A time-delay control-based internet-of-things agent multi-task scheduling method and system

Abstract

The application provides a time-delay control-based Internet-of-things agent multi-task scheduling method and system, which comprises the following steps: obtaining information of multiple power services arriving at a resource pool of an Internet-of-things agent; inputting the information into a pre-constructed time-delay control-based multi-task deployment and resource allocation optimization problem model to obtain a deployment mode of the power services and a resource allocation mode of the Internet-of-things agent; and scheduling the power services according to the deployment mode of the power services and the resource allocation mode of the Internet-of-things agent; wherein the time-delay control-based multi-task deployment and resource allocation optimization problem model is constructed based on the minimum total time delay of the multiple power services as an objective; by solving the time-delay control-based multi-task deployment and resource allocation optimization problem model, the application completes the cooperative deployment of multiple power tasks and realizes the efficient allocation of resources for time-delay control.

Description

Technical Field

[0001] This invention belongs to the field of power Internet of Things (IoT) technology, specifically relating to an IoT agent multi-task scheduling method and system based on time delay control. Background Technology

[0002] With the continuous advancement of the construction of the power Internet of Things (IoT), the scale of the power system is constantly expanding, the number of smart terminals is growing rapidly, and power services are showing a trend towards diversification and real-time processing. Accompanying the increasingly widespread application of IoT technology, new sensor technology, and artificial intelligence technology, power grid operation is shifting towards machine intelligence, perceptual intelligence, and computational intelligence, resulting in massive amounts of heterogeneous data. Traditional cloud computing architectures cannot efficiently meet all the business needs of the smart grid. Edge computing models add task execution, data caching, and analysis capabilities to network devices, migrating some or all of the computing tasks of the original cloud computing model to network edge devices. This reduces the computing load on cloud computing centers, alleviates network bandwidth pressure, and improves data processing efficiency, becoming a new power business data processing solution.

[0003] However, the following problems still exist in the current power Internet of Things: 1. The wide variety of power terminal equipment, complex and diverse physical interfaces, and large differences in underlying connection protocols lead to problems such as long business application development cycles, difficulty in expansion, and strong dependence on terminal manufacturers. At the same time, it also restricts the interconnection and data sharing between different business systems; 2. All information collected by the business terminals of the power Internet of Things is sent to the back-end business system for processing. This centralized processing mode causes performance degradation of some latency-sensitive applications and also puts huge processing pressure on the backbone communication network and the back-end business system. The latency performance of real-time services cannot be guaranteed.

[0004] Invention CN112988285B, entitled "Task Unloading Method and Apparatus, Electronic Device and Storage Medium," relates to the field of task unloading technology. The task unloading method is applied to an electronic device, which is communicatively connected to a task unloading system. The task unloading system includes a second device and at least one first device. The task unloading method includes: first, acquiring tasks to be processed from at least one first device; second, inputting the tasks to be processed into a preset task unloading model to obtain a task unloading strategy; and then, sending the task unloading strategy to at least one first device, so that at least one first device unloads the target task to the second device based on the task unloading strategy, and the second device performs execution processing on the target task. This method can improve the efficiency of task unloading.

[0005] Invention CN113553165A, entitled "A Game Theory-Based Method for Task Offloading and Resource Scheduling in Mobile Edge Computing," discloses a game theory-based method for task offloading and resource scheduling in mobile edge computing. This method aims to jointly minimize the energy consumption and user latency of mobile edge computing servers. It models the user task offloading and resource scheduling problem as a specific optimization problem, constructs a multi-user task offloading computing system with different task offloading priorities, and establishes a migration model by offloading multiple user tasks to an edge base station. In this specific optimization problem, it uses game theory to solve for transmission rate and cost coefficients as constraints, and designs a task offloading method with the ultimate goal of minimizing server energy consumption. This method effectively balances the interests of users and the system, providing a guarantee for implementing task offloading in mobile edge computing systems.

[0006] Invention CN113590232A, entitled "A Task Offloading Method for Relay Edge Networks Based on Digital Twins," discloses a task offloading method for relay edge networks based on digital twins. The method includes: building a task offloading strategy model for the relay edge network; updating the states of corresponding parts in the digital twin environment; inputting the parameters of the digital twin into a simulated task offloading system for iterative training to obtain the optimal task offloading strategy model; transmitting the optimal task offloading strategy model to a simulated manual control interface for backup; transmitting the current digital twin parameter training model and the optimal task offloading strategy model to the digital twin environment cache, and then forwarding them to each relay node by the real-world edge server, which in turn forwards them to the user terminals communicating with it; the user terminals and relay nodes perform corresponding task offloading according to the optimal task offloading strategy model. This invention can reduce the trial-and-error costs in the deployment of real-world 5G edge computing technology and improve deployment efficiency.

[0007] Technical Solution 1 first acquires at least one task to be processed from a first device; second, it inputs the task to be processed into a preset task unloading model to obtain a task unloading strategy; then, it sends the task unloading strategy to at least one first device, so that at least one first device unloads the target task to a second device based on the task unloading strategy, and the second device performs the target task processing. Although this solution improves task unloading efficiency, it only considers task unloading in a single task collaboration scenario.

[0008] Technical Solution 2: This method aims to jointly minimize the energy consumption and user latency of mobile edge computing servers. It models the user task unloading and resource scheduling problem as a specific optimization problem. However, this solution does not consider the data routing optimization problem between multiple tasks.

[0009] Technical solution 3 transmits the current digital twin parameter training model and the optimal task offloading strategy model to the digital twin environment cache, and then forwards them to each relay node by the real-world edge server. The relay nodes then forward them to the user terminals they communicate with. The user terminals and relay nodes offload tasks according to the optimal task offloading strategy model. Similar to solution 1, this solution only considers tasks in a single task collaboration scenario and lacks universality for specific scenarios. Summary of the Invention

[0010] To overcome the shortcomings of the prior art, this invention proposes a multi-task scheduling method for IoT agents based on time delay control, comprising:

[0011] Obtain information about the resource pool where multiple power services arrive at the IoT agent;

[0012] The information is input into a pre-built optimization model for multi-task deployment and resource allocation with time delay control to solve the deployment method of power business and the resource allocation method of IoT agent.

[0013] The power service is scheduled according to the deployment method of the power service and the resource allocation method of the IoT agent;

[0014] The optimization model for multi-task deployment and resource allocation of the delay control is constructed with the goal of minimizing the total delay of multiple power services.

[0015] Preferably, the construction of the optimization problem model for multi-task deployment and resource allocation in latency control includes:

[0016] The objective function is constructed with the goal of minimizing the total latency of multiple power services deployed in the resource pools of various IoT agents and the resource allocation method.

[0017] The constraints are constructed based on the resource constraints of IoT agents;

[0018] Based on the objective function and constraints, construct an optimization problem model for multi-task deployment and resource allocation with time delay control;

[0019] The resources include computing resources and bandwidth resources; the information includes at least one or more of the following: the sequence of containers in the resource pool required for power service requests, the priority of power services, the parameters of the Poisson process corresponding to the time interval distribution of power service data packets leaving the resource pool, the processing speed of power services per unit of computing resources in the resource pool, and the transmission speed of power services per unit of bandwidth resources in the resource pool.

[0020] Preferably, the objective function is calculated as follows:

[0021]

[0022] In the formula, F represents the total latency of multiple power services deployed and allocated in the resource pools of various IoT agents, and K represents the total latency. S ω represents the number of power services in the same resource pool during the same time period. s D represents the priority of the electricity service s. s Let S represent the total delay of the power service s, and let S represent the set of all power services.

[0023] Total delay D of power service s s The calculation formula is as follows:

[0024]

[0025] In the formula, |N s | represents the number of containers in the resource pool that process the power service s, D s,z This represents the average latency of power service s at resource pool z. This indicates the link latency of the power service s after it leaves the resource pool z;

[0026] The average latency D of the power service s at resource pool z s,z The calculation formula is as follows:

[0027]

[0028] In the formula, c s,1 c represents the processing speed for the power service s after resource allocation. s,2 P represents the transmission speed for the power service s after resource allocation. s,z λ represents the non-empty probability of power service s in the processing queue of resource pool z. s The parameters represent the Poisson process parameters corresponding to the time interval distribution of data packets leaving the resource pool for power service s;

[0029] The link delay of the power service s after leaving the resource pool z The calculation formula is as follows:

[0030]

[0031] In the formula, n z This indicates the size of the data to be transmitted at resource pool z.

[0032] Preferably, the formula for calculating the resource constraint is as follows:

[0033]

[0034]

[0035]

[0036] In the formula, K B κ represents the total number of power services deployed on the resource pool. z,i ν represents the proportion of computing resources allocated from resource pool z to power service i. z,i σ represents the proportion of bandwidth resources allocated from resource pool z to power service i, and σ represents the maximum amount of resources that the power service can obtain from the resource pool. This represents the set of resources available to the power business in the resource pool. This represents the processing speed of the power service when the containers in resource pool z use all computing resources. This represents the transmission speed of power services when containers in resource pool z use all bandwidth resources.

[0037] Preferably, the step of inputting the information into a pre-constructed optimization model for multi-task deployment and resource allocation with time delay control to obtain the deployment method of power services and the resource allocation method of IoT agents includes:

[0038] The information is input into a pre-built optimization problem model for multi-task deployment and resource allocation with time delay control. An improved ant colony algorithm based on dynamic pheromone evaporation coefficient is used to solve the optimization problem model for multi-task deployment and resource allocation with time delay control, so as to obtain the deployment mode of each container of multiple power tasks in the resource pool of IoT agent.

[0039] Based on the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, the near-end strategy optimization algorithm is used to solve the optimization problem model of the multi-task deployment and resource allocation of the latency control, so as to obtain the resource allocation method of the IoT agent's resource pool to each container.

[0040] Preferably, the improved ant colony algorithm based on the dynamic pheromone evaporation coefficient is used to solve the optimization problem model of multi-task deployment and resource allocation for latency control, resulting in the deployment method of multiple power tasks in each container of the IoT agent's resource pool, including:

[0041] The computing resources in the resource pool of the IoT agent are used to initialize the pheromone of the ant colony algorithm;

[0042] The disappearance of pheromone trajectories in each resource pool is calculated based on the dynamic pheromone evaporation coefficient.

[0043] The probability of selecting each resource pool for task deployment is calculated based on the disappearance of the pheromone trajectory, and the next resource pool for deploying the power task is selected based on the probability.

[0044] Based on the disappearance of the pheromone trajectory and the selection of the next power task, the resource pool is iteratively calculated using ant colonies to obtain the most frequent path planning in the ant colony path planning, which serves as the deployment method for multiple power tasks in each container of the IoT agent's resource pool.

[0045] Preferably, the formula for calculating the pheromone in the initial ant colony algorithm is as follows:

[0046]

[0047] In the formula, τ 0,j The initial pheromone corresponding to resource pool j is represented by r'. m,j This represents the available memory in resource pool j, where the subscript m represents memory, and r... m,j ψ represents the total memory of resource pool j. m r' represents the total memory of all resource pools. p,j This indicates the available CPUs in resource pool j, where the subscript p represents the CPU, and r p,j ψ represents the total CPU in resource pool j. p This represents the total CPU across all resource pools.

[0048] Preferably, the formula for calculating the disappearance of the pheromone trajectory is as follows:

[0049]

[0050] In the formula, τ(t,j) represents the disappearance of the pheromone trajectory corresponding to resource pool j at time t, τ 0,j Let represent the initial pheromone corresponding to resource pool j, Δ represent the change in pheromone, P(k) represent the resource pool that can be selected in the kth iteration, and ρ represent the dynamic pheromone evaporation coefficient.

[0051] The formula for calculating the dynamic pheromone volatile coefficient ρ is as follows:

[0052]

[0053] Based on the same inventive concept, this application also provides an IoT agent multi-task scheduling system based on time delay control, including: a data acquisition module, a solution module and a scheduling module;

[0054] The data acquisition module is used to obtain information about the resource pool of multiple power services arriving at the IoT agent;

[0055] The solution module is used to input the information into a pre-built optimization problem model for multi-task deployment and resource allocation of time delay control, and solve it to obtain the deployment method of power business and the resource allocation method of IoT agent.

[0056] The scheduling module is used to schedule power services according to the deployment method of power services and the resource allocation method of IoT agents;

[0057] The optimization model for multi-task deployment and resource allocation of the delay control is constructed with the goal of minimizing the total delay of multiple power services.

[0058] Preferably, the solution module is further configured to construct an objective function with the goal of minimizing the total latency of multiple power services deployed in the resource pools of each IoT agent and the resource allocation method.

[0059] The constraints are constructed based on the resource constraints of IoT agents;

[0060] Based on the objective function and constraints, construct an optimization problem model for multi-task deployment and resource allocation with time delay control;

[0061] The resources include computing resources and bandwidth resources; the information includes at least one or more of the following: the sequence of containers in the resource pool required for power service requests, the priority of power services, the parameters of the Poisson process corresponding to the time interval distribution of power service data packets leaving the resource pool, the processing speed of power services per unit of computing resources in the resource pool, and the transmission speed of power services per unit of bandwidth resources in the resource pool.

[0062] Preferably, the objective function constructed by the solution module is calculated as follows:

[0063]

[0064] In the formula, F represents the total latency of multiple power services deployed and allocated in the resource pools of various IoT agents, and K represents the total latency. S ω represents the number of power services in the same resource pool during the same time period. s D represents the priority of the electricity service s. s Let S represent the total delay of the power service s, and let S represent the set of all power services.

[0065] Total delay D of power service s s The calculation formula is as follows:

[0066]

[0067] In the formula, |N s | represents the number of containers in the resource pool that process the power service s, D s,z This represents the average latency of power service s at resource pool z. This indicates the link latency of the power service s after it leaves the resource pool z;

[0068] The average latency D of the power service s at resource pool z s,z The calculation formula is as follows:

[0069]

[0070] In the formula, c s,1 c represents the processing speed for the power service s after resource allocation. s,2 P represents the transmission speed for the power service s after resource allocation. s,z λ represents the non-empty probability of power service s in the processing queue of resource pool z. s The parameters represent the Poisson process parameters corresponding to the time interval distribution of data packets leaving the resource pool for power service s;

[0071] The link delay of the power service s after leaving the resource pool z The calculation formula is as follows:

[0072]

[0073] In the formula, n z This indicates the size of the data to be transmitted at resource pool z.

[0074] Preferably, the calculation formula for the resource constraints constructed by the solution module is as follows:

[0075]

[0076]

[0077]

[0078] In the formula, K B κ represents the total number of power services deployed on the resource pool. z,i ν represents the proportion of computing resources allocated from resource pool z to power service i. z,i σ represents the proportion of bandwidth resources allocated from resource pool z to power service i, and σ represents the maximum amount of resources that the power service can obtain from the resource pool. This represents the set of resources available to the power business in the resource pool. This represents the processing speed of the power service when the containers in resource pool z use all computing resources. This represents the transmission speed of power services when containers in resource pool z use all bandwidth resources.

[0079] Preferably, the solution module is specifically used for:

[0080] The information is input into a pre-built optimization problem model for multi-task deployment and resource allocation with time delay control. An improved ant colony algorithm based on dynamic pheromone evaporation coefficient is used to solve the optimization problem model for multi-task deployment and resource allocation with time delay control, so as to obtain the deployment mode of each container of multiple power tasks in the resource pool of IoT agent.

[0081] Based on the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, the near-end strategy optimization algorithm is used to solve the optimization problem model of the multi-task deployment and resource allocation of the latency control, so as to obtain the resource allocation method of the IoT agent's resource pool to each container.

[0082] Preferably, the solution module uses an improved ant colony algorithm based on the dynamic pheromone evaporation coefficient to solve the optimization problem model of multi-task deployment and resource allocation for the delay control, obtaining the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, including:

[0083] The computing resources in the resource pool of the IoT agent are used to initialize the pheromone of the ant colony algorithm;

[0084] The disappearance of pheromone trajectories in each resource pool is calculated based on the dynamic pheromone evaporation coefficient.

[0085] The probability of selecting each resource pool for task deployment is calculated based on the disappearance of the pheromone trajectory, and the next resource pool for deploying the power task is selected based on the probability.

[0086] Based on the disappearance of the pheromone trajectory and the selection of the next power task, the resource pool is iteratively calculated using ant colonies to obtain the most frequent path planning in the ant colony path planning, which serves as the deployment method for multiple power tasks in each container of the IoT agent's resource pool.

[0087] Preferably, the formula for calculating the pheromone in the ant colony algorithm initialized by the solution module is as follows:

[0088]

[0089] In the formula, τ 0,j The initial pheromone corresponding to resource pool j is represented by r'. m,j This represents the available memory in resource pool j, where the subscript m represents memory, and r... m,j ψ represents the total memory of resource pool j. m r' represents the total memory of all resource pools. p,j This indicates the available CPUs in resource pool j, where the subscript p represents the CPU, and r p,j ψ represents the total CPU in resource pool j. p This represents the total CPU across all resource pools.

[0090] Preferably, the calculation formula for the disappearance of the pheromone trajectory by the solution module is as follows:

[0091]

[0092] In the formula, τ(t,j) represents the disappearance of the pheromone trajectory corresponding to resource pool j at time t, τ 0,j Let represent the initial pheromone corresponding to resource pool j, Δ represent the change in pheromone, P(k) represent the resource pool that can be selected in the kth iteration, and ρ represent the dynamic pheromone evaporation coefficient.

[0093] The formula for calculating the dynamic pheromone volatile coefficient ρ is as follows:

[0094]

[0095] The present invention also provides a computer device, comprising: one or more processors;

[0096] Memory, used to store one or more programs;

[0097] When the one or more programs are executed by the one or more processors, the IoT agent multi-task scheduling method based on latency control as described above is implemented.

[0098] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the IoT agent multi-task scheduling method based on latency control as described above. Compared with the closest prior art, this invention has the following beneficial effects:

[0099] This invention provides a method and system for multi-task scheduling of IoT agents based on latency control, comprising: acquiring information on the resource pool of multiple power services arriving at the IoT agent; inputting the information into a pre-constructed optimization problem model for multi-task deployment and resource allocation with latency control for solution, thereby obtaining the deployment method of the power services and the resource allocation method of the IoT agent; scheduling the power services according to the deployment method of the power services and the resource allocation method of the IoT agent; wherein, the optimization problem model for multi-task deployment and resource allocation with latency control is constructed with the objective of minimizing the total latency of multiple power services; by solving the optimization problem model for multi-task deployment and resource allocation with latency control, this invention completes the collaborative deployment of multiple power tasks and achieves efficient allocation of resources oriented towards latency control. Attached Figure Description

[0100] Figure 1 A schematic diagram of a multi-task scheduling method for IoT agents based on time delay control provided by the present invention;

[0101] Figure 2A schematic diagram illustrating the design flow of the multi-task deployment and resource allocation algorithm provided by this invention;

[0102] Figure 3 A schematic diagram of the structure of an IoT agent multi-task scheduling system based on time delay control is provided for this invention;

[0103] Figure 4 This is a schematic diagram illustrating the structure of an example of a power IoT agent scheduling system provided by the present invention. Detailed Implementation

[0104] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0105] Example 1:

[0106] A flowchart illustrating a multi-task scheduling method for IoT agents based on delay control provided by this invention is shown below. Figure 1 As shown, it includes:

[0107] Step 1: Obtain information on the resource pool of multiple power services arriving at the IoT agent;

[0108] Step 2: Input the information into the pre-built optimization problem model of multi-task deployment and resource allocation for delay control, and solve it to obtain the deployment method of power business and the resource allocation method of IoT agent;

[0109] Step 3: Schedule the power service according to the deployment method of the power service and the resource allocation method of the IoT agent;

[0110] Among them, the optimization problem model for multi-task deployment and resource allocation of delay control is constructed with the goal of minimizing the total delay of multiple power services.

[0111] This invention leverages the high abstraction of lightweight containers to achieve high reusability of hardware and software. IoT agents, as limited resources, can play multiple roles in a time-sharing manner, performing different functions such as communication and computing. A virtual switch software used for communication or a virtual network function can be placed as a subtask in a remote repository. Each edge IoT agent can download and run the subtask to support edge power services. To facilitate the description of the container's resource allocation process, this invention first constructs an edge node resource allocation model and, based on this model, establishes an optimization problem model for multi-task deployment and resource allocation with latency control. For this model, this invention proposes an MTDRA (Multi-task Deployment and Resource Allocation) algorithm, the design process of which is as follows: Figure 2Specifically, this includes: constructing an optimization problem model for multi-task deployment and resource allocation with latency control; decomposing the solution of the optimization problem model into two sequentially solved sub-problems; designing an improved ant colony algorithm based on the dynamic pheromone evaporation coefficient to solve the multi-task deployment results; and designing a resource allocation algorithm based on PPO (Proximal Policy Optimization) according to the task deployment results to allocate the computing and communication resources (i.e., bandwidth resources) of the IoT agent.

[0112] First, the model used in this invention will be described.

[0113] I-Power Edge Network Model.

[0114] This invention divides edge IoT agent resources into several public resource pools. This invention uses z to represent a resource pool, and K... Z This represents the number of resource pools. Since resource pools are connected via different data links and have different uplink and downlink bandwidths, they are represented using C. z B z These represent the computing and bandwidth resources possessed by resource pool z, respectively. In actual production scenarios, the combination of service requests will change at different times. When a power service (or service) request s arrives at the dispatch system, K is used... S This represents the number of business requests from the same resource pool within the same time period.

[0115] This invention assumes that requests arrive at the resource pool randomly at different time intervals. This invention uses Ω. s The information represents the arrival of service s in the resource pool, where Ω s =(N s ,λ s C s,1 C s,2 ,ω s In this stage, the present invention considers the arrival process of the service request stream data packet as having a parameter λ. s The Poisson process, where N s Represents the sequence of containers required for a business request, |N s | represents the number of containers, ω s C represents the priority of the electricity service s. s,1 C represents the processing speed of a unit of computing resource in the resource pool for a service data packet s. s,2 This represents the transmission speed of s data per unit bandwidth resource in the resource pool, where

[0116] This invention uses The processing speed obtained by a container of business s in resource pool z using all resources. This represents the data packet transmission speed obtained by s using the bandwidth resources in resource pool z. To find an approximate optimal solution for resource pool allocation, a discrete resource allocation strategy is used to allocate the resource pool and its bandwidth resources. The allocation strategy π(s,a)={κ z ,ν z},in:

[0117]

[0118]

[0119]

[0120] In the above formula, K S K represents the number of business requests from the same resource pool within the same time period. B κ represents the total number of power services deployed on the resource pool. z ,ν z These represent the resource allocation of resource pool z to the business request chain, κ and κ respectively. z,i ν represents the proportion of computing resources allocated from resource pool z to power service i. z,i This represents the proportion of bandwidth resources allocated from resource pool z to power service i. It is a finite discrete set representing the set of resources available to the power business in the resource pool. Specifically, the resource allocation is divided into 10 discrete resource blocks. This significantly reduces the size of the action space when defining resource allocation actions, lowering the difficulty of algorithm convergence. σ represents the maximum amount of resources that service request s can obtain from the resource pool. This is done to prevent a single request from consuming too many resources and affecting the operation of other services. The processing and transmission rates of the current service request can be obtained based on the resource allocation strategy.

[0121]

[0122]

[0123] In the above formula, This indicates the processing speed of the containers in resource pool z for service s after resource allocation. This indicates the transmission speed of containers in resource pool z for service s after resource allocation.

[0124] Within the same business request, the data packet processing speed and transmission speed within the same resource pool should meet c as much as possible. s,1 =c s,2Only in this way can data packets be transmitted through the data link without queuing time after being processed by the resource pool. Resource allocation must consider both maximizing and balancing resource allocation to avoid unnecessary waste. Therefore, the following conditions must be met when allocating resources in the resource pool.

[0125] II-Latency Model: A model for optimizing multi-task deployment and resource allocation with latency control.

[0126] To accurately describe end-to-end latency in multi-task collaborative scenarios, this invention first establishes a latency model based on a serial queuing model, which improves the accuracy of network latency assessment.

[0127] This invention will discuss the process of service request data packets passing through container 1 and subsequent containers separately. This invention first analyzes the departure process of the first container, focusing on the time interval between the departure of two adjacent data packets within the same service. When the arrival rate of service request stream data packets is much lower than the resource pool processing rate, the time interval between data packets leaving the resource pool conforms to the parameter λ. s In the Poisson process, when the arrival rate equals the resource pool processing rate, the time interval between packets leaving the resource pool is approximately equal to a parameter c. s,1 The determination process. This invention obtains a complete concatenated queuing model for container packet processing and packet transmission by analyzing the processing latency, transmission latency, and queuing backlog latency of the service request flow in container 1, which may be caused by differences in arrival rate and processing rate. The average packet latency of the service request flow S in the first container is as follows:

[0128]

[0129] In the formula, c s,1 c represents the processing speed for the power service s after resource allocation. s,2 P represents the transmission speed for the power service s after resource allocation. s,1 This represents the probability that the power service s is not empty in the processing queue of container 1.

[0130] Based on the above analysis, this invention further analyzes the link transmission latency after data leaves container 1. If container 1 and container 2 are in the same resource pool, then the data link transmission latency does not need to be considered. If container 2 is in another resource pool, then the transmission latency is mainly determined by the number of switches and links traversed by the link. Let K be defined. n The number of switches traversed is K, and the number of links is also K. n Then the link transmission delay is:

[0131]

[0132] In the formula, n1 represents the size of the data to be transmitted at container 1.

[0133] Next, this invention studies the average total latency of data packets in container 2 after passing through container 1. Since the arrival of data packets in container 2 is affected by the processing rate and transmission rate of container 1, it is difficult to analyze the time interval Z between consecutive arrivals of service flow data packets in container 2. i To eliminate the impact of the processing and transmission rate of each data packet at container 1 on its arrival at container 2, this invention approximates the arrival process of a data stream S of a certain service at container 2 as having an arrival rate of λ. s The Poisson process is used, and the same M / D / 1 queuing model is established. Based on the obtained packet arrival process, the sum of the average delays of packets at container 2 is as follows:

[0134]

[0135] In the formula, c s,1 c represents the processing speed for the power service s after resource allocation. s,2 This represents the transmission speed for power service s after resource allocation, when c s,1 <c s,2 At that time, due to the slower processing rate of data packets in container 2, a queuing backlog occurs at container 2, causing a delay. s,1 ≥c s,2 At that time, the data packet uses c s,1 The processing rate is achieved through container 2, and in c s,2 The transmission rate leaves the edge computing node. Therefore, the total latency can be calculated from the resource pool and data link through which the service request data stream passes:

[0136]

[0137] In the above formula, D s,z This represents the processing rate and transmission rate of resource pool z (i.e., the average latency of power service s at resource pool z). D represents the data link transmission delay after resource pool z. s,z and The calculation formula is as follows:

[0138]

[0139]

[0140] In the formula, n z This indicates the size of the data to be transmitted at resource pool z.

[0141] In summary, the purpose of this invention is to optimize the end-to-end latency of data packets in the service request stream of a container cluster. The optimization problem model established by this invention is shown in the following equation:

[0142]

[0143]

[0144]

[0145]

[0146] III-Latency-controlled multi-task deployment and dynamic allocation algorithm for CPU computing resources.

[0147] In response to the complex environment and diverse services of the power Internet of Things, this invention first proposes a multi-task deployment algorithm for delay control based on an improved ant colony algorithm.

[0148] A service consists of one or more tasks, which users initiate by submitting tasks to the scheduling module. This invention designs an improved ant colony optimization (IACO) algorithm and uses the scheduling module to select a set of nodes that meet specified constraints and deploy tasks to these nodes (i.e., containers in a resource pool).

[0149] The goal of the scheduling module is to allocate tasks to available resources. Available resources are used during each scheduling process. An artificial ant randomly selects resources by observing the pheromone trails of each resource. The formula for calculating each computational resource of a given node is:

[0150]

[0151] In the above formula: R(j) represents the resources of node j, and r' m,j r is the available memory for node j. m,j r is the total memory of the node p,j ' is the available CPU of the node, r p,j ψ represents the total CPU of the node. m ψ is the memory size of all nodes. p The CPU size for all nodes. Where n is the number of tasks to be allocated memory, ψ i This indicates the ratio of memory allocated.

[0152] To initialize the pheromone trajectory for each node, the loop-greedy algorithm uses R(j), which simply places each task on a node in the loop pattern, τ 0,j=R(j) represents the initial pheromone level for each node. The formula for calculating the probability of selecting the current node for task deployment is:

[0153]

[0154] In the above formula, η e Let α be the heuristic value of node e, β be the information heuristic factor, β be the expected heuristic factor, and m be the total number of nodes.

[0155] Calculate the deployment probability p(t,j) for each node, and select the next node j.

[0156]

[0157] Where j∈P(k), q0 is the preset exploration rate threshold, and q is the exploration rate of the ant colony algorithm. P(k) represents the resource nodes that can be selected to deploy tasks in the k-th iteration. N is the set of all nodes.

[0158] The next step is to calculate pheromone disappearance. The formula for calculating the disappearance of the pheromone trajectory for each node is:

[0159]

[0160] In the formula, τ(t,j) represents the disappearance of the pheromone trajectory corresponding to resource pool j at time t, τ 0,j Let ρ represent the initial pheromone level corresponding to resource pool j, Δ represent the pheromone variation, and ρ represent the dynamic pheromone evaporation coefficient. When a task is placed into a specific resource by the scheduler, the value of Δ is always less than 0. Pheromones are calculated using the above formula. To significantly reduce the pheromone level of the selected node, making it less important for the next task, the remaining tasks are allocated to the entire resource.

[0161] In ant colony optimization (ACO) algorithms, the pheromone evaporation coefficient plays a crucial role in the algorithm's convergence and search performance. In traditional ACO algorithms, ρ is typically a fixed value between 0 and 1. Because the pheromone concentration on the path is relatively low at the beginning of the iterations, even though pheromones evaporate with each iteration, the pheromone concentration still increases. Therefore, using a fixed pheromone evaporation coefficient leads to excessive pheromone evaporation in the early stages and less evaporation in the later stages, severely limiting the performance of the ACO algorithm. To address this issue, this invention introduces the Sigmoid function, commonly used as an activation function in neural networks. This function is monotonically increasing and maps the value of ρ to the range (0,1). This reduces pheromone evaporation in the early stages and increases it in the later stages. The formula for ρ becomes:

[0162]

[0163] In the above formula, k represents the number of iterations of the algorithm.

[0164] The best planning method is to select the most frequent plan using the following formula.

[0165]

[0166] In the formula, x is the value of p(k) when it reaches its maximum, and this value is assigned to p. w .

[0167] By improving the traditional ant colony algorithm, this invention solves the problem of poor performance caused by the use of a fixed information evaporation coefficient in the traditional ant colony algorithm, and completes the collaborative deployment of multiple tasks.

[0168] Secondly, after completing the deployment of multi-task services on containers, this invention designs a resource allocation algorithm oriented towards latency control to guide the dynamic allocation of CPU computing resources and bandwidth resources in IoT agents, thereby ensuring the latency requirements of services.

[0169] Because the environment of edge computing networks is complex and dynamic, it is necessary to use reliable and scalable learning algorithms to learn in this challenging environment. Since the PPO algorithm ensures stability by binding the range of parameter updates to a trust region, this invention considers using this algorithm for resource allocation. To solve the problem obtained in the previous section using deep reinforcement learning algorithms, this invention first transforms the problem into an MDP (Markov decision process). This process contains four elements: agent, input, action, and reward.

[0170] Intelligent agent: In this invention, the intelligent agent is an edge IoT agent.

[0171] Input: The system state is represented as x after environmental information changes. t =[N t,z A t The current system state is input into the model, where N... z,t The container flow group representing resource pool z In the formula, with For example, This represents the resource allocation of resource pool z for task 1 at time t. Resource allocation strategy for all resource pools.

[0172] Action: At time t, resource pool z can only use one resource allocation strategy a. t,z =[κ t,z ,ν t,z ],in

[0173] Reward Function: When an agent selects a certain action, the corresponding system state will change and a reward value will be obtained. In this invention, the end-to-end latency difference of the data packets in the container service request stream between the current state and the next state is used as the reward. As shown in the following formula:

[0174]

[0175] In the above formula, S and S′ represent the power services processed by the container group in the resource pool services of the current state and the next state, respectively. A positive reward value is obtained when the end-to-end latency obtained in the next state is lower, and a negative reward value is obtained otherwise.

[0176] The PPO algorithm is a deep reinforcement learning algorithm based on the actor-critic algorithm framework. The PPO architecture designed in this invention includes two Actor networks, Actor1 and Actor2. Actor1 represents the current state policy π and interacts with the edge network environment, selecting task deployment actions based on the current environment state. The critic evaluates the current policy based on the reward obtained after executing the deployment action and updates the parameters in the critic network through backpropagation of the loss function. Actor2 represents the old policy π. old Every time the agent is trained, it updates Actor2 using the parameters in Actor1. This process is repeated until the PPO algorithm converges, at which point a well-trained edge multi-task deployment model based on the AC framework is obtained.

[0177] The PPO algorithm requires the similarity value r between the old and new strategies. t (θ) is used to impose restrictions.

[0178] L CLIP (θ)=E t [min(r t (θ)B t ,clip(r t (θ)),1-ε,1+ε)B t ]

[0179] In the above formula, ε∈[0,1] is a hyperparameter, and clip() will change r t The value of (θ) is constrained within the interval [1-ε, 1+ε].

[0180] The above formula represents the similarity between the old and new strategies, where π represents the similarity between the old and new strategies. θ(a|s) To calculate the probability of taking action a under the corresponding state of power business s using strategy π, r is substituted into the calculation. t Substitute (θ) into the above equation. π θ(a|s)There is no specific calculation formula; the specific result is output by the neural network; B t Let E be the dominant function at time t. t Let be the expected value at time t.

[0181] The schedulable resources in the edge network include the available computing resources and available bandwidth resources of the edge IoT agent. The algorithm generates an immediate reward for the agent that takes a certain action through a reward function. The algorithm learns continuously based on the reward and thus obtains the optimal resource allocation strategy.

[0182] The PPO algorithm's actor network consists of two neural networks, Actor1 and Actor2. Actor1 guides the agent to interact with the environment, obtains transition samples, and caches them. The policy parameters in Actor2 represent the old policy; after each iteration, the parameters in Actor2 are updated using the parameters from Actor1. The critic network consists of a single neural network. The specific training steps for the task deployment model are as follows:

[0183] a) Input the current state into the Actor1 network, and the agent bases its actions on policy π. old Choose one action, i.e., a l =π(s) l Repeat the above process, with the agent continuously interacting with the edge network for T time steps, collecting and caching historical interaction information.

[0184] b) Calculate the advantage function for each time step using the following formula, where γ is the discount factor, V is the state value function, φ is the commentator network parameter, and l is the time step.

[0185] B t =∑ l＞t γ l-t rw l -V φ (s t )

[0186] c) Calculate the loss function of the critic network using the following formula, and update the critic network parameters φ based on backpropagation using this function.

[0187]

[0188] d) Using L CLIP (θ) and the advantage function update the parameters of the actor network.

[0189] e) Repeat step d, and after a certain number of steps, update the parameters of Actor2 using the network parameters in Actor1.

[0190] f) Repeat step af.

[0191] The PPO algorithm based on the Actor-Critic framework is obtained. Then, the agent outputs the next action according to the given task queue based on the actor network, and the critic network gives the corresponding evaluation. The process is iterated until the allocation of resources for time-latency sensitive tasks is completed.

[0192] This invention employs a PPO-based resource allocation algorithm to achieve efficient allocation of computing and communication resources oriented towards delay control.

[0193] Example 2:

[0194] Based on the same inventive concept, this invention also provides an IoT agent multi-task scheduling system based on time delay control, the system structure of which is as follows: Figure 3 As shown, it includes: a data acquisition module, a solution module, and a scheduling module;

[0195] The data acquisition module is used to obtain information about the resource pool of multiple power services arriving at the IoT agent.

[0196] The solution module is used to solve a pre-built optimization problem model for multi-task deployment and resource allocation of time delay control, and obtain the deployment method of power business and the resource allocation method of IoT agent.

[0197] The scheduling module is used to schedule power services according to the deployment method of power services and the resource allocation method of IoT agents;

[0198] Among them, the optimization problem model for multi-task deployment and resource allocation of delay control is constructed with the goal of minimizing the total delay of multiple power services.

[0199] The solution module is also used to construct an objective function with the goal of minimizing the total latency of multiple power services deployed in the resource pools of various IoT agents and the resource allocation method.

[0200] The constraints are constructed based on the resource constraints of IoT agents;

[0201] An optimization model for multi-task deployment and resource allocation with time delay control is constructed based on the objective function and constraints.

[0202] The resources include computing resources and bandwidth resources; the information includes at least one or more of the following: the sequence of containers in the resource pool required for the power service request, the priority of the power service, the parameters of the Poisson process corresponding to the time interval distribution of the power service data packets leaving the resource pool, the processing speed of the power service per unit of computing resources in the resource pool, and the transmission speed of the power service per unit of bandwidth resources in the resource pool.

[0203] The objective function constructed by the solution module is calculated as follows:

[0204]

[0205] In the formula, F represents the total latency of multiple power services deployed and allocated in the resource pools of various IoT agents, and K represents the total latency. S ω represents the number of power services in the same resource pool during the same time period. s D represents the priority of the electricity service s. s Let S represent the total delay of the power service s, and let S represent the set of all power services.

[0206] Total delay D of power service s s The calculation formula is as follows:

[0207]

[0208] In the formula, |N s | represents the number of containers in the resource pool that process the power service s, D s,z This represents the average latency of power service s at resource pool z. This indicates the link latency of the power service s after it leaves the resource pool z;

[0209] The average delay D of power service s at resource pool z s,z The calculation formula is as follows:

[0210]

[0211] In the formula, c s,1 c represents the processing speed for the power service s after resource allocation. s,2 P represents the transmission speed for the power service s after resource allocation. s,z λ represents the non-empty probability of power service s in the processing queue of resource pool z. s The parameters represent the Poisson process parameters corresponding to the time interval distribution of data packets leaving the resource pool for power service s;

[0212] Link latency of power service s after leaving resource pool z The calculation formula is as follows:

[0213]

[0214] In the formula, n z This indicates the size of the data to be transmitted at resource pool z.

[0215] The calculation formula for the resource constraints constructed by the solution module is as follows:

[0216]

[0217]

[0218]

[0219] In the formula, K B κ represents the total number of power services deployed on the resource pool. z,i ν represents the proportion of computing resources allocated from resource pool z to power service i. z,i σ represents the proportion of bandwidth resources allocated from resource pool z to power service i, and σ represents the maximum amount of resources that the power service can obtain from the resource pool. This represents the set of resources available to the power business in the resource pool. This represents the processing speed of the power service when the containers in resource pool z use all computing resources. This represents the transmission speed of power services when containers in resource pool z use all bandwidth resources.

[0220] The solution module is specifically used for:

[0221] The information is input into a pre-built optimization problem model for multi-task deployment and resource allocation with time delay control. An improved ant colony algorithm based on dynamic pheromone evaporation coefficient is used to solve the optimization problem model for multi-task deployment and resource allocation with time delay control, and the deployment mode of each container in the resource pool of the IoT agent for multiple power tasks is obtained.

[0222] Based on the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, a near-end strategy optimization algorithm is used to solve the optimization problem model of multi-task deployment and resource allocation with latency control, and the resource allocation method of the IoT agent's resource pool to each container is obtained.

[0223] The solution module employs an improved ant colony algorithm based on the dynamic pheromone evaporation coefficient to solve the optimization problem model of multi-task deployment and resource allocation for latency control. This yields the deployment methods of multiple power tasks within the resource pool of the IoT agent, including:

[0224] The computing resources in the resource pool of the IoT agent are used to initialize the pheromone of the ant colony algorithm;

[0225] The disappearance of pheromone trajectories in each resource pool is calculated based on the dynamic pheromone evaporation coefficient.

[0226] The probability of selecting each resource pool for task deployment is calculated based on the disappearance of pheromone trajectories, and the next resource pool for deploying power tasks is selected based on the probability.

[0227] Based on the disappearance of pheromone trajectories and the selection of the next deployment power task, the resource pool uses ant colony iteration to calculate and obtain the most frequent path planning in the ant colony path planning as the deployment method of multiple power tasks in each container of the IoT agent's resource pool.

[0228] The formula for calculating the pheromone for initializing the ant colony algorithm in the solution module is as follows:

[0229]

[0230] In the formula, τ 0,j The initial pheromone corresponding to resource pool j is represented by r'. m,j This represents the available memory in resource pool j, where the subscript m represents memory, and r... m,j ψ represents the total memory of resource pool j. m r' represents the total memory of all resource pools. p,j This indicates the available CPUs in resource pool j, where the subscript p represents the CPU, and r p,j ψ represents the total CPU in resource pool j. p This represents the total CPU across all resource pools.

[0231] The calculation formula for the disappearance of pheromone trajectories in the solution module is as follows:

[0232]

[0233] In the formula, τ(t,j) represents the disappearance of the pheromone trajectory corresponding to resource pool j at time t, τ 0,j Let represent the initial pheromone corresponding to resource pool j, Δ represent the change in pheromone, P(k) represent the resource pool that can be selected in the kth iteration, and ρ represent the dynamic pheromone evaporation coefficient.

[0234] The formula for calculating the dynamic pheromone volatility coefficient ρ is as follows:

[0235]

[0236] Example 3:

[0237] The power IoT agent scheduling system constructed in this invention is as follows: Figure 4 As shown. Business terminals include various smart meters, photovoltaic panels, and charging piles, among other power service equipment. The power communication network includes industrial Ethernet, EPON (Ethernet Passive Optical Network), 2G / 3G / 4G, wireless LAN, and wireless WAN. The IoT agent layer consists of software and hardware layers, connecting business terminals to the platform layer to achieve various functions such as data acquisition, message adaptation, edge computing, and transmission. Based on lightweight container technology, the IoT agent uses a built-in resource scheduling engine to achieve flexible scheduling of edge-side virtualized computing and network resources, enabling unified access, multi-dimensional perception, unified modeling, and resource orchestration. The IoT management platform manages the IoT agent using RESTful / OpenFlow / SNMP.

[0238] Example 4:

[0239] Based on the same inventive concept, this invention also provides a computer device, which includes a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement corresponding method flows or corresponding functions, thereby realizing the steps of the IoT agent multi-task scheduling method based on latency control in the above embodiments.

[0240] Example 5:

[0241] Based on the same inventive concept, this invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of the latency-controlled IoT agent multi-task scheduling method in the above embodiments.

[0242] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0243] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0244] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0245] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0246] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the scope of protection of the claims pending approval.

Claims

1. A multi-task scheduling method for IoT agents based on time delay control, characterized in that, include: Obtain information about the resource pool where multiple power services arrive at the IoT agent; The information is input into a pre-built optimization model for multi-task deployment and resource allocation with time delay control to solve the problem and obtain the deployment method of power business and the resource allocation method of IoT agent. The power service is scheduled according to the deployment method of the power service and the resource allocation method of the IoT agent; The optimization problem model for multi-task deployment and resource allocation of the delay control is constructed based on the objective of minimizing the total delay of multiple power services. The construction of the optimization problem model for multi-task deployment and resource allocation in latency control includes: The objective function is constructed with the goal of minimizing the total latency of multiple power services deployed in the resource pools of various IoT agents and the resource allocation method. The constraints are constructed based on the resource constraints of IoT agents; Based on the objective function and constraints, construct an optimization problem model for multi-task deployment and resource allocation with time delay control; The resources include computing resources and bandwidth resources; the information includes at least one or more of the following: the container sequence in the resource pool required for power service requests, the priority of power services, the parameters of the Poisson process corresponding to the time interval distribution of power service data packets leaving the resource pool, the processing speed of power services per unit of computing resources in the resource pool, and the transmission speed of power services per unit of bandwidth resources in the resource pool. The objective function is calculated as follows: In the formula, F This represents the total latency of multiple power services deployed in the resource pools of various IoT agents and the resource allocation method. This indicates the number of power transactions within the same resource pool during the same time period. Indicates electricity business s priority, Indicates electricity business s Total latency, S It represents the set of all electricity-related businesses; Power business s Total latency The calculation formula is as follows: In the formula, Indicates the electricity business in the resource pool s The number of containers being processed. Indicates electricity business s The average latency at resource pool z Indicates electricity business s Link latency after leaving resource pool z; The power business s Average latency at resource pool z The calculation formula is as follows: In the formula, This indicates the allocation of resources for electricity services. s Processing speed This indicates the allocation of resources for electricity services. s transmission speed, Indicates electricity business s The probability of a non-empty queue in the processing queue of resource pool z. Indicates electricity business s The parameters of the Poisson process corresponding to the distribution of time intervals at which data packets leave the resource pool; The power business s Link latency after leaving resource pool z The calculation formula is as follows: In the formula, This indicates the size of the data to be transmitted at resource pool z.

2. The method as described in claim 1, characterized in that, The formula for calculating the resource constraint is as follows: In the formula, This represents the total number of power services deployed on the resource pool. This indicates that resource pool z is allocated to the power business. i The proportion of computing resources This indicates that resource pool z is allocated to the power business. i The proportion of bandwidth resources, This indicates the maximum amount of resources that the power business can obtain from the resource pool. This represents the set of resources available to the power business in the resource pool. This represents the processing speed of the power service when the containers in resource pool z use all computing resources. This represents the transmission speed of power services when containers in resource pool z utilize all bandwidth resources. This indicates that resource pool z is allocated to the power business. s The proportion of computing resources This indicates that resource pool z is allocated to the power business. s The proportion of bandwidth resources.

3. The method as described in claim 1, characterized in that, The process involves inputting the information into a pre-built optimization model for multi-task deployment and resource allocation with time-delay control, and solving it to obtain the deployment method for power services and the resource allocation method for IoT agents, including: The information is input into a pre-built optimization problem model for multi-task deployment and resource allocation with time delay control. An improved ant colony algorithm based on dynamic pheromone evaporation coefficient is used to solve the optimization problem model for multi-task deployment and resource allocation with time delay control, so as to obtain the deployment mode of each container of multiple power tasks in the resource pool of IoT agent. Based on the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, the near-end strategy optimization algorithm is used to solve the optimization problem model of the multi-task deployment and resource allocation of the latency control, so as to obtain the resource allocation method of the IoT agent's resource pool to each container.

4. The method as described in claim 3, characterized in that, The improved ant colony algorithm based on the dynamic pheromone evaporation coefficient is used to solve the optimization problem model of multi-task deployment and resource allocation for latency control, resulting in the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, including: The computing resources in the resource pool of the IoT agent are used to initialize the pheromone of the ant colony algorithm; The disappearance of pheromone trajectories in each resource pool is calculated based on the dynamic pheromone evaporation coefficient. The probability of selecting each resource pool for task deployment is calculated based on the disappearance of the pheromone trajectory, and the next resource pool for deploying the power task is selected based on the probability. Based on the disappearance of the pheromone trajectory and the selection of the next power task, the resource pool is iteratively calculated using ant colonies to obtain the most frequent path planning in the ant colony path planning, which serves as the deployment method for multiple power tasks in each container of the IoT agent's resource pool.

5. The method as described in claim 4, characterized in that, The formula for calculating the pheromone in the initialization ant colony algorithm is as follows: In the formula, Indicates resource pool j The corresponding initial pheromone, Indicates resource pool j Available memory, subscript m Represents memory, Indicates resource pool j Total memory, This represents the total memory of all resource pools. Indicates resource pool j The available CPUs, where the subscript p indicates the CPU. Indicates resource pool j Total CPU, This represents the total CPU across all resource pools.

6. The method as described in claim 4, characterized in that, The formula for calculating the disappearance of the pheromone trajectory is as follows: In the formula, express t Time Resource Pool j The corresponding pheromone trajectory disappeared. Indicates resource pool j The corresponding initial pheromone, Indicates changes in pheromones, Indicates the first k Resource pools available for selection in the next iteration Indicates the dynamic pheromone evaporation coefficient; The dynamic pheromone volatility coefficient The calculation formula is as follows: 。 7. A multi-task scheduling system for IoT agents based on time delay control, characterized in that, include: Data acquisition module, solution module, and scheduling module; The data acquisition module is used to obtain information about the resource pool of multiple power services arriving at the IoT agent; The solution module is used to input the information into a pre-built optimization problem model for multi-task deployment and resource allocation of time delay control, and solve it to obtain the deployment method of power business and the resource allocation method of IoT agent. The scheduling module is used to schedule power services according to the deployment method of power services and the resource allocation method of IoT agents; The optimization problem model for multi-task deployment and resource allocation of the delay control is constructed based on the objective of minimizing the total delay of multiple power services. The solution module is also used to construct an objective function with the goal of minimizing the total latency of multiple power services deployed in the resource pools of various IoT agents and the resource allocation method. The constraints are constructed based on the resource constraints of IoT agents; Based on the objective function and constraints, construct an optimization problem model for multi-task deployment and resource allocation with time delay control; The resources include computing resources and bandwidth resources; the information includes at least one or more of the following: the container sequence in the resource pool required for power service requests, the priority of power services, the parameters of the Poisson process corresponding to the time interval distribution of power service data packets leaving the resource pool, the processing speed of power services per unit of computing resources in the resource pool, and the transmission speed of power services per unit of bandwidth resources in the resource pool. The objective function constructed by the solution module is calculated as follows: In the formula, F This represents the total latency of multiple power services deployed in the resource pools of various IoT agents and the resource allocation method. This indicates the number of power transactions within the same resource pool during the same time period. Indicates electricity business s priority, Indicates electricity business s Total latency, S It represents the set of all electricity-related businesses; Power business s Total latency The calculation formula is as follows: In the formula, Indicates the electricity business in the resource pool s The number of containers being processed. Indicates electricity business s The average latency at resource pool z Indicates electricity business s Link latency after leaving resource pool z; The power business s Average latency at resource pool z The calculation formula is as follows: In the formula, This indicates the allocation of resources for electricity services. s Processing speed This indicates the allocation of resources for electricity services. s transmission speed, Indicates electricity business s The probability of a non-empty queue in the processing queue of resource pool z. Indicates electricity business s The parameters of the Poisson process corresponding to the distribution of time intervals at which data packets leave the resource pool; The power business s Link latency after leaving resource pool z The calculation formula is as follows: In the formula, This indicates the size of the data to be transmitted at resource pool z.

8. The system as described in claim 7, characterized in that, The calculation formula for the resource constraints constructed by the solution module is as follows: In the formula, This represents the total number of power services deployed on the resource pool. This indicates that resource pool z is allocated to the power business. i The proportion of computing resources This indicates that resource pool z is allocated to the power business. i The proportion of bandwidth resources, This indicates the maximum amount of resources that the power business can obtain from the resource pool. This represents the set of resources available to the power business in the resource pool. This represents the processing speed of the power service when the containers in resource pool z use all computing resources. This represents the transmission speed of power services when containers in resource pool z utilize all bandwidth resources. This indicates that resource pool z is allocated to the power business. s The proportion of computing resources This indicates that resource pool z is allocated to the power business. s The proportion of bandwidth resources.

9. The system as described in claim 7, characterized in that, The solution module is specifically used for: The information is input into a pre-built optimization problem model for multi-task deployment and resource allocation with time delay control. An improved ant colony algorithm based on dynamic pheromone evaporation coefficient is used to solve the optimization problem model for multi-task deployment and resource allocation with time delay control, so as to obtain the deployment mode of each container of multiple power tasks in the resource pool of IoT agent. Based on the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, the near-end strategy optimization algorithm is used to solve the optimization problem model of the multi-task deployment and resource allocation of the latency control, so as to obtain the resource allocation method of the IoT agent's resource pool to each container.

10. The system as described in claim 9, characterized in that, The solution module employs an improved ant colony algorithm based on the dynamic pheromone evaporation coefficient to solve the optimization problem model of multi-task deployment and resource allocation for latency control, obtaining the deployment methods of multiple power tasks in each container of the IoT agent's resource pool, including: The computing resources in the resource pool of the IoT agent are used to initialize the pheromone of the ant colony algorithm; The disappearance of pheromone trajectories in each resource pool is calculated based on the dynamic pheromone evaporation coefficient. The probability of selecting each resource pool for task deployment is calculated based on the disappearance of the pheromone trajectory, and the next resource pool for deploying the power task is selected based on the probability. Based on the disappearance of the pheromone trajectory and the selection of the next power task, the resource pool is iteratively calculated using ant colonies to obtain the most frequent path planning in the ant colony path planning, which serves as the deployment method for multiple power tasks in each container of the IoT agent's resource pool.

11. The system as claimed in claim 10, characterized in that, The formula for calculating the pheromone in the ant colony algorithm initialization of the solution module is as follows: In the formula, Indicates resource pool j The corresponding initial pheromone, Indicates resource pool j Available memory, subscript m Represents memory, Indicates resource pool j Total memory, This represents the total memory of all resource pools. Indicates resource pool j The available CPUs, where the subscript p indicates the CPU. Indicates resource pool j Total CPU, This represents the total CPU across all resource pools.

12. The system as described in claim 10, characterized in that, The calculation formula for the disappearance of pheromone trajectories by the solution module is as follows: In the formula, express t Time Resource Pool j The corresponding pheromone trajectory disappeared. Indicates resource pool j The corresponding initial pheromone, Indicates changes in pheromones, Indicates the first k Resource pools available for selection in the next iteration Indicates the dynamic pheromone evaporation coefficient; The dynamic pheromone volatility coefficient The calculation formula is as follows: 。 13. A computer device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the IoT agent multi-task scheduling method based on latency control as described in any one of claims 1 to 6 is implemented.

14. A computer-readable storage medium, characterized in that, It contains a computer program, which, when executed, implements the IoT agent multi-task scheduling method based on time delay control as described in any one of claims 1 to 6.

Images