Dynamic resource allocation method for industrial internet scenario and related device

Abstract

The application provides a dynamic resource allocation method for an industrial internet scene and related equipment. Considering that the industrial internet scene has mobile position production equipment with time-varying geographical positions, the total time delay and total energy consumption of the scene are analyzed, the total resource consumption is obtained through resource consumption algorithm based on the total time delay and total energy consumption, and then whether the target edge server executes resource reallocation within the communication range is guided according to the error between the predicted total resource energy consumption and the actual total resource energy consumption. Then, the optimal resource allocation scheme in the considered scene is adaptively solved as the target resource allocation strategy based on this by using the proximal policy optimization algorithm, and resource allocation is performed according to the target resource allocation strategy, which can meet the production demand changes of the industrial internet scene, adaptively optimize the resource allocation strategy, and improve the overall efficiency and reliability of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, and in particular to a dynamic resource allocation method for an industrial internet scenario. BACKGROUND

[0002] Industrial Internet of Things (IIoT) is crucial for the development of intelligent manufacturing. Due to the heterogeneity of industrial network systems, industrial production devices with different workloads compete for limited communication resources, thus requiring optimization of resource allocation to achieve efficient resource utilization.

[0003] However, the industrial internet environment has high dynamicity and uncertainty, and the geographical location of industrial production devices will change over time, making it difficult for traditional resource allocation strategies to adapt to changing demands. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a dynamic resource allocation method for an industrial internet scenario and related equipment to solve the above technical problems.

[0005] To achieve the above purpose, the first aspect of the present application provides a dynamic resource allocation method for an industrial internet scenario, applied to a dynamic resource allocation system, the system comprising a plurality of edge servers, any edge server in the plurality of edge servers being taken as a target edge server, the target edge server being in communication connection with at least one fixed location production device and at least one mobile location production device respectively; the method comprising:

[0006] determining a total latency corresponding to the target edge server in the industrial internet scenario, and determining a total energy consumption corresponding to the target edge server in the industrial internet scenario;

[0007] processing based on the total latency and the total energy consumption through a resource consumption algorithm to obtain an actual total resource consumption;

[0008] obtaining a historical latency and a historical energy consumption corresponding to the target edge server, obtaining a predicted total latency and a predicted total energy consumption based on the historical latency and the historical energy consumption through a pre-trained prediction model, and processing based on the predicted total latency and the predicted total energy consumption through a resource consumption algorithm to obtain a predicted total resource consumption;

[0009] determining an error between the actual total resource consumption and the predicted total resource consumption, determining whether the target edge server needs to be re-allocated resources based on the error, and obtaining a determination result;

[0010] based on the determination result, the actual total resource consumption is processed by a proximal policy optimization algorithm to obtain a target resource allocation strategy, and the fixed location production equipment and the mobile location production equipment in communication with the target edge server are allocated resources according to the target resource allocation strategy.

[0011] Based on the same inventive concept, the second aspect of the present application provides a dynamic resource allocation device for an industrial internet scene. The device is arranged in a dynamic resource allocation system, and the system includes a plurality of edge servers. Any edge server in the plurality of edge servers is taken as a target edge server, and the target edge server is in communication connection with at least one fixed location production equipment and at least one mobile location production equipment. The device includes:

[0012] A determination module is configured to determine a total time delay corresponding to the target edge server in an industrial internet scene, and determine a total energy consumption corresponding to the target edge server in an industrial internet scene.

[0013] A resource consumption processing module is configured to process the total time delay and the total energy consumption by a resource consumption algorithm to obtain an actual total resource consumption.

[0014] A prediction module is configured to obtain a historical time delay and a historical energy consumption corresponding to the target edge server, obtain a predicted total time delay and a predicted total energy consumption according to the historical time delay and the historical energy consumption by a pre-trained prediction model, and process the predicted total time delay and the predicted total energy consumption by a resource consumption algorithm to obtain a predicted total resource consumption.

[0015] A reallocation determination module is configured to determine an error between the actual total resource consumption and the predicted total resource consumption, determine whether the target edge server needs to be reallocated resources based on the error, and obtain a determination result.

[0016] An optimization module is configured to process the actual total resource consumption by a proximal policy optimization algorithm based on the determination result to obtain a target resource allocation strategy, and allocate resources to the fixed location production equipment and the mobile location production equipment in communication with the target edge server according to the target resource allocation strategy.

[0017] Based on the same inventive concept, the third aspect of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable by the processor. When the computer program is executed by the processor, the method of the first aspect is implemented.

[0018] Based on the same inventive concept, the fourth aspect of the present application provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the method of the first aspect above.

[0019] As can be seen from the above, the dynamic resource allocation method for an industrial internet scene and related equipment provided by the present application considers that there are mobile location production devices with time-varying geographical positions in the industrial internet scene, therefore, the total time delay and total energy consumption of the scene are analyzed, and based on the total time delay and total energy consumption, the actual total resource consumption is obtained through a resource consumption algorithm, then whether the target edge server executes resource reallocation within the communication range is guided according to the error between the predicted total resource energy consumption and the actual total resource energy consumption, and then the optimal resource allocation scheme in the considered scene is adaptively solved as the target resource allocation strategy based on this using a proximal policy optimization algorithm, and resource allocation is performed according to the target resource allocation strategy, which can meet the production demand changes of the industrial internet scene, adaptively optimize the resource allocation strategy, and improve the overall efficiency and reliability of the system. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present application or related art, the drawings needed to be used in the embodiments or related art description will be briefly introduced as follows. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1A The cooperation optimization framework of the embodiments of the present application is shown in the figure;

[0022] Figure 1B The model architecture corresponding to the dynamic resource allocation system of the embodiments of the present application is shown in the figure;

[0023] Figure 1C The flowchart of the proximal policy optimization algorithm processing of the embodiments of the present application is shown in the figure;

[0024] Figure 1D The bandwidth release and allocation diagram due to the movement of the mobile location production device of the embodiments of the present application is shown in the figure;

[0025] Figure 2 The flowchart of the dynamic resource allocation method for an industrial internet scene of the embodiments of the present application is shown in the figure;

[0026] Figure 3 The structural block diagram of the dynamic resource allocation device for an industrial internet scene of the embodiments of the present application is shown in the figure;

[0027] Figure 4A schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the embodiments and the accompanying drawings.

[0029] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be understood as their common meanings to those of ordinary skill in the art to which the present application pertains. The terms "first", "second", and similar terms used in the embodiments of the present application do not denote any order, quantity, or importance, but are merely used to distinguish different components. The terms "include", "contain", and similar terms mean that the elements or objects before the terms encompass the elements or objects listed after the terms and their equivalents, and do not exclude other elements or objects. The terms "connect" or "connected" and similar terms do not mean physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "upper", "lower", "left", "right", and the like are merely used to indicate relative positional relationships, and when the absolute positions of the described objects are changed, the relative positional relationships can also be changed accordingly.

[0030] It can be understood that, before using the technical solutions of the embodiments of the present application, the user will be informed of the types of personal information involved, the scope of use, the use scenarios, and the like by appropriate means, and the authorization of the user will be obtained.

[0031] For example, in response to receiving an active request of the user, prompt information is sent to the user to explicitly prompt the user that the operation requested to be performed will require the acquisition and use of personal information of the user. Thus, the user can voluntarily choose whether to provide personal information to the electronic device, application program, server, or storage medium, and the like software or hardware that performs the operation of the technical solutions of the present application according to the prompt information.

[0032] As an optional but non-limiting implementation manner, in response to accepting the active request of the user, the manner of sending prompt information to the user may, for example, be a pop-up window manner, and the prompt information may, for example, be presented in the form of text in the pop-up window. In addition, the pop-up window may, for example, carry a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0033] It can be understood that the above notification and user authorization process is merely illustrative, and does not limit the implementation manners of the present application, and other manners that meet the relevant laws and regulations can also be applied to the implementation manners of the present application.

[0034] Industrial Internet of Things (IIoT) is crucial for the development of smart manufacturing. However, the heterogeneity and complexity of industrial systems pose significant challenges to dynamic and effective resource allocation. First, existing resource allocation strategies often lack in-depth analysis of the impact of resource reallocation in dynamic scenarios: mobile industrial production equipment in smart factories undergo regular movements to connect various production processes, and the resource allocation strategy of mobile equipment is affected by its location changes. In addition, existing resource allocation cannot linguistically understand unstructured data in the factory, such as factory production plans and state information of various devices in the factory. This makes it difficult for allocation strategies to adapt to changing needs. Therefore, this application considers both problems and introduces Large Language Models (LLM) into IIoT to propose a new framework and analyze the dynamic resource allocation problem in smart factories with time-varying geographical location characteristics. The dynamic resource allocation framework determined in this application provides a new perspective for flexible decision-making and resource allocation in smart factories. The solutions to the problems listed are as follows:

[0035] 1) Existing allocation schemes lack language understanding ability and cannot achieve flexible decision-making. In current research, many resource allocation strategies cannot fully understand unstructured data in the factory, such as various state information in production plan files, which need to be interpreted and converted by technical personnel. This lack of language understanding ability makes existing models unable to flexibly adjust their strategies to respond to changes in factory production processes or production priorities, thereby affecting overall production efficiency and resource utilization. Traditional deep reinforcement learning (Deep Reinforce Learning, DRL) models have difficulty directly utilizing unstructured data and need to rely on pre-defined fixed optimization goals. When the production process or production priority of the factory changes, these models cannot flexibly adjust their strategies to respond to new needs. Therefore, this application proposes a fusion framework based on LLM and deep reinforcement learning DRL. This framework uses the natural language processing capabilities of LLM to interpret unstructured production data, combined with the strategy exploration capabilities of DRL, to achieve intelligent decision-making for dynamic resource allocation problems. In this way, we can respond to changes in factory production needs while optimizing resource allocation strategies to improve the overall efficiency and reliability of the system.

[0036] 2) Lack of analysis of the impact of resource reallocation in dynamic scenarios. The devices in a smart factory are not evenly distributed within the coverage range of the edge nodes. In most cases, the devices in the factory are arranged according to the production capacity and production process of the factory, so their distribution has the characteristics of hot spot distribution and concentration distribution. In addition, the devices in a smart factory are not all fixed, and there are mobile industrial production devices (M-IPD) to assist industrial production. M-IPD also needs to transmit data with edge nodes at all times, and its transmission delay requirement is more stringent than that of fixed devices (i.e., fixed position production devices). Due to the mobility of M-IPD, there will inevitably be two processes of moving away from the edge node (target edge server) and moving closer to the edge node over time, which is not the case for fixed devices. If the resource allocation is considered according to the case where M-IPD is far away from the edge node, more bandwidth needs to be allocated to M-IPD in order to guarantee the delay requirement, but when M-IPD gradually approaches the edge node, M-IPD does not need so much bandwidth to guarantee the production delay requirement, and the excess bandwidth is allocated, which inevitably leads to a decrease in bandwidth utilization; the resource allocation scheme in the case where M-IPD is close to the edge node is difficult to meet the link delay requirement when M-IPD is far away from the edge node. In order to guarantee the real-time position of M-IPD, the bandwidth allocation scheme needs to be adjusted dynamically. However, if the bandwidth allocation scheme is adjusted at every time slot of M-IPD movement, it will inevitably cause a large amount of calculation energy consumption of the edge node, and the ideal situation should be that the node makes autonomous decision based on the actual situation whether the allocation scheme needs to be adjusted to achieve the balance between system delay and energy consumption. Therefore, based on the proposed fusion framework, a dynamic resource allocation method for industrial internet scenarios is provided, and a dynamic resource allocation strategy based on adaptive language information near-end policy optimization is designed.

[0037] In addition, IIoT gradually evolves into a key technology platform for intelligent manufacturing. Through network interconnection, data exchange, and system interoperation, IIoT realizes precise control of production processes, reasonable scheduling of manufacturing processes, flexible allocation of resources, and the creation of an application-centric industrial ecosystem. Typical IIoT applications include network resource scheduling, fault diagnosis, and device state prediction. In these application scenarios, industrial production device IPDs generate a large number of computing tasks, which are usually sensitive to latency or consume a large amount of energy. IPDs usually have limited power, storage, and computing capabilities to process and transmit large amounts of production data. In addition, traditional cloud computing has significant latency delays, making it difficult to meet the various Quality of Experience (QoE) or Quality of Service (QoS) requirements of IIoT applications. To solve this problem, edge computing (EC) is introduced in the Internet of Things, allowing industrial production device IPDs to transmit computing tasks directly to nearby edge servers (ES), thereby reducing task latency and device energy consumption compared to centralized cloud computing.

[0038] To achieve more efficient and intelligent resource management, related technologies optimize power control and computing resource allocation for mobile edge computing, effectively reducing task latency. For example, a DRL-based offloading decision and time allocation algorithm is used to solve mixed integer nonlinear programming problems and optimize resource allocation. Joint optimization of sub-channel allocation, offloading decision, and computing resource allocation is used to minimize latency; a joint optimization problem is proposed to minimize energy consumption while meeting latency requirements. An intelligent-driven 5G network Internet of Things green resource allocation mechanism is proposed, which uses deep reinforcement learning for efficient resource management; a DRL-based resource allocation and IIoT device scheduling joint strategy is used to minimize latency, energy consumption, and model accuracy. The impact of mobile devices on resource allocation is analyzed, and a new mobile-aware scheduling method based on DRL is proposed to minimize the maximum task latency.

[0039] In addition, in order to better solve different natural language processing tasks in different fields, the related technology uses a pre-trained LLM, which can process natural language processing tasks under a unified pre-training-fine-tuning paradigm. In the pre-training phase, the model usually learns general language representation through large-scale unlabeled text. In the fine-tuning phase, the pre-trained model is supervised using annotated and task-specific datasets to perform specific downstream tasks. For example, a lightweight model based on the pre-trained model (BERT) architecture is trained for network threat detection in the Internet of Things; a framework is proposed to endow large language models with optimal planning capabilities (LLM+P), which combines the advantages of LLM and classic planners to solve planning problems; hierarchical goal decomposition and structured action planning using LLM significantly improve the success rate of completing complex tasks.

[0040] The problems of the above related technology are as follows:

[0041] 1) Due to the heterogeneity of industrial network systems, industrial production devices (IPDs) with different workloads compete for limited communication resources, thus requiring optimization of resource allocation to achieve efficient resource utilization. However, effective resource allocation for IIoT faces many challenges. First, the access of a large number of devices leads to a highly distributed and complex IIoT environment, increasing the demand for industrial network processing capacity and data transmission capacity. Second, the IIoT environment has high dynamicity and uncertainty, with the geographical location of devices changing over time, leading to frequent changes in communication paths and network topology. This further complicates the management and optimization of network resources, requiring the factory network to quickly respond and adjust resource allocation strategies. The resource allocation problem is a non-linear integer programming problem, which is non-deterministic polynomial (NP-hard), making it unrealistic to find a global optimal solution in polynomial time. The low complexity of traditional iterative optimization methods leads to a trade-off in model performance, especially for systems with multi-dimensional inputs, making it difficult to solve the above problems.

[0042] 2) Although the related technology considers the characteristics of continuous production tasks in factories and dynamic wireless environments, it only discusses resource allocation strategies from the perspective of devices and tasks, without considering the impact of frequent resource allocation in dynamic scenarios on the system. This impact needs to be quantified to build a more reasonable allocation strategy.

[0043] 3) The existing DRL-based methods have two main limitations. First, the lack of language understanding ability. Traditional DRL models have difficulty directly utilizing unstructured data, such as production plan documents and various state information in IIoT, which require manual interpretation and conversion. Second, existing models rely on pre-defined and fixed optimization objectives. When the production process or production priority of the factory changes, these models cannot flexibly adjust the strategy to respond to new demands, thereby affecting overall productivity and resource utilization.

[0044] To solve the above problems, the LLM is introduced into the IIoT, giving the IIoT system more powerful natural language interaction capabilities, and providing a new perspective for intelligent decision-making in complex manufacturing environments. Specifically, the present application: 1) designs a fusion framework, namely the LLM-driven collaborative optimization (LCOP) framework, which aims to combine the natural language understanding and generation capabilities of LLM with the strategy exploration capabilities of DRL. In addition, LCOP is applied to dynamic resource allocation to demonstrate its specific workflow. 2) analyzes the resource allocation problem in dynamic industrial scenarios characterized by time-varying geographic locations. Considering the impact of frequent resource reallocation in dynamic scenarios on system energy consumption, a reallocation decision variable is proposed to guide the ES to reallocate resources within the communication range. The long short-term memory (LSTM) network is used to analyze data changes and affect the threshold of the decision variable. 3) Under the proposed framework, a dynamic resource allocation method for industrial internet scenarios is proposed, and an adaptive language-adaptive language (ALIP-DRA) algorithm based on language information is designed to solve the optimal allocation scheme in the considered scenario.

[0045] Based on large language models, edge computing, dynamic resource allocation, and deep reinforcement learning, the present application proposes an LLM-driven joint optimization framework and applies it to time-varying industrial internet scenarios to achieve a dynamic resource allocation scheme for decision automation. Specifically, the present application: 1) combines the natural language understanding and generation capabilities of LLM with the strategy exploration capabilities of DRL; 2) considering the impact of frequent resource reallocation caused by the movement of mobile devices, a reallocation decision variable is proposed to guide the edge node to reallocate resources; 3) based on the above scenario, deep reinforcement learning is introduced to solve the scheme.

[0046] Among them, the collaborative optimization (LCOP) framework proposed by the present application is as shown in Figure 1A :

[0047] Goal: Minimizing factory production cost with limited resources (GOAL: Minimizing factory production cost with limited resources).

[0048] Step 1: Optimization suggestion: Collect relevant data such as production plan and factory status information, clean and normalize them, and fine-tune the general LLM to adapt to the characteristics and needs of the industrial scene. The standard input format is {Instruction: Plan, Information}. The instruction represents the input prompt, representing the decision direction of the LLM, which can be cost minimization, equipment condition prediction, safety risk warning, fault elimination, etc. The LLM provides corresponding output (Output) according to the input, such as optimization goals. In the scenario considered by the present application, the instruction is assumed to be cost minimization in the production process, and the output is the joint optimization of delay and energy consumption.

[0049] Step 2: Data prediction: Based on the LLM output, use LSTM to predict and integrate the predicted value (X Pre ) from the historical values of delay and energy consumption (Historcal data) and analyze (Evaluate) the actual value (X obs ) to express the difference between the data in the form of error and directly affect the threshold of the re-allocation decision variable, thereby affecting the action strategy of the DRL agent, where the error can be the root mean square error (root means square error, RMSE).

[0050] Step 3: Agent action: The agent executes the adaptive language-adaptive language (adaptive language-adaptive language, ALIP-DRA) algorithm based on language information to explore the optimal allocation strategy.

[0051] Based on a reallocation decision, an optimization goal under constraints is determined, wherein the constraints include delay constraints, energy constraints, and bandwidth constraint, and the optimization goal is to minimize system cost.

[0052] Embodiments of the present application provide a dynamic resource allocation method for an industrial internet scenario. Considering that the industrial internet scenario involves mobile location production equipment with time-varying geographical positions, the total time delay and total energy consumption of the scenario are analyzed, and the actual total resource consumption is obtained through a resource consumption algorithm based on the total time delay and total energy consumption. Then, whether the target edge server performs resource reallocation within the communication range is guided according to the error between the predicted total resource energy consumption and the actual total resource energy consumption. Then, the optimal resource allocation scheme in the considered scenario is adaptively solved as a target resource allocation strategy by using a proximal policy optimization algorithm, and resource allocation is performed according to the target resource allocation strategy. The production demand changes of the industrial internet scenario can be met, and the resource allocation strategy can be adaptively optimized to improve the overall efficiency and reliability of the system.

[0053] The method of the embodiments is applied to a dynamic resource allocation system, and the system includes a plurality of edge servers. Any edge server in the plurality of edge servers is taken as a target edge server, and the target edge server is in communication connection with at least one fixed location production equipment and at least one mobile location production equipment.

[0054] Figure 1B As shown, a three-layer model of a smart factory for LCOP deployment corresponding to the system is shown, which includes a strategy and optimization (SO) layer, an intelligent processing (IP) layer, and a perception layer. The fine-tuned LLM is deployed in the SO layer to provide optimization suggestions for the lower layer. A plurality of edge servers ES are located in the IP layer, and the DRL algorithm is executed. The agent takes actions according to the state information of the factory according to the suggestions provided by the upper layer. The perception layer includes a plurality of production equipment IPDs.

[0055] The present application defines the production equipment with a fixed geographical position (i.e., the fixed location production equipment) IPD as an F-IPD, and the total number of the F-IPDs is M. The set of the F-IPDs is M = {fIPD1, fIPD2,..., fIPD i ,..., fIPD M} are represented. However, in actual factories, not all IPDs are located at fixed positions. There are IPDs in motion (i.e., mobile position production devices), denoted by a variable mIPD, which need to move within an area to complete production tasks. IPDs exchange data with ESs (edge servers), the number of which is N, denoted by a set N = {ES1, ES2,..., ES i ,..., ES N}. Each edge server ES has a certain communication range and is able to communicate with IPDs within its coverage. For a target edge server ES i , the set of F-IPDs within its coverage is It is known that A time slot model {1, 2,..., t,..., T} is used to describe a complete production process in an industrial internet scenario. Fixed position production devices and mobile position production devices mIPD upload data to the target edge server ES i at time slot t in amounts of and Since the amount of data for downlink transmission is usually small, only uplink from production devices IPD to edge servers ES is considered.

[0056] wherein production devices IPDs and edge servers ESs are modeled using a two-dimensional coordinate system, wherein fixed position production devices mobile position production devices mIPD and target edge server ES i are represented by coordinates and (x i , y i ), respectively.

[0057] As shown in Figure 2 , the method comprises:

[0058] Step 201, determining the total latency corresponding to the target edge server in the industrial internet scenario, and determining the total energy consumption corresponding to the target edge server in the industrial internet scenario.

[0059] In this step, there are fixed production devices within the coverage of the target edge server ES i in the industrial internet scenario, and at time slot t, mobile position production devices mIPD enter the coverage of the target edge server ES i and communicate with it.

[0060] Considering that there are mobile position production devices with time-varying geographical positions in the industrial internet scene, total time delay and total energy consumption of the scene are analyzed as the basis for determining actual total resource consumption.

[0061] In step 202, the actual total resource consumption is obtained by processing based on the total time delay and the total energy consumption through a resource consumption algorithm.

[0062] In this step, the total time delay and the total energy consumption are comprehensively considered, and the speed and accuracy of obtaining the actual total resource consumption can be ensured through the resource consumption algorithm.

[0063] In step 203, the historical time delay and the historical energy consumption corresponding to the target edge server are obtained, the predicted total time delay and the predicted total energy consumption are obtained according to the historical time delay and the historical energy consumption through a pre-trained prediction model, and the predicted total resource consumption is obtained based on the predicted total time delay and the predicted total energy consumption through a resource consumption algorithm.

[0064] In this step, the prediction model introduces an LSTM network, which uses a memory block and a gate mechanism to maintain the internal state of the network, explores the internal abstract features and underlying structure of the data, and has strong prediction ability for time series.

[0065] It can map the historical sequence x to the predicted sequence y=(y1, y2,..., y Tpre) :

[0066] τ t =σ(W τ ·[h t-1 ,x t ]+b τ );

[0067]

[0068] f gt =σ(W f ·[h t-1 ,x t ]+b f );

[0069]

[0070] e t =σ(W e ·[hx -1 ,x t ]+b e );

[0071] h t =e t ⊙tanh(C t);

[0072] y t = φ(W y h t + b y ).

[0073] where τ, C, fg, and e are the input gate, cell activation, forge gate, and output gate vectors, respectively. h t and y t are the candidate state, hidden layer state, and final output, respectively. W τ , W f , W c , W e , and W y are the weight matrices, b τ , b f , b c , b e , and b y are the corresponding bias vectors. σ(·) is the Sigmoid function, is the element-wise product of vectors, and φ is the network output activation function.

[0074] The LSTM is used to establish prediction models for the two sub-sequences of time delay T all，t and energy consumption E all，t , to obtain predicted total time delay and predicted total energy consumption , which can more accurately capture the change rule of the data. Finally, the predicted total time delay and the predicted total energy consumption are comprehensively considered, and the speed and accuracy of the predicted total resource consumption are ensured through the resource consumption algorithm.

[0075] In step 204, the error between the actual total resource consumption and the predicted total resource consumption is determined, and based on the error, it is determined whether the target edge server performs resource reallocation, to obtain a determination result.

[0076] In this step, the difference between the actual total resource consumption and the predicted total resource consumption is evaluated through the root mean square error function, to guide whether the target edge server performs resource reallocation within the communication range.

[0077] In step 205, based on the determination result, the actual total resource consumption is optimized through a proximal policy optimization algorithm, to obtain a target resource allocation strategy, and the fixed location production equipment and the mobile location production equipment that communicate with the target edge server are allocated resources according to the target resource allocation strategy.

[0078] In this step, different allocation schemes have different effects on the system cost. Under the constraints of bandwidth resources, communication delay, etc., the optimal resource allocation scheme is obtained to balance the delay and energy consumption.

[0079] The present application establishes an optimization problem O t , that is,

[0080]

[0081] The constraint condition corresponding to the optimization problem needs to meet the following:

[0082]

[0083] wherein, represents the bandwidth allocation scheme (i.e. the target resource allocation strategy). C1 represents the maximum delay constraint condition. Constraint condition C2 indicates that the target edge server ES i has a limited allocable bandwidth B i . In addition, only when the mobile location production device mIPD communicates with the target edge server ES i , the bandwidth corresponding to the mobile location production device is allocated. and mIPD. represents the bandwidth corresponding to the fixed location production device. is used to indicate whether to perform resource reallocation.

[0084] Since the state of the system changes with time, the optimization problem O t that minimizes the actual total resource consumption is expressed as a Markov decision process (MDP). The state transition probability matrix of the MDP is also difficult to obtain. However, a model-free reinforcement learning algorithm can obtain a suboptimal solution without any prior knowledge.

[0085] Proximal Policy Optimization (PPO) algorithm is a new policy gradient (PG) algorithm, which designs a new objective function to realize the minimum batch update, solves the problem that the PG algorithm is sensitive to the step length and is difficult to determine a reasonable step length. The PPO algorithm originates from the Trust Region Policy Optimization (TRPO) algorithm, but it is easier to solve than the TRPO algorithm. Compared with the TRPO algorithm, the PPO algorithm uses a scissor agent target instead of a constraint function, which reduces the complexity of the algorithm. The key of reinforcement learning is to set the state space, action space and reward, which is very important for the algorithm to quickly converge to the correct result. The process of the PPO algorithm is shown in Figure 1C

[0086] For the optimization problem proposed in this application, the state information of all production equipment IPDs, including geographic location, transmission data volume, etc. is taken as the state space of reinforcement learning, and the bandwidth allocation scheme of the target edge server ESs is taken as the action space. Therefore, the optimization goal of this application is to minimize the average delay, but reinforcement learning generally maximizes the reward, so the complex value of the average delay is taken as the reward, that is:

[0087]

[0088] Through the above scheme, considering that the industrial internet scene exists a mobile location production equipment with time-varying geographic location, the total delay and total energy consumption of the scene are analyzed, and the total resource consumption is obtained based on the total delay and total energy consumption by using the resource consumption algorithm. Then, according to the error between the predicted total resource energy consumption and the actual total resource energy consumption, it is guided whether the target edge server executes resource reallocation within the communication range, and then the proximal policy optimization algorithm is used to adaptively solve the optimal resource allocation scheme in the considered scene as the target resource allocation strategy, and resource allocation is performed according to the target resource allocation strategy. It can meet the production demand changes of the industrial internet scene, and adaptively optimize the resource allocation strategy, improve the overall efficiency and reliability of the system.

[0089] In some embodiments, the system further comprises a policy optimization layer.

[0090] In step 201, the total delay corresponding to the target edge server in the industrial internet scene is determined, including:

[0091] In step A1, the transmission power of the mobile location production equipment communicating with the target edge server is obtained ​Gaussian noise power N0 of the mobile location production device, channel path loss from the mobile location production device to the target edge server Small-scale fading H caused by mobility of the mobile location production device t Distance from the mobile location production device to the target edge server Signal interference of the fixed location production device in communication with the target edge server

[0092] Step A2, based on the The The The H t , the N0 and the The first signal-to-noise ratio of the mobile location production device is obtained by the following formula

[0093]

[0094] Step A3, obtaining the data amount of the mobile location production device uploaded to the target edge server at time slot t And the communication bandwidth of the mobile location production device and the target edge server

[0095] Step A4, based on the The And the The first delay of the mobile location production device in communication with the target edge server is obtained by the following formula

[0096]

[0097] Step A5, obtaining the transmission power of the fixed location production device in communication with the target edge server Gaussian noise power N0 of the fixed location production device, channel path loss from the fixed location production device to the target edge server PL(d ij，i ), distance d ij，i from the fixed location production device to the target edge server Signal interference of the mobile location production device in communication with the target edge server

[0098] Step A6, based on the The d ij，i , the PL(d ij，j ), the N0 and the A second signal-to-noise ratio of the fixed-location production equipment is obtained by the following formula

[0099]

[0100] Step A7, obtaining an amount of data uploaded by the fixed-location production equipment to the target edge server at time slot t and a communication bandwidth of the fixed-location production equipment and the target edge server

[0101] Step A8, based on the the and the A second latency of the fixed-location production equipment communicating with the target edge server is obtained by the following formula

[0102]

[0103] Step A9, obtaining an amount of data transmitted by the target edge server and a wired link transmission rate R between the target edge server and the policy optimization layer w .

[0104] Step A10, based on the and the R w A third latency of the target edge server communicating with the policy optimization layer is obtained by the following formula

[0105]

[0106] Step A11, obtaining a number of central processor cycles c required for processing the amount of data transmitted by all the target edge servers n，t and a computing capability f of the policy optimization layer SO .

[0107] Step A12, based on the c n，t and the f SO A fourth latency T of the policy optimization layer performing calculation is obtained by the following formula SO，t：

[0108]

[0109] Step A13, based on the the and the T SO，t A total latency T is obtained by the following formula all，t：

[0110]

[0111] In the above scheme, the uplink transmission latency of the mobile location production device mIPD (i.e., the first latency) is:

[0112] The signal to noise ratio (Signal to Noise Ratio, SNR) at the mobile location production device mIPD (i.e., the first signal to noise ratio) can be given by:

[0113]

[0114] where the distance from the mobile location production device to the target edge server is It can be obtained in the following way:

[0115]

[0116] The coordinates of the mobile location production device mIPD and the target edge server ES i are denoted as and (x i , y i ), respectively.

[0117] In addition, the channel path loss from the mobile location production device to the target edge server can be obtained in the following way:

[0118]

[0119] where PL(d) is a function of d, which can be calculated by:

[0120]

[0121] where f c is the carrier center frequency, d0 represents the reference distance, c is the speed of light, equal to 3 × 108m / s. n loss is the distance-dependent index, and X σ represents that the shadow fading follows a normal distribution with a standard deviation of σ.

[0122] The uplink transmission latency of the fixed location production device F-IPD (i.e., the second latency) is:

[0123] The signal to noise ratio (Signal to Noise Ratio, SNR) at the target edge server ES i can be given by:

[0124]

[0125] where the distance from the fixed location production device to the target edge server is dij，i The channel path loss PL(d) from the fixed location production device to the target edge server can be obtained as follows:

[0126]

[0127] Fixed location production device and target edge server ES i The coordinates are represented as (x ij，i , y ij，i ) and (x i , y i ) respectively.

[0128] In addition, the channel path loss PL(d ij，i ) from the fixed location production device to the target edge server can be obtained as follows:

[0129]

[0130] where PL(d) is a function of d, which can be calculated by:

[0131]

[0132] where f c is the carrier center frequency, d0 represents the reference distance, c is the speed of light, equal to 3 x 108m / s. n loss is the distance-dependent index, X σ represents that the shadow fading follows a normal distribution with a standard deviation of σ.

[0133] The communication delay between the target edge server ES and the policy optimization (SO) layer (i.e., the third delay) is:

[0134] At this stage, the target edge server ES i will upload the key information of all production devices IPDs in the coverage area collected to the SO layer. The transmission delay can be calculated as:

[0135]

[0136] The SO layer calculates the delay (i.e., the fourth delay) as:

[0137] The calculation delay of the SO layer processing the data uploaded by the target edge server ES is given by:

[0138]

[0139] Based on and T SO，t The total delay T all，t can be quickly obtained by the following formula:

[0140]

[0141] In some embodiments, the system further comprises a policy optimization layer.

[0142] In step 201, the total energy consumption corresponding to the target edge server in the industrial internet scenario is determined, including:

[0143] Step B1, obtaining the transmission power of the mobile location production equipment communicating with the target edge server and the first time delay of the mobile location production equipment communicating with the target edge server

[0144] Step B2, based on the and the determine the first energy consumption of the fixed location production equipment data upload by the following formula

[0145]

[0146] Step B3, obtaining the transmission power of the fixed location production equipment communicating with the target edge server and the second time delay of the fixed location production equipment communicating with the target edge server

[0147] Step B4, based on the and the determine the second energy consumption of the fixed location production equipment data upload by the following formula

[0148]

[0149] Step B5, obtaining the transmission power of the target edge server communicating with the policy optimization layer and the third time delay of the target edge server communicating with the policy optimization layer

[0150] Step B6, based on the and the determine the third energy consumption of the target edge server data upload by the following formula

[0151]

[0152] Step B7, obtaining the energy consumed by the target edge server in each central processor cycle and the number of central processor cycles required for each resource reallocation of the target edge server

[0153] Step B8, based on the and stated The fourth energy consumption E for resource reallocation by the target edge server is determined by the following formula: re，t :

[0154]

[0155] Step B9, based on the described described and the E re，t The total energy consumption E is determined by the following formula all，t :

[0156]

[0157] in, Indicates whether the target edge server needs to reallocate resources in time slot t. Resource reallocation is performed when No resource reallocation is performed.

[0158] In the above scheme, the energy consumption during data exchange (including the first energy consumption, the second energy consumption and the third energy consumption) is:

[0159] Fixed-position production equipment The energy consumption of data upload (i.e. the second energy consumption) is given by the following formula

[0160]

[0161] Similarly, the mobile location production device mIPD and the target edge server ES can be obtained i The upload energy consumption is (i.e. first energy consumption) and (i.e. the third energy consumption).

[0162] Calculation consumption:

[0163] make (in J / Gcycle) represents the energy consumed per CPU cycle, calculated at the SO layer The energy consumption is as follows:

[0164]

[0165] in, (J / Gcycle) indicates the target edge server ES i The energy consumed in each CPU cycle, the target edge server ES iThe number of CPU cycles required for each resource reallocation

[0166] The energy consumption in the resource reconfiguration process (i.e., the fourth energy consumption) is:

[0167] The computing energy consumption of the edge server ES for each resource reallocation can be calculated by the following formula:

[0168]

[0169] Based on and E re，t The total energy consumption E all，t can be quickly determined by the following formula:

[0170]

[0171] In some embodiments, step 202 includes:

[0172] Based on the total latency T all，t and the total energy consumption E all，t , the actual total resource consumption O t is determined by the following formula:

[0173] O t = w1·A·T all，t+ w2 · B·E all，t

[0174] wherein w1 and w2 are preset weighting parameters of the latency and energy consumption corresponding to the resource consumption, satisfying w1+w2=1, w1∈(0,1), w2∈(0,1), and A and B are preset values for normalizing the latency and energy consumption units into resource consumption units.

[0175] In the above scheme, by comprehensively considering the total latency T all，t and the total energy consumption E all，t , and normalizing both into values of resource consumption units, the accuracy of the actual total resource consumption O t obtained can be improved, and it is also convenient for subsequent gap analysis with the predicted total resource consumption.

[0176] In some embodiments, in step 203, the predicted total resource consumption is obtained by processing the predicted total latency and the predicted total energy consumption by a resource consumption algorithm, including:

[0177] Based on the predicted total latency and the predicted total energy consumption , the predicted total resource consumption is determined by the following formula:

[0178]

[0179] wherein w1 and w2 are preset weighting parameters corresponding to the resource consumption of the time delay and the energy consumption satisfying w1+w2=1, w1∈(0, 1), w2∈(0, 1), A and B are preset values for normalizing the time delay and the energy consumption into the resource consumption unit.

[0180] In the above scheme, by comprehensively considering the predicted total time delay and the predicted total energy consumption and normalizing both into the value of the resource consumption unit, both the accuracy of the obtained predicted total resource consumption can be improved, and the gap analysis with the actual total resource consumption can be facilitated.

[0181] In some embodiments, in step 204, the error between the actual total resource consumption and the predicted total resource consumption is determined, including:

[0182] based on the actual total resource consumption O t and the predicted total resource consumption the error RMSE is determined by the following formula:

[0183]

[0184] wherein T represents all time slots of the entire production process corresponding to the industrial internet scene, and t represents time slot t.

[0185] In the above scheme, the error of the present application is preferably the root mean square error, which can effectively measure the deviation between the predicted total resource consumption and the actual total resource consumption O t and provide sensitivity to large errors, which makes the RMSE more suitable in application scenarios that require strict control of large errors.

[0186] In some embodiments, in step 204, the target edge server is determined whether to perform resource reallocation based on the error, to obtain a determination result, including:

[0187] Step C1, obtaining the location of the target edge server and the location of the mobile location production equipment in communication with the target edge server

[0188] Step C2, based on the and the the distance d Δ between the target edge server and the mobile location production equipment is determined by the following formula:

[0189]

[0190] Step C3, obtaining the communication bandwidth of the mobile production device and the target edge server and the time interval Δτ.

[0191] Step C4, based on the and the Δτ, determining the change amount φ of the communication bandwidth occupation time in the time interval Δτ by the following formula Δ :

[0192]

[0193] Step C5, based on the d Δ and the φ Δ , determining the resource reallocation variable Γ(φ Δ , d Δ ) by the following formula

[0194] Γ(φ Δ , d Δ ) = log2[(1 + arctan φ Δ )(1 + arctan d Δ )].

[0195] Step C6, based on the Γ(φ Δ , d Δ ) and the error RMSE, determining whether the target edge server performs resource reallocation by the following formula, to obtain a determination result:

[0196]

[0197] wherein Γ0 represents a preset decision threshold, represents the determination result, when resource reallocation is performed, and when resource reallocation is not performed.

[0198] In the above scheme, an example of releasing and allocating bandwidth due to the movement of the mobile production device mIPD is shown in Figure 1D , which needs to reallocate bandwidth in the coverage of the target edge server ES1 and the target edge server ES2 to minimize the O t during its movement.

[0199] The reallocation of bandwidth resources makes the latency of the system decrease, but the resource reallocation makes the energy consumption of the system increase. Therefore, a resource reallocation variable Γ(φ Δ , d Δ ) is proposed to guide the target edge server ESi Is it necessary to adjust the allocation plan? Δ , d Δ ) can be expressed as:

[0200] Γ(φ Δ , d Δ )=log2[(1+arctanφ Δ )(1+arctand Δ )];

[0201] Among them, due to d Δ and φ Δ The minimum, maximum, and distribution functions of are unknown and unstable, so the inverse tangent function is chosen as the normalization function.

[0202] merge and Γ(φ Δ , d Δ )=log2[(1+arctanφ Δ )(1+arctan d Δ )], we can get the following redistribution decision criterion:

[0203]

[0204] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied in a distributed scenario and performed by multiple devices working together. In such a distributed scenario, one of the multiple devices may only perform one or more steps of the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the method.

[0205] It should be noted that the above description is limited to some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in an order different from that described in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0206] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a dynamic resource allocation device for industrial Internet scenarios.

[0207] refer to Figure 3The device is arranged in a dynamic resource allocation system, the system includes a plurality of edge servers, any edge server in the plurality of edge servers is taken as a target edge server, the target edge server is respectively in communication connection with at least one fixed position production equipment and at least one mobile position production equipment; the device includes:

[0208] The determining module 301 is configured to determine a total time delay corresponding to the target edge server in an industrial internet scene, and determine a total energy consumption corresponding to the target edge server in the industrial internet scene;

[0209] The resource consumption processing module 302 is configured to process based on the total time delay and the total energy consumption through a resource consumption algorithm to obtain an actual total resource consumption;

[0210] The prediction module 303 is configured to obtain a historical time delay and a historical energy consumption corresponding to the target edge server, obtain a predicted total time delay and a predicted total energy consumption through a prediction model pre-trained according to the historical time delay and the historical energy consumption, and process based on the predicted total time delay and the predicted total energy consumption through the resource consumption algorithm to obtain a predicted total resource consumption;

[0211] The reallocation determining module 304 is configured to determine an error between the actual total resource consumption and the predicted total resource consumption, determine whether the target edge server performs resource reallocation based on the error to obtain a determination result;

[0212] The optimization module 305 is configured to perform optimization processing on the actual total resource consumption through a proximal policy optimization algorithm based on the determination result to obtain a target resource allocation strategy, and perform resource allocation on the fixed position production equipment and the mobile position production equipment in communication with the target edge server according to the target resource allocation strategy.

[0213] For the convenience of description, the above device is described as various modules respectively described in function. Of course, in the implementation of the present application, the functions of each module can be implemented in one or more software and / or hardware.

[0214] The device of the above embodiment is used to implement the corresponding dynamic resource allocation method for the industrial internet scene in any of the above embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be described here.

[0215] Corresponding to the method of any of the above embodiments based on the same inventive concept, the application further provides an electronic device, comprising 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 method for dynamic resource allocation facing an industrial internet scene according to any of the above embodiments.

[0216] Figure 4 A more specific hardware structure of an electronic device provided by the embodiment is shown, which can include a processor 401, a memory 402, an input / output interface 403, a communication interface 404, and a bus 405. The processor 401, the memory 402, the input / output interface 403, and the communication interface 404 are connected to each other through the bus 405 for internal communication.

[0217] The processor 401 can be implemented in the form of a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, etc., for executing related programs to implement the technical solutions provided by the embodiments of the present specification.

[0218] The memory 402 can be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 402 can store an operating system and other application programs, and when the technical solutions provided by the embodiments of the present specification are implemented by software or firmware, the related program codes are stored in the memory 402 and called and executed by the processor 401.

[0219] The input / output interface 403 is used to connect input / output modules to realize information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. The input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0220] The communication interface 404 is used to connect a communication module (not shown in the figure) to realize the communication interaction between the device and other devices. The communication module can realize communication through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.).

[0221] The bus 405 includes a path for transferring information between the various components (for example, the processor 401, the memory 402, the input / output interface 403, and the communication interface 404) of the device.

[0222] It should be noted that although the above device only shows the processor 401, the memory 402, the input / output interface 403, the communication interface 404, and the bus 405, in the specific implementation process, the device can also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device can also only contain the components necessary to implement the embodiments of the present application, and does not have to contain all the components shown in the figure.

[0223] The electronic device of the above embodiment is used to implement the dynamic resource allocation method for the industrial internet scene in any of the preceding embodiments, and has the beneficial effects of the corresponding method embodiments, which are not repeated here.

[0224] Based on the same inventive concept, corresponding to the method of any of the above embodiments, the present application also provides a non-transitory computer readable storage medium storing computer instructions for causing the computer to execute the dynamic resource allocation method for the industrial internet scene as described in any of the above embodiments.

[0225] The computer readable medium of the present embodiment includes permanent and non-permanent, removable and non-removable media, which can be implemented by any method or technology to store information. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device.

[0226] The computer instructions stored in the storage medium of the above embodiment are used to cause the computer to execute the dynamic resource allocation method for the industrial internet scene as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which are not repeated here.

[0227] Those skilled in the art will understand that the foregoing discussion of any of the embodiments is merely illustrative of the application and is not intended to limit the scope of the application as described in the appended claims. Numerous modifications, variations and adaptations will be apparent to those skilled in the art in light of the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) can use the embodiments discussed.

[0228] In addition, to simplify the description and discussion, and so as not to make the embodiments of the application difficult to understand, the well-known power / ground connections to integrated circuit (IC) chips and other components can or can not be shown in the provided drawings. Further, devices can be shown in block diagram form so as not to make the embodiments of the application difficult to understand, and this also takes into account the fact that details regarding implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the application are to be implemented (i.e., these details should be well within the understanding of one of skill in the art). Where specific details (e.g., circuitry) are set forth in order to describe an illustrative embodiment of the application, it should be apparent to those skilled in the art that the embodiments of the application can be practiced without, or with variation of, these specific details. The description is thus to be considered as illustrative and not restrictive, and the scope of the application should be determined not with reference to the description, but rather with reference to the appended claims.

[0229] While the application has been described in connection with specific embodiments thereof, it will be understood that many modifications, variations and adaptations will be apparent to those skilled in the art in light of the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) can use the embodiments discussed.

[0230] It is intended to cover all such modifications, variations and adaptations as fall within the scope of the application. Therefore, it should be understood that any of the embodiments of the application can include one, another or several of the features discussed above, and combinations thereof. Accordingly, any and all modifications, variations or equivalent arrangements that fall within the scope of this application should be considered as being within the scope of the application.

Claims

1. A dynamic resource allocation method for industrial Internet scenarios, characterized in that: Applied to a dynamic resource allocation system, the system includes multiple edge servers, any one of the multiple edge servers is used as a target edge server, and the target edge server is respectively communicatively connected to at least one fixed-location production device and at least one mobile-location production device; the method includes: Determine a total latency corresponding to the target edge server in an industrial internet scenario, and determine a total energy consumption corresponding to the target edge server in an industrial internet scenario; Processing the total delay and the total energy consumption through a resource consumption algorithm to obtain actual total resource consumption; Obtaining historical delays and historical energy consumption corresponding to the target edge server, obtaining a predicted total delay and a predicted total energy consumption based on the historical delays and the historical energy consumption using a pre-trained prediction model, and processing the predicted total delay and the predicted total energy consumption using a resource consumption algorithm to obtain a predicted total resource consumption; determining an error between the actual total resource consumption and the predicted total resource consumption, and determining whether to reallocate resources to the target edge server based on the error, to obtain a determination result; Based on the determination result, the actual total resource consumption is optimized by a proximal policy optimization algorithm to obtain a target resource allocation strategy, and resources are allocated to the fixed-location production equipment and the mobile-location production equipment that communicate with the target edge server according to the target resource allocation strategy; The system also includes a strategy optimization layer; Determining the total delay corresponding to the target edge server in the industrial Internet scenario includes: Obtain the transmission power of the mobile location production device communicating with the target edge server , the Gaussian noise power of the mobile production equipment , the channel path loss from the mobile location production device to the target edge server Small-scale fading caused by the mobility of the mobile location production equipment , the distance from the mobile location production equipment to the target edge server Signal interference with fixed-location production equipment communicating with the target edge server ; Based on the 、 、 、 、 and stated The first signal-to-noise ratio of the mobile production equipment is obtained by the following formula : ； Get the mobile location of the production equipment in the time slot The amount of data uploaded to the target edge server , and the communication bandwidth between the mobile location production device and the target edge server ; Based on the 、 and stated The first delay of communication between the mobile location production device and the target edge server is obtained by the following formula : ； Obtain the transmission power of the fixed-location production equipment communicating with the target edge server , the Gaussian noise power of the fixed position production equipment , the channel path loss from the fixed location production equipment to the target edge server , the distance from the fixed location production equipment to the target edge server Signal interference with mobile location production equipment communicating with the target edge server ; Based on the 、 、 、 and stated The second signal-to-noise ratio of the fixed-position production equipment is obtained by the following formula : ； Get the fixed position production equipment in the time slot The amount of data uploaded to the target edge server , and the communication bandwidth between the fixed-location production equipment and the target edge server ; Based on the 、 and stated The second delay for the communication between the fixed-position production equipment and the target edge server is obtained by the following formula : ； Get the amount of data transmitted by the target edge server , and the wired link transmission rate between the target edge server and the policy optimization layer ; Based on the and stated The third delay for the target edge server to communicate with the policy optimization layer is obtained by the following formula: : ； Get the number of CPU cycles required to process the amount of data transmitted by all target edge servers , and the computing power of the strategy optimization layer ; Based on the and stated The fourth delay calculated by the strategy optimization layer is obtained by the following formula: : ； Based on the 、 、 and stated The total delay is obtained by the following formula : , where N is the number of edge servers, M i The number of production devices in the target edge server coverage area; The determining whether to reallocate resources on the target edge server based on the error to obtain a determination result includes: Get the location of the target edge server , and the location of the mobile location production device communicating with the target edge server ; Based on the and stated The distance between the target edge server and the mobile location production equipment is determined by the following formula : ； Obtain the communication bandwidth between the mobile location production device and the target edge server , and the time interval ; Determine communication bandwidth The occupancy time in the time interval The change in : ； Based on the and stated The resource reallocation variable is determined by the following formula : ； Based on the and the error The following formula is used to determine whether the target edge server should reallocate resources and obtain the result: in, represents the preset decision threshold, Indicates the result of determination. Resource reallocation is performed when No resource reallocation is performed.

2. The method according to claim 1, characterized in that The system also includes a strategy optimization layer; The determining of the total energy consumption corresponding to the target edge server in the industrial Internet scenario includes: Obtain the transmission power of the mobile location production device communicating with the target edge server , and a first delay for the mobile location production device to communicate with the target edge server ; Based on the and stated The first energy consumption of the mobile location production equipment data upload is determined by the following formula : ； Obtain the transmission power of the fixed-location production equipment communicating with the target edge server , and a second delay for the fixed location production device to communicate with the target edge server ; Based on the and stated The second energy consumption of the fixed position production equipment data upload is determined by the following formula : ； Obtain the transmission power for the target edge server to communicate with the policy optimization layer , and a third delay for the target edge server to communicate with the policy optimization layer ; Based on the and stated The third energy consumption of data upload to the target edge server is determined by the following formula: : ； Get the energy consumed by the target edge server in each CPU cycle , and the number of CPU cycles required for each resource reallocation of the target edge server ; Based on the and stated The fourth energy consumption of the target edge server for resource reallocation is determined by the following formula: : ； Based on the 、 、 and stated The total energy consumption is determined by the following formula : in, Indicates whether the target edge server needs to be in the time slot Reallocate resources when Resource reallocation is performed when No resource reallocation is performed.

3. The method according to claim 1, characterized in that The processing based on the total delay and the total energy consumption by a resource consumption algorithm to obtain actual total resource consumption includes: Based on the total delay and the total energy consumption The actual total resource consumption is determined by the following formula : in, and It is the presupposition of satisfaction The weighted parameters corresponding to the resource consumption of delay and energy consumption, , , and It is a preset value that normalizes the latency and energy consumption units to resource consumption units.

4. The method according to claim 1, wherein The processing based on the predicted total delay and the predicted total energy consumption by a resource consumption algorithm to obtain the predicted total resource consumption includes: Based on the predicted total delay and the predicted total energy consumption The predicted total resource consumption is determined by the following formula : in, and It is the presupposition of satisfaction The weighted parameters corresponding to the resource consumption of delay and energy consumption, , , and It is a preset value that normalizes the latency and energy consumption units to resource consumption units.

5. The method according to claim 1, wherein Determining the error between the actual total resource consumption and the predicted total resource consumption includes: Based on the actual total resource consumption and the predicted total resource consumption The error is determined by the following formula : in, It represents all time slots of the entire production process corresponding to the industrial Internet scenario. Represented as time slots .

6. A dynamic resource allocation device for industrial Internet scenarios, characterized in that: The device is provided in a dynamic resource allocation system, the system including a plurality of edge servers, any one of the plurality of edge servers being used as a target edge server, the target edge server being respectively communicatively connected to at least one fixed-location production device and at least one mobile-location production device; the device comprising: A determination module is configured to determine a total latency corresponding to the target edge server in an industrial Internet scenario, and to determine a total energy consumption corresponding to the target edge server in an industrial Internet scenario; a resource consumption processing module configured to process the total delay and the total energy consumption using a resource consumption algorithm to obtain actual total resource consumption; a prediction module configured to obtain historical delays and historical energy consumption corresponding to the target edge server, obtain a predicted total delay and a predicted total energy consumption based on the historical delays and the historical energy consumption using a pre-trained prediction model, and process the predicted total delay and the predicted total energy consumption using a resource consumption algorithm to obtain a predicted total resource consumption; a reallocation determination module configured to determine an error between the actual total resource consumption and the predicted total resource consumption, and determine whether to perform resource reallocation on the target edge server based on the error, to obtain a determination result; an optimization module configured to optimize the actual total resource consumption using a proximal policy optimization algorithm based on the determination result to obtain a target resource allocation strategy, and allocate resources to the fixed-location production equipment and the mobile-location production equipment that communicate with the target edge server according to the target resource allocation strategy; The system further includes a strategy optimization layer; a determination module, specifically configured to: Obtain the transmission power of the mobile location production device communicating with the target edge server , the Gaussian noise power of the mobile production equipment , the channel path loss from the mobile location production device to the target edge server Small-scale fading caused by the mobility of the mobile location production equipment , the distance from the mobile location production equipment to the target edge server Signal interference with fixed-location production equipment communicating with the target edge server ; Based on the 、 、 、 、 and stated The first signal-to-noise ratio of the mobile production equipment is obtained by the following formula : ； Get the mobile location of the production equipment in the time slot The amount of data uploaded to the target edge server , and the communication bandwidth between the mobile location production device and the target edge server ; Based on the 、 and stated The first delay of communication between the mobile location production device and the target edge server is obtained by the following formula : ； Obtain the transmission power of the fixed-location production equipment communicating with the target edge server , the Gaussian noise power of the fixed position production equipment , the channel path loss from the fixed location production equipment to the target edge server , the distance from the fixed location production equipment to the target edge server Signal interference with mobile location production equipment communicating with the target edge server ; Based on the 、 、 、 and stated The second signal-to-noise ratio of the fixed-position production equipment is obtained by the following formula : ； Get the fixed position production equipment in the time slot The amount of data uploaded to the target edge server , and the communication bandwidth between the fixed-location production equipment and the target edge server ; Based on the 、 and stated The second delay for the communication between the fixed-position production equipment and the target edge server is obtained by the following formula : ； Get the amount of data transmitted by the target edge server , and the wired link transmission rate between the target edge server and the policy optimization layer ; Based on the and stated The third delay for the target edge server to communicate with the policy optimization layer is obtained by the following formula: : ； Get the number of CPU cycles required to process the amount of data transmitted by all target edge servers , and the computing power of the strategy optimization layer ; Based on the and stated The fourth delay calculated by the strategy optimization layer is obtained by the following formula: : ； Based on the 、 、 and stated The total delay is obtained by the following formula : , where N is the number of edge servers, M i The number of production devices in the target edge server coverage area; The reallocation determination module is specifically configured as follows: Get the location of the target edge server , and the location of the mobile location production device communicating with the target edge server ; Based on the and stated The distance between the target edge server and the mobile location production equipment is determined by the following formula : ； Obtain the communication bandwidth between the mobile location production device and the target edge server , and the time interval ; Determine communication bandwidth The occupancy time in the time interval The change in : ； Based on the and stated The resource reallocation variable is determined by the following formula : ； Based on the and the error The following formula is used to determine whether the target edge server should reallocate resources and obtain the result: in, represents the preset decision threshold, Indicates the result of determination. Resource reallocation is performed when No resource reallocation is performed.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 5.

Images