Intelligent security edge computing resource allocation method and system

Abstract

The invention discloses an intelligent security edge computing resource allocation method and system, and belongs to the technical field of intelligent security edge computing and swarm intelligence optimization, and the method comprises the following steps: S1, constructing a multi-target resource scheduling optimization model of an optimization target; s2, initializing mixed population parameters; s3, adaptive parameter adjustment based on nonlinear attenuation; s4, fusing mixed position updating guided by a velocity vector, and generating a candidate resource allocation scheme; s5, calculating the fitness of the candidate resource allocation scheme, updating the individual position, the individual historical optimal position and the global optimal position by adopting a greedy strategy until the maximum number of iterations is reached, and outputting the optimal resource allocation scheme corresponding to the global optimal position; according to the scheme, quick response of tasks to tasks and the requirement for the resource utilization rate of the intelligent security edge computing system can be met at the same time, and the actual requirements of the intelligent security system for high reliability and low delay are met.

Description

Technical Field

[0001] This invention relates to the field of intelligent security edge computing and swarm intelligence optimization technology, specifically to an intelligent security edge computing resource allocation method and system. Background Technology

[0002] With the development of the Internet of Things (IoT) and 5G communication technologies, various mobile and fixed terminal devices are connected to the Internet, resulting in high-concurrency computing tasks. Traditional cloud computing uploads data from various terminals to the cloud, where it performs unified computing and processing, and finally distributes the processed data back to the corresponding terminals. This leads to problems such as limited public network bandwidth and congestion, unavoidable latency, and unsuitability for ultra-low latency scenarios. In contrast, in the context of Mobile Edge Computing (MEC), computing tasks are distributed to various terminals for processing, which can reduce the computing and storage resources of network edge nodes such as base stations or roadside units, shorten task response time, and reduce the pressure on the core network. Therefore, in scenarios with high concurrency and low latency, how to ensure the shortest task completion time while also ensuring load balancing of edge nodes is a major problem that urgently needs to be solved.

[0003] Especially in scenarios involving high concurrency and latency sensitivity, such as large-scale urban video surveillance, real-time public safety early warning systems, and facial recognition access control terminals, terminal devices generate concurrent task requests. If the resource allocation strategy is unreasonable, nodes with weak computing power will be idle, while some nodes with strong computing power will experience task queuing and timeouts due to excessive load, resulting in long task response times. This load imbalance not only severely impacts the user experience but also leads to low overall system resource utilization and may even cause service crashes.

[0004] Currently, metaheuristic algorithms (such as genetic algorithms, particle swarm optimization (PSO), and artificial fish swarm algorithm (AFSA) are commonly used to solve NP-hard problems. While AFSA has strong global search capabilities and robustness and has a certain theoretical basis, it has shortcomings in practical applications. 1. Low search efficiency due to fixed parameters: In the traditional AFSA algorithm, the field of view (Visual) and the step size (Step) are both fixed values. In the initial stage of the algorithm's iteration, the small field of view will affect the global search speed, resulting in a low speed. However, in the later stage of the iteration, the step size may be too large, which will cause the algorithm to easily get trapped in local optima and fail to guarantee good convergence accuracy.

[0005] 2. Prone to getting trapped in local optima: Although Particle Swarm Optimization (PSO) has a high convergence speed, it is prone to "premature convergence" due to the loss of population diversity. This can lead to some edge nodes being overloaded while other nodes are idle, thus failing to achieve true load balancing.

[0006] 3. Lack of Adaptive Mechanisms: Current improved algorithms are mostly single-method implementations, lacking state optimization methods for the iterative process, making it difficult to effectively balance "global exploration" and "local development." Therefore, it is necessary to develop edge computing resource allocation algorithms that can adaptively adjust search strategies and balance convergence speed and solution accuracy. Summary of the Invention

[0007] To address the aforementioned shortcomings of existing technologies, this invention provides an intelligent security edge computing resource allocation method and system, which solves the problems of slow algorithm convergence speed, low optimization accuracy, susceptibility to local optima, and uneven node load distribution in existing edge computing resource scheduling.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Firstly, a method for allocating intelligent security edge computing resources is provided, which includes the following steps: S1: Obtain the task set and edge node set of the intelligent security edge computing system, and construct a multi-objective resource scheduling optimization model with task completion time, total system energy consumption and node load variance as optimization objectives; S2: Initialize the mixed population parameters, including population size, maximum number of iterations, initial values ​​of artificial fish's field of vision and movement step size, initial values ​​of individual historical best position and global best position, randomly generated initial artificial fish position, and initial velocity of particle swarm optimization algorithm; S3: Enter the iteration process and dynamically adjust the artificial fish's field of vision and movement step size using a nonlinear adaptive strategy based on the current iteration number; S4: Perform AFSA behavior selection for each artificial fish, and introduce the velocity vector guidance mechanism of particle swarm algorithm in the position update. Combine the individual historical optimal solution and the global optimal solution to correct the movement direction of the artificial fish and generate candidate resource allocation schemes. S5: Calculate the fitness of candidate resource allocation schemes, update individual positions, individual historical best positions, and global best positions using a greedy strategy. If the maximum number of iterations has not been reached, return to step S3. Continue until the maximum number of iterations is reached, and output the optimal resource allocation scheme corresponding to the global best position.

[0009] Secondly, an intelligent security edge computing resource allocation system is provided, based on an intelligent security edge computing resource allocation method, which includes: a task perception module: real-time collection of security task data and edge node status; a parameter optimization module: dynamic adjustment of algorithm parameters according to the iteration progress; a fusion computing module: running a hybrid artificial fish swarm and particle swarm algorithm to generate a mapping scheme for the optimal resource allocation scheme; and a scheduling execution module: receiving the mapping scheme and distributing tasks to the corresponding edge nodes for execution.

[0010] The beneficial effects of this invention are as follows: 1. This scheme accelerates the convergence speed of the artificial fish swarm algorithm and avoids premature convergence; it introduces the velocity vector guidance mechanism of the particle swarm algorithm into the artificial fish swarm algorithm, uses the information of the global optimum to guide the search direction, overcomes the defect of blind search in the artificial fish swarm algorithm, accelerates the convergence speed of the algorithm, and solves the problem that the particle swarm algorithm is prone to getting trapped in local optima.

[0011] 2. This scheme balances the capabilities of global exploration and local development. It uses the nonlinear exponential decay method of Visual and Step, which allows the population search process to adaptively adjust the global exploration and local development processes according to the algorithm's own adaptability. The fixed step size in the early stage ensures that the population can quickly cover the entire solution space. In the later stage, the step size is adjusted for the area near the optimal solution, which is conducive to the algorithm's high-precision mining of the area near the optimal solution and obtaining a better resource allocation scheme.

[0012] 3. This solution achieves load balancing of edge nodes by incorporating node load variance into a multi-objective optimization function. In addition to reducing the total system latency, it can also better ensure that there are no overloaded nodes and that there are no idle nodes after an overloaded node, thereby improving the resource utilization and stability of the edge computing system. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The above and other objects, features, and advantages of the present invention will become clearer through the accompanying drawings. The same reference numerals indicate the same parts in all the drawings. The drawings are not intentionally drawn to scale to actual dimensions; the focus is on illustrating the main points of the invention.

[0014] Figure 1 This is a flowchart illustrating the resource allocation method for edge computing in intelligent security.

[0015] Figure 2 This is a comparison chart of convergence performance in the simulation experiment.

[0016] Figure 3 This is a comparison chart of load balancing in the simulation experiment.

[0017] Figure 4 This is a simulation comparison of resource scheduling between ARM nodes and FPGA nodes in edge computing based on an improved algorithm. Detailed Implementation

[0018] 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, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0021] Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0022] like Figure 1 As shown, the intelligent security edge computing resource allocation method of this solution includes the following steps: S1: Obtain the task set and edge node set of the intelligent security edge computing system, and construct a multi-objective resource scheduling optimization model with task completion time, total system energy consumption and node load variance as optimization objectives; Specifically, the model construction includes: assuming the edge computing system contains a task set T = {t1, t2, ..., t} with m tasks to be assigned. m} and a set E = {e1, e2, ..., e} of n edge server nodes n}. Each task t i With data size D i And computational complexity C i Each edge node e j With CPU processing power P j and current bandwidth resources Bj ; To balance user experience and system stability, an expression for the fitness of a multi-objective resource scheduling optimization model is established, aiming to minimize the weighted sum of the following three indicators, as expressed in the following expression:

[0023] in, X Let the vector represent the current resource allocation scheme. ω 1. ω 2. ω 3 represents the weighting coefficient, which in this embodiment is taken as 0.5, 0.3, and 0.2 respectively, and ω1+ω2+ω3= 1. T makespan , E total , load These represent the total completion time of all tasks, the total energy consumption of the system in processing all tasks, and the mean square deviation of the load on each edge node, respectively.

[0024] S2: Initialize the mixed population parameters, including population size, maximum number of iterations, initial values ​​of artificial fish's field of vision and movement step size, initial values ​​of individual historical best position and global best position, randomly generated initial artificial fish position, and initial velocity of particle swarm optimization algorithm; In this embodiment, the population size N=50 and the maximum number of iterations G are set. max =200; Encoding strategy: Integer encoding is used; Position X of each artificial fish i Given a vector of length m, the value at the k-th position represents the edge node number (range 1 to n) to which the k-th task is assigned; Initial Visual start =10, initial step size Step start =2; Initialize Particle Swarm Optimization (PSO) parameters: learning factor c1=c2=2.0, inertia weight ω=0.8, and randomly generate initial velocity vectors. V i .

[0025] S3: Adaptive Parameter Adjustment Based on Nonlinear Decay: Entering the iterative process, a nonlinear adaptive strategy is used to dynamically adjust the artificial fish's field of vision and movement step size based on the current iteration number. This ensures the algorithm has strong global exploration capabilities in the early stages and strong local exploration capabilities in the later stages; Field of Vision Visual ( t ) and moving step size Step ( t The formula for calculating ) is:

[0026] Among them, the starting value of the visible range Visual start and initial step size Step start The initial value; the minimum visible range. Visual min and minimum step size Step min This is the default value; μ This is the attenuation adjustment coefficient; t This indicates the number of steps in this iteration. G max Indicates the maximum number of iterations; sets the Visual... min =1.0, Step min =0.1, attenuation adjustment coefficient μ=3.0.

[0027] S4: Hybrid position update guided by velocity vector integration: AFSA behavior selection is performed on each artificial fish, and the velocity vector guidance mechanism of particle swarm algorithm is introduced into the position update. The movement direction of the artificial fish is corrected by combining the individual historical optimal solution and the global optimal solution, and candidate resource allocation schemes are generated. The method by which each artificial fish performs AFSA behavior selection is as follows: based on the crowding factor, the artificial fish autonomously chooses to perform foraging, grouping, or tail-chasing behaviors, and calculates the candidate displacement based on AFSA. X AFSA The calculation formula is as follows:

[0028] in, Rand Represents a random numerator. Step This represents the step size, which is the maximum distance the artificial fish can move in one iteration; X traget This indicates the selected target direction of the fish's behavior; if it's a clustering: X traget Location of the Partner Center X center If it's a rear-end collision: X traget The location of the best neighbor X best_neighbor If it's foraging: X traget For better random points within the field of view X random , X i ( t )for t The first iteration i The position vector of the artificial fish; The velocity update formula for the velocity vector guidance mechanism of the particle swarm optimization algorithm is:

[0029] in, ω For inertial weights, c 1. c 2 is the learning factor. r 1. r 2 is a random number; P best For an individual, the best historical position; G best The globally optimal position; V i ( t )for t The first iteration i The velocity vector of the artificial fish; Candidate displacement X AFSA The expression for the new position, obtained by weighted fusion with the velocity vector, is:

[0030] Where α is the fusion factor, which is set to 0.6 in this embodiment.

[0031] S5: Calculate the fitness of candidate resource allocation schemes, and update the individual position, the individual's historical best position, and the global best position using a greedy strategy. If the maximum number of iterations has not been reached, return to step S3; continue until the maximum number of iterations is reached, and output the optimal resource allocation scheme corresponding to the global best position; the method for updating the individual position, the individual's historical best solution, and the global best solution using a greedy strategy is as follows: Calculate the fitness of the artificial fish in its new location. F ( X i ( t +1), if the artificial fish's fitness in the new location F ( X i ( t +1)) is less than the fitness F( of the artificial fish's position in the previous iteration. X i ( t If the artificial fish's position is not updated, then update the artificial fish's position to the new position; otherwise, do not update the artificial fish's position. If the artificial fish's fitness level at the new position is... F ( X i ( t +1)) is less than the fitness of the individual's historical best position. F ( P best If the fitness of the artificial fish's new position is positive, then update the individual's historical best position; otherwise, do not update the individual's historical best position.F ( X i ( t +1)) is less than the fitness of the global optimum. F ( G best If the global optimal position is updated, then the global optimal position is updated; otherwise, the global optimal position is not updated.

[0032] This solution also provides an intelligent security edge computing resource allocation system based on an intelligent security edge computing resource allocation method. The system includes: an intelligent security task perception module, used to monitor task requests in the intelligent security monitoring network in real time, collect the size and latency sensitivity attributes of the data to be processed, and simultaneously obtain the computing resource occupancy rate and hardware type status of each edge computing node; and an adaptive parameter tuning module, used to dynamically adjust the field of view and movement step size of the artificial fish according to the current search progress and a preset nonlinear decay strategy during algorithm iteration, in order to balance the system's global search and local development capabilities. The swarm intelligence fusion computing module is used to run the artificial fish swarm algorithm that incorporates a particle swarm velocity vector guidance mechanism. When the artificial fish perform foraging, swarming, and tail-chasing behaviors, the particle velocity vector is introduced to calculate the optimal resource allocation scheme for the current security task. The heterogeneous resource scheduling and execution module is used to receive the optimal resource allocation scheme and accurately distribute the massive concurrent security monitoring tasks to the corresponding ARM or FPGA edge nodes for processing.

[0033] This scheme improves convergence speed and accuracy: by introducing the velocity vector guidance mechanism of the particle swarm optimization algorithm, the search direction is guided by global optimal information, significantly accelerating the convergence speed of the algorithm, while avoiding getting trapped in local optima and improving the optimization accuracy of the resource allocation scheme; this scheme achieves adaptive global and local balance: by adopting a nonlinear exponential decay strategy to dynamically adjust the field of view and step size, the algorithm has strong global exploration capabilities in the early stage of iteration, and focuses on fine local search in the later stage, effectively balancing the "exploration" and "development" processes; this scheme significantly improves load balancing: by taking the node load variance as one of the optimization objectives, it ensures that the load of each edge node is uniform, avoids some nodes being overloaded or idle, and improves the system resource utilization and overall stability.

[0034] The simulation experiment verification process of this scheme is given below: To verify the effectiveness of this method, a comparative experiment was conducted on a simulation platform.

[0035] Experimental environment settings: Number of edge nodes: 10 (computing power randomly distributed between 1000 MIPS and 3000 MIPS); Number of tasks: 100-500 (task length randomly distributed between 500 MI and 2000 MI); Comparison algorithms: standard Artificial Fish Swarm Algorithm (AFSA), standard Particle Swarm Algorithm (PSO), and the algorithm of this invention.

[0036] Analysis of experimental results: Convergence performance comparison: such as Figure 2 As shown, the standard AFSA algorithm only stabilizes around generation 120, with relatively low convergence accuracy; however, the improved algorithm can converge stably and quickly around generation 60, and the final fitness value is about 18% lower than that of the standard AFSA. This demonstrates that introducing the velocity guidance mechanism of the standard particle swarm optimization algorithm can effectively accelerate the convergence speed of the algorithm.

[0037] Load balancing comparison: such as Figure 3 As shown, when using the standard PSO algorithm to process 500 concurrent tasks, some nodes with better performance will be overloaded (CPU utilization greater than 95%), and the load variance will be high. However, after multi-objective optimization and adaptive step size update rule, a good balance is achieved among the nodes, and the load variance is only 0.12.

[0038] like Figure 4 As shown, the verification environment uses an FPGA to simulate an edge computing scheduling center, communicating with a PC host computer via a UART serial port. The host computer software adopts a real-time dynamic visualization design, divided into left and right channels for comparative monitoring. The left-hand channel displays the scheduling status using the traditional standard PSO algorithm. From Figure 4 As can be seen on the left, in the real-time data stream, the CPU load of some nodes frequently exceeds the 90% overload threshold, and the loads of each node are uneven, indicating that the system is in an unstable state. The right channel shows the scheduling status using the multi-strategy improved algorithm proposed in this scheme. Figure 4 As can be seen on the right, thanks to FPGA hardware acceleration and the adaptive step size mechanism of this algorithm, the load of all edge nodes is always maintained in the healthy range of 60%-70%, and the height of each node is basically the same, achieving an excellent load balancing effect; the bottom status bar shows that the system frame rate and hardware connection status are normal, verifying that the algorithm of this solution has the ability to run in real time and process high-concurrency tasks on low-power embedded hardware.

[0039] In summary, this method can simultaneously meet the requirements of rapid task-to-task response and high resource utilization of intelligent security edge computing systems. It is particularly suitable for edge computing environments with high task concurrency and high latency requirements, such as intelligent security and video surveillance, and can achieve system load balancing while ensuring response speed. With the support of embedded hardware acceleration such as FPGA, it can run in real time on low-power edge devices, meeting the actual needs of intelligent security systems for high reliability and low latency.

[0040] Although the specific embodiments of the invention have been described in detail with reference to the accompanying drawings, this should not be construed as limiting the scope of protection of this patent; various modifications and variations that can be made by a person skilled in the art without inventive effort within the scope described in the claims are still within the scope of protection of this patent.

Claims

1. A method for allocating intelligent security edge computing resources, characterized in that, Includes the following steps: S1: Obtain the task set and edge node set of the intelligent security edge computing system, and construct a multi-objective resource scheduling optimization model with task completion time, total system energy consumption and node load variance as optimization objectives; S2: Initialize the mixed population parameters, including population size, maximum number of iterations, initial values ​​of artificial fish's field of vision and movement step size, initial values ​​of individual historical best position and global best position, randomly generated initial artificial fish position, and initial velocity of particle swarm optimization algorithm; S3: Enter the iteration process and dynamically adjust the artificial fish's field of vision and movement step size using a nonlinear adaptive strategy based on the current iteration number; S4: Perform AFSA behavior selection for each artificial fish, and introduce the velocity vector guidance mechanism of particle swarm algorithm in the position update. Combine the individual historical optimal solution and the global optimal solution to correct the movement direction of the artificial fish and generate candidate resource allocation schemes. S5: Calculate the fitness of candidate resource allocation schemes, update individual positions, individual historical best positions, and global best positions using a greedy strategy. If the maximum number of iterations has not been reached, return to step S3. Continue until the maximum number of iterations is reached, and output the optimal resource allocation scheme corresponding to the global best position.

2. The intelligent security edge computing resource allocation method according to claim 1, characterized in that, The fitness expression for the multi-objective resource scheduling optimization model in step S1 is: in, X Let the vector represent the current resource allocation scheme. ω 1. ω 2. ω 3 is the weighting coefficient, and ω 1+ ω 2+ ω 3 = 1, T makespan , E total , load These represent the total completion time of all tasks, the total energy consumption of the system in processing all tasks, and the mean square deviation of the load on each edge node, respectively.

3. The intelligent security edge computing resource allocation method according to claim 2, characterized in that, In step S3, a nonlinear adaptive strategy dynamically adjusts the field of vision of the artificial fish. Visual ( t ) and moving step size Step ( t The formula for calculating ) is: Among them, the starting value of the visible range Visual start and initial step size Step start The initial value; the minimum visible range. Visual min and minimum step size Step min This is the default value; μ This is the attenuation adjustment coefficient; t This indicates the number of steps in this iteration. G max This indicates the maximum number of iterations.

4. The intelligent security edge computing resource allocation method according to claim 3, characterized in that, In step S4, each artificial fish selects its AFSA behavior as follows: based on the crowding factor, the artificial fish autonomously chooses to perform foraging, grouping, or tail-chasing behaviors, and calculates the candidate displacement based on AFSA. X AFSA The calculation formula is as follows: in, Rand Represents a random numerator. Step This represents the step size, which is the maximum distance the artificial fish can move in one iteration; X traget This indicates the selected target direction of the fish's behavior; if it's a clustering: X traget Location of the Partner Center X center If it's a rear-end collision: X traget The location of the best neighbor X best_neighbor If it's foraging: X traget For better random points within the field of view X random , X i ( t )for t The first iteration i The position vector of the artificial fish.

5. The intelligent security edge computing resource allocation method according to claim 4, characterized in that, The velocity update formula for the velocity vector guidance mechanism of the particle swarm algorithm in step S4 is: in, ω For inertial weights, c 1. c 2 is the learning factor. r 1. r 2 is a random number; P best For an individual, the best historical position; G best The globally optimal position; V i ( t )for t The first iteration i The velocity vector of the artificial fish.

6. The intelligent security edge computing resource allocation method according to claim 5, characterized in that, In step S4, when correcting the artificial fish's movement direction, the expression for generating the new position is: Where α is the fusion factor.

7. The intelligent security edge computing resource allocation method according to claim 6, characterized in that, The method for updating the individual position, the individual's historical best solution, and the global best solution using a greedy strategy in step S5 is as follows: Calculate the fitness of the artificial fish in its new location. F ( X i ( t +1), if the artificial fish's fitness in the new location F ( X i ( t +1)) is less than the fitness F( of the artificial fish's position in the previous iteration. X i ( t If the artificial fish's position is not updated, then update the artificial fish's position to the new position; otherwise, do not update the artificial fish's position. If the artificial fish's fitness level at the new position is... F ( X i ( t +1)) is less than the fitness of the individual's historical best position. F ( P best If the fitness of the artificial fish's new position is positive, then update the individual's historical best position; otherwise, do not update the individual's historical best position. F ( X i ( t +1)) is less than the fitness of the global optimum. F ( G best If the global optimal position is updated, then the global optimal position is updated; otherwise, the global optimal position is not updated.

8. An intelligent security edge computing resource allocation system, based on the intelligent security edge computing resource allocation method according to any one of claims 1-7, characterized in that, The system includes: Task awareness module: Collects security task data and edge node status in real time; Parameter tuning module: Dynamically adjusts algorithm parameters according to the iteration progress; Fusion computing module: Runs a hybrid artificial fish swarm and particle swarm algorithm to generate a mapping scheme for the optimal resource allocation scheme; Scheduling and execution module: Receives the mapping scheme and distributes the tasks to the corresponding edge nodes for execution.

Images