Method and system for filling articles in limited space
Through the combination of tower loading heuristic algorithm, ant colony algorithm and genetic algorithm, the problems of inefficiency and insufficient dynamic adaptability in the three-dimensional packaging problem are solved, and an efficient and flexible cargo loading process is achieved, suitable for complex loading scenarios and improving The operational efficiency of the logistics distribution center.
Patent Information
- Application Number
- CN202410981545.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-22
AI Technical Summary
When handling large and diverse types of goods, the three-dimensional packaging problem is inefficient and dynamic adaptability is insufficient, making it difficult to generate the optimal filling solution within a reasonable time.
Combining tower loading heuristic algorithm, ant colony algorithm and genetic algorithm, through multi-objective optimization of the fitness function, the loading scheme is gradually optimized to achieve dynamic updates and efficient loading.
It significantly improves loading efficiency, stability and adaptability, reduces operational difficulty and time costs, and is suitable for complex loading scenarios, reduces transportation vehicles, reduces operating costs, and improves the overall operational efficiency of logistics and distribution centers.
Smart Images

Figure CN118941015A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of three-dimensional packaging, and in particular to a method and system for filling articles in a limited space. Background Art
[0002] The loading and transportation of goods is the most critical step in the logistics industry. Considering the efficiency of cargo packing, cargo packing is usually converted into a three-dimensional packing problem. The three-dimensional packing problem is a classic combinatorial optimization problem, and the optimal packing solution cannot be solved within the polynomial time complexity. If an exact algorithm is used to solve the problem, it can only solve the three-dimensional packing problem with a small number of cargo types and quantities. When the number of cargo types and quantities increases, the number of feasible solutions in the solution space will increase exponentially. Therefore, it is necessary to design an optimization algorithm to solve the three-dimensional packing problem in a reasonable time.
[0003] The existing technology generally uses the ant algorithm or simulated annealing algorithm alone to solve the three-dimensional packing problem. It can load simple and small quantities of similar goods, but it is not applicable to the loading of goods in large logistics distribution centers. The goods in large logistics centers flow frequently and the loading demand may change at any time. The dynamic adaptability of this method is insufficient. Summary of the invention
[0004] In view of this, the present invention proposes a method and system for loading items in a limited space. Through the combination of tower loading heuristic algorithm, ant colony algorithm and genetic algorithm, it is possible to generate an efficient loading plan within a reasonable time and realize dynamic update plan, thereby solving the problems of low efficiency and insufficient dynamic adaptability of traditional methods when processing large quantities and multiple types of goods.
[0005] The technical solution of the present invention is implemented as follows: On the one hand, the present invention provides a method for filling objects in a limited space, comprising the following steps:
[0006] S1, reading the filling space data of the filling space, and constructing a filling space coordinate system based on the filling space data;
[0007] S2, obtaining the three-dimensional data of all the objects to be loaded, and constructing a plurality of three-dimensional models of the objects based on the three-dimensional data of the objects;
[0008] S3, sorting the plurality of three-dimensional object models by a tower loading heuristic algorithm to generate an initial loading solution set;
[0009] S4, performing a global search on the multiple initial loading solutions by using an ant colony algorithm to obtain a first loading solution set;
[0010] S5, constructing a genetic fitness function, screening the first filling scheme set by using the genetic fitness function to obtain a second filling scheme set, performing a crossover operation and a mutation operation on the second filling scheme set to obtain a third filling scheme set;
[0011] S6, constructing a multi-objective optimization fitness function, and performing effect evaluation and screening on the third filling scheme set by using the multi-objective optimization fitness function to obtain an optimal filling scheme;
[0012] S7, acquiring updated three-dimensional data of the newly added objects to be loaded in real time, and adjusting the optimal loading solution based on the updated three-dimensional data of the objects.
[0013] On the basis of the above technical solution, preferably, step S3 includes:
[0014] A plurality of loading strategies are formulated through a tower loading heuristic algorithm. The loading strategies use volume as a first reference indicator, and one of the center of gravity, length, width or height of the items to be loaded as a second and a third reference indicator. Based on the plurality of loading strategies, the three-dimensional models of the plurality of items are sorted respectively to obtain a plurality of loading schemes, which are summarized as an initial loading scheme set.
[0015] Based on the above technical solution, preferably, step S4 includes:
[0016] S41, initialize various parameters of the ant algorithm and construct the problem graph structure;
[0017] S42, ant path selection is performed based on the probability transfer rule, a loading plan is obtained and evaluated;
[0018] S43, iteratively optimize the ant colony algorithm and update the pheromone;
[0019] S44, recording the best packing solution found by the ants in each iteration, summarizing the best packing solution corresponding to each ant, and forming a first packing solution set.
[0020] Based on the above technical solution, preferably, step S41 includes:
[0021] A problem graph structure is constructed with the loading space and the items to be loaded, the number of ants in the ant colony, the initial pheromone concentration and the pheromone volatility coefficient are set, and the initial loading scheme set is used as the initial solution. The nodes in the problem graph structure represent the items to be loaded, and each edge in the problem graph structure represents each loading position of the items to be loaded.
[0022] Based on the above technical solution, preferably, step S42 includes:
[0023] Each ant selects a path order based on the probability transfer rule to load the items to be loaded in sequence, records the loading path selected by each ant, obtains multiple loading schemes, calculates the fitness of each loading scheme, and evaluates the multiple loading schemes. The calculation formula of the probability transfer rule is:
[0024]
[0025] Among them, P ab is the probability of an ant moving from node a to node b, T ab is the pheromone concentration on edge ab, η ab is the heuristic information of edge ab, α is the adjustment parameter of pheromone concentration, β is the adjustment parameter of heuristic information, T ak is the pheromone concentration on edge ak, η ak is the heuristic information of edge ak, and K is the number of all nodes reachable by node a.
[0026] Based on the above technical solution, preferably, step S43 includes:
[0027] The pheromone is volatilized according to the preset volatilization coefficient, the pheromone concentration on each edge of the problem graph structure is updated, and the pheromone concentration on the loading path is increased:
[0028] T ij '=(1-ρ)T ij ;
[0029] T ij =T ij '+ΔT ij ;
[0030] Among them, T ij is the pheromone concentration on edge ij, ρ is the volatility coefficient, T ij ' is the pheromone concentration on the ij after volatilization, T ij ” is the pheromone concentration after the pheromone concentration is added on edge ij, ΔT ij is the concentration of the newly added pheromone on edge ij;
[0031] The ant colony algorithm is iteratively optimized, and the iterative stopping conditions of the iterative optimization include: the pheromone concentration reaches a preset value, the fitness reaches a preset value, and the number of iterations reaches a preset value.
[0032] Based on the above technical solution, preferably, step S5 includes:
[0033] S51, selecting multiple loading schemes from the first loading scheme as an initial population, and setting a crossover operation probability and a mutation operation probability;
[0034] S52, constructing a genetic fitness function, performing fitness evaluation on the initial population based on the genetic fitness function, selecting the initial population with a fitness higher than a preset value in the initial population as the second loading scheme set to enter the breeding pool, randomly selecting two loadings in the second loading scheme set as parents, judging whether to crossover according to the crossover operation probability of the two loading schemes, and when judging to crossover, selecting a crossover point, exchanging the parts of the two loading schemes at the crossover point, and generating an initial offspring loading scheme;
[0035] S53, randomly selecting an initial offspring filling scheme and a mutation position according to the mutation operation probability, changing the filling order or filling strategy of the mutation position of the selected initial offspring filling scheme, and generating an offspring filling scheme;
[0036] S54, iteratively train the offspring loading scheme generated each time, repeat steps S51-S53, and obtain a third loading scheme set. The iterative stopping conditions of the iterative training include: the number of iterations reaches a preset maximum value and the fitness converges.
[0037] On the basis of the above technical solution, preferably, step S6 includes:
[0038] The multi-objective optimization fitness function includes a comprehensive fitness function and a multi-objective optimization algorithm. The multi-objective optimization algorithm adopts a Pareto frontier optimization algorithm. The construction process of the comprehensive fitness function includes: selecting an optimization target, the optimization target includes space utilization, loading stability, and loading operation convenience, constructing multiple single fitness functions according to the optimization target, and comprehensively weighting the multiple single fitness functions to obtain a comprehensive fitness function;
[0039] The third filling scheme is used as the initial population of the multi-objective optimization fitness function, and an evolutionary operation is performed on the initial population. The evolutionary operation includes crossover, mutation and selection. The fitness of the new initial population generated by each generation of evolution is evaluated, and the optimal solution of each generation of evolution is retained. The stopping conditions of the evolutionary operation include: the number of evolutions reaches a preset value, the fitness converges, and after the evolutionary operation iteration stops, the optimal filling scheme is obtained.
[0040] Based on the above technical solution, preferably, step S7 includes:
[0041] The updated three-dimensional data of the newly added objects to be loaded are read by a 3D laser scanner, and the updated three-dimensional data of the objects are filtered, denoised and gridded. A three-dimensional model of the newly added objects to be loaded is constructed based on the updated three-dimensional data of the objects. The multi-objective optimization fitness function is updated by the three-dimensional model of the newly added objects to be loaded to obtain a first multi-objective optimization fitness function. The fitness is recalculated based on the first multi-objective optimization fitness function, and the multi-objective optimization is re-performed. According to the recalculated fitness and the optimized structure, a dynamically updated loading plan is generated.
[0042] On the other hand, the present invention also provides a system for filling objects in a limited space, the system comprising:
[0043] A coordinate system construction module, used for reading the filling space data of the filling space, and constructing the filling space coordinate system based on the filling space data;
[0044] An item model building module, used for acquiring item 3D data of all items to be loaded, and building a plurality of item 3D models based on the item 3D data;
[0045] A heuristic sorting module, used for sorting the plurality of three-dimensional object models by a tower loading heuristic algorithm to generate an initial loading solution set;
[0046] An ant colony algorithm module, configured to perform a global search on the multiple initial loading solutions by using an ant colony algorithm to obtain a first loading solution set;
[0047] A genetic algorithm module, used for constructing a genetic fitness function, screening the first filling scheme set by using the genetic fitness function to obtain a second filling scheme set, performing a crossover operation and a mutation operation on the second filling scheme set to obtain a third filling scheme set;
[0048] A multi-objective optimization fitness module, used to construct a multi-objective optimization fitness function, and to evaluate and screen the third filling scheme set through the multi-objective optimization fitness function to obtain an optimal filling scheme;
[0049] The real-time scheme updating module is used to obtain updated three-dimensional data of newly added items to be loaded in real time, and adjust the optimal loading scheme based on the updated three-dimensional data of the items.
[0050] The method and system for loading articles in a limited space of the present invention have the following beneficial effects compared with the prior art:
[0051] (1) Through 3D laser scanning technology to obtain accurate loading space and cargo three-dimensional data, combined with tower loading heuristic algorithm, ant colony algorithm, genetic algorithm and multi-objective optimization algorithm, the loading plan is gradually optimized to achieve an efficient, flexible and stable cargo loading process. It not only improves space utilization and loading efficiency, but also has intelligent adaptability and real-time dynamic adjustment capabilities, significantly improving loading efficiency, stability and safety, reducing operational difficulty and time cost, and is suitable for various complex loading scenarios. Ultimately, it reduces the number of transport trips, reduces operating costs, and improves the overall operating efficiency and service quality of the logistics distribution center;
[0052] (2) The tower loading heuristic algorithm is used to sort multiple three-dimensional models of items and generate multiple sets of initial loading solutions, which are based on parameters such as volume, center of gravity, length, width, and height, and preliminarily improve the utilization of the loading space;
[0053] (3) Constructing a genetic fitness function, screening the first set of loading solutions to obtain the second set of loading solutions, and generating the third set of loading solutions through crossover and mutation operations, which significantly improves the adaptability and diversity of the loading solutions;
[0054] (4) The third set of loading solutions is evaluated and screened through a multi-objective optimization fitness function to obtain the optimal loading solution, taking into account space utilization, loading stability, and operational convenience to ensure comprehensive optimization of the loading solution. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0056] Figure 1 This is a flow chart of a method for loading items in a limited space according to the present invention;
[0057] Figure 2 This is a structural diagram of a system for loading items in a limited space according to the present invention. DETAILED DESCRIPTION
[0058] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0059] See also Figure 1 This embodiment provides a method for loading objects in a limited space, comprising the following steps:
[0060] S1, reading the filling space data of the filling space, and constructing a filling space coordinate system based on the filling space data;
[0061] S2, obtaining the three-dimensional data of all the objects to be loaded, and constructing a plurality of three-dimensional models of the objects based on the three-dimensional data of the objects;
[0062] S3, sorting the plurality of three-dimensional object models by a tower loading heuristic algorithm to generate an initial loading solution set;
[0063] S4, performing a global search on the multiple initial loading solutions by using an ant colony algorithm to obtain a first loading solution set;
[0064] S5, constructing a genetic fitness function, screening the first filling scheme set by using the genetic fitness function to obtain a second filling scheme set, performing a crossover operation and a mutation operation on the second filling scheme set to obtain a third filling scheme set;
[0065] S6, constructing a multi-objective optimization fitness function, and performing effect evaluation and screening on the third filling scheme set by using the multi-objective optimization fitness function to obtain an optimal filling scheme;
[0066] S7, acquiring updated three-dimensional data of the newly added objects to be loaded in real time, and adjusting the optimal loading solution based on the updated three-dimensional data of the objects.
[0067] This embodiment combines the tower loading heuristic algorithm, the ant colony algorithm and the genetic algorithm to generate an efficient loading plan within a reasonable time and realize dynamic update of the plan, thus solving the problem of low efficiency of traditional methods when processing large quantities and multiple types of goods.
[0068] A multi-objective optimization fitness function is adopted to comprehensively consider multiple optimization objectives such as space utilization, loading stability and loading operation convenience to ensure that the generated loading plan is optimal in multiple aspects.
[0069] The three-dimensional data of newly added items to be loaded are acquired in real time, and the optimal loading plan is adjusted based on these data, ensuring the flexibility and adaptability of the loading plan.
[0070] Using the ant colony algorithm for global search and the genetic algorithm for local optimization can effectively avoid local optimal solutions and improve the overall quality of the loading solution.
[0071] Specifically, step S1 includes:
[0072] The filling space data of the filling space is read by a 3D laser scanner, and the filling space data is filtered, denoised and gridded using point cloud processing software. The filling space data includes length, width, height, volume and right rear lower corner point. A spatial coordinate system is constructed with the length as the X-axis, the width as the Y-axis, the height as the Z-axis, and the right rear lower corner point as the origin O.
[0073] In step S1 of this embodiment, detailed data of the loading space is read by a 3D laser scanner, and point cloud processing software is used for filtering, denoising and gridding, thereby ensuring the accuracy and clarity of the data. By obtaining key parameters such as the length, width, height, volume and right rear lower corner point of the loading space, and constructing a spatial coordinate system with these parameters, an accurate and standardized spatial reference frame is provided. Step S1 solves the problems of unreasonable loading and space waste caused by inaccurate spatial data in traditional methods, and significantly improves loading efficiency and space utilization.
[0074] Specifically, step S2 includes:
[0075] Reading the three-dimensional data of all the objects to be loaded by a 3D laser scanner, and building a three-dimensional model of the objects based on the three-dimensional data of the objects;
[0076] For the non-standard shaped objects to be filled, the simulation process is performed to obtain the most similar standard shaped objects, and the specific steps include:
[0077] Connecting a pair of first vertices and a second vertices that are farthest from each other among the vertices of the objects to be loaded to form a first dividing line;
[0078] The vertices of the object to be filled are divided into a point set on the left side of the first dividing line and a point set on the right side of the first dividing line by the first dividing line, and the third vertex in the point set on the left side of the first dividing line, which is farthest from the first dividing line, is selected, and the third vertex is connected to the first vertex and the third vertex to the second vertex, respectively, to form a second dividing line and a third dividing line;
[0079] The dividing lines are formed repeatedly until the most similar standard shape object is formed.
[0080] Step S2 uses a 3D laser scanner to read the 3D data of all items to be loaded, and builds a 3D model of the items based on these data. For items with non-standard shapes, they are simulated and converted into the most similar standard-shaped objects. The specific simulation processing steps include: connecting the farthest pair of vertices among the vertices of the items to form a dividing line, and gradually approaching the shape of the items to the standard shape through multiple segmentations. Step S2 ensures that the 3D data of all items to be loaded are accurate and standardized, solves the problem of difficulty in loading non-standard-shaped items, improves the accuracy and operability of the loading plan, and optimizes the loading efficiency and space utilization.
[0081] Specifically, step S3 includes:
[0082] A plurality of loading strategies are formulated through a tower loading heuristic algorithm. The loading strategies use volume as a first reference indicator, and one of the center of gravity, length, width or height of the items to be loaded as a second and a third reference indicator. Based on the plurality of loading strategies, the three-dimensional models of the plurality of items are sorted respectively to obtain a plurality of loading schemes, which are summarized as an initial loading scheme set.
[0083] Step S3 formulates multiple loading strategies through the tower loading heuristic algorithm, takes volume as the first reference index, and takes the center of gravity, length, width or height of the items to be loaded as the second and third reference indexes, sorts the three-dimensional models of multiple items, generates multiple loading plans and summarizes them into an initial loading plan set. Step S3 ensures the diversity and rationality of the loading strategy, and can optimize the sorting according to the characteristics of different items, thereby improving the scientificity and effectiveness of the loading plan, maximizing the use of space, and improving the overall loading efficiency and stability.
[0084] Specifically, step S4 includes:
[0085] S41, initialize various parameters of the ant algorithm and construct the problem graph structure;
[0086] S42, ant path selection is performed based on the probability transfer rule, a loading plan is obtained and evaluated;
[0087] S43, iteratively optimize the ant colony algorithm and update the pheromone;
[0088] S44, recording the best packing solution found by the ants in each iteration, summarizing the best packing solution corresponding to each ant, and forming a first packing solution set.
[0089] Step S4 forms the first set of loading solutions by initializing the parameters of the ant algorithm, constructing the problem graph structure, selecting paths based on probability transfer rules, iteratively optimizing the pheromone concentration, and recording the best loading solution in each iteration. By using the global search capability of the ant algorithm, it is possible to find efficient loading paths in complex loading problems and avoid local optimal solutions.
[0090] Specifically, step S41 includes:
[0091] A problem graph structure is constructed with the loading space and the items to be loaded, the number of ants in the ant colony, the initial pheromone concentration and the pheromone volatility coefficient are set, and the initial loading scheme set is used as the initial solution. The nodes in the problem graph structure represent the items to be loaded, and each edge in the problem graph structure represents each loading position of the items to be loaded.
[0092] Step S41 constructs the problem graph structure, sets the number of ants, the initial pheromone concentration and the volatility coefficient, and uses the initial loading solution set as the initial solution, thereby ensuring the accuracy of the basic settings of the ant algorithm and the problem modeling. By mapping the loading space and the items to be loaded into the graph structure, the meaning of each node and edge is clarified, and the optimal solution problem of the loading solution is converted into a probabilistic object for the path selection of the ant algorithm.
[0093] Specifically, step S42 includes:
[0094] Each ant selects a path order based on the probability transfer rule to load the items to be loaded in sequence, records the loading path selected by each ant, obtains multiple loading schemes, calculates the fitness of each loading scheme, and evaluates the multiple loading schemes. The calculation formula of the probability transfer rule is:
[0095]
[0096] Among them, P ab is the probability of an ant moving from node a to node b, T ab is the pheromone concentration on edge ab, η ab is the heuristic information of edge ab, α is the adjustment parameter of pheromone concentration, β is the adjustment parameter of heuristic information, T ak is the pheromone concentration on edge ak, η ak is the heuristic information of edge ak, and K is the number of all nodes reachable by node a.
[0097] In step S42, each ant selects a path sequence based on the probability transfer rule to load the items to be loaded, records the loading path and calculates the fitness, and evaluates multiple loading schemes. The application of the probability transfer rule ensures the diversity and rationality of the path selection, and can explore different loading paths and evaluate their fitness, thereby screening out a better loading scheme, improving the optimization effect and adaptability of the loading scheme.
[0098] Specifically, step S43 includes:
[0099] The pheromone is volatilized according to the preset volatilization coefficient, the pheromone concentration on each edge of the problem graph structure is updated, and the pheromone concentration on the loading path is increased:
[0100] T ij '=(1-ρ)T ij ;
[0101] T ij =T ij '+ΔT ij ;
[0102] Among them, T ijis the pheromone concentration on edge ij, ρ is the volatility coefficient, T ij ' is the pheromone concentration on the ij after volatilization, T ij ” is the pheromone concentration after the pheromone concentration is added on edge ij, ΔT ij is the concentration of the newly added pheromone on edge ij;
[0103] The ant colony algorithm is iteratively optimized, and the iterative stopping conditions of the iterative optimization include: the pheromone concentration reaches a preset value, the fitness reaches a preset value, and the number of iterations reaches a preset value.
[0104] Step S43 performs iterative optimization by volatilizing the pheromone according to the preset volatilization coefficient, updating the pheromone concentration on each edge of the problem graph structure, and increasing the pheromone concentration on the loading path. The dynamic update and volatilization mechanism of pheromone enhances the adaptive ability of the algorithm, avoids excessive concentration of pheromone on certain paths, and ensures the extensiveness and balance of the search process. The stop conditions of iterative optimization (pheromone concentration, fitness, number of iterations) further ensure the convergence and efficiency of the algorithm, and finally obtain an optimized loading solution.
[0105] Specifically, step S5 includes:
[0106] S51, selecting multiple loading schemes from the first loading scheme as an initial population, and setting a crossover operation probability and a mutation operation probability;
[0107] S52, constructing a genetic fitness function, performing fitness evaluation on the initial population based on the genetic fitness function, selecting the initial population with a fitness higher than a preset value in the initial population as the second loading scheme set to enter the breeding pool, randomly selecting two loadings in the second loading scheme set as parents, judging whether to crossover according to the crossover operation probability of the two loading schemes, and when judging to crossover, selecting a crossover point, exchanging the parts of the two loading schemes at the crossover point, and generating an initial offspring loading scheme;
[0108] S53, randomly selecting an initial offspring filling scheme and a mutation position according to the mutation operation probability, changing the filling order or filling strategy of the mutation position of the selected initial offspring filling scheme, and generating an offspring filling scheme;
[0109] S54, iteratively train the offspring loading scheme generated each time, repeat steps S51-S53, and obtain a third loading scheme set. The iterative stopping conditions of the iterative training include: the number of iterations reaches a preset maximum value and the fitness converges.
[0110] Step S5 selects solutions with high space utilization through fitness evaluation, adopts crossover operation to produce a more stable loading order, adopts mutation operation to find a more efficient loading strategy, adds an iterative optimization process to adapt to goods of various shapes and sizes, and reduces the number and time of loading required by optimizing the loading plan, thereby reducing costs.
[0111] Step S51 selects multiple filling schemes from the first filling scheme set as the initial population, sets the crossover operation probability and the mutation operation probability, considers the filling problem from multiple angles, avoids falling into the local optimum, reasonably sets the crossover and mutation probabilities, speeds up the process of finding high-quality solutions, and achieves a balance between exploring new solutions and using known good solutions through probability settings.
[0112] Step S52 selects high-quality loading solutions through fitness evaluation, uses crossover operation to generate new solutions with higher space utilization, combines the advantages of different parent solutions to generate child solutions with higher loading efficiency, randomly selects parents for crossover, and generates more diverse child solutions.
[0113] Step S53 explores new possibilities by changing the loading order or strategy, adopts mutation operations to generate loading plans that are more suitable for goods of special shapes or weights, and increases the adaptability of the plan to different loading situations through random mutation, thereby finding a better loading order and improving the overall loading efficiency.
[0114] Step S54 continuously improves the space utilization through multiple iterations. In the repeated optimization process, a more stable loading scheme is screened out. Through multiple iterations, this embodiment can handle more complex loading situations. By setting reasonable stop conditions, the reliability and practicality of the final scheme are ensured. Through the automated iterative process, manual intervention is reduced and efficiency is improved.
[0115] Specifically, step S6 includes:
[0116] The multi-objective optimization fitness function includes a comprehensive fitness function and a multi-objective optimization algorithm. The multi-objective optimization algorithm adopts a Pareto frontier optimization algorithm. The construction process of the comprehensive fitness function includes: selecting an optimization target, the optimization target includes space utilization, loading stability, and loading operation convenience, constructing multiple single fitness functions according to the optimization target, and comprehensively weighting the multiple single fitness functions to obtain a comprehensive fitness function;
[0117] The third filling scheme is used as the initial population of the multi-objective optimization fitness function, and an evolutionary operation is performed on the initial population. The evolutionary operation includes crossover, mutation and selection. The fitness of the new initial population generated by each generation of evolution is evaluated, and the optimal solution of each generation of evolution is retained. The stopping conditions of the evolutionary operation include: the number of evolutions reaches a preset value, the fitness converges, and after the evolutionary operation iteration stops, the optimal filling scheme is obtained.
[0118] Step S6 further optimizes the third loading solution set by constructing a multi-objective optimization fitness function and using the Pareto frontier optimization algorithm. The specific effects include:
[0119] Comprehensive optimization of multiple objectives: Through the comprehensive fitness function, multiple optimization objectives such as space utilization, loading stability and loading operation convenience are weighted and integrated to ensure that the final loading solution performs well in multiple dimensions.
[0120] Efficiently screen the optimal solution: Using the Pareto frontier optimization algorithm, it is possible to effectively screen out loading solutions that perform well on multiple objectives, avoiding over-optimization of a single objective while ignoring other important factors.
[0121] Strong dynamic adaptability: Through evolutionary operations (crossover, mutation and selection), new initial populations are continuously generated, and their fitness is evaluated and optimized to ensure that the scheme can dynamically adapt to different filling requirements and changes.
[0122] Avoiding local optimality: The iterative process of evolutionary operations, through multiple crossovers and mutations, can effectively prevent the algorithm from falling into the local optimal solution and gradually approach the global optimal solution.
[0123] Improve solution quality: The new population generated by each generation of evolution undergoes fitness evaluation and retains the optimal solution, ensuring that the quality of the solution continues to improve after each iteration. The final loading solution performs well in all optimization objectives.
[0124] Stable convergence: Set reasonable evolution stop conditions (such as the number of evolutions reaches the preset value, fitness converges), to ensure that the algorithm stops in time when a satisfactory optimization effect is achieved, avoid over-calculation, and improve efficiency.
[0125] In step S6, under the framework of multi-objective optimization, space utilization, loading stability and operational convenience are comprehensively considered, and the optimal loading solution is finally selected to meet the diverse needs in practical applications.
[0126] Specifically, step S7 includes:
[0127] The updated three-dimensional data of the newly added objects to be loaded are read by a 3D laser scanner, and the updated three-dimensional data of the objects are filtered, denoised and gridded. A three-dimensional model of the newly added objects to be loaded is constructed based on the updated three-dimensional data of the objects. The multi-objective optimization fitness function is updated by the three-dimensional model of the newly added objects to be loaded to obtain a first multi-objective optimization fitness function. The fitness is recalculated based on the first multi-objective optimization fitness function, and the multi-objective optimization is re-performed. According to the recalculated fitness and the optimized structure, a dynamically updated loading plan is generated.
[0128] The 3D laser scanner reads the 3D data of the newly added items to be loaded in real time, and filters, denoises and meshes the data to build a 3D model of the new item. The multi-objective optimization fitness function is updated through the updated 3D model, the fitness is recalculated and multi-objective optimization is performed to generate a dynamically updated loading plan. The specific technical effects include:
[0129] Real-time adaptation to new items: 3D laser scanners are used to obtain 3D data of newly added items in real time, ensuring that the loading solution can adapt to the shape and size changes of new items in a timely manner.
[0130] Improve data accuracy: Filtering, denoising and gridding processing steps improve the accuracy of 3D data, ensuring that the constructed 3D model is accurate and reliable, and providing high-quality basic data for subsequent optimization.
[0131] Dynamically update the optimization function: Based on the three-dimensional model of the new object, update the multi-objective optimization fitness function so that the optimization process can dynamically adapt to the characteristics of the new object, ensuring the effectiveness and accuracy of the optimization results.
[0132] Enhanced flexibility of loading plans: By recalculating fitness and performing multi-objective optimization, the generated loading plan can flexibly respond to the addition of new items and keep the loading process efficient and stable.
[0133] Improve loading efficiency and stability: The dynamically updated loading plan can better utilize space and improve loading efficiency while ensuring loading stability and ease of operation.
[0134] Reduced manual intervention: Automated data processing and optimization processes reduce the need for manual intervention and improve the automation and operating efficiency of the filling system.
[0135] Step S7 ensures that the filling system can dynamically adapt to the addition of new items, update the optimization plan in real time, maintain the efficiency, stability and flexibility of the filling process, and meet the changing needs in practical applications.
[0136] In a specific embodiment, the background is: in a large logistics distribution center, a large number of goods of different sizes and shapes need to be processed every day. These goods need to be efficiently loaded into transport vehicles to maximize space utilization, ensure loading stability and convenient operation. The specific steps of a method for loading items in a limited space in this embodiment are as follows:
[0137] (1) Initial data collection
[0138] Loading space data: Use a 3D laser scanner to scan the loading space of the transport vehicle, and collect data including length, width, height, volume, and internal structure characteristics;
[0139] Data of items to be loaded: 3D scan each item to be loaded, and collect data including length, width, height, volume, weight, shape characteristics, and center of gravity position;
[0140] (2) Data processing
[0141] Loading space data processing: remove noise points through filtering, smooth surface data through denoising, convert point cloud data into a three-dimensional mesh model through gridding, and establish a three-dimensional coordinate system with the lower right rear corner of the car as the origin;
[0142] Data processing of items to be loaded: filtering, denoising, and gridding the data of items to be loaded, simulating non-standard shapes of items, simplifying complex shapes into the most approximate standard geometric shapes, and constructing a three-dimensional model of the item, including information such as size, center of gravity, and rotation angle;
[0143] Data obtained during the optimization process: initial filling solution set, including multiple initial solutions based on different filling strategies; first filling solution set, a solution set optimized by global search of ant colony algorithm; second filling solution set, a high fitness solution set screened by genetic algorithm; third filling solution set, an optimized solution set after crossover mutation of genetic algorithm; optimal filling solution, the final solution after multi-objective optimization;
[0144] (3) Detailed description of the optimal loading solution
[0145] Space utilization: The filling space utilization rate reaches 92%, and the remaining space is evenly distributed, which is convenient for subsequent filling of small items;
[0146] Loading sequence: List in detail the loading sequence, location and direction of each item of cargo;
[0147] Cargo 1 (large box) → position (0,0,0), direction (long side parallel to X axis) Cargo 2 (medium cylinder) → position (200,0,0), direction (axis parallel to Y axis);
[0148] Stability considerations: center of gravity distribution, the overall center of gravity is located below the center of the carriage; support structure, large stable cargo as the foundation, small or irregular cargo on top; anti-dumping measures, wedging and fixing point suggestions between adjacent cargo;
[0149] Convenience of operation: Optimized loading and unloading sequence, following the first-in-last-out principle; reserved aisle, with a 50cm wide operating aisle reserved in the middle of the carriage; partitioned loading, loading by destination or cargo type;
[0150] Special cargo handling: fragile goods are placed on the upper layer, with special cushioning packaging recommendations, and heavy goods are concentrated in the front of the carriage near the cab;
[0151] Dynamic adjustment capability: 10% of the space is reserved for temporary addition of cargo; 3 alternative loading options are provided to cope with temporary changes in cargo;
[0152] Loading efficiency: Estimated loading time is reduced by 30% compared with traditional methods; the number of loaders is reduced from 3 to 2.
[0153] Through the above optimal loading solutions, the logistics distribution center can significantly improve loading efficiency, ensure the safety of goods, and have the ability to flexibly respond to changes, thereby optimizing the overall operational results.
[0154] See also Figure 2 Based on the above method embodiments, the present invention further provides a system for filling objects in a limited space, the system comprising:
[0155] A coordinate system construction module, used for reading the filling space data of the filling space, and constructing the filling space coordinate system based on the filling space data;
[0156] An item model building module, used for acquiring item 3D data of all items to be loaded, and building a plurality of item 3D models based on the item 3D data;
[0157] A heuristic sorting module, used for sorting the plurality of three-dimensional object models by a tower loading heuristic algorithm to generate an initial loading solution set;
[0158] An ant colony algorithm module, configured to perform a global search on the multiple initial loading solutions by using an ant colony algorithm to obtain a first loading solution set;
[0159] A genetic algorithm module, used for constructing a genetic fitness function, screening the first filling scheme set by using the genetic fitness function to obtain a second filling scheme set, performing a crossover operation and a mutation operation on the second filling scheme set to obtain a third filling scheme set;
[0160] A multi-objective optimization fitness module, used to construct a multi-objective optimization fitness function, and to evaluate and screen the third filling scheme set through the multi-objective optimization fitness function to obtain an optimal filling scheme;
[0161] The real-time scheme updating module is used to obtain updated three-dimensional data of newly added items to be loaded in real time, and adjust the optimal loading scheme based on the updated three-dimensional data of the items.
[0162] The above system embodiments and method embodiments correspond one to one, and for a brief description of the system embodiments, please refer to the method embodiments.
[0163] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for loading objects in a limited space, characterized in that: The following steps are involved: S1, reading the filling space data of the filling space, and constructing a filling space coordinate system based on the filling space data; S2, obtaining the three-dimensional data of all the objects to be loaded, and constructing a plurality of three-dimensional models of the objects based on the three-dimensional data of the objects; S3, sorting the plurality of three-dimensional object models by a tower loading heuristic algorithm to generate an initial loading solution set; S4, performing a global search on the multiple initial loading solutions by using an ant colony algorithm to obtain a first loading solution set; S5, constructing a genetic fitness function, screening the first filling scheme set by using the genetic fitness function to obtain a second filling scheme set, performing a crossover operation and a mutation operation on the second filling scheme set to obtain a third filling scheme set; S6, constructing a multi-objective optimization fitness function, and performing effect evaluation and screening on the third filling scheme set by using the multi-objective optimization fitness function to obtain an optimal filling scheme; S7, acquiring updated three-dimensional data of the newly added objects to be loaded in real time, and adjusting the optimal loading solution based on the updated three-dimensional data of the objects.
2. The method for loading objects into a limited space as claimed in claim 1, characterized in that: Step S3 includes: A plurality of loading strategies are formulated through a tower loading heuristic algorithm. The loading strategies use volume as a first reference indicator, and one of the center of gravity, length, width or height of the items to be loaded as a second and a third reference indicator. Based on the plurality of loading strategies, the three-dimensional models of the plurality of items are sorted respectively to obtain a plurality of loading schemes, which are summarized as an initial loading scheme set.
3. The method for loading objects into a limited space as claimed in claim 1, characterized in that: Step S4 includes: S41, initialize various parameters of the ant algorithm and construct the problem graph structure; S42, ant path selection is performed based on the probability transfer rule, a loading plan is obtained and evaluated; S43, iteratively optimize the ant colony algorithm and update the pheromone; S44, recording the best packing solution found by the ants in each iteration, summarizing the best packing solution corresponding to each ant, and forming a first packing solution set.
4. The method for loading objects into a limited space as claimed in claim 1, characterized in that: Step S41 includes: A problem graph structure is constructed with the loading space and the items to be loaded, the number of ants in the ant colony, the initial pheromone concentration and the pheromone volatility coefficient are set, and the initial loading scheme set is used as the initial solution. The nodes in the problem graph structure represent the items to be loaded, and each edge in the problem graph structure represents each loading position of the items to be loaded.
5. The method for loading objects into a limited space as claimed in claim 4, characterized in that: Step S42 includes: Each ant selects a path order based on the probability transfer rule to load the items to be loaded in sequence, records the loading path selected by each ant, obtains multiple loading schemes, calculates the fitness of each loading scheme, and evaluates the multiple loading schemes. The calculation formula of the probability transfer rule is: Among them, P ab is the probability of an ant moving from node a to node b, T ab is the pheromone concentration on edge ab, η ab is the heuristic information of edge ab, α is the adjustment parameter of pheromone concentration, β is the adjustment parameter of heuristic information, T ak is the pheromone concentration on edge ak, η ak is the heuristic information of edge ak, and K is the number of all nodes reachable by node a.
6. The method for loading objects into a limited space as claimed in claim 5, characterized in that: Step S43 includes: The pheromone is volatilized according to the preset volatilization coefficient, the pheromone concentration on each edge of the problem graph structure is updated, and the pheromone concentration on the loading path is increased: T ij '=(1-ρ)T ij ; T ij ”=T ij '+ΔT ij ; Among them, T ij is the pheromone concentration on edge ij, ρ is the volatility coefficient, T ij ' is the pheromone concentration on the ij after volatilization, T ij ” is the pheromone concentration after the pheromone concentration is added on edge ij, ΔT ij is the concentration of the newly added pheromone on edge ij; The ant colony algorithm is iteratively optimized, and the iterative stopping conditions of the iterative optimization include: the pheromone concentration reaches a preset value, the fitness reaches a preset value, and the number of iterations reaches a preset value.
7. The method for loading objects into a limited space as claimed in claim 1, characterized in that: Step S5 includes: S51, selecting multiple loading schemes from the first loading scheme as an initial population, and setting a crossover operation probability and a mutation operation probability; S52, constructing a genetic fitness function, performing fitness evaluation on the initial population based on the genetic fitness function, selecting the initial population with a fitness higher than a preset value in the initial population as the second loading scheme set to enter the breeding pool, randomly selecting two loadings in the second loading scheme set as parents, judging whether to crossover according to the crossover operation probability of the two loading schemes, and when judging to crossover, selecting a crossover point, exchanging the parts of the two loading schemes at the crossover point, and generating an initial offspring loading scheme; S53, randomly selecting an initial offspring filling scheme and a mutation position according to the mutation operation probability, changing the filling order or filling strategy of the mutation position of the selected initial offspring filling scheme, and generating an offspring filling scheme; S54, iteratively train the offspring loading scheme generated each time, repeat steps S51-S53, and obtain a third loading scheme set. The iterative stopping conditions of the iterative training include: the number of iterations reaches a preset maximum value and the fitness converges.
8. The method for loading objects into a limited space as claimed in claim 1, characterized in that: Step S6 includes: The multi-objective optimization fitness function includes a comprehensive fitness function and a multi-objective optimization algorithm. The multi-objective optimization algorithm adopts a Pareto frontier optimization algorithm. The construction process of the comprehensive fitness function includes: selecting an optimization target, the optimization target includes space utilization, loading stability, and loading operation convenience, constructing multiple single fitness functions according to the optimization target, and comprehensively weighting the multiple single fitness functions to obtain a comprehensive fitness function; The third filling scheme is used as the initial population of the multi-objective optimization fitness function, and an evolutionary operation is performed on the initial population. The evolutionary operation includes crossover, mutation and selection. The fitness of the new initial population generated by each generation of evolution is evaluated, and the optimal solution of each generation of evolution is retained. The stopping conditions of the evolutionary operation include: the number of evolutions reaches a preset value, the fitness converges, and after the evolutionary operation iteration stops, the optimal filling scheme is obtained.
9. The method for loading objects into a limited space as claimed in claim 8, characterized in that: Step S7 includes: The updated three-dimensional data of the newly added objects to be loaded are read by a 3D laser scanner, and the updated three-dimensional data of the objects are filtered, denoised and gridded. A three-dimensional model of the newly added objects to be loaded is constructed based on the updated three-dimensional data of the objects. The multi-objective optimization fitness function is updated by the three-dimensional model of the newly added objects to be loaded to obtain a first multi-objective optimization fitness function. The fitness is recalculated based on the first multi-objective optimization fitness function, and the multi-objective optimization is re-performed. According to the recalculated fitness and the optimized structure, a dynamically updated loading plan is generated.
10. A system for loading items in a limited space, characterized in that: The system comprises: A coordinate system construction module, used for reading the filling space data of the filling space, and constructing the filling space coordinate system based on the filling space data; An item model building module, used for acquiring item 3D data of all items to be loaded, and building a plurality of item 3D models based on the item 3D data; A heuristic sorting module, used for sorting the plurality of three-dimensional object models by a tower loading heuristic algorithm to generate an initial loading solution set; An ant colony algorithm module, configured to perform a global search on the multiple initial loading solutions by using an ant colony algorithm to obtain a first loading solution set; A genetic algorithm module, used for constructing a genetic fitness function, screening the first filling scheme set by using the genetic fitness function to obtain a second filling scheme set, performing a crossover operation and a mutation operation on the second filling scheme set to obtain a third filling scheme set; A multi-objective optimization fitness module, used to construct a multi-objective optimization fitness function, and to evaluate and screen the third filling scheme set through the multi-objective optimization fitness function to obtain an optimal filling scheme; The real-time scheme updating module is used to obtain updated three-dimensional data of newly added items to be loaded in real time, and adjust the optimal loading scheme based on the updated three-dimensional data of the items.
Citation Information
Patent Citations
Method and system for optimizing loading layout of three-dimensional containers
CN102214321A
Three-dimensional encasement novel genetic algorithm model under multi-constrain condition
CN104680237A
Method for solving two-dimensional loading constraint logistics vehicle dispatch problem
CN107578199A
Logistics transportation scheme optimization method and system
CN112633555A
Computerized system and method for developing optimized cargo transportation solutions
US20220156693A1
Cited By
Transportation intelligent assignment scheduling method and system
CN119990945A
