A multi-fidelity ant colony algorithm-based section scheduling method for gantry cranes in a yard

Abstract

This invention discloses a segmented scheduling method for gantry cranes in a storage yard based on a multi-fidelity ant colony algorithm. First, the ant colony pheromones are initialized based on a benchmark evaluation threshold. Next, in the low-fidelity module, a solution sequence is constructed based on a priority partitioning strategy, and the low-fidelity fitness is calculated. Then, elite ant individuals are selected to enter the high-fidelity evaluation module for evaluation. Finally, based on the benchmark evaluation threshold filtering mechanism of the high-fidelity evaluation result set, the pheromones are locally and globally updated, iterating continuously until the optimal scheduling scheme is output. This invention, through a benchmark evaluation threshold filtering mechanism combined with evaluations in both the low-fidelity and high-fidelity modules, and with elite ant individuals guiding the multi-fidelity ant colony algorithm search, significantly reduces computational redundancy and improves the feasibility and robustness of the gantry crane scheduling problem under complex physical constraints.

Description

Technical Field

[0001] This invention relates to the technical field of ship assembly and automated scheduling, and in particular to a segmented scheduling method for yard gantry cranes based on a multi-fidelity ant colony algorithm. Background Technology

[0002] In modern ship assembly, the scheduling efficiency of yard gantry cranes for sections directly impacts the overall production schedule. With the increasing size of ships and the growing complexity of section structures, the scheduling problem of gantry crane sections in yard operations is becoming increasingly prominent. Some challenges stem from the complexity of the site layout, while others are unique to the operational constraints of gantry cranes during operation. The dynamic interference of yard gantry cranes operating on the same track or in adjacent areas determines that it involves more constraints than conventional logistics scheduling. However, current research on gantry crane scheduling is mostly based on single high-precision simulations. But when facing large-scale section scheduling, the computational cost of algorithms is enormous and the convergence speed is slow because complex conflict detection is required for each iteration.

[0003] Ant colony optimization, as a classic heuristic search algorithm, has unique advantages in solving combinatorial optimization problems. However, in the segmented scheduling of shipyards, achieving a balance between real-time constraint processing and global optimal search is crucial. Traditional algorithms often struggle to find a scheduling sequence that satisfies both constraints and achieves the highest operational efficiency within a limited timeframe. Segmented scheduling of gantry cranes in shipyards needs to consider crane movement and is also subject to strict constraints such as segment priority and safe distances between gantry cranes. Therefore, developing a segmented scheduling method for shipyard gantry cranes that can adaptively adjust evaluation accuracy while balancing computational cost and feasibility is an important means to improve the level of automation in shipbuilding. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings and deficiencies of the prior art and propose a segmented scheduling method for gantry cranes in storage yards based on a multi-fidelity ant colony algorithm. By using a benchmark evaluation threshold filtering mechanism, combined with evaluation by low-fidelity and high-fidelity modules, elite ants guide the multi-fidelity ant colony algorithm search, which significantly reduces computational redundancy and improves the feasibility and robustness of the gantry crane scheduling problem under complex physical constraints.

[0005] To achieve the above objectives, the technical solution provided by this invention is: a segmented scheduling method for yard gantry cranes based on a multi-fidelity ant colony algorithm, wherein the multi-fidelity ant colony algorithm is an improved ant colony algorithm, the improvement of which includes: first, a pheromone initialization and pheromone evaluation threshold filtering mechanism based on a benchmark evaluation threshold, wherein the pheromone initialization uses the operating parameter information under the ship assembly scenario to evaluate and obtain a benchmark evaluation threshold, and the benchmark evaluation threshold is mathematically calculated to assign an initial value to the pheromone, providing an initial starting point; the benchmark evaluation threshold filtering mechanism uses the benchmark evaluation threshold as a threshold for local and global pheromone updates. Second, a low-fidelity module and a high-fidelity module are used for evaluation to improve the efficiency of the algorithm evaluation. The low-fidelity module selects the target segment based on a priority partitioning strategy and estimates the gantry crane's movement distance using the Manhattan distance. The high-fidelity module, based on the Manhattan distance, incorporates the avoidance distance caused by gantry crane operation conflicts to obtain a corrected movement distance. Third, elite ant individuals guide the multi-fidelity ant colony algorithm search. After all ant individuals in the colony are evaluated by the low-fidelity module, the solution sequence and low-fidelity fitness are obtained. Then, based on the low-fidelity fitness, elite ant individuals are selected.

[0006] The specific implementation of the segmented scheduling method for the yard gantry cranes includes the following steps:

[0007] S1: Based on the benchmark evaluation threshold, the pheromone is initialized, the storage yard operation parameter information is obtained, the benchmark evaluation threshold is calculated, and the pheromone of the multi-fidelity ant colony algorithm is initially assigned using the benchmark evaluation threshold.

[0008] S2: The initialized pheromone is used to guide the selection of the next segment in the low-fidelity module during the ant colony iteration process. At the same time, the solution sequence and corresponding low-fidelity fitness of all individual ants are output in the low-fidelity module.

[0009] S3: Summarize the solution sequences and low-fidelity fitness of all ant individuals in the low-fidelity module, sort them according to the size of the low-fidelity fitness, and select elite ant individuals;

[0010] S4: Input the solution sequence of the elite ant individual into the high-fidelity module to obtain the high-fidelity fitness of the elite ant individual. Then, use the benchmark evaluation threshold filtering mechanism to compare the high-fidelity fitness of the elite ant individual with the benchmark evaluation threshold. Perform pheromone local update and global update on the elite ant individuals that pass the benchmark evaluation threshold.

[0011] S5: Execute the multi-fidelity ant colony algorithm in a loop, and select the ant individual with the minimum high-fidelity fitness from the elite ant individuals of each generation as the global optimal ant individual. When the preset iteration stopping condition is met, output the solution sequence and high-fidelity fitness of the global optimal ant individual as the segmented scheduling scheme of the yard gantry crane.

[0012] Furthermore, the specific steps of step S1 are as follows:

[0013] S11: Operational parameter information in the ship assembly scenario is used to configure the gantry crane and site layout, providing a basic assessment environment;

[0014] S12: After configuring the basic evaluation environment using the runtime parameter information, perform an evaluation to obtain the baseline evaluation threshold. ;

[0015] S13: Obtaining the benchmark evaluation threshold Subsequently, the benchmark evaluation threshold is used to initially assign values ​​to the pheromones in the multi-fidelity ant colony algorithm:

[0016] ;

[0017] In the formula, and Represents a segmented index, and ,in The total number of segments, Indicates segmentation and The pheromone values ​​between them.

[0018] Furthermore, in step S2, the execution logic of the low-fidelity evaluation module is as follows:

[0019] S21: In the low-fidelity module, individual ants use a priority partitioning strategy to select the next segment, thereby constructing a solution sequence. The priority partitioning strategy sorts the segments according to their priority. A directed acyclic graph is constructed between the segments to represent the priority relationship between them, and priority groups are formed according to the topological sorting of the graph. Segments without incoming edges are assigned to the highest priority group. After the priority groups are determined, the segments will be dynamically divided into three categories according to the execution progress of the gantry crane: the set of scheduled segments, the set of schedulable segments, and the set of unscheduled segments.

[0020] S22: In the low-fidelity module, individual ants select the next segment from the set of schedulable segments. The selection of the next segment by an individual ant follows a pseudo-random proportional rule, which aims to balance local development and global exploration. The probability of selecting the next segment is determined by the following formula:

[0021] ;

[0022] In the formula, It is a random number drawn from a uniform distribution of 0-1. It is a pseudo-random scaling factor. , and The size relationship determines how the next segment is selected:

[0023] when hour, The next segment selected represents the one that satisfies... The segment index with the largest value, where, Time-sharing The set of schedulable segments at time For the schedulable segment set, the first The index values ​​of each segment, and , The total number of segments, For time-sliced ​​indexes, , The total number of time segments. and These represent the last segment that has been processed and the next segment to be selected, respectively. This indicates that the last segment has been processed. Segments in the schedulable segment set The pheromone values ​​between them The representative is in time slices At that time, gantry crane Segments in the set of schedulable segments The heuristic information value at time, where For gantry crane index, , Total number of gantry cranes, parameters This is used to adjust the weight of heuristic information on decision-making; a larger value means a higher probability of selection. The calculation method is as follows:

[0024] ;

[0025] In the formula, , Represents gantry crane Time Slicing Horizontal and vertical coordinates; , Representative segment The initial horizontal and vertical coordinates located in the stockyard;

[0026] when At that time, a roulette wheel selection method will be used to determine the next segment, calculated using the following formula:

[0027] ;

[0028] In the formula, For random variables, The next segment selected is The probability of that time. For all schedulable segments Segmentation in Summation, Indicates the last segment that has been processed. With the next segment The pheromone values ​​between them The representative is in time slices At that time, gantry crane Schedule the next segment Heuristic information value;

[0029] S23: Gantry Crane The next segment was selected during the operation. Subsequently, the crane will move, resulting in a travel distance. In the low-fidelity module, the avoidance operations performed by the gantry crane due to conflicts during operation are ignored. The travel distance of the gantry crane when moving to the beginning and end positions of the segment is simplified to the Manhattan travel distance, calculated as follows:

[0030] ;

[0031] ;

[0032] In the formula, , Gantry crane Time Slicing Proceed to the next segment Manhattan distance traveled between the starting and ending positions; , Represents the next segment The initial horizontal and vertical coordinates located in the stockyard; , Represents the next segment The end horizontal and end vertical coordinates located in the stockyard;

[0033] S24: Each ant in the low-fidelity module will repeat S21 to S23 in each time slice, thereby gradually constructing a solution sequence and calculating the corresponding low-fidelity fitness. This refers to the gantry crane's travel distance in the low-fidelity module.

[0034] Furthermore, in step S3, the selection of elite ant individuals will be based on low-fidelity fitness. Sort by size from smallest to largest, the smaller the low-fidelity fitness value, the better the ant is, combined with the preset proportion of elite ants. Elite ant individuals are selected.

[0035] Furthermore, the specific steps of step S4 are as follows:

[0036] S41: The high-fidelity module performs a high-fidelity evaluation on the selected elite ant individuals. This evaluation considers avoidance caused by conflict, thus obtaining high-fidelity fitness. The high-fidelity module considers the avoidance caused by conflicts during gantry crane operation, requiring additional calculation of avoidance distances. High-fidelity adaptability... The calculation method is as follows:

[0037] ;

[0038] In the formula, Given a time partition In the middle, there is a gantry crane Schedule a segment The Manhattan movement distance generated at that time This gives the corresponding time slice. In the middle, by the gantry crane Scheduling Segments The additional avoidance distance incurred to resolve conflict, the movement distance within a time slice, should be determined by... and The summation yields the result; if no conflict occurs, no additional avoidance distance needs to be added for calculating high-fidelity fitness. , This represents the sum of the travel distances of all gantry cranes completing all segmented scheduling tasks across all time segments, considering conflict and avoidance operations in the entire problem. The total number of time segments. The total number of gantry cranes. This represents the total number of segments;

[0039] S42: Utilize a benchmark evaluation threshold filtering mechanism to filter the high-fidelity fitness obtained by the high-fidelity module. Compared with the benchmark evaluation threshold Comparison, only when Less than Only when the condition is met will the pheromone local update and global update be performed on the solution sequence. Greater than or equal to In this case, the local and global pheromone updates are skipped;

[0040] S43: When Less than When the condition is met, a local pheromone update is performed. The formula for the local pheromone update is:

[0041] ;

[0042] In the formula, , This is part of the solution sequence constructed by elite ant individuals, located in adjacent positions. The solution sequence constructed by individual ants can be considered as... , The solution sequence constructed using segmented indices can essentially be viewed as a sequence of natural numbers. One rearrangement, in which The total number of segments is given on the right side of the equation. View as segments and The pheromone value before the local update, on the left side of the equation It can be viewed as segmentation. and The pheromone value after local update The local pheromone evaporation coefficient. The initial pheromone value is determined by a benchmark evaluation threshold. The calculation method is as follows: ;

[0043] S44: When Less than When the condition is met, a global pheromone update is performed. The formula for the global pheromone update is:

[0044] ;

[0045] In the formula, , This is part of the solution sequence constructed by the globally optimal ant individual, located in adjacent positions. The solution sequence constructed by the globally optimal ant individual can be considered as... , The solution sequence constructed using segmented indices can essentially be viewed as a sequence of natural numbers. A rearrangement, the right side of the equation It can be viewed as segmentation. and The pheromone value before the global update, on the left side of the equation. It can be viewed as segmentation. and The pheromone value after global update The global pheromone evaporation coefficient. The globally optimal ant individual in this iteration is in the segmentation. and The increment of pheromones released between them is calculated as follows:

[0046] ;

[0047] In the formula, The high-fidelity fitness value is the value of the globally optimal ant individual.

[0048] Furthermore, the specific steps of step S5 are as follows:

[0049] S51: Compare the high-fidelity fitness of each elite ant individual with the high-fidelity fitness of the globally optimal ant individual, and select the ant individual with the minimum high-fidelity fitness from the elite ant individuals of each generation as the globally optimal ant individual.

[0050] S52: After comparing the elite ant individual with the globally optimal ant individual, enter the next iteration loop of the multi-fidelity ant colony algorithm to continuously update the globally optimal ant individual;

[0051] S53: After the iteration ends, output the solution sequence and high-fidelity fitness of the globally optimal ant individual as a segmented scheduling scheme for the gantry crane in the yard.

[0052] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0053] 1. Significantly improved computational efficiency: The method of this invention quickly filters solution sequences through a low-fidelity module and performs high-fidelity evaluation only on elite ant individuals, greatly reducing the number of time-consuming high-fidelity evaluations.

[0054] 2. Enhanced Search Robustness: A benchmark evaluation threshold filtering mechanism is adopted, which performs pheromone local and global updates only on elite ant individuals that are better than the benchmark evaluation threshold. This effectively suppresses the interference of inferior paths, guides the ant colony to converge quickly to high-quality regions, and avoids the algorithm getting stuck in local optima.

[0055] 3. Search space compression: A directed acyclic graph is constructed using segmented priority relationships, and an infeasible solution sequence is filtered out during the solution sequence construction process using a priority partitioning strategy, which reduces the search space from the source and improves the solution efficiency in complex scenarios.

[0056] 4. Flexible and adaptive operation response: By utilizing dynamic heuristic information to provide real-time feedback on the dynamic relationship between the gantry crane position and segmented tasks, the scheduling scheme can be adaptively adjusted according to the real-time status of the yard, which has significant engineering practical value. Attached Figure Description

[0057] Figure 1 This is a flowchart outlining the framework of the method of this invention.

[0058] Figure 2 It is a priority-based strategy.

[0059] Figure 3 It is the logic for building low-fidelity modules.

[0060] Figure 4 It is the logic for building high-fidelity modules. Detailed Implementation

[0061] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0062] This embodiment discloses a segmented scheduling method for yard gantry cranes based on a multi-fidelity ant colony algorithm. It assumes the following key parameters are considered in a ship assembly scenario: gantry crane operating parameters, segment attributes, and site parameters. The multi-fidelity ant colony algorithm is an improved ant colony algorithm, with improvements including: First, a pheromone initialization and filtering mechanism based on a benchmark evaluation threshold. The pheromone initialization uses operating parameter information from the ship assembly scenario to evaluate and obtain a benchmark evaluation threshold. This benchmark evaluation threshold is mathematically calculated to assign an initial value to the pheromone, providing an initial starting point. The benchmark evaluation threshold filtering mechanism uses the benchmark evaluation threshold as the pheromone. The first step involves setting thresholds for local and global updates. The second step uses both low-fidelity and high-fidelity modules for evaluation to improve algorithm efficiency. The low-fidelity module selects target segments based on a priority partitioning strategy and estimates the gantry crane's movement distance using Manhattan distance. The high-fidelity module, based on the Manhattan distance, incorporates the avoidance distance caused by gantry crane operation conflicts to obtain a corrected movement distance. The third step involves elite ant individuals guiding the multi-fidelity ant colony algorithm search. After evaluation by the low-fidelity module, all ant individuals in the colony obtain solution sequences and low-fidelity fitness. Then, based on the low-fidelity fitness, elite ant individuals are selected.

[0063] like Figures 1 to 4 As shown, the specific implementation of the sectional scheduling method for the yard gantry crane includes the following steps:

[0064] S1: Based on the benchmark evaluation threshold, pheromone initialization is performed by obtaining the stockpile operation parameter information, calculating the benchmark evaluation threshold, and using the benchmark evaluation threshold to initially assign pheromone values ​​to the multi-fidelity ant colony algorithm; the specific operation steps are as follows:

[0065] S11: Operational parameter information in the ship assembly scenario is used to configure the gantry crane and site layout, providing a basic assessment environment;

[0066] S12: After configuring the basic evaluation environment using the runtime parameter information, an evaluation can be performed to obtain the baseline evaluation threshold. ;

[0067] S13: Obtaining the benchmark evaluation threshold Subsequently, the benchmark evaluation threshold is used to initially assign values ​​to the pheromones in the multi-fidelity ant colony algorithm:

[0068] ;

[0069] In the formula, and Represents a segmented index, and ,in The total number of segments, Indicates segmentation and The pheromone values ​​between them.

[0070] S2: The initialized pheromone is used to guide the selection of the next segment in the low-fidelity module during the ant colony iteration process. Simultaneously, the low-fidelity module outputs the solution sequence constructed by all individual ants and their corresponding low-fidelity fitness; for example... Figure 2 and Figure 3 As shown, the execution logic of the low-fidelity evaluation module is as follows:

[0071] S21: In the low-fidelity module, individual ants use a priority partitioning strategy to select the next segment, thereby constructing a solution sequence. The priority partitioning strategy sorts the segments according to their priority. A directed acyclic graph is constructed between the segments to represent the priority relationship between them, and priority groups are formed according to the topological sorting of the graph. Segments without incoming edges are assigned to the highest priority group. After the priority groups are determined, the segments will be dynamically divided into three categories according to the execution progress of the gantry crane: the set of scheduled segments, the set of schedulable segments, and the set of unscheduled segments.

[0072] S22: In the low-fidelity module, individual ants select the next segment from the set of schedulable segments. The selection of the next segment by an individual ant follows a pseudo-random proportional rule, which aims to balance local development and global exploration. The probability of selecting the next segment is determined by the following formula:

[0073] ;

[0074] In the formula, It is a random number drawn from a uniform distribution of 0-1. It is a pseudo-random scaling factor. , and The size relationship determines how the next segment is selected:

[0075] when hour, The next segment selected represents the one that satisfies... The segment index with the largest value, where, Time-sharing The set of schedulable segments, where For the schedulable segment set, the first The index values ​​of each segment, and , The total number of segments, For time-sliced ​​indexes, , The total number of time segments. and These represent the last segment that has been processed and the next segment to be selected, respectively. This indicates that the last segment has been processed. Segments in the schedulable segment set The pheromone values ​​between them The representative is in time slices At that time, gantry crane Segments in the set of schedulable segments The heuristic information value at time, where For gantry crane index, , Total number of gantry cranes, parameters This is used to adjust the weight of heuristic information on decision-making; a larger value means a higher probability of selection. The calculation method is as follows:

[0076] ;

[0077] In the formula, , Represents gantry crane Time Slicing Horizontal and vertical coordinates; , Representative segment The initial horizontal and vertical coordinates located in the stockyard;

[0078] when At that time, a roulette wheel selection method will be used to determine the next segment, calculated using the following formula:

[0079] ;

[0080] In the formula, For random variables, The next segment selected is The probability of that time. For all schedulable segments Segmentation in Summation, Indicates the last segment that has been processed. With the next segment The pheromone values ​​between them The representative is in time slices At that time, gantry crane Schedule the next segment Heuristic information value;

[0081] S23: Gantry Crane The next segment was selected during the operation. Subsequently, the crane will move, resulting in a travel distance. In the low-fidelity module, the avoidance operations performed by the gantry crane due to conflicts during operation are ignored. The travel distance of the gantry crane when moving to the beginning and end positions of the segment is simplified to the Manhattan travel distance, calculated as follows:

[0082] ;

[0083] ;

[0084] In the formula, Gantry crane Time Slicing Proceed to the next segment Manhattan distance traveled between the starting and ending positions; , Represents the next segment The initial horizontal and vertical coordinates located in the stockyard; , Represents the next segment The end horizontal and end vertical coordinates located in the stockyard;

[0085] S24: Each ant in the low-fidelity module will repeat S21 to S23 in each time slice, thereby gradually constructing a solution sequence and calculating the corresponding low-fidelity fitness. This refers to the gantry crane's travel distance in the low-fidelity module.

[0086] S3: Summarize the solution sequences and low-fidelity fitness of all ant individuals in the low-fidelity module, sort them according to their low-fidelity fitness, and select elite ant individuals; wherein, the selection of elite ant individuals will be based on low-fidelity fitness. Sort by size from smallest to largest, the smaller the low-fidelity fitness value, the better the ant is, combined with the preset proportion of elite ants. Elite ant individuals are selected.

[0087] S4: Input the solution sequence of the elite ant individual into the high-fidelity module to obtain the high-fidelity fitness of the elite ant individual. Then, use the benchmark evaluation threshold filtering mechanism to compare the high-fidelity fitness of the elite ant individual with the benchmark evaluation threshold. Perform pheromone local update and global update on the elite ant individuals that pass the benchmark evaluation threshold. The specific operation steps are as follows:

[0088] S41: The high-fidelity module performs a high-fidelity evaluation on the selected elite ant individuals. This evaluation considers avoidance caused by conflict, thus obtaining high-fidelity fitness. The high-fidelity module considers the avoidance caused by conflicts during gantry crane operation, requiring additional calculation of avoidance distances. High-fidelity adaptability... The calculation method is as follows:

[0089] ;

[0090] In the formula, Given a time partition In the middle, there is a gantry crane Schedule a segment The Manhattan movement distance generated at that time This gives the corresponding time slice. In the middle, by the gantry crane Scheduling Segments The additional avoidance distance incurred to resolve conflict, the movement distance within a time slice, should be determined by... and The summation yields the result; if no conflict occurs, no additional avoidance distance needs to be added for calculating high-fidelity fitness. , This represents the sum of the travel distances of all gantry cranes completing all segmented scheduling tasks across all time segments, considering conflict and avoidance operations in the entire problem. The total number of time segments. The total number of gantry cranes. This represents the total number of segments;

[0091] S42: Utilize a benchmark evaluation threshold filtering mechanism to filter the high-fidelity fitness obtained by the high-fidelity module. Compared with the benchmark evaluation threshold Comparison, only when Less than Only when the condition is met will the pheromone local update and global update be performed on the solution sequence. Greater than or equal to In this case, the local and global pheromone updates are skipped;

[0092] S43: When Less than When the condition is met, a local pheromone update is performed. The formula for the local pheromone update is:

[0093] ;

[0094] In the formula, , This is part of the solution sequence constructed by elite ant individuals, located in adjacent positions. The solution sequence constructed by individual ants can be considered as... , The solution sequence constructed using segmented indices can essentially be viewed as a sequence of natural numbers. One rearrangement, in which The total number of segments is given on the right side of the equation. View as segments and The pheromone value before the local update, on the left side of the equation It can be viewed as segmentation. and The pheromone value after local update The local pheromone evaporation coefficient. The initial pheromone value is determined by a benchmark evaluation threshold. The calculation method is as follows: ;

[0095] S44: When Less than When the condition is met, a global pheromone update is performed. The formula for the global pheromone update is:

[0096] ;

[0097] In the formula, , This is part of the solution sequence constructed by the globally optimal ant individual, located in adjacent positions. The solution sequence constructed by the globally optimal ant individual can be considered as... , The solution sequence constructed using segmented indices can essentially be viewed as a sequence of natural numbers. A rearrangement, the right side of the equation It can be viewed as segmentation. and The pheromone value before the global update, on the left side of the equation. It can be viewed as segmentation. and The pheromone value after global update The global pheromone evaporation coefficient. The globally optimal ant individual in this iteration is in the segmentation. and The increment of pheromones released between them is calculated as follows:

[0098] ;

[0099] In the formula, The high-fidelity fitness value is the value of the globally optimal ant individual.

[0100] S5: Execute the multi-fidelity ant colony algorithm repeatedly, and select the ant individual with the minimum high-fidelity fitness from the elite ant individuals of each generation as the globally optimal ant individual. When the preset iteration stopping condition is met, output the solution sequence and high-fidelity fitness of the globally optimal ant individual as the segmented scheduling scheme for the gantry crane in the yard; the specific operation steps are as follows:

[0101] S51: Compare the high-fidelity fitness of each elite ant individual with the high-fidelity fitness of the globally optimal ant individual, and select the ant individual with the minimum high-fidelity fitness from the elite ant individuals of each generation as the globally optimal ant individual.

[0102] S52: After comparing the elite ant individual with the globally optimal ant individual, enter the next iteration loop of the multi-fidelity ant colony algorithm to continuously update the globally optimal ant individual;

[0103] S53: After the iteration ends, output the solution sequence and high-fidelity fitness of the globally optimal ant individual as a segmented scheduling scheme for the gantry crane in the yard.

[0104] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A segmented scheduling method for yard gantry cranes based on a multi-fidelity ant colony algorithm, characterized in that, The multi-fidelity ant colony algorithm is an improved ant colony algorithm, and its improvements include: First, a pheromone initialization and pheromone evaluation threshold filtering mechanism based on a benchmark evaluation threshold. The pheromone initialization uses operational parameter information in the ship assembly scenario to evaluate and obtain a benchmark evaluation threshold. This benchmark evaluation threshold is mathematically calculated to assign initial values ​​to the pheromone, providing an initial starting point. The benchmark evaluation threshold filtering mechanism uses the benchmark evaluation threshold as a threshold for local and global pheromone updates. Second, low-fidelity and high-fidelity modules are used for evaluation to improve the algorithm's evaluation efficiency. The low-fidelity module selects target segments based on a priority partitioning strategy and estimates the gantry crane's movement distance using Manhattan distance. The high-fidelity module, based on the Manhattan distance, incorporates the avoidance distance caused by gantry crane operation conflicts to obtain a corrected movement distance. Third, elite ant individuals guide the multi-fidelity ant colony algorithm search. After all ant individuals in the colony are evaluated by the low-fidelity module, they obtain a solution sequence and low-fidelity fitness. Then, they are sorted according to their low-fidelity fitness to select elite ant individuals. The specific implementation of the segmented scheduling method for the yard gantry cranes includes the following steps: S1: Based on the benchmark evaluation threshold, the pheromone is initialized, the storage yard operation parameter information is obtained, the benchmark evaluation threshold is calculated, and the pheromone of the multi-fidelity ant colony algorithm is initially assigned using the benchmark evaluation threshold. S2: The initialized pheromone is used to guide the selection of the next segment in the low-fidelity module during the ant colony iteration process. At the same time, the solution sequence and corresponding low-fidelity fitness of all individual ants are output in the low-fidelity module. S3: Summarize the solution sequences and low-fidelity fitness of all ant individuals in the low-fidelity module, sort them according to the size of the low-fidelity fitness, and select elite ant individuals; S4: Input the solution sequence of the elite ant individual into the high-fidelity module to obtain the high-fidelity fitness of the elite ant individual. Then, use the benchmark evaluation threshold filtering mechanism to compare the high-fidelity fitness of the elite ant individual with the benchmark evaluation threshold. Perform pheromone local update and global update on the elite ant individuals that pass the benchmark evaluation threshold. S5: Execute the multi-fidelity ant colony algorithm in a loop, and select the ant individual with the minimum high-fidelity fitness from the elite ant individuals of each generation as the global optimal ant individual. When the preset iteration stopping condition is met, output the solution sequence and high-fidelity fitness of the global optimal ant individual as the segmented scheduling scheme of the yard gantry crane.

2. The method for segmented scheduling of yard gantry cranes based on multi-fidelity ant colony algorithm according to claim 1, characterized in that, The specific steps for step S1 are as follows: S11: Operational parameter information in the ship assembly scenario is used to configure the gantry crane and site layout, providing a basic assessment environment; S12: After configuring the basic evaluation environment using the runtime parameter information, perform an evaluation to obtain the baseline evaluation threshold. ; S13: Obtaining the benchmark evaluation threshold Subsequently, the benchmark evaluation threshold is used to initially assign values ​​to the pheromones in the multi-fidelity ant colony algorithm: ； In the formula, and Represents a segmented index, and ,in The total number of segments, Indicates segmentation and The pheromone values ​​between them.

3. The method for segmented scheduling of yard gantry cranes based on multi-fidelity ant colony algorithm according to claim 2, characterized in that, In step S2, the execution logic of the low-fidelity evaluation module is as follows: S21: In the low-fidelity module, individual ants use a priority partitioning strategy to select the next segment, thereby constructing a solution sequence. The priority partitioning strategy sorts the segments according to their priority. A directed acyclic graph is constructed between the segments to represent the priority relationship between them, and priority groups are formed according to the topological sorting of the graph. Segments without incoming edges are assigned to the highest priority group. After the priority groups are determined, the segments will be dynamically divided into three categories according to the execution progress of the gantry crane: the set of scheduled segments, the set of schedulable segments, and the set of unscheduled segments. S22: In the low-fidelity module, individual ants select the next segment from the set of schedulable segments. The selection of the next segment by an individual ant follows a pseudo-random proportional rule, which aims to balance local development and global exploration. The probability of selecting the next segment is determined by the following formula: ； In the formula, It is a random number drawn from a uniform distribution of 0-1. It is a pseudo-random scaling factor. , and The size relationship determines how the next segment is selected: when hour, The next segment selected represents the one that satisfies... The segment index with the largest value, where, Time-sharing The set of schedulable segments at time For the schedulable segment set, the first The index values ​​of each segment, and , The total number of segments, For time-sliced ​​indexes, , The total number of time segments. and These represent the last segment that has been processed and the next segment to be selected, respectively. This indicates that the last segment has been processed. Segments in the schedulable segment set The pheromone values ​​between them The representative is in time slices At that time, gantry crane Segments in the set of schedulable segments The heuristic information value at time, where For gantry crane index, , Total number of gantry cranes, parameters This is used to adjust the weight of heuristic information on decision-making; a larger value means a higher probability of selection. The calculation method is as follows: ； In the formula, , Represents gantry crane Time Slicing Horizontal and vertical coordinates; , Representative segment The initial horizontal and initial vertical coordinates located in the stockyard; when At that time, a roulette wheel selection method will be used to determine the next segment, calculated using the following formula: ； In the formula, For random variables, The next segment selected is The probability of that time. For all schedulable segments Segmentation in Summation, Indicates the last segment that has been processed. With the next segment The pheromone values ​​between them The representative is in time slices At that time, gantry crane Schedule the next segment Heuristic information value; S23: Gantry Crane The next segment was selected during the operation. Subsequently, the crane will move, resulting in a travel distance. In the low-fidelity module, the avoidance operations performed by the gantry crane due to conflicts during operation are ignored. The travel distance of the gantry crane when moving to the beginning and end positions of the segment is simplified to the Manhattan travel distance, calculated as follows: ； ； In the formula, , Gantry crane Time Slicing Proceed to the next segment Manhattan distance traveled between the starting and ending positions; , Represents the next segment The initial horizontal and initial vertical coordinates located in the stockyard; , Represents the next segment The end horizontal and end vertical coordinates located in the stockyard; S24: Each ant in the low-fidelity module will repeat S21 to S23 in each time slice, thereby gradually constructing a solution sequence and calculating the corresponding low-fidelity fitness. This refers to the gantry crane's travel distance in the low-fidelity module.

4. The method for segmented scheduling of yard gantry cranes based on multi-fidelity ant colony algorithm according to claim 3, characterized in that, In step S3, the selection of elite ant individuals will be based on low-fidelity fitness. Sort by size from smallest to largest, the smaller the low-fidelity fitness value, the better the ant is, combined with the preset proportion of elite ants. Elite ant individuals are selected.

5. A method for segmented scheduling of yard gantry cranes based on multi-fidelity ant colony algorithm according to claim 4, characterized in that, The specific steps for step S4 are as follows: S41: The high-fidelity module performs a high-fidelity evaluation on the selected elite ant individuals. This evaluation considers avoidance caused by conflict, thus obtaining high-fidelity fitness. The high-fidelity module considers the avoidance caused by conflicts during gantry crane operation, requiring additional calculation of avoidance distances. High-fidelity adaptability... The calculation method is as follows: ； In the formula, Given a time partition In the middle, there is a gantry crane Schedule a segment The Manhattan movement distance generated at that time This gives the corresponding time slice. In the middle, by the gantry crane Scheduling Segments The additional avoidance distance incurred to resolve conflict, the movement distance within a time slice, should be determined by... and The summation yields the result; if no conflict occurs, no additional avoidance distance needs to be added for calculating high-fidelity fitness. , This represents the sum of the travel distances of all gantry cranes completing all segmented scheduling tasks across all time segments, considering conflict and avoidance operations in the entire problem. The total number of time segments. The total number of gantry cranes. This represents the total number of segments; S42: Utilize a benchmark evaluation threshold filtering mechanism to filter the high-fidelity fitness obtained by the high-fidelity module. Compared with the benchmark evaluation threshold Comparison, only when Less than Only when the condition is met will pheromone local and global updates be performed on the solution sequence. Greater than or equal to In this case, the local and global pheromone updates are skipped; S43: When Less than When the condition is met, a local pheromone update is performed. The formula for the local pheromone update is: ； In the formula, , This is part of the solution sequence constructed by elite ant individuals, located in adjacent positions. The solution sequence constructed by individual ants can be considered as... , The solution sequence constructed using segmented indices can essentially be viewed as a sequence of natural numbers. One rearrangement, in which The total number of segments is given on the right side of the equation. View as segments and The pheromone value before the local update, on the left side of the equation It can be viewed as segmentation. and The pheromone value after local update The local pheromone evaporation coefficient. The initial pheromone value is determined by a benchmark evaluation threshold. The calculation method is as follows: ; S44: When Less than When the condition is met, a global pheromone update is performed. The formula for the global pheromone update is: ； In the formula, , This is part of the solution sequence constructed by the globally optimal ant individual, located in adjacent positions. The solution sequence constructed by the globally optimal ant individual can be considered as... , The solution sequence constructed using segmented indices can essentially be viewed as a sequence of natural numbers. A rearrangement, the right side of the equation It can be viewed as segmentation. and The pheromone value before the global update, on the left side of the equation. It can be viewed as segmentation. and The pheromone value after global update The global pheromone evaporation coefficient. The globally optimal ant individual in this iteration is in the segmentation. and The increment of pheromones released between them is calculated as follows: ； In the formula, The high-fidelity fitness value is the value of the globally optimal ant individual.

6. The segmented scheduling method for gantry cranes based on multi-fidelity ant colony algorithm according to claim 5, characterized in that, The specific steps for step S5 are as follows: S51: Compare the high-fidelity fitness of each elite ant individual with the high-fidelity fitness of the globally optimal ant individual, and select the ant individual with the minimum high-fidelity fitness from the elite ant individuals of each generation as the globally optimal ant individual. S52: After comparing the elite ant individual with the globally optimal ant individual, enter the next iteration loop of the multi-fidelity ant colony algorithm to continuously update the globally optimal ant individual; S53: After the iteration ends, output the solution sequence and high-fidelity fitness of the globally optimal ant individual as the segmented scheduling scheme for the gantry crane in the yard.

Images