End-side cloud resource and task real-time dynamic matching method and system based on adaptive evolution algorithm

Abstract

The invention relates to the field of computers, in particular to an end-side cloud resource and task real-time dynamic matching method and system based on a self-adaptive evolution algorithm, and the method comprises the steps: A, data collection and system modeling: obtaining state environment data of a terminal side, an edge side and a cloud side, and defining calculation-intensive, delay-sensitive and data-local tasks according to the obtained data; step B, designing an adaptive evolution algorithm: adaptively adjusting an evolution parameter crossover rate and a mutation rate in real time according to the environmental change of the end-side cloud equipment, and calculating a fitness value; step C, layered optimization and real-time decision making: layered optimization and real-time matching decision making are carried out according to the fitness value of dynamic matching of the end edge cloud equipment resources and the tasks in the step B; d, performing dynamic adjustment and fault-tolerant processing; the method can adapt to dynamic matching of end-side cloud equipment resources and tasks, and has high robustness for network fluctuation.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular to a method and system for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm. Background Art

[0002] With the development of the Internet of Things (IoT), 5G, and artificial intelligence (AI), end-edge-cloud collaborative computing and evolution have become key architectures supporting low-latency, energy-efficient computing. However, existing technologies have the problem of static resource allocation. Traditional methods (such as heuristic rules and static optimization) cannot adapt to scenarios such as dynamic device joining / exiting and network fluctuations. There are also problems with multi-objective conflicts in the end-edge-cloud: optimization goals such as latency, energy consumption, and cost are difficult to meet simultaneously, and there is a lack of adaptive trade-off mechanisms. The end-edge-cloud also has poor real-time performance: traditional resource and task matching methods converge slowly and cannot meet real-time requirements. There is an urgent need for dynamic optimization methods for end-edge-cloud resource and task collaboration to achieve real-time dynamic matching of task offloading, resource allocation, and energy consumption optimization. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the existing technology and propose a real-time dynamic matching method and system for end-edge cloud resources and tasks based on an adaptive evolutionary algorithm. The present invention can adapt to the dynamic matching of end-edge cloud device resources and tasks, and is highly robust to network fluctuations.

[0004] The purpose of the present invention is to achieve the following: a method for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm, comprising the following steps:

[0005] Step A: Data collection and system modeling:

[0006] Obtain state and environmental data on the terminal, edge, and cloud sides, and define compute-intensive, latency-sensitive, and data-local tasks based on the acquired data;

[0007] Step B: Adaptive evolutionary algorithm design:

[0008] According to the changes in the edge-cloud device environment, the evolution parameter crossover rate and mutation rate are adaptively adjusted in real time, the fitness value is calculated, and the algorithm convergence speed is improved;

[0009] Step C: Hierarchical optimization and real-time decision-making:

[0010] Based on the fitness values of the dynamic matching of end-edge-cloud device resources and tasks in step B, hierarchical optimization and real-time matching decisions are performed. Lightweight rules are executed on the terminal side, and the main adaptive evolutionary algorithm is deployed and run on the edge side. The offload strategy is updated every 5 seconds.

[0011] Step D: Dynamic adjustment and fault tolerance:

[0012] If the device status is offline, the task is migrated to the nearby edge node or cloud; if the device status is network interrupted, it is downgraded to the end side for local processing.

[0013] Preferably, in step A, data collection includes: terminal side: collecting terminal device computing power, power, and network status; edge side: monitoring edge computing server load, GPU / CPU utilization, and task queue; cloud side: obtaining global resources and task status as well as consumption costs.

[0014] Preferably, the step B specifically includes the following steps:

[0015] Step B1: Determine a dynamic encoding scheme: Use binary encoding to represent the task offload location, where 00 = client-side execution, 01 = edge-side execution, 10 = cloud-side execution, and 11 is a vacant position. Use real number encoding to represent the resource allocation ratio. Each resource has a real number gene, and the length can dynamically increase or decrease with the resource type. GPU resources can be added by expanding the dimension.

[0016] Step B2: Define the fitness function: Fitness = αT + βE + γC

[0017] T is the total task delay, E is the total energy consumption on the edge, and C is the cloud service cost. A user preference model is introduced, and the weights α, β, and γ are dynamically adjusted based on the specific user's edge-cloud application scenario.

[0018] Step B3: Online adaptive evolutionary operation mechanism: According to the state of the edge cloud device, the adaptive evolutionary algorithm is used to dynamically adjust the evolutionary operation factor crossover rate p c , mutation rate p m , p c and p m Automatically changes with fitness. When the fitness of individuals in the population tends to be consistent or locally optimal, p c and p m Increase, and when the group fitness is more dispersed, make p c and p m reduce;

[0019] Step B4: Fitness value calculation: Calculate the fitness value of the dynamic matching of end-edge cloud device resources and tasks, and select the solution with high fitness based on the selection strategy; enter the next round of evolution operation until the set fitness threshold or number of iterations is met, and terminate the evolution operation.

[0020] Preferably, in step B3, p c and p m Automatic adjustment is performed according to the following formula:

[0021]

[0022] Among them, f max is the maximum fitness value in the group, f avg is the average fitness value of each generation group, f￠ is the larger fitness value of the two individuals to be crossed, f is the fitness value of the mutant individual, p c1 is the maximum crossover probability, p m1 is the maximum mutation probability.

[0023] Preferably, the crossover rate is increased when the terminal device network is stable, and the mutation rate is increased when the device battery is low; mutation is triggered according to the end-edge cloud device environment, and when the edge is overloaded, the unloading gene of the associated task is forced to mutate.

[0024] Preferably, the global parameters are optimized on the cloud side for a long time, and the values of the weights α, β, and γ are updated and adjusted.

[0025] A system for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm, comprising a processor, wherein the processor executes a computer program to implement the steps of a method for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm as described in any one of claims 1 to 6.

[0026] Beneficial effects of the present invention: The present invention realizes real-time task offloading decision-making, resource adaptive matching, and multi-objective resource optimization (delay, energy consumption, and cost trade-off) of the three-layer structure of end, edge, and cloud through dynamic coding, adaptive evolution, online adaptation, and hierarchical optimization. It can adapt to the dynamic matching of end-edge cloud device resources and tasks, and is highly robust to network fluctuations. Compared with traditional evolutionary algorithms and static resource allocation methods, the algorithm of the present invention reduces the average delay from 122ms to 88ms, a delay reduction of about 28%, and the end-side energy consumption decreases by about 30%. At the same time, the real-time decision-making of dynamic matching of resources and tasks is improved by 23%, and the dynamic adaptation performance is significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Processing flow chart of the present invention;

[0028] Figure 2 This is a diagram of the device-edge-cloud collaborative architecture;

[0029] Figure 3 This is the flow chart of the adaptive evolution algorithm. DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] Example 1: Figure 1 、 Figure 2 、 Figure 3 As shown, a method for real-time dynamic matching of edge-cloud resources and tasks based on an adaptive evolutionary algorithm includes the following steps:

[0032] Step A: Data collection and system modeling:

[0033] Obtain state and environmental data on the terminal, edge, and cloud sides, and define compute-intensive, latency-sensitive, and data-local tasks based on the acquired data;

[0034] Specifically, data collection includes: terminal side: collecting terminal device computing power, power consumption, and network status; edge side: monitoring edge computing server load, GPU / CPU utilization, and task queue; cloud side: obtaining global resources and task status as well as consumption costs.

[0035] Step B: Adaptive evolutionary algorithm design:

[0036] According to the changes in the edge-cloud device environment, the evolution parameter crossover rate and mutation rate are adaptively adjusted in real time, the fitness value is calculated, and the algorithm convergence speed is improved;

[0037] Specifically, the method includes the following steps:

[0038] Step B1: Determine a dynamic encoding scheme: Use binary encoding to represent the task offload location, where 00 = end-side execution, 01 = edge-side execution, 10 = cloud-side execution, and 11 is a vacant position. The encoding length can be dynamically expanded if the end-edge-cloud layer needs to be increased. Use real number encoding to represent the resource allocation ratio (CPU utilization, bandwidth allocation). Each resource has a real number gene, and the length can be dynamically increased or decreased depending on the resource type. GPU resources can be added by expanding the dimension.

[0039] Step B2: Define the fitness function: Fitness = αT + βE + γC

[0040] T is the total task delay (including transmission delay and computation delay), E is the total energy consumption on the edge, and C is the cloud service cost. A user preference model is introduced, and the weights α, β, and γ are dynamically adjusted based on the specific user's edge-cloud application scenario (e.g., medical applications prioritize latency, while IoT prioritizes energy consumption).

[0041] Step B3: Online adaptive evolutionary operation mechanism: According to the state of the edge cloud device, the adaptive evolutionary algorithm is used to dynamically adjust the evolutionary operation factor crossover rate p c , mutation rate p m , p c and p m Automatically changes with fitness. When the fitness of individuals in the population tends to be consistent or locally optimal, p c and p m Increase, and when the group fitness is more dispersed, make p c and p m decreases; therefore, the adaptively adjusted p c and p m Can provide the best p relative to a certain solution c and p m , which not only ensures the diversity of the population, but also ensures the convergence of the algorithm.

[0042] p c and p m Automatic adjustment is performed according to the following formula:

[0043]

[0044] Among them, f max is the maximum fitness value in the group, f avg is the average fitness value of each generation group, f￠ is the larger fitness value of the two individuals to be crossed, f is the fitness value of the mutant individual, p c1 is the maximum crossover probability, p m1 The maximum mutation probability is maximal. The crossover rate is increased when the terminal device network is stable, and the mutation rate is increased when the device battery is low. Mutations are triggered based on the edge-cloud device environment. When the edge is overloaded, the offload genes of the associated tasks are forced to mutate.

[0045] Step B4: Fitness value calculation: Calculate the fitness value of the dynamic matching of end-edge cloud device resources and tasks, and select the solution with high fitness based on the selection strategy; enter the next round of evolution operation until the set fitness threshold or number of iterations is met, and terminate the evolution operation.

[0046] Step C: Hierarchical optimization and real-time decision-making:

[0047] Based on the fitness value of the dynamic matching of end-edge-cloud device resources and tasks in step B, hierarchical optimization and real-time matching decisions are performed. Lightweight rules are executed on the terminal side (unloading is rejected when the battery level is <20%), and the main adaptive evolutionary algorithm is deployed and run on the edge side. The unloading strategy is updated every 5 seconds. Global parameters are optimized on the cloud side for a long time, and the values of weights α, β, and γ are updated and adjusted.

[0048] Step D: Dynamic adjustment and fault tolerance:

[0049] If the device status is offline, the task is migrated to the nearby edge node or cloud; if the device status is network interrupted, it is downgraded to the end side for local processing.

[0050] like Figure 3 As shown in the figure, the adaptive evolutionary algorithm process is as follows: according to the collected parameters of the edge-cloud device environment change (such as device power and network status), the evolution population is initialized, and resource and task matching evolution is performed according to the edge-cloud device topology structure. Through crossover and mutation operations, the edge-cloud topology node with the highest fitness is selected. If the resource and task matching conditions are met, the result is output; if the optimal solution conditions are not met, the suboptimal solution group is selected as the new evolutionary population according to the roulette rule. At the same time, the evolution factor is automatically adjusted according to the adaptive evolutionary algorithm, the evolutionary population is updated, and new edge-cloud device topology nodes are continued to be selected for fitness solution. The next round of evolutionary operations is performed until the optimal matching conditions of resource and task fitness are met, the evolutionary operation is ended, and the result is output.

[0051] This method achieves low-latency, energy-efficient distributed resource and task scheduling through dynamic coding, adaptive evolution, online adaptation, and hierarchical optimization. Compared to traditional methods, it significantly reduces latency and energy consumption, and offers strong robustness. It is suitable for scenarios requiring adaptive matching of end-edge cloud resources and tasks, such as the Industrial Internet of Things, smart healthcare, and autonomous driving.

[0052] The performance improvement was demonstrated through simulation and field data. By using an adaptive evolutionary algorithm to search for end-edge cloud resource and task matching paths, the optimal end-edge cloud node resource matching latency was reduced by 12%, node search energy consumption was reduced by 17%, and the search algorithm convergence speed was increased by 20% compared to traditional evolutionary algorithms and 14% compared to particle swarm search algorithms.

[0053] The present invention has the following characteristics:

[0054] (1) Dynamic coding and online adaptation mechanism: According to the changes in the edge-cloud device environment (such as device power and network status), the evolution parameters (crossover rate and mutation rate) are adaptively adjusted in real time to improve the algorithm convergence speed.

[0055] (2) Hierarchical optimization architecture: The hierarchical optimization architecture of "lightweight rules on the end side + primary optimization algorithms on the edge side + global parameter adjustment on the cloud side" enables efficient distributed decision-making.

[0056] (3) Dynamic trade-offs among multiple objectives: Introducing a user preference model to dynamically adjust the weights of optimization objectives (e.g., medical applications prioritize latency, IoT prioritizes energy consumption).

[0057] (4) Incremental evolution and memory mechanism: Only the evolutionary genes of the evolving individuals affected by environmental changes are re-optimized to reduce computational overhead.

[0058] A system for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm includes a processor, which executes a computer program to implement the steps of the above-mentioned method for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm, and can achieve the same effect as the method for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm.

[0059] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for real-time dynamic matching of edge-cloud resources and tasks based on an adaptive evolutionary algorithm, characterized by: The following steps are involved: Step A: Data collection and system modeling: Obtain state and environmental data on the terminal, edge, and cloud sides, and define compute-intensive, latency-sensitive, and data-local tasks based on the acquired data; Step B: Adaptive evolutionary algorithm design: According to the changes in the edge-cloud device environment, the evolution parameter crossover rate and mutation rate are adaptively adjusted in real time, the fitness value is calculated, and the algorithm convergence speed is improved; Step C: Hierarchical optimization and real-time decision-making: Based on the fitness values of the dynamic matching of end-edge-cloud device resources and tasks in step B, hierarchical optimization and real-time matching decisions are performed. Lightweight rules are executed on the terminal side, and the main adaptive evolutionary algorithm is deployed and run on the edge side. The offload strategy is updated every 5 seconds. Step D: Dynamic adjustment and fault tolerance: If the device status is offline, the task is migrated to the nearby edge node or cloud; if the device status is network interrupted, it is downgraded to the end side for local processing.

2. The method for real-time dynamic matching of edge-cloud resources and tasks based on an adaptive evolutionary algorithm according to claim 1 is characterized by: In step A, data collection includes: terminal side: collecting terminal device computing power, power, and network status; edge side: monitoring edge computing server load, GPU / CPU utilization, and task queue; cloud side: obtaining global resources and task status as well as consumption costs.

3. The method for real-time dynamic matching of edge-cloud resources and tasks based on an adaptive evolutionary algorithm according to claim 1 is characterized by: The step B specifically includes the following steps: Step B1: Determine a dynamic encoding scheme: Use binary encoding to represent the task offload location, where 00 = client-side execution, 01 = edge-side execution, 10 = cloud-side execution, and 11 is a vacant position. Use real number encoding to represent the resource allocation ratio. Each resource has a real number gene, and the length can dynamically increase or decrease with the resource type. GPU resources can be added by expanding the dimension. Step B2: Define fitness function: Fitness = αT + βE + γC T is the total task delay, E is the total energy consumption on the edge, and C is the cloud service cost. A user preference model is introduced, and the weights α, β, and γ are dynamically adjusted based on the specific user's edge-cloud application scenario. Step B3: Online adaptive evolutionary operation mechanism: According to the state of the edge cloud device, the adaptive evolutionary algorithm is used to dynamically adjust the evolutionary operation factor crossover rate p c , mutation rate p m , p c and p m Automatically changes with fitness. When the fitness of individuals in the population tends to be consistent or locally optimal, p c and p m Increase, and when the group fitness is more dispersed, make p c and p m reduce; Step B4: Fitness value calculation: Calculate the fitness value of the dynamic matching of end-edge cloud device resources and tasks, and select the solution with high fitness based on the selection strategy; enter the next round of evolution operation until the set fitness threshold or number of iterations is met, and terminate the evolution operation.

4. The method for real-time dynamic matching of edge-to-edge cloud resources and tasks based on an adaptive evolutionary algorithm according to claim 3 is characterized by: In step B3, p c and p m Automatic adjustment is performed according to the following formula: Among them, f max is the maximum fitness value in the group, f avg is the average fitness value of each generation group, f￠ is the larger fitness value of the two individuals to be crossed, f is the fitness value of the mutant individual, p c1 is the maximum crossover probability, p m1 is the maximum mutation probability.

5. The method for real-time dynamic matching of edge-to-edge cloud resources and tasks based on an adaptive evolutionary algorithm according to claim 4 is characterized by: The crossover rate is increased when the terminal device network is stable, and the mutation rate is increased when the device battery is low. Mutations are triggered based on the end-edge cloud device environment. When the edge is overloaded, the offload genes of the associated tasks are forced to mutate.

6. The method for real-time dynamic matching of edge-to-edge cloud resources and tasks based on an adaptive evolutionary algorithm according to claim 3 is characterized by: On the cloud side, global parameters are optimized over a long period of time, and the values of weights α, β, and γ are updated and adjusted.

7. A real-time dynamic matching system for edge-cloud resources and tasks based on an adaptive evolutionary algorithm, comprising a processor, characterized in that: The processor executes a computer program to implement the steps of the method for real-time dynamic matching of end-edge cloud resources and tasks based on an adaptive evolutionary algorithm as described in any one of claims 1-6.

Images