Robot manufacturing workshop scheduling method and device based on coalition game

Abstract

The invention provides a robotized manufacturing workshop scheduling method and device based on a coalition game. The method provided by the invention comprises the following steps: establishing an optimization model taking minimization of maximum completion time as a single target; performing process sorting optimization based on the optimization model and particle swarm optimization, encoding the process sorting into discrete particles, updating the particle speed through an exchange operation, evaluating a process sequence corresponding to the particles with the maximum completion time, iteratively updating individual and global optimal sequences, and obtaining an initial process execution sequence; based on the optimization model and the initial process execution sequence, adopting an alliance formation game to carry out robot task allocation optimization, and through iterative optimization of a task utility function and a robot utility function, controlling a robot to adjust alliance members according to utility to obtain a robot alliance allocation scheme; and based on the initial process execution sequence and the robot alliance allocation scheme, dynamically adjusting the process execution sequence and the robot alliance allocation scheme through an iterative loop to obtain an optimal scheduling scheme.

Description

Technical Field

[0001] This application relates to the field of workshop robot scheduling technology, and in particular to a robotized manufacturing workshop scheduling method and apparatus based on alliance game theory. Background Technology

[0002] With the rapid development of intelligent manufacturing, automobile assembly workshops are transforming from traditional rigid assembly lines to a fully mobile collaborative assembly model. In this process, the importance of robot scheduling is becoming increasingly prominent. This is because mobile robots with different characteristics need to be dynamically combined to form temporary assembly units to collaboratively complete key processes such as battery pack installation and door assembly. Efficient scheduling can ensure that the feature set of the robot alliance is strictly matched with the assembly task requirements, which is the core link to ensure smooth production and improve collaborative assembly efficiency.

[0003] Currently, mainstream practices employ mobile robot alliances to complete all assembly tasks. This involves mobile robots with different characteristics (such as heavy-load handling robots and high-precision positioning robots) dynamically combining to form temporary assembly units, collaboratively completing key processes. For scheduling methods, mainstream approaches such as particle swarm optimization, genetic algorithms, or mixed-integer programming are commonly used, and these can be extended to multi-objective optimization. However, existing robot scheduling methods have significant limitations: First, they lack feature constraints, failing to establish a quantitative matching model between robot features (such as end-effector load stiffness and visual positioning capabilities) and assembly task requirements, potentially leading to optimization schemes failing due to feature mismatches. Second, they fail to consider alliance effects, assuming independent robot operation and neglecting the alliance effect of complementary robot features, thus missing the optimization space for improving task completion through alliance reconstruction. Third, they separate placement and scheduling, optimizing robot placement planning and task scheduling separately, failing to consider the impact of mobile robot position changes on alliance formation, resulting in frequent spatiotemporal conflicts between placement and scheduling schemes, and a significant decrease in actual assembly cycle time compared to theoretical values. These problems lead to frequent resource mismatches in actual production, such as "robot overcapacity" or "task inability to be executed," which restricts the improvement of collaborative assembly efficiency.

[0004] Therefore, there is an urgent need for a method to achieve three-dimensional collaboration of layout planning, scheduling optimization and feature matching, so as to provide an efficient collaborative assembly solution for automobile assembly workshops in static scenarios. Summary of the Invention

[0005] In view of this, this application provides a robotized manufacturing workshop scheduling method and device based on alliance game theory, which can realize three-dimensional collaboration of layout planning, scheduling optimization and feature matching, and provide an efficient collaborative assembly solution for automobile final assembly workshops in static scenarios.

[0006] Specifically, this application is implemented through the following technical solution:

[0007] The first aspect of this application provides a method for scheduling robotic manufacturing workshops based on alliance game theory, the method comprising:

[0008] An optimization model is established with the single objective of minimizing the maximum completion time. The optimization model includes an objective function and constraints, including process sequence constraints, time logic constraints, robot allocation constraints, and robot task sequence constraints.

[0009] Based on the optimization model, the process ordering optimization is performed based on particle swarm optimization. The process ordering is encoded as discrete particles. The particle velocity is updated through swapping operations. The process sequence corresponding to the particle is evaluated with the maximum completion time as the fitness function. The individual and global optimal sequences are iteratively updated to obtain the initial process execution sequence.

[0010] Based on the optimization model and the initial process execution sequence, a coalition formation game is used to optimize robot task allocation. A coalition formation game model is constructed, and through iterative optimization of the task utility function and the robot utility function, the robot is controlled to adjust coalition members according to utility, thereby obtaining a robot coalition allocation scheme that matches the initial process execution sequence.

[0011] Based on the initial process execution sequence and the robot alliance allocation scheme, the process execution sequence and the robot alliance allocation scheme are dynamically adjusted through iterative loops until the optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained.

[0012] The second aspect of this application provides a robotic manufacturing workshop scheduling device based on alliance game theory, the device comprising an establishment module, a processing module, and an adjustment module;

[0013] The establishment module is used to establish an optimization model with the single objective of minimizing the maximum completion time. The optimization model includes an objective function and constraints, including process sequence constraints, time logic constraints between processes and features, and robot resource allocation constraints.

[0014] The processing module is used to optimize the process order based on the optimization model and particle swarm optimization. The process order is encoded as discrete particles, the particle velocity is updated through swapping operations, the process sequence corresponding to the particle is evaluated with the maximum completion time as the fitness function, and the individual and global optimal sequences are iteratively updated to obtain the initial process execution sequence.

[0015] The processing module is further configured to optimize robot task allocation using alliance formation game based on the optimization model and the initial process execution sequence, construct an alliance formation game model, and control the robot to adjust alliance members according to utility through iterative optimization of task utility function and robot utility function, thereby obtaining a robot alliance allocation scheme that matches the initial process execution sequence.

[0016] The adjustment module is used to dynamically adjust the process execution sequence and the robot alliance allocation scheme through iterative loops based on the initial process execution sequence and the robot alliance allocation scheme, until the optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained.

[0017] The robotic manufacturing workshop scheduling method and apparatus based on coalition game theory provided in this application form a closed-loop optimization system through a complete process of "modeling-initial sorting-coalition allocation-dynamic iteration". First, based on a "single-objective optimization model", the core objective of minimizing the maximum completion time is clearly defined, and the scheduling feasibility boundary is defined through constraints such as process sequence and time logic. This setting ensures from the source that the scheduling scheme conforms to the basic rules of workshop production (such as process sequence relationships and robot task capability limitations), avoiding infeasible scheduling and providing a valid solution space for subsequent optimization. Second, a "particle swarm optimization for process sorting" method is adopted, encoding processes as discrete particles and updating their speed through exchange operations. Iterative optimization is performed using the maximum completion time as the fitness function. This process quickly generates the initial process execution sequence, solving the combinatorial optimization problem of process sorting. Compared with random sorting, it is easier to approach a better solution, providing an efficient initial process for subsequent robot allocation. The first step involves a sequence framework. Secondly, through "alliance-based game theory for task allocation," a game model is constructed based on the initial process sequence. Iterative optimization of task and robot utility functions dynamically adjusts alliance members, ensuring that the allocation scheme meets both process skill requirements and improves execution efficiency through alliance collaboration. This step achieves precise matching between robot resources and process needs, avoiding efficiency losses caused by resource mismatch. Finally, using "iterative cyclical dynamic adjustment" as the core approach, the process sequence and robot alliance are continuously optimized through alternating operation of particle swarm optimization and alliance game theory until the preset target of maximum completion time is met. This closed-loop mechanism overcomes the static limitations of the initial scheme, enabling the scheduling scheme to dynamically adapt to the collaborative relationship between processes and resources, continuously approaching the global optimum. Overall, this embodiment achieves efficient (minimum maximum completion time), precise (skill and task matching), and adaptive (dynamically adaptable to different order sizes) scheduling of workshop robots through a progressive approach of "constraint modeling to ensure feasibility, intelligent algorithms to improve optimization efficiency, alliance game theory to achieve precise resource matching, and dynamic iteration to approach the global optimum." This provides a feasible solution for intelligent scheduling of workshop production, significantly improving production efficiency and resource utilization. Attached Figure Description

[0018] Figure 1 A flowchart of the robotized manufacturing workshop scheduling method based on alliance game theory provided in Embodiment 1 of this application;

[0019] Figure 2This is a schematic diagram of the structure of the robotized manufacturing workshop scheduling device based on alliance game theory provided in Embodiment 2 of this application. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0021] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0022] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0023] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0024] Figure 1 The flowchart illustrates the robotized manufacturing workshop scheduling method based on alliance game theory provided in Embodiment 1 of this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0025] S101. Establish an optimization model with the single objective of minimizing the maximum completion time.

[0026] The optimization model includes an objective function and constraints, including process sequence constraints, time logic constraints, robot allocation constraints, and robot task sequence constraints.

[0027] Specifically, the optimization model refers to an abstract model constructed mathematically to describe the objective and constraint relationships of robot scheduling problems in automotive assembly workshops. By clearly defining "what objectives need to be achieved" and "what rules restrict them," the optimization model transforms the scheduling requirements in actual production into a quantifiable and solvable logical framework, providing a standardized decision-making basis for subsequent algorithm optimization. In the workshop scenario, the objective function (minimizing the maximum completion time) provides core optimization indicators for robot scheduling, ensuring that all scheduling strategies revolve around "improving overall production efficiency." Constraints eliminate solutions that do not conform to production logic (such as reversed process order or mismatched robot capabilities), ensuring that the final scheduling solution is executable in the actual workshop environment. The optimization model provides a unified evaluation standard (such as whether constraints are satisfied and whether the objective function value is better) for process order optimization in particle swarm optimization and robot allocation optimization in coalition game theory, enabling the algorithm to efficiently search for the optimal solution within a reasonable range.

[0028] Furthermore, the maximum completion time (makespan) refers to the maximum time taken to complete all products (or all processes) in a workshop. For example, if a workshop is producing three car models simultaneously, with car A taking 80 minutes, car B taking 90 minutes, and car C taking 75 minutes, then the maximum completion time is 90 minutes. The maximum completion time directly reflects the overall production efficiency of the workshop and is a core indicator for evaluating the quality of a scheduling scheme. Constraints refer to the production rules and logical restrictions that must be followed during the scheduling process to ensure that the scheduling scheme meets process requirements, resource capabilities, and time logic. The constraints in this application include the following four categories: process sequence constraints, time logic constraints, robot allocation constraints, and robot task sequence constraints.

[0029] The process sequence constraint represents the order of processes within the same product, as well as the relationship between the process and the final completion time of the product; the time logic constraint represents the time inclusion relationship between the process and its subordinate features; the robot allocation constraint restricts the matching of robot capabilities, the uniqueness of feature allocation, and the matching of alliance size and feature quantity; the robot task sequence constraint represents the time order in which the same robot executes multiple features. The process sequence constraints characterize the order of processes within the same product, as well as the relationship between processes and the final completion time of the product. These constraints include: the completion time of a preceding process for the same product is no greater than the start time of the following process; the completion time of any process for a product is no greater than the final completion time of the product. The time logic constraints include: the start time of a process for a product is no greater than the start time of any feature included in that process; the completion time of any feature included in a process for a product is no greater than the overall completion time of the process; and all time parameters are non-negative. The time parameters include at least the start time, completion time, and final completion time. The robot allocation constraints include: matching the capability identifier and allocation identifier of the robot's execution features; each feature of each process being executed by one robot; and the size of the robot consortium not exceeding the number of features in the process. The robot task sequence constraints include: matching the robot's execution time of features with the time window of the process; the start time of the same robot executing a preceding feature being less than the start time of the following feature; and the same robot executing multiple features having the same allocation identifier.

[0030] For example, in one embodiment, the objective function of the optimization model can be expressed as:

[0031] ;

[0032] Among them, the The maximum completion time; The completion time; the stated Index for product type, .

[0033] The constraints of the optimization model can be expressed as:

[0034] ;

[0035] ;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] ;

[0046] in, , Index for product task types; , For products Process index; , For products Process Feature index; Index for robots; For products The total number of processes; For products Process The number of features; For products Process Features Available robots The value is 1 if the action is executed, and 0 otherwise. For robots Execution Product Process Chinese characteristics The time consumption index; For products Process Chinese characteristics Standard assembly time; The average moving speed of the chassis AGV carrying the products; For robots Average moving speed; For products The direction vector of movement; For loading products The starting position of the chassis AGV; For a sufficiently large number; For products Process Features Assembly start time; For products Process Features Assembly completion time; For products Process Assembly start time; For products Process Assembly completion time; For products The final completion time; For products Process Features By robots The value is 1 if the assembly is complete, and 0 otherwise. If the robot Assemble the product first Process Features Reprocessed products Operation Features If it is 1, then it is 1; otherwise, it is 0.

[0047] The method provided in this embodiment uses an optimization model with the objective function of minimizing the maximum completion time, which includes constraints on process sequence, time logic, robot allocation, and robot task sequence. First, the objective function provides a unique and quantifiable core indicator for scheduling optimization, giving a unified standard for evaluating the merits of process sequencing in particle swarm optimization and robot allocation in coalition games, thus avoiding directional confusion caused by multi-objective conflicts. The four types of constraints define the feasible boundaries of the scheduling scheme from a production logic perspective. Process sequence constraints and time logic constraints ensure that scheduling conforms to the process flow, robot allocation constraints prevent resource misallocation, and robot task sequence constraints avoid robot conflicts, significantly improving the algorithm's search efficiency. At the effect level, constraints filter out schemes that do not conform to actual production logic, ensuring the actual executability of the scheduling scheme. The objective function drives the algorithm to search in a more efficient direction, while constraints ensure that efficiency is not at the expense of feasibility, achieving a balance between efficiency and feasibility. Furthermore, the optimization model treats shop floor scheduling as a holistic system, focusing on global indicators and relating the dependencies between various links, pushing the algorithm out of local optima traps and ultimately obtaining the globally optimal solution. In summary, the optimization model, through "objective function defining direction and constraints defining boundaries," provides clear criteria for the algorithm, ensuring the executability of the scheme and achieving a unity of "optimal efficiency" and "production feasibility," providing a scientific and reliable decision-making basis for collaborative assembly in automotive assembly workshops.

[0048] S102. Based on the optimization model, the process order optimization is performed based on particle swarm optimization. The process order is encoded as discrete particles. The particle velocity is updated through exchange operations. The process sequence corresponding to the particle is evaluated with the maximum completion time as the fitness function. The individual and global optimal sequences are iteratively updated to obtain the initial process execution sequence.

[0049] Specifically, Particle Swarm Optimization (PSO) searches for optimal solutions in the solution space by simulating information sharing and cooperation among individuals in a swarm. In this application, PSO is used to optimize the process sequencing of assembly lines for a workshop robot. PSO is closely related to the optimization model; the optimization model provides the constraints and objectives for PSO, while PSO searches and optimizes the process sequencing within the framework of the optimization model. The fitness function (maximum completion time) used to evaluate particle performance and the rules followed during the search process both originate from the objective function and constraints of the optimization model.

[0050] Furthermore, the initial process execution sequence refers to the first process execution order scheme obtained through iterative search using the particle swarm optimization algorithm, under the premise of satisfying the constraints of the optimization model. It serves as the basis for subsequent robot alliance allocation. It is the globally optimal sequence based on the current optimization stage, providing an initial process execution framework for subsequent robot task allocation optimization.

[0051] In specific implementation, obtaining the initial process execution sequence includes:

[0052] (1) Encode the execution order of all processes into discrete particles; the position of each particle corresponds to a global process execution sequence, and the velocity of each particle represents the tendency of process sequence to be exchanged.

[0053] Specifically, all processes to be executed are arranged by number or identifier to form an ordered sequence as the code for discrete particles; the position of each particle directly corresponds to a complete sequence of process execution, and the velocity of the particles represents the probability of exchange between different process positions through a preset probability parameter.

[0054] (2) Initialize the position and velocity of the particle swarm; the position is a randomly generated sequence of operations, and the velocity is an initial exchange probability parameter.

[0055] Specifically, determine the particle swarm size (e.g., 50 particles, i.e., 50 different process sequences); for each particle, randomly select all unique identifiers from all processes (e.g., numbers 1 to N, where N is the total number of processes), shuffle them to form an unordered sequence, and use this as the initial position of the particle. For example, if there are 6 processes, the initial position of a particle might be randomly generated as [4, 1, 6, 2, 5, 3], indicating that the initial process execution order for this particle is process 4 → process 1 → process 6 → process 2 → process 5 → process 3; repeat the above process to generate random process sequences consistent with the particle swarm size, ensuring that the initial positions (process sequences) of each particle are different or as diverse as possible. Furthermore, each particle is configured with a velocity array of the same length as the position sequence (length equal to the total number of processes); each element in the velocity array is an initial exchange probability parameter, which can be generated by preset rules: for example, uniformly set to 0.5 (i.e., 50% initial exchange probability), or randomly generated in a uniform distribution within the interval [0, 0.8] (e.g., the velocity array of a certain particle is [0.3, 0.6, 0.2, 0.7, 0.4]).

[0056] (3) Based on the constraints of the optimization model, calculate the maximum completion time of the process sequence corresponding to each particle, and use the maximum completion time as the fitness function value.

[0057] Specifically, extract the process sequence corresponding to the current particle and clarify the execution order of all processes (e.g., process 1 → process 3 → process 2 → ...). Check each sequence to see if it meets the four types of constraints of the optimization model: Verify process sequence constraints: confirm that the processes of the same product are arranged in the preset process order (e.g., "chassis installation" must precede "body assembly"); Verify time logic constraints: ensure the time inclusion relationship between the process and its subordinate features (e.g., the start time of the process is not later than the start time of its first feature, and the completion time of the process is not earlier than the completion time of its last feature); Verify robot allocation constraints: pre-allocate robots according to the default rules (e.g., each feature is assigned to any robot that meets the skill requirements), ensuring that no feature is assigned to a mismatched robot; Verify robot task sequence constraints: ensure that multiple features pre-allocated to the same robot are arranged in chronological order without overlap or conflict. For a sequence that satisfies all constraints, the maximum completion time is simulated and calculated: based on the standard working time of each process, the pre-allocated robot movement time, and the characteristic operation time, the start time and completion time of each process are calculated in sequence; the completion time of all processes is tracked, and the time with the largest value is determined as the maximum completion time of the sequence, and the maximum completion time is directly determined as the fitness function value of the current particle; if the sequence does not satisfy the constraints, an extremely large fitness value is assigned (considered an invalid solution).

[0058] (4) By retaining the historical exchange tendency, combining the learning factors of the individual optimal sequence and the global optimal sequence, a random number within a fixed interval is introduced to dynamically adjust the exchange tendency, update the particle velocity, and generate new velocity parameters.

[0059] Specifically, the process of dynamically adjusting the exchange tendency by retaining historical exchange tendencies, combining the learning factors of the individual optimal sequence and the global optimal sequence, and introducing random numbers within a fixed interval to update the particle velocity and generate new velocity parameters includes: retaining the historical process sequence exchange tendency represented by the particle's current velocity, and controlling the degree of retention of the historical process sequence exchange tendency through an inertia weight parameter; calculating the individual learning factor corresponding to the difference between the particle's current position and the individual optimal sequence; calculating the social learning factor corresponding to the difference between the particle's current position and the global optimal sequence; introducing two random numbers that follow a uniform distribution within a fixed interval to adjust the individual learning factor and the social learning factor respectively; and fusing the updated historical process sequence exchange tendency, the adjusted individual learning factor, and the adjusted social learning factor to generate new velocity parameters.

[0060] In specific implementation, the historical process sequence exchange tendency represented by the particle's current velocity is retained. The current velocity parameter is multiplied by a preset inertia weight parameter, and the product is the retained historical exchange tendency portion. The process sequence corresponding to the particle's current position is compared with the individual optimal sequence to identify process pairs with different positions. The first difference of these process pairs is calculated, and the first difference is multiplied by a first adjustment amount (the first adjustment amount adjusts the intensity of the exchange probability update based on the individual optimal sequence) to obtain the individual learning factor. Similarly, the process sequence corresponding to the particle's current position is compared with the global optimal sequence to identify process pairs with different positions. The second difference of these process pairs is calculated, and the second difference is multiplied by a second adjustment amount (the second adjustment amount adjusts the intensity of the exchange probability update based on the global optimal sequence) to obtain the social learning factor. Further, two random numbers uniformly distributed within the interval [0, 1] are generated and multiplied by the individual learning factor and the social learning factor respectively to obtain the adjusted individual learning factor and social learning factor. The retained historical exchange tendency portion, the adjusted individual learning factor, and the adjusted social learning factor are added together to obtain the new velocity parameter.

[0061] For example, in one embodiment, the speed update process can be represented as:

[0062] ;

[0063] Among them, the The updated velocity parameters; The inertia weight parameter; The tendency to exchange historical process sequences; the aforementioned This is the first adjustment amount; For the individual optimal sequence; the The current position of the particle; , Two random numbers; This is the second adjustment amount; This is the globally optimal sequence; For individual learning factors; It is a social learning factor.

[0064] The method provided in this embodiment retains historical exchange tendencies by using inertial weight parameters to preserve the historical exchange tendencies in the current velocity of particles. A larger weight indicates stronger continuity of the historical adjustment pattern, preventing particles from frequently and drastically changing the process sequence during iterations. This avoids the scheduling scheme deviating from the potential optimal solution direction due to over-adjustment, ensuring the stability of the scheduling optimization process. Simultaneously, when the inertial weight is appropriate, it can both retain effective adjustment experience from the previous stage and leave room for the introduction of new information, thus achieving a balance between continuing reasonable experience and accepting new optimization directions in the robot scheduling scheme. Combining the exchange probabilities of individuals and the global optimum, the method calculates the differences between the current sequence and the individual optimal and global optimal sequences to obtain individual learning factors and social learning factors. The learning factor enables each particle to focus on its own historical optimal scheduling logic, avoiding repeated trial and error and accelerating local optimization. The social learning factor promotes all particles to move towards the direction recognized by the group, guiding the scheduling scheme to converge towards the global optimum and reducing ineffective exploration. The combination of the two allows the robot scheduling scheme to absorb individual experience and learn from the wisdom of the group, improving the efficiency of the final scheme. The introduction of random numbers in a fixed interval, using uniformly distributed random numbers in [0,1] to adjust the individual learning factor and the social learning factor, introduces randomness, breaks the limitation of particles adjusting according to a fixed pattern, gives rise to new combinations of process sequences, increases the diversity of scheduling schemes, avoids the particle swarm converging to local optima too early, ensures that the algorithm explores a wider range of scheduling possibilities, and determines a better robot collaboration scheme. In summary, the speed update mechanism, through the combination of "inertia to preserve stability, learning to guide the optimization direction, and randomness to enhance diversity," enables the particle swarm algorithm to both stably converge to an efficient scheduling direction and avoid getting trapped in local optima in the optimization of robot scheduling in the workshop. The final generated scheduling scheme is more likely to achieve reasonable allocation of robot resources and efficient and smooth process connection, thereby shortening the total processing cycle and improving workshop production efficiency.

[0065] (5) Calculate the exchange probability of the process sequence based on the updated velocity parameters, and exchange the process sequence corresponding to the particle position based on the exchange probability to obtain a new process execution sequence.

[0066] Specifically, the step of calculating the exchange probability of the process sequence based on the updated velocity parameters, and exchanging the process sequences corresponding to the particle positions based on the exchange probabilities to obtain a new process execution sequence, includes: performing a probability transformation on the updated velocity parameters to map the velocity parameters to a unified range to obtain the exchange probability of the process sequence; for the current process sequence corresponding to the particle position, filtering process position pairs to be exchanged according to the exchange probabilities; process position pairs with exchange probabilities higher than a preset threshold are selected first; performing an exchange operation on the selected process position pairs, swapping the processes at the two positions to generate a temporary process sequence; checking whether the temporary process sequence satisfies the process sequence constraints in the optimization model, if it does, then determining the temporary process sequence as the updated particle position; if it does not, then re-filtering the exchange position pairs and performing the exchange, until a new process execution sequence that meets the process sequence constraints is generated.

[0067] In practice, the updated velocity parameters (e.g., velocity values ​​in array form) are transformed using the Sigmoid function with equal probability, mapping each velocity value to the (0, 1) interval. The resulting value represents the exchange probability of the corresponding process position. For the particle's current process sequence (e.g., [3, 1, 5, 2, 4]), all possible process position pairs (e.g., (1, 2), (1, 3)) are traversed, and the exchange probability corresponding to each position pair is extracted. A preset threshold (e.g., 0.5) is set, and process position pairs with exchange probabilities higher than this threshold are included in the priority selection range. From this range, position pairs to be exchanged are randomly selected or sorted by probability. The selected process position pairs (e.g., position 2 and position 1) are exchanged, swapping the processes at the two positions in the original sequence (e.g., [3, 2, 5, 1, 4]), generating a temporary process sequence. Based on the process sequence constraints in the optimization model (e.g., the preset order of processes for the same product), the temporary process sequence is checked to see if it meets the requirements. If the temporary process sequence meets the constraints, it is directly used as the updated particle position; otherwise, a new position pair is selected from the position pairs with a swap probability higher than the preset threshold for swapping, and the operation is repeated until a new process execution sequence that meets the process sequence constraints is generated.

[0068] For example, in one embodiment, the process of performing a probability transformation on the updated velocity parameters can be represented as follows:

[0069] ;

[0070] Among them, the The Sigmoid activation function; The updated velocity parameters; Let be the probability of exchanging the process order.

[0071] The method provided in this embodiment first maps the updated speed parameters to a unified range through probability transformation, converting the abstract "exchange tendency" into a specific "exchange probability," providing a quantitative basis for process adjustment. Next, it filters the pairs of positions to be exchanged based on the exchange probability, prioritizing combinations with high probabilities to focus adjustments on process combinations more likely to improve efficiency. Then, it performs exchanges on the selected positions to generate a temporary sequence, and checks the process order constraints to ensure the new sequence conforms to the process logic; if it does not, the operation is repeated until compliance is achieved. In terms of effectiveness, probability transformation avoids the blindness of process adjustment, ensuring each exchange is based on a clear quantitative probability, thus improving the targeting of optimization. Prioritizing high-probability position pairs accelerates the algorithm. By converging towards a better solution, reducing ineffective exploration, and shortening the time to determine an efficient process sequence, and by strictly checking constraints, the generated process sequence is fundamentally guaranteed to conform to the actual production process sequence in the workshop (such as "assemble the chassis first and then the body"), eliminating assembly errors or production interruptions caused by process reversal. In summary, this update method not only ensures that process adjustments are directed towards improving efficiency through probability guidance, but also guarantees the actual executability of the sequence through constraint checks. The final generated process execution sequence can shorten the overall assembly cycle and adapt to the physical logic of workshop production, providing a reasonable and feasible basic framework for subsequent robot alliance allocation, thereby improving the efficiency and reliability of the entire workshop robot scheduling scheme.

[0072] (6) Compare the fitness function values ​​of the new process execution sequence with the historical best sequence of the particles, update the individual best sequence, compare all individual best sequences, and update the global best sequence.

[0073] Specifically, the fitness function value of the new process sequence is compared with the fitness value of the particle's historical best sequence. If the new value is better, the individual best sequence is updated. The individual best sequences of all particles are summarized, and the sequence with the best fitness value is selected and updated as the global best sequence.

[0074] (7) Repeat the steps from calculating to comparing fitness function values ​​until the preset maximum number of iterations is reached, and use the final global optimal sequence as the initial process execution sequence.

[0075] Specifically, repeat steps (3) to (6), recalculate the fitness function value, update the speed and position, and adjust the optimal sequence in each iteration until the number of iterations reaches the preset maximum value. The global optimal sequence at this time is the initial process execution sequence.

[0076] The method provided in this embodiment first encodes the process execution sequence as discrete particles, transforming the abstract scheduling problem into a quantifiable algorithm optimization problem. The position and velocity of each particle correspond to the process sequence and exchange tendency, respectively, providing a basic framework for subsequent iterations. Next, the positions and velocities of the particle swarm are initialized. By randomly generating process sequences and initial exchange probabilities, the diversity of initial solutions is ensured, avoiding the algorithm from getting trapped in local optima. Then, the maximum completion time is calculated as the fitness function based on the constraints of the optimization model, providing a clear criterion for judging the quality of particles. By retaining historical exchange tendencies, combining individual and global optimal information, and introducing random numbers to update the velocity, particles can both continue effective adjustment experience and absorb collective wisdom while maintaining exploration diversity. When adjusting the process sequence based on the exchange probability, probability quantification and constraint checks ensure that the adjustment direction is reasonable and conforms to the process logic. Finally, by iteratively updating the individual and global optimal sequences, the optimal solution is gradually approached. In terms of effectiveness, the encoding method achieves algorithmic adaptation for scheduling problems, enabling complex process sequencing to be solved efficiently through particle swarm optimization. The diversity of initial solutions expands the search range, laying the foundation for determining the global optimum. Using the maximum completion time as the fitness function ensures that optimization always revolves around the core objective of shortening the production cycle. The speed update mechanism balances the algorithm's exploration and convergence capabilities, avoiding blind adjustments while quickly moving towards a better solution. The probabilistic approach and constraint checks for process sequence adjustments ensure that the sequence generated in each iteration has both optimization potential and conforms to actual production logic. The iterative update mechanism, through continuous optimization, ultimately yields an initial process execution sequence that is both efficient and feasible. In summary, this setup, through a complete process of "problem transformation—diversified initialization—directional optimization—constraint guarantee—iterative convergence," provides a scientific, efficient, and practical initial process framework for workshop robot scheduling that meets actual production needs. It ensures that subsequent robot allocation has a reasonable process foundation and directly improves scheduling efficiency by optimizing the maximum completion time, providing key support for ultimately achieving efficient and collaborative workshop robot scheduling.

[0077] S103. Based on the optimization model and the initial process execution sequence, the robot task allocation is optimized using alliance formation game theory. An alliance formation game model is constructed. Through iterative optimization of the task utility function and the robot utility function, the robot is controlled to adjust the alliance members according to the utility, thereby obtaining a robot alliance allocation scheme that matches the initial process execution sequence.

[0078] Specifically, alliance formation game theory is a game theory model that refers to the process by which multiple entities (such as robots) autonomously choose partners to form a collaborative group (alliance), and adjust alliance members based on preset utility rules (such as task completion efficiency, resource consumption, etc.) to eventually reach a stable state. In this model, each entity decides whether to participate in the alliance based on the changes in benefits (utility) brought about by joining or leaving the alliance. The final alliance must satisfy the stability condition that "no entity has an incentive to leave unilaterally." In workshop robot scheduling, alliance formation game theory is used to solve the problem of "how to allocate tasks (processes and subordinate characteristics) in the initial process execution sequence to robots." By constructing a game theory model, robots can autonomously adjust alliance members based on task utility (such as task completion time and difficulty) and their own utility (such as load balancing and capability matching), achieving efficient matching between tasks and robots, and ensuring that the initial process sequence can be executed optimally in robot collaboration (such as shortening the total project duration and reducing resource waste).

[0079] Furthermore, an alliance refers to a collaborative group of multiple robots tasked with completing a specific task (such as several related processes or features) within an initial process execution sequence. For example, to complete the "door assembly" process, an alliance might be formed consisting of robot A (with welding capabilities), robot B (responsible for handling), and robot C (with high-precision positioning capabilities), working together to complete all the feature operations of this process. The initial process execution sequence defines the sequential order of all processes (e.g., process 1 → process 2 → process 3), while the alliance is formed based on the specific process tasks within that sequence—the alliance's task scope and collaborative rhythm must strictly match the process order and time logic of the initial sequence (e.g., the alliance must complete process 1 before starting process 2). In other words, the initial process execution sequence is the "task framework" formed by the alliance, while the alliance is the "execution vehicle" that realizes the sequence. The robot alliance allocation scheme refers to the robot collaboration scheme that is ultimately determined through a game of alliance formation and perfectly matches the initial process execution sequence. The robot alliance allocation scheme includes which robot alliances are responsible for executing each process (and its subordinate features), the specific division of labor among the robots within the alliance (e.g., who performs welding, who is responsible for positioning), and the task connection time between alliances.

[0080] In specific implementation, based on the optimization model and the initial process execution sequence, a coalition formation game is used to optimize robot task allocation. A coalition formation game model is constructed, and through iterative optimization of the task utility function and the robot utility function, the robot is controlled to adjust coalition members according to utility, resulting in a robot coalition allocation scheme that matches the initial process execution sequence, including:

[0081] (1) For each process in the initial process execution sequence, select robots that meet the skill requirements of each process to form a candidate robot set.

[0082] Specifically, each process in the initial process execution sequence is traversed, the skill requirements for each process are extracted (such as welding, handling, positioning, etc.), and robots with the corresponding skills are selected from all robots and these robots are classified into the candidate robot set for that process.

[0083] (2) For each robot in the candidate robot set, freely form an alliance; the size of the alliance shall not exceed the feature quantity constraint of the corresponding process.

[0084] Specifically, for each process's candidate robot set, robots are allowed to autonomously select partners to form alliances, while ensuring that the number of robots in each alliance does not exceed the number of features contained in that process (e.g., if a process has 3 features, its alliance size can be up to 3 robots).

[0085] (3) Design a task utility function to evaluate the overall performance of each alliance's execution process.

[0086] Specifically, a task utility function is a mathematical model or calculation rule used to quantitatively evaluate the overall performance of a consortium when executing a specific process. Its input typically includes various indicators related to process execution (such as completion time, resource consumption, collaboration efficiency, and error rate). By setting weights or functional relationships, these indicators are transformed into a single utility value to intuitively reflect the consortium's performance in executing the process. In this embodiment, the task utility function provides an objective basis for selecting the optimal consortium: by calculating the utility values ​​of different candidate consortia executing the same process, the overall performance of each consortium can be quantitatively compared, thereby accurately selecting the consortium with the highest utility value as the optimal execution consortium for the current process. This ensures that process execution achieves a superior level in multiple dimensions such as efficiency, cost, and quality, laying the foundation for the subsequent formation of a robot consortium allocation scheme.

[0087] In specific implementation, the task utility function is designed to evaluate the overall performance of each alliance's execution process, including: determining a first time parameter and a second time parameter of the task utility function; the first time parameter is the time length from the start of the entire assembly process to the alliance completing the current process; the second time parameter is the time length from the earliest completion time of the previous task among all robots participating in the current process to the alliance completing the current process; different constant weight values ​​are configured for the first time parameter and the second time parameter respectively; the weight value of the first time parameter is used to adjust the degree of influence of the entire assembly process on the alliance performance, and the weight value of the second time parameter is used to adjust the degree of influence of robot movement and operation time on the alliance performance; the weighted combination of the first time parameter and the second time parameter with the corresponding constant weight value is determined as the task utility function.

[0088] Specifically, the time from the start of the entire assembly process to the end of the corresponding step completed by the current alliance is recorded and used as the first time parameter. Similarly, the earliest completion time of the previous task among all robots participating in the current step is determined, and the time from that point to the end of the current step is calculated; this time is used as the second time parameter. Further, a constant weight value (e.g., 0.6) is assigned to the first time parameter to adjust the impact of the entire assembly process time on the alliance's performance evaluation; another constant weight value (e.g., 0.4) is assigned to the second time parameter to adjust the impact of robot movement and operation time on the alliance's performance evaluation. The first time parameter is multiplied by its corresponding constant weight value to obtain the first weighted term; the second time parameter is multiplied by its corresponding constant weight value to obtain the second weighted term; the two weighted terms are summed to obtain the task utility function.

[0089] For example, in one embodiment, the task utility function can be expressed as:

[0090] ;

[0091] Among them, the For the task utility function; the , The weight values ​​are constants; The first time parameter; This is the second time parameter.

[0092] The method provided in this embodiment incorporates a first time parameter (the time from the start of assembly to the completion of the current process) and a second time parameter (the time from the earliest completion of the previous task by the participating robot to the completion of the current process), and assigns different constant weights to both, constructing a weighted combined evaluation model. The first time parameter weight adjusts the impact of the overall process time, while the second time parameter weight adjusts the impact of robot movement and operation time. The introduction of the first time parameter allows the alliance evaluation to focus on the overall assembly progress, encouraging the alliance to choose solutions that tend to shorten the total production cycle and avoid situations where local efficiency is high but overall delays occur. The second time parameter... The introduction focuses on robot movement and operation issues to improve equipment utilization. The differentiated configuration of weights allows for adjustments to the evaluation emphasis based on actual production needs (such as prioritizing timely completion or prioritizing efficient equipment utilization), making the utility function more aligned with actual scheduling goals. Overall, this task utility function not only controls assembly progress globally but also optimizes resource utilization locally. Through quantitative evaluation, it guides the alliance to choose a better solution that balances total time and equipment idle time, providing a scientific basis for subsequent selection of the optimal alliance. Ultimately, the robot alliance allocation scheme can ensure overall assembly efficiency while reducing resource waste and improving the overall benefits of workshop robot scheduling.

[0093] (4) Calculate the task utility value of all alliances based on the task utility function, and select the alliance with the largest task utility value as the current optimal alliance.

[0094] Specifically, all the formed alliances are substituted into the task utility function to calculate their respective task utility values. The alliance with the largest value is then selected as the current optimal alliance for the current process.

[0095] (5) Calculate the utility of each robot for the current best alliance.

[0096] Specifically, the utility of each robot refers to a quantitative indicator of the benefit or value gained by a single robot participating in the execution of tasks within the current optimal alliance. The task utility value of the current optimal alliance is a quantitative assessment of the overall performance of the alliance in executing tasks, and the utility of each robot is usually related to the task utility value of the alliance—this may be the distribution of task utility value among alliance members, or it may be derived from the task utility value by combining factors such as the robot's own costs and contributions. For example, a robot's utility may equal the product of its contribution percentage in the alliance and its task utility value, minus its own participation cost. The core idea is to link the overall utility of the alliance with the benefits of individual robots to determine whether an individual is suitable to remain in the alliance.

[0097] In specific implementation, calculating the utility of each robot for the current optimal alliance includes: calculating the total utility of the current optimal alliance using the task utility function; for each robot in the alliance, removing each robot from the current optimal alliance to form a new robot set, obtaining the remaining alliance after each robot leaves; calculating the total utility of each remaining alliance using the task utility function; and determining the utility of each robot by the difference between the total utility of the current optimal alliance and the total utility of the remaining alliance after each robot leaves.

[0098] Specifically, the task utility function is called, inputting relevant parameters of the current optimal alliance (such as the first time parameter and the second time parameter), to calculate the total utility value of the alliance performing the corresponding process. For each robot in the current optimal alliance, that robot is removed from the alliance in turn, forming a new set of robots that does not contain that robot, thus obtaining the remaining alliance after the robot's departure. For each remaining alliance, the task utility function is called again, inputting its corresponding first time parameter and second time parameter, to calculate the total utility value of each remaining alliance. Further, the total utility value of the current optimal alliance is subtracted from the total utility value of the remaining alliances after each robot's departure; the difference is the robot's utility in the current optimal alliance.

[0099] For example, in one embodiment, the utility of each robot can be expressed as:

[0100] ;

[0101] Among them, the The utility of each robot; For an alliance that includes robots; the The remaining alliance after the robot leaves.

[0102] (6) Adjust the corresponding alliance members based on the utility of each robot. If the robot's utility is positive, it shall remain in the alliance; if it is negative, it shall leave the alliance and a new alliance shall be formed.

[0103] Specifically, the utility value of each robot in the current best alliance is iterated in turn. If the utility value is positive, the robot remains in the alliance; if the utility value is negative, the robot leaves the alliance. Then, a new alliance is formed by selecting robots from the candidate robot set.

[0104] (7) Repeat the steps from setting the task utility function to adjusting the alliance members until no robot leaves or joins the alliance due to utility issues, forming a robot alliance allocation scheme that matches the initial process execution sequence.

[0105] Specifically, repeat steps (3) to (6), that is, recalculate the task utility value of the new alliance, select the optimal alliance, evaluate the individual utility of the robot and adjust the alliance members until the alliance members are stable (no robot leaves or joins due to positive or negative changes in utility). The alliance combination formed at this time is the robot alliance allocation scheme that matches the initial process execution sequence.

[0106] The method provided in this embodiment first selects candidate robots with corresponding skills for each process step to ensure that the basic capabilities of alliance members are matched; allows robots to freely form alliances and limits the size to no more than the number of process features to ensure that the alliance structure is adapted to the task complexity; evaluates the alliance performance quantitatively through a task utility function and selects the alliance with the highest utility as the current optimal alliance; calculates the utility of each robot in the alliance, adjusts members according to the positive or negative utility (positive utility is retained, negative utility is removed and reorganized), and stabilizes the alliance members through iteration. Skill selection ensures that the alliance is capable of completing the process tasks and avoids task failure due to insufficient robot capabilities; free formation and size restrictions allow the alliance to flexibly adapt to process requirements without causing collaborative redundancy due to too many members; the task utility function provides an objective standard for the quality of the alliance, promoting the selection of alliances with better overall performance and improving process execution efficiency; robot utility calculation and member adjustment ensure that each robot in the alliance can make a positive contribution and avoid inefficient members dragging down the whole, while the iterative stabilization mechanism enables the alliance to finally reach a state of efficient collaboration. In summary, this setup, through a process of "capability matching, structural adaptation, quantitative evaluation, dynamic optimization, and stable convergence," creates a robot alliance allocation scheme that can accurately match the task requirements of the initial process execution sequence, achieve efficient allocation and collaboration of robot resources, and ultimately improve the overall efficiency, feasibility, and stability of robot scheduling in the workshop, ensuring a smooth and efficient production process.

[0107] S104. Based on the initial process execution sequence and the robot alliance allocation scheme, the process execution sequence and the robot alliance allocation scheme are dynamically adjusted through iterative loops until the optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained.

[0108] Specifically, the optimal scheduling scheme refers to the scheme that, after iterative and dynamic adjustments during the scheduling of robots in the workshop, ultimately satisfies two core conditions: skill matching: that is, the robot alliance participating in each process has all the skills required to perform the corresponding process, ensuring that the process can be effectively completed; minimum maximum completion time: the maximum completion time refers to the total time from the start of the entire production process to the completion of the last process. The optimal scheme should minimize this total time to maximize overall production efficiency.

[0109] In specific implementation, based on the initial process execution sequence and the robot alliance allocation scheme, the process execution sequence and the robot alliance allocation scheme are dynamically adjusted through iterative loops until the optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained. This includes: calculating the maximum completion time of the current initial process execution sequence and the corresponding robot alliance allocation scheme; determining whether the current iteration count has reached the maximum iteration count; if not, adjusting the process execution sequence based on the particle swarm optimization algorithm, and optimizing the current maximum completion time as the fitness function to obtain a new process execution sequence; based on the new process execution sequence, reconstructing the robot alliance allocation scheme through alliance formation game to form a new robot alliance allocation scheme that matches the new process execution sequence; repeating the steps from calculating the maximum completion time to reconstructing the robot alliance allocation scheme until the current iteration count reaches the maximum iteration count, and determining the scheduling scheme with the minimum maximum completion time as the optimal scheduling scheme.

[0110] Specifically, the total time from the start of the entire production process to the completion of the final step is calculated, and this time is the maximum completion time of the current robot alliance allocation scheme. In each iteration, the current iteration number is determined and compared with the maximum iteration number. If the current iteration number has not reached the maximum iteration number, the following process is repeated: Adjusting the process execution sequence based on the particle swarm optimization algorithm: Using the current maximum completion time as the fitness function, a new process execution sequence is obtained through operations such as particle velocity updates and position adjustments (process sequence swapping) in the particle swarm optimization algorithm. For each step in the new process execution sequence, candidate robots are re-selected, an alliance is formed, task utility and robot utility are evaluated, and alliance members are adjusted to form a new robot alliance allocation scheme that matches the new process execution sequence. This process continues until the current iteration number reaches the maximum iteration number. Then, the maximum completion time corresponding to the scheduling scheme under each iteration is traversed sequentially, and the scheduling scheme with the minimum maximum completion time is determined as the optimal scheduling scheme.

[0111] The method provided in this embodiment uses "iterative cyclic adjustment" as its core approach. Through a cyclical process of "calculating the maximum completion time → determining whether the maximum number of iterations has been reached → adjusting the process sequence → reconstructing the robot alliance," the initial scheduling scheme is continuously modified. The particle swarm optimization algorithm is used to dynamically adjust the execution order of processes, and the alliance formation game is used to rematch the robot alliance based on the new sequence. This solves the static limitations of the initial scheme, enabling the optimal scheduling scheme to respond in real time to the dynamic changes in process connection and robot status in workshop production. It avoids scheduling failures caused by the rigidity of the initial scheme and achieves dynamic matching between processes and the alliance, ensuring that the execution of each process under the new sequence still meets the skill requirements and guarantees production feasibility. With "dual objective constraints" as the optimization guide, "skill matching" and "minimum maximum completion time" are used as core constraints. The iteration termination is controlled by the maximum number of iterations, ensuring a balance between production feasibility and efficiency. This avoids process failures due to insufficient robot skills and directly improves production efficiency. At the same time, the quantitative optimization standard reduces subjective errors and ensures the stability and repeatability of the final scheme.

[0112] The method provided in this embodiment, firstly, from the overall process of "establishing an optimization model - generating an initial process sequence - allocating a robot alliance - dynamic iterative optimization", constructs an optimization model with multiple constraints aimed at minimizing the maximum completion time, providing a unified goal orientation and feasible boundary for subsequent scheduling optimization; generating an initial process sequence based on particle swarm optimization solves the combinatorial optimization problem of process sequencing, laying an efficient process framework for robot allocation; using alliance formation game theory for task allocation achieves precise matching of robot resources and process requirements; and dynamically adjusting both through iterative cycles keeps the solution approaching the global optimum. This process forms a closed loop of "constraints ensuring feasibility, algorithms improving optimization efficiency, game theory achieving resource matching, and iterative convergence to the optimum", ultimately achieving a more efficient, precise, and adaptive scheduling scheme, significantly improving workshop production efficiency and resource utilization.

[0113] Secondly, by encoding the work processes as discrete particles, the abstract scheduling problem is transformed into a quantifiable algorithm optimization problem. Initializing the particle swarm's position and velocity ensures the diversity of initial solutions, avoiding getting trapped in local optima. Using the maximum completion time as the fitness function clarifies the optimization direction. By retaining historical exchange tendencies, combining individual and global optimal information, and introducing random number updates, the algorithm's exploration and convergence capabilities are balanced. Adjusting the work process sequence based on individual and social learning factors and checking constraints ensures the sequence's optimization potential and practical feasibility. These methods make the generated initial work process sequence both efficient and compliant, providing a scientific and reasonable foundation for subsequent robot alliance allocation and directly improving the initial quality of scheduling.

[0114] Thirdly, by screening candidate robots that meet the skill requirements, the alliance is ensured to possess the basic capabilities to execute processes; robots are allowed to freely form alliances with limited size, making the alliance structure compatible with task complexity; a task utility function incorporating first and second time parameters is designed to quantitatively evaluate alliance performance while considering overall progress and robot idleness; the utility of each robot is calculated and alliance members are dynamically adjusted to ensure that each robot in the alliance can make a positive contribution. These methods enable the resulting robot alliance allocation scheme to accurately match the task requirements of the initial process sequence, improve execution efficiency through collaboration, avoid resource misallocation and collaborative redundancy, and ensure a smooth and efficient production process.

[0115] Corresponding to the aforementioned embodiment of a robotized manufacturing workshop scheduling method based on alliance game theory, this application also provides an embodiment of a robotized manufacturing workshop scheduling device based on alliance game theory.

[0116] Figure 2 This is a schematic diagram of the robotic manufacturing workshop scheduling device based on alliance game theory provided in Embodiment 2 of this application. Please refer to... Figure 2 The apparatus provided in this embodiment includes an establishment module 210, a processing module 220, and an adjustment module 230;

[0117] The establishment module 210 is used to establish an optimization model with the single objective of minimizing the maximum completion time. The optimization model includes an objective function and constraints, including process sequence constraints, time logic constraints between processes and features, and robot resource allocation constraints.

[0118] The processing module 220 is used to optimize the process ordering based on the optimization model and particle swarm optimization. The process ordering is encoded as discrete particles. The particle velocity is updated through swapping operations. The process sequence corresponding to the particle is evaluated with the maximum completion time as the fitness function. The individual and global optimal sequences are iteratively updated to obtain the initial process execution sequence.

[0119] The processing module 220 is further configured to optimize robot task allocation based on the optimization model and the initial process execution sequence by employing alliance formation game theory, constructing an alliance formation game model, and controlling the robot to adjust alliance members according to utility through iterative optimization of task utility function and robot utility function, thereby obtaining a robot alliance allocation scheme that matches the initial process execution sequence.

[0120] The adjustment module 230 is used to dynamically adjust the process execution sequence and the robot alliance allocation scheme through iterative loops based on the initial process execution sequence and the robot alliance allocation scheme until the optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained.

[0121] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0122] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0123] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0124] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A robot manufacturing workshop scheduling method based on alliance game theory, characterized in that, The method includes: An optimization model is established with the single objective of minimizing the maximum completion time. The optimization model includes an objective function and constraints, including process sequence constraints, time logic constraints, robot allocation constraints, and robot task sequence constraints. Based on the optimization model, the process ordering optimization is performed based on particle swarm optimization. The process ordering is encoded as discrete particles. The particle velocity is updated through swapping operations. The process sequence corresponding to the particle is evaluated with the maximum completion time as the fitness function. The individual and global optimal sequences are iteratively updated to obtain the initial process execution sequence. Based on the optimization model and the initial process execution sequence, a coalition formation game is used to optimize robot task allocation. A coalition formation game model is constructed, and through iterative optimization of the task utility function and the robot utility function, the robot is controlled to adjust coalition members according to utility, thereby obtaining a robot coalition allocation scheme that matches the initial process execution sequence. Based on the initial process execution sequence and the robot alliance allocation scheme, the process execution sequence and the robot alliance allocation scheme are dynamically adjusted through iterative loops until the optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained.

2. The method according to claim 1, characterized in that, The process of obtaining the initial execution sequence includes: The execution order of all processes is encoded as discrete particles; the position of each particle corresponds to a global process execution sequence, and the velocity of each particle represents the tendency to exchange process sequences. Initialize the position and velocity of the particle swarm; the position is a randomly generated sequence of operations, and the velocity is an initial exchange probability parameter; Based on the constraints of the optimization model, the maximum completion time of the process sequence corresponding to each particle is calculated, and the maximum completion time is used as the fitness function value. By preserving historical exchange tendencies, combining the learning factors of individual optimal sequences and global optimal sequences, random numbers within a fixed interval are introduced to dynamically adjust the exchange tendencies, update particle velocities, and generate new velocity parameters. The exchange probability of the process sequence is calculated based on the updated velocity parameters. The process sequence corresponding to the particle position is then exchanged based on the exchange probability to obtain a new process execution sequence. Compare the fitness function values ​​of the new process execution sequence with the historical best sequence of the particles, update the individual best sequence, compare the best sequences of all individuals, and update the global best sequence; Repeat the steps from calculating to comparing fitness function values ​​until the preset maximum number of iterations is reached, and use the final globally optimal sequence as the initial process execution sequence.

3. The method according to claim 2, characterized in that, The process involves retaining historical exchange tendencies, combining the learning factors of individual optimal sequences and global optimal sequences, and dynamically adjusting the exchange tendencies by introducing random numbers within a fixed interval to update particle velocities and generate new velocity parameters. This includes: The tendency to exchange historical process sequences, as represented by the current velocity of the particle, is preserved, and the degree to which this tendency is preserved is controlled by an inertial weight parameter. Calculate the individual learning factor corresponding to the difference between the particle's current position and the individual's optimal sequence; Calculate the social learning factor corresponding to the difference between the particle's current position and the global optimal sequence; Two random numbers that follow a uniform distribution within a fixed interval are introduced to adjust the individual learning factor and the social learning factor, respectively. The updated historical process sequence exchange tendency, the adjusted individual learning factor, and the adjusted social learning factor are integrated to generate a new speed parameter.

4. The method according to claim 2, characterized in that, The step of calculating the exchange probability of the process sequence based on the updated velocity parameters, and exchanging the process sequences corresponding to the particle positions based on the exchange probability to obtain a new process execution sequence includes: The updated speed parameters are subjected to probability transformation to map the speed parameters to a unified range, thereby obtaining the exchange probability of the process sequence; For the current process sequence corresponding to the particle position, process position pairs to be exchanged are selected according to the exchange probability; process position pairs with an exchange probability higher than a preset threshold are selected first. Perform a swap operation on the selected process positions, interchange the processes in the two positions, and generate a temporary process sequence; Check whether the temporary process sequence satisfies the process order constraint in the optimization model. If it does, the temporary process sequence is determined as the updated particle position. If it does not, re-select the swap position pairs and perform the swap until a new process execution sequence that meets the process order constraint is generated.

5. The method according to claim 1, characterized in that, Based on the optimization model and the initial process execution sequence, a coalition formation game is used to optimize robot task allocation. A coalition formation game model is constructed, and through iterative optimization of the task utility function and the robot utility function, the robot is controlled to adjust coalition members according to utility, resulting in a robot coalition allocation scheme that matches the initial process execution sequence, including: For each step in the initial process execution sequence, robots that meet the skill requirements of each step are selected to form a candidate robot set; For each robot in the candidate robot set, an alliance can be freely formed; the size of the alliance shall not exceed the feature number constraint of the corresponding process. Design a task utility function to evaluate the overall performance of each alliance's execution process; Calculate the task utility value of all alliances based on the task utility function, and select the alliance with the largest task utility value as the current optimal alliance; For the current optimal alliance, calculate the utility of each robot; The alliance members are adjusted based on the utility of each robot. If the robot's utility is positive, it remains in the alliance; if it is negative, it leaves the alliance and a new alliance is formed. Repeat the steps from setting the task utility function to adjusting the alliance members until no robot leaves or joins the alliance due to utility issues, thus forming a robot alliance allocation scheme that matches the initial process execution sequence.

6. The method according to claim 5, characterized in that, The design task utility function evaluates the overall performance of each alliance's execution process, including: Determine the first and second time parameters of the task utility function; the first time parameter is the time from the start of the entire assembly process to the completion of the current process by the alliance; the second time parameter is the time from the earliest completion of the previous task among all robots participating in the current process to the completion of the current process by the alliance. Different constant weight values ​​are configured for the first time parameter and the second time parameter respectively; the weight value of the first time parameter is used to adjust the degree of influence of the entire assembly process on the alliance performance, and the weight value of the second time parameter is used to adjust the degree of influence of robot movement and operation time on the alliance performance. The weighted combination of the first time parameter, the second time parameter, and the corresponding constant weight value is determined as the task utility function.

7. The method according to claim 5, characterized in that, The calculation of the utility of each robot for the current optimal alliance includes: The total utility of the current optimal alliance is calculated using the task utility function. For each robot in the alliance, remove each robot from the current best alliance to form a new set of robots, and obtain the remaining alliance after each robot leaves; The total utility of each remaining coalition is calculated using the task utility function. The utility of each robot is determined by the difference between the total utility of the current best coalition and the total utility of the remaining coalition after each robot leaves.

8. The method according to claim 1, characterized in that, Based on the initial process execution sequence and the robot alliance allocation scheme, the process execution sequence and robot alliance allocation scheme are dynamically adjusted iteratively until an optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained, including: Calculate the maximum completion time of the current initial process execution sequence and the corresponding robot alliance allocation scheme; Determine whether the current iteration count has reached the maximum iteration count. If the maximum iteration count has not been reached, adjust the process execution sequence based on the particle swarm optimization algorithm and optimize the new process execution sequence using the current maximum completion time as the fitness function. Based on the new process execution sequence, the robot alliance allocation scheme is reconstructed through alliance formation game theory, forming a new robot alliance allocation scheme that matches the new process execution sequence. Repeat the steps from calculating the maximum completion time to reconstructing the robot alliance allocation scheme until the current iteration count reaches the maximum iteration count. Then, determine the scheduling scheme with the minimum maximum completion time as the optimal scheduling scheme.

9. The method according to claim 1, characterized in that, The process sequence constraint represents the order of processes within the same product, as well as the relationship between the process and the final completion time of the product; the time logic constraint represents the temporal inclusion relationship between the process and its subordinate features; the robot allocation constraint restricts the matching of robot capabilities, the uniqueness of feature allocation, and the matching of alliance size and feature quantity; the robot task sequence constraint represents the temporal order in which the same robot executes multiple features.

10. A robotic manufacturing workshop scheduling device based on alliance game theory, characterized in that, The device includes an establishment module, a processing module, and an adjustment module; The establishment module is used to establish an optimization model with the single objective of minimizing the maximum completion time. The optimization model includes an objective function and constraints, including process sequence constraints, time logic constraints between processes and features, and robot resource allocation constraints. The processing module is used to optimize the process order based on the optimization model and particle swarm optimization. The process order is encoded as discrete particles, the particle velocity is updated through swapping operations, the process sequence corresponding to the particle is evaluated with the maximum completion time as the fitness function, and the individual and global optimal sequences are iteratively updated to obtain the initial process execution sequence. The processing module is further configured to optimize robot task allocation using alliance formation game based on the optimization model and the initial process execution sequence, construct an alliance formation game model, and control the robot to adjust alliance members according to utility through iterative optimization of task utility function and robot utility function, thereby obtaining a robot alliance allocation scheme that matches the initial process execution sequence. The adjustment module is used to dynamically adjust the process execution sequence and the robot alliance allocation scheme through iterative loops based on the initial process execution sequence and the robot alliance allocation scheme, until the optimal scheduling scheme that satisfies skill matching and minimizes the maximum completion time is obtained.

Images