Task unloading and computing power distribution optimization method oriented to agent communication network

Abstract

The invention relates to a task unloading and computing power distribution optimization method for an agent communication network, and the method comprises the steps: obtaining user equipment parameters and assistant equipment parameters in a task unloading and computing power distribution system, calculating the first time delay and the first energy consumption of user equipment, and the second time delay and the second energy consumption of assistant equipment according to the user equipment parameters and the assistant equipment parameters, constructing a target optimization function of the total overhead of the system, taking the core of assistant equipment as a virtual assistant node, linearizing the maximum time delay item of the target optimization function to obtain a mixed integer linear programming problem, and converting the mixed integer linear programming problem into a minimum cost flow problem based on a pre-constructed delay bottleneck set; according to the method, the network flow graph comprising the virtual source node, the target node, the split user node and the virtual assistant node is constructed, and the network flow graph is solved to obtain the optimal allocation optimization scheme, so that the problem of relatively low allocation scheme solving efficiency caused by high algorithm complexity in related technologies is solved, and task unloading and computing power allocation efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method for optimizing task offloading and computing power allocation in intelligent agent communication networks. Background Technology

[0002] Intelligent agent communication networks are a crucial component of future 6G mobile communications. Their goal is to enable collaborative computing and information sharing among intelligent agents to support scenarios such as unmanned systems, industrial automation, and immersive virtual interaction. Typically, a central intelligent agent, acting as a steward, breaks down the user-submitted task into public and private components, distributing them to multiple user or assistant intelligent agents. Users need to compute their private tasks locally while deciding whether to offload the public task portion to a nearby assistant intelligent agent. In practical applications, assistant intelligent agents are often equipped with multi-core Central Processing Units (CPUs), allowing them to autonomously adjust the amount of computing power used to serve users, rather than constantly running at full power, thus improving energy efficiency and making efficient use of multi-core resources.

[0003] Traditional task offloading research in related technologies focuses on the asymmetric computing power structure of "user-server". It assumes that the server has a fixed computing power that is much higher than that of the user, and thus simplifies the task scheduling problem into a single decision problem of whether the user chooses to offload. It also assumes that the server's computing power is constant and can be allocated arbitrarily. At the same time, there are two main technical methods for computing power allocation: one is a simple allocation method based on task size, and the other is a high-precision algorithm based on exhaustive search and pruning.

[0004] However, the joint optimization problem of task unloading and computing power allocation is essentially a discrete mixed integer optimization problem, involving the optimization of two highly coupled discrete variable matrices, and there are complex coupling constraints between them, which makes it extremely difficult to solve directly. The computing power allocation methods of related technologies have obvious shortcomings. Allocation methods based on task size cannot guarantee the optimal solution, while high-precision algorithms based on exhaustive search and pruning result in excessively high computational complexity, which urgently needs to be addressed. Summary of the Invention

[0005] This invention provides an optimization method for task offloading and computing power allocation in intelligent agent communication networks, which solves the problem of low efficiency in solving allocation schemes due to high algorithm complexity in related technologies, and improves the efficiency of task offloading and computing power allocation.

[0006] The first aspect of this invention provides a method for optimizing task offloading and computing power allocation in an intelligent agent communication network, comprising the following steps: obtaining user device parameters and assistant device parameters in the task offloading and computing power allocation system; calculating a first latency and a first energy consumption of the user device, and a second latency and a second energy consumption of the assistant device based on the user device parameters and the assistant device parameters; constructing a target optimization function for the total system overhead based on the first latency, the first energy consumption, the second latency, and the second energy consumption; taking the core of the assistant device as a virtual assistant node; linearizing the maximum latency term of the target optimization function for the total system overhead to obtain a mixed-integer linear programming problem; transforming the mixed-integer linear programming problem into a minimum cost flow problem based on a pre-constructed set of latency bottlenecks; constructing a network flow graph including virtual source nodes, target nodes, split user nodes, and virtual assistant nodes; and using the network flow graph to solve the minimum cost flow problem to obtain the optimal allocation optimization scheme for the task offloading and computing power allocation system.

[0007] Further, in some embodiments, the step of obtaining user device parameters and assistant device parameters in the task offloading and computing power allocation system, and calculating the first latency and first energy consumption of the user device, and the second latency and second energy consumption of the assistant device based on the user device parameters and the assistant device parameters, includes: obtaining the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, the user task allocation ratio, the public task volume, and the private task volume; based on the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, and a preset binary offloading decision matrix and computing power allocation matrix, offloading some tasks of the user device to the assistant device according to the user task allocation ratio, the public task volume, and the private task volume, to obtain the first task execution strategy of the user device and the second task execution strategy of the assistant device; calculating the first latency and first energy consumption of the user device according to the first task execution strategy, and calculating the second latency and second energy consumption of the assistant device according to the second task execution strategy.

[0008] Further, in some embodiments, the step of using the core of the assistant device as a virtual assistant node and linearizing the maximum delay term of the objective optimization function of the total system overhead to obtain a mixed-integer linear programming problem, and transforming the mixed-integer linear programming problem into a minimum cost flow problem based on a pre-constructed set of latency bottlenecks, includes: generating a virtual assistant matrix based on the task unloading decision matrix, and generating a virtual assistant set based on the virtual assistant matrix and the CPU cores of the assistant device, and using the nodes in the virtual assistant set as the virtual assistant nodes; linearizing the maximum delay term of the objective optimization function of the total system overhead based on preset linear auxiliary variables to obtain a linearized objective optimization function of the total system overhead, and obtaining the mixed-integer linear programming problem based on the objective optimization function of the total system overhead; constructing a set of latency bottlenecks, and transforming the mixed-integer linear programming problem into the minimum cost flow problem based on the parameters in the set of latency bottlenecks and the virtual assistant nodes.

[0009] Furthermore, in some embodiments, constructing a network flow graph including virtual source nodes, virtual target nodes, user nodes, and virtual assistant nodes includes: based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem, splitting the user nodes into a first user node representing a sending flow and a second user node representing computing power; based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem, generating a first type of edge of the network flow graph according to the virtual source node and the first user node, generating a second type of edge of the network flow graph according to the first user node and the second user node, generating a third type of edge of the network flow graph according to the first user node and the virtual assistant node, and generating a fourth type of edge of the network flow graph according to the virtual assistant node and the virtual target node, to obtain the network flow graph.

[0010] Furthermore, in some embodiments, the preset linear auxiliary variables include multiple auxiliary variables. The step of using the network flow graph to solve the minimum cost flow problem to obtain the optimal allocation optimization scheme for the task unloading and computing power allocation system includes: using the network simplex method in the network flow graph, solving the minimum cost flow problem based on the multiple auxiliary variables to obtain the unloading decision matrix and objective function value for each auxiliary variable; determining the target auxiliary variable and its unloading decision matrix based on the objective function value of each auxiliary variable; obtaining the computing power allocation matrix based on the target auxiliary variable and its unloading decision matrix; and obtaining the optimal allocation optimization scheme based on the unloading decision matrix of the target auxiliary variable, the computing power allocation matrix, and the total system overhead.

[0011] According to an embodiment of the present invention, a method for optimizing task offloading and computing power allocation in intelligent agent communication networks is provided. First, user device parameters and assistant device parameters in the task offloading and computing power allocation system are used to calculate the latency and energy consumption of users and assistants. A total cost objective function is constructed, and with the assistant core as a virtual node, the maximum latency term is linearized to obtain a mixed integer programming problem, which is then transformed into a minimum cost flow problem by combining the latency bottleneck set. A network flow graph containing virtual source, target, split user, and virtual assistant nodes is constructed, and the optimal solution is obtained. This method solves the problem of low efficiency in solving allocation schemes due to high algorithm complexity in related technologies, thereby improving the efficiency of task offloading and computing power allocation.

[0012] A second aspect of the present invention provides a task offloading and computing power allocation optimization device for intelligent agent communication networks, wherein the device includes: The acquisition module is used to acquire user device parameters and assistant device parameters in the task unloading and computing power allocation system, and calculate the first latency and first energy consumption of the user device, and the second latency and second energy consumption of the assistant device based on the user device parameters and the assistant device parameters. The mathematical analysis module is used to construct a target optimization function for the total system cost based on the first delay, the first energy consumption, the second delay, and the second energy consumption. The core of the assistant device is used as a virtual assistant node, and the maximum delay term of the target optimization function for the total system cost is linearized to obtain a mixed integer linear programming problem. Based on a pre-constructed set of delay bottlenecks, the mixed integer linear programming problem is transformed into a minimum cost flow problem. The solution module is used to construct a network flow graph containing virtual source nodes, target nodes, split user nodes, and virtual assistant nodes, and use the network flow graph to solve the minimum cost flow problem to obtain the optimal allocation optimization scheme of the task unloading and computing power allocation system.

[0013] Further, in some embodiments, the acquisition module is specifically used to: acquire the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, the user task allocation ratio, the public task volume, and the private task volume; based on the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, a preset binary offloading decision matrix and computing power allocation matrix, and according to the user task allocation ratio, the public task volume, and the private task volume, offload some tasks of the user device to the assistant device to obtain the first task execution strategy of the user device and the second task execution strategy of the assistant device; calculate the first latency and first energy consumption of the user device according to the first task execution strategy, and calculate the second latency and second energy consumption of the assistant device according to the second task execution strategy.

[0014] Further, in some embodiments, the mathematical analysis module is specifically used for: generating a virtual assistant matrix based on the task unloading decision matrix, and generating a virtual assistant set based on the virtual assistant matrix and the CPU cores of the assistant device, and using the nodes in the virtual assistant set as the virtual assistant nodes; obtaining a linearized system total overhead objective optimization function by linearizing the maximum delay term of the objective optimization function based on preset linear auxiliary variables, and obtaining the mixed-integer linear programming problem based on the objective optimization function of the system total overhead; constructing a delay bottleneck set, and transforming the mixed-integer linear programming problem into the minimum cost flow problem based on the parameters in the delay bottleneck set and the virtual assistant nodes.

[0015] Further, in some embodiments, the solution module is specifically used to: based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem, split the user node into a first user node representing the sending flow and a second user node representing the computing power; based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem, generate a first type of edge of the network flow graph according to the virtual source node and the first user node, generate a second type of edge of the network flow graph according to the first user node and the second user node, generate a third type of edge of the network flow graph according to the first user node and the virtual assistant node, and generate a fourth type of edge of the network flow graph according to the virtual assistant node and the virtual target node, so as to obtain the network flow graph.

[0016] Furthermore, in some embodiments, the preset linear auxiliary variables include multiple auxiliary variables, and the solution module is further configured to: solve the minimum cost flow problem based on the network simplex method in the network flow graph to obtain the unloading decision matrix and objective function value of each auxiliary variable; determine the target auxiliary variable and the unloading decision matrix of the target auxiliary variable when the objective function value is minimized based on the objective function value of each auxiliary variable; obtain the computing power allocation matrix based on the target auxiliary variable and the unloading decision matrix of the target auxiliary variable; and obtain the optimal allocation optimization scheme based on the unloading decision matrix of the target auxiliary variable, the computing power allocation matrix, and the total system overhead.

[0017] According to an embodiment of the present invention, an intelligent agent communication network task offloading and computing power allocation optimization device is provided. First, user equipment parameters and assistant equipment parameters in the task offloading and computing power allocation system are used to calculate the latency and energy consumption of users and assistants. A total cost objective function is constructed, and with the assistant core as a virtual node, the maximum latency term is linearized to obtain a mixed integer programming problem, which is then transformed into a minimum cost flow problem by combining the latency bottleneck set. A network flow graph containing virtual source, target, split user and virtual assistant nodes is constructed, and the optimal solution is obtained. This solves the problem of low efficiency in solving allocation schemes due to high algorithm complexity in related technologies, and improves the efficiency of task offloading and computing power allocation.

[0018] A third aspect of the present invention provides an electronic device, including: 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 the task offloading and computing power allocation optimization method for intelligent agent communication networks described in the above embodiments.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a task offloading and computing power allocation optimization method for intelligent agent communication networks as described in the above embodiments.

[0020] A fifth aspect of the present invention provides a computer program product, including a computer program that is executed to implement a task offloading and computing power allocation optimization method for intelligent agent communication networks as described in the above embodiments.

[0021] Additional aspects and advantages of the invention 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 the invention. Attached Figure Description

[0022] The above and / or additional aspects and advantages of the present invention 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 task offloading and computing power allocation optimization method for intelligent agent communication networks according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a network flow diagram for virtual assistant task unloading and computing power allocation according to a specific embodiment of the present invention; Figure 3 This is a flowchart illustrating a method for optimizing task offloading and computing power allocation in an intelligent agent communication network according to a specific embodiment of the present invention. Figure 4 This is a block diagram of a task offloading and computing power allocation optimization device for intelligent agent communication networks according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0023] Embodiments of the present invention are described in detail below, examples of which are illustrated 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 the present invention, and should not be construed as limiting the present invention.

[0024] The following description, with reference to the accompanying drawings, illustrates an embodiment of the present invention for optimizing task offloading and computing power allocation in intelligent agent communication networks. Addressing the problem of low efficiency in solving allocation schemes due to high algorithm complexity in related technologies mentioned in the background, the present invention provides a method for optimizing task offloading and computing power allocation in intelligent agent communication networks. First, user device parameters and assistant device parameters in the task offloading and computing power allocation system are used to calculate the latency and energy consumption of users and assistants. A total cost objective function is constructed, using the assistant core as a virtual node. The maximum latency term is linearized to obtain a mixed integer programming problem, which is then transformed into a minimum cost flow problem by incorporating the latency bottleneck set. A network flow graph containing virtual source, target, split user, and virtual assistant nodes is constructed, and the optimal solution is obtained. This solves the problem of low efficiency in solving allocation schemes due to high algorithm complexity in related technologies, thereby improving the efficiency of task offloading and computing power allocation.

[0025] Specifically, Figure 1 This is a flowchart illustrating a task offloading and computing power allocation optimization method for intelligent agent communication networks according to an embodiment of the present invention.

[0026] like Figure 1 As shown, this method for optimizing task offloading and computing power allocation in intelligent agent communication networks includes the following steps: In step S101, the user device parameters and assistant device parameters in the task unloading and computing power allocation system are obtained, and the first latency and first energy consumption of the user device, as well as the second latency and second energy consumption of the assistant device are calculated based on the user device parameters and assistant device parameters.

[0027] The first latency of the user equipment refers to the time consumed by each user equipment when processing its own private task locally, and the first energy consumption of the user equipment refers to the energy consumption generated by each user equipment during the process of processing its private task locally. The second latency of the assistant device refers to the time from the transmission of the task to the completion of processing after each user equipment offloads the public task to the assistant device, and the second energy consumption of the assistant device refers to the energy consumption generated during the offloading of each task.

[0028] Further, in some embodiments, obtaining user device parameters and assistant device parameters in the task offloading and computing power allocation system, and calculating the first latency and first energy consumption of the user device, and the second latency and second energy consumption of the assistant device based on the user device parameters and assistant device parameters, includes: obtaining the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, the user task allocation ratio, the public task volume, and the private task volume; based on the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, and a preset binary offloading decision matrix and computing power allocation matrix, offloading some tasks of the user device to the assistant device according to the user task allocation ratio, the public task volume, and the private task volume, to obtain the first task execution strategy of the user device and the second task execution strategy of the assistant device; calculating the first latency and first energy consumption of the user device according to the first task execution strategy, and calculating the second latency and second energy consumption of the assistant device according to the second task execution strategy.

[0029] Among them, the first computing power of the user device refers to the computing and processing capabilities of the user device itself; the second computing power of the assistant device refers to the total computing power of the assistant device through multi-core collaboration; the third energy consumption of the user device refers to the time consumed by each user device when processing its own private tasks locally; and the fourth energy consumption of the assistant device refers to the energy consumption generated during the unloading process of each task.

[0030] As one possible implementation, the task offloading and computing power allocation system of this embodiment includes a steward intelligent agent. individual user devices and Each assistant device has [number] devices. The first antenna represents the CPU computing power of the user equipment (i.e., the primary computing power). The CPU computing power (i.e., secondary computing power) of each assistant device. The energy consumption coefficient of user equipment is and Initialize user task allocation ratio. Public workload and private task volume The energy consumption of each user device (i.e., the third energy consumption) is determined by the user device energy consumption coefficient. It is determined that the energy consumption of each assistant device (i.e., the fourth energy consumption) is determined by the user device energy consumption coefficient. Sure.

[0031] Furthermore, initialize the binary unloading decision matrix. Initialize the computing power allocation matrix ,in Assistant Assigned to user The CPU ratio is used to calculate the latency of each user executing private tasks locally. (i.e., the first latency) and the energy consumption of each user executing private tasks locally (the first energy consumption). Calculate the latency for user device m to be unloaded to assistant device n. (i.e., the second delay) and the total power consumption of the assistant device. (i.e., the second energy consumption).

[0032] For example, suppose the task offloading and computing power allocation system of this embodiment includes one steward agent and There are 1 intelligent agents, among which A device with a private task is called a user device, and the rest... Each assistant device has [a certain feature]. One antenna, computing power is Power consumption coefficient The proportion of total tasks assigned by the administrator to the user Public task volume and one's own private task volume .

[0033] Users can either execute tasks locally or choose to unload the decision matrix via binary. Unload the task to the assistant agent for execution, where Indicates user Uninstall to Assistant Perform the task. Indicates user Choose local uninstallation. Assistant AI Equipped with Each CPU core's total computing power is determined by its frequency parameter. Characterization. This invention introduces a computing power allocation variable. , indicating assistant Assigned to user The CPU ratio was uniformly written into the matrix. During the computing power allocation process, if the user chooses local computing, the latency will be... Energy consumption If the user chooses to uninstall, then To Assistant Total latency Unloading delay includes unloading transmission delay. And computational delay, i.e. Ultimately, the latency of the parallel unloading process is taken. Latency compared to local computing private tasks The maximum value. User equipment. To assistant device Total energy consumption is

[0034] In step S102, a target optimization function for the total system overhead is constructed based on the first delay, the first energy consumption, the second delay, and the second energy consumption. The core of the assistant device is taken as a virtual assistant node, and the maximum delay term of the target optimization function for the total system overhead is linearized to obtain a mixed integer linear programming problem. Based on the pre-constructed set of delay bottlenecks, the mixed integer linear programming problem is transformed into a minimum cost flow problem.

[0035] Among them, the target optimization function of total system overhead refers to the core indicator function used to measure the overall performance of the task unloading and computing power allocation system, the core of the assistant device refers to the CPU core of the assistant device, and the maximum latency term refers to the maximum value between the latency of the user device processing private tasks locally and the total latency of task unloading to the assistant device for processing.

[0036] Specifically, the objective function for the total system overhead is: ; This leads to the construction of the problem that needs to be solved. : ; Calculate system overhead based on the combination of user equipment and assistant equipment. Task unloading decision matrix Expanding into a virtual assistant matrix Each assistant's CPU core is treated as a virtual assistant, forming a set of virtual assistants. Introducing auxiliary variables By linearizing the maximum delay term, a mixed-integer linear programming problem is established, and finally, a set of delay bottlenecks is constructed. For any This transforms the MILP problem into a minimum cost flow problem.

[0037] Furthermore, in some embodiments, the core of the assistant device is used as a virtual assistant node, and the maximum delay term of the objective optimization function of the total system overhead is linearized to obtain a mixed-integer linear programming problem. Based on a pre-constructed set of delay bottlenecks, the mixed-integer linear programming problem is transformed into a minimum cost flow problem, including: generating a virtual assistant matrix based on the task unloading decision matrix, and generating a virtual assistant set based on the virtual assistant matrix and the CPU cores of the assistant device, and using the nodes in the virtual assistant set as virtual assistant nodes; linearizing the maximum delay term of the objective optimization function of the total system overhead based on preset linear auxiliary variables to obtain a linearized objective optimization function of the total system overhead, and obtaining a mixed-integer linear programming problem based on the objective optimization function of the total system overhead; constructing a set of delay bottlenecks, and transforming the mixed-integer linear programming problem into a minimum cost flow problem based on the parameters in the set of delay bottlenecks and the virtual assistant nodes.

[0038] It should be noted that, and For all user-assistant pairs Generally speaking, they are all constants, but when the computing power allocation matrix... Given that, for any user-assistant pair System overhead This would then become a completely known constant. Therefore, each possible computation frequency can be viewed as a virtual assistant device, whose computation frequency is proportional to that of the original assistant and whose position is the same as that of the original assistant device. As one possible implementation, embodiments of the present invention use a task offloading decision matrix. Expand to elements matrix And define the virtual assistant set as Then, auxiliary variables that satisfy the constraints are introduced. :

[0039] Then all terms are loaded as linear terms, transforming the problem into a mixed-integer linear programming problem: ; As the number of user devices and assistant devices increases, the computational load grows exponentially. Therefore, it is necessary to define a latency bottleneck set to constrain and control the algorithm's complexity. The latency bottleneck set is... For any auxiliary variable The problem can be simplified to: ; This simplifies to a binary allocation problem with capacity constraints.

[0040] In step S103, a network flow graph containing virtual source nodes, target nodes, split user nodes, and virtual assistant nodes is constructed, and the minimum cost flow problem is solved using the network flow graph to obtain the optimal allocation optimization scheme for the task unloading and computing power allocation system.

[0041] It should be noted that, Due to the global monomodular property of its constraint matrix, the problem can be modeled as a minimum flow cost problem by constructing a network flow graph.

[0042] Furthermore, in some embodiments, constructing a network flow graph including virtual source nodes, virtual target nodes, user nodes, and virtual assistant nodes includes: based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem, splitting the user nodes into a first user node representing a sending flow and a second user node representing computing power; based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem, generating a first type of edge of the network flow graph according to the virtual source node and the first user node, generating a second type of edge of the network flow graph according to the first user node and the second user node, generating a third type of edge of the network flow graph according to the first user node and the virtual assistant node, and generating a fourth type of edge of the network flow graph according to the virtual assistant node and the virtual target node, to obtain the network flow graph.

[0043] Specifically, the network flow graph contains virtual source nodes. Virtual target node User nodes and virtual assistant nodes Each user node is split into representations of the sending stream. and representing computational ability It also includes four types of edges , , and .

[0044] For example, Figure 2 This is a schematic diagram of the network flow for virtual assistant task unloading and computing power allocation according to a specific embodiment of the present invention, such as... Figure 2 As shown, there is a virtual source node in the network flow graph. A virtual target node A group of user nodes and a group of virtual assistant nodes To represent local computation and avoid loops that may form due to backflow in a directed graph, this embodiment of the invention splits each user node into two nodes, the first... This only represents users who send streams but do not have computing power; the second one... This only represents the user's computing power. The four edges in the network flow graph are: one with capacity of 1 and one with cost of 0. That is, the edge from the virtual source node to the first user node; with a capacity of 1 and a cost of of That is, the edge from the first user node to the second user node, if and only if the following condition is met: Local computation for the user; capacity is 1, overhead is of That is, the edge from the first user node to the virtual assistant node, if and only if the following condition is met: Time user Uninstall to Virtual Assistant ,as well as All edges from the helper node to the destination node.

[0045] It should be noted that when When a quantity is adopted as a known quantity in this round, it needs to be... To ensure that this edge will definitely appear, the entire network from Flow direction The flow rate is This ensures that each user is served by at most one assistant.

[0046] Furthermore, in some embodiments, the preset linear auxiliary variables include multiple auxiliary variables. The optimal allocation optimization scheme of the task unloading and computing power allocation system is obtained by solving the minimum cost flow problem using a network flow graph, including: based on the network simplex method in the network flow graph, solving the minimum cost flow problem according to multiple auxiliary variables to obtain the unloading decision matrix and objective function value of each auxiliary variable; based on the objective function value of each auxiliary variable, determining the target auxiliary variable and the unloading decision matrix of the target auxiliary variable when the objective function value is minimized; obtaining the computing power allocation matrix according to the target auxiliary variable and the unloading decision matrix of the target auxiliary variable; and obtaining the optimal allocation optimization scheme according to the unloading decision matrix of the target auxiliary variable, the computing power allocation matrix, and the total system overhead.

[0047] Specifically, for considering the maximum delay The edges are forced to be mandatory to ensure that bottleneck latency is considered. The network simplex method is used to solve the problem, and each edge is obtained. Corresponding unloading decision matrix and the objective function value, for all Sort the corresponding objective function values ​​and select the unloading decision matrix that minimizes the objective function. and auxiliary variables ,according to Map back to the original user-assistant combination to identify each user. Uninstall Assistant and computing power allocation ratio Output unloading decision Computing power allocation matrix By considering the total system overhead, the optimal allocation and optimization scheme can be obtained.

[0048] For example, for each possible The solution is obtained using the network simplex method, with a time complexity of O(n log n). In the set of latency bottlenecks Each of them After iteration, a set of unloading decision matrices can be obtained. and its related objective function values For each Sort the results and find the ones that minimize the objective function value. Yes, we obtain the optimal unloading decision matrix. Finally, the optimal allocation scheme is obtained.

[0049] To enable those skilled in the art to better understand the task offloading and computing power allocation optimization method for intelligent agent communication networks according to embodiments of the present invention, the following will explain and illustrate it in conjunction with specific embodiments.

[0050] Figure 3 This is a flowchart illustrating a method for optimizing task offloading and computing power allocation in an intelligent agent communication network according to a specific embodiment of the present invention. Figure 3 As shown, the process first initializes the user's public task volume, private task volume, offloading decision matrix, and computing power allocation matrix. Then, it calculates the latency, energy consumption, and system overhead of local execution and offloading to the assistant. Next, it expands the virtual assistant matrix to form a virtual assistant set, introduces an auxiliary variable τ to linearize the maximum latency term, and constructs a latency bottleneck set, establishing a mixed-integer linear programming problem. Afterward, it constructs a network flow graph containing source nodes, target nodes, user nodes (split into sending flow and computing power nodes), virtual assistant nodes, and four types of edges, forcibly defining bottleneck latency edges, and using the network simplex method to solve the minimum cost flow problem. Then, it sorts the objective function values ​​corresponding to all τ, selects the offloading matrix with the smallest objective function value, and τ*. Subsequently, it maps back to the original user-assistant combination to determine the offloading assistant and computing power allocation. Finally, it outputs the optimal offloading decision matrix, computing power allocation matrix, and total system overhead.

[0051] According to an embodiment of the present invention, a method for optimizing task offloading and computing power allocation in intelligent agent communication networks is provided. First, user device parameters and assistant device parameters in the task offloading and computing power allocation system are used to calculate the latency and energy consumption of users and assistants. A total cost objective function is constructed, and with the assistant core as a virtual node, the maximum latency term is linearized to obtain a mixed integer programming problem, which is then transformed into a minimum cost flow problem by combining the latency bottleneck set. A network flow graph containing virtual source, target, split user, and virtual assistant nodes is constructed, and the optimal solution is obtained. This method solves the problem of low efficiency in solving allocation schemes due to high algorithm complexity in related technologies, thereby improving the efficiency of task offloading and computing power allocation.

[0052] Next, referring to the accompanying drawings, an embodiment of the present invention provides a task offloading and computing power allocation optimization device for intelligent agent communication networks.

[0053] Figure 4 This is a block diagram of a task offloading and computing power allocation optimization device for intelligent agent communication networks according to an embodiment of the present invention.

[0054] like Figure 4As shown, the task offloading and computing power allocation optimization device for intelligent agent communication networks includes: an acquisition module 100, a mathematical analysis module 200, and a solution module 300.

[0055] The acquisition module 100 is used to acquire user device parameters and assistant device parameters in the task unloading and computing power allocation system, and calculate the first latency and first energy consumption of the user device, and the second latency and second energy consumption of the assistant device based on the user device parameters and assistant device parameters; the mathematical analysis module 200 is used to construct the objective optimization function of the total system cost based on the first latency, first energy consumption, second latency, and second energy consumption, take the core of the assistant device as a virtual assistant node, and linearize the maximum latency term of the objective optimization function of the total system cost to obtain a mixed integer linear programming problem. Based on the pre-constructed latency bottleneck set, the mixed integer linear programming problem is transformed into a minimum cost flow problem; the solution module 300 is used to construct a network flow graph containing virtual source nodes, target nodes, split user nodes, and virtual assistant nodes, and use the network flow graph to solve the minimum cost flow problem to obtain the optimal allocation optimization scheme of the task unloading and computing power allocation system.

[0056] Further, in some embodiments, the acquisition module 100 is specifically used to: acquire the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, the user task allocation ratio, the public task volume, and the private task volume; based on the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, and a preset binary offloading decision matrix and computing power allocation matrix, offload some tasks of the user device to the assistant device according to the user task allocation ratio, the public task volume, and the private task volume, to obtain the first execution task strategy of the user device and the second execution task strategy of the assistant device; calculate the first latency and first energy consumption of the user device according to the first execution task strategy, and calculate the second latency and second energy consumption of the assistant device according to the second execution task strategy.

[0057] Furthermore, in some embodiments, the mathematical analysis module 200 is specifically used for: generating a virtual assistant matrix based on the task unloading decision matrix, and generating a virtual assistant set based on the virtual assistant matrix and the CPU cores of the assistant device, and using the nodes in the virtual assistant set as virtual assistant nodes; obtaining the linearized system total overhead objective optimization function by linearizing the maximum delay term of the objective optimization function based on preset linear auxiliary variables, and obtaining a mixed-integer linear programming problem based on the objective optimization function of the system total overhead; constructing a delay bottleneck set, and transforming the mixed-integer linear programming problem into a minimum cost flow problem based on the parameters in the delay bottleneck set and the virtual assistant nodes.

[0058] Furthermore, in some embodiments, the solution module 300 is specifically used for: splitting the user node into a first user node representing the sending flow and a second user node representing the computing capacity based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem; generating a first type of edge of the network flow graph based on the virtual source node and the first user node, generating a second type of edge of the network flow graph based on the first user node and the second user node, generating a third type of edge of the network flow graph based on the first user node and the virtual assistant node, and generating a fourth type of edge of the network flow graph based on the virtual assistant node and the virtual target node, so as to obtain the network flow graph.

[0059] Furthermore, in some embodiments, the preset linear auxiliary variables include multiple auxiliary variables. The solution module 300 is also used to: solve the minimum cost flow problem based on the network simplex method in the network flow graph to obtain the unloading decision matrix and objective function value of each auxiliary variable; determine the target auxiliary variable and the unloading decision matrix of the target auxiliary variable when the objective function value is minimized based on the objective function value of each auxiliary variable; obtain the computing power allocation matrix based on the target auxiliary variable and the unloading decision matrix of the target auxiliary variable; and obtain the optimal allocation optimization scheme based on the unloading decision matrix of the target auxiliary variable, the computing power allocation matrix, and the total system overhead.

[0060] It should be noted that the above explanation of an embodiment of a task offloading and computing power allocation optimization method for intelligent agent communication networks also applies to a task offloading and computing power allocation optimization device for intelligent agent communication networks in the same embodiment, and will not be repeated here.

[0061] According to an embodiment of the present invention, a task offloading and computing power allocation optimization device for intelligent agent communication networks is provided. First, user equipment parameters and assistant equipment parameters in the task offloading and computing power allocation system are used to calculate the latency and energy consumption of users and assistants. A total cost objective function is constructed, and with the assistant core as a virtual node, the maximum latency term is linearized to obtain a mixed integer programming problem, which is then transformed into a minimum cost flow problem by combining the latency bottleneck set. A network flow graph containing virtual source, target, split user and virtual assistant nodes is constructed, and the optimal solution is obtained. This solves the problem of low efficiency in solving allocation schemes due to high algorithm complexity in related technologies, and improves the efficiency of task offloading and computing power allocation.

[0062] Figure 5 This is a schematic diagram of an electronic device provided according to an embodiment of the present invention. 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.

[0063] When the processor 502 executes the program, it implements the task offloading and computing power allocation optimization method for intelligent agent communication networks provided in the above embodiments.

[0064] Furthermore, the electronic device also includes: Communication interface 503 is used for communication between memory 501 and processor 502.

[0065] The memory 501 is used to store computer programs that can run on the processor 502.

[0066] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0067] 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 Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, 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.

[0068] 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.

[0069] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0070] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for optimizing task offloading and computing power allocation for intelligent agent communication networks.

[0071] In addition, embodiments of the present invention also provide a computer program product, including a computer program, which is executed to implement the above-described method for optimizing task offloading and computing power allocation for intelligent agent communication networks.

[0072] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. 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.

[0073] 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 indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0074] Any process or method description 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 preferred embodiments of the invention 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 the invention pertain.

[0075] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented in 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 of the following techniques known in the art, or a combination thereof: 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 (PGAs), field-programmable gate arrays (FPGAs), etc.

[0076] 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.

Claims

1. A method for optimizing task offloading and computing power allocation in intelligent agent communication networks, characterized in that, Includes the following steps: Obtain user device parameters and assistant device parameters in the task unloading and computing power allocation system, and calculate the first latency and first energy consumption of the user device, and the second latency and second energy consumption of the assistant device based on the user device parameters and the assistant device parameters; Based on the first delay, the first energy consumption, the second delay, and the second energy consumption, a target optimization function for the total system overhead is constructed. The core of the assistant device is taken as a virtual assistant node, and the maximum delay term of the target optimization function for the total system overhead is linearized to obtain a mixed integer linear programming problem. Based on the pre-constructed set of delay bottlenecks, the mixed integer linear programming problem is transformed into a minimum cost flow problem. A network flow graph containing virtual source nodes, target nodes, split user nodes, and virtual assistant nodes is constructed, and the minimum cost flow problem is solved using the network flow graph to obtain the optimal allocation optimization scheme for the task unloading and computing power allocation system.

2. The method according to claim 1, characterized in that, The process of obtaining user device parameters and assistant device parameters in the task unloading and computing power allocation system, and calculating the first latency and first energy consumption of the user device, and the second latency and second energy consumption of the assistant device based on the user device parameters and the assistant device parameters, includes: Acquire the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, the user task allocation ratio, the amount of public tasks and the amount of private tasks; Based on the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, and the preset binary offloading decision matrix and computing power allocation matrix, some tasks of the user device are offloaded to the assistant device according to the user task allocation ratio, the amount of public tasks and the amount of private tasks, so as to obtain the first task execution strategy of the user device and the second task execution strategy of the assistant device. The user device's first latency and first energy consumption are calculated according to the first task execution strategy, and the assistant device's second latency and second energy consumption are calculated according to the second task execution strategy.

3. The method according to claim 1, characterized in that, The process involves using the core of the assistant device as a virtual assistant node and linearizing the maximum delay term of the objective function for the total system overhead to obtain a mixed-integer linear programming problem. Based on a pre-built set of delay bottlenecks, this mixed-integer linear programming problem is transformed into a minimum-cost flow problem, including: A virtual assistant matrix is ​​generated based on the task unloading decision matrix, and a virtual assistant set is generated based on the virtual assistant matrix and the CPU cores of the assistant device, with the nodes in the virtual assistant set being used as the virtual assistant nodes. Based on preset linear auxiliary variables, the maximum time delay term of the objective optimization function of the total system cost is linearized to obtain the objective optimization function of the total system cost, and the mixed integer linear programming problem is obtained according to the objective optimization function of the total system cost. Construct a set of latency bottlenecks, and based on the parameters in the set of latency bottlenecks and the virtual assistant node, transform the mixed integer linear programming problem into the minimum cost flow problem.

4. The method according to claim 1, characterized in that, The construction of the network flow graph, which includes virtual source nodes, virtual target nodes, user nodes, and virtual assistant nodes, includes: Based on the global monomodular property of the constraint matrix in the capacity-constrained binary allocation problem, the user node is split into a first user node representing the transmission flow and a second user node representing the computing power. Based on the global monomodular property of the constraint matrix of the capacity-constrained binary allocation problem, the network flow graph is generated by generating the first type of edges based on the virtual source node and the first user node, the second type of edges based on the first user node and the second user node, the third type of edges based on the first user node and the virtual assistant node, and the fourth type of edges based on the virtual assistant node and the virtual target node, so as to obtain the network flow graph.

5. The method according to claim 1, characterized in that, The preset linear auxiliary variables include multiple auxiliary variables. The step of using the network flow graph to solve the minimum cost flow problem to obtain the optimal allocation optimization scheme for the task unloading and computing power allocation system includes: Based on the network simplex method in the network flow graph, the minimum cost flow problem is solved according to the multiple auxiliary variables to obtain the unloading decision matrix and objective function value of each auxiliary variable; Based on the objective function value of each auxiliary variable, determine the target auxiliary variable when the objective function value is minimized and the unloading decision matrix of the target auxiliary variable; The computing power allocation matrix is ​​obtained based on the target auxiliary variable and the unloading decision matrix of the target auxiliary variable, and the optimal allocation optimization scheme is obtained based on the unloading decision matrix of the target auxiliary variable, the computing power allocation matrix, and the total system overhead.

6. A task offloading and computing power allocation optimization device for intelligent agent communication networks, characterized in that, The device includes: The acquisition module is used to acquire user device parameters and assistant device parameters in the task unloading and computing power allocation system, and calculate the first latency and first energy consumption of the user device, and the second latency and second energy consumption of the assistant device based on the user device parameters and the assistant device parameters. The mathematical analysis module is used to construct a target optimization function for the total system cost based on the first delay, the first energy consumption, the second delay, and the second energy consumption. The core of the assistant device is used as a virtual assistant node, and the maximum delay term of the target optimization function for the total system cost is linearized to obtain a mixed integer linear programming problem. Based on a pre-constructed set of delay bottlenecks, the mixed integer linear programming problem is transformed into a minimum cost flow problem. The solution module is used to construct a network flow graph containing virtual source nodes, target nodes, split user nodes, and virtual assistant nodes, and to use the network flow graph to solve the minimum cost flow problem to obtain the optimal allocation optimization scheme of the task unloading and computing power allocation system.

7. The apparatus according to claim 6, characterized in that, The acquisition module is specifically used to: acquire the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, the user task allocation ratio, the amount of public tasks and the amount of private tasks; Based on the first computing power and third energy consumption of the user device, the second computing power and fourth energy consumption of the assistant device, and the preset binary offloading decision matrix and computing power allocation matrix, some tasks of the user device are offloaded to the assistant device according to the user task allocation ratio, the amount of public tasks and the amount of private tasks, so as to obtain the first task execution strategy of the user device and the second task execution strategy of the assistant device. The user device's first latency and first energy consumption are calculated according to the first task execution strategy, and the assistant device's second latency and second energy consumption are calculated according to the second task execution strategy.

8. 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 task offloading and computing power allocation optimization method for intelligent agent communication networks as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement a task offloading and computing power allocation optimization method for intelligent agent communication networks as described in any one of claims 1-5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements a task offloading and computing power allocation optimization method for intelligent agent communication networks as described in any one of claims 1-5.

Images