Task allocation and unloading joint optimization method for agent communication network
By decomposing the task offloading problem into subproblems and optimizing them alternately in the intelligent agent communication network, and combining minimum cost flow and convex approximation algorithms, the problems of high computational complexity and poor optimization effect of task offloading are solved, and low latency and low energy consumption optimization are achieved in large-scale scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING INST OF TECH
- Filing Date
- 2025-12-01
- Publication Date
- 2026-04-24
AI Technical Summary
In intelligent agent communication networks, the task offloading problem presents itself as a non-convex mixed-integer nonlinear programming NP-hard problem. Existing technologies are difficult to solve effectively in large-scale scenarios, especially when the number of user and assistant agents increases, resulting in high computational complexity, poor optimization performance, and insufficient convergence.
The task unloading problem is decomposed into task unloading subproblems and resource allocation subproblems using a block coordinate descent framework. Each subproblem is optimized by alternating iterations. The minimum cost flow and convex approximation algorithms are combined to reduce computational complexity and ensure convergence. A steward agent is used for task allocation and resource allocation.
It significantly reduces task execution latency and energy consumption, improves computing power and scalability, and can stably converge to near-optimal solutions in large-scale intelligent agent networks.
Smart Images
Figure CN121924497A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of communication technology, and in particular to a joint optimization method for task allocation and offloading in intelligent agent communication networks. Background Technology
[0002] Unlike Mobile Edge Computing (MEC) in related technologies, agents in Agent Communication Networks (ACNs) typically have similar computing power and equal status. There is no significant difference in computing power or master-slave relationship like in MEC's "user-edge server" model. Furthermore, users only need to compute their pre-defined private tasks, and there is no task negotiation between users. This gives the task offloading problem entirely new characteristics in ACNs. In an ACN environment, the steward agent needs to simultaneously determine the allocation ratio of public tasks and the wireless resource allocation strategy, distributing the assigned tasks to users. Users then choose, based on the task offloading matrix, whether to have other idle assistant agents around them help with the offloading computation. This problem involves multiple continuous and discrete variables coupled together, making it an NP-hard mixed-integer nonlinear programming problem with nonconvex properties.
[0003] As the number of users and assistant agents increases, exhaustive search or branch-and-bound methods become virtually infeasible due to the exponential growth in computational load. For example, even in small-scale scenarios, such as a combination of two users and three assistants, accurately solving the problem using the BnB algorithm can consume significant time and computational resources. In related technologies, researchers have proposed offloading optimization schemes based on convex optimization, game theory, and deep reinforcement learning. However, these schemes typically only consider the joint optimization of wireless resource allocation and offloading decisions, often assuming that the task size is predetermined and neglecting the optimization process of allocating a proportion of common tasks. This assumption does not hold true in the ACN scenario because common tasks need to be actively allocated to users by the steward agent, a step that has a decisive impact on the latency and energy consumption of the entire system.
[0004] Furthermore, the related technologies suffer from two prominent problems: first, poor scalability—as the number of users or assistants increases, the precise branch-and-bound algorithm exhibits exponential complexity, making it difficult to adapt to large-scale agent networks; second, insufficient convergence—while deep reinforcement learning-based methods can reduce complexity to some extent, they typically lack stable convergence and global optimum guarantees; game theory-based equilibrium solutions often deviate significantly from the optimal solution. Moreover, when continuous variables such as task proportions, bandwidth, and power coexist with discrete variables such as offloading decisions, how to reduce complexity while maintaining optimization accuracy remains a pressing issue. Summary of the Invention
[0005] This application provides a joint optimization method for task allocation and offloading in intelligent agent communication networks to solve the problems of high computational complexity and poor optimization effect of task offloading. It can still significantly reduce task execution latency and energy consumption in large-scale ACN deployment scenarios.
[0006] A first aspect of this application provides a joint optimization method for task allocation and offloading in an intelligent agent communication network. The method is applied to an intelligent agent communication network, which includes a steward intelligent agent, at least one user intelligent agent, and at least one assistant intelligent agent. The method includes the following steps: Obtain the total task load to be allocated, the current offload matrix, the current proportion of common tasks, the current bandwidth, the current power allocation vector, and the current total number of iterations; The system utility function of the current iteration round is calculated based on the current unloading matrix, the current common task ratio, the current bandwidth, and the current power allocation vector. Based on the system utility function of the current iteration round, the system utility function of the previous iteration round, and the current total number of iterations, it is determined whether the preset iteration termination condition is met. If the preset iteration termination condition is met, based on the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector, the total task load is allocated to the at least one user agent through the steward agent, so that each user agent determines the target execution mode of the assigned task based on the current offload matrix, wherein the target execution mode includes executing the assigned task locally or executing the assigned task through the at least one assistant agent.
[0007] Optionally, in some embodiments, obtaining the current offload matrix, current common task ratio, current bandwidth, and current power allocation vector includes: Get the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector; Based on the previous common task ratio, the previous bandwidth, and the previous power allocation vector, the current offloading matrix is determined through a preset task offloading optimization model; Based on the current offload matrix, the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector are determined through a preset resource allocation model.
[0008] Optionally, in some embodiments, obtaining the previous offload matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector includes: Determine whether the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector exist; If the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector do not exist, then the initial common task ratio is used as the previous common task ratio, the initial bandwidth is used as the previous bandwidth, and the initial power allocation vector is used as the previous power allocation vector.
[0009] Optionally, in some embodiments, after determining whether a preset iteration termination condition is met, the process includes: If the preset iteration termination condition is not met, a new unloading matrix is determined based on the current common task ratio, current bandwidth, and current power allocation vector through a preset task unloading optimization model, and the current total number of iterations is accumulated. Based on the new offload matrix, a new proportion of common tasks, new bandwidth, and new power allocation vector are determined through a preset resource allocation model. The system utility function for the current iteration is calculated based on the new unloading matrix, the new common task ratio, the new bandwidth, and the new power allocation vector, until the preset iteration termination condition is met.
[0010] Optionally, in some embodiments, the preset task unloading optimization model is:
[0011] in, To unload the decision matrix, The weighting coefficients for system delay are: For system latency, The weighting factor for system energy consumption. For system energy consumption, This is the fifth constraint. For the number of user agents, For the user's intelligent agent, For user intelligent agents Uninstall to Assistant Smart Agent To perform the task, For the number of assistant agents, As an assistant intelligent agent, This is the sixth constraint. This is the seventh constraint.
[0012] Optionally, in some embodiments, the preset resource allocation model is:
[0013] in, Assign a proportional vector to the common tasks of user agents. For bandwidth, For power allocation vector, For user intelligent agents The proportion of public tasks allocated, For user intelligent agents bandwidth, For user intelligent agents downlink transmission power, This represents the total downlink transmit power. The total system bandwidth, This is the first constraint condition. This is the second constraint. This is the third constraint. This is the fourth constraint. .
[0014] Optionally, in some embodiments, the preset iteration termination condition is: The absolute value of the difference between the system utility function of the current iteration and the system utility function of the previous iteration is less than a first preset threshold, and / or the current total number of iterations is less than or equal to a second preset threshold.
[0015] A second aspect of this application provides a task allocation and offloading joint optimization apparatus for an intelligent agent communication network. The apparatus is applied to an intelligent agent communication network, which includes a steward intelligent agent, at least one user intelligent agent, and at least one assistant intelligent agent. The apparatus includes: The acquisition module is used to obtain the total task load to be allocated, the current offload matrix, the current proportion of common tasks, the current bandwidth, the current power allocation vector, and the current total number of iterations; The judgment module is used to calculate the system utility function of the current iteration round based on the current unloading matrix, the current common task ratio, the current bandwidth and the current power allocation vector, and to determine whether the preset iteration termination condition is met based on the system utility function of the current iteration round, the system utility function of the previous iteration round and the current total number of iterations. An execution module is configured to, upon satisfying the preset iteration termination condition, allocate the total task load to at least one user agent through the steward agent based on the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector, such that each user agent determines the target execution method of the assigned task based on the current offload matrix, wherein the target execution method includes executing the assigned task locally or executing the assigned task through the at least one assistant agent.
[0016] Optionally, in some embodiments, the acquisition module includes: The acquisition unit is used to acquire the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector. The first determining unit is used to determine the current unloading matrix based on the previous common task ratio, the previous bandwidth, and the previous power allocation vector, using a preset task unloading optimization model. The second determining unit is used to determine the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector based on the current offload matrix and through a preset resource allocation model.
[0017] Optionally, in some embodiments, the acquisition unit includes: The judgment subunit is used to determine whether the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector exist. A sub-unit is defined to, in the absence of the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector, use the initial common task ratio as the previous common task ratio, the initial bandwidth as the previous bandwidth, and the initial power allocation vector as the previous power allocation vector.
[0018] Optionally, in some embodiments, after determining whether a preset iteration termination condition is met, the determining module includes: The third determining unit is used to determine a new unloading matrix based on the current common task ratio, current bandwidth and current power allocation vector, and a preset task unloading optimization model, and to accumulate the current total number of iterations, if the preset iteration termination condition is not met. The fourth determining unit is used to determine a new common task ratio, a new bandwidth, and a new power allocation vector based on the new offloading matrix and a preset resource allocation model. The calculation unit is used to calculate the system utility function of the current iteration round based on the new unloading matrix, the new common task ratio, the new bandwidth, and the new power allocation vector, until the preset iteration termination condition is met.
[0019] Optionally, in some embodiments, the preset task unloading optimization model is:
[0020] in, To unload the decision matrix, The weighting coefficients for system delay are: For system latency, The weighting factor for system energy consumption. For system energy consumption, This is the fifth constraint. For the number of user agents, For the user's intelligent agent, For user intelligent agents Uninstall to Assistant Smart Agent To perform the task, For the number of assistant agents, As an assistant intelligent agent, This is the sixth constraint. This is the seventh constraint.
[0021] Optionally, in some embodiments, the preset resource allocation model is:
[0022] in, Assign a proportional vector to the common tasks of user agents. For bandwidth, For power allocation vector, For user intelligent agents The proportion of public tasks allocated, For user intelligent agents bandwidth, For user intelligent agents downlink transmission power, This represents the total downlink transmit power. The total system bandwidth, This is the first constraint condition. This is the second constraint. This is the third constraint. This is the fourth constraint. This represents the proportion of total public tasks to all tasks.
[0023] Optionally, in some embodiments, the preset iteration termination condition is: The absolute value of the difference between the system utility function of the current iteration and the system utility function of the previous iteration is less than a first preset threshold, and / or the current total number of iterations is less than or equal to a second preset threshold.
[0024] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement a joint optimization method for task allocation and offloading in an intelligent agent communication network as described in the above embodiments.
[0025] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a joint optimization method for task allocation and offloading in an intelligent agent communication network as described in the above embodiments.
[0026] Therefore, this application has the following beneficial effects: (1) Strong computational power: Joint optimization problems in related technologies are usually non-convex large-scale mixed integer problems, which are extremely difficult to solve directly. This application divides the problem into task unloading subproblems and resource allocation subproblems based on the block coordinate descent framework and optimizes them alternately. Each time, only local subproblems are processed, which greatly reduces the computational complexity of each step.
[0027] (2) Support for alternating iteration and convergence control: This application implements alternating iteration of task unloading and resource allocation through a block coordinate descent scheme, and uses the minimum cost flow and convex approximation algorithm to ensure that subproblems converge in each iteration. Compared with the one-time global optimization method that requires updating all variables at the same time, block coordinate descent optimizes only one variable block at a time, decomposing the original problem into a series of low-dimensional subproblems. Under appropriate conditions, it can stably converge to the near-optimal solution or critical point, and has the advantages of simple implementation, high memory efficiency and easy parallelization.
[0028] (3) Strong scalability: By using a block coordinate descent-based scheme, it supports adding users and server / resource dimensions without destroying the subproblem structure. Combining different optimization techniques such as convex approximation and minimum cost flow, it keeps the computational complexity controllable in complex systems and reduces system latency and energy consumption to a certain extent.
[0029] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description
[0030] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a joint optimization method for task allocation and offloading in an intelligent agent communication network according to an embodiment of this application; Figure 2 This is a schematic diagram of an ACN task unloading system model provided according to an embodiment of this application; Figure 3 This is a schematic diagram illustrating the principle of a joint optimization method for task allocation and offloading in an intelligent agent communication network according to an embodiment of this application; Figure 4This is a schematic diagram of a task allocation and offloading joint optimization device for an intelligent agent communication network according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation
[0031] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.
[0032] The following describes a joint optimization method for task allocation and offloading in an agent communication network, based on embodiments of the present application, with reference to the accompanying drawings. Addressing the issues of high computational complexity and poor optimization effect in task offloading mentioned in the background art, this application provides a joint optimization method for task allocation and offloading in an agent communication network. In this method, the total task load to be allocated, the current offloading matrix, the current common task ratio, the current bandwidth, the current power allocation vector, and the current total number of iterations are obtained. The system utility function for the current iteration round is calculated based on the current offloading matrix, the current common task ratio, the current bandwidth, and the current power allocation vector. Based on the system utility function for the current iteration round, the system utility function for the previous iteration round, and the current total number of iterations, it is determined whether a preset iteration termination condition is met. If the preset iteration termination condition is met, the total task load is allocated to at least one user agent through a steward agent, based on the current offloading matrix, the current common task ratio, the current bandwidth, and the current power allocation vector. Each user agent determines the target execution method for the allocated task based on the current offloading matrix. The target execution method includes executing the allocated task locally or executing the allocated task through at least one assistant agent. This solves the problems of high computational complexity and poor optimization effect of task unloading, and can still significantly reduce task execution latency and energy consumption in large-scale ACN deployment scenarios.
[0033] Specifically, Figure 1 This is a flowchart illustrating a joint optimization method for task allocation and unloading in an intelligent agent communication network, as provided in an embodiment of this application.
[0034] It should be noted that the task allocation and offloading joint optimization method for intelligent agent communication networks in this application embodiment is applied to an intelligent agent communication network, which includes a steward intelligent agent, at least one user intelligent agent, and at least one assistant intelligent agent.
[0035] Before introducing a joint optimization method for task allocation and offloading in intelligent agent communication networks according to an embodiment of this application, the principle of the embodiment of this application can be introduced first.
[0036] The embodiments of this application can be deployed in an FDMA (Frequency Division Multiple Access) transmission system model, such as... Figure 2 As shown, the model includes one butler agent and There are 10 intelligent agents, among which... Among the intelligent agents, there are A device with a private task is called a user device, and the rest... Each device is an assistant device, and each device has... The root antenna. The steward agent needs to be responsible for allocating the total task load. According to the proportion of public tasks and the proportion of private tasks Assigned to users, where each user has a percentage of private tasks. This is predetermined; the administrator needs to collaboratively optimize the shared task allocation ratio for each user during the task distribution phase. ,bandwidth and downlink transmission power After the task is assigned to the user, the user can either execute the task locally or choose to unload the decision matrix via binary. The task is offloaded to the assistant AI agent for execution, where... Indicates user Uninstall to Assistant Perform the task. Indicates user Choose to uninstall locally.
[0037] In the task unloading model of this application embodiment, Represents the total downlink transmit power. Represents the total system bandwidth. and These represent the weighting coefficients for system latency and energy consumption, respectively. For each user, latency is divided into uplink latency and uplink latency. downlink latency and users -assistant Delay For the overall system, system latency For each user, energy consumption is also divided into uplink energy consumption. Downlink energy consumption and users -assistant Energy consumption Therefore, the total energy consumption of the system is Therefore, the total system delay Total energy consumption The joint optimization problem can be expressed as:
[0038] Among them, constraints , 、 This represents the sum of constraints on the calculation results, task ratio, allocated bandwidth, and downlink transmit power. The proportion of tasks assigned is a continuous variable, while This indicates that the elements of the unloading decision matrix are binary, and the constraints are... Each assistant can only help one user, a constraint. This indicates that users can either compute tasks locally or offload the entire task to a maximum of one assistant.
[0039] The embodiments of this application can decompose the original problem into two sub-problems: First, the first sub-problem It is a task unloading subproblem of optimizing discrete variables, that is, given continuous variables... Under the condition of optimizing the unloading decision matrix ,Right now:
[0040] And when Given that uplink and downlink latency and power consumption are known values for each user, Uninstall to Assistant Between and It is also a known value. individual users and Each assistant will generate a total of In the case of unloading / local computation, by enumerating the latency of these edges, we first determine whether the constraints can be satisfied. , and Maximum delay edge and the corresponding set of all possible unloading matrices The network is constructed as a minimum flow problem, and the optimal solution is found using the network simplex method. And the final system utility function Sort the matrices and select the one with the smallest utility function, which is the optimal unloading matrix. .
[0041] Secondly, the second sub-problem It is a subproblem of optimizing resource allocation with continuous variables, that is, given discrete variables... In this case, optimize the proportion of public tasks. Bandwidth allocation ratio and downlink transmission power ,Right now:
[0042] Among them, uplink latency downlink latency and users -assistant Delay Uplink energy consumption and users -assistant Energy consumption All terms are nonconvex. An approximate function and a quadratic penalty term are constructed through a first-order Taylor expansion. The convex approximation method is used to solve the problem in each iteration to ensure monotonic convergence.
[0043] The block coordinate descent algorithm framework can iteratively optimize subproblems. The problem of succubi Update the continuous and discrete variables until the objective function converges.
[0044] Based on the above principles, the following is combined with Figure 1 This application introduces a joint optimization method for task allocation and offloading in intelligent agent communication networks.
[0045] like Figure 1 As shown, the joint optimization method for task allocation and offloading in intelligent agent communication networks includes the following steps: In step S101, the total task load to be allocated, the current offload matrix, the current common task ratio, the current bandwidth, the current power allocation vector, and the current total number of iterations are obtained.
[0046] Furthermore, in some embodiments, obtaining the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector includes: obtaining the previous offload matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector; determining the current offload matrix based on the previous common task ratio, the previous bandwidth, and the previous power allocation vector using a preset task offload optimization model; and determining the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector based on the current offload matrix using a preset resource allocation model.
[0047] Optionally, in some embodiments, obtaining the previous offload matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector includes: determining whether the previous offload matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector exist; if the previous offload matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector do not exist, then the initial common task ratio is used as the previous common task ratio, the initial bandwidth is used as the previous bandwidth, and the initial power allocation vector is used as the previous power allocation vector.
[0048] Specifically, in combination Figure 3 As shown, embodiments of this application can establish a system as follows: Figure 2 The FDMA transmission system model shown includes one steward agent and There are 10 intelligent agent nodes, including individual user devices and Each device has an assistant device. The root antenna. The steward agent is responsible for allocating the overall task load. Divide it into public task proportions Ratio of private tasks and allocate downlink power to each user. and bandwidth .
[0049] Combination Figure 3 As shown, after establishing the FDMA transmission system model, system parameters can be initialized, including: maximum downlink transmit power. Total system bandwidth Weighting coefficients and Set the initial common task ratio. ,bandwidth and power distribution Vectors, and the initial unloading matrix .
[0050] The embodiments of this application can first solve the task unloading subproblem, as follows: Step 201: Given continuous variables Under the given conditions, a pre-defined task offloading optimization model is established, with the objective of minimizing the system utility function while satisfying the constraints. .in, For the total system latency, This represents the total energy consumption of the system.
[0051] The preset task unloading optimization model is as follows:
[0052] in, To unload the decision matrix, The weighting coefficients for system delay are: For system latency, The weighting factor for system energy consumption. For system energy consumption, This is the fifth constraint. For the number of user agents, For the user's intelligent agent, For user intelligent agents Uninstall to Assistant Smart Agent To perform the task, For the number of assistant agents, As an assistant intelligent agent, This is the sixth constraint. This is the seventh constraint.
[0053] Step 202: Calculate the latency and energy consumption for each user under different uninstallation options, including uplink latency. Downlink latency ,user -assistant Delay and the corresponding energy consumption , , .
[0054] Step 203: Enumeration individual users and Given all possible combinations of offload / local computation for each assistant, we obtain the set of edges for user-assistant pairs and their corresponding latency and energy costs.
[0055] Step 204: Construct the minimum cost flow problem: Establish virtual source nodes, user nodes, and virtual assistant nodes, with the cost being the calculated energy consumption index.
[0056] Step 205: Solve the minimum cost flow problem using the network simplex method to obtain all feasible unloading matrices. gather .
[0057] Step 206: Based on the system utility function Sort the candidate solutions to the unloading matrix and select the one with the lowest utility. As the optimal uninstallation solution.
[0058] Then, the resource allocation sub-problem can be solved in this embodiment of the application, as follows: Step 301: Given the unloading matrix Under the given conditions, a pre-defined resource allocation optimization model is established, with the goal of optimizing the proportion of common tasks while satisfying power, bandwidth, and task ratio constraints. Bandwidth allocation and downlink power allocation .
[0059] The default resource allocation model is:
[0060] in, Assign a proportional vector to the common tasks of user agents. For bandwidth, For power allocation vector, For user intelligent agents The proportion of public tasks allocated, For user intelligent agents bandwidth, For user intelligent agents downlink transmission power, This represents the total downlink transmit power. The total system bandwidth, This is the first constraint condition. This is the second constraint. This is the third constraint. This is the fourth constraint. This represents the proportion of total public tasks to all tasks.
[0061] Step 302: Perform a first-order Taylor expansion on the non-convex terms in the energy consumption and time delay formulas to obtain an approximate convex function. Introduce a quadratic penalty term constraint in the objective function to ensure that the update of continuous variables satisfies the feasible region condition.
[0062] Step 303: Iteratively solve the optimization problem using the successive convex approximation method. Each iteration performs a linear approximation within the convex domain and updates the... .
[0063] Step 304: Determine the convergence of the objective function: If the improvement in two consecutive iterations is less than the threshold... If the iteration stops, the resource allocation result will be output.
[0064] In step S102, the system utility function of the current iteration round is calculated based on the current unloading matrix, the current common task ratio, the current bandwidth, and the current power allocation vector. Based on the system utility function of the current iteration round, the system utility function of the previous iteration round, and the current total number of iterations, it is determined whether the preset iteration termination condition is met.
[0065] Optionally, in some embodiments, after determining whether a preset iteration termination condition is met, the process includes: if the preset iteration termination condition is not met, determining a new offloading matrix based on the current common task ratio, current bandwidth, and current power allocation vector using a preset task offloading optimization model, and accumulating the current total number of iterations; determining a new common task ratio, new bandwidth, and new power allocation vector based on the new offloading matrix using a preset resource allocation model; calculating the system utility function for the current iteration round based on the new offloading matrix, new common task ratio, new bandwidth, and new power allocation vector, until the preset iteration termination condition is met.
[0066] The preset iteration termination conditions are: the absolute value of the difference between the system utility function of the current iteration and the system utility function of the previous iteration is less than a first preset threshold, and / or the current total number of iterations is less than or equal to a second preset threshold.
[0067] Specifically, in combination Figure 3 As shown, the embodiments of this application can be iteratively optimized based on the block coordinate descent iterative optimization algorithm framework, as detailed below: Step 401: Initialize the iteration count Set the maximum number of iterations. and convergence threshold .
[0068] Step 402: In the In this iteration, steps 201-206 are executed to obtain the unloading matrix. .
[0069] Step 403: After obtaining Under the given conditions, execute steps 301-304 to obtain the solution for continuous variables. .
[0070] Step 404: Calculate the system utility function .like or If the iteration terminates, then the iteration ends; otherwise... Repeat steps 402 to 403.
[0071] In step S103, if the preset iteration termination condition is met, the total task load is allocated to at least one user agent through the steward agent based on the current offload matrix, the current common task ratio, the current bandwidth and the current power allocation vector, so that each user agent determines the target execution method of the assigned task based on the current offload matrix. The target execution method includes executing the assigned task locally or executing the assigned task through at least one assistant agent.
[0072] Specifically, in combination Figure 3 As shown, in this embodiment of the application, an unloading matrix is output when a preset iteration termination condition is met. With resource allocation solution As the final optimization result, based on the final optimization result, the total task load is distributed to at least one user agent through the steward agent, so that each user agent determines the target execution method of the assigned task based on the current offload matrix. The target execution method includes executing the assigned task locally or executing the assigned task through at least one assistant agent.
[0073] According to an embodiment of this application, a joint optimization method for task allocation and offloading in intelligent agent communication networks (ACNs) is proposed. This method obtains the total task load to be allocated, the current offloading matrix, the current common task ratio, the current bandwidth, the current power allocation vector, and the current total number of iterations. It then calculates the system utility function for the current iteration based on these parameters. Based on the system utility function of the current iteration, the system utility function of the previous iteration, and the current total number of iterations, it determines whether a preset iteration termination condition is met. If the preset termination condition is met, the total task load is allocated to at least one user agent through a steward agent, based on the current offloading matrix, the current common task ratio, the current bandwidth, and the current power allocation vector. Each user agent determines the target execution method for the allocated task based on the current offloading matrix. The target execution method includes executing the allocated task locally or through at least one assistant agent. This solves the problems of high computational complexity and poor optimization effect in task offloading, and can significantly reduce task execution latency and energy consumption even in large-scale ACN deployment scenarios.
[0074] Next, referring to the accompanying drawings, an apparatus for a joint optimization method of task allocation and unloading for intelligent agent communication networks is described according to an embodiment of this application.
[0075] Figure 4 This is a block diagram illustrating a task allocation and offloading joint optimization device for an intelligent agent communication network according to an embodiment of this application. It should be noted that the task allocation and offloading joint optimization device for an intelligent agent communication network is applied to an intelligent agent communication network, which includes a steward intelligent agent, at least one user intelligent agent, and at least one assistant intelligent agent.
[0076] like Figure 4 As shown, the task allocation and offloading joint optimization device 10 for intelligent agent communication networks includes: an acquisition module 100, a judgment module 200, and an execution module 300.
[0077] The acquisition module 100 is used to acquire the total task load to be allocated, the current offload matrix, the current common task ratio, the current bandwidth, the current power allocation vector, and the current total number of iterations.
[0078] The judgment module 200 is used to calculate the system utility function of the current iteration round based on the current unloading matrix, the current common task ratio, the current bandwidth and the current power allocation vector, and to determine whether the preset iteration termination condition is met based on the system utility function of the current iteration round, the system utility function of the previous iteration round and the current total number of iterations.
[0079] The execution module 300 is used to distribute the total task load to at least one user agent through the steward agent, based on the current offload matrix, the current common task ratio, the current bandwidth and the current power allocation vector, when the preset iteration termination conditions are met. This allows each user agent to determine the target execution method of the assigned task based on the current offload matrix. The target execution method includes executing the assigned task locally or executing the assigned task through at least one assistant agent.
[0080] Optionally, in some embodiments, the acquisition module 100 includes: an acquisition unit, a first determination unit, and a second determination unit.
[0081] The acquisition unit is used to acquire the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector.
[0082] The first determining unit is used to determine the current unloading matrix based on the previous common task ratio, the previous bandwidth, and the previous power allocation vector, using a preset task unloading optimization model.
[0083] The second determining unit is used to determine the current offload matrix, the current proportion of common tasks, the current bandwidth, and the current power allocation vector based on the current offload matrix and a preset resource allocation model.
[0084] Optionally, in some embodiments, the acquisition unit includes: a judgment subunit and a determination subunit.
[0085] The judgment sub-unit is used to determine whether the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector exist.
[0086] The sub-unit is determined to use the initial common task ratio as the previous common task ratio, the initial bandwidth as the previous bandwidth, and the initial power allocation vector as the previous power allocation vector when the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector are not available.
[0087] Optionally, in some embodiments, after determining whether a preset iteration termination condition is met, the determination module 200 includes: a third determination unit, a fourth determination unit, and a calculation unit.
[0088] The third determining unit is used to determine a new unloading matrix based on the current common task ratio, current bandwidth, and current power allocation vector, and to accumulate the current total number of iterations, when the preset iteration termination condition is not met.
[0089] The fourth determining unit is used to determine the new proportion of common tasks, the new bandwidth, and the new power allocation vector based on the new offloading matrix and a preset resource allocation model.
[0090] The computational unit is used to calculate the system utility function of the current iteration based on the new unloading matrix, the new common task ratio, the new bandwidth, and the new power allocation vector, until the preset iteration termination condition is met.
[0091] Optionally, in some embodiments, the preset task unloading optimization model is:
[0092] in, To unload the decision matrix, The weighting coefficients for system delay are: For system latency, The weighting factor for system energy consumption. For system energy consumption, This is the fifth constraint. For the number of user agents, For the user's intelligent agent, For user intelligent agents Uninstall to Assistant Smart Agent To perform the task, For the number of assistant agents, As an assistant intelligent agent, This is the sixth constraint. This is the seventh constraint.
[0093] Optionally, in some embodiments, the preset resource allocation model is:
[0094] in, Assign a proportional vector to the common tasks of all user agents. For bandwidth, For power allocation vector, For user intelligent agents The proportion of public tasks allocated, For user intelligent agents bandwidth, For user intelligent agents downlink transmission power, This represents the total downlink transmit power. The total system bandwidth, This is the first constraint condition. This is the second constraint. This is the third constraint. This is the fourth constraint. This represents the proportion of total public tasks to all tasks.
[0095] Optionally, in some embodiments, the preset iteration termination condition is: the absolute value of the difference between the system utility function of the current iteration and the system utility function of the previous iteration is less than a first preset threshold, and / or the current total number of iterations is less than or equal to a second preset threshold.
[0096] It should be noted that the foregoing explanation of an embodiment of a task allocation and offloading joint optimization method for intelligent agent communication networks also applies to an embodiment of a task allocation and offloading joint optimization device for intelligent agent communication networks, and will not be repeated here.
[0097] According to an embodiment of this application, a joint optimization device for task allocation and offloading in intelligent agent communication networks (ACNs) is proposed. This device acquires the total task load to be allocated, the current offloading matrix, the current common task ratio, the current bandwidth, the current power allocation vector, and the current total number of iterations. It then calculates the system utility function for the current iteration based on these parameters. Based on the system utility function of the current iteration, the system utility function of the previous iteration, and the current total number of iterations, it determines whether a preset iteration termination condition is met. If the preset termination condition is met, the device allocates the total task load to at least one user agent through a steward agent, based on the current offloading matrix, the current common task ratio, the current bandwidth, and the current power allocation vector. Each user agent determines the target execution method for the allocated task based on the current offloading matrix. The target execution method includes executing the allocated task locally or through at least one assistant agent. This solves the problems of high computational complexity and poor optimization effect in task offloading, and can significantly reduce task execution latency and energy consumption even in large-scale ACN deployment scenarios.
[0098] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.
[0099] When the processor 502 executes the program, it implements the task allocation and unloading joint optimization method for intelligent agent communication networks provided in the above embodiments.
[0100] Furthermore, electronic devices also include: Communication interface 503 is used for communication between memory 501 and processor 502.
[0101] The memory 501 is used to store computer programs that can run on the processor 502.
[0102] The memory 501 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.
[0103] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0104] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.
[0105] The processor 502 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.
[0106] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described joint optimization method for task allocation and unloading in an intelligent agent communication network.
[0107] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0108] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0109] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.
[0110] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.
[0111] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.
[0112] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.
Claims
1. A joint optimization method for task allocation and offloading in intelligent agent communication networks, characterized in that, The method is applied to an intelligent agent communication network, the intelligent agent communication network including a butler intelligent agent, at least one user intelligent agent, and at least one assistant intelligent agent, the method including the following steps: Obtain the total task load to be allocated, the current offload matrix, the current proportion of common tasks, the current bandwidth, the current power allocation vector, and the current total number of iterations; The system utility function of the current iteration round is calculated based on the current unloading matrix, the current common task ratio, the current bandwidth, and the current power allocation vector. Based on the system utility function of the current iteration round, the system utility function of the previous iteration round, and the current total number of iterations, it is determined whether the preset iteration termination condition is met. If the preset iteration termination condition is met, based on the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector, the total task load is allocated to the at least one user agent through the steward agent, so that each user agent determines the target execution method of the assigned task based on the current offload matrix, wherein the target execution method includes executing the assigned task locally or executing the assigned task through the at least one assistant agent.
2. The method according to claim 1, characterized in that, Obtain the current unloading matrix, current common task ratio, current bandwidth, and current power allocation vector, including: Get the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector; Based on the previous common task ratio, the previous bandwidth, and the previous power allocation vector, the current offloading matrix is determined through a preset task offloading optimization model; Based on the current offload matrix, the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector are determined through a preset resource allocation model.
3. The method according to claim 2, characterized in that, The process of obtaining the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector includes: Determine whether the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector exist; If the previous unloading matrix, the previous common task ratio, the previous bandwidth, and the previous power allocation vector do not exist, then the initial common task ratio is used as the previous common task ratio, the initial bandwidth is used as the previous bandwidth, and the initial power allocation vector is used as the previous power allocation vector.
4. The method according to claim 1, characterized in that, After determining whether the preset iteration termination condition is met, the process includes: If the preset iteration termination condition is not met, a new unloading matrix is determined based on the current common task ratio, current bandwidth, and current power allocation vector through a preset task unloading optimization model, and the current total number of iterations is accumulated. Based on the new offload matrix, a new proportion of common tasks, new bandwidth, and new power allocation vector are determined through a preset resource allocation model. The system utility function for the current iteration is calculated based on the new unloading matrix, the new common task ratio, the new bandwidth, and the new power allocation vector, until the preset iteration termination condition is met.
5. The method according to claim 2, characterized in that, The preset task unloading optimization model is as follows: in, To unload the decision matrix, The weighting coefficients for system delay are: For system latency, The weighting factor for system energy consumption. For system energy consumption, This is the fifth constraint. For the number of user agents, For the user's intelligent agent, For user intelligent agents Uninstall to Assistant Smart Agent To perform the task, For the number of assistant agents, As an assistant intelligent agent, This is the sixth constraint. This is the seventh constraint.
6. The method according to claim 2, characterized in that, The preset resource allocation model is as follows: in, Assign a proportional vector to the common tasks of user agents. For bandwidth, For power allocation vector, For user intelligent agents The proportion of public tasks allocated, For user intelligent agents bandwidth, For user intelligent agents downlink transmission power, This represents the total downlink transmit power. The total system bandwidth, This is the first constraint condition. This is the second constraint. This is the third constraint. This is the fourth constraint. This represents the proportion of total public tasks to all tasks.
7. The method according to claim 1, characterized in that, The preset iteration termination condition is: The absolute value of the difference between the system utility function of the current iteration and the system utility function of the previous iteration is less than a first preset threshold, and / or the current total number of iterations is less than or equal to a second preset threshold.
8. A joint optimization device for task allocation and offloading in intelligent agent communication networks, characterized in that, The device is applied to an intelligent agent communication network, the intelligent agent communication network including a butler intelligent agent, at least one user intelligent agent, and at least one assistant intelligent agent, the device comprising: The acquisition module is used to obtain the total task load to be allocated, the current offload matrix, the current proportion of common tasks, the current bandwidth, the current power allocation vector, and the current total number of iterations; The judgment module is used to calculate the system utility function of the current iteration round based on the current unloading matrix, the current common task ratio, the current bandwidth and the current power allocation vector, and to determine whether the preset iteration termination condition is met based on the system utility function of the current iteration round, the system utility function of the previous iteration round and the current total number of iterations. An execution module is configured to, upon satisfying the preset iteration termination condition, allocate the total task load to at least one user agent through the steward agent based on the current offload matrix, the current common task ratio, the current bandwidth, and the current power allocation vector, such that each user agent determines the target execution method of the assigned task based on the current offload matrix, wherein the target execution method includes executing the assigned task locally or executing the assigned task through the at least one assistant agent.
9. An electronic device, characterized in that, include: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a joint optimization method for task allocation and offloading in an intelligent agent communication network as described in any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement a joint optimization method for task allocation and offloading in an agent communication network as described in any one of claims 1-7.