Complex task collaborative twinborn deduction method based on virtual-real fusion

Abstract

The invention discloses a virtual-real fusion-based complex task collaborative twinborn deduction method, which comprises the following steps of: firstly, establishing a multi-dimensional collaborative relation model between a task and equipment, and constructing a scheduling modeling basis from three aspects of equipment topology, resource mapping and task dependence; secondly, constructing a task collaborative deduction and optimization model based on a dynamic game, simulating a state evolution process of multiple tasks in a virtual space, and performing path optimization by taking a system efficiency function as a target; and finally, fusing an annealing mechanism and a virtual gradient perturbation design collaborative strategy optimization method, guiding a path adjustment direction based on efficiency feedback, and realizing multi-round iterative evolution of a task strategy. According to the method, the global optimization capability of the scheduling strategy and the stability of system operation are effectively improved, and the method has wide application prospects in automobile part flexible production task scheduling and collaborative optimization scenes.

Description

Technical Field

[0001] The present invention belongs to the field of digital twin and complex task system scheduling optimization, and specifically relates to a technical solution based on the digital twin concept, which realizes the collaborative scheduling of complex tasks in a flexible manufacturing environment through virtual-reality fusion modeling, game deduction and strategy optimization methods. Background Art

[0002] Digital twinning is a technology that dynamically maps and interacts with physical systems through digital models, enabling high-precision simulation of their structure, behavior, and operational status in virtual space. This technology enables simultaneous deduction and predictive analysis of actual systems, demonstrating powerful modeling and optimization capabilities in a variety of fields, including industrial manufacturing, energy management, and urban operations. In the area of collaborative scheduling and optimization of complex tasks, digital twins can map multiple elements of a production system, such as tasks, equipment, and resources, into virtual space, constructing a dynamic evolution model to support the deduction and performance evaluation of scheduling paths.

[0003] Traditional methods for collaborative scheduling of complex tasks often rely on static modeling and heuristic optimization, which have many limitations. First, the modeling dimension is single: traditional methods often simplify the relationship between tasks and resources, making it difficult to depict the dependencies between tasks and the topological coupling between devices in real systems. Second, lack of adaptability: facing dynamic changes in the production environment, such as task insertion and resource fluctuations, traditional scheduling algorithms lack a response mechanism, resulting in rigid scheduling paths and high adjustment costs. Third, the optimization strategy is highly localized, making it difficult to find the global optimal solution in the complex game of resource conflicts and multi-task collaboration. It is easy to fall into local optimality and affect overall performance.

[0004] In contrast, a Twin-Based Collaborative Task Deduction Method for Complex Systems Integrating Physical and Virtual Spaces (TCDPV) leverages the advantages of digital twin technology in virtual deduction and real-time mapping. It can dynamically reconstruct the relationship between tasks and resources and combine strategy evolution with path feedback mechanisms for multiple rounds of iterative optimization. This method not only effectively captures state changes and conflict risks during task collaboration but also enables global strategy adjustment capabilities, improving the intelligence and optimization performance of scheduling systems in scenarios involving multi-task parallelism, resource reconfiguration, and scheduling dynamics. Summary of the Invention

[0005] Purpose of the invention: In response to the problems of complex task scheduling constraints, frequent resource conflicts, and low collaborative efficiency in complex flexible production systems, the present invention provides a complex task collaborative twin deduction method based on virtual-real fusion. The present invention is based on the idea of digital twins and effectively improves the overall performance of collaborative scheduling between tasks by constructing a virtual-real fusion task system mapping model, solving the problem of insufficient scheduling optimization capabilities of traditional methods in scenarios with strong dynamics and high coupling.

[0006] Technical solution: To achieve the above purpose, the technical solution adopted by the present invention is: A complex task collaborative twin deduction method based on virtual-reality fusion includes the following steps: Step 1: Obtain the task and equipment information of the flexible production system of automobile parts, and build a multi-dimensional collaborative relationship model of tasks and equipment in the flexible production system of automobile parts.

[0007] Step 2: Construct a complex task collaborative deduction model based on the multi-dimensional collaborative relationship model, deduce the state evolution process of multiple tasks under resource constraints and collaborative relationships, and form a dynamic evolution path for task scheduling.

[0008] Step 3: Build a complex task collaborative optimization model based on the dynamic evolution path, concretize the system performance into an efficiency function including maximum completion time, total energy consumption and task benefit, and perform performance evaluation and optimization selection on the task path under multiple types of scheduling constraints.

[0009] Step 4: Based on the optimized task path, a collaborative strategy optimization method that integrates the annealing mechanism and the virtual gradient is designed. The strategy perturbation direction is constructed based on the efficiency feedback of the current path, and the annealing mechanism is introduced to adjust the perturbation intensity and acceptance probability to achieve multi-round iterative optimization of the task collaborative path.

[0010] Preferred: The complex task collaborative optimization model constructed in step 3 is as follows: stX(t+1)=F(X(t),U(t)),t=0,1,...,N-1 Among them, E represents system efficiency, P represents deduction path, and E total represents production efficiency, T max represents the maximum completion time, Represents each task T j The execution progress at time t, M represents the total number of devices, A ij (t) represents task T j Is device D occupied at time t? i , represents the state of the device at time t, Ctotal represents the total production energy consumption, T represents the task set, N represents the number of tasks, Represents task T i , F represents the system evolution function, X represents the system state space set, U represents the strategy space set, C task-task represents the task-task constraint. task-device Represents the task-device constraint. C device-device Represents a task-task constraint.

[0011] Preferred: The method for constructing the complex task collaborative optimization model in step 3 is as follows: For the measurement of the maximum completion time, considering the actual execution time of all tasks in the task set T, the following specific expression is defined: Among them, T max represents the maximum completion time, Represents each task T j The execution progress at time t, M represents the total number of devices, A ij (t) represents task T j Is device D occupied at time t? i , Represents the state of the device at time t.

[0012] For the total production energy consumption, under the premise of knowing a specific deduction path, it can be expressed as follows: Among them, C total represents the total production energy consumption, T represents the task set, and N represents the number of tasks. Represents task T i implementation decisions.

[0013] By combining the system state evolution law and scheduling constraints, a complex task collaborative optimization model can eventually be constructed.

[0014] Optimal: Step 3 is to obtain the next round of deduction path P (k+1) , a new initial state X needs to be constructed (k+1) (0). With the help of the change information of path efficiency, a virtual gradient perturbation model based on the target value difference is constructed, which is specifically expressed as follows: Among them, λ (k) is the perturbation step size control factor, E (k) represents the k-th deduction path system efficiency, virtual represents the virtual gradient, is the virtual gradient direction constructed based on efficiency feedback. In actual calculation, considering the non-differentiable nature of the efficiency function, this term can be approximated by difference: where sign(·) is a sign function that indicates the direction of change in the efficiency of the current path.

[0015] Therefore, the new initial state update formula is expressed as: X (l+1) (0) = X (k) (0)+λ (k) ·sign(E (k) -E (k-1) ·(X (k) (0)-X (k-1) (0)) Among them, X (k+1) represents the starting state of the k+1th deduction, X (k) represents the starting state of the kth deduction, X (k-1) represents the starting state of the k-1th deduction, E (k) represents the efficiency of the k-th deduction path system, E (k-1) Indicates the efficiency of the k-1th deduction path system.

[0016] Preferably: the collaborative strategy optimization method integrating the annealing mechanism and the virtual gradient in step 4 includes the following steps: assuming that the current k-th deduction path is P (k) , whose initial state is X (k) (0), the corresponding efficiency is E (k) According to the virtual gradient concept, the direction guidance item is constructed by the state difference and efficiency change between historical paths: ΔX (k) =sign(E (k) -E (k-1) )·(X (k) (0)-X (k-1) (0)) Where ΔX (k) Indicates a direction guide item.

[0017] Consider introducing the perturbation control function in the annealing mechanism, combining the direction guidance and annealing perturbation to construct the update formula of the initial state of the path: X (k+1) (0) = X (k) (0)+λ (k) ΔX (k) +ε (k) Among them, ε (k) is the disturbance term, and the statistical characteristics of the disturbance term are related to the current temperature T (k) Related to, satisfying: Where N represents the normal distribution, represents the variance of the disturbance term, β>0 is the disturbance amplitude control parameter, T (k) Indicates the current temperature.

[0018] Newly generated path P (k+1) It is generated through the collaborative deduction model of complex tasks, and its performance value E is calculated by the efficiency function (k+1) To control the path acceptance strategy, the Metropolis criterion in simulated annealing is introduced to determine the acceptance probability of the current path based on whether the efficiency is improved or decreased: in, Indicates the acceptance probability of the current path.

[0019] The temperature is updated according to an exponential decay strategy: T (k+1) =α·T (k) , 0<α<1 During the multi-round path update process, the following evolutionary strategy is used to find the most efficient path: Among them, P * Indicates the optimal deduction path obtained by the final convergence, and the initial state X corresponding to this path * (0), the strategy sequence {U(t)}, the state sequence {X(t)} and the task-equipment collaborative relationship all achieve the optimal efficiency configuration, constituting the optimal scheduling execution scheme in the flexible production system.

[0020] Preferably, the method for constructing a multi-dimensional collaborative relationship model between tasks and equipment in the flexible production system of automobile parts in step 1 comprises the following steps: Step 11: Modeling device-device relationships In the flexible manufacturing process of automotive parts, the equipment types in the flexible production system include CNC machine tools, automatic testing equipment, robotic arms, logistics conveyor lines, and material handling robots. The equipment set in the flexible production process is defined as follows: V={D1,D2,...,D i ,...,D M} Among them, V represents the set of devices in the system, D i represents the i-th device, and M represents the total number of devices.

[0021] In flexible production systems, the interactions between devices can be unidirectional, bidirectional, or non-unidirectional. The following uses directed edges to uniformly describe this feature and define the edge set E: If (D i , D j )∈E, it means there is a slave device D i Flow to device D j For a pair of devices that have both bidirectional connections, the set E can contain (D i , D j ) and (D j , D i ).

[0022] If you only need to determine whether the connection is feasible, you can use the adjacency matrix A = [a ij ] M×M Represents a binary relationship: Define the capacity matrix C: C=[c ij ] M×M in, When c ij >0 and c ji =0 indicates a one-way connection. ij >0 and c ji > 0 indicates a bidirectional connection. If both are 0, there is no connection.

[0023] For the establishment of topology and directed graph, based on the equipment set and connectivity definition, the flexible production system as a whole is represented as a triple: G=(V,E,C) Where G is the triplet representation of the flexible production system, V is the set of devices in the system, E is the edge set, which is used to represent the feasible transmission directions between devices, and C is the capacity matrix.

[0024] The following direction function is defined to distinguish the connection direction between devices: Wherein, dir(i, j) represents the direction function.

[0025] Step 12: Task-device relationship modeling First, we analyze the correlation between task sets and devices. Assume that there are N tasks in the system, which constitute the task set T: T={T1,T2,...,T N} Each task T k The device association degree is defined as the normalized weight w of the number of devices required for the task k : satisfy For task T k The desired collection of device nodes.

[0026] Task T k The device connection relationship can be modeled as a weighted directed graph G k =(D k ,E k ,W k ), is the set of directed edges between devices. is the edge weight matrix: Formula α ij Representation device arrive The benchmark transmission efficiency. t is the task execution time, β, γ are time decay parameters. ij is the physical distance between devices, and λ is the spatial attenuation coefficient.

[0027] Task T k The execution path must satisfy the following constraints: First, the topological connectivity, i.e., the task subgraph G k There must be at least one directed path from the starting point to the end point for any device node in . Second, the capacity constraint is that the amount of parallel transmission between devices does not exceed the limit of the global capacity matrix C, which is expressed as follows: Where P is the number of parallel transmission channels, C k (i, j, p) is task T k The instantaneous capacity on the p-th channel.

[0028] Device D i The resource occupancy rate at time t must satisfy: Among them, ρ i (t) is the device D i The resource occupancy rate at time t, ò is the nonlinear damping coefficient, ρ max The maximum load threshold of the device.

[0029] Step 13: Task-task relationship modeling. There are N kinds of tasks in the system, and their relationships are expressed by the fourth-order tensor R∈R N ×N×M×M Indicates that the tensor element R(i, j, p, q) describes the task t i With T j In device D p to D q The interaction strength on a link is defined as: Where D i ∩D j For task T i With T j σ is the difference tolerance coefficient, θ is the nonlinear adjustment parameter, C k (i, j, q) is task T k The instantaneous capacity on the qth channel.

[0030] The overall conflict intensity between tasks is expressed by the matrix K∈R N×N Indicates that its element K(i, j) is calculated as: Where: A i For task T i The adjacency matrix, A j For task T j The adjacency matrix of .

[0031] The collaborative potential between tasks is expressed by the three-dimensional matrix S∈R N×N×L Description, its elements S(i, j, t) are defined as: Where L is the number of time windows, W i (p,·) is the task T i The output weight, W j (·, p) is the task T j The input weight, W i (p,·)·W j (·, p) is the task T i The output weights and T j The dot product of the input weights.

[0032] The final relationship between tasks can be expressed by the decision matrix F(i, j) that integrates conflict and collaboration: Among them, F(i, j) is the decision matrix. When F(i, j)>0, the task Ti and T j It mainly manifests as resource competition. When F(i, j)<0, task T i With T j Possesses potential for collaborative optimization.

[0033] Preferably: Step 2 of the method for constructing a complex task collaborative deduction model includes the following steps: Step 211, state space equipment state: The resources of the flexible production system are various types of equipment. For any equipment D i ∈V, the state of the device at time t Defined as: The overall device status is described as follows: Task execution status: In the flexible production system, production tasks constitute the decision-making subject in the dynamic game. The task set is T = {T1, T 2, ..., T N Each task T j The execution progress at time t is represented by a continuous value: The overall status of all tasks is expressed as: Task-device association status: Task execution is inseparable from device resources. Define any task T j The state of association with the device at time t is: The overall relationship between all tasks and equipment occupancy is expressed in matrix form as follows: A T，V (t)=[A j，i (t)] M×N Then the overall system state space can be uniformly represented as a set of triples: X(t)={S V (t), S T (t), A T，V (t)} Among them, the state space X(t) provides a comprehensive real-time information basis for strategy formulation and interactive decision-making of each task in the dynamic game.

[0034] Step 212, strategy space Equipment occupancy strategy: Each task is a decision-making subject in the dynamic game. At time t, any task T j For device D i The occupancy decision is defined as: The overall equipment occupancy strategy matrix is expressed as: U V (t)=[U j，t (t)] M×N Task execution strategy: Each task is an independent decision-making entity. At any time t, task T j The execution decision of is defined as: The overall vector of all task execution strategies is: Then, the strategy space can be uniformly expressed as: U(t)={U V (t), U T (t)} Preferably, the method for constructing a complex task collaborative deduction model in step 2 includes an effectiveness evaluation index, and the effectiveness evaluation index method is as follows: Starting from the maximum completion time, total production energy consumption and production efficiency, a dynamic path-dependent performance evaluation system is constructed, and the system performance index E is defined accordingly.

[0035] For the maximum completion time T max In the dynamic collaborative scheduling process, the task completion time is not fixed, but changes dynamically with the task execution order, device parallelism, and resource conflicts, with the following relationship: Among them, F T (·) is the time aggregation map.

[0036] For the total production energy consumption C total , the total production energy consumption can satisfy the following relationship: Among them, F C (·) is the production energy consumption aggregation function.

[0037] For production efficiency ε total The set of production tasks T is predefined, with clear objectives and fixed boundaries. The production benefits that each task can generate are fixed values: Among them, ε j For task T j The corresponding fixed benefit value.

[0038] Then, the system performance indicators have the following relationship: Among them, E is the system efficiency index.

[0039] Preferably: The state evolution process of multiple tasks under resource constraints and collaborative relationships in step 2 includes the following steps: defining the main point intersection It is used to describe the structural conflict intensity when two tasks have overlapping key equipment during execution. For any two tasks x1, x2, if their principal points intersect to form an empty set, that is, This indicates that the sets of devices used by the two tasks at any given moment do not overlap. If the principal point intersection is at a particular principal step point, k, then the other task can only run in parallel with the benchmark task after it has completed step k. During game scheduling, by adjusting the execution order and starting offsets of tasks, the principal point intersection structure can be optimized, enabling multiple tasks to transition from serial to parallel execution as quickly as possible, improving overall scheduling resource utilization and system performance.

[0040] Based on this, the game equilibrium and strategy update mechanism are unified and abstracted into the following strategy update formula, which is used to describe the collaborative strategy evolution process of the task set under the iteration round r: Among them, U (r+1) represents the joint strategy set of round r+1. E(U) represents the system efficiency function of the task set under the current strategy combination U. Represents task x i with x j The principal point intersection conflict cost under the current policy. represents the policy perturbation degree. λ1 and λ2 represent control parameters.

[0041] The entire deduction output is shown below: sU (r) Reach game equilibrium and maximize E; Among them, X(T * ) represents the set of system state spaces at the time of reaching the optimal deduction path, T * Indicates the moment when the optimal deduction path is reached, P * represents the best deduction path, F represents the system evolution function, U (r) (t) represents the set of joint strategies in round r at time t.

[0042] Another object of the present invention is to provide a complex task collaborative twin deduction system based on virtual-reality fusion, which is used to implement the complex task collaborative twin deduction method based on virtual-reality fusion, including an input unit, a multi-dimensional collaborative relationship model unit, a complex task collaborative deduction model unit, a complex task collaborative optimization model unit, and a collaborative strategy optimization unit, wherein: the input unit is used to input the task and equipment information of the flexible production system of automotive parts.

[0043] The multidimensional collaborative relationship model unit is used to construct a multidimensional collaborative relationship model of tasks and equipment in the flexible production system of automobile parts based on the tasks and equipment information of the flexible production system of automobile parts.

[0044] The complex task collaborative deduction model unit is used to construct a complex task collaborative deduction model based on a multi-dimensional collaborative relationship model, deduce the state evolution process of multiple tasks under resource constraints and collaborative relationships, and form a dynamic evolution path for task scheduling.

[0045] The complex task collaborative optimization model unit is used to construct a complex task collaborative optimization model based on a dynamic evolution path, concretize the system performance into an efficiency function including maximum completion time, total energy consumption and task benefit, and perform performance evaluation and optimization selection on the task path under multiple types of scheduling constraints.

[0046] The collaborative strategy optimization unit is used to design a collaborative strategy optimization method that integrates an annealing mechanism and a virtual gradient according to the optimized selected task path, construct a strategy perturbation direction based on the efficiency feedback of the current path, and introduce an annealing mechanism to adjust the perturbation intensity and acceptance probability, thereby realizing multi-round iterative optimization of the task collaborative path.

[0047] Compared with the prior art, the present invention has the following beneficial effects: First, by constructing a multi-dimensional collaborative relationship model between tasks and equipment, the present invention can fully reflect the resource topology and scheduling constraints in the flexible production system, thereby improving the accuracy of task modeling and the integrity of the scheduling basis.

[0048] Second, the present invention introduces dynamic game and virtual gradient annealing optimization mechanism to realize the deduction and strategy evolution of multi-task collaborative paths, which can achieve efficient optimization of scheduling strategies in complex scheduling scenarios and significantly improve the overall operating efficiency and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is the overall flow chart of the complex task collaborative twin deduction method (TCDPV) based on virtual-reality fusion.

[0050] Figure 2 This is the scheduling result diagram of the TCDPV method under normal circumstances.

[0051] Figure 3 Iterative curve of maximum completion time and comprehensive energy consumption of TCDPV under normal scheduling.

[0052] Figure 4 This is the scheduling result diagram of the TCDPV method under the condition of key equipment failure.

[0053] Figure 5 Iterative curve of maximum completion time and comprehensive energy consumption of TCDPV method under the condition of key equipment failure.

[0054] Figure 6 The following is a comparison chart of the maximum completion time and energy consumption performance of each algorithm in two scenarios. DETAILED DESCRIPTION

[0055] The present invention is further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, modifications of various equivalent forms of the present invention made by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0056] In order to solve the problem of complex task scheduling constraints and difficult collaborative optimization in flexible production systems, this embodiment provides a complex task collaborative twin deduction method based on virtual-real fusion, and constructs a virtual-real mapping model based on the idea of digital twins to achieve dynamic correspondence and synchronous deduction between production tasks and virtual scheduling environments. First, a multi-dimensional collaborative relationship model between tasks and equipment is established, and a scheduling modeling foundation is constructed from three aspects: equipment topology, resource mapping, and task dependency; secondly, a task collaborative deduction and optimization model based on dynamic game is constructed to simulate the state evolution process of multiple tasks in a virtual space, and perform path optimization with the system efficiency function as the goal; finally, a collaborative strategy optimization method is designed by integrating the annealing mechanism and virtual gradient perturbation, and the path adjustment direction is guided based on efficiency feedback to achieve multiple rounds of iterative evolution of task strategies. Figure 1 As shown, the specific steps include: Step 1: Obtain the task and equipment information of the flexible production system of automobile parts, and build a multidimensional collaborative relationship model of tasks and equipment in the flexible production system of automobile parts. The multidimensional collaborative relationship model includes the equipment topology structure, the mapping relationship between tasks and resources, and the dependency constraints between tasks, forming the basic structure of task scheduling modeling.

[0057] Step 2: Construct a complex task collaborative deduction model based on the multi-dimensional collaborative relationship model, define the system's state space, strategy space, and performance evaluation indicators, deduce the state evolution process of multiple tasks under resource constraints and collaborative relationships, and form a dynamic evolution path for task scheduling.

[0058] Step 3: Based on the dynamic evolution path, a complex task collaborative optimization model is constructed to concretize the system performance into an efficiency function including maximum completion time, total energy consumption and task benefit, and the task path is evaluated and optimized under multiple scheduling constraints.

[0059] Step 4: Design a collaborative strategy optimization method that integrates the annealing mechanism and virtual gradient based on the optimized task path. Construct the strategy perturbation direction based on the efficiency feedback of the current path, and introduce the annealing mechanism to adjust the perturbation intensity and acceptance probability, so as to realize multi-round iterative optimization of the task collaborative path and improve the global performance and convergence stability of the scheduling strategy.

[0060] In step 1, the process of building a multidimensional collaborative relationship model includes the following steps: Step 11: Modeling device-device relationships In the flexible manufacturing process of automotive parts, the production system usually includes multiple types of key equipment to complete tasks such as processing, testing, handling and transportation of parts. Equipment types include CNC machine tools, automatic testing equipment, robotic arms, logistics conveyor lines, and material handling robots. Different types of equipment have diverse connection and interaction modes, which are the basis for achieving task collaboration and dynamic resource scheduling. The equipment collection in the flexible production process can be defined as follows: V={D1,D2,...,D M} This formula represents the set of devices in the system, where M represents the total number of devices. i It can represent a specific physical device or a pool of devices with the same functions.

[0061] In flexible production systems, the interactions between devices can be unidirectional, bidirectional, or non-interactive. This feature is described uniformly using directed edges, defining the edge set E: If (D i , D j )∈E, it means there is a slave device D i Flow to device D j For a pair of devices that have both bidirectional connections, the set E can contain (D i , D j ) and (D j , D i ).

[0062] If you only need to determine whether the connection is feasible, you can use the adjacency matrix A = [a ij ] M×M Represents a binary relationship: However, in flexible production systems, it is often necessary to distinguish between differences in parallelism or channel capacity. i It may be possible to have the ability to transfer materials to multiple subsequent devices in parallel. For this purpose, a capacity matrix can be defined: C=[c ij ] M×M in, When c ij >0 and c ji =0 indicates a one-way connection. ij >0 and c ji> 0 indicates a bidirectional connection. If both are 0, there is no connection.

[0063] For the establishment of topology and directed graph, based on the equipment set and connectivity definition, the unmanned flexible production system can be represented as a triple: G=(V,E,C) Where V represents the set of nodes (devices). E is the set of directed edges, which is used to represent the feasible transmission directions between devices. C is the capacity matrix (or channel capacity function), which is denoted as C = [c ij ] M×M , used to quantify the slave device D i To device D j The maximum degree of parallelism or number of channels.

[0064] To further reflect the difference between the connection direction and the serial constraint and bidirectional feasibility, the following direction function can be defined to distinguish the connection direction between devices: Step 12: Task-device relationship modeling In order to accurately understand the equipment requirements of each task, we first need to analyze the correlation between the task set and the equipment. Assume that there are N types of tasks in the system, which constitute the task set: T={T1,T2,...,T N} Each task T k The device relevance is defined as the normalized weight of the number of devices required for the task: The weight reflects the proportion of system device resources occupied by the task, satisfying For task T k The desired collection of device nodes.

[0065] Task T k The device connection relationship can be modeled as a weighted directed graph G k =(D k , E k , W k ).in, is the set of directed edges between devices. is the edge weight matrix: Formula α ij Representation device arrive The benchmark transmission efficiency. t is the task execution time, β, γ are time decay parameters. ij is the physical distance between devices, and λ is the spatial attenuation coefficient.

[0066] Task T k The execution path must satisfy the following constraints: First, the topological connectivity, i.e., the task subgraph G k There must be at least one directed path from the starting point to the end point for any device node in . Second, the capacity constraint is that the amount of parallel transmission between devices does not exceed the limit of the global capacity matrix C, which is expressed as follows: Where P is the number of parallel transmission channels, C k (i,j,p) is task T k The instantaneous capacity on the p-th channel.

[0067] Device D i The resource occupancy rate at time t must satisfy: Where ò is the nonlinear damping coefficient, ρ max is the maximum load threshold of the device. This constraint ensures the rationality of system resource scheduling and the stability of device operation under concurrent execution of multiple tasks.

[0068] Step 13: Task-task relationship modeling Assume that there are N kinds of tasks in the system, and their relationships can be expressed through a fourth-order tensor Represented by, where M is the total number of global devices. The tensor element R(i, j, p, q) describes the task T i With T j In device D p to D q The interaction strength on a link is defined as: Where D i ∩D j For task T i With T j σ is the difference tolerance coefficient, which adjusts the overall sensitivity of the interaction strength of shared devices, and θ is the nonlinear adjustment parameter.

[0069] The overall conflict intensity between tasks can be expressed by the matrix Indicates that its element k(i, j) is calculated as: Where: A i For task T i The adjacency matrix of molecule A i (p, q)A j (q, p) captures bidirectional path conflicts such as T i Requirement D p →D q , and T j Requirement D q→D p ).

[0070] The synergy potential between tasks is expressed through a three-dimensional matrix Description (L is the number of time windows), its element S(i, j, t) is defined as: Where: W i (p,·)·W j (·, p) is the task T i The output weights and T j The dot product of the input weights reflects the device D p supply and demand matching.

[0071] The final relationship between tasks can be expressed by the decision matrix F(i, j) that integrates conflict and collaboration: The first term is the normalized conflict term, and the second term is the normalized coordination term. When F(i, j)>0, task T i With T j It mainly manifests as resource competition. When F(i, j)<0, task T i With T j Possesses potential for collaborative optimization.

[0072] In step 2, to build a complex task collaborative deduction model, it is necessary to define the state space, strategy space and effectiveness evaluation indicators. The specific definitions are as follows: Step 21, State Space Equipment status: The resources of the flexible production system are mainly various types of equipment. For any equipment D i ∈V, the state is defined as: The overall device status is described as: Task execution status: The production tasks in the system constitute the decision-making body in the dynamic game. The task set is T = {T1, T2, ..., T N}. Each task T j The execution progress at time t is represented by a continuous value: The overall status of all tasks is expressed as: Task-device association status: Task execution is inseparable from device resources. Define any task T j The state of association with the device at time t is: The overall relationship between all tasks and equipment occupancy is expressed in matrix form as follows: A T，V (t)=[A j，i (t)] M×N Therefore, through the state definitions of the above three aspects, the overall system state space can be uniformly represented as a set of triples: X(t)={S V (t), S T (t), A T，V (t)} The state space X(t) provides a comprehensive real-time information basis for strategy formulation and interactive decision-making of each task in the dynamic game.

[0073] Step 22, strategy space Equipment occupancy strategy: Each task is a decision-making subject in the dynamic game. At time t, any task T j For device D i The occupancy decision is defined as: The overall equipment occupancy strategy matrix is expressed as: U V (t)=[U j，i (t)] M×N Task execution strategy: Each task is an independent decision-making entity. At any time t, task T j The execution decision of is defined as: The overall vector of all task execution strategies is: In summary, the strategy space can be uniformly expressed as: U(t)={U V (t), U T (t)} Step 23: Performance evaluation indicators In a complex task collaborative deduction model, the quality of the path directly determines the system's operational efficiency. Therefore, this method constructs a dynamic path-dependent performance evaluation system based on three aspects: maximum completion time, total production energy consumption, and production efficiency. This system's performance indicator, E, is defined accordingly.

[0074] For the maximum completion time T max In the dynamic collaborative scheduling process, the task completion time is not fixed, but changes dynamically with factors such as task execution order, device parallelism, resource conflicts, etc., with the following relationship: Among them FT (·) is the time aggregation map.

[0075] For the total production energy consumption C total , the system energy consumption mainly comes from the energy expenditure caused by resource consumption, equipment start-up and shutdown, path switching and coordinated scheduling during task execution. The total production energy consumption can satisfy the following relationship: Among them F C (·) is the production energy consumption aggregation function.

[0076] For production efficiency total The set of production tasks T is predefined, with clear objectives and fixed boundaries. The production benefits that each task can generate are fixed values: where ε j For task T j The corresponding fixed benefit value.

[0077] Based on the above analysis, the system performance index E shows the following relationship: The complex task collaborative deduction model constructed in step 2 can deduce the state evolution process of multiple tasks under resource constraints and collaborative relationships, forming a dynamic evolution path for task scheduling. The specific process is as follows: In order to characterize the execution structure relationship between tasks, the main point intersection is defined It is used to describe the structural conflict intensity when two tasks have overlapping key equipment during execution. For any two tasks x1, x2, if their principal points intersect to form an empty set (i.e. ), indicating that the sets of devices used by the two tasks at any given moment do not overlap. If the principal point intersection is at a particular principal step point k, then the other task can only run in parallel after the benchmark task has executed to step k. During game scheduling, by adjusting the execution order and starting offset times of tasks, the principal point intersection structure can be optimized, enabling multiple tasks to transition from serial to parallel execution as quickly as possible, improving overall scheduling resource utilization and system performance.

[0078] Based on this, the game equilibrium and strategy update mechanism are unified and abstracted into the following strategy update formula, which is used to describe the collaborative strategy evolution process of the task set under the iteration round r: Among them U (r+1) represents the joint strategy set of round r+1. E(U) represents the system efficiency function of the task set under the current strategy combination U. Represents task x i with x jThe principal point intersection conflict cost under the current policy. represents the policy perturbation, which is used to control the policy convergence speed. λ1 and λ2 are control parameters, which are used to balance the game objectives between efficiency, conflict and policy stability.

[0079] The entire deduction output is shown below: sU (r) The game equilibrium is reached and E is maximized. In step 3, the specific process of constructing the complex task collaborative optimization model is as follows.

[0080] In the construction of a collaborative optimization model for complex tasks, in order to achieve quantitative optimization of system efficiency, it is necessary to base it on a known specific deduction path, relying on the clear task state evolution and resource allocation information in the path, and concretize the abstract indicators in step 2 into a clearly structured and computable optimization objective function.

[0081] For the measurement of the maximum completion time, the actual execution time of all tasks in the task set T is considered. The specific expression is as follows: Among them, A ij (t) represents task T j Whether device D is occupied at time t i .

[0082] For the total production energy consumption, under the premise of knowing a specific deduction path, it can be expressed as follows: in, Represents task T i implementation decisions.

[0083] By combining the system state evolution law and scheduling constraints, the following complex task collaborative optimization model can be constructed: stX(t+1)=F(X(t),U(t)),t=0,1,...,N-1 Among them C task-task represents the task-task constraint. task-device Represents the task-device constraint. C device-device Represents a task-task constraint.

[0084] On this basis, in order to obtain the next round of deduction path P (k+1) , a new initial state X needs to be constructed (k+1)(0). This method uses the change information of path efficiency to construct a virtual gradient perturbation model based on the target value difference, which is specifically expressed as follows: Among them, λ (k) is the perturbation step size control factor, is the virtual gradient direction constructed based on efficiency feedback. In actual calculations, considering the non-differentiable nature of the efficiency function, this term can be approximated by difference: where sign(·) is a sign function that indicates the direction of change in the efficiency of the current path.

[0085] Therefore, the new initial state update formula can be expressed as: X (k+1) (0) = X (k) (0)+λ (k) ·sign(E (k )-E (k-1) )·(X (k) (0)-X (k-1 )(0)) In this formula, if the current path efficiency E (k) Higher than the previous round of path efficiency E (k-1) , it indicates that the current disturbance direction has a lifting effect.

[0086] In step 4, this method integrates the annealing mechanism and the collaborative strategy optimization method of virtual gradient. Based on the efficiency feedback of the current path, the strategy perturbation direction is constructed, and the annealing mechanism is introduced to adjust the perturbation intensity and acceptance probability, thereby realizing multi-round iterative optimization of the task collaborative path. The method process is as follows: Let the current k-th deduction path be P (k) , whose initial state is X (k) (0), the corresponding efficiency is E (k) According to the virtual gradient concept, the direction guidance item is constructed by the state difference and efficiency change between historical paths: ΔX (k) =sign(E (k) -E (k-1) )·(X (k) (0)-X (k-1) (0)) Consider introducing the perturbation control function in the annealing mechanism, combining the direction guidance and annealing perturbation to construct the update formula of the initial state of the path: X (k+1) (0) = X (k) (0)+λ (k) ΔX (k) +ε (k) Among them, ε(k) is a disturbance term, and its statistical characteristics are related to the current temperature T (k) Related to, satisfying: Where β>0 is the perturbation amplitude control parameter. This construction method makes the perturbation intensity gradually converge with temperature, thus making the search process gradually transition from the initial large-scale exploration to the later fine convergence.

[0087] Newly generated path P (k+1) It is generated through the task collaborative deduction model, and its performance value E is calculated by the efficiency function (k+1) To control the path acceptance strategy, the algorithm introduces the Metropolis criterion in simulated annealing, which determines the acceptance probability of the current path based on whether the efficiency is improved or decreased: The temperature is updated according to an exponential decay strategy: T (k+1) =α·T (k) , 0<α<1 Through the above mechanism, the system always accepts a better path when efficiency improves, and retains a certain probability of accepting a suboptimal solution when efficiency decreases. Even if the new path performs slightly worse in the current round than in the previous round, there is a certain probability of being accepted. This makes it perturbation-resistant and fault-tolerant, preventing the optimization process from being constrained by early local optimal points.

[0088] During the multi-round path update process, the entire algorithm uses the following evolutionary strategies to find the most efficient path: Among them, P * Indicates the optimal deduction path obtained by the final convergence, and the initial state X corresponding to this path * (0), the strategy sequence {U(t)}, the state sequence {X(t)} and the task-equipment collaborative relationship all achieve the optimal efficiency configuration, constituting the optimal scheduling execution scheme in the flexible production system.

[0089] In another embodiment of the present invention, a complex task collaborative twin deduction system based on virtual-reality fusion is provided, which is used to implement the complex task collaborative twin deduction method based on virtual-reality fusion, including an input unit, a multi-dimensional collaborative relationship model unit, a complex task collaborative deduction model unit, a complex task collaborative optimization model unit, and a collaborative strategy optimization unit, wherein: the input unit is used to input task and equipment information of the flexible production system of automotive parts.

[0090] The multidimensional collaborative relationship model unit is used to construct a multidimensional collaborative relationship model of tasks and equipment in the flexible production system of automobile parts based on the tasks and equipment information of the flexible production system of automobile parts.

[0091] The complex task collaborative deduction model unit is used to construct a complex task collaborative deduction model based on a multi-dimensional collaborative relationship model, deduce the state evolution process of multiple tasks under resource constraints and collaborative relationships, and form a dynamic evolution path for task scheduling.

[0092] The complex task collaborative optimization model unit is used to construct a complex task collaborative optimization model based on a dynamic evolution path, concretize the system performance into an efficiency function including maximum completion time, total energy consumption and task benefit, and perform performance evaluation and optimization selection on the task path under multiple types of scheduling constraints.

[0093] The collaborative strategy optimization unit is used to design a collaborative strategy optimization method that integrates an annealing mechanism and a virtual gradient according to the optimized selected task path, construct a strategy perturbation direction based on the efficiency feedback of the current path, and introduce an annealing mechanism to adjust the perturbation intensity and acceptance probability, thereby realizing multi-round iterative optimization of the task collaborative path.

[0094] The scheduling results of the complex task collaborative twin deduction method based on virtual-real fusion in the present invention under normal circumstances are as follows: Figure 2 As shown in the figure, the maximum completion time and comprehensive energy consumption iteration curve of the TCDPV of the present invention under normal circumstances are as follows: Figure 3 As shown, the scheduling results of the TCDPV method of the present invention in the case of key equipment failure are as follows Figure 4 As shown in the figure, the maximum completion time and comprehensive energy consumption iteration curve of the TCDPV method of the present invention in the case of key equipment failure are as follows: Figure 5 As shown in the figure, the maximum completion time and energy consumption performance of each algorithm in the two scenarios of the present invention are compared. Figure 6 The present invention leverages the virtual deduction and feedback mechanism driven by digital twins to effectively enhance the global optimization capability of the scheduling strategy and the stability of system operation. It has broad application prospects in the flexible production task scheduling and collaborative optimization scenarios of automotive parts in the field of intelligent manufacturing.

[0095] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A complex task collaborative twin deduction method based on virtual-real fusion, characterized by: The following steps are involved: Step 1: Obtain the task and equipment information of the flexible production system of automobile parts, and build a multi-dimensional collaborative relationship model of tasks and equipment in the flexible production system of automobile parts; Step 2: Build a complex task collaborative deduction model based on the multi-dimensional collaborative relationship model, deduce the state evolution process of multiple tasks under resource constraints and collaborative relationships, and form a dynamic evolution path for task scheduling; Step 3: Build a complex task collaborative optimization model based on the dynamic evolution path, concretize the system performance into an efficiency function including maximum completion time, total energy consumption, and task benefits, and perform performance evaluation and optimization selection of task paths under multiple scheduling constraints; Step 4: Based on the optimized task path, a collaborative strategy optimization method that integrates the annealing mechanism and the virtual gradient is designed. The strategy perturbation direction is constructed based on the efficiency feedback of the current path, and the annealing mechanism is introduced to adjust the perturbation intensity and acceptance probability to achieve multi-round iterative optimization of the task collaborative path.

2. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 1 is characterized by: The complex task collaborative optimization model constructed in step 3 is as follows: Where E represents the system efficiency, represents the deduction path, ε total Indicates production efficiency, represents the maximum completion time, S Tj (t) represents each task T j The execution progress at time t, M represents the total number of devices, A ij (t) represents task T j Is device D occupied at time t? i , S Di (t) represents the state of the device at time t, C total represents the total production energy consumption, T represents the task set, N represents the number of tasks, U Ti (t) represents task T i implementation decisions, represents the system evolution function, X represents the system state space set, U represents the strategy space set, C task-task represents the task-task constraint; C task-device represents the task-device constraint; C device-device Represents a task-task constraint.

3. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 2 is characterized by: The method for constructing the complex task collaborative optimization model in step 3 is as follows: For the measurement of the maximum completion time, considering the actual execution time of all tasks in the task set T, the following specific expression is defined: in, represents the maximum completion time, S Tj (t) represents each task T j The execution progress at time t, M represents the total number of devices, A ij (t) represents task T j Is device D occupied at time t? i , S Di (t) represents the state of the device at time t; For the total production energy consumption, under the premise of knowing a specific deduction path, it can be expressed as follows: Among them, C total represents the total production energy consumption, T represents the task set, N represents the number of tasks, and U Ti (t) represents task T i By combining the system state evolution law and scheduling constraints, a complex task collaborative optimization model can be constructed.

4. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 3 is characterized by: Step 3 is to obtain the next round of deduction path A new initial state X needs to be constructed (k+1) (0); With the help of the change information of path efficiency, a virtual gradient perturbation model based on the target value difference is constructed, which is specifically expressed as follows: Among them, λ (k) is the perturbation step size control factor, E (k) represents the k-th deduction path system efficiency, virtual represents the virtual gradient, is the virtual gradient direction constructed based on efficiency feedback. In actual calculation, considering the non-differentiable nature of the efficiency function, this term can be approximated by difference: Where sign(·) is a sign function that indicates the direction of change in the efficiency of the current path; Therefore, the new initial state update formula is expressed as: X (k+1) (0)=X (k) (0)+λ (k) ·sign(E (k) -E (k-1) )·(X (k) (0)-X (k-1) (0)) Among them, X (k+1) represents the starting state of the k+1th deduction, X (k) represents the starting state of the kth deduction, X (k-1) represents the starting state of the k-1th deduction, E (k) represents the efficiency of the k-th deduction path system, E (k-1) Indicates the efficiency of the k-1th deduction path system.

5. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 4 is characterized by: The collaborative strategy optimization method integrating the annealing mechanism and the virtual gradient in step 4 includes the following steps: Assume that the current k-th deduction path is Its initial state is X (k) (0), the corresponding efficiency is E (k) According to the virtual gradient concept, the direction guidance item is constructed through the state difference and efficiency change between historical paths: ΔX (k) =sign(E (k) -E (k-1) )·(X (k) (0)-X (k-1) (0)) Where ΔX (k) Indicates a direction guide item; Consider introducing the perturbation control function in the annealing mechanism, combining the direction guidance and annealing perturbation to construct the update formula of the initial state of the path: X (k+1) (0)=X (k) (0)+λ (k) ·ΔX (k) +e (k) Among them, ε (k) is the disturbance term, and the statistical characteristics of the disturbance term are related to the current temperature T (k) Related to, satisfying: in, represents a normal distribution, represents the variance of the disturbance term, β>0 is the disturbance amplitude control parameter, T (k) Indicates the current temperature; New generation path It is generated through the collaborative deduction model of complex tasks, and its performance value E is calculated by the efficiency function (k+1) To control the path acceptance strategy, the Metropolis criterion in simulated annealing is introduced to determine the acceptance probability of the current path based on whether the efficiency is improved or decreased: in, represents the acceptance probability of the current path; The temperature is updated according to an exponential decay strategy: T (k+1) =α·T (k) ,0<α<1 During the multi-round path update process, the following evolutionary strategy is used to find the most efficient path: in, Indicates the optimal deduction path obtained by the final convergence, and the initial state X corresponding to this path * (0), the strategy sequence {U(t)}, the state sequence {X(t)} and the task-equipment collaborative relationship all achieve the optimal efficiency configuration, constituting the optimal scheduling execution scheme in the flexible production system.

6. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 5 is characterized by: The method for constructing a multi-dimensional collaborative relationship model between tasks and equipment in the flexible production system of automobile parts in step 1 includes the following steps: Step 11: Modeling device-device relationships In the flexible manufacturing process of automotive parts, the equipment types in the flexible production system include CNC machine tools, automatic testing equipment, robotic arms, logistics conveyor lines, and material handling robots. The equipment set in the flexible production process is defined as follows: V={D1,D2,...,D i ,...,D M } Among them, V represents the set of devices in the system, D i represents the i-th device, and M represents the total number of devices; In a flexible production system, the interactions between devices can be unidirectional, bidirectional, or non-unidirectional. This feature is described uniformly using directed edges, and the edge set E is defined as follows: If (D i ,D j )∈E, it means there is a slave device D i Flow to device D j For a pair of devices that have both bidirectional connections, the set E can contain (D i ,D j ) and (D j ,D i ); If you only need to determine whether the connection is feasible, you can use the adjacency matrix A = [a ij ] M×M Represents a binary relationship: Define the capacity matrix C: C=[c ij ] M×M in, When c ij >0 and c ji =0 indicates a one-way connection; when c ij >0 and c ji >0 indicates a two-way connection; if both are 0, there is no connection; For the establishment of topology and directed graph, based on the equipment set and connectivity definition, the flexible production system as a whole is represented as a triple: G=(V,E,C) Among them, G is the triple representation of the flexible production system, V is the set of equipment in the system, E is the edge set used to characterize the feasible transmission direction between equipment; C is the capacity matrix; The following direction function is defined to distinguish the connection direction between devices: Among them, dir(i,j) represents the direction function; Step 12: Task-device relationship modeling First, we analyze the correlation between task sets and devices. Assume that there are N tasks in the system, which constitute the task set T: T={T1,T2,...,T N } Each task T k The device association degree is defined as the normalized weight w of the number of devices required for the task k : satisfy For task T k The required device node set; Task T k The device connection relationship can be modeled as a weighted directed graph G k =(D k ,ε k ,W k ), is the set of directed edges between devices; is the edge weight matrix: Formula α ij Representation device arrive The benchmark transmission efficiency; t is the task execution time, β, γ are time decay parameters; d ij is the physical distance between devices, λ is the spatial attenuation coefficient; Task T k The execution path must satisfy the following constraints: First, the topological connectivity, i.e., the task subgraph G k There must be at least one directed path from the starting point to the end point for any device node in . Second, the capacity constraint is that the amount of parallel transmission between devices does not exceed the limit of the global capacity matrix C, which is expressed as follows: Where P is the number of parallel transmission channels, C k (i,j,p) is task T k The instantaneous capacity on the pth channel; Device D i The resource occupancy rate at time t must satisfy: Among them, ρ i (t) is the device D i The resource occupancy rate at time t, ∈ is the nonlinear damping coefficient, ρ max is the maximum load threshold of the device; Step 13: Task-task relationship modeling There are N kinds of tasks in the system, and their relationships are expressed through the fourth-order tensor R∈R N×N×M×M Indicates that the tensor element R(i,j,p,q) describes the task T i With T j In device D p to D q The interaction strength on a link is defined as: Where D i ∩D j For task T i With T j The shared device set; σ is the difference tolerance coefficient, θ is the nonlinear adjustment parameter, C k (i,j,q) is task T k The instantaneous capacity on the qth channel; The overall conflict intensity between tasks is expressed by the matrix K∈R N×N Indicates that its element k(i,j) is calculated as: Where: A i For task T i The adjacency matrix, A j For task T j The adjacency matrix of The collaborative potential between tasks is expressed by the three-dimensional matrix S∈R N×N×L Description, its elements S(i,j,t) are defined as: Where L is the number of time windows, W i (p,·) is the task T i The output weight, W j (·,p) is the task T j The input weight, W i (p,·)·W j (·,p) is the task T i The output weights and T j The dot product of the input weights; The final relationship between tasks can be expressed by the decision matrix F(i,j) that integrates conflict and collaboration: Among them, F(i,j) is the decision matrix; when F(i,j)>0, task T i With T j It mainly manifests as resource competition; when F(i,j)<0, task T i With T j Possesses potential for collaborative optimization.

7. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 6 is characterized by: Step 2: The method for constructing a complex task collaborative deduction model includes the following steps: Step 211, state space Equipment status: The resources of the flexible production system are various types of equipment. For any equipment D i ∈V, the state S of the device at time t Di (t) is defined as: The overall device status is described as follows: S V (t)=[S D1 (t),S D2 (t),...,S DM (t)] T Task execution status: In the flexible production system, production tasks constitute the decision-making subject in the dynamic game. The task set is T = {T1, T2, ..., T N Each task T j The execution progress at time t is represented by a continuous value: S Tj (t)∈[0,1],j=1,2,…,N The overall status of all tasks is expressed as: S T (t)=[S T1 (t),S T2 (t),...,S TN (t)] T Task-device association status: Task execution is inseparable from device resources. Define any task T j The state of association with the device at time t is: The overall relationship between all tasks and equipment occupancy is expressed in matrix form as follows: A T,V (t)=[A j,i (t)] M×N Then the overall system state space can be uniformly represented as a set of triples: X(t)={S V (t),S T (t),A T,V (t)} Among them, the state space X(t) provides a comprehensive real-time information basis for strategy formulation and interactive decision-making of each task in the dynamic game; Step 212, strategy space Equipment occupancy strategy: Each task is a decision-making subject in the dynamic game. At time t, any task T j For device D i The occupancy decision is defined as: The overall equipment occupancy strategy matrix is expressed as: U V (t)=[U j,i (t)] M×N Task execution strategy: Each task is an independent decision-making entity. At any time t, task T j The execution decision of is defined as: The overall vector of all task execution strategies is: Then, the strategy space can be uniformly expressed as: U(t)={U V (t),U T (t)}。 8. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 7 is characterized by: Step 2: The method for constructing a complex task collaborative deduction model includes effectiveness evaluation indicators. The effectiveness evaluation indicator method is as follows: Starting from the maximum completion time, total production energy consumption and production efficiency, a dynamic path-dependent performance evaluation system is constructed, and the system performance index E is defined accordingly. For the maximum completion time In the dynamic collaborative scheduling process, the task completion time is not fixed, but changes dynamically with the task execution order, device parallelism, and resource conflicts, with the following relationship: in, is the time aggregation map; For total production energy consumption The total production energy consumption can satisfy the following relationship: in, is the production energy consumption aggregation function; For production efficiency ε total The set of production tasks T is predefined, with clear objectives and fixed boundaries. The production benefits that each task can generate are fixed values: Among them, ε j For task T j The corresponding fixed benefit value; Then, the system performance indicators have the following relationship: Among them, E is the system efficiency index.

9. The complex task collaborative twin deduction method based on virtual-reality fusion according to claim 8 is characterized by: The state evolution process of multiple tasks under resource constraints and collaborative relationships in step 2 includes the following steps: Define principal point intersection It is used to describe the structural conflict intensity when two tasks have overlapping key equipment during execution. For any two tasks x1, x2, if their principal points intersect to form an empty set, that is, This means that the sets of devices used by the two tasks at any given time do not overlap. If the principal point intersection is a certain principal step point k, then the other task can only run in parallel with the benchmark task after it has executed to step k. During the game scheduling process, by adjusting the execution order and starting offset time of the tasks, the structure of the principal point intersection can be optimized, so that multiple tasks can be switched from serial to parallel as early as possible, thereby improving the resource utilization and system efficiency of the overall scheduling. Based on this, the game equilibrium and strategy update mechanism are unified and abstracted into the following strategy update formula, which is used to describe the task set in the iteration round r The collaborative strategy evolution process under: Among them, U (r+1) represents the joint strategy set of round r+1; E(U) represents the system effectiveness function of the task set under the current strategy combination U; Represents task x i with x j Principal point intersection conflict cost under the current strategy; represents the policy disturbance degree; λ1, λ2 represent the control parameters; The entire deduction output is shown below: sU (r) Reach game equilibrium and maximize E; Among them, X(T * ) represents the set of system state spaces at the time of reaching the optimal deduction path, T * Indicates the moment when the optimal deduction path is reached, represents the best deduction path, represents the system evolution function, U (r) (t) represents the set of joint strategies in round r at time t.

10. A deduction system for implementing the complex task collaborative twin deduction method based on virtual-reality fusion as described in claim 1, characterized in that: It includes input unit, multi-dimensional collaborative relationship model unit, complex task collaborative deduction model unit, complex task collaborative optimization model unit, and collaborative strategy optimization unit, among which: The input unit is used to input task and equipment information of the flexible production system of automobile parts; The multidimensional collaborative relationship model unit is used to construct a multidimensional collaborative relationship model of tasks and equipment in the flexible production system of automobile parts according to the tasks and equipment information of the flexible production system of automobile parts; The complex task collaborative deduction model unit is used to construct a complex task collaborative deduction model based on the multi-dimensional collaborative relationship model, deduce the state evolution process of multiple tasks under resource constraints and collaborative relationships, and form a dynamic evolution path for task scheduling; The complex task collaborative optimization model unit is used to construct a complex task collaborative optimization model based on a dynamic evolution path, concretize system performance into an efficiency function including maximum completion time, total energy consumption and task benefit, and perform performance evaluation and optimization selection on task paths under multiple types of scheduling constraints; The collaborative strategy optimization unit is used to design a collaborative strategy optimization method that integrates an annealing mechanism and a virtual gradient according to the optimized selected task path, construct a strategy perturbation direction based on the efficiency feedback of the current path, and introduce an annealing mechanism to adjust the perturbation intensity and acceptance probability, thereby realizing multi-round iterative optimization of the task collaborative path.

Images