Exhibition area automatic optimization layout method

Abstract

The invention discloses an exhibition area automatic optimization layout method, which relates to the technical field of space intelligent planning, and comprises the following steps: dispersing an exhibition area into a grid space, and abstracting an exhibition position into a shape matrix; and constructing a constraint relationship between the booths. And clustering and randomly generating a plurality of layout schemes based on the topic categories. And defining a target function, and calculating a target value by integrating the space utilization rate, the topic correlation degree and the constraint violation penalty. The optimal scheme is marked as the optimal layout, and the rest are conventional layouts; calculating the average layout, and updating the conventional layout by using the difference between the optimal layout and the average layout. Randomly combining all the layout schemes in pairs, and updating the scheme with the low target value by using the scheme with the high target value; the booths are limited in the grids, and overlapping is eliminated; retaining K schemes with the highest target values; and after iteration is finished, outputting an optimal layout. According to the method, global optimization is carried out by utilizing theme clustering and utilizing a group evolution mechanism which learns from excellent individuals, and high-quality layout meeting multiple constraints is realized by combining cooperative movement and physical conflict resolution.

Description

Technical Field

[0001] This invention relates to the field of spatial intelligent planning technology, and in particular to a method for automatically optimizing the layout of an exhibition area. Background Technology

[0002] In the curation and brand promotion of large-scale events in the power industry, offline exhibitions serve as a crucial window for showcasing corporate technological innovation, business achievements, and social responsibility. Exhibition layout, as a core element of the curatorial process, essentially involves the scientific arrangement of several booths with specific shapes, sizes, and content attributes within a limited physical space. An excellent exhibition layout not only maximizes the use of the venue area but also ensures smooth visitor flow and strictly adheres to the logical connections between the exhibits. With the increasing richness of exhibits and the widespread application of multimedia interactive devices, booth shapes have evolved from standard rectangles to complex irregular structures such as L-shapes and U-shapes, making layout design extremely complex.

[0003] Current exhibition area layout design primarily relies on manual experience or general-purpose 2D drawing software. Designers typically employ a trial-and-error approach, initially arranging booths based on experience and then repeatedly adjusting them to eliminate overlaps or meet aisle width requirements. However, the booth layout problem mathematically combines a typical 2D bin packing problem with a constraint satisfaction problem, exhibiting NP-hard characteristics. As the number of exhibitors increases, the solution space explodes exponentially, making it difficult to find the globally optimal solution that maximizes space utilization and satisfies all business logic constraints in a short time using only manual experience. Furthermore, if curatorial needs change, manual re-layout is time-consuming and labor-intensive, severely impacting curatorial efficiency.

[0004] In academia and industry, metaheuristic algorithms such as genetic algorithms and particle swarm optimization have been explored for solving layout optimization problems. These methods search for optimal solutions by simulating natural evolution or group behavior, and possess a certain degree of versatility. Meanwhile, with the rise of deep learning technology, layout generation methods based on generative adversarial networks, graph neural networks, and deep reinforcement learning are also gradually being applied. These methods attempt to directly generate visually appealing layout schemes by learning the feature distribution of large amounts of historical layout data.

[0005] However, when applied to the specific scenario of curating large-scale events in the power industry, the aforementioned existing technologies still have the following significant shortcomings, making it difficult to directly meet actual engineering needs: 1. Deep learning methods suffer from data dependency and hard constraint failure: Deep learning-based methods typically require massive amounts of labeled data for model training. However, power exhibitions are low-frequency, highly customized events, lacking large-scale historical layout datasets, which makes it difficult for deep models to converge or generalize effectively.

[0006] 2. Traditional heuristic algorithms suffer from parameter sensitivity and premature convergence: Traditional genetic algorithms or particle swarm optimization algorithms heavily rely on fine-tuning of hyperparameters. When dealing with strong constraints (such as the requirement that certain nodes must be adjacent), these algorithms often lack effective guidance mechanisms, leading to low search efficiency. More importantly, information about the worst-case solution is often ignored during the optimization process, making the convergence process overly randomized.

[0007] 3. Lack of an effective mechanism for handling semantic-spatial mapping: Most existing automatic layout technologies focus on geometric rectangular arrangement, that is, only concerned with how to cram in more objects. However, in a curatorial scenario, booths are not just geometric shapes, but also nodes that carry business logic. Existing technologies often ignore the semantic relationships of the content behind the booths, resulting in a layout that, while compact, has a logically fragmented visitor experience.

[0008] Therefore, there is an urgent need to provide an automatic layout method that does not require a large amount of training data, has strong parameter adaptability, and can deeply integrate geometric constraints and business logic constraints to solve the above-mentioned technical problems. Summary of the Invention

[0009] To address the problems existing in the prior art, this invention provides an automatic optimization layout method for exhibition areas. By constructing a digital mapping model between exhibition areas and booths, it performs global optimization by combining three perspectives: space utilization, constraint violation penalties, and relevance of exhibition area themes. Furthermore, by combining a collaborative movement mechanism and a physical conflict resolution strategy, it achieves a high-quality automatic layout that satisfies multiple constraints.

[0010] The technical solution of this invention is implemented as follows: An automatic optimization layout method for an exhibition area, wherein the exhibition area is used to arrange multiple booths; each booth is associated with a theme category and display content; the method includes the following steps: T1. Construct a digital mapping model of exhibition area and booth: Based on the actual physical dimensions, transform the exhibition area into a two-dimensional, discrete grid space, and abstract the booth into a shape matrix that can be mapped to the grid space; construct the constraint relationship between exhibition areas based on the content displayed in the booth. The digital mapping model is the foundation for automated layout and calculation. The size of the grid space is obtained by dividing the actual physical dimensions of the exhibition area by the grid precision. The choice of grid precision needs to balance computational efficiency and layout accuracy; an optimal value of 0.5 meters or 1 meter is recommended. The shape matrix is ​​used to accurately describe the geometric features of the booth, especially for non-rectangular irregular booths such as L-shaped, U-shaped, or circular booths.

[0011] Furthermore, based on the content displayed in each booth, a content association constraint graph is constructed between the booths. The content association constraint graph is represented as follows: , This refers to the booth in question. This indicates a constraint that two booths must be adjacent to each other or maintain a specified distance.

[0012] Content association constraint graphs are used to mathematically represent business logic; booths are nodes, and constraints are edges. Furthermore, the weight of each edge is set to the expected Euclidean distance between nodes. Specifically, establishing constraints requires parsing the curatorial document and connecting sections with business flow relationships through edges, such as connecting booths related to green power generation and clean power transmission.

[0013] T2. Initialize the population: Cluster the population according to the theme category, and randomly generate multiple initial layout schemes based on the clustering; wherein, the layout scheme is encoded as a multi-dimensional state vector, which includes the position, rotation angle and theme category of each shape matrix in the grid space; The initial population size is set between 30 and 100 individuals. Too small a size will result in insufficient diversity, while too large a size will increase computational overhead.

[0014] Specifically, a semantic evaluation method based on a large model can be used. By calculating the semantic confidence between topic categories, categories with confidence scores higher than a preset threshold are grouped into the same category, resulting in multiple topic categories.

[0015] Specifically, during the initialization process, a central area is defined within the grid space; several theme categories are randomly selected, and within each theme category, a booth (shape matrix) is randomly selected and placed within the central area.

[0016] Furthermore, the shape matrix is ​​represented using a binary mask, with the value "1" indicating that the booth entity occupies a grid in the grid space, and the value "0" indicating that it does not occupy a grid in the grid space; for example, a 3x3 L-shaped booth matrix can be represented as the first column being all 1s, the third row being all 1s, and the rest being 0s.

[0017] Furthermore, to accelerate algorithm convergence, a cluster-based generation strategy is introduced in the initialization phase. This involves pre-scanning the content association constraint graph and treating booths with constraint relationships as a virtual booth cluster. When generating the initial solution, priority is given to ensuring that the relative positions of booths within the cluster satisfy the constraints, rather than completely independent random placement.

[0018] T3. Define the objective function; the objective function is calculated based on the space utilization rate, constraint violation penalty, and theme relevance of the layout scheme; use the objective function to calculate the target value for each layout scheme; the objective function is the sole criterion for evaluating the quality of the layout, and its core logic is "reward compactness and punish conflict". The higher the space utilization rate, the lower the penalty score, and the higher the similarity of activity content between exhibition areas (i.e., the higher the theme relevance score), the larger the target value, indicating a better layout scheme.

[0019] T4. Layout-oriented optimization: Among all layout schemes, the layout scheme with the highest objective value is recorded as the optimal layout, and the rest are regular layouts; obtain the vector mean of all layout schemes to obtain the average layout; use the difference between the optimal layout and the average layout to update each regular layout. The vector mean is the mean calculated for each dimension of all vectors.

[0020] T5. Layout Interaction Optimization: Randomly combine all layout schemes in pairs; within the same combination, use the layout scheme with the higher target value to update the layout scheme with the lower target value. T6. Layout Correction and Conflict Resolution: For the updated layout scheme, each booth is restricted to a grid space, and booth overlap is eliminated; T7. Update the population: Based on the principle of survival of the fittest, retain the K layout schemes with the highest target values; T8. Iterate from T4 to T7 until the iteration termination condition is met; output the optimal layout.

[0021] The iteration termination condition is that the number of iterations reaches a preset threshold, or the optimal layout has not been updated for several consecutive iterations, or the variance of the target value of the optimal layout is less than a preset threshold for several consecutive iterations.

[0022] Furthermore, the optimal layout is parsed into geometric data, generating a structured data file (such as JSON or DXF format), containing each booth ID, precise center point coordinates, length and width dimensions, rotation angle, theme category, and a diagram of connections between related booths. This file can be directly imported into CAD software.

[0023] This invention discretizes the exhibition area into a grid space and abstracts the booths into shape matrices. Combined with a multi-dimensional state vector encoding layout scheme, it can efficiently handle complex spatial layout constraints. By guiding the update of conventional layouts through optimal layout and mutual learning between layout schemes, it achieves a balance between global search and local development, effectively avoiding getting trapped in local optima. At the same time, it ensures the physical feasibility of layout schemes through overlap elimination and boundary constraints, and retains high-quality schemes by using a target value sorting and filtering mechanism. Thus, while satisfying the constraints between booths, it significantly improves space utilization, reduces layout conflicts, and ultimately quickly obtains high-quality optimal layout schemes.

[0024] As a further optimization of the above scheme, in T2, the position is the anchor point coordinate, which is used to locate the shape matrix in the grid space, such as locating the upper left corner or center of the shape matrix to the anchor point coordinate; During the initialization process, the layout scheme ensures that all shape matrices are within the grid space using a boundary check function; the boundary check function includes range restrictions on the anchor point coordinates.

[0025] Boundary checking functions are used to ensure the feasibility of each initial solution (layout scheme). Further, the top-left corner of the shape matrix is ​​used to locate the anchor point coordinates. For the... A shape, with anchor point coordinates as... Its projection width and projection height in the grid space at the current rotation angle are respectively and The height and width of the grid space are denoted as follows: and Then the boundary check function satisfies: and .

[0026] Specifically, the multidimensional state vector is represented as , Indicates the coordinates of the anchor point. Indicates the rotation angle, This indicates the topic category.

[0027] As a further optimization of the above scheme, the rotation angle includes only 4 orthogonal directions, corresponding to 0° for no rotation, 90° to the left, 90° to the right, and 180° rotation respectively; the default rotation angle of the shape matrix in the grid space is no rotation.

[0028] The four orthogonal directions are used to meet the standardized requirements for booth construction. Furthermore, the rotation angle is discretized, with 0, 1, 2, and 3 corresponding to 0°, 90° (right-hand rotation), 180°, and 270° (left-hand rotation), respectively.

[0029] As a further optimization of the above scheme, in T3, the sum of the areas of all the shape matrices is calculated; the ratio between the sum of the areas and the area of ​​the grid space is the space utilization rate. Obtain the expected distance between two booths that have the aforementioned constraint relationship, and the actual distance between the two booths in the current layout scheme; subtract the expected distance from the actual distance to obtain the distance difference; obtain the sum of all distance differences greater than 0, which is the penalty for violating the constraint; For any given shape matrix, based on a preset radius, find the number of adjacent shape matrices with the same topic category, and record this as the topic relevance. Within the same grid space, calculate the average of the topic relevance of all shape matrices, which is the degree of topic relevance.

[0030] As a further optimization of the above scheme, in T3, the space utilization rate, the constraint violation penalty, and the topic relevance are weighted and calculated to obtain the target value.

[0031] Furthermore, with To represent a layout scheme, its target value is... The calculation is as follows: ;in, This indicates the space utilization rate. Indicates the degree of relevance of the topic. This indicates the penalty for violating the stated constraint. , , These correspond to the weight values ​​of the three.

[0032] As a further optimization of the above scheme, in T4, the update process of the conventional layout is expressed as follows: calculate the difference vector between the optimal layout and the average layout, and the difference vector between the worst layout and the average layout, and update the position of the conventional layout by combining the random step size.

[0033] Furthermore, the update process is represented as follows: .in, , These correspond to the regular layout before and after the update, respectively. , All values ​​are random step sizes, ranging from (0,1), but they are not necessarily the same.

[0034] In the same iteration, the optimal layout, worst layout, and average layout are denoted as follows: , , The layout with the highest target value is the optimal layout, while the layout with the lowest target value is the worst layout.

[0035] The value is the intensity coefficient, which is randomly assigned to 1 or 2. A value of 1 indicates a lower update intensity to maintain the stability of the algorithm, while a value of 2 indicates a higher update intensity to enhance the exploratory nature of the algorithm.

[0036] The solution vector is updated by utilizing the offset directions of the current optimal and worst solutions relative to the average state of all layout schemes. This update strategy significantly improves the efficiency and accuracy of the search process by simultaneously strengthening the approach force to the optimal solution and the repulsion force against the worst solution, guiding the solution to move towards a better region in the solution space.

[0037] In the above formulas, since the calculation involves multi-dimensional vectors, the multi-dimensional parameters in each layout scheme are calculated one-to-one according to their dimensions, and only the position part is calculated in the T4 stage.

[0038] Furthermore, since each dimension parameter in the vector is a discrete integer, the calculation result may result in a decimal. Therefore, only the integer part of the calculation result is taken.

[0039] As a further optimization of the above scheme, the update process of T5 is expressed as follows: calculate the difference vector between the two layout schemes participating in the comparison, and update the layout scheme with the lower target value by combining the random step size.

[0040] Furthermore, the two layout schemes with high and low target values ​​are denoted as follows: , , The representations before and after the update are respectively and The update process is then represented as: .in, The step size is random, and its value ranges from (0,1).

[0041] As a further optimization of the above scheme, in T5, a perturbation operation is performed on the updated layout scheme; the perturbation operation includes one of rotation operation and jump operation, or one of them randomly, or a combination thereof; The rotation operation is as follows: for the updated layout scheme, randomly select a shape matrix and adjust the rotation angle in a random direction, with the adjustment range being a fixed 90°. The jump operation is as follows: for the updated layout scheme, randomly select a shape matrix that does not have the aforementioned constraint relationship, and move the shape matrix to a random empty position in the grid space.

[0042] The purpose of perturbation is to introduce random disturbances, prevent the algorithm from getting stuck in local optima, enhance the diversity of layout schemes, and thus expand the global search capability.

[0043] For irregularly shaped booths such as L-shaped ones, rotation operations can effectively resolve spatial deadlocks and are an important means of finding the global optimum. Deadlock is a locally stalled layout state, characterized by the inability of iterative updates based on local adjustments to continue reducing conflict metrics or optimizing objectives when conflicts still exist or objectives are not met, leading to stagnation or cyclical oscillations in layout updates.

[0044] Furthermore, the disturbance operation is triggered randomly or under specified conditions.

[0045] The specified conditions are as follows: if the target value of a layout scheme does not increase within a specified number of consecutive iterations, then a perturbation is applied to the current layout scheme; or if a layout scheme has overlapping slots within a specified number of consecutive iterations, then a perturbation is applied to the current layout scheme. Both specified conditions can also be implemented simultaneously.

[0046] Regarding the identification of overlap, if two shape matrices are both assigned a value of 1 in the same grid space, the grid is an overlapping grid.

[0047] As a further optimization of the above scheme, a collaborative movement mechanism is also executed when updating in T4 and T5: when the layout scheme is updated, if there is a position update of booth A, and there is a constraint relationship between booth B and booth A, then booth B only follows booth A to move its position.

[0048] The cooperative movement mechanism is a key improvement of this invention, designed to protect satisfied constraints from being disrupted by random updates. Specifically, it calculates the position vector of booth B relative to booth A. When booth A is updated, the location update for booth B does not use the previous formula; instead, the coordinates of booth B are directly added to the location vector. The algorithm adds a small random perturbation to obtain the new position of booth B. This mechanism fixes connected booths together for overall movement, maintaining the local topology during the global search, thus significantly improving the convergence efficiency of the algorithm under strong constraints. Specifically, the shape matrix corresponding to the booth is used as the calculation parameter.

[0049] As a further optimization of the above scheme, T6 includes performing boundary verification on each of the layout schemes to identify booths that exceed the grid space boundary; and forcibly constraining the positions of booths exceeding the boundary to within the grid space; specifically, for booths that exceed the boundary, the system will not delete them, but will perform a "forced repositioning" operation, directly modifying the coordinate values ​​of the booth that exceed the boundary to the boundary values, so that it is located inside the boundary of the grid space. The implementation process refers to the truncation method: ;at this time, The x-coordinate represents the anchor point coordinates. Indicates the width of the grid space. This indicates the width of the current booth.

[0050] The system also resolves conflicts for each layout scheme. If two booths overlap in the grid space, a repulsive force vector is calculated for the overlapping area. Based on this repulsive force vector, the two booths are moved with minimum distance until the overlapping area is eliminated. Specifically, conflict resolution is based on the separating axis theorem. The repulsive force vector is the orientation of the line connecting the centers of the two booths. The two booths are moved in opposite directions along the repulsive force vector, with the moving distance proportional to the overlap depth.

[0051] If one of the booths is marked as a fixed booth, such as a themed booth at the entrance, then only the other booth will be moved. If neither booth is a fixed booth, then each booth will be moved in half.

[0052] The conflict resolution process is executed in an iterative loop until the total overlap area of ​​the system is reduced to zero or the maximum number of loops is reached, thereby generating a conflict-free, constructible layout.

[0053] Alternatively, pixel-level bitmask detection can be used for collision resolution.

[0054] Compared with the prior art, the present invention achieves the following beneficial effects: This invention provides an automatic optimization layout method for exhibition areas. By discretizing the exhibition area into a grid space and abstracting the booths into shape matrices, combined with a multi-dimensional state vector encoding layout scheme, it can efficiently handle complex spatial layout constraints. By guiding the update of conventional layouts through optimal layout and mutual learning between layout schemes, it achieves a balance between global search and local development, effectively avoiding getting trapped in local optima. At the same time, it ensures the physical feasibility of the layout scheme through overlap elimination and boundary constraints, and retains high-quality schemes by using a target value sorting and filtering mechanism. Thus, while satisfying the constraints between booths, it significantly improves space utilization, reduces layout conflicts, and ultimately quickly obtains a high-quality optimal layout scheme.

[0055] Clustering based on thematic categories and evaluating target values ​​according to the thematic relevance of adjacent booths solves the problem of traditional layout methods that only focus on physical space utilization while neglecting the relevance of booth content. By clustering booths with similar themes, the overall thematic coherence of the exhibition area can be effectively improved, enhancing the visitor's viewing experience and content comprehension efficiency. At the same time, prioritizing the adjacent placement of booths with related content effectively promotes information flow and collaborative display effects within the exhibition area, achieving a dual-objective synergistic optimization of space utilization and content organization. This results in a layout scheme that satisfies physical constraints and possesses good semantic rationality. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the overall process of an automatic optimization layout method for exhibition areas provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the effect of mapping the shape matrix of the L-shaped booth in the grid space according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the layout correction and conflict resolution process provided in the embodiments of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0058] like Figures 1 to 3 As shown, this embodiment provides a method for automatically optimizing the layout of an exhibition area, where multiple booths are arranged; each booth is associated with a theme category and display content; the method includes the following steps: T1. Construct a digital mapping model of exhibition area and booths: The exhibition area is used to arrange multiple booths; based on the actual physical dimensions, the exhibition area is transformed into a two-dimensional, discrete grid space, and the booths are abstracted into shape matrices that can be mapped onto the grid space. The grid accuracy is set to 0.5 meters.

[0059] Shape matrices are used to accurately describe the geometric features of booths, especially for non-rectangular, irregularly shaped booths such as L-shaped, U-shaped, or circular booths. In this embodiment, a binary mask matrix is ​​used for high-fidelity modeling of each booth. A value of "1" indicates that the booth entity occupies a portion of the grid space, while a value of "0" indicates no occupation of the grid space. Therefore, for regular rectangular booths, the mask matrix is ​​all 1s; for irregularly shaped booths commonly seen in power exhibitions, they are embedded within a minimum bounding rectangle, with the occupied area marked as 1 and the empty area marked as 0. For example, a 3x3 L-shaped booth matrix can be represented as a first column all 1s, a third row all 1s, and the rest as 0s. In a 2x2 square booth matrix, all positions are represented by 1s.

[0060] This pixel-based matrix representation allows the algorithm to identify the recessed areas of irregularly shaped booths during subsequent overlap calculations, enabling other small exhibits to be embedded within them, thereby greatly improving space utilization.

[0061] Based on the content displayed in each booth, construct the constraint relationships between the booth areas. Specifically, based on the content displayed in each booth, construct a content relationship constraint diagram between the booth areas. The content relationship constraint diagram is represented as follows: , Indicates the booth. This represents a constraint relationship where two booths must be adjacent or maintain a specified distance. The content association constraint graph is used to mathematically represent business logic; booths are nodes, and constraints are edges. Furthermore, the weight of each edge is set to the expected Euclidean distance between nodes. During implementation, establishing these constraints requires parsing the curatorial document and connecting sections with business flow relationships through edges, such as connecting booths related to green power generation and clean power transmission. This ensures that the algorithm not only focuses on geometric non-overlap but also on the continuity of business flow.

[0062] The digital mapping model is the foundation for automatic layout and computation. The size of the grid space is obtained by dividing the actual physical dimensions of the exhibition area by the grid precision. One of the innovative aspects of this invention is the transformation of the complex two-dimensional bin packing problem into a multi-objective optimization problem on a discrete grid.

[0063] T2. Initialize the population: Cluster the population according to the theme category and randomly generate 50 initial layout schemes based on the clustering.

[0064] Specifically, by calculating the semantic confidence between topic categories, categories with confidence scores higher than a preset threshold are grouped into the same category, resulting in multiple topic categories. During initialization, a central region is defined within the grid space; several topic categories are randomly selected, and within each topic category, a booth (shape matrix) is randomly selected and placed within the central region. Here, "several" refers to an arbitrary number of booths.

[0065] The layout scheme is encoded as a multi-dimensional state vector, which includes the position, rotation angle, and theme category of each shape matrix in the grid space.

[0066] In this embodiment, the position refers to anchor point coordinates, used to locate the shape matrix within the grid space. Specifically, in this embodiment, the top-left corner of the shape matrix is ​​located at anchor point coordinates; In this embodiment, the rotation angle includes only four orthogonal directions, corresponding to 0° (no rotation), 90° left rotation, 90° right rotation, and 180° rotation. The default rotation angle for the shape matrix within the grid space is no rotation. The limitation of four orthogonal directions is to meet the standardized requirements of booth construction. To further simplify coding and facilitate calculation, the rotation angle values ​​are discretized, with 0, 1, 2, and 3 corresponding to 0°, 90° (90° right rotation), 180°, and 270° (90° left rotation), respectively.

[0067] Therefore, the multidimensional state vector in this embodiment is represented as follows: , Indicates the anchor point coordinates. Indicates the rotation angle. Indicates the topic category.

[0068] During the initialization process, the layout scheme ensures that all shape matrices are within the grid space through a boundary check function to avoid generating invalid solutions; the boundary check function includes range restrictions on anchor point coordinates.

[0069] Boundary checking functions are used to ensure the feasibility of each initial solution (layout scheme). Specifically, for the... A shape, with anchor point coordinates as... Its projection width and projection height in the grid space at the current rotation angle are respectively and The height and width of the grid space are denoted as follows: and Then the boundary check function satisfies: and .

[0070] In addition, to accelerate algorithm convergence, a cluster-based generation strategy is introduced in the initialization phase. This involves pre-scanning the content association constraint graph and treating booths with constraint relationships as a virtual booth cluster. When generating the initial solution, priority is given to ensuring that the relative positions of booths within the cluster satisfy the constraints, rather than completely independent random scattering.

[0071] T3. Define the objective function; the objective function is calculated based on the space utilization rate of the layout scheme, the penalty for constraint violation, and the degree of theme relevance.

[0072] Specifically, calculate the sum of the areas of all shape matrices; the ratio between the sum of the areas and the area of ​​the grid space is the space utilization rate.

[0073] Obtain the expected distance between two booths with a constraint relationship, and the actual distance between the two booths in the current layout scheme; subtract the expected distance from the actual distance to get the distance difference; obtain the sum of all distance differences greater than 0, which is the constraint violation penalty.

[0074] For any shape matrix, based on a preset radius, find the number of adjacent shape matrices with the same topic category, and record it as the topic relevance; within the same grid space, calculate the average of the topic relevance of all shape matrices, which is the topic relevance degree.

[0075] In this embodiment, with To represent a layout scheme, its target value is... The calculation is as follows: ;in, Indicates space utilization rate, Indicates the degree of relevance to the topic. This indicates the penalty for violating the restrictions. , , These correspond to the weight values ​​of the three.

[0076] The objective function is the sole criterion for evaluating the quality of a layout, and its core logic is "rewarding compactness and penalizing conflict." The higher the space utilization rate, the lower the penalty score, and the higher the similarity of activities within the exhibition areas (i.e., the higher the score for the degree of thematic relevance), the larger the objective value, indicating a better layout scheme.

[0077] Calculate the target value for each layout scheme using the objective function.

[0078] T4. Layout-Oriented Optimization: Among all layout schemes, the layout scheme with the highest objective value is denoted as the optimal layout. The layout scheme with the lowest objective value is designated as the worst layout. The rest are standard layouts; the average layout is obtained by taking the vector mean of all layout schemes. The vector mean is calculated by taking the mean of each dimension of all vectors.

[0079] Calculate the difference vector between the optimal and average layouts, and the difference vector between the worst and average layouts. Then, using a random step size, update the position of the regular layout, as follows: .

[0080] in, , These correspond to the regular layout before and after the update, respectively. , All values ​​are random step sizes, ranging from (0,1), but they are not necessarily the same.

[0081] The value is the intensity coefficient, which is randomly assigned to 1 or 2. A value of 1 indicates a lower update intensity to maintain the stability of the algorithm, while a value of 2 indicates a higher update intensity to enhance the exploratory nature of the algorithm.

[0082] In the above formula, since the calculation involves multidimensional vectors, the multidimensional parameters in each layout scheme are calculated one-to-one according to their dimensions, and only the position part is involved in the calculation, excluding the rotation angle part.

[0083] Furthermore, since each dimension parameter in the vector is a discrete integer, the calculation result may result in a decimal. Therefore, only the integer part of the calculation result is taken; the same applies to subsequent calculations.

[0084] In stage T4, the offset directions of the current best and worst solutions relative to the population average state are used to guide the update of the solution vector. By simultaneously strengthening the approach force to the optimal solution and the repulsion force against the worst solution, the efficiency and accuracy of the search process are significantly improved, guiding the solution to move towards a better region in the solution space.

[0085] T5. Layout interaction optimization includes: randomly pairing all layout schemes; within the same pair, using the layout scheme with the higher target value to update the layout scheme with the lower target value. Specifically, the difference vector between the two layout schemes being compared is calculated, and combined with a random step size, the layout scheme with the lower target value is updated.

[0086] In this embodiment, the two layout schemes with high and low target values ​​are denoted as follows: , , The representations before and after the update are respectively and The update process is then represented as: .in, The step size is random, and its value ranges from (0,1).

[0087] Specifically, in T5, a perturbation operation is performed on the updated layout scheme, and this perturbation operation is triggered by specified conditions. There are two types of specified conditions.

[0088] If the target value of a layout scheme does not increase within a specified number of consecutive iterations, then a perturbation is applied to the current layout scheme.

[0089] If a layout scheme has overlapping spaces in a specified number of consecutive iterations, then the current layout scheme is perturbed. Regarding the identification of overlap, if two shape matrices are both assigned a value of 1 to the same grid in the grid space, that grid is considered an overlapping grid.

[0090] The determination of the implementation of the specified conditions is achieved by maintaining one or more stall counters.

[0091] In this embodiment, the perturbation operation is a random one of rotation and jump operations.

[0092] Specifically, the rotation operation is as follows: for the updated layout scheme, randomly select a shape matrix and adjust its rotation angle in a random direction, with a fixed range of 90°; the jump operation is as follows: for the updated layout scheme, randomly select a shape matrix without constraints and move the shape matrix to a random empty position in the grid space.

[0093] The purpose of perturbation is to introduce random disturbances, prevent the algorithm from getting stuck in local optima, enhance the diversity of layout schemes, and thus expand the global search capability.

[0094] For irregularly shaped booths such as L-shaped ones, rotation operations can effectively resolve spatial deadlocks and are an important means of finding the global optimum. Deadlock is a locally stalled layout state, characterized by the inability of iterative updates based on local adjustments to continue reducing conflict metrics or optimizing objectives when conflicts still exist or objectives are not met, leading to stagnation or cyclical oscillations in layout updates.

[0095] The T5 stage update calculations include not only the position component but also the rotation angle component.

[0096] In particular, the update process for T4 and T5 aims to bring conventional layout schemes closer to better layout schemes (i.e., schemes with higher target values), but simple numerical approximation may disrupt the good topology already formed between booths.

[0097] Therefore, in this embodiment, the update process also executes a cooperative movement mechanism: when the layout scheme is updated, if the position of booth A is updated, and there is a constraint relationship between booth B and booth A, then booth B only follows booth A in moving position. That is, booth B does not additionally execute the predetermined scheme update process.

[0098] The cooperative movement mechanism is a key improvement of this invention, designed to protect satisfied constraints from being disrupted by random updates. Specifically, it calculates the position vector of booth B relative to booth A. When booth A is updated, the location update for booth B does not use the previous formula; instead, the coordinates of booth B are directly added to the location vector. The algorithm adds a small random perturbation to obtain the new position of booth B. This mechanism virtually binds connected booths into a rigid body for overall movement, maintaining the local topology during the global search. This effectively prevents random update operations from disrupting the business chain, thus significantly improving the convergence efficiency of the algorithm under strong constraints. Specifically, the shape matrix corresponding to the booth is used as the calculation parameter.

[0099] T6. Layout Correction and Conflict Resolution: For the updated layout scheme, each booth is restricted to a grid space and booth overlap is eliminated.

[0100] Specifically, layout correction includes boundary verification for each layout scheme to identify booths that exceed the grid space boundary; and forcibly constraining the positions of these out-of-bounds booths to within the grid space, i.e., out-of-bounds handling. During implementation, for out-of-bounds booths, the system does not delete them, but instead performs a "forced repositioning" operation, directly modifying the coordinate values ​​of the out-of-bounds booth to the boundary values, placing it inside the grid space boundary. The implementation process follows the truncation method. ;at this time, The x-coordinate represents the anchor point coordinates. Indicates the width of the grid space. This indicates the width of the current booth.

[0101] The system also resolves conflicts for each layout scheme. If two booths overlap in the grid space, the repulsive force vector of the overlapping area is calculated. Based on this vector, the two booths are moved with minimum distance until the overlap is eliminated. Specifically, conflict resolution is based on the separating axis theorem. The repulsive force vector is the orientation of the line connecting the centers of the two booths. Each booth is moved in the opposite direction along the repulsive force vector, with the moving distance proportional to the overlap depth.

[0102] If one of the booths is marked as a fixed booth, such as a themed booth at the entrance, then only the other booth will be moved. If neither booth is a fixed booth, then each booth will be moved in half.

[0103] The conflict resolution process is executed in an iterative loop until the total overlap area of ​​the system is reduced to zero or the maximum number of loops is reached, thereby generating a conflict-free, constructible layout.

[0104] In other embodiments, pixel-level bitmask detection can also be used for conflict resolution.

[0105] T7. Update the population: Retain the K layout schemes with the highest target values; T8. Iterate from T4 to T7 until the iteration termination condition is met; output the optimal layout.

[0106] The iteration termination condition is that the number of iterations reaches a preset threshold, or the optimal solution has not been updated for several consecutive iterations, or the variance of the target value of the optimal layout is less than a preset threshold for several consecutive iterations.

[0107] Furthermore, the optimal layout is parsed into geometric data, generating a structured data file (such as JSON or DXF format) containing each booth ID, precise center point coordinates, length and width dimensions, rotation angle, theme category, and a diagram showing the connections between related booths. This file can be directly imported into CAD software for on-site construction teams to perform precise layout.

[0108] This invention discretizes the exhibition area into a grid space and abstracts the booths into shape matrices. Combined with a multi-dimensional state vector encoding layout scheme, it can efficiently handle complex spatial layout constraints. By guiding the update of conventional layouts through optimal layout and mutual learning between layout schemes, it achieves a balance between global search and local development, effectively avoiding getting trapped in local optima. At the same time, it ensures the physical feasibility of layout schemes through overlap elimination and boundary constraints, and retains high-quality schemes by using a target value sorting and filtering mechanism. Thus, while satisfying the constraints between booths, it significantly improves space utilization, reduces layout conflicts, and ultimately quickly obtains high-quality optimal layout schemes.

[0109] Clustering based on thematic categories and evaluating target values ​​according to the thematic relevance of adjacent booths solves the problem of traditional layout methods that only focus on physical space utilization while neglecting the relevance of booth content. By clustering booths with similar themes, the overall thematic coherence of the exhibition area can be effectively improved, enhancing the visitor's viewing experience and content comprehension efficiency. At the same time, prioritizing the adjacent placement of booths with related content effectively promotes information flow and collaborative display effects within the exhibition area, achieving a dual-objective synergistic optimization of space utilization and content organization. This results in a layout scheme that satisfies physical constraints and possesses good semantic rationality.

[0110] Based on the disclosure and teachings of the foregoing specification, those skilled in the art can make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and some modifications and changes to the present invention should also fall within the protection scope of the claims of the present invention. Furthermore, although some specific terms are used in this specification, these terms are only for convenience of explanation and do not constitute any limitation on the present invention.

Claims

1. A method for automatically optimizing the layout of an exhibition area, wherein the exhibition area is used to arrange multiple booths; each booth is associated with a theme category and display content; characterized in that, Including the following steps: T1. Based on the actual physical dimensions, the exhibition area is transformed into a two-dimensional, discrete grid space, and the booths are abstracted into shape matrices that can be mapped onto the grid space; based on the content to be displayed, the constraint relationships between the exhibition areas are constructed. T2. Cluster the elements according to the theme categories, and randomly generate multiple initial layout schemes based on the clustering; wherein, the layout scheme is encoded as a multi-dimensional state vector, which includes the position, rotation angle and theme category of each shape matrix; T3. Define the objective function; the objective function is calculated based on the space utilization rate of the layout scheme, the penalty for constraint violation, and the degree of theme relevance; use the objective function to calculate the target value for each layout scheme; T4. Among all layout schemes, the layout scheme with the highest objective value is recorded as the optimal layout, and the rest are regular layouts; obtain the vector mean of all layout schemes to obtain the average layout; use the difference between the optimal layout and the average layout to update each regular layout. T5. Randomly combine all layout schemes in pairs; within the same combination, update the layout scheme with the lower target value using the layout scheme with the higher target value. T6. For the updated layout scheme, each booth will be confined to a grid space, and booth overlap will be eliminated; T7. Retain the K layout schemes with the highest target values; T8. Iterate from T4 to T7 until the iteration termination condition is met; output the optimal layout.

2. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, In T2, the position is the anchor point coordinate, used to locate the shape matrix within the grid space; During the initialization process, the layout scheme ensures that all shape matrices are within the grid space using a boundary check function; the boundary check function includes range restrictions on the anchor point coordinates.

3. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, The rotation angle includes only four orthogonal directions, corresponding to no rotation, 90° left rotation, 90° right rotation, and 180° rotation respectively; the default rotation angle of the shape matrix in the grid space is no rotation.

4. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, In step T3, the sum of the areas of all the shape matrices is calculated; the ratio between the sum of the areas and the area of ​​the grid space is the space utilization rate. Obtain the expected distance between two booths that have the aforementioned constraint relationship, and the actual distance between the two booths in the current layout scheme; subtract the expected distance from the actual distance to obtain the distance difference; obtain the sum of all distance differences greater than 0, which is the penalty for violating the constraint; For any given shape matrix, based on a preset radius, find the number of adjacent shape matrices with the same topic category, and record this as the topic relevance. Within the same grid space, calculate the average of the topic relevance of all shape matrices, which is the degree of topic relevance.

5. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, In step T3, the space utilization rate, the constraint violation penalty, and the topic relevance are weighted and calculated to obtain the target value.

6. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, In T4, the update process of the conventional layout is as follows: calculate the difference vector between the optimal layout and the average layout, and the difference vector between the worst layout and the average layout, and update the position of the conventional layout by combining the random step size.

7. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, The update process of T5 is as follows: calculate the difference vector between the two layout schemes being compared, and update the layout scheme with the lower target value by combining the random step size.

8. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, In T5, a perturbation operation is performed on the updated layout scheme; the perturbation operation includes one of rotation operation and jump operation, or one of them randomly, or a combination thereof. The rotation operation is as follows: for the updated layout scheme, randomly select a shape matrix and adjust the rotation angle in a random direction, with the adjustment range being a fixed 90°. The jump operation is as follows: for the updated layout scheme, randomly select a shape matrix that does not have the aforementioned constraint relationship, and move the shape matrix to a random empty position in the grid space.

9. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, During updates in T4 and T5, a collaborative movement mechanism is also executed: when the layout scheme is updated, if the position of booth A is updated, and there is a constraint relationship between booth B and booth A, then booth B only follows booth A to move its position.

10. The automatic optimization layout method for exhibition areas according to claim 1, characterized in that, The T6 includes performing boundary verification on each of the layout schemes to identify booths that exceed the boundaries of the grid space; and forcibly constraining the positions of booths that exceed the boundaries to within the grid space. The system also resolves conflicts for each of the layout schemes; if two booths form an overlapping area in the grid space, the repulsive force vector of the overlapping area is calculated, and the two booths are moved by the minimum distance according to the repulsive force vector until the overlapping area is eliminated.

Images