Calculation network intelligent agent system based on distributed collaboration and resource dynamic scheduling method thereof

Abstract

The invention provides a distributed collaboration-based computing network intelligent agent system and a resource dynamic scheduling method thereof, and belongs to the technical field of computing network integration. The system comprises an edge agent used for sensing local computing power, network bandwidth and task load in real time, predicting task demand fluctuation by using a lightweight neural network, adjusting resource allocation weight in real time in combination with network topology change, and executing a preliminary task scheduling decision; the regional collaborative agent is used for aggregating multiple edge node states based on federated learning, generating a cross-node resource scheduling strategy, verifying the credibility of a computing power transaction smart contract and determining a cross-domain resource allocation scheme; and the cloud management agent is used for constructing a global resource portrait model according to the information provided by the edge agent and the regional collaborative agent, performing long-term strategy optimization, issuing global strategy information, and constructing and updating a computing power transaction smart contract based on a preset computing power transaction smart contract template. According to the invention, multi-level refined scheduling of computing network resources is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of algorithm and network integration, and more particularly to an algorithm and network intelligent agent system based on distributed collaboration and a resource dynamic scheduling method thereof. BACKGROUND

[0002] In today's rapid development of information technology, efficient management and scheduling of computing power and network resources have become key problems to be solved. The traditional cloud computing mode adopts a centralized scheduling mechanism, which leads to the concentration of computing power in large data centers. This not only results in high response delay on the edge side, but also makes it difficult to meet the needs of industrial Internet of Things, autonomous driving and other application scenarios that require real-time performance, thereby seriously restricting the development and application of related technologies.

[0003] At the same time, although edge computing nodes can share the computing power pressure to some extent, their own resources are limited. In a complex scenario with multiple concurrent tasks, resource allocation conflicts may occur in edge computing nodes. The current technology lacks effective global optimization capabilities and cannot achieve reasonable allocation and efficient use of resources, resulting in resource waste and performance bottlenecks.

[0004] In addition, in the current algorithm and network resource management, the three key elements of network, computing power and data are often managed separately, and there is insufficient collaboration between them, resulting in low cross-domain collaboration efficiency. In a multi-operator environment or distributed computing scenario, resource sharing and task scheduling between different domains face many challenges, such as the lack of trust mechanisms, the complexity of resource trading processes, etc. These problems make it difficult to form an efficient and unified algorithm and network resource scheduling system, greatly limiting the performance of the overall system and the improvement of resource utilization.

[0005] Existing research and practice also have some limitations. For example, some solutions propose an algorithm resource scheduling method based on software-defined networks (SDN), but do not introduce an artificial intelligence (AI) dynamic decision-making mechanism, making it difficult to adapt to complex and changing real-world application scenarios, especially in terms of accurate prediction and real-time adjustment of resource needs. Some methods attempt to use reinforcement learning to optimize task allocation, but are often limited to single-layer edge nodes and do not fully consider the collaborative cooperation between nodes in a multi-layer distributed architecture, making it difficult to achieve efficient resource scheduling in a large-scale algorithm and network environment and fundamentally solving the complex problems faced by existing algorithm and network resource scheduling. SUMMARY

[0006] In view of the above problems, the purpose of the present application is to provide a distributed collaborative based on the algorithm network intelligent agent system and its resource dynamic scheduling method, through edge, regional, cloud agent hierarchical collaborative scheduling, using lightweight neural network, improved auction algorithm, double deep Q network algorithm and other intelligent algorithm optimization resource matching, constructing global resource portrait model for long-term planning, based on the block chain technology to build a trusted computing power transaction smart contract, using dynamic bidding mode to realize the dynamic balance of price and supply and demand and cross-domain resource automatic allocation charging, so as to realize the multi-level fine scheduling of algorithm network resources, efficient utilization, scientific configuration, trusted transaction and efficient circulation.

[0007] In order to achieve the above purpose, the following technical solutions are adopted: In a first aspect, the present application provides a distributed collaborative based on the algorithm network intelligent agent system, comprising: The edge intelligent agent is deployed on the edge node, is used for real-time sensing of local computing power, network bandwidth and task load, and uses lightweight neural network to predict task demand fluctuation, combines network topology change to adjust resource allocation weight in real time, and executes preliminary task scheduling decision; The regional collaborative intelligent agent is in communication connection with a plurality of edge intelligent agents, is used for aggregating multi-edge node state based on federated learning, generating cross-node resource scheduling strategy, and verifying the credibility of the computing power transaction smart contract and determining the cross-domain resource allocation scheme; The cloud management intelligent agent is in communication connection with the regional collaborative intelligent agent, is used for constructing global resource portrait model according to the information provided by the edge intelligent agent and the regional collaborative intelligent agent, performing long-term strategy optimization, publishing global strategy information, and constructing and updating the computing power transaction smart contract based on the preset computing power transaction smart contract template.

[0008] In an optional embodiment, in the global resource portrait model, the resource service quality is comprehensively evaluated by the following formula: QoE(t) = w1(t)•C + w2(t)•B - w3(t)•E; Wherein, QoE(t) is the service quality experience of the user at t time, C is the normalized computing power, B is the available bandwidth, E is the energy consumption, w1(t), w2(t), w3(t) are respectively the weight factor dynamically adjusted according to the historical load at t time, w1(t)+w2(t)+w3(t)=1.

[0009] In an optional embodiment, when the edge intelligent agent executes the preliminary task scheduling decision, if the local node resource cannot meet the task demand, the improved auction algorithm is used to perform task decomposition and distribution to other edge nodes for collaborative processing according to the cross-node resource scheduling strategy issued by the regional collaborative intelligent agent; The improved auction algorithm introduces a load balancing factor, and a calculation formula of the load balancing factor is:

[0010] wherein, is an average load in a region, is a current load of a target edge node.

[0011] In an optional embodiment, the regional coordination agent generates a cross-node resource scheduling strategy by using a double deep Q network algorithm. A reward function of the double deep Q network algorithm is: R=a×RU-b×DT; wherein, RU is resource utilization, DT is task response time delay, and a and b are weight coefficients.

[0012] In an optional embodiment, the computing power transaction smart contract is constructed based on a blockchain technology, and the computing power transaction smart contract includes computing power specifications of a computing power provider, a pricing model, a service availability commitment, and task demand description, budget limitation and default processing terms of a computing power user.

[0013] In an optional embodiment, the pricing model adopts a dynamic bidding mode, and a calculation formula of the dynamic bidding mode is:

[0014] wherein, P(t) is a computing power price at time t, P0 is an initial price, k is a dynamic adjustment coefficient, and D(t) is a difference value between computing power demand and supply at time t. When D(t)>0, it indicates that the computing power demand is greater than the supply, and the price rises. When D(t)<0, it indicates that the computing power supply is greater than the demand, and the price falls.

[0015] In a second aspect, the embodiments of the present application further provide a resource dynamic scheduling method of a distributed coordination-based computing network intelligent agent system, and the method comprises: The local computing power, network bandwidth and task load information are monitored and reported by the edge agent in real time, the task demand fluctuation is predicted by using a lightweight neural network, the resource allocation weight adjustment demand is preliminarily judged in combination with network topology changes, and the information is reported to the regional coordination agent; The information reported by the edge agent is received by the regional coordination agent, the states of multiple edge nodes are aggregated based on federated learning, a comprehensive state model is generated, a cross-node resource scheduling strategy is formulated by using an improved double deep Q network algorithm, and the strategy is distributed to the edge agent; The task scheduling is executed by the edge agent according to the cross-node resource scheduling strategy, the task is disassembled and distributed to other edge nodes for cooperative processing if the local resource is insufficient, and the feedback information is generated by monitoring the task execution state and resource usage. The cloud management agent collects the resource state, task scheduling and task execution result of the edge agent, constructs a global resource portrait model, optimizes the long-term planning and allocation strategy of the algorithm network resource by using a long-term strategy optimization algorithm, and issues the optimized global strategy information; The cloud management agent receives the idle computing power information published by the operator in real time, the regional collaborative agent verifies the credibility of the computing power transaction according to the smart contract, determines the cross-domain resource allocation scheme according to the computing task demand, completes the computing power transaction and automatically charges by using the block chain ledger.

[0016] In an optional implementation, the prediction of task demand fluctuation by using a lightweight neural network comprises: A time series prediction model is constructed based on historical task data, short-term features are extracted by a convolutional neural network, long-term dependencies are captured by using a long short-term memory network, and a task demand fluctuation prediction value and a confidence interval in a future preset time period are output.

[0017] In a third aspect, the embodiments of the present application also provide an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the resource dynamic scheduling method of the distributed collaborative algorithm network agent system according to any one of the above when executing the program.

[0018] In a fourth aspect, the embodiments of the present application also provide a storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the steps of the resource dynamic scheduling method of the distributed collaborative algorithm network agent system according to any one of the above.

[0019] As can be seen from the above technical solutions, the present application has the following advantages: The present application realizes multi-level and intelligent algorithm network resource management. Among them, the edge agent realizes real-time sensing of local resources and preliminary scheduling, the regional collaborative agent generates a cross-node scheduling strategy based on federated learning aggregation of multi-node states, and the cloud management agent realizes global resource portrait and long-term strategy optimization; the system adopts various innovative algorithms, such as lightweight neural network prediction of task demand, improved auction algorithm and double deep Q network algorithm optimization of resource scheduling, dynamic bidding mode realizes reasonable pricing, and blockchain technology guarantees the credibility of computing power transaction; such distributed collaborative architecture and intelligent scheduling mechanism can accurately perceive resource changes, efficiently optimize resource allocation, improve resource utilization and service quality, reduce energy consumption, and at the same time ensure the fairness and credibility of computing power transaction, effectively meet diversified task demand, realize dynamic and efficient configuration of algorithm network resources and sustainable development.

[0020] The application adopts layered cooperation of edges, regions and cloud agents, the edge agent perceives local resources in real time and preliminarily schedules, the regional collaborative agent aggregates multi-node states to generate cross-node strategies, and the cloud management agent performs global optimization, so as to realize multi-level and refined dynamic scheduling of algorithm network resources and improve resource utilization efficiency.

[0021] The application uses a lightweight neural network to predict task demand fluctuations, improves an auction algorithm combined with a load balancing factor to realize reasonable disassembly and distribution of tasks, a double deep Q network algorithm optimizes cross-node resource scheduling strategies, accurately matches resources and tasks through intelligent algorithms, reduces task response time delay, and improves the overall performance of the system.

[0022] The application constructs a global resource portrait model through a cloud management agent, comprehensively evaluates resource service quality, uses a long-term strategy optimization algorithm to make long-term planning and distribution strategy adjustment of algorithm network resources, ensures the scientificity and sustainability of resource configuration, and adapts to long-term development needs of business.

[0023] The application constructs a computing power transaction smart contract based on a blockchain technology, clearly defines rights and interests and responsibilities of both sides of computing power supply and demand, covers terms such as computing power specifications, pricing, service commitment and default handling, verifies transaction credibility through a smart contract, guarantees fair and just computing power transactions, and promotes healthy development of the computing power market.

[0024] The application uses a dynamic bidding mode pricing model, adjusts prices in real time according to the difference between computing power supply and demand, realizes dynamic balance between prices and market supply and demand, and at the same time, the cloud management agent cooperates with the regional collaborative agent to determine a cross-domain resource allocation scheme, completes computing power transactions and automatically charges, and improves resource circulation efficiency and economic benefits. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the present application, the drawings needed to be used in the description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0026] Figure 1 The structure schematic diagram of the algorithm network agent system based on distributed cooperation provided by the present application is shown.

[0027] Figure 2 The flowchart of the resource dynamic scheduling method of the algorithm network agent system based on distributed cooperation provided by the present application is shown.

[0028] Figure 3 The structure schematic diagram of the electronic device provided by the present application is shown. DETAILED DESCRIPTION

[0029] Various embodiments of the disclosure will be described in detail below with reference to the accompanying drawings. The various embodiments of the disclosure can have various modifications, and all possible modifications should be construed as falling within the scope of the various embodiments of the disclosure. It should be understood that there is no intent to limit the various embodiments of the disclosure to particular embodiments disclosed herein, but on the contrary, the disclosure is to cover all modifications, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of the disclosure.

[0030] Hereinafter, the term "include" or "may include" used in the various embodiments of the disclosure indicates the presence of the disclosed functions, operations, or elements and does not limit the addition of one or more functions, operations, or elements. Also, as used in the various embodiments of the disclosure, the terms "include", "have", and their conjugates merely indicate the presence of specific features, numbers, steps, operations, elements, components, or combinations thereof, and should not be construed as excluding the presence or addition of one or more other features, numbers, steps, operations, elements, components, or combinations thereof.

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0032] Please refer to Figure 1 The structure of a distributed collaborative computing network agent system in an embodiment is shown in the figure. The system includes edge agents, regional collaborative agents, and cloud management agents. The edge agents, regional collaborative agents, and cloud management agents exchange information through a secure communication protocol to ensure the confidentiality, integrity, and availability of data transmission.

[0033] The edge agent is deployed on the edge node to perceive the local computing power, network bandwidth, and task load in real time, predict task demand fluctuations using a lightweight neural network, adjust resource allocation weights in real time in combination with network topology changes, and execute preliminary task scheduling decisions.

[0034] When the edge agent executes preliminary task scheduling decisions, if the local node resources cannot meet the task demand, the edge agent uses an improved auction algorithm to disassemble and distribute the task to other edge nodes for collaborative processing according to the cross-node resource scheduling strategy issued by the regional collaborative agent. The improved auction algorithm introduces a load balancing factor, and the calculation formula of the load balancing factor is:

[0035] wherein, is the average load in the region, is the current load of the target edge node. The value range of F is [0, 1], and the larger F is, the more the target edge node meets the load balancing requirement.

[0036] The regional collaborative agent is in communication connection with the plurality of edge agents, and is configured to aggregate multi-edge node states based on federated learning, generate a cross-node resource scheduling strategy, and verify the credibility of the computing power transaction smart contract and determine a cross-domain resource allocation scheme.

[0037] The regional collaborative agent generates a cross-node resource scheduling strategy by using a double deep Q network algorithm. The reward function of the double deep Q network algorithm is: R = a x RU-b x DT; wherein, RU is resource utilization, DT is task response time delay, a and b are weight coefficients, and the weight coefficients are dynamically adjusted according to actual application scenarios and performance requirements.

[0038] The cloud management agent is in communication connection with the regional collaborative agent, and is configured to construct a “computing power-network-energy consumption” three-dimensional global resource portrait model according to information provided by the edge agent and the regional collaborative agent, perform long-term strategy optimization, support computing power network elastic expansion and contraction, publish global strategy information, and construct and update a computing power transaction smart contract based on a preset computing power transaction smart contract template.

[0039] In the global resource portrait model, the resource service quality is comprehensively evaluated by the following formula: QoE(t) = w1(t)•C + w2(t)•B - w3(t)•E; wherein, QoE(t) is the service quality experience of a user at time t, C is standardized computing power, B is available bandwidth, E is energy consumption, w1(t), w2(t), and w3(t) are weight factors dynamically adjusted according to historical load at time t, and w1(t)+w2(t)+w3(t)=1.

[0040] When performing long-term strategy optimization, the cloud management agent comprehensively considers the energy consumption cost of the computing network system, the supply and demand balance of computing power resources, and the evolution trend of the network topology, and adopts a multi-objective optimization algorithm to make the system achieve performance optimization and cost benefit balance in the long-term running process.

[0041] In addition, the computing power transaction smart contract adopted is constructed based on a blockchain technology, and contains computing power specifications of a computing power provider, a pricing model, a service availability commitment, and task demand description, budget limitation, and default processing terms of a computing power user, etc. Through an automatic execution mechanism of the smart contract, automation, credibility and high efficiency of the computing power transaction are realized.

[0042] The pricing model adopts a dynamic bidding mode, and a calculation formula thereof is as follows:

[0043] wherein P(t) is a computing power price at time t, P0 is an initial price, k is a dynamic adjustment coefficient, and D(t) is a difference between computing power demand and supply at time t. When D(t) > 0, it indicates that the computing power demand is greater than the supply, and the price rises. When D(t) < 0, it indicates that the computing power supply is greater than the demand, and the price falls.

[0044] In the embodiment, through the layered collaborative work of the edge, regional and cloud intelligent agents, in combination with the technologies such as the lightweight neural network, the improved auction algorithm and the double deep Q network algorithm, real-time monitoring, accurate prediction and efficient scheduling of computing power and network resources are realized. At the same time, a trusted computing power transaction smart contract is constructed by using the blockchain technology, and flexible pricing is realized in a dynamic bidding mode. This not only effectively improves the resource utilization rate and reduces the task response time delay, but also guarantees the fairness and credibility of the computing power transaction, and finally achieves the optimized allocation and efficient circulation of computing network resources, thereby providing a better service experience for users.

[0045] As shown in FIG. 1, Figure 2 The following is an embodiment of a resource dynamic scheduling method of a distributed collaborative computing network intelligent agent system provided by the embodiment of the present disclosure, which belongs to the same inventive concept as the distributed collaborative computing network intelligent agent system described above. Details not described in the embodiment of the resource dynamic scheduling method of the distributed collaborative computing network intelligent agent system can be referred to the embodiment of the distributed collaborative computing network intelligent agent system.

[0046] A resource dynamic scheduling method of a distributed collaborative computing network intelligent agent system, comprising the following steps: S1: Real-time monitoring and reporting of local computing power, network bandwidth and task load information by the edge intelligent agent, prediction of task demand fluctuation by the lightweight neural network, preliminary judgment of resource allocation weight adjustment demand in combination with network topology change, and reporting to the regional collaborative intelligent agent.

[0047] In the specific embodiment, the edge agent perceives the computing power available amount, network bandwidth usage, and current task load information of the local node in real time according to the preset monitoring frequency on the respective deployed edge node, and uses a lightweight neural network to predict task demand fluctuations, combines network topology changes, preliminarily judges the potential adjustment demand of the resource allocation weight, and regularly reports the above information to the regional collaborative agent.

[0048] For example, in this step, the specific method of the edge agent using a lightweight neural network to predict task demand fluctuations is as follows: A time series prediction model is constructed based on historical task data, the input features include historical time task request quantity, task type, task required computing power scale, and network bandwidth occupation situation and other information, short-term time series features are extracted through a convolutional neural network (CNN), and long-term time dependence is captured through a long short-term memory network (LSTM), and finally the task demand fluctuation prediction value in the future preset time period is output, and the confidence interval of the prediction value is estimated through a Bayesian statistical method to measure the credibility of the prediction, and provide an accurate reference for resource allocation weight adjustment.

[0049] S2: The regional collaborative agent receives the information reported by the edge agent, aggregates the multi-edge node states based on federated learning, generates a comprehensive state model, uses an improved double deep Q network algorithm to formulate a cross-node resource scheduling strategy and issues it to the edge agent.

[0050] In the specific embodiment, after the regional collaborative agent receives the information reported by each edge agent, it aggregates the state data of the multiple edge nodes based on federated learning technology, generates a comprehensive state model that integrates the computing power, network status, and task status of multiple edge nodes, uses an improved double deep Q network (DDQN) algorithm, takes resource utilization rate as the maximization target, and considers factors such as delay constraints to formulate a cross-edge node resource scheduling strategy, and issues the strategy to the corresponding edge agent.

[0051] S3: The edge agent executes task scheduling according to the cross-node resource scheduling strategy, and if the local resources are insufficient, the task is disassembled and distributed to other edge nodes for collaborative processing, while monitoring the task execution state and resource usage and generating feedback information. In the specific embodiment, after the edge agent receives the resource scheduling strategy issued by the regional collaborative agent, it schedules and allocates local tasks according to the strategy, and if the local node resources can meet the task demand, the task is executed locally; if the local resources cannot meet the demand, the task is disassembled and distributed to other edge nodes for collaborative processing according to the scheduling strategy, and the task execution state and resource usage are continuously monitored during the task execution process to ensure the smooth progress of the task, and the task execution result is fed back to the regional collaborative agent.

[0052] S4: The cloud management agent collects the resource state, task scheduling situation and task execution result of the edge agent, constructs a global resource portrait model, optimizes and adjusts the long-term planning and distribution strategy of the computing network resource by using a long-term strategy optimization algorithm, and issues the optimized global strategy information.

[0053] In the specific embodiment, the cloud management agent collects various information from the regional collaborative agent, including the resource state, task scheduling situation and execution result of each edge node, constructs a "computing power-network-energy consumption" three-dimensional global resource portrait model based on these data, optimizes and adjusts the long-term planning and distribution strategy of the computing network resource by using a long-term strategy optimization algorithm, and periodically issues the optimized global strategy information to the regional collaborative agent and the edge agent to guide the resource management and scheduling of the entire computing network system in the long-term running process.

[0054] S5: The cloud management agent receives the idle computing power information issued by the operator in real time, controls the regional collaborative agent to verify the credibility of the computing power transaction according to the smart contract, determines the cross-domain resource allocation scheme according to the computing task demand, completes the computing power transaction and automatically charges by using the blockchain ledger.

[0055] In the specific embodiment, in a multi-operator environment, after the cloud management agent receives the idle computing power information issued by the operator, it first performs preliminary screening and integration on the computing power information, the regional collaborative agent verifies the credibility of the computing power transaction according to the smart contract, and after verification, determines the cross-domain resource allocation scheme according to the demand of the emergency computing task and the characteristics of the computing power of each operator, completes the computing power transaction, and automatically completes the charging process by using the blockchain ledger in the settlement stage.

[0056] The resource dynamic scheduling method of the computing network intelligent agent system based on distributed collaboration provided in the embodiment, through the real-time monitoring and reporting of resource information by the edge agent and the prediction of task demand, the regional collaborative agent uses the federated learning to aggregate the state, uses the improved double deep Q network algorithm to generate the scheduling strategy, the edge agent executes the scheduling and feedbacks accordingly, the cloud management agent constructs the global resource portrait model and optimizes the long-term strategy, and simultaneously relies on the blockchain smart contract to realize the credible computing power transaction and automatic charging; by using various algorithms such as lightweight neural network, the whole-process intelligent management of the computing network resource from local perception, collaborative scheduling to global optimization is realized, the resource utilization efficiency is effectively improved, the time delay is reduced, the transaction credibility is guaranteed, the dynamic and efficient resource configuration and sustainable development are achieved.

[0057] Figure 3 A hardware structure schematic diagram of an electronic device for implementing various embodiments of the present application.

[0058] The resource dynamic scheduling method of the algorithm network intelligent agent system based on distributed cooperation provided by the embodiments of the application can be applied to an electronic device. Those skilled in the art can understand that the structure of the electronic device involved in the embodiments of the application does not constitute a limitation on the electronic device, and the electronic device can include more or fewer components than the illustration, or combine certain components, or different component arrangements. In the embodiments of the application, the electronic device includes but is not limited to a laptop computer, a desktop computer, a workstation, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples, and are not intended to limit the implementation of the embodiments of the application described herein and / or claimed.

[0059] The electronic device can include a processor, an external memory interface, an internal memory, a universal serial bus (USB) interface, a charge management module, a power management module, a battery, a wireless communication module, an audio module, a speaker, a microphone, a sensor module, a key, a camera, a display screen, and a SIM card interface, and the like.

[0060] The processor can include one or more processing units, such as: the processor can include a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU), and the like. Among them, different processing units can be independent devices, or can be integrated in one or more processors.

[0061] Among them, the processor can be the nerve center and command center of the electronic device. The controller can generate operation control signals according to instruction operation codes and timing signals to complete the control of fetching and executing instructions.

[0062] The processor can further include a memory that stores instructions and data. In some embodiments, the memory in the processor is a cache memory. The memory can hold instructions or data that the processor has recently used or is likely to use again. If the processor needs to use the instructions or data again, it can call them directly from the memory. This avoids repeated access and reduces the processor's latency, thus improving system efficiency.

[0063] The external memory interface can be used to connect an external memory card, such as a MicroSD card, to extend the storage capacity of the electronic device. The external memory card communicates with the processor through the external memory interface to implement data storage functions. For example, music, video, and other files are saved in the external memory card.

[0064] The internal memory can be used to store computer executable program code, which includes instructions. The processor executes various functional applications and data processing of the electronic device by running the instructions stored in the internal memory. The internal memory can include a program storage area and a data storage area. The internal memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, a universal flash storage (UFS), etc.

[0065] The wireless communication function of the electronic device can be implemented through an antenna, a wireless communication module, a modem processor, and a baseband processor, etc.

[0066] The wireless communication module can provide a wireless communication solution including wireless local area networks (WLAN) (such as a wireless fidelity (Wi-Fi) network), Bluetooth (BT), a global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), infrared technology (IR), etc. applied to the electronic device.

[0067] The electronic device can implement audio functions, etc. through an audio module, a speaker, a receiver, a microphone, an earphone interface, and an application processor, etc.

[0068] The electronic device can implement a shooting function through an ISP, a camera, a video codec, a GPU, a display screen, and an application processor, etc.

[0069] The electronic device can realize display function through GPU, display screen and application processor.

[0070] The GPU is a microprocessor for image processing, connected to the display screen and the application processor. The GPU is used to perform mathematical and geometric calculations for graphics rendering. The processor can include one or more GPUs that execute program instructions to generate or change display information.

[0071] The display screen is used to display images, videos, etc. The display screen includes a display panel.

[0072] The above electronic device realizes the resource dynamic scheduling method of the distributed collaborative algorithm network intelligent agent system provided in the application. The edge intelligent agent monitors and reports resource information in real time, and predicts task demand fluctuation by using a lightweight neural network. The regional collaborative intelligent agent aggregates the states of multiple nodes based on federated learning, and uses an improved double deep Q network algorithm to formulate a cross-node resource scheduling strategy. The edge intelligent agent executes the scheduling accordingly, and uses an improved auction algorithm to disassemble and distribute tasks when local resources are insufficient. The cloud management intelligent agent constructs a global resource portrait model for long-term strategy optimization, and relies on blockchain technology to build a computing power transaction smart contract to realize trusted transaction and automatic charging. The beneficial effects of efficient dynamic configuration of algorithm network resources, improved resource utilization, reduced task response time, optimized service quality, and fair and trusted computing power transaction are achieved.

[0073] In the storage medium provided in the application, a program product capable of realizing the resource dynamic scheduling method of the distributed collaborative algorithm network intelligent agent system is stored.

[0074] The resource dynamic scheduling method of the distributed collaborative algorithm network intelligent agent system comprises: The edge intelligent agent monitors and reports local computing power, network bandwidth and task load information in real time, predicts task demand fluctuation by using a lightweight neural network, preliminarily judges the need for resource allocation weight adjustment in combination with network topology changes, and reports to the regional collaborative intelligent agent; The regional collaborative intelligent agent receives the information reported by the edge intelligent agent, aggregates the states of multiple edge nodes based on federated learning, generates a comprehensive state model, uses an improved double deep Q network algorithm to formulate a cross-node resource scheduling strategy, and issues the strategy to the edge intelligent agent; The edge intelligent agent executes task scheduling according to the cross-node resource scheduling strategy, disassembles tasks and distributes them to other edge nodes for collaborative processing if local resources are insufficient, and monitors task execution status and resource usage and generates feedback information; The cloud management intelligent agent collects resource states, task scheduling conditions and task execution results of the edge intelligent agent, constructs a global resource portrait model, optimizes and adjusts long-term planning and allocation strategies of algorithm network resources by using a long-term strategy optimization algorithm, and issues the optimized global strategy information. The idle computing power information issued by the operator is received in real time by the cloud management intelligent agent, the cross-domain resource allocation scheme is determined according to the computing task demand by the control area collaborative intelligent agent according to the smart contract to verify the credibility of the computing power transaction, the computing power transaction is completed, and the blockchain ledger is used for automatic charging. In some possible implementation manners, the resource dynamic scheduling method of the distributed collaboration-based algorithm network intelligent agent system of the present disclosure can be implemented in the form of a program product, which includes program codes for causing a terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of the present specification when the program product is run on the terminal device.

[0075] The storage medium of the present disclosure can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium may, for example, be but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include an electrical connection with one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0076] The above description of disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A distributed coordination-based algorithmic network agent system, characterized in that, Comprise: Edge agents deployed at edge nodes for real-time perception of local computing power, network bandwidth and task load, and prediction of task demand fluctuations using lightweight neural networks, real-time adjustment of resource allocation weights combined with network topology changes, and execution of preliminary task scheduling decisions; Regional collaborative agents in communication connection with multiple edge agents for aggregating multi-edge node states based on federated learning, generating cross-node resource scheduling strategies, and verifying the credibility of computing power transaction smart contracts and determining cross-domain resource allocation schemes; Cloud management agents in communication connection with the regional collaborative agents for constructing a global resource portrait model based on information provided by edge agents and regional collaborative agents, performing long-term strategy optimization, publishing global strategy information, and constructing and updating computing power transaction smart contracts based on a pre-set computing power transaction smart contract template.

2. The distributed-collaborative-based algorithmic mesh agent system of claim 1, wherein, In the global resource portrait model, the quality of service is comprehensively evaluated by the following formula: QoE(t) = w1(t)•C + w2(t)•B - w3(t)•E; Wherein, QoE(t) is the service quality experience of the user at time t, C is the normalized computing power, B is the available bandwidth, E is the energy consumption, w1(t), w2(t), w3(t) are weight factors dynamically adjusted according to historical load at time t, w1(t)+w2(t)+w3(t)=1.

3. The distributed-collaborative-based algorithmic mesh agent system of claim 1, wherein, When the edge agent executes the preliminary task scheduling decision, if the local node resources cannot meet the task demand, the improved auction algorithm is used to disassemble and distribute the task to other edge nodes for collaborative processing according to the cross-node resource scheduling strategy issued by the regional collaborative agent; The improved auction algorithm introduces a load balancing factor, and the calculation formula of the load balancing factor is: wherein, is the average load in the region, is the current load of the target edge node.

4. The distributed-collaborative-based algorithmic mesh agent system of claim 1, wherein, The regional collaborative agent generates a cross-node resource scheduling strategy using a double deep Q network algorithm; The reward function of the double deep Q network algorithm is: R = a x RU - b x DT; Wherein, RU is the resource utilization rate, DT is the task response time delay, a and b are weight coefficients.

5. The distributed-collaborative-based algorithmic mesh agent system of claim 1, wherein, The computing power transaction smart contract is constructed based on blockchain technology, and the computing power transaction smart contract includes the computing power specifications, pricing model, service availability commitment of the computing power provider, and the task demand description, budget limit and default processing clause of the computing power user.

6. The distributed-collaborative-based algorithmic mesh agent system of claim 5, wherein, The pricing model adopts a dynamic bidding mode, and the calculation formula is: Wherein, P(t) is the computing power price at time t, P0 is the initial price, k is the dynamic adjustment coefficient, D(t) is the difference between the computing power demand and supply at time t, when D(t)>0, the price rises when the demand is greater than the supply; when D(t)<0, the price falls when the supply is greater than the demand.

7. A resource dynamic scheduling method based on a distributed cooperative algorithm network intelligent agent system, characterized in that, The method uses the distributed collaborative-based algorithm network agent system according to any one of claims 1 to 6; The method comprises: Real-time monitoring and reporting of local computing power, network bandwidth and task load information by edge agents, prediction of task demand fluctuations using lightweight neural networks, preliminary judgment of resource allocation weight adjustment requirements combined with network topology changes, and reporting to regional collaborative agents; The regional coordination agent receives the information reported by the edge agent, aggregates the state of multiple edge nodes based on federated learning, generates a comprehensive state model, uses an improved double deep Q network algorithm to formulate a cross-node resource scheduling strategy, and issues the strategy to the edge agent; The edge agent executes task scheduling according to the cross-node resource scheduling strategy, and if local resources are insufficient, the agent disassembles the task and distributes it to other edge nodes for collaborative processing, while monitoring the task execution state and resource usage and generating feedback information; The cloud management agent collects the resource state, task scheduling, and task execution results of the edge agent, constructs a global resource portrait model, optimizes the long-term planning and allocation strategy of the algorithm network resources using a long-term strategy optimization algorithm, and issues the optimized global strategy information; The cloud management agent receives idle computing power information published by the operator in real time, controls the regional coordination agent to verify the credibility of the computing power transaction according to the smart contract, determines the cross-domain resource allocation scheme according to the computing task demand, completes the computing power transaction, and automatically charges using the blockchain ledger.

8. The resource dynamic scheduling method of the distributed coordination-based algorithm network intelligent agent system according to claim 7, characterized in that, The light neural network is used to predict the task demand fluctuation, which includes: Based on historical task data, a time series prediction model is constructed, short-term features are extracted through a convolutional neural network, long-term dependencies are captured using a long short-term memory network, and predicted values and confidence intervals of task demand fluctuations in a future preset time period are output.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the steps of the resource dynamic scheduling method of the distributed collaborative based algorithm network agent system according to any one of claims 7-8.

10. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the steps of the resource dynamic scheduling method of the distributed collaborative based algorithm network agent system according to any one of claims 7-8.

Images