Multi-target AGV task allocation optimization method

Abstract

The invention discloses a multi-target AGV task allocation optimization method, and the method comprises the steps: obtaining the position information, load information, task information and related constraint conditions of an AGV according to a manufacturing workshop environment and order production demands; performing initialization coding, and generating a task relation matrix, a task data matrix and an AGV state matrix by adopting integer coding for representing a task allocation scheme; constructing a multi-target AGV task allocation optimization model by taking the total transportation distance, the total energy consumption and the load time deviation as optimization targets; solving the multi-objective optimization model by adopting an improved hunting optimization algorithm to obtain an optimal task allocation optimization scheme; according to the method provided by the invention, the problems of load unbalance and the like in AGV task allocation can be effectively solved, the balance and efficiency of task allocation are improved, the task completion time is shortened, the energy consumption of the system is reduced, and the overall operation efficiency of the multi-AGV system is improved.

Description

Technical Field

[0001] The present invention belongs to the field of multi-AGV task allocation in automated workshops, and in particular relates to a multi-objective AGV task allocation optimization method. Background Art

[0002] Smart manufacturing has become one of the key driving forces behind industrial upgrading. An increasing number of small and medium-sized enterprises are turning to intelligent production models to improve efficiency and reduce costs. In this process, the automation of cargo transportation is particularly important, especially by replacing manual transportation with automated guided vehicles (AGVs). AGVs have become an essential component of modern manufacturing workshops. However, with the expansion of AGV systems and the surge in complexity of their operating environments, as well as the multiple variables and constraints involved in task allocation, traditional optimization methods struggle to find the optimal solution within a reasonable timeframe. Therefore, efficiently coordinating the task allocation of multiple AGVs has become a research priority. Efficient task allocation methods can significantly improve workshop production efficiency.

[0003] Prior art approaches have addressed multi-AGV task allocation by minimizing average execution time and maximizing the number of tasks. However, these optimization designs have limitations in their goal setting, failing to fully consider the cost of AGV travel distance during task execution and overlooking the impact of AGV load balancing on system efficiency. Therefore, efficient optimization of multi-AGV task allocation has been achieved through the particle swarm optimization (PSO) algorithm. However, when solving multi-AGV task allocation problems in complex scenarios, such as multi-AGV task allocation using traditional metaheuristic algorithms, the algorithm is prone to slow convergence and becoming stuck in local optima. Specifically, multi-AGV task allocation suffers from the following major issues: 1) The objective functions of existing AGV task allocation models often overly focus on single or one-sided optimization objectives, such as minimizing task completion time or maximizing task throughput, while neglecting key factors such as energy management. This increases maintenance costs and limits overall system efficiency. 2) Existing optimization algorithms often employ traditional metaheuristic algorithms to solve multi-AGV task allocation problems. However, as the number of AGVs, task scale, and computational requirements increase, these algorithms are prone to slow convergence and becoming stuck in local optima. 3) Existing mathematical models for AGV task allocation are incomplete in their scenario assumptions and constraint settings, particularly regarding load balancing. Traditional methods typically use the number of tasks as a criterion for load balancing. However, when AGV task execution times vary significantly, the number of tasks cannot accurately reflect the actual AGV load, resulting in distorted balance assessments. Summary of the Invention

[0004] In response to the problems existing in the prior art, the present invention provides a multi-objective AGV task allocation optimization method, which aims to effectively address problems such as load imbalance in AGV task allocation, improve the balance and efficiency of task allocation, shorten task completion time, reduce system energy consumption, and improve the overall operating efficiency of the multi-AGV system.

[0005] In order to achieve the above technical objectives, the present invention provides the following technical solutions:

[0006] A multi-objective AGV task allocation optimization method specifically includes the following steps:

[0007] S1. Obtain the AGV's location information, load information, task information, and related constraints based on the manufacturing workshop environment and order production requirements;

[0008] S2. Perform initialization coding. Based on the acquired AGV information and task information, integer coding is used to generate the task relationship matrix, task data matrix, and AGV status matrix to represent the task allocation plan.

[0009] S3. Taking total transport distance, total energy consumption and load time deviation as optimization targets, a multi-objective AGV task allocation optimization model is constructed;

[0010] S4. The improved Cheetah optimization algorithm is used to solve the multi-objective AGV task allocation optimization model, and the optimal AGV task allocation solution is iteratively generated by combining the various matrices generated by integer coding.

[0011] Furthermore, step S1 specifically includes:

[0012] S11. According to the map of the manufacturing workshop and the requirements of the order, obtain AGV status information and task information. The AGV status information includes the starting position and load quantity of the AGV, and the task information includes the task sequence number, the starting position of the task, and the end position of the task;

[0013] S12. Set constraints for the AGV to perform tasks, including:

[0014] Constraint 1:

[0015] Constraint 2:

[0016] Constraint 3:

[0017] Constraint 4: L min ≤N j ≤L max ,j∈{1,2,...,m};

[0018] Among them, constraint 1 indicates the distribution, xij To assign an indication, if x ij =1, it means the i-th task H i Assigned to the jth AGV, the jth AGV is recorded as AGV a j ; On the contrary x ij =0, it means the i-th task H i Not assigned to the jth AGVa j Constraint 2 constrains each task to be executed only once by one AGV to avoid repeated execution of tasks; Constraint 3 constrains the total number of tasks in the system to be equal to the total number of assigned tasks to ensure that no tasks are missed, n and m are the total number of tasks and the total number of AGVs respectively; Constraint 4 constrains the number of tasks per AGV load, N j Indicates AGVa j The number of load tasks, L min and L max They represent the upper and lower limits of the number of tasks assigned to each AGV, respectively, to ensure that the task load of each AGV is within a reasonable range.

[0019] Furthermore, step S2 specifically includes:

[0020] S21. Define the task relationship matrix M to represent the task allocation scheme. The formula is:

[0021]

[0022] The element x corresponding to the i-th row and j-th column in the task relationship matrix ij is the allocation indicator, which indicates whether the i-th task is assigned to the j-th AGV. There are n tasks and m AGVs in total.

[0023] S22. Define a task data matrix G to represent the starting coordinates, ending coordinates, and waiting time of each task. Each row of the matrix G corresponds to a task, and the formula is:

[0024]

[0025] Among them, (x i S ,y i S ) represents the starting position of the i-th task; (x i E ,y i E ) represents the end point of the i-th task; represents the waiting time required for the i-th task, that is, the time required for the AGV to load and unload goods;

[0026] S23. Define the AGV state matrix S to reflect the starting position of each AGV and the number of current load tasks. The formula is expressed as:

[0027]

[0028] in, Indicates the starting point of the jth AGV, N j Indicates the number of load tasks of the j-th AGV.

[0029] Furthermore, step S3 specifically includes:

[0030] S31. Before constructing the optimization model, set the following assumptions for the optimization model in advance:

[0031] (1) The location information of the machine tools, AGV starting points, obstacles, etc. in the manufacturing workshop is known;

[0032] (2) Each AGV travels at a constant speed and has sufficient power, without considering the impact of turning, collision, and power on driving speed

[0033] (3) The driving distance, total energy consumption, driving time and load time deviation are all in uniform unit values;

[0034] (4) All tasks have the same priority and there are no urgent tasks;

[0035] (5) Each AGV can only perform one task at a time, and each task can only be performed once by one AGV. At the same time, the number of tasks is greater than the number of AGVs.

[0036] (6) All AGVs start from their respective starting points, and the waiting time of the AGV is the loading and unloading time of the cargo. The loading and unloading time is only related to the weight of the cargo. Different weights will result in different loading and unloading times.

[0037] S32, based on the assumptions in step S31, a multi-objective AGV task allocation optimization model is constructed, taking into account the total driving distance, total energy consumption and load time deviation, and using a weighted approach to calculate the objective function; the total driving distance is respectively denoted as D total , record the total energy consumption as E total , let the load time deviation be U, then the formula of the total objective function is expressed as:

[0038] Z=w1·D total +w2·E total +w3·U;

[0039] Where Z is the total objective function; w1, w2, and w3 represent the weight coefficients of the total driving distance, total energy consumption, and load time deviation, respectively, and w1+w2+w3=1; the optimization goal of the model is to minimize the total objective function.

[0040] Furthermore, the specific calculation method of each item in the total objective function is:

[0041] Total distance traveled

[0042] in, and For AGVa j Execute Task H i The no-load stroke and load stroke are calculated as follows:

[0043]

[0044] in, For AGVa j The last task assigned H k The end position of (x i S ,y i S ) and (x i E ,y i E ) represents task H i the starting and ending points;

[0045] Total energy consumption

[0046] Among them, e empty , e loaded and e wait They represent no-load energy consumption coefficient, load energy consumption coefficient and waiting energy consumption coefficient respectively; The time required to load and unload cargo for the i-th task;

[0047] Load time deviation

[0048] Among them, T max is the total transportation time of AGV, that is, the longest AGV transportation time; T j is the transportation time of the jth AGV; the calculation formulas are:

[0049]

[0050] T max =max{T1,T2,...,T m}.

[0051] Furthermore, step S4 specifically includes the following steps:

[0052] S41. Based on the traditional cheetah optimization algorithm, the step size update and position update strategies are improved by introducing the hyperbolic tangent step size strategy; the formula is expressed as:

[0053]

[0054] Among them, S t is the step size at the tth iteration; S max and S min is the maximum and minimum value of the step size; T is the maximum number of iterations;

[0055] S42. Introduce a hierarchical simulated annealing strategy for position update; implement a global simulated annealing strategy in the early stages of the algorithm iteration to maintain a high temperature to increase the probability of accepting inferior solutions and enhance exploration capabilities; implement a local simulated annealing strategy in the later stages of the algorithm iteration to control the temperature to decay faster and reduce the acceptance rate of inferior solutions, so as to conduct a detailed search and improve the quality of the solution;

[0056] The total objective function of the multi-objective AGV task allocation optimization model is then used as the fitness function of the improved cheetah optimization algorithm. The improved cheetah optimization algorithm selects the optimal task allocation solution by comparing the fitness functions.

[0057] S43. Initialize the algorithm in combination with the coding matrix of step S2, initialize the population parameters, and balance global exploration and local development through the hyperbolic tangent step strategy; position updates are performed according to the search, wait, and attack strategies in the traditional cheetah optimization algorithm, and the temperature and solution acceptance probability are dynamically adjusted in combination with the hierarchical simulated annealing strategy to perform global and local optimization; the algorithm iteratively updates the solution acceptance probability, compares the fitness values ​​of the cheetahs at different positions, continuously optimizes the task allocation plan, and outputs the optimal task allocation plan until the termination condition is met.

[0058] Furthermore, at the initial stage of the algorithm, a cheetah population is established based on the coding matrix in step S2. The initial cheetah position is defined by the sequential coding vector of each AGV performing the task. The sequential coding vector formula is as follows:

[0059]

[0060] Among them, O t (j) represents the position of the j-th AGV cheetah after the t-th iteration, i.e., the sequential encoding vector; t ji represents the execution order of the i-th task on the j-th AGV, which serves as the initial layout of task allocation;

[0061] In the exploration phase, a global simulated annealing method is used to set a high initial temperature, gradually lower the temperature through the attenuation factor, adjust the wide search range of the cheetah position, and update the sequence encoding with the hyperbolic tangent step size strategy to explore more global task sequence combinations;

[0062] During the waiting and attacking stages, a local simulated annealing strategy is implemented, a low initial temperature is set, the attenuation factor is adjusted to accelerate the temperature drop, the hyperbolic tangent step size strategy is used to update the sequential encoding, refine the local task order, and seek the optimal allocation plan.

[0063] Based on the above technical solution, the present invention has at least the following beneficial effects:

[0064] 1. This invention employs an efficient task allocation encoding method. Combining AGV and task information in the workshop, this method uses integer encoding to generate a task relationship matrix, a task data matrix, and an AGV status matrix, systematically representing the task allocation scheme. This encoding method boasts a simple structure and complete data, effectively improving the computational efficiency and optimization accuracy of task allocation.

[0065] 2. This invention introduces a load time deviation indicator. Compared with the traditional method that only evaluates load balancing based on the number of tasks, this indicator comprehensively considers the difference between the system's maximum transportation time and the task execution time of each AGV, accurately reflecting the actual load situation of each AGV, thereby optimizing the balance of task allocation and significantly improving the operating efficiency of the multi-AGV system.

[0066] 3. This paper proposes a multi-objective AGV task allocation model, which takes the total transportation distance, total energy consumption and load time deviation as optimization targets, comprehensively optimizes transportation efficiency, energy consumption cost and load balance, effectively shortens the AGV operation path, reduces system energy consumption, and improves task load balance.

[0067] 4. The present invention proposes an improved Cheetah optimization algorithm, which effectively improves the algorithm convergence speed and enhances the ability to escape local optimality by introducing the hyperbolic tangent step size strategy and the hierarchical simulated annealing strategy, thereby optimizing the generation efficiency and solution quality of different task allocation schemes in multi-AGV task allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is an overall flow chart of a multi-objective AGV task allocation optimization method proposed in the present invention;

[0069] Figure 2 Layout drawing of the manufacturing workshop;

[0070] Figure 3 This is a flow chart of the improved cheetah optimization algorithm proposed in the present invention.

[0071] Figure 4Improved Cheetah optimization algorithm fitness value curve;

[0072] Figure 5 Gantt chart of the traditional Cheetah optimization algorithm;

[0073] Figure 6 Improved Gantt chart of Cheetah optimization algorithm. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0075] Although the steps in the present invention are arranged with numbers, they are not intended to limit the order of the steps. Unless the order of the steps is clearly stated or the execution of a step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" used herein refers to and covers any and all possible combinations of one or more of the associated listed items.

[0076] like Figure 1 FIG. 1 shows a multi-objective AGV task allocation optimization method proposed by the present invention, which specifically includes the following steps:

[0077] S1. Obtain the AGV's location information, load information, task information, and related constraints based on the manufacturing workshop environment and order production requirements;

[0078] As a preferred embodiment, step S1 specifically includes:

[0079] S11. According to the map of the manufacturing workshop and the requirements of the order, obtain AGV status information and task information. The AGV status information includes the starting position and load quantity of the AGV, and the task information includes the task sequence number, the starting position of the task, and the end position of the task;

[0080] In this embodiment, 18 tasks and 3 AGVs are set. The initial positions of the 3 AGVs are as follows: Figure 2 As shown, Figure 2 The positions of each machine tool correspond to the starting and ending points of different tasks. Table 1 below lists the starting coordinates of the AGV and the specific task information of each task:

[0081] Table 1 Task information

[0082]

[0083]

[0084] S12. Set constraints for the AGV to perform tasks, including:

[0085] Constraint 1:

[0086] Constraint 2:

[0087] Constraint 3:

[0088] Constraint 4: L min ≤N j ≤L max ,j∈{1,2,...,m};

[0089] Among them, constraint 1 indicates the distribution, x ij To assign an indication, if x ij =1, it means the i-th task H i Assigned to the jth AGV, the jth AGV is recorded as AGV a j ; On the contrary x ij =0, it means the i-th task H i Not assigned to the jth AGVa j Constraint 2 constrains each task to be executed only once by one AGV to avoid repeated execution of tasks; Constraint 3 constrains the total number of tasks in the system to be equal to the total number of assigned tasks to ensure that no tasks are missed, n and m are the total number of tasks and the total number of AGVs respectively; Constraint 4 constrains the number of tasks per AGV load, N j Indicates AGVa j The number of load tasks, L min and L max Respectively represent the upper and lower limits of the number of tasks assigned to each AGV, ensuring that the task load of each AGV is within a reasonable range. In this embodiment, set L min 3, L max is 7, when the number of tasks assigned to an AGV is less than L min or more than L max When , a penalty value (-100000) is added to its corresponding fitness value to reduce the fitness of unreasonable task allocation schemes, thereby guiding the optimization algorithm to give priority to allocation schemes with a balanced number of tasks and promote the improvement of overall scheduling efficiency.

[0090] S2. Perform initialization coding. Based on the acquired AGV information and task information, integer coding is used to generate the task relationship matrix, task data matrix, and AGV status matrix to represent the task allocation plan.

[0091] As a preferred embodiment, step S2 specifically includes:

[0092] S21. Define the task relationship matrix M to represent the task allocation scheme. The formula is:

[0093]

[0094] The element x corresponding to the i-th row and j-th column in the task relationship matrix ij is the allocation indicator, which indicates whether the i-th task is assigned to the j-th AGV. There are n tasks and m AGVs in total.

[0095] In this embodiment, the matrix X can be represented as 18 rows and 3 columns corresponding to 18 tasks and 3 AGVs. The number of matrices X is related to the number of cheetah individuals set when the algorithm is initialized;

[0096] S22. Define a task data matrix G to represent the starting coordinates, ending coordinates, and waiting time of each task. Each row of the matrix G corresponds to a task, and the formula is:

[0097]

[0098] Among them, (x i S ,y i S ) represents the starting position of the i-th task; (x i E ,y i E ) represents the end point of the i-th task; represents the waiting time required for the i-th task, that is, the time required for the AGV to load and unload goods;

[0099] In this embodiment, the matrix G is:

[0100]

[0101] S23. Define the AGV state matrix S to reflect the starting position of each AGV and the number of current load tasks. The formula is expressed as:

[0102]

[0103] Among them, (x j a ,y j a ) represents the starting point of the jth AGV, N j represents the number of load tasks of the j-th AGV;

[0104] In this embodiment, the matrix S is specifically:

[0105]

[0106] S3. Taking total transport distance, total energy consumption and load time deviation as optimization targets, a multi-objective AGV task allocation optimization model is constructed;

[0107] As a preferred embodiment, step S3 specifically includes:

[0108] S31. Before constructing the optimization model, set the following assumptions for the optimization model in advance:

[0109] (1) The location information of the machine tools, AGV starting points, obstacles, etc. in the manufacturing workshop is known;

[0110] (2) Each AGV travels at a constant speed and has sufficient power, without considering the impact of turning, collision, and power on driving speed

[0111] (3) The driving distance, total energy consumption, driving time and load time deviation are all in uniform unit values;

[0112] (4) All tasks have the same priority and there are no urgent tasks;

[0113] (5) Each AGV can only perform one task at a time, and each task can only be performed once by one AGV. At the same time, the number of tasks is greater than the number of AGVs.

[0114] (6) All AGVs start from their respective starting points, and the waiting time of the AGV is the loading and unloading time of the cargo. The loading and unloading time is only related to the weight of the cargo. Different weights will result in different loading and unloading times.

[0115] S32, based on the assumptions in step S31, a multi-objective AGV task allocation optimization model is constructed, taking into account the total driving distance, total energy consumption and load time deviation, and using a weighted approach to calculate the objective function; the total driving distance is respectively denoted as D total , record the total energy consumption as E total , let the load time deviation be U;

[0116] More specifically, the specific calculation method for each item in the total objective function is:

[0117] Total distance traveled

[0118] in, and For AGVa j Execute Task H i The no-load stroke and load stroke are calculated as follows:

[0119]

[0120] in, For AGVa jThe last task assigned H k The end position of (x i S ,y i S ) and (x i E ,y i E ) represents task H i the starting and ending points;

[0121] Total energy consumption

[0122] Among them, e empty , e loaded and e wait They represent no-load energy consumption coefficient, load energy consumption coefficient and waiting energy consumption coefficient respectively; The time required to load and unload cargo for the i-th task;

[0123] In this implementation case, the values ​​of the three energy consumption coefficients are e empty =4,e loaded =5 and e wait =1, which means the energy consumed by the AGV for each step. Since the energy consumed by the AGV is the highest when it is loaded, the load energy consumption coefficient is set to the maximum.

[0124] Load time deviation Compared with the limitation of traditional methods that only evaluate load balancing based on the number of tasks, this application considers the load time deviation indicator. By comprehensively considering the difference between the system's maximum transportation time and the task execution time of each AGV, it accurately reflects the actual load situation of each AGV, thereby optimizing the balance of task allocation and significantly improving the operating efficiency of the multi-AGV system.

[0125] Among them, T max is the total transportation time of AGV, that is, the longest AGV transportation time; T j is the transportation time of the jth AGV; the calculation formulas are:

[0126]

[0127] T max =max{T1,T2,...,T m};

[0128] The formula of the overall objective function is expressed as:

[0129] Z=w1·D total +w2·E total +w3·U;

[0130] Among them, Z is the total objective function; w1, w2, and w3 represent the weight coefficients of the total driving distance, total energy consumption, and load time deviation, respectively, and w1+w2+w3=1; the optimization goal of the model is to minimize the total objective function. In this embodiment, the weight priorities are total transportation distance (w1=0.5), total energy consumption (w2=0.3), and load time deviation (w3=0.2), in order to reduce driving distance, optimize energy consumption, and balance task load.

[0131] S4. Use the improved Cheetah optimization algorithm to solve the multi-objective AGV task allocation optimization model, combine the various matrices generated by integer coding, and iteratively generate the optimal AGV task allocation solution;

[0132] As a preferred embodiment, step S4 specifically includes:

[0133] S41. Based on the traditional cheetah optimization algorithm, the step size update and position update strategies are improved by introducing the hyperbolic tangent step size strategy; the formula is expressed as:

[0134]

[0135] Among them, S t is the step size at the tth iteration; S max and S min is the maximum and minimum value of the step size; T is the maximum number of iterations;

[0136] S42. Introduce a hierarchical simulated annealing strategy for position update; implement a global simulated annealing strategy in the early stages of the algorithm iteration to maintain a high temperature to increase the probability of accepting inferior solutions and enhance exploration capabilities; implement a local simulated annealing strategy in the later stages of the algorithm iteration to control the temperature to decay faster and reduce the acceptance rate of inferior solutions, so as to conduct a detailed search and improve the quality of the solution;

[0137] The global simulated annealing strategy is as follows:

[0138] During the global search phase, maintain a high temperature To increase the probability of accepting inferior solutions P ag , enhancing global search capabilities. The specific calculation formula is as follows:

[0139]

[0140] in, represents the global strategy temperature at the tth iteration, which is set to a higher value; r g represents the global cooling rate, which is close to 1; Δf is the fitness difference between the current solution and the candidate solution; the random number k∈[0,1], if k <P ag Then accept the inferior solution;

[0141] The local simulated annealing strategy is as follows:

[0142] In the local search phase, the temperature decays faster, reducing the acceptance rate of inferior solutions, focusing on local fine search, and improving the quality of the solution. The specific calculation formula is as follows:

[0143]

[0144] in, represents the local strategy temperature at the tth iteration, which is set to a smaller value; r l Indicates the local cooling rate, which is smaller; P al The probability of the local strategy accepting an inferior solution. The global strategy slowly cools down in the early stage to enhance the exploration ability, and the local strategy quickly cools down in the later stage to finely optimize the solution, thereby effectively balancing global search and local development.

[0145] The total objective function of the multi-objective AGV task allocation optimization model is then used as the fitness function of the improved cheetah optimization algorithm. The improved cheetah optimization algorithm selects the optimal task allocation solution by comparing the fitness functions.

[0146] S43: Initialize the algorithm based on the encoding matrix of step S2, initialize the population parameters, and balance global exploration and local development through the hyperbolic tangent step strategy; position update is performed according to the search, wait, and attack strategies in the traditional cheetah optimization algorithm:

[0147] The search strategy simulates the cheetah's active exploration behavior in the early stages of hunting. The cheetah's position is searched globally using a random step size, and the update formula is:

[0148]

[0149] in, represents the w-th dimension position (w=1,2,3,…,D) of individual v (v=1,2,3,…,N) at the t-th iteration, for The updated position; N is the number of cheetahs in the population; D is the dimension of the problem; r i,j is a standard normal random number, is the step size factor, which helps to control the search range;

[0150] The static strategy simulates a cheetah remaining stationary in an appropriate position, waiting for the optimal opportunity. This strategy prevents the algorithm from over-searching by maintaining the current position, thereby improving convergence stability. The formula is as follows:

[0151]

[0152] It also combines a hierarchical simulated annealing strategy to dynamically adjust the temperature and solution acceptance probability, performing global and local optimization; the algorithm iteratively updates the solution acceptance probability, compares the fitness values ​​of cheetahs at different positions, and continuously optimizes the task allocation plan until the termination condition is met, outputting the optimal task allocation plan.

[0153] The attack strategy simulates a cheetah's decision to attack its prey. It will rush towards the target at maximum speed and adjust its position according to the prey's escape path. The cheetah's position update formula is as follows:

[0154]

[0155] in, is the current position of the prey in the wth row; θ v,w and Represent the steering factor and interaction factor of cheetah v in the arrangement w respectively; It can be calculated by the position difference of adjacent cheetahs, r v,w is a random number that follows a standard normal distribution (i.e., with a variance of 1 and a mean of 0).

[0156] Then, the temperature and solution acceptance probability are dynamically adjusted in combination with the hierarchical simulated annealing strategy to perform global and local optimization. More specifically, at the initial stage of the algorithm, a cheetah population is established based on the coding matrix in step S2. The initial cheetah position is defined by the sequential coding vector of each AGV's task execution. The sequential coding vector formula is as follows:

[0157]

[0158] Among them, O t (j) represents the position of the j-th AGV cheetah after the t-th iteration, i.e., the sequential encoding vector; t ji represents the execution order of the i-th task on the j-th AGV, which serves as the initial layout of task allocation;

[0159] In the exploration phase, a global simulated annealing method is used to set a high initial temperature, gradually lower the temperature through the attenuation factor, adjust the wide search range of the cheetah position, and update the sequence encoding with the hyperbolic tangent step size strategy to explore more global task sequence combinations;

[0160] During the waiting and attack phases, a local simulated annealing strategy is implemented, setting a low initial temperature, adjusting the decay factor to accelerate the temperature drop, and using the hyperbolic tangent step size strategy to update the sequence code, refine the local task order, and seek the optimal allocation solution.

[0161] The present invention uses a collaborative mechanism of hierarchical temperature control and sequential coding position update. In the initial stage, the coding matrix is ​​relied on to lay the foundation for task allocation. In the search phase, high temperature and a larger step size based on the hyperbolic tangent step size strategy are used to explore the global task sequence combination. In the static and attack phases, low temperature and a smaller step size are used to refine the local task order. Finally, the objective function of the multi-objective optimization model is used as the evaluation basis, and the fitness of the cheetah position is continuously compared until the termination condition is met to obtain the optimal task allocation plan.

[0162] In this embodiment, the overall process of the improved Cheetah optimization algorithm is as follows: Figure 3 The initialization parameters of the improved cheetah optimization algorithm are shown in Table 2.

[0163] Table 2 Initialization parameters

[0164]

[0165] In order to verify the effectiveness of the improved algorithm, a simulation analysis was conducted based on the aforementioned algorithm parameters, AGV information and task information. Figure 4 As shown, the task allocation Gantt chart is as follows Figure 5 and Figure 6 shown.

[0166] Experimental results show that the improved Cheetah Optimization Algorithm significantly outperforms the original Cheetah Optimization Algorithm in terms of convergence speed and achieves a lower fitness value, indicating that both optimization objectives have been effectively improved. Gantt chart analysis shows that the AGV task allocation plan generated by the improved algorithm takes less time and achieves a more balanced load distribution among the AGVs. Therefore, the improved Cheetah Optimization Algorithm can significantly improve the efficiency of task allocation and generate more reasonable allocation plans.

[0167] In summary, the present invention provides a multi-objective AGV task allocation optimization method, which can effectively deal with problems such as load imbalance in AGV task allocation, improve the balance and efficiency of task allocation, shorten task completion time, reduce system energy consumption, and improve the overall operating efficiency of the multi-AGV system.

[0168] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0169] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A multi-objective AGV task allocation optimization method, characterized in that: The specific steps include: S1. Obtain the AGV's location information, load information, task information, and related constraints based on the manufacturing workshop environment and order production requirements; S2. Perform initialization coding. Based on the acquired AGV information and task information, integer coding is used to generate the task relationship matrix, task data matrix, and AGV status matrix to represent the task allocation plan. S3. Taking total transport distance, total energy consumption and load time deviation as optimization targets, a multi-objective AGV task allocation optimization model is constructed; S4. The improved Cheetah optimization algorithm is used to solve the multi-objective AGV task allocation optimization model, and the optimal AGV task allocation solution is iteratively generated by combining the various matrices generated by integer coding.

2. A multi-objective AGV task allocation optimization method according to claim 1, characterized in that: Step S1 specifically includes: S11. According to the map of the manufacturing workshop and the requirements of the order, obtain AGV status information and task information. The AGV status information includes the starting position and load quantity of the AGV, and the task information includes the task sequence number, the starting position of the task, and the end position of the task; S12. Set constraints for the AGV to perform tasks, including: Constraint 1: Constraint 2: Constraint 3: Constraint 4: L min ≤N j ≤L max ,j∈{1,2,...,m}; Among them, constraint 1 indicates the distribution, x ij To assign an indication, if x ij =1, it means the i-th task H i Assigned to the jth AGV, the jth AGV is recorded as AGV a j ; On the contrary x ij =0, it means the i-th task H i Not assigned to the jth AGV a j Constraint 2 constrains each task to be executed only once by one AGV to avoid repeated execution of tasks; Constraint 3 constrains the total number of tasks in the system to be equal to the total number of assigned tasks to ensure that no tasks are missed, n and m are the total number of tasks and the total number of AGVs respectively; Constraint 4 constrains the number of tasks per AGV load, N j Indicates AGVa j The number of load tasks, L min and L max They represent the upper and lower limits of the number of tasks assigned to each AGV, respectively, to ensure that the task load of each AGV is within a reasonable range.

3. The multi-objective AGV task allocation optimization method according to claim 1 is characterized in that: Step S2 specifically includes: S21. Define the task relationship matrix M to represent the task allocation scheme. The formula is: The element x corresponding to the i-th row and j-th column in the task relationship matrix ij is the allocation indicator, which indicates whether the i-th task is assigned to the j-th AGV. There are n tasks and m AGVs in total. S22. Define a task data matrix G to represent the starting coordinates, ending coordinates, and waiting time of each task. Each row of the matrix G corresponds to a task, and the formula is: in, Indicates the starting position of the i-th task; Indicates the end point of the i-th task; represents the waiting time required for the i-th task, that is, the time required for the AGV to load and unload goods; S23. Define the AGV state matrix S to reflect the starting position of each AGV and the number of current load tasks. The formula is expressed as: in, Indicates the starting point of the jth AGV, N j Indicates the number of load tasks of the j-th AGV.

4. A multi-objective AGV task allocation optimization method according to claim 3, characterized in that: Step S3 specifically includes: S31. Before constructing the optimization model, set the following assumptions for the optimization model in advance: (1) The location information of the machine tools, AGV starting points, obstacles, etc. in the manufacturing workshop is known; (2) Each AGV travels at a constant speed and has sufficient power, without considering the impact of turning, collision, and power on driving speed (3) The driving distance, total energy consumption, driving time and load time deviation are all in uniform unit values; (4) All tasks have the same priority and there are no urgent tasks; (5) Each AGV can only perform one task at a time, and each task can only be performed once by one AGV. At the same time, the number of tasks is greater than the number of AGVs. (6) All AGVs start from their respective starting points, and the waiting time of the AGV is the loading and unloading time of the cargo. The loading and unloading time is only related to the weight of the cargo. Different weights will result in different loading and unloading times. S32, based on the assumptions in step S31, a multi-objective AGV task allocation optimization model is constructed, taking into account the total driving distance, total energy consumption and load time deviation, and using a weighted approach to calculate the objective function; the total driving distance is respectively denoted as D total , record the total energy consumption as E total , let the load time deviation be U, then the formula of the total objective function is expressed as: Z=w1·D total +w2·E total +w3·U; Where Z is the total objective function; w1, w2, and w3 represent the weight coefficients of the total driving distance, total energy consumption, and load time deviation, respectively, and w1+w2+w3=1; the optimization goal of the model is to minimize the total objective function.

5. A multi-objective AGV task allocation optimization method according to claim 4, characterized in that: The specific calculation method of each item in the total objective function is: Total distance traveled in, and For AGVa j Execute Task H i The no-load stroke and load stroke are calculated as follows: in, For AGV a j The last task assigned H k The end position of and Represents task H i the starting and ending points; Total energy consumption Among them, e empty , e loaded and e wait They represent no-load energy consumption coefficient, load energy consumption coefficient and waiting energy consumption coefficient respectively; The time required to load and unload cargo for the i-th task; Load time deviation Among them, T max is the total transportation time of AGV, that is, the longest AGV transportation time; T j is the transportation time of the jth AGV; the calculation formulas are: T max =max{T1,T2,...,T m }。 6. A multi-objective AGV task allocation optimization method according to claim 1, characterized in that: Step S4 specifically includes the following steps: S41. Based on the traditional cheetah optimization algorithm, the step size update and position update strategies are improved by introducing the hyperbolic tangent step size strategy; the formula is expressed as: Among them, S t is the step size at the tth iteration; S max and S min is the maximum and minimum value of the step size; T is the maximum number of iterations; S42. Introduce a hierarchical simulated annealing strategy for position update; implement a global simulated annealing strategy in the early stages of the algorithm iteration to maintain a high temperature to increase the probability of accepting inferior solutions and enhance exploration capabilities; implement a local simulated annealing strategy in the later stages of the algorithm iteration to control the temperature to decay faster and reduce the acceptance rate of inferior solutions, so as to conduct a detailed search and improve the quality of the solution; The total objective function of the multi-objective AGV task allocation optimization model is then used as the fitness function of the improved cheetah optimization algorithm. The improved cheetah optimization algorithm selects the optimal task allocation solution by comparing the fitness functions. S43. Initialize the algorithm in combination with the coding matrix of step S2, initialize the population parameters, and balance global exploration and local development through the hyperbolic tangent step strategy; position updates are performed according to the search, wait, and attack strategies in the traditional cheetah optimization algorithm, and the temperature and solution acceptance probability are dynamically adjusted in combination with the hierarchical simulated annealing strategy to perform global and local optimization; the algorithm iteratively updates the solution acceptance probability, compares the fitness values ​​of the cheetahs at different positions, continuously optimizes the task allocation plan, and outputs the optimal task allocation plan until the termination condition is met.

7. A multi-objective AGV task allocation optimization method according to claim 6, characterized in that: The location update in step S43 is specifically as follows: At the initial stage of the algorithm, a cheetah population is established based on the coding matrix in step S2. The initial cheetah position is defined by the sequential coding vector of each AGV performing the task. The sequential coding vector formula is as follows: Among them, O t (j) represents the position of the j-th AGV cheetah after the t-th iteration, i.e., the sequential encoding vector; t ji represents the execution order of the i-th task on the j-th AGV, which serves as the initial layout of task allocation; In the exploration phase, a global simulated annealing method is used to set a high initial temperature, gradually lower the temperature through the attenuation factor, adjust the wide search range of the cheetah position, and update the sequence encoding with the hyperbolic tangent step size strategy to explore more global task sequence combinations; During the waiting and attacking stages, a local simulated annealing strategy is implemented, a low initial temperature is set, the attenuation factor is adjusted to accelerate the temperature drop, the hyperbolic tangent step size strategy is used to update the sequential encoding, refine the local task order, and seek the optimal allocation plan.

Images