defined. Using solution The probability is The problem can then be transformed into the following equivalent form:

[0061] in, regarded as probability The corresponding weights, in physical terms, represent the adoption of feasible solutions. The global objective function value of the agricultural IoT system. This problem can be further transformed using the logarithmic-exponential approximation method. First, a convex logarithmic-exponential function is introduced. Optimize the objective function for agricultural sensor data tasks An approximation is made, and its definition is as follows: .in, For positive integers, satisfy the following relationship: .

[0062] Among them, the upper bound of the approximate performance gap of agricultural Internet of Things systems is Logarithmic-Sum-Exponential Functions for Agricultural IoT Systems The value of is equivalent to the optimal solution to the following optimization problem: .

[0063] In other words, the problem can be transformed into: The probability distribution of the optimal solution can be obtained from this. Its expression is as follows: .

[0064] The detailed descriptions listed above are merely specific illustrations of feasible embodiments of the present invention and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A joint optimization method for sampling rate control and computational resource allocation based on Lyapunov, characterized in that: Includes the following steps: Step S1: Collect the edge server set: The server control system collects the set of all edge servers. ,in, This is the first traffic entry. This is the second traffic entry. This is the third traffic entry. Refers to the k-th traffic item; Step S2, compute edge server In the time slot Computational workload: ; in This represents the number of CPU cycles required to compute each bit of agricultural sensor data. The computational resource requirements for the preprocessing of agricultural sensing data; Step S3: Calculate the energy consumption generated during the sensor data calculation task. ; in, For effective switched capacitors in cloud computing centers, This indicates the CPU computing speed of the cloud computing center (unit: cycles / second). For time slots Assigned to edge server The proportion of cloud computing resources, For cloud computing centers in time slots Processing from edge servers The computation latency corresponding to agricultural sensor data computation tasks; Step S4: Derive the three-dimensional joint optimization formula for agricultural sensor sampling rate, edge server preprocessing mode, and edge-cloud communication and computing resources; Step S5: Define and optimize the function: ; in, Indicates time slot Inner The deviation between the service latency of an edge server's agricultural sensor data computing task and the long-term latency requirements of agricultural applications is initially set to... ; This is a precision-loss queue.

2. The method for joint optimization of sampling rate control and computational resource allocation based on Lyapunov as described in claim 1, characterized in that: The three-dimensional joint optimization formula in step S4 is: ； (a)； (b)； (c)； (d)； (e)； (f)； in, Represents the total number of system cycles. Represents the dimensions of decision variables. The set of decision variables for the i-th sensor at time step t is used to determine... The value of ; This refers to the energy consumption of edge server i within time slot t. It is the bandwidth allocation ratio of edge server i. The cloud computing resource allocation ratio of edge server i, and the adaptive decision variable for agricultural sensor data sampling rate. Selection variables for edge preprocessing mode All are discrete binary variables. and These represent the long-term latency and accuracy requirements for edge server application i, respectively. and Describe the edge server respectively The latency and accuracy performance; Equation (a) represents the bandwidth allocation constraint; Equation (b) represents the cloud computing resource allocation constraint; Equation (c) represents the preprocessing mode selection constraint; Equation (d) represents the sampling rate decision constraint; Constraints (e) and (f) represent the edge servers for specific agricultural application needs (such as crop soil moisture monitoring and pest and disease early warning). The corresponding long-term average service latency constraints and processing accuracy constraints, among which and Threshold requirements related to agricultural application scenarios.

3. The joint optimization method for sampling rate control and computational resource allocation based on Lyapunov as described in claim 1, characterized in that: Description of edge server The latency performance, i.e. ; Size is Agricultural sensor data computing tasks are performed from edge servers. The transmission latency to the cloud computing center can be expressed as: ,in This represents the size of the agricultural sensor data computation task. Representing time slots Inner edge server Transmission speed between the cloud computing center and the edge server; The energy consumption for transmitting computational task data can be expressed as: ,in For edge servers The preset transmission power, Computing agricultural sensor data from edge servers Transmission latency to the cloud computing center; cloud computing center in time slots Processing from edge servers When performing agricultural sensor data computation tasks, the corresponding computation latency can be expressed as: ,in, This indicates the CPU computing speed of the cloud computing center (unit: cycles / second). The number of CPU cycles required for the cloud center to complete the computation of each bit of agricultural sensor data. For time slots Assigned to edge server The proportion of cloud computing resources, this parameter must meet the following constraints: ,and .

4. The method for joint optimization of sampling rate control and computational resource allocation based on Lyapunov as described in claim 1, characterized in that: In step 5, The Lyapunov drift penalty function can be expressed as: , ; in, The overall optimization objective for each time step is composed of a weighted average of "system stability cost" and "instant performance cost"; For the condition of Lyapunov drift, It is a non-negative constant used to adjust the priority between "queue stability" and "instant performance cost". Let be the conditional expectation operator, representing the state of the system at the current time t. Under the premise of eliminating random factors, the mathematical expectation of the expression within the parentheses is calculated. Let a feasible solution to the optimization problem be: ,in Let the set of all feasible solutions be defined; let the time slot be defined. Using solution The probability is The problem can then be transformed into the following equivalent form: ; in, regarded as probability The corresponding weights, in physical terms, represent the adoption of feasible solutions. The global objective function value of the agricultural IoT system is obtained; the problem can be further transformed using the logarithmic-supplier approximation method; firstly, a convex logarithmic-supplier function is introduced. Optimize the objective function for agricultural sensor data tasks An approximation is made, and its definition is as follows: ;in, For positive integers, satisfy the following relationship: ; Among them, the upper bound of the approximate performance gap of agricultural Internet of Things systems is Logarithmic-Sum-Exponential Functions for Agricultural IoT Systems The value of is equivalent to the optimal solution to the following optimization problem: ; It can be converted into: ; The probability distribution of the optimal solution can then be obtained. Its expression is as follows: .

Images