Multi-industrial chain network task allocation method based on industrial agent dynamic resources and reliability

Abstract

The invention discloses a multi-industrial-chain network task allocation method based on industrial agent dynamic resources and reliability, which comprises the following steps of: firstly, modeling a multi-industrial-chain network based on the industrial agent dynamic resources and the reliability, and adaptively changing a reliability value along with continuous allocation of tasks; the same industrial agent has different reliability degrees in different network layers, but the reliability degrees influence each other; the target of task allocation is that each task in the task flow is optimally allocated so as to minimize the task evaluation function and minimize the total execution time of the tasks; the reliability of the industrial agent is dynamically adjusted through a reliability cascade effect model; solving an initial population of a task allocation problem by using a problem context knowledge driven meta-heuristic method, wherein each individual in the population is a feasible task allocation scheme; by means of the method, the tasks in the task flow are allocated to the better industrial agent alliance, and the optimal task allocation scheme is achieved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent computing technology of multi-industrial chain networks, and mainly relates to a task allocation method for multi-industrial chain networks based on the dynamic resources and reliability of industrial intelligent agents. Background Technology

[0002] In modern industrial environments, cross-regional collaboration among industrial agents leads to multi-layered industrial chain networks. These networks represent various interactions between industrial agents across different industrial chains through multiple network layers, including logistics, supplier relationships, and technological cooperation. Compared to traditional single-layer industrial chain networks, multi-layered networks exhibit significant complexity and heterogeneity. Generally, traditional task allocation only considers cooperation among industrial agents within a single chain to complete the task. However, multi-layered industrial chain networks require effective handling of the complex interactions between industrial agents across multiple industrial chains.

[0003] In the actual operation of industrial chains, complex tasks need to be assigned to multiple industrial agents to complete, and the industrial agents responsible for the tasks form a task alliance. The task alliance is initiated by one industrial agent, which autonomously seeks out cooperating industrial agents to jointly complete the complex task. Compared with other types of task allocation methods, task alliances demonstrate the process by which industrial agents spontaneously seek cooperative industrial agents in actual tasks. The coupling and heterogeneity of multi-industry chain networks bring significant challenges to the task allocation problem. Specifically, the challenges are as follows: 1) Compared with single-layer industrial chain networks, the structure of multi-industry chain networks means that industrial agents can seek task partners across different industrial chains. Therefore, the candidate pool of cooperative industrial agents is enormous. 2) Multi-industry chain networks contain various types of resources. An industrial agent possesses multiple resources, which can be utilized by multiple industrial chains. Therefore, resource overlap leads to resource coupling in multi-industry chain networks. 3) The quality evaluation of industrial agents in multi-industry chain networks is affected by reliability, task resource satisfaction, and task processing efficiency. In multi-layered industrial supply chain networks, the reliability, resource quantity, and resource processing capacity of industrial agents differ across network layers, and these factors influence each other. Therefore, it is difficult to comprehensively measure the production capacity of industrial agents using information from multiple network layers.

[0004] Because of the large number of industrial agents in a multi-chain industrial network, there are numerous feasible task allocation schemes that can meet the task requirements. Therefore, the solution space for the task allocation problem in a multi-chain industrial network is enormous and complex. After a task in the task flow of a multi-chain industrial network is completed, the industrial agent will replenish its own resources and update its reliability based on the reliability reward of the task to prepare for the next stage of the task. Therefore, the reliability and resource availability of an industrial agent differ when performing different tasks, and both resources and the reliability of the industrial agent are affected by the task allocation results.

[0005] From a mathematical perspective, the task allocation problem in multi-industrial chain networks is an NP-hard problem with a highly complex solution space. In engineering, for complex resource tasks of industrial agents, forming high-quality industrial agent alliances can significantly reduce project time and task execution costs, playing a crucial role in improving the economic efficiency of industrial agents and promoting socio-economic development. In recent years, the improvement of socio-economic levels and the intensification of market competition have placed higher demands on the capabilities of industrial agents. Effective task alliance control can greatly improve the execution efficiency and market competitiveness of industrial agents. Therefore, research on the formation and adjustment of alliances in multi-industrial chain networks has both strong practical relevance and significant academic value. Those skilled in the art are also delving into the characteristics of multi-industrial chain networks, combining industrial agent resource allocation and objective constraints, hoping to design effective solution methods. Summary of the Invention

[0006] This invention addresses the challenge of existing heavy industrial chain networks failing to meet the constraints of multiple couplings and dynamic changes in the resources and reliability of industrial agents during task processing. It proposes a task allocation method for multiple industrial chain networks based on the dynamic resources and reliability of industrial agents. First, the multiple industrial chain network is modeled based on the dynamic resources and reliability of industrial agents. The reliability value adaptively changes as tasks are allocated. The reliability of the same industrial agent differs across different network layers but influences each other. The goal of task allocation is to achieve optimal allocation for each task in the task flow, minimizing the task evaluation function and the total execution time. Then, the reliability of industrial agents is dynamically adjusted using a reliability cascade effect model. A metaheuristic method driven by problem context knowledge is used to solve the initial population of the task allocation problem, where each individual represents a feasible task allocation scheme. Finally, the method of this invention allocates tasks in the task flow to a better alliance of industrial agents, achieving the optimal task allocation scheme.

[0007] To achieve the above objectives, the technical solution adopted by this invention is: a multi-industrial chain network task allocation method based on dynamic resources and reliability of industrial intelligent agents, comprising the following steps:

[0008] S1. Multi-industrial chain network modeling: The model is based on the dynamic resources and reliability of industrial agents. Each layer of the multi-industrial chain network represents a connected, undirected weighted graph of multiple industrial chains. The l-th layer of the multi-industrial chain network is represented as N. l = <V l E l >, among which In network layer N l The set of industrial intelligent agent nodes mapped in the middle. For network layer N l The set of edges, where nc is the total number of industrial intelligent agents; industrial intelligent agents C i In network layer N l The reliability in is The reliability value will change adaptively as tasks are continuously assigned. The reliability of the same industrial intelligent agent will be different in different network layers, but they will affect each other.

[0009] S2. Target Calculation: The target of the task allocation is to achieve optimal allocation for each task in the task flow, in order to minimize the task evaluation function and minimize the total execution time of the tasks.

[0010]

[0011] in, It is to solve task T k The best industrial intelligent agent alliance, where nt is the number of tasks in task set T; express Solve the current task T k The task evaluation function value, express Solve the current task T k Execution time.

[0012] S3. Dynamic Reliability and Resource Adjustment: The reliability of the industrial agent is dynamically adjusted through a reliability cascade effect model. The resources of the industrial agent are dynamically adjusted based on its resource contribution to the task and the resource replenishment amount per task cycle. In the reliability cascade effect model, task T... k The reliability bonus is defined as:

[0013]

[0014] Where, sumR j Represents resource r in a multi-industry chain network j Total amount, tr k,j T represents k For resource r j Demand, Ratio k It is task Tk The ratio of resource demand to total resources, nc l It is network N l The total number of industrial intelligent agents, μ is the weight of the reliability reward;

[0015] S4. Problem Solving: The initial population of the task allocation problem is solved using a metaheuristic method driven by problem context knowledge. Each individual in the population represents a feasible task allocation scheme. During the solution of the initial population, the performance of the industrial intelligent agent is evaluated using task allocation priority and context task processing capability. The task allocation priority is determined based on context reliability and context resource satisfaction.

[0016] As an improvement of the present invention, in step S1, the multi-industrial chain network model also includes task alliance constraints, whereby the industrial agents in the industrial agent alliance are interconnected in the multi-industrial chain network and the resources provided by the industrial agent alliance meet the resource requirements of the task; the interconnection includes intra-layer connections in the network and inter-layer connections between industrial agents with different mappings.

[0017] As an improvement of the present invention, the task evaluation function F(CA) in step S2 k Specifically:

[0018]

[0019] Where β1 and β2 represent the values ​​assigned to task T, respectively. k The weights of reliability and resource satisfaction of the industrial intelligent agent alliance, ACol k For industrial intelligent agents within the Industrial Intelligent Agent Alliance to target task T k The cooperation cost, ARep k For task T k The sum of the reliability of the industrial intelligent agents within the alliance, ASat k This indicates that the mapping of industrial intelligent agents within the alliance to task T k The sum of resource satisfaction.

[0020] As another improvement of the present invention, in step S2, the alliance cooperation cost ACol k The calculation method is as follows:

[0021]

[0022] The reliability ARep of the industrial intelligent agent alliance k The calculation method is as follows:

[0023]

[0024] The alliance resource satisfaction level ASat k The calculation method is as follows:

[0025]

[0026] in, To map industrial intelligent bodies For task T k Resource satisfaction.

[0027] As an improvement of the present invention, the execution time function Tmax of the task in step S2 k (CA k Specifically:

[0028]

[0029] Among them, CA k To satisfy task T k In a demand-based alliance, nr is the number of resource types. TP k,j Task T for the Industrial Intelligent Agents Consortium k Required resources r j The total time consumed is calculated as follows:

[0030]

[0031] in, For the Industrial Intelligent Agents Alliance Contributing unit resources r j Required processing time yes For task T k Contribute resources r j The amount.

[0032] As another improvement to the present invention, in the reliability cascade effect model of step S3, according to the Industrial Intelligent Agents Consortium (CA) k Industrial intelligent agents for task T k Resource satisfaction will be rewarded with reliability bonuses. k Assigned to each alliance member, task T k After execution, the Industrial Intelligent Agents Alliance (CA) k Chinese Industrial Intelligent Body Reliability rewards obtained as follows:

[0033]

[0034] in, Indicates members within the alliance For task T k Resource satisfaction, Rewardk Reliability reward;

[0035] Each alliance member The reliability reward obtained is propagated to neighboring industrial agents that are not in the alliance through intra-layer links. Receive corresponding reliability rewards as follows:

[0036]

[0037] in, Member of the alliance The set of neighboring intelligent agents on the same floor, γ l It is network N l Reliability influencing factors For members of the alliance With neighbors on the same floor Communication costs;

[0038] Each alliance member Reliability reward The mapping of industrial agents is propagated to other networks, thus the industrial agents Reliability rewards obtained across layers for:

[0039]

[0040] Among them, CL i For agent C i The set of network layers, γ l It is network N l The reliability influencing factor.

[0041] As a further improvement of the present invention, in step S3, the industrial intelligent agent is task T. k Contributed resource amount The calculation method is as follows:

[0042]

[0043] in, Indicates that for task T k Industrial intelligent agents that have joined the alliance For intelligent agents Having resources r j The amount, tr k,j For task T k Required resources;

[0044] After the current task is completed, the intelligent agent The resources will be dynamically adjusted as follows:

[0045]

[0046] in, For intelligent agents Having resources r j The amount, For intelligent agents For task T k Contributed resources r j The amount, For each task cycle, the agent Supplementary resources j The amount.

[0047] As a further improvement of the present invention, in step S4, the method for calculating the situation reliability in task allocation priority is as follows:

[0048]

[0049] in, Represents mapping industrial intelligent agents Contextual reliability in multi-industry chain networks, CL i It is related to industrial intelligent body C i Corresponding mapping industrial intelligent body The set of network layer indices where it is located Industrial intelligent agents representing mapping In network N l Contextual reliability; mapped industrial intelligent agents In network N l Contextual reliability The definition is as follows:

[0050]

[0051] in, Industrial intelligent agents and neighbors on the same floor The minimum cooperation cost between them, γ l It is network N l Reliability influencing factors;

[0052] The method for calculating the contextual resource satisfaction in task allocation priority is as follows:

[0053]

[0054] in, Industrial intelligent agents For task T k Contextual resource satisfaction, CN i It is an industrial intelligent agent C in a multi-industry chain network. iThe corresponding set of adjacent industrial intelligent agents, where α represents the collaborative resource influencing factor, and CV i′ It is an industrial intelligent agent C in a multi-industry chain network. i′ The corresponding set of mapping industrial intelligent agents.

[0055] Industrial intelligent bodies For task T k Task allocation priority is calculated as follows:

[0056]

[0057] Where β1 and β2 represent the values ​​assigned to task T, respectively. k The weights of the reliability and resource satisfaction of the industrial intelligent agent alliance.

[0058] The calculation method for contextual task processing ability is as follows:

[0059]

[0060] Where ω is the collaborative resource processing factor, To possess resources j Quantity, And not equal to 0, Represents mapping industrial intelligent agents The task processing capability is calculated as follows:

[0061]

[0062] As another improvement of the present invention, after the initial population is generated in step S4, each feasible task allocation scheme in the population is perturbed by operation operators. The operation operators include, but are not limited to, destruction and reconstruction operations based on context reliability, destruction and reconstruction operations based on context resource satisfaction, destruction and reconstruction operations based on task allocation priority, and destruction and reconstruction operations based on context task processing capability.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] (1) Optimize resource utilization efficiency and task processing time in dynamic environment: The method of the present invention allocates tasks by real-time sensing of the resource status and reliability level of industrial intelligent agents, which overcomes the limitation of traditional static resource scheduling that cannot adapt to dynamic environment changes, thereby significantly improving the overall utilization rate of system resources and reducing task processing time.

[0065] (2) Enhancing the reliability of the task processing alliance: The unique feature of this invention is that it considers the constraints of multiple network structures on task allocation, and that the same industrial agent has different resources and reliability at different levels, that is, the industrial agent nodes are heterogeneous in different industrial chain networks. In addition, this method carefully considers the reliability cascading effect in multiple networks during the reliability adjustment stage. That is, a change in the reliability of one industrial agent will not only affect the reliability of other industrial agents in the current layer, but also have a chain reaction on the reliability of industrial agents in other layers. At the same time, it also considers the dynamic changes in industrial agent resources.

[0066] (3) Enhance the reliability and robustness of task allocation: The method of this invention introduces a reliability threshold constraint and a dynamic update mechanism, which effectively avoids the task execution risk caused by low reliability industrial intelligent agents and solves the problem of high task risk and poor system stability caused by only considering resources and ignoring reliability in the previous allocation strategy. Attached Figure Description

[0067] Figure 1 This is a schematic diagram illustrating the process of multi-industrial chain network modeling, reliability cascade effect modeling, and resource adjustment.

[0068] Figure 2 A schematic diagram of the initialization process in a multi-industrial chain network task allocation method that considers the dynamic resources and reliability of industrial intelligent agents;

[0069] Figure 3 This is a schematic diagram of the method of the present invention. Detailed Implementation

[0070] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0071] Example 1

[0072] The method of this invention is designed for a given multi-industry chain network structure N and task T. k Resource requirements for the task k And the resource set CR of each industrial intelligent agent to complete task T k The optimal allocation of resources and reliability for an industrial agent is achieved by dynamically adjusting these resources based on task completion. During task execution, the industrial agent contributes its resources; after task completion, it replenishes its resources and updates its reliability based on task reliability rewards. Therefore, the reliability and resource availability of an industrial agent differ depending on the task being executed. The goal of this invention is to achieve optimal allocation for each task in the task flow, minimizing the task evaluation function and the total execution time of the tasks.

[0073] In actual industrial chain operations, tasks arrive sequentially in the form of task flows. A task flow indicates that the number of tasks is finite, and they arrive in a specific order. When executing different tasks, the reliability and resources of industrial agents are affected by the previous task allocation and execution. Therefore, in multi-chain industrial networks, the reliability and resources of industrial agents vary depending on the completion status of tasks.

[0074] A multi-industry chain network consists of various industry chain networks with different link types. Each layer represents an industry chain and is a connected, undirected, weighted graph. In a multi-industry chain network, the same industrial agent may exist in multiple industry chain networks. Due to the complexity of tasks within the industry chain, a single industrial agent cannot complete the task alone; usually, cooperation among industrial agents in multiple industry chains is required. During task allocation, tasks tend to be assigned to industrial agents with higher reliability and resource satisfaction, and shorter task processing time. Since the performance of an industrial agent in a multi-industry chain network is affected by the reliability, resource quantity, and resource processing capacity of industrial agents in different network layers, as well as the reliability, resource quantity, and resource processing capacity of neighboring industrial agents within the same network layer, contextual knowledge is used to measure the impact of industrial agents across layers and with neighboring industrial agents. Contextual knowledge includes contextual reliability, contextual resource satisfaction, task allocation priority, and contextual task processing capacity.

[0075] This invention introduces a reliability cascading effect model to reflect the selection trend of industrial agents and the change in their reliability as tasks are completed. After a task is completed, the reliability of certain industrial agents needs to be adjusted before assigning the next task. Due to the coupling of multiple industrial chain networks, changes in the reliability of industrial agents affect the reliability of adjacent industrial agents within the network, and industrial agents in other industrial chain network layers are also affected by the reliability cascading effect. Specifically, after a task is completed, the reliability reward for that task increases the reliability of industrial agents in the task alliance. Simultaneously, the reliability of industrial agents in the task alliance is fed back to neighboring industrial agents, increasing their reliability as well. Regarding resource consumption in multiple industrial chains, there is a characteristic of replenishment throughout the task cycle; industrial agent resources are replenished after each task execution cycle to prepare for the next task.

[0076] Based on the problem definition, a mathematical model for task allocation in a multi-industrial chain network considering the dynamic resources and reliability of industrial agents is established. In actual task allocation, the formed task alliances are generally connected. Therefore, when allocating task execution in the task flow, the constraints of the multi-industrial chain network structure are considered to ensure that the task alliances are connected within the multi-industrial chain network. For problems of different scales, contextual knowledge is used to solve the problem. The contextual knowledge measures the influence of neighboring industrial agents on the selection of cooperative industrial agents in the cooperative alliance.

[0077] A multi-industry chain network task allocation method based on dynamic resources and reliability of industrial intelligent agents, such as Figure 2 As shown, it includes the following steps:

[0078] Step S1: Modeling multiple industrial chains;

[0079] In multi-chain industrial networks, complex tasks require industrial agents to form alliances and cooperate. For task T... k Industrial Intelligence Consortium (CA) k This indicates that task T is satisfied in a multi-industry chain network. k A set of mapped industrial intelligent agents represents the resource requirements. The Industrial Intelligent Agent Alliance (CA) is spontaneously formed by industrial intelligent agents. When a single industrial intelligent agent encounters a complex task that it cannot complete independently, it will cooperate with neighboring industrial intelligent agents to form an industrial intelligent agent alliance to solve the complex task. As neighboring industrial intelligent agents join the alliance, the alliance continuously expands, attracting more industrial intelligent agents to provide services for the task until it is completed. Mapped Industrial Intelligent Agents Joining the alliance means that the industrial intelligent agent C i Contribution in network layer N l Resources within. The Industrial Agent Consortium (CA) is a subset of the set of industrial agents mapped in a multi-chain industrial network.

[0080] N = {B 1 N 2 ,…,N L} is a set of multi-industry chain networks, where L represents the total number of layers in the multi-industry chain network. The multi-industry chain network consists of various multi-industry chain networks with different link types. Each layer represents a connected, undirected, weighted graph of a multi-industry chain. C = {C1, C2, ..., C} nc Let} be the set of industrial agents in a multi-industry chain network, where nc is the total number of industrial agents. In a multi-industry chain network, the same industrial agent may exist in multiple multi-industry chain networks. The mapping relationship between industrial agents and multi-industry chain networks is represented by a matrix. express. Industrial intelligent agent C i In network layer N l There is a mapping in it. Industrial intelligent agent C i In network layer N l There is no mapping in the data. The l-th layer multi-industry chain network is represented as N. l = <V l E l >, among which In network layer N l The set of industrial intelligent agent nodes mapped in the middle. For network layer N l A set of edges. An edge represents a connection between two industrial intelligent agents that can cooperate to perform a task; the weight of the edge is... Represents mapping industrial intelligent agent nodes and mapping industrial intelligent body nodes Inter-layer cooperation costs. Network layer N l elements in the matrix Specifically:

[0081]

[0082] Industrial Intelligent Agent C i The set of industrial intelligent agents mapped at different network layers is as follows: in Industrial intelligent agent C i In network layer N l The mapping of industrial intelligent agents in the model. Industrial intelligent agent C i The set of network layer numbers Industrial Intelligent Agent C i Set of neighboring industrial agents in a multi-chain industrial network Industrial Intelligent Agent C i In network layer N l The collection of neighboring industrial intelligent agents Industrial Intelligent Agent C i In network layer N l The reliability in is The reliability value adapts and changes adaptively as tasks are assigned. The higher the reliability value of an industrial agent, the higher its probability of being selected when forming a coalition, and vice versa. The reliability of the same industrial agent varies across different network layers, but they influence each other. Network layer N l The reliability impact factor among industrial intelligent agents is Y. l Reliability Influence Factor Υ l Represents network layer N lThe degree to which the reliability of an internal industrial agent is affected by the reliability of its neighboring industrial agents.

[0083] The resource set in a multi-industry chain network is R = {r1, r2, ..., r}. nr}, where nr is the number of resource types. Industrial Intelligent Agent C i In network layer N l The current set of resources Industrial intelligent agent C i In network layer N l Having resources r j The number of tasks. The task set T = {T1, T2, ..., T} nk}, where nk represents the number of tasks. For task T k Its resource requirements TR k ={tr k,1 ,tr k,2 ,…,tr k,nr Each item in the set represents task T. k The quantity required for a certain resource, such as tr k,j This represents the quantity of resource type j required by task k. If T... k A certain type of resource is not needed; the required quantity of that type of resource is 0. Network layer N l Industrial intelligent agents C i Join the alliance as a mission tank k The contributed resource set is

[0084] The amount of resources replenished by the industrial intelligent agent in each task cycle is Industrial intelligent agent C i In network layer N l Supplementary resources j The quantity. The resource capacity of industrial intelligent agents is If C i If an industrial agent can produce a certain type of resource, then the capacity of that type of resource is 0. The amount of each type of resource an industrial agent possesses cannot exceed its resource capacity. The time taken by an industrial intelligent agent to process a unit of resource when performing a task is Industrial intelligent agents Contributing unit resources r j Required processing time.

[0085] In multi-chain industrial networks, complex tasks require industrial agents to form alliances and cooperate. For task T... kSelf-organizing Industrial Intelligence Alliance (CA) k This indicates that task T is satisfied in a multi-industry chain network. k A self-organizing industrial agent alliance (CA) is a set of mapped industrial agents that represent resource requirements. When a single industrial agent encounters a complex task it cannot complete independently, it will collaborate with neighboring industrial agents to form an alliance to solve the task. As neighboring industrial agents join, the alliance expands, attracting more agents to provide services for the task until it is completed. The CA is a subset of the mapped industrial agent set in a multi-chain industrial network. In a multi-chain industrial network, CS represents the set of industrial agents (CA). Specific constraints are as follows:

[0086] (1) The industrial agents in the Industrial Agent Consortium are interconnected within a multi-industry chain network. The connectivity of the Industrial Agent Consortium is defined as follows. For Path exists for The connectivity of the Industrial Agent Alliance consists of two parts: intra-layer connectivity within the network and inter-layer connectivity between different mappings of industrial agents.

[0087] (2) The resources provided by the Industrial Intelligent Agents Consortium meet the resource requirements of the task. For

[0088] Step S2: Target calculation;

[0089] Due to the complexity of tasks within multiple industrial chains, a single industrial agent cannot complete the task alone; typically, cooperation among industrial agents across multiple industrial chains is required. In the actual production phase of a multi-industrial chain, tasks tend to be assigned to industrial agents with higher reliability and resource availability. Therefore, when forming an alliance, it is necessary to select industrial agents with high reliability to join the alliance. The reliability of an industrial agent alliance is measured by ARep. k For task T k The reliability of an industrial agent alliance is the sum of the reliability of the industrial agents within the alliance. The reliability of an industrial agent alliance is directly proportional to the overall reliability of the alliance. The calculation of the reliability of an industrial agent alliance is as follows:

[0090]

[0091] The resource satisfaction of an industrial intelligent agent with a task needs to be measured by the quantity and quality of resources it possesses. In a multi-layered industrial chain network, this involves mapping industrial intelligent agents. For task T k Resource satisfaction is defined as follows:

[0092]

[0093] in Represents mapping industrial intelligent agents The quality of resources.

[0094] Alliance resource satisfaction ASat k This indicates that the mapping of industrial intelligent agents within the alliance to task T k The sum of resource satisfaction. The calculation of alliance resource satisfaction is as follows:

[0095]

[0096] Alliance cooperation cost ACol k For industrial intelligent agents within the Industrial Intelligent Agent Alliance to target task T k The cooperation costs. The alliance cooperation costs are described below:

[0097]

[0098] The goal of solving the task allocation problem in a multi-industry chain network that considers the dynamic resources and reliability of industrial intelligent agents is to complete task T in the multi-industry chain network. k The optimal allocation of tasks aims to optimize the task evaluation function and minimize the total execution time of the tasks. The task evaluation function is used to measure the performance of the industrial agent consortium. It considers consortium reliability, resource satisfaction, and cooperation costs. Minimizing the task evaluation function means maximizing the industrial agent consortium's reliability, maximizing resource satisfaction, and minimizing cooperation costs. The task evaluation function F(CA) k The definition is as follows:

[0099]

[0100] Where β1 and β2 represent the values ​​assigned to task T, respectively. k The weights of the reliability and resource satisfaction of the industrial intelligent agent alliance.

[0101] The task completion time of an industrial intelligence agent alliance is crucial to task collaboration efficiency, competitiveness, and task responsiveness. Therefore, the task processing speed of these agents must be considered when selecting them for alliance membership. The industrial intelligence agent alliance processes task T. k Required resources r j Cumulative Time Consumption (TP) k,j as follows:

[0102]

[0103] When the mission ended, the Alliance CA k For task Tk The time consumed is the maximum cumulative time consumed for each resource, task T. k The total execution time is calculated as follows:

[0104]

[0105] The formation of an alliance requires comprehensive consideration of both the alliance's quality and the task's completion time. Therefore, finding task T... k Optimal Industrial Intelligent Agent Alliance The objective function is shown below.

[0106] Minimize f1 = F(CA) k ),f2=Tmax k (CA k )

[0107] Due to task T k The optimal industrial agent alliance needs to consider both alliance quality and total task execution time. Let the set be a Pareto set. The resulting alliance must satisfy the following constraints: (1) Ensure that any two alliance members can access each other through internal paths within the alliance. (2) Ensure that the resulting alliance can meet the resource requirements of the task.

[0108] For a given set of n tasks T = {T1, T2, ..., T...} nk The task set is used as the objective function for the multi-industry chain network task allocation method problem, which is the sum of the evaluation values ​​of the industrial intelligent agent consortium for each task and the sum of the completion times of each task. The objective function is as follows:

[0109]

[0110] in It is to solve task T k The best industrial intelligent agent alliance, where nt is the number of tasks in the task set T.

[0111] Step S3: Dynamic reliability and resource adjustment;

[0112] After the current task is completed, the reliability of some industrial agents needs to be dynamically adjusted before assigning the next task. Due to the coupling nature of multiple industrial chain networks, changes in the reliability of a single industrial agent not only affect the reliability of its neighboring industrial agents in the network, but also trigger cascading effects in other industrial chain network layers through cross-chain associations, further affecting the reliability of industrial agents on a larger scale. Therefore, a reliability cascading effect model is proposed to reflect the dynamic propagation and mutual influence of the reliability of industrial agents in multiple industrial chain networks. Task T k The reliability reward is defined as follows:

[0113]

[0114] Where sumR j Represents resource r in a multi-industry chain network j Total, Ratio k It is task T k The ratio of resource demand to total resources. l It is network N l The total number of industrial intelligent agents. μ is the weight of the reliability reward.

[0115] In this embodiment, the depth of reliability diffusion within the multi-industry chain network is set to 1. Reliability changes in industrial agents within the network layer will only propagate to neighboring industrial agents. The main steps of the reliability cascading adjustment model in the multi-industry chain network are as follows:

[0116] (1) Reliability Adjustment within the Industrial Agent Consortium: Since the industrial agents in the Industrial Agent Consortium contribute their own resources to the task, according to CA... k Industrial intelligent agents for task T k The degree of resource satisfaction will determine the Reward k Assigned to each alliance member. Task T k After execution, the Industrial Intelligent Agents Alliance (CA) k Chinese Industrial Intelligent Body The reliability bonuses obtained are as follows:

[0117]

[0118] (2) Network N l Internal reliability adjustments: for each alliance member The reliability rewards obtained are propagated to neighboring industrial agents that are not in the alliance through intra-layer links. The reliability rewards obtained by neighboring industrial agents... As shown below:

[0119]

[0120] Among them, Υ l It is network N l Reliability influencing factors. Neighboring industrial intelligent agents. Outside of a consortium, the maximum reliability reward propagated by a consortium member is awarded by neighboring industrial agents. get.

[0121] (3) Reliability reward propagation across network layers: each alliance member Reliability reward The mapping of industrial intelligent agents is propagated to other networks. Industrial intelligent agents The reliability bonus obtained across layers is represented as follows:

[0122]

[0123] To prevent the continuous increase in the reliability of industrial intelligent agents, reliability normalization is introduced into the reliability cascade effect model. Each mapped industrial intelligent agent... The reliability value is expressed as Normalized to range [Rep] lb Rep ub [Inside, Rep] lb Rep represents the lower limit of the reliability of the industrial intelligent agent, set to 0.1. ub This represents the upper limit, set to 0.9. The initial reliability of an industrial intelligent agent is related to its own resource quality and the total amount of resources it possesses, calculated as follows:

[0124]

[0125] in Industrial intelligent agents The quality coefficient of the resources provided.

[0126] Because the reliability of industrial intelligent agents must meet reliability range constraints, the initial reliability needs to be normalized to [Rep]. lb Rep ub ]. In completing task T k Subsequently, the reliability of the industrial intelligent agent is adjusted in a cascade manner according to the following:

[0127]

[0128] The reliability of all industrial agents is within acceptable limits. [Rep] lb Rep ub Adjustments will be made within the [internal framework]. Mapping industrial intelligent bodies. Reliability value Normalization adjustments are performed using the following method:

[0129]

[0130] The amount of resources contributed by an industrial intelligent agent is related to the order in which it joins the alliance. If an industrial intelligent agent joins the alliance, its current resource r is... j If the remaining resources are not used to fulfill the task, the industrial agent will contribute all of its available resources. j Resources. If an industrial agent joins the alliance, the resources r currently possessed by the industrial agent... j If the remaining resources are sufficient to meet the task requirements, the industrial intelligent agent will fill the resource gaps in the task. The calculation is shown below.

[0131]

[0132] in Indicates that for task T k Industrial agents that have joined the alliance contribute their resources to the task. After the task is completed, the industrial agents replenish their resources for the next stage of manufacturing. The dynamic changes in the resources of industrial agents are shown below:

[0133]

[0134] in It is an industrial intelligent body C i Within each task cycle, at network layer N l Supplementary resources j The quantity.

[0135] Step S4: Problem-Context Knowledge-Driven Metaheuristic Approach Flow;

[0136] (1) For each task in the task flow, an initial population for solving the task allocation problem in a multi-industrial chain network is generated using a stochastic heuristic method, such as... Figure 2 As shown, each individual in the population represents a feasible task allocation scheme. During the initial population generation process, task allocation priority and contextual task processing capabilities are used to evaluate the performance of the industrial agent. Task allocation priority comprehensively considers contextual reliability and contextual resource satisfaction; contextual reliability considers the network N... l The impact of neighboring industrial agents on the reliability of industrial agents. Mapped industrial agents. In network N l Contextual reliability The definition is as follows:

[0137]

[0138] in Industrial intelligent agents and The minimum cooperation cost between them, Υ l It is network N l The reliability influencing factor.

[0139] Because industrial agents are mapped into a multi-layered industrial chain network, the reliability of each mapped industrial agent is affected by the reliability of other mapped industrial agents within the industrial chain network. If industrial agent C in the multi-layered industrial chain network... i If there are multiple mapping industrial intelligent agents, then Represents mapping industrial intelligent agents Contextual reliability in multi-layered industrial supply chains It is expressed as follows:

[0140]

[0141] Among them CL i It is related to industrial intelligent body C i Corresponding mapping industrial intelligent body The set of network layer indices. In a multi-industry chain network, the contextual resource satisfaction of an industrial agent is influenced by its neighboring industrial agents. For task T k The contextual resource satisfaction is calculated as follows.

[0142]

[0143] Among them, CN i It is an industrial intelligent agent C in a multi-industry chain network. i The corresponding set of adjacent industrial intelligent agents, where α represents the collaborative resource influencing factor. Industrial intelligent agents For task T k The alliance priority is calculated as follows.

[0144]

[0145] Where β1 and β2 represent task T respectively. k The weights for the reliability and resource satisfaction of the industrial intelligent agent alliance are determined. The method for calculating the contextual task processing capability is as follows:

[0146]

[0147] Where ω is the collaborative resource processing factor. To possess resources j Quantity, And it is not 0. Represents mapping industrial intelligent agents The task processing capability is calculated as follows:

[0148]

[0149] The heuristic process method used in this step is as follows:

[0150] ① Calculate any two industrial agents in a multi-industry chain network structure Minimum cooperation cost between Minimum cooperation cost The shortest path is obtained through the aggregation network, which is represented as follows:

[0151]

[0152] ② Randomly select task allocation priority or situational task processing capability as the evaluation index for industrial intelligent agents. Calculate the performance of each industrial intelligent agent. The index value.

[0153] ③ For each industrial intelligent agent Set accessibility flag =1; Initial task alliance set CA k and the mission alliance formation sequence π (CA) k ) is an empty set, and the set of candidate industrial intelligent agents, CanSet, is an empty set.

[0154] ④ Sort all industrial intelligent agents from highest to lowest according to their indicator values, and randomly select one industrial intelligent agent from the top E-ratio%. As the initiator of the mission alliance, [they will...]. Accessibility flag Set to 0 and update CA. k and π(CA) k ).

[0155] ⑤ The initiator of the alliance Unvisited neighboring industrial agents are added to the candidate industrial agent set CanSet. Unvisited neighboring industrial agents represent 1). and This industrial intelligence agent is connected to other industrial intelligence agents at different network layers. The neighboring industrial intelligence agents are represented by C. i′ ∈CN i , Unvisited neighbor industrial agents and

[0156] ⑥ Determine whether the resources of the current task alliance meet the resource requirements of the task. If they do, output the task alliance formation result (CA). k and π(CA) k If the conditions are not met, then continue to expand the alliance using steps ⑦ and ⑧.

[0157] ⑦ Sort the industrial intelligent agents in the CanSet set according to their index values ​​from high to low, and randomly select one industrial intelligent agent from the top 30% of the industrial intelligent agents. Join the mission alliance. Accessibility flag Set to 0 and update CA. k and π(CA) k ).

[0158] ⑧ New industrial intelligent agents joining the alliance Unvisited neighboring industrial agents are added to the candidate industrial agent set CanSet.

[0159] (2) After the initial population is generated, four types of operation operators can be used to perturb each feasible task allocation scheme in the method population to obtain a better feasible scheme. The operation operators include destruction and reconstruction operations based on context reliability, destruction and reconstruction operations based on context resource satisfaction, destruction and reconstruction operations based on task allocation priority, and destruction and reconstruction operations based on context task processing capability. In this process, each individual in the population randomly selects an operation operator to update the feasible allocation scheme. Since the agents within the alliance are interconnected, the formation of the alliance follows a chain structure according to the order in which the mapping agents join the alliance. The destruction and reconstruction process of the operation operators is described below.

[0160] ① Disruption operation process: Randomly select a position Des in the coalition formation sequence π (CA). pos And Des pos ∈[1,size(π(CA))]. The mapped agent at the selected position and all subsequent agents will be deleted. The order of alliance formation after the disruption operation is represented as π(CA)′.

[0161] ② Reconstruction process: If the location Des is destroyed pos If the value equals 1, then a new initiator is selected for the alliance, and the entire alliance is rebuilt according to the selected operation operator. Destruction and reconstruction operations based on context reliability rebuild the entire alliance using context reliability as the indicator. Destruction and reconstruction operations based on context resource satisfaction rebuild the entire alliance using context satisfaction as the indicator. Destruction and reconstruction operations based on task allocation priority rebuild the entire alliance using task allocation priority as the indicator. Destruction and reconstruction operations based on context task processing capacity rebuild the entire alliance using context task processing capacity as the indicator. If the destruction location Des... pos If the value is not equal to 1, then elite cooperative agents are selected from the CanSet of π(CA)′. Agents in the CanSet are ranked based on their knowledge of the specific problem context, and the top E-ratio% agents in the CanSet are considered elite cooperative agents. An elite agent is randomly selected from the CanSet to join the alliance, and the Const is updated until the alliance is able to complete the task.

[0162] (3) Determine if the iteration has ended. If the iteration has not ended, return to step (2) to continue the destruction and reconstruction operation. If the iteration has ended, perform a fast non-dominated sort on the current population and randomly select an optimal task allocation scheme on the Pareto front to execute the current task.

[0163] (4) Based on the selected task allocation scheme, dynamically adjust the reliability and resources of the multi-industrial chain network according to the task reward, resource consumption and resource replenishment to carry out the next task.

[0164] (5) Determine whether a task in the task flow is completed; otherwise, return to step (1) to execute the next task. If completed, end the task allocation process and output the optimal allocation result for each task in the task flow.

[0165] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. A method for multi-industrial chain network task allocation based on dynamic resources and reliability of industrial agents, characterized in that, Comprising the following steps: S1. Multi-industrial chain network modeling: The model is based on the dynamic resources and reliability of industrial agents. Each layer of the multi-industrial chain network represents a connected, undirected weighted graph of multiple industrial chains. The l-th layer of the multi-industrial chain network is represented as N. l = <V l E l >, among which In network layer N l The set of industrial intelligent agent nodes mapped in the middle. For network layer N l The set of edges, where nc is the total number of industrial intelligent agents; industrial intelligent agents C i In network layer N l The reliability in is The reliability value will change adaptively as tasks are continuously assigned. The reliability of the same industrial intelligent agent will be different in different network layers, but they will affect each other. S2, target calculation: the target of the task allocation is to complete the optimal allocation of each task in the task flow to minimize the task evaluation function and minimize the total execution time of the task: wherein is the optimal industrial agent alliance solving task T k , and nt is the number of tasks in task set T; denotes the task evaluation function value of solving current task T k , denotes the execution time of solving current task T k . S3, dynamic reliability and resource adjustment: after the completion of the current task, the reliability of the industrial agent is dynamically adjusted through the reliability cascade effect model, and the resources of the industrial agent are dynamically adjusted through the resource contribution of the industrial agent to the task and the resource supplement of each task period; the reliability reward of the task T k in the reliability cascade effect model is defined as: Where sumR j Represents resource r in a multi-industry chain network j Total amount, tr k,j T represents k For resource r j Demand, Ratio k It is task T k The ratio of resource demand to total resources, nc l It is network N l The total number of industrial intelligent agents, μ is the weight of the reliability reward; S4, problem solving: using a problem context knowledge driven meta-heuristic method to solve the initial population of the task allocation problem, each individual in the population is a feasible task allocation scheme; in the solving process of the initial population, the performance of the industrial agent is evaluated by using the task allocation priority and the context task processing capacity, the task allocation priority is determined by the context reliability and the context resource satisfaction degree.

2. The method of claim 1, wherein the method further comprises: determining a reliability of each of the plurality of industrial agents; and determining a reliability of each of the plurality of industrial agents based on the reliability of each of the plurality of industrial agents. In the step S1, the multi-industrial chain network model further includes task alliance constraints, the industrial agents in the industrial agent alliance are interconnected in the multi-industrial chain network, and the resources provided by the industrial agent alliance meet the resource demand of the task; the connection includes intra-layer connection in the network and inter-layer connection between different mapping industrial agents of the industrial agent. 3.The method of claim 1, wherein the method further comprises: determining a reliability of each of the industrial agents; and determining a reliability of each of the industrial agents based on the reliability of each of the industrial agents. The task evaluation function F(CA in the step S2 is specifically: k ) is specifically: wherein β1 and β2 represent weights of reliability of industrial agent alliance and resource satisfaction degree of alliance for task T k respectively, ACol k is cooperation cost between industrial agents in industrial agent alliance for task T k k is the sum of reliability of industrial agents in the alliance for task T k k represents the sum of resource satisfaction degree of mapping industrial agents for task T k in the alliance.​​ 4. The method of claim 3, wherein the method further comprises: determining a reliability of each of the plurality of industrial agents; and determining a reliability of each of the plurality of industrial agents based on the reliability of each of the plurality of industrial agents. In the step S2, the alliance cooperation cost ACol k is calculated in the following way: The industrial agent alliance reliability ARep k The calculation method is as follows: The alliance resource satisfaction degree ASat k is calculated as follows: wherein, to map industrial agents to tasks T k resource satisfaction.

5. The method of claim 1, wherein the method further comprises: determining a reliability of each of the plurality of industrial agents; and determining a reliability of each of the plurality of industrial agents based on the reliability of each of the plurality of industrial agents. The execution time function Tmax of the task in step S2 k (CA k ) is specifically: where CA k To meet the task T k requirements of the alliance, nr is the number of resource types, TP k,j is the cumulative consumption time of the industrial agent alliance processing task T k required resources r j is calculated as follows: wherein, is a contribution of a unit resource r to the industrial agent alliance j the required processing time, is is a contribution of a resource r k to the task T j the amount.

6. The method of claim 1, wherein the method further comprises: determining a reliability of each of the plurality of industrial agents; and determining a reliability of each of the plurality of industrial agents based on the reliability of each of the plurality of industrial agents. In the reliability cascade effect model of step S3, according to the Industrial Intelligent Agents Consortium (CA) k Industrial intelligent agents for task T k Resource satisfaction will be rewarded with reliability bonuses. k Assigned to each alliance member, task T k After execution, the Industrial Intelligent Agents Alliance (CA) k Chinese Industrial Intelligent Body Reliability rewards obtained as follows: wherein, represents the members within the alliance satisfies the degree of resource, Reward k for the task T k is the reliability reward. Each alliance member transmits the reliability reward obtained through the intra-layer link to the neighbor industrial agent not in the alliance, the adjacent industrial agent obtains the corresponding reliability reward as follows: wherein, is a set of same-layer neighbor agents of the member within the coalition, γ l is a reliability influence factor of the network N l , is a communication cost of the member within the coalition with the same-layer neighbor ​ Each league member rewards reliability to other mapping industrial agents in the network, thus the industrial agents rewards reliability obtained across layers for: where CL i is the set of network layers that the agent C i belongs to, and γ l is the reliability influence factor of the network N l .

7. The method of claim 5, wherein the method further comprises: determining a reliability of each of the plurality of industrial agents; and determining a reliability of each of the plurality of industrial agents based on the reliability of each of the plurality of industrial agents. In the step S3, during the execution of the current task, the agent contributes to the task T k the amount of resources r j is: ​ wherein, represents the amount of resource r k that has joined the coalition, is the amount of resource r owned by the agent j ; and k,j is the amount of resource r k required for task T j . After the current task is completed, the agent The dynamic adjustment of resources is as follows: wherein, is an agent having resources r j in an amount, is an industrial agent for a task T k contributing resources r j in an amount, is an industrial agent supplementing resources r j in an amount.

8. The method of claim 1, wherein the method further comprises: determining a reliability of each of the plurality of industrial agents; and determining a reliability of each of the plurality of industrial agents based on the determined reliability of each of the plurality of industrial agents. In the step S4, the calculation method of the context reliability in the task allocation priority is: wherein, represents a mapped industrial agent context reliability in a multi-industrial chain network, CL i is the context reliability in the network N i corresponding to the industrial agent C the network layer index set where the mapped industrial agent represents a mapped industrial agent context reliability in the network N l corresponding to the mapped industrial agent context reliability in the network N l context reliability in the network N is defined as follows: wherein, represents the minimum cooperation cost between the industrial agent and the same-layer neighbor γ l is the reliability influence factor of the network N l ; The calculation method of the context resource satisfaction degree in the task allocation priority is: wherein, represents an industrial agent satisfies the context resource of the task T k , CN i is the industrial agent C i corresponding to the adjacent industrial agent set, and α represents the collaborative resource influence factor, CV i′ is the industrial agent C i′ corresponding to the mapping industrial agent set; Industrial agent To task T k The task assignment priority for task T is calculated as follows: wherein β1 and β2 represent the weights of the industrial agent alliance reliability and the alliance resource satisfaction degree assigned to the task T k respectively. The industrial agent alliance reliability and the alliance resource satisfaction degree assigned to the task T The calculation method of the context task processing capacity is as follows: Where ω is the collaborative resource processing factor, To possess resources j Quantity, And not equal to 0; Represents mapping industrial intelligent agents The task processing capability is calculated as follows: 9.The method of claim 7, wherein the method further comprises: determining a reliability of each of the industrial agents; and determining a reliability of each of the industrial agents based on the reliability of each of the industrial agents. After the initial population is generated in the step S4, each feasible task allocation scheme in the population is disturbed by an operation operator, the operation operator includes but is not limited to a destruction and reconstruction operation based on the context reliability, a destruction and reconstruction operation based on the context resource satisfaction degree, a destruction and reconstruction operation based on the task allocation priority, and a destruction and reconstruction operation based on the context task processing capacity.

Images