Task allocation method and system based on hybrid game genetic algorithm

Abstract

The invention discloses a task allocation method and system based on a hybrid game genetic algorithm, and belongs to the field of intelligent manufacturing, and the method comprises the following steps: obtaining task allocation data based on an industrial Internet of Things and an edge computing technology; constructing a task allocation-path planning-energy consumption control collaborative optimization model based on the task allocation data; the collaborative optimization model comprises a three-dimensional optimization objective function and a constraint system; the constraint system comprises an equipment health state constraint, an AGV dynamic motion constraint, a process connection space-time constraint and a multi-agent game equilibrium constraint; and solving the collaborative optimization model based on a hybrid game genetic algorithm to obtain a task allocation optimal solution. Through deep coupling of a multi-agent game and a dynamic evolutionary algorithm, the problem of decision dimension splitting of traditional single-target static scheduling is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent manufacturing, and particularly relates to a task allocation method and system based on a hybrid game genetic algorithm. Background Art

[0002] Currently, task allocation in textile workshops mainly relies on traditional scheduling algorithms (such as heuristic algorithms, genetic algorithms, etc.). However, there are the following core bottlenecks in the scenario of multi-objective collaborative optimization: ① Local optimal trap: The traditional genetic algorithm uses fixed crossover / mutation probabilities and is prone to falling into local optimal solutions. It is difficult to balance various indicators, especially in multi-objective optimization (such as completion time, energy consumption, and empty driving rate). ② Unbalanced resource allocation: The task allocation is separated from the equipment selection, resulting in uneven equipment loads; the AGV path planning is independent of the production plan, with a high empty driving rate and significant energy consumption redundancy. ③ Insufficient dynamic adaptability: There is a lack of real-time data feedback mechanism and it is unable to cope with dynamic disturbances such as order insertion and equipment failures; the charging strategy is fixed, and the AGV interrupts transportation due to insufficient battery power. ④ Failure of multi-objective optimization: Traditional algorithms only focus on a single objective (such as minimizing the completion time), ignoring the collaborative optimization of energy consumption and logistics efficiency; the path planning does not consider the dynamic occupation of warehouse resources, resulting in warehouse conflicts.

[0003] The limitations of the existing technologies essentially stem from the design paradigm of "single-objective orientation" and "static decision-making". Therefore, the present invention provides a task allocation method based on a hybrid game genetic algorithm. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a task allocation method and system based on a hybrid game genetic algorithm to solve the problems existing in the above prior art.

[0005] To achieve the above object, the present invention provides a task allocation method based on a hybrid game genetic algorithm, including:

[0006] Obtaining task allocation data based on industrial Internet of Things and edge computing technologies;

[0007] Constructing a collaborative optimization model of task allocation - path planning - energy consumption control based on the task allocation data; the collaborative optimization model includes a three-dimensional optimization objective function and a constraint system; the constraint system includes: equipment health status constraints, AGV dynamic motion constraints, process connection spatio-temporal constraints, and multi-agent game equilibrium constraints;

[0008] Solving the collaborative optimization model based on a hybrid game genetic algorithm to obtain the optimal task allocation solution.

[0009] Optionally, the task allocation data includes: equipment status data, AGV charge status data, warehouse topology, and order dynamic data.

[0010] Optionally, the expression of the three-dimensional optimization objective function is:

[0011]

[0012]

[0013]

[0014] In the formula, represents the maximum operation completion time, represents the completion time of the m-th operation, m represents the operation sequence number, and M represents the total number of operations, represents the total energy consumption, N represents the total number of devices, represents the power of the i-th device, represents the processing duration of the i-th device, represents the total number of AGVs, represents the energy consumption during the operation of the j-th AGV, represents the empty driving distance of the j-th AGV, represents the total driving distance of the j-th AGV, represents the AGV empty driving rate.

[0015] Optionally, the expression of the device health status constraint is:

[0016]

[0017] The expression of the AGV dynamic motion constraint is:

[0018]

[0019] In the formula, represents the health index of the i-th device, represents the path node conflict detection function, represents the indicator function, taking 1 in the charging state, represents the health index of the i-th device at time t, represents the absolute value of the acceleration of the j-th AGV at time t, represents the state of the AGV, represents the state of charge of the j-th AGV at time t, k represents the k-th task, represents the path node of the j-th AGV at time t when executing the k-th task, represents the path node of the l-th AGV at time t when executing the m-th task, represents the AGV number, represents the time, Indicates the AGV number, Indicates the state of the j-th AGV at time t, Indicates the total number of charging piles.

[0020] Optionally, the expression of the spatio-temporal constraint for process connection is:

[0021]

[0022] The expression of the multi-agent game equilibrium constraint is:

[0023]

[0024] In the formula, Indicates the start time of task k, Indicates the arrival time of the predecessor material p, where p belongs to the set of predecessor materials of task k , Indicates the set of predecessor materials of task k, Indicates the total number of materials currently involved, Indicates the indicator function, Indicates the volume of material i, Indicates the maximum capacity of storage location c, Indicates the end time of the m-th process at the equipment end, Indicates the time when the AGV arrives at the m-th process, Indicates the standard processing duration of the current process, Indicates the Shapley value of the equipment agent, Indicates the Shapley value of the AGV agent, Indicates the Shapley value of the Energy Agent, Indicates the revenue matrix of the three-party agents at time t, Indicates the mixed strategy space, Indicates the utility function of agent i, Indicates the Frobenius norm of the revenue matrix at time t and (t - 1) times, Indicates the mixed strategy Nash equilibrium point, Indicates that agent i takes the equilibrium strategy when other agents 、and itself takes the strategy The value of the utility function at this time.

[0025] Optionally, the process of solving the collaborative optimization model based on the hybrid game genetic algorithm includes:

[0026] Dividing the chromosome into equipment selection gene segments, path planning gene segments, energy consumption gene segments, and game strategy gene segments to obtain a four-segment dynamic coding structure;

[0027] Construct a multi-agent game model, and based on the multi-agent game model, update the payoff matrix and the Shapley value to dynamically adjust the population corresponding to the four-stage dynamic coding structure to move towards the Pareto front;

[0028] Adjust the crossover operator and the dynamic mutation operator during the process of the population moving towards the Pareto front through an adaptive crossover and mutation mechanism to obtain the optimal task allocation solution.

[0029] Optionally, the calculation expression of the Shapley value is:

[0030]

[0031] In the formula, represents the agent for which the Shapley value is currently being calculated, represents the Shapley value corresponding to agent a, represents the set of three parties of agents {MA, AGV, EA}, represents the set composed of all the remaining agents after removing agent a from set A, represents any subset of represents the marginal contribution of agent a to coalition S.

[0032] Optionally, the adaptive crossover and mutation mechanism includes Pareto front detection crossover and dynamic mutation.

[0033] Optionally, a three-level mutation intensity is adopted to control the adaptive crossover and mutation mechanism;

[0034] Among them, the three-level mutation intensity includes: the elite layer, the middle layer, and the worst layer.

[0035] The present invention also provides a task allocation system based on a hybrid game genetic algorithm for implementing the task allocation method, and the system includes:

[0036] A multi-source data perception layer for obtaining task allocation data based on industrial Internet of Things and edge computing technologies;

[0037] A collaborative optimization model layer for constructing a collaborative optimization model of task allocation - path planning - energy consumption control based on the task allocation data;

[0038] A hybrid game genetic algorithm engine for integrating the Nash bargaining mechanism and an improved genetic algorithm, and designing a dynamic chromosome coding strategy, a Pareto front detection operator, and a multi-agent payoff matrix;

[0039] The real-time feedback control layer is used for the digital twin platform to implement simulation verification of the scheduling scheme and issue dynamic adjustment instructions through the edge computing nodes.

[0040] Compared with the prior art, the present invention has the following advantages and technical effects: The present invention constructs a three-dimensional collaborative optimization model of task assignment - path planning - energy consumption control. Through the deep coupling of multi-agent game and dynamic evolution algorithm, it breaks through the problem of decision dimension fragmentation in traditional single-objective static scheduling, integrates the objective functions, defines a three-dimensional optimization objective function, incorporates the equipment processing time, AGV transportation energy consumption and empty running rate into the same optimization system, realizes multi-objective collaborative optimization through dynamic weights, establishes a dynamic correlation model between energy consumption and efficiency, and reduces the AGV energy consumption redundancy through the collaborative optimization of acceleration constraint and charging strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0042] Figure 1 It is the system framework diagram of the embodiment of the present invention;

[0043] Figure 2 It is the algorithm mechanism flow chart of the embodiment of the present invention;

[0044] Figure 3 It is the self-adaptive crossover and mutation flow chart of the embodiment of the present invention;

[0045] Figure 4 It is the game strategy feedback of the embodiment of the present invention;

[0046] Figure 5 It is the energy consumption optimization result diagram of the embodiment of the present invention;

[0047] Figure 6 It is the change of the AGV empty running rate of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0049] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0050] Embodiment 1.

[0051] The limitations of the existing technologies essentially stem from the design paradigms of "single - goal orientation" and "static decision - making", which are specifically manifested in the following aspects: ① Lack of multi - goal coordination: The independent optimization of equipment scheduling and logistics transportation leads to an increase in the waiting time of AGVs and a relatively high proportion of energy consumption during equipment idling; the lack of coupled optimization between charging decision - making and transportation tasks results in a large proportion of non - working energy consumption of AGVs in the total energy consumption. ② Weak ability to handle dynamic constraints: Traditional algorithms rely on offline parameter tuning and cannot adapt to real - time changes in the workshop (such as equipment status, order priority fluctuations); the path planning uses the static A* algorithm, which cannot avoid sudden congestion and has a relatively high path conflict rate. ③ Imperfect multi - agent cooperation mechanism: There is a lack of effective negotiation among multiple agents (Machine Agent / AGV Agent, etc.), resulting in resource preemption conflicts; the solution of the Nash equilibrium stays in the theoretical model and does not form a closed - loop optimization with the genetic algorithm, and the strategy stability is poor. ④ There are inherent limitations in the evolutionary operators of traditional genetic algorithms: Rigid crossover / mutation strategies: Fixed probability parameters lead to an imbalance between exploration and exploitation, being prone to falling into local optima in the early stage and losing population diversity in the later stage, resulting in being unable to jump out of the sub - optimal solution. Single elitist retention strategy: Only relying on the elitist selection mechanism based on fitness ranking, ignoring the Pareto dominance relationship among multiple goals, resulting in an uneven distribution of the solution set. Lagging environmental response: Real - time working condition data (such as AGV battery level, equipment fault signals) cannot be incorporated during the iterative process, leading to a large deviation between the offline optimization results and the actual scenario. ⑤ The fragmented optimization of task assignment, equipment scheduling, and logistics transportation causes systematic losses: Spatiotemporal resource conflicts: The path planning of AGVs is not synchronized with the machining time sequence of machines, resulting in a relatively high proportion of waiting time of AGVs beside machines and an increase in machine idle rate. Energy coordination blind spots: No associated model between energy consumption and efficiency is established, with too high acceleration / deceleration frequencies of AGVs and redundant motor energy consumption. Mismatch in charging scheduling: The charging pile allocation strategy is decoupled from transportation tasks, resulting in an increase in the task delay rate of AGVs due to charging interruptions. There are essential bottlenecks in the ability of existing technologies to respond to emergencies, that is, the collaborative optimization under the distributed decision - making framework fails: Lack of negotiation mechanism: Machine Agent and AGV Agent adopt a simple priority preemption strategy, and the time consumption for resolving resource conflicts accounts for a relatively high proportion. Distorted game equilibrium: The traditional Nash equilibrium solution relies on a static payoff matrix and does not consider the state transition probability of the workshop, with a relatively large strategy stability error.

[0052] In the context of the intelligent upgrading of the textile industry, for the complex production scenarios of multi-process coupling (spinning - weaving - printing and dyeing), heterogeneous equipment clusters (more than 500 processing units), and large-scale AGV scheduling (more than 100 transportation units), traditional single-objective static scheduling methods have systematic defects such as fragmented decision-making dimensions (independent optimization of processes / equipment / logistics), sluggish dynamic response, and extensive energy efficiency control (a relatively high proportion of ineffective energy consumption). The innovation of this invention lies in constructing a three-dimensional collaborative optimization model of task allocation - path planning - energy consumption control. By designing a chromosome dynamic coding strategy guided by the Nash bargaining mechanism (the coding dimension includes equipment working conditions, AGV charge status, and warehouse topology information), an adaptive crossover and mutation operator based on Pareto front detection, and a population evolution direction correction mechanism driven by the multi-agent revenue matrix, multi-objective Pareto optimization of process connection efficiency, logistics turnover rate, and unit energy consumption output is achieved, especially breaking through the real-time decision-making bottleneck of traditional algorithms under the scale of more than 200 concurrent tasks.

[0053] The creative points of this invention are as follows: (1) Multi-agent strategy gene fusion coding: Breaking through the limitation that traditional chromosomes only map physical parameters, embedding the game strategy into the gene structure.

[0054] Four-dimensional dynamic gene segments: Fusing dynamic weight coding in the equipment health gene segment to trigger the weight reconstruction of high-precision processes; using spatio-temporal conflict hot zone coding in the path planning segment, dynamically generating congestion attenuation coefficients based on the RFID grid occupancy rate, and introducing the exclusive OR logic of charge status and charging pile occupancy to achieve gene-level responses for conflict prediction and path avoidance; in the energy consumption segment, through the joint modeling of acceleration energy consumption field strength and equipment health loss, encoding the energy efficiency parameters as field strength coupling factors to drive the energy consumption balance optimization of path and equipment selection; the game strategy segment uses Shapley value - Nash equilibrium dual-mode coding, recording the proportion of agent contribution and equilibrium convergence flag at the gene level, and adjusting the expression intensity of the strategy anchoring area through real-time feedback of edge nodes, enabling the chromosome to have an adaptive multi-objective game navigation ability.

[0055] (2) Hierarchical perception evolution regulator: Breaking through the static parameter limitations of traditional evolution operators, proposing a dynamic perception and intelligent regulation mechanism based on the Pareto front hierarchy, and achieving the evolutionary control of multi-objective optimization through triple innovations of hierarchical threshold decision-making, collaborative probability decay, and iterative self-adaptation.

[0056] ① Dynamic threshold decision-making mechanism: Frontier-level intelligent perception: Real-time monitor the distribution density of the population in the target space, and dynamically adjust the crossover trigger threshold according to the total number of frontier levels to avoid interference from low-quality levels to high-quality solutions. ② Hierarchical regulation strategy: Divide the population into a global exploration area, a transition optimization area, and an elite protection area, and design different evolutionary rules for different levels. ③ Iterative feedback-compensation mechanism: Dynamically adjust the threshold parameters and attenuation rate. When the population diversity is lower than the critical threshold, perform directional mutation on the individuals in the transition area to avoid evolutionary stagnation.

[0057] (3)Game-evolution dynamic cooperation engine: Construct a deep coupling mechanism of multi-agent game and genetic algorithm, and realize real-time evolutionary navigation through dynamic revenue feedback and strategy anchoring technology.

[0058] Dynamic revenue matrix construction: Generate a revenue matrix in real time based on equipment utilization rate, AGV effective transportation rate, and energy efficiency index to quantify the collaborative contribution of agents; Shapley value gradient optimization: Dynamically adjust the weights of the objective function through the calculation of the marginal contribution of the coalition to guide the population to move towards the Pareto front; Nash equilibrium convergence flag embedding: Use the equilibrium error as the gene flag bit to trigger the adjustment of the strategy anchoring area of the edge node.

[0059] (4)Three-dimensional collaborative optimization model: Construct a three-dimensional collaborative optimization model of task allocation - path planning - energy consumption control. Through the deep coupling of multi-agent game and dynamic evolutionary algorithm, break through the decision-making dimension fragmentation problem of traditional single-objective static scheduling, integrate the objective function, define a three-dimensional optimization objective function, incorporate equipment processing time, AGV transportation energy consumption, and empty running rate into the same optimization system, realize multi-objective collaborative optimization through dynamic weights, establish a dynamic correlation model between energy consumption and efficiency, and reduce the AGV energy consumption redundancy through the collaborative optimization of acceleration constraints and charging strategies.

[0060] As Figure 1 shown, the present invention constructs a three-dimensional collaborative optimization framework based on a hybrid game genetic algorithm, and realizes the collaborative optimization of task allocation, path planning, and energy consumption control in an intelligent manufacturing workshop through the deep coupling of multi-agent game and dynamic evolutionary algorithm.

[0061] This embodiment provides a task assignment system based on a hybrid game genetic algorithm. The system adopts a four-layer architecture design and realizes dynamic optimization through the closed-loop control of data-driven and intelligent decision-making: 1. Multi-source data perception layer: Integrate industrial Internet of Things (IoT) and edge computing technologies to collect three types of key data in real time: equipment status (vibration / temperature sensors), AGV charge status (BMS data), warehouse topology (RFID positioning), and order dynamics (MES system) data. 2. Collaborative optimization model layer: Construct a three-dimensional objective function for task assignment-path planning-energy consumption control, and define process connection constraints, AGV charging time window constraints, and equipment load balancing constraints. 3. Hybrid game genetic algorithm engine: Integrate the Nash bargaining mechanism and the improved genetic algorithm, and design a dynamic chromosome coding strategy, a Pareto front detection operator, and a multi-agent revenue matrix. 4. Real-time feedback control layer: Based on the digital twin platform, realize the simulation verification of the scheduling scheme, and issue dynamic adjustment instructions through the edge computing node.

[0062] As a specific implementation manner of this embodiment, this embodiment also provides a task assignment method based on a hybrid game genetic algorithm, which specifically includes the following steps: Obtain task assignment data based on industrial Internet of Things and edge computing technologies; construct a collaborative optimization model for task assignment-path planning-energy consumption control based on the task assignment data; the collaborative optimization model includes a three-dimensional optimization objective function and a constraint system; the constraint system includes: equipment health status constraints, AGV dynamic motion constraints, process connection spatio-temporal constraints, and multi-agent game equilibrium constraints; solve the collaborative optimization model based on the hybrid game genetic algorithm to obtain the optimal task assignment solution.

[0063] Multi-source data perception layer: Realize the fusion of three types of data through industrial Internet of Things edge nodes. Among them, equipment condition data: The sensor types include piezoelectric vibration sensors (range 0-20g) and infrared temperature sensors (range 0-150°C).

[0064] The key parameters are as follows: ① Vibration root mean square (RMS): Sampling rate 10kHz, detect abnormal machining vibration; ② Bearing temperature: Accuracy ±0.5°C, monitor overheating risk; ③ Extract the equipment health index (HI): 。

[0065] AGV dynamic data: SOC monitoring: Accuracy ±1%, sampling period 1 second; Represents the root mean square value of the current operating parameters of the equipment (such as vibration amplitude, current, etc.), reflecting the real-time operating state of the equipment; Represents the root mean square threshold of the corresponding parameters when the equipment is operating normally, serving as a reference value for judging whether the equipment is normal; Represents the current cumulative operating time of the equipment, recording the operating duration after the equipment is put into use; Represents the maximum cumulative operating time (or cycle) of the equipment design, which represents the theoretically longest effective operating duration of the equipment.

[0066] Real-time position: UWB positioning (accuracy ±10 cm); Motion state: speed (0 - 2 m / s), acceleration (±0.3 m / s²); Charging strategy model: 。

[0067] Represents the charging time of the AGV, which is used to measure the duration required to charge the AGV; Represents the current state of charge (SOC) of the AGV, which indicates the proportion of the remaining battery power; Represents the capacity of the AGV battery, that is, the total amount of electricity that the battery can store; Represents the power during AGV charging, which reflects the amount of electricity charged into the battery per unit time.

[0068] Order and warehousing data: Obtain order priorities (calculated based on delivery urgency and customer level) and real-time inventory coordinates (RFID grid resolution 0.5 m × 0.5 m) from the MES system.

[0069] Collaborative optimization model layer: Multi-objective modeling, where a three-dimensional optimization objective function is defined.

[0070]

[0071]

[0072]

[0073] In the formula, Represents the maximum operation completion time, Represents the completion time of the m-th operation, m represents the operation sequence number, and M represents the total number of operations, Represents the total energy consumption, N represents the total number of equipment, Represents the power of the i-th equipment, Represents the processing duration of the i-th equipment, Represents the total number of AGVs, Represents the energy consumption during the operation of the j-th AGV, Represents the empty driving distance of the j-th AGV, Represents the total driving distance of the j-th AGV (including empty driving and loaded driving), Represents the AGV empty driving rate.

[0074] The constraint system includes: 1. Equipment health state constraint (based on the HI index).

[0075] 。

[0076] Constraint ①: Ensure that the equipment for high-precision processing (such as cotton yarn above 60 counts) is in excellent condition; Constraint ②: Model based on the degradation synergy effect of the equipment group to prevent cluster failures; Constraint ③: Detect abnormal degradation (such as sudden bearing wear) through a sliding time window.

[0077] In the formula, represents the health index of the i-th equipment, represents the path node conflict detection function, represents the indicator function, taking 1 in the charging state, represents the health index of the i-th equipment at time t, represents the absolute value of the acceleration of the j-th AGV at time t, represents the state of the AGV (such as charging, running), represents the state of charge of the j-th AGV at time t. In the acceleration limit formula, it ensures that the AGV is more flexible when the power is sufficient, and limits the acceleration when the power is low, avoiding equipment failures or task interruptions due to insufficient power. k represents the k-th task, represents the path node of the j-th AGV when performing the k-th task at time t, represents the path node of the l-th AGV when performing the m-th task at time t, represents the AGV number (the identification of another AGV different from j), represents the time (time parameter), represents the AGV number (the j-th AGV), represents the state of the j-th AGV at time t, represents the total number of charging piles.

[0078] 2. AGV dynamic motion constraints (integrating BMS and UWB data).

[0079] 。

[0080] represents the path node conflict detection function (grid resolution 0.5m); represents the total number of charging piles (dynamically adjustable quantity); represents the indicator function, taking 1 in the charging state; is the acceleration limit; is the path conflict avoidance; is the charging pile occupancy.

[0081] 3. Process connection space-time constraints (based on RFID warehouse topology).

[0082] 。

[0083] Denote the set of precursor materials for task k Denote the maximum capacity of storage location c Denote the standard processing duration of the current process Denote the payoff matrix of the tripartite agent at time t Denote the mixed strategy space Denote the utility function of agent i Denote the start time of task k Denote the arrival time of precursor material p, where p belongs to the set of precursor materials for task k , Denote the total number of materials currently involved Denote the indicator function, which takes the value of 1 if material i is at storage location c at time t, otherwise 0 Denote the volume of material i Denote the end time of the m-th process at the equipment end Denote the time when the AGV arrives at the m-th process is the material completeness is the storage capacity limit is the AGV connection timing

[0084] 4. Multi-agent game equilibrium constraints

[0085] 。

[0086] The Shapley value of the Machine Agent Denote the Shapley value of the AGV agent Denote the Shapley value of the Energy Agent Denote the Frobenius norm of the payoff matrix at time t and (t - 1), used to measure the matrix change amplitude Denote the mixed strategy Nash equilibrium point Denote that when other agents of agent i adopt equilibrium strategies and agent i itself adopts strategy the value of the utility function is the Shapley value allocation is the payoff matrix stability is the Nash equilibrium existence

[0087] Hybrid game genetic algorithm engine: The process of the hybrid game genetic algorithm is as follows Figure 2As shown below. Dynamic chromosome coding strategy: The chromosome adopts a four - segment dynamic coding structure, mapping the real - time status of the workshop and game strategies through multi - dimensional genes. The specific expression is as follows: .

[0088] 1. Equipment selection gene segment ( ): Integer coding represents equipment ID + health status weight, and each gene bit is defined as: .

[0089] 2. Path planning gene segment ( ): Binary coding based on the spatio - temporal topology matrix, dynamically integrating warehousing occupancy information: .

[0090] Among them, , β is updated through the game, represents the logical AND weighted integration of the path information for each task to generate a binary gene segment. k is a loop variable, representing the serial number of the currently processed task, traversing each task from 1 to K. K is the total number of tasks and serves as the upper limit of k, indicating that a total of K tasks participate in the logical AND weighted integration, represents the indicator function. If path j belongs to the set of feasible paths of task k , the function value is 1; if not feasible, the function value is 0, which is used to screen the valid paths of task k, represents the congestion degree of path j at time t, and the weight will be dynamically adjusted according to the real - time congestion status of the path.

[0091] 3. Energy consumption gene segment ( ). Floating - point coding represents the energy - efficiency decision parameter: .

[0092] 4. Game strategy gene segment ( ). Hybrid coding stores the Shapley value ratio of multi - agents and the Nash equilibrium convergence flag: , among which, .

[0093] Fitness function: .

[0094] Among them, , represents the Euclidean distance between the k - th generation and the (k - 1) - th generation of the strategy vector p, measuring the change amplitude of the two - generation strategies. The larger the distance, the more significant the strategy change; the smaller the distance, the more stable the strategy. The distance is normalized by a constant to make the value of this item within a reasonable range. The Shapley value affects the multi - objective item through the weight : .

[0095] Adaptive crossover and mutation mechanism: As Figure 3 shown. ① Pareto front detection crossover includes front rank division: The population is divided into front ranks according to non-dominated sorting (Rank1 is the optimal front), and the threshold is adaptively adjusted according to the total number of population front levels: .

[0096] Dynamic adjustment of crossover probability: .

[0097] Gene fragment exchange rule: Through dynamic threshold adjustment (setting the threshold according to the total number of population front levels), non-linear exponential decay strategy (when ΔRank ≥ threshold, maintain a high crossover probability of 0.8 to promote global search, and when the difference is medium, quickly reduce the crossover probability by 0.7e^{-0.1ΔRank} to reduce the perturbation of high-quality solutions in adjacent levels), and strengthening the protection of the same layer (when ΔRank = 0, the crossover probability drops to 0.5 to avoid destroying elite solutions), the balance between exploration and exploitation is achieved, enhancing the local development stability while ensuring the global search ability, especially suitable for industrial optimization scenarios such as multi-objective workshop scheduling that require both diversity and convergence speed.

[0098] ② Dynamic mutation includes an adaptive mutation formula based on the Pareto front, and the adaptive mutation formula is: .

[0099] Adaptive rule for mutation step size:

[0100]

[0101]

[0102] In the initial stage of iteration (t → 0): β ≈ 0.3 provides strong perturbation, and α ≈ 0 maintains a large step size; in the later stage of iteration (t → T): β ≈ 0 weakens the perturbation, and α ≈ 0.1 accelerates the step size decay; where: represents the basic mutation probability; represents the individual belonging to the front rank; : The total number of population front levels at present; represents the maximum mutation step size.

[0103] Three-level mutation intensity control includes the elite layer (Rank1): Adopt a decaying protection mutation probability that decreases linearly with iteration, and the step size decay speed doubles; the middle layer (Rank2~N-1): Standard adaptive mutation maintains the Rank-sensitive characteristics of the original formula; the worst layer (RankN): Strengthen the perturbation mutation probability lower limit to 2 times the basic value (upper limit 0.8), and the step size increases with the dynamic strengthening coefficient β(t).

[0104] Multi-agent game optimization process: Construct a three-party game model of Machine Agent, AGV Agent, and Energy Agent, as shown in Table 1.

[0105] Table 1 Three-party game model.

[0106]

[0107] 1. Construction of the payoff matrix: 。

[0108] 2. Optimization of Shapley value calculation: 。

[0109] represents the agent currently calculating the Shapley value (such as Machine Agent, AGV Agent, Energy Agent); represents the set of three agents {MA, AGV, EA}; represents any subset (coalition) of, representing the combination of participants without agent a, is used to calculate the marginal contribution of a single agent to the coalition. The Shapley value of agent a needs to consider all subsets S that do not contain a, that is, S is any subset of, and these subsets S represent coalitions without a's participation, represents the set composed of all remaining agents after removing agent a from set A. represents the marginal contribution of agent a to coalition S, that is, the increment of the payoff after a joins.

[0110] Define the coalition value function as the total payoff of coalition , calculated based on the real-time payoff matrix : 。 Through the factorial term balance the contributions of coalitions of different sizes, S = 100.

[0111] 3. Solving for Nash equilibrium: 。

[0112] Combine the Shapley value with the gradient of the objective function to ensure that the strategy moves towards the Pareto front, decay with the solution change rate to accelerate convergence. Termination conditions: and 。 After each iteration, update the payoff matrix according to the current strategy : , and the game strategy feedback is as Figure 4as shown

[0113] Based on the above technical solutions, the present invention has the following beneficial effects: Systematic improvement of multi-dimensional collaborative optimization ability: By integrating the game equilibrium and dynamic evolution mechanism, the synergy between task allocation, path planning, and energy consumption control is significantly enhanced. The coupled optimization of equipment utilization rate and logistics turnover efficiency effectively reduces resource idleness. At the same time, through the dynamic modeling of energy efficiency to optimize energy allocation, efficient scheduling of global resources is achieved in complex workshop scenarios, significantly reducing the systematic losses in the production process.

[0114] Enhanced robustness and real-time performance in dynamic environments: Based on the real-time data-driven game strategy update mechanism, it can quickly respond to emergencies such as order changes and equipment failures. Through the combination of edge computing and digital twin technology, the optimization solution can be dynamically adjusted according to the workshop state, shortening the decision-making delay and ensuring the scheduling stability and execution reliability in complex disturbance scenarios.

[0115] Breakthrough in algorithm convergence efficiency and solution set quality: Through the adaptive crossover and mutation operator and the game equilibrium guiding mechanism, the algorithm convergence is accelerated while maintaining population diversity. The Pareto front detection technology effectively balances the trade-off between multiple objectives, generates a uniformly distributed high-quality solution set, and significantly improves the comprehensive performance of the scheduling scheme.

[0116] Optimization of resource allocation fairness and energy efficiency synergy: Based on the Shapley value-based contribution quantification mechanism, it avoids resource preemption or "free-riding" problems in traditional scheduling, and realizes fair synergy of equipment, AGV, and energy. Through the dynamic correlation modeling of energy consumption-efficiency, the energy allocation of equipment processing and logistics transportation is optimized, significantly reducing ineffective energy consumption while ensuring production efficiency. As Figures 5 - 6 shown

[0117] Adaptive expansion in complex constraint scenarios: Through spatio-temporal topology coding and dynamic gene mapping technology, it deals with equipment health state constraints, warehousing conflict avoidance, and process connection timing requirements, enhancing the scheduling ability of the algorithm for large-scale heterogeneous resources. The strategy negotiation mechanism under the hybrid game framework effectively resolves resource conflicts among multiple agents and improves the feasibility of the solution in complex constraint scenarios.

[0118] The above is only the preferred specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A task allocation method based on a hybrid game genetic algorithm, characterized in that: The following steps are involved: Obtain task allocation data based on industrial Internet of Things and edge computing technology; Constructing a collaborative optimization model of task allocation-path planning-energy consumption control based on the task allocation data; The collaborative optimization model includes a three-dimensional optimization objective function and a constraint system; the constraint system includes: equipment health status constraints, AGV dynamic motion constraints, process connection time and space constraints and multi-agent game equilibrium constraints; The collaborative optimization model is solved based on a hybrid game genetic algorithm to obtain the optimal solution for task allocation.

2. The task allocation method based on hybrid game genetic algorithm according to claim 1 is characterized in that: The task allocation data includes: equipment status data, AGV charge status data, warehouse topology and order dynamic data.

3. The task allocation method based on hybrid game genetic algorithm according to claim 1 is characterized in that: The expression of the three-dimensional optimization objective function is: In the formula, represents the maximum process completion time, represents the completion time of the mth process, m represents the process number, and M represents the total number of processes. represents the total energy consumption, N represents the total number of devices, represents the power of the i-th device, represents the processing time of the i-th device, Indicates the total number of AGVs, represents the energy consumption during the operation of the jth AGV, represents the idle driving distance of the jth AGV, represents the total driving distance of the jth AGV, Indicates the AGV idle rate.

4. The task allocation method based on hybrid game genetic algorithm according to claim 3 is characterized in that: The expression of the equipment health status constraint is: The expression of the AGV dynamic motion constraint is: In the formula, represents the health index of the ith device, represents the path node conflict detection function, Indicates the indicator function, which takes 1 when in charging state. represents the health index of the ith device at time t, represents the absolute value of the acceleration of the jth AGV at time t, Indicates the status of the AGV. represents the charge state of the j-th AGV at time t, k represents the k-th task, It represents the path node at time t when the j-th AGV performs the k-th task. It represents the path node at time t when the lth AGV performs the mth task. Indicates the AGV number. Indicates the time, Indicates the AGV number, represents the state of the j-th AGV at time t, Indicates the total number of charging piles.

5. The task allocation method based on hybrid game genetic algorithm according to claim 4 is characterized in that: The expression of the time and space constraints of the process connection is: The expression of the multi-agent game equilibrium constraint is: In the formula, represents the start time of task k, Represents the arrival time of the precursor material p, p belongs to the set of precursor materials of task k , represents the set of precursor materials for task k, Indicates the total number of materials currently involved. represents the indicator function, represents the volume of material i, represents the maximum capacity of storage location c, Indicates the end time of the mth process on the equipment side, It indicates the time when AGV reaches the mth process. Indicates the standard processing time of the current process. represents the Shapley value of the device agent, represents the Shapley value of the AGV agent, Indicates the Shapley value of Energy Agent, represents the profit matrix of the three-party agents at time t, represents the mixed strategy space, represents the utility function of agent i, represents the Frobenius norm of the profit matrix at time t and time (t-1), represents the mixed strategy Nash equilibrium point, Indicates that agent i adopts a balanced strategy among other agents , adopt your own strategy The value of the utility function when .

6. The task allocation method based on hybrid game genetic algorithm according to claim 5 is characterized in that: The process of solving the collaborative optimization model based on the hybrid game genetic algorithm includes: The chromosome is divided into the equipment selection gene segment, the path planning gene segment, the energy consumption gene segment and the game strategy gene segment to obtain a four-segment dynamic coding structure; Constructing a multi-agent game model, and dynamically adjusting the population corresponding to the four-stage dynamic coding structure to move toward the Pareto frontier based on updating the payoff matrix and Shapley value of the multi-agent game model; The optimal solution for task allocation is obtained by adjusting the crossover operator and dynamic mutation operator in the process of the population moving toward the Pareto front through an adaptive crossover-mutation mechanism.

7. The task allocation method based on hybrid game genetic algorithm according to claim 6 is characterized in that: The calculation expression of the Shapley value is: In the formula, represents the agent currently calculating the Shapley value, represents the Shapley value corresponding to agent a, represents the set of three intelligent agents {MA, AGV, EA}, represents the set of all remaining agents after removing agent a from set A, express Any subset of Represents the marginal contribution of agent a to the alliance S.

8. The task allocation method based on hybrid game genetic algorithm according to claim 6 is characterized in that: The adaptive crossover-mutation mechanism includes Pareto frontier detection crossover and dynamic mutation.

9. The task allocation method based on hybrid game genetic algorithm according to claim 8 is characterized in that: Adopting three-level mutation intensity control adaptive crossover mutation mechanism; Among them, the three levels of variation strength include: elite layer, middle layer and worst layer.

10. A task allocation system based on a hybrid game genetic algorithm, characterized in that: For implementing the task allocation method according to claim 1, the system comprises: Multi-source data perception layer, used to obtain task allocation data based on industrial Internet of Things and edge computing technology; A collaborative optimization model layer, used to construct a collaborative optimization model of task allocation-path planning-energy consumption control based on the task allocation data; Hybrid game genetic algorithm engine, used to integrate Nash bargaining mechanism and improved genetic algorithm, design dynamic chromosome encoding strategy, Pareto frontier detection operator, and multi-agent benefit matrix; The real-time feedback control layer is used to implement simulation verification of scheduling schemes on the digital twin platform, and to issue dynamic adjustment instructions through edge computing nodes.

Images