Digital home design system and method based on artificial intelligence

Abstract

The invention discloses a digital home design system and method based on artificial intelligence, and belongs to the field of intelligent home design, and the system comprises a demand collection module, a space planning module, a style recommendation module, an equipment configuration module, a material selection module, an environment optimization module, a home integration module, a cost prediction module and a display interaction module. According to the invention, the privacy data of the user can be effectively protected, the intelligence and adaptive ability of the home equipment are enhanced, the precision of real-time decision and intelligent control is improved, the bandwidth consumption can be obviously saved, and the network transmission burden is reduced; according to the technical scheme, real-time cost optimization and budget monitoring can be achieved, the situation that the budget is hyper-branched due to design changes can be effectively avoided, the number of times and cost of later modification are reduced, the design changes are flexibly coped with, different parts in the scheme are adjusted in time to adapt to new requirements, and user experience and satisfaction are improved.

Description

Technical Field

[0001] The present invention relates to the field of smart home design, and in particular to a digital home design system and method based on artificial intelligence. Background Art

[0002] With the continuous progress of modern technology, home design is no longer a single decoration task, but a complex process involving multiple disciplines and various requirements. In recent years, the concept of smart home has been rising day by day, and the wide application of smart home devices has brought unprecedented improvement to the comfort, security and convenience of family life. With the progress of technology and the continuous improvement of people's demand for the quality of living environment, digital home design has become an important part of the development of smart home systems. Digital home design is not only traditional decoration design, but also a new way to deeply optimize and intelligently transform home space through technical means such as artificial intelligence, Internet of Things (IoT), and big data analysis. However, how to effectively integrate these smart devices into home design, how to achieve optimal function within a limited budget, and how to carry out personalized customization according to the needs and living habits of each family member are still major challenges in the field of home design. At the same time, traditional home design methods often rely on the experience and intuition of designers, lack flexibility and the ability of personalized customization, and have low efficiency in the design process. With the popularization of smart devices and the personalization of user needs, the traditional home design mode is difficult to meet the improvement of modern people's multi-dimensional needs such as comfort, functionality, and artistry.

[0003] After retrieval, Chinese Patent No. CN118153142B discloses a home design method and system based on virtual reality. This invention has the advantages of simple immersive operation, high design efficiency, strong simulation and realistic sense, and good user experience. However, it cannot protect the privacy data of users, has poor intelligence and adaptability of home devices, reduces the accuracy of real-time decision-making and intelligent control; in addition, the existing digital home design systems and methods cannot perform real-time cost optimization and budget monitoring, and are prone to budget overruns caused by design changes, increasing the number and cost of later modifications. Therefore, we propose a digital home design system and method based on artificial intelligence. Summary of the Invention

[0004] The purpose of the present invention is to solve the defects existing in the prior art, and to propose a digital home design system and method based on artificial intelligence.

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

[0006] An artificial intelligence-based digital home design system, including a requirement collection module, a space planning module, a style recommendation module, a device configuration module, a material selection module, an environmental optimization module, a home integration module, a cost prediction module, and a display interaction module;

[0007] The requirement collection module is used to collect the specific requirements of users for home design;

[0008] The space planning module is used to divide the room functions and design the space layout based on the user requirement data;

[0009] The style recommendation module is used to generate personalized style recommendations according to user preferences and space characteristics;

[0010] The device configuration module is used to automatically select furniture and home appliances according to the space planning results and style recommendations through the Internet of Things database;

[0011] The material selection module is used to recommend material options that meet the budget and have the best cost performance for users;

[0012] The environmental optimization module is used to simulate the propagation path of light and the trajectory of air flow, and design the best lighting layout and ventilation plan;

[0013] The home integration module is used to automatically configure the intelligent system in the home and provide a personalized smart home control solution;

[0014] The cost prediction module is used to predict the budget for user requirements and dynamically adjust the design plan according to the budget set by the user;

[0015] The display interaction module is used to provide 3D models and renderings of the user's home design plan.

[0016] As a further solution of the present invention, the space planning module collects the room dimensions, functional requirements, and user's personalized requirements from the information input by the user, the existing home design case library, and each data source of environmental data, preprocesses each group of collected raw data, then converts each group of processed data into a coded form, and inputs each group of converted data into the pre-trained CNN model. The convolutional layer of the CNN extracts local features on each group of data by sliding the convolutional kernel to generate a new feature map. Then, the pooling layer performs max-pooling processing on the feature map to reduce the dimension of the convolutional result. Through multiple layers of convolution and pooling processing, the CNN model continuously divides the functional areas according to the space requirements input by the user and optimizes the space layout. The features extracted through multiple layers of convolution and pooling are passed to the fully connected layer. Then, the fully connected layer analyzes the relationships between different rooms and the space layout requirements and generates the optimal space layout. After the CNN model completes the space layout optimization, it outputs the final space layout plan and design drawings through the output layer. At the same time, each cell in the drawings represents the functional area of each room and is labeled.

[0017] As a further solution of the present invention, the specific steps for the environmental optimization module to design the best lighting layout and ventilation plan are as follows:

[0018] S1.1: The specific steps for the environmental optimization module to design the best lighting layout and ventilation plan are as follows:

[0019] S1.1: The environmental optimization module maximally utilizes the natural light of the positions of the windows, light sources, and the room layout as the lighting optimization goal, and takes the smooth air circulation in the space, avoiding dead corners, and improving the air exchange efficiency as the ventilation optimization goal. Based on the lighting optimization goal and the ventilation optimization goal, it designs the lighting optimization objective function and the ventilation optimization objective function, and at the same time adds corresponding constraint conditions to the two groups of objective functions;

[0020] S1.2: Taking the latest space layout plan generated by the space planning module, the style recommendation module, the device configuration module, and the material selection module as the root node, it calculates the lighting and ventilation effects of the current space layout plan using the lighting optimization objective function and the ventilation optimization objective function. By recursively searching for various different configurations in the space layout and taking the results of the recursive search as the child nodes, a search tree is constructed. At the same time, it calculates the lighting and ventilation effects of each child node and initializes the return values and access times of the root node and the child nodes;

[0021] S1.3: Calculate the UCT values of each node in the search tree using the Upper Confidence Bound algorithm, and layer by layer, select the child node with the highest UCT value. When the selection phase reaches a leaf node that has not been fully expanded, enter the expansion phase. Adjust the layout based on the known information, assign a new evaluation status to each new layout node, and at the same time, add this spatial layout as a new child node to the search tree;

[0022] S1.4: After the expansion is completed, establish a radiative transfer model and a fluid dynamics model and randomly adjust the final effects of simulated lighting and ventilation, and evaluate the comprehensive effects of lighting and ventilation under the current layout according to the model results. Propagate the evaluation results obtained in the simulation phase back to each node on the same path in the search tree, and update the return value and the number of visits of each node, where L a represents the lighting intensity of the a-th room, k represents the number of all light sources, S b represents the brightness of light source b, A b represents the effective irradiation area of light source b, d b represents the distance from light source b to the room, θ b represents the angle between light source b and the room surface, V a represents the ventilation flow rate of the a-th room, C represents the efficiency coefficient of the ventilation equipment, B represents the cross-sectional area of the air duct or window, ΔP represents the pressure difference on both sides of the air duct or window, and ρ represents the density of air;

[0023] S1.5: Repeat the processes of selection, expansion, simulation, and backtracking until the changes in the return values of each node in the search tree converge to a preset range. Traverse each node in the search tree and select the node with the largest return value as the optimal solution, and output the best lighting layout and ventilation plan.

[0024] As a further solution of the present invention, the specific calculation formula of the lighting optimization objective function described in S1.1 is as follows;

[0025]

[0026] In the formula, L represents the sum of lighting intensities; L i represents the lighting intensity of the i-th room; n represents the number of rooms in the space; L min represents the minimum lighting requirement for each room;

[0027] The specific calculation formula of the ventilation optimization objective function described in S1.1 is as follows:

[0028]

[0029] In the formula, V represents the total ventilation efficiency; V irepresents the ventilation flow rate of the i-th room; m represents the number of rooms in the space; V min represents the minimum ventilation requirement for each room;

[0030] The specific evaluation formula for the comprehensive effect of lighting and ventilation described in S1.4 is as follows:

[0031]

[0032] In the formula, Q(v) represents the average effect value of this layout; N(v) represents the number of times node v is visited; R(v j ) represents the lighting and ventilation effect score of the j-th simulation result.

[0033] As a further solution of the present invention, the specific steps for the home integration module to automatically configure the intelligent system in the home are as follows:

[0034] S2.1: The furniture integration module receives the information of each home device selected by the device configuration module, and identifies and models the intelligent furniture devices. When the devices are connected, each device reports its current status to the system according to its own functions and characteristics, and initializes a local model. At the same time, the furniture integration module initializes a set of global models based on the multiple functions of each smart home.

[0035] S2.2: After the connection of each group of smart home devices is completed, verify the number of connected devices. If there are missing connections, reconnect the missing devices manually or automatically. Then, the furniture integration module distributes the global model to each smart home device. Each smart home device receives the global model and replaces the architecture and parameters of the original local model according to the architecture and parameters of the global model.

[0036] S2.3: Each smart home device collects data according to the environment and user needs, and trains the local model based on the local data collected by itself. With the goal of accurately predicting and adjusting the device parameters for optimization, a corresponding objective function is constructed Use the backpropagation algorithm to calculate the gradient and update the model parameters. After each device completes multiple rounds of local training, it transmits the final model parameters updated by the local training to the furniture integration module, where W q (ω q ) represents the loss function of device q in local training, H q represents the number of training data points on device q, W(f(ω q ,x z ),y z ) represents the prediction error, f(ω q ,x z ) represents the predicted value of the device model at the data point x z , y z represents the true label, ωq Represent the local model parameters of device q;

[0037] S2.4: The furniture integration module calculates the model parameters of each smart home device through the weighted average method to aggregate into a global model, and then updates the global model through mean iteration. After the global model is updated, the furniture integration module broadcasts the new global model to all smart home devices and updates the parameter information of the local models of each smart home device;

[0038] S2.5: Each smart home device continues to train locally according to the new global model and performs the next round of mean iteration. When the change in the global model parameters updated by the furniture integration module is less than the preset threshold, the change in the loss function converges to the preset threshold for multiple consecutive rounds, or the maximum number of training rounds is reached, the training ends;

[0039] S2.6: The furniture integration module distributes the final global model to all smart home devices. Each device adjusts its local configuration according to the global model. At the same time, each smart home device makes adaptive adjustments according to the user's personal habits and needs. After the automatic adjustment is completed, it monitors the actual performance of the device in real time and evaluates whether the device meets the expected effect by collecting feedback data. If not, it makes adjustments again.

[0040] The digital home design method based on artificial intelligence, and the specific steps of this design method are as follows:

[0041] Ⅰ. Collect the user's home design requirements, and based on the space data provided by the user and the user's style preferences, conduct space function division and layout design;

[0042] Ⅱ. Generate multiple design schemes that meet the user's taste according to the user's requirements, space characteristics, and personal preferences;

[0043] Ⅲ. Based on the space layout, style recommendation results, the user's budget, and space requirements, select furniture, household appliances, and materials that meet the user's needs;

[0044] Ⅳ. Optimize the lighting and ventilation effects in the space according to the actual layout of the room, the position of the windows, the lighting conditions of the surrounding environment, and the air flow situation;

[0045] Ⅴ. Generate a personalized smart home control solution, update the control solution in real time according to the user's habits, and dynamically adjust the design solution according to the budget set by the user;

[0046] Ⅵ. Present the design results to the user in a three-dimensional form, generate construction drawings and installation guidance plans, analyze the user's evaluations and changing needs, and adjust and improve the design solution;

[0047] Ⅶ. After the home design is completed, the changes in the home environment are detected in real time through Internet of Things sensors to provide various suggestions for later maintenance and upgrading.

[0048] As a further solution of the present invention, the specific steps of the dynamic adjustment design solution in step V are as follows:

[0049] S3.1: Each design element e in the design solution u is regarded as a design variable, and each design element has a corresponding cost parameter. The design solution consists of multiple design elements to form a complete home design, generating multiple optimization bodies. Each optimization body represents a design solution S = (e 1 , e 2 ,... e J ) u ∈ J that meets the user requirements. Then, the design solutions represented by each optimization body are initialized by random selection, and the pheromone concentration of each design element is initialized.

[0050] S3.2: Calculate the selection probability of each design element in the design solution. The optimization body adjusts by selecting the next design element according to the probability. After each optimization body completes a round of scheme selection, the costs of each design element in the corresponding design solution of each optimization body are summed to obtain the total cost of the scheme where C(S t ) represents the total cost of the scheme S of the optimization body t, and c t (e g ) represents the cost of selecting e tg on the gth design element; tg

[0051] S3.3: Compare the total cost of each calculated optimization body with the user's budget. If the total cost exceeds the budget, it means that the design solution of this optimization body is an infeasible solution, and the total pheromone of each design solution is updated according to the preset pheromone evaporation factor and the cost of the design solution.

[0052] S3.4: After the design solution selection is completed, through local search, the scheme is adjusted secondarily, the cost after adjustment is calculated, and it is evaluated whether the total cost can be reduced. Repeatedly perform design solution selection and local search until the predetermined maximum number of iterations is reached or the cost converges to the preset threshold, then stop scheme update, and select the scheme with the smallest final cost and meeting the budget constraint as the final design.

[0053] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0054] 1. The present invention receives the information of each home appliance device selected by the device configuration module through the furniture integration module, identifies and models the intelligent furniture devices among them. After the connection of each group of smart home devices is completed, the number of connected devices is verified. If there is a missed connection, the missed connection device is reconnected manually or automatically. Then, the furniture integration module distributes the global model to each smart home device. Each smart home device replaces the architecture and parameters of the original local model according to the architecture and parameters of the global model, and trains the local model based on the local data collected by itself. After each device completes multiple rounds of local training, it transmits the final model parameters updated by the local training to the furniture integration module. The furniture integration module calculates the model parameters of each smart home device by the weighted average method to aggregate a global model, and then updates the global model through mean iteration. After the global model is updated, the furniture integration module broadcasts the new global model to all smart home devices and updates the parameter information of the local models of each smart home device. The model update is repeated. Then, the furniture integration module distributes the final global model to all smart home devices. Each device adjusts its own local configuration according to the global model. At the same time, each smart home device makes an adaptive adjustment according to the user's personal habits and needs. After the automatic adjustment is completed, the actual performance of the device is monitored in real time, and it is evaluated whether the device meets the expected effect by collecting feedback data. If not, readjustment is carried out. It can effectively protect the user's privacy data, enhance the intelligence and adaptive ability of home appliances, improve the accuracy of real-time decision-making and intelligent control, and can significantly save bandwidth consumption and reduce the network transmission burden.

[0055] 2. In the present invention, each design element in the design scheme is regarded as a design variable, and each design element has a corresponding cost parameter. The design scheme consists of multiple design elements to form a complete home design, generating multiple optimization bodies. Each optimization body represents a design scheme that meets the user's requirements. Then, the design schemes represented by each optimization body are initialized by random selection, and the pheromone concentration of each design element is initialized. The selection probability of each design element in the design scheme is calculated. The optimization body adjusts by selecting the next design element according to the probability. After each optimization body completes a round of scheme selection, the costs of each design element in the design scheme corresponding to each optimization body are summed up to obtain the total cost of the scheme. The total costs of each optimization body calculated are compared with the user's budget. If the total cost exceeds the budget, it means that the design scheme of this optimization body is an infeasible scheme, and the total pheromone of each design scheme is updated according to the preset pheromone evaporation factor and the cost of the design scheme. After the design scheme selection is completed, through local search, the scheme is adjusted secondly, the cost after adjustment is calculated, and it is evaluated whether the total cost can be reduced. The design scheme selection and local search are repeatedly carried out until the predetermined maximum number of iterations is reached or the cost converges to the preset threshold, then the scheme update is stopped, and the scheme with the minimum final cost and meeting the budget constraint is selected as the final design, which can realize real-time cost optimization and budget monitoring, effectively avoid budget overrun caused by design changes, reduce the number and cost of later modifications, flexibly respond to design changes, and timely adjust different parts of the scheme to meet new requirements, improving the user experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention.

[0057] Figure 1 It is a system block diagram of a digital home design system based on artificial intelligence proposed by the present invention;

[0058] Figure 2 It is a flow block diagram of a digital home design method based on artificial intelligence proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0060] Embodiment 1

[0061] Refer to Figure 1, a digital home design system based on artificial intelligence, including a requirement collection module, a space planning module, a style recommendation module, a device configuration module, a material selection module, an environmental optimization module, a home integration module, a cost prediction module, and a display and interaction module.

[0062] The requirement collection module is used to collect the specific requirements of users for home design; the space planning module is used to divide the room functions and design the space layout based on the user requirement data.

[0063] Specifically, the space planning module collects the room dimensions, functional requirements, and user's personalized requirements from various data sources such as the information input by the user, the existing home design case library, and environmental data, and preprocesses the collected groups of raw data. Then, the processed groups of data are converted into encoded forms, and the converted groups of data are input into the pre-trained CNN model. The convolutional layer of the CNN extracts local features on each group of data by sliding the convolutional kernel to generate a new feature map. Then, the pooling layer performs max-pooling processing on the feature map to reduce the dimension of the convolutional result. Through multiple layers of convolution and pooling processing, the CNN model continuously divides the functional areas according to the space requirements input by the user and optimizes the space layout. The features extracted through multiple layers of convolution and pooling are passed to the fully connected layer. Then, the fully connected layer analyzes the relationships between different rooms and the space layout requirements and generates the optimal space layout. After the CNN model completes the space layout optimization, it outputs the final space layout plan and design drawings through the output layer. At the same time, each cell in the drawings represents the functional area of each room and is marked.

[0064] The style recommendation module is used to generate personalized style recommendations according to user preferences and space characteristics; the device configuration module is used to automatically select furniture and home appliances through the Internet of Things database based on the space planning results and style recommendations.

[0065] The material selection module is used to recommend material options that meet the budget and have the best cost performance for users; the environmental optimization module is used to simulate the propagation path of light and the trajectory of air flow and design the best lighting layout and ventilation plan.

[0066] Specifically, the environmental optimization module maximizes the use of natural light for the positions of windows and light sources and the layout of the room as the lighting optimization goal, makes the air circulation in the space smooth, avoids dead corners, and improves the air exchange efficiency as the ventilation optimization goal. Based on the lighting optimization goal and the ventilation optimization goal, it designs the lighting optimization objective function and the ventilation optimization objective function, and at the same time adds corresponding constraint conditions to the two sets of objective functions. Taking the latest space layout plan generated by the space planning module, the style recommendation module, the equipment configuration module, and the material selection module as the root node, it calculates the lighting and ventilation effects of the current space layout plan using the lighting optimization objective function and the ventilation optimization objective function, searches recursively for various different configurations in the space layout, and takes the results of the recursive search as child nodes to construct a search tree. At the same time, it calculates the lighting and ventilation effects of each child node, initializes the return values and visit counts of the root node and the child nodes, uses the upper confidence bound algorithm to calculate the UCT values of each node in the search tree, and selects the child node with the highest UCT value layer by layer. When the selection stage reaches an unexpanded leaf node, it enters the expansion stage, adjusts the layout according to the known information, assigns a new evaluation status to each new layout node, and at the same time adds this space layout as a new child node to the search tree. After the expansion is completed, according to the current layout configuration, a radiation transfer model and a fluid dynamics model are established, and the final effects of simulated lighting and ventilation are adjusted randomly, and the comprehensive effects of lighting and ventilation under the current layout are evaluated according to the model results. The evaluation results obtained in the simulation stage are backpropagated to each node on the same path in the search tree, and the return values and visit counts of each node are updated, where L a represents the lighting intensity of the a-th room, k represents the number of all light sources, S b represents the brightness of light source b, A b represents the effective irradiation area of light source b, d b represents the distance from light source b to the room, θ b represents the angle between light source b and the room surface, V a represents the ventilation flow rate of the a-th room, C represents the efficiency coefficient of the ventilation equipment, B represents the cross-sectional area of the air duct or window, ΔP represents the pressure difference between both sides of the air duct or window, ρ represents the density of air. The selection, expansion, simulation, and backtracking are repeated until the changes in the return values of each node in the search tree converge to the preset range. Each node in the search tree is traversed, and the node with the largest return value is selected as the optimal solution, and the best lighting layout and ventilation plan are output.

[0067] In this embodiment, the specific calculation formula of the lighting optimization objective function is as follows;

[0068]

[0069] Wherein, L represents the total sum of light intensity; L i represents the light intensity of the i-th room; n represents the number of rooms in the space; L min represents the minimum light requirement for each room;

[0070] The specific calculation formula of the ventilation optimization objective function is as follows:

[0071]

[0072] Wherein, V represents the total ventilation efficiency; V i represents the ventilation flow rate of the i-th room; m represents the number of rooms in the space; V min represents the minimum ventilation requirement for each room;

[0073] The specific evaluation formula for the comprehensive effect of light and ventilation is as follows:

[0074]

[0075] Wherein, Q(v) represents the average effect value of this layout; N(v) represents the number of times the node v is visited; R(v j ) represents the score of the light and ventilation effects of the j-th simulation result.

[0076] The home integration module is used to automatically configure the intelligent system in the home and provide a personalized smart home control solution.

[0077] Specifically, the furniture integration module receives the information of each home device selected by the device configuration module, and identifies and models the intelligent furniture devices among them. When the devices are connected, each device reports its current status to the system according to its own functions and characteristics, and initializes a local model. At the same time, the furniture integration module initializes a set of global models according to the multiple functions of each smart home. When the connection of each group of smart home devices is completed, verify the number of connected devices. If there are missing connections, reconnect the missing devices manually or automatically. After that, the furniture integration module distributes the global model to each smart home device. Each smart home device receives the global model, and replaces the architecture and parameters of the original local model according to the architecture and parameters of the global model. Each smart home device collects data according to the environment and user needs, and trains the local model according to the local data collected by itself, aiming at accurately predicting and adjusting the device parameters to optimize the target, and constructs the corresponding objective function Use the backpropagation algorithm to calculate the gradient and update the model parameters. After each device completes multiple rounds of local training, it transmits the final model parameters updated by the local training to the furniture integration module, where W q (ω q ) represents the loss function of device q in local training, H qrepresents the number of training data points on device q, W(f(ω q , x z ), y z ) represents the prediction error, f(ω q , x z ) represents the predicted value of the device model at the data point x z , y z represents the true label, ω q represents the local model parameters of device q. The furniture integration module calculates the model parameters of each smart home device by the weighted average method and aggregates them into a global model. Then, the global model is updated iteratively by the mean value. After the global model update is completed, the furniture integration module broadcasts the new global model to all smart home devices and updates the parameter information of the local models of each smart home device. Each smart home device continues to train locally according to the new global model and performs the next round of mean value iteration. When the change in the update of the global model parameters of the furniture integration module is less than the preset threshold, the change in the loss function converges to the preset threshold for multiple consecutive rounds, or the maximum number of training rounds is reached, the training ends. The furniture integration module distributes the final global model to all smart home devices, and each device adjusts its local configuration according to the global model. At the same time, each smart home device makes adaptive adjustments according to the user's personal habits and needs. After the automatic adjustment is completed, the actual performance of the device is monitored in real time, and whether the device meets the expected effect is evaluated by collecting feedback data. If not, readjustment is performed.

[0078] The cost prediction module is used to perform budget prediction on user requirements and dynamically adjust the design scheme according to the budget set by the user; the display and interaction module is used to display the 3D model and renderings of the user's home design scheme.

[0079] Embodiment 2

[0080] Referring to Figure 2 , a digital home design method based on artificial intelligence, the specific steps of the design method are as follows:

[0081] Collect the user's home design requirements, and perform space function division and layout design according to the space data provided by the user and the user's style preference.

[0082] Generate multiple design schemes that meet the user's taste according to the user's requirements, space characteristics, and personal preferences.

[0083] Based on the space layout, style recommendation results, the user's budget, and space requirements, select furniture, household appliances, and materials that meet the user's needs.

[0084] Optimize the lighting and ventilation effects in the space according to the actual layout of the room, the position of the window, the lighting conditions of the surrounding environment, and the air flow situation.

[0085] Generate personalized smart home control solutions, update the control solutions in real time according to user habits, and dynamically adjust the design solutions according to the budget set by the user.

[0086] Specifically, each design element in the design plan u As a design variable, each design element has a corresponding cost parameter. The design scheme is composed of multiple design elements to form a complete home design, generating multiple groups of optimization bodies, each of which represents a design scheme that meets the user's requirements S = (e 1 ,e 2 ,...e J )u∈J, and then initialize the design scheme represented by each optimization body by random selection, and initialize the pheromone concentration of each design element, calculate the selection probability of each design element in the design scheme, and the optimization body selects the next design element for adjustment according to the probability. After each optimization body completes a round of scheme selection, the cost of each design element in the corresponding design scheme of each optimization body is added up to obtain the total cost of the scheme Where C(S t ) represents the solution S for optimizing volume t t The total cost, c g (e tg ) represents the selection of e on the g-th design element tg The calculated total cost of each optimization body is compared with the user budget. If the total cost exceeds the budget, it means that the design of the optimization body is an infeasible solution. The total pheromone of each design solution is updated according to the preset pheromone evaporation factor and the cost of the design solution. After the design solution is selected, the solution is adjusted twice through local search, the adjusted cost is calculated, and it is evaluated whether the total cost can be reduced. The design solution selection and local search are repeated until the predetermined maximum number of iterations is reached or the cost converges to the preset threshold. The solution update is stopped, and the solution with the minimum final cost and meeting the budget constraint is selected as the final design.

[0087] The design results are presented to users in three-dimensional form, and construction drawings and installation instructions are generated. The user's evaluation and demand changes are analyzed to adjust and improve the design plan.

[0088] After the home design is completed, the IoT sensors will be used to detect changes in the home environment in real time to provide suggestions for subsequent maintenance and upgrades.

Claims

1. A digital home design system based on artificial intelligence, characterized by: It includes demand collection module, space planning module, style recommendation module, equipment configuration module, material selection module, environment optimization module, home integration module, cost prediction module and display interaction module; The demand collection module is used to collect users' specific demands for home design; The space planning module is used to divide room functions and design space layout based on user demand data; The style recommendation module is used to generate personalized style recommendations based on user preferences and space characteristics; The equipment configuration module is used to automatically select furniture and home appliances through the Internet of Things database based on space planning results and style recommendations; The material selection module is used to recommend material options that meet the user's budget and have the best cost-effectiveness; The environmental optimization module is used to simulate the propagation path of light and the trajectory of air flow, and design the best lighting layout and ventilation plan; The home integration module is used to automatically configure the smart system in the home and provide a personalized smart home control solution; The cost prediction module is used to make budget predictions for user needs and dynamically adjust the design plan according to the budget set by the user; The display interaction module is used to provide the 3D model and renderings of the user's home design plan.

2. The digital home design system based on artificial intelligence according to claim 1 is characterized in that: The specific steps of the environmental optimization module to design the optimal lighting layout and ventilation plan are as follows: S1.1: The environment optimization module takes the position of windows, light sources and the layout of the room to maximize the use of natural light as the lighting optimization target, and takes the smooth air flow in the space, avoiding dead corners and improving air exchange efficiency as the ventilation optimization target. Based on the lighting optimization target and the ventilation optimization target, the lighting optimization objective function and the ventilation optimization objective function are designed, and corresponding constraints are added to the two sets of objective functions. S1.2: The latest space layout scheme generated by the space planning module, style recommendation module, equipment configuration module and material selection module is used as the root node. The lighting and ventilation effects of the current space layout scheme are calculated using the lighting optimization objective function and the ventilation optimization objective function. Different configurations in the space layout are recursively searched, and the results of the recursive search are used as child nodes to build a search tree. The lighting and ventilation effects of each child node are calculated at the same time, and the return value and visit count of the root node and child nodes are initialized. S1.3: Use the upper confidence bound algorithm to calculate the UCT value of each node in the search tree, and select the child node with the highest UCT value layer by layer. When the selection stage reaches a leaf node that is not fully expanded, enter the expansion stage, adjust the layout based on the known information, and assign a new evaluation state to each new layout node. At the same time, add the spatial layout as a new child node to the search tree; S1.4: After the expansion is completed, establish a radiation transfer model based on the current layout configuration and fluid dynamics models The final effect of lighting and ventilation is simulated by random adjustment, and the comprehensive effect of lighting and ventilation under the current layout is evaluated according to the model results. The evaluation results obtained in the simulation stage are back-propagated to each node of the same path in the search tree, and the reward value and the number of visits of each node are updated, where L a represents the light intensity of the ath room, k represents the number of all light sources, S b represents the brightness of light source b, A b represents the effective irradiation area of ​​light source b, d b represents the distance from the light source b to the room, θ b represents the angle between the light source b and the room surface, V a represents the ventilation flow rate of the ath room, C represents the efficiency coefficient of the ventilation equipment, B represents the cross-sectional area of ​​the duct or window, ΔP represents the pressure difference on both sides of the duct or window, and ρ represents the density of the air; S1.5: Repeat selection, expansion, simulation and backtracking until the change in the reward value of each node in the search tree converges to the preset range, traverse each node of the search tree, and select the node with the largest reward value as the optimal solution, and output the best lighting layout and ventilation solution.

3. The digital home design system based on artificial intelligence according to claim 2 is characterized in that: The specific calculation formula of the illumination optimization objective function described in S1.1 is as follows: Where, L represents the total light intensity; L i represents the light intensity of the i-th room; n represents the number of rooms in the space; L min Represents the minimum lighting requirements for each room; The specific calculation formula of the ventilation optimization objective function described in S1.1 is as follows: Where V represents the total ventilation efficiency; V i represents the ventilation flow rate of the i-th room; m represents the number of rooms in the space; V min Represents the minimum ventilation requirements for each room; The specific evaluation formula for the comprehensive effect of lighting and ventilation described in S1.4 is as follows: In the formula, Q(v) represents the average effect value of the layout; N(v) represents the number of times node v is visited; R(v j ) represents the lighting and ventilation effect scores of the j-th simulation result.

4. The digital home design system based on artificial intelligence according to claim 1 is characterized in that: The specific steps of the home integration module automatically configuring the home's intelligent system are as follows: S2.1: The furniture integration module receives the information of each home appliance selected by the device configuration module, and identifies and models the smart furniture appliances. When the appliances are connected, each appliance reports its current status to the system according to its own functions and characteristics, and initializes a local model. At the same time, the furniture integration module initializes a set of global models according to the multiple functions of each smart home. S2.2: After each group of smart home devices is connected, the number of connected devices is verified. If there is any missing device, the missing device is reconnected manually or automatically. After that, the furniture integration module sends the global model to each smart home device. Each smart home device receives the global model and replaces the architecture and parameters of the original local model according to the architecture and parameters of the global model. S2.3: Each smart home device collects data according to the environment and user needs, and trains the local model based on the local data collected by itself, with the optimization goal of accurately predicting and adjusting device parameters, and constructing the corresponding objective function The back propagation algorithm is used to calculate the gradient and update the model parameters. After completing multiple rounds of local training, each device transmits the final model parameters updated by local training to the furniture integration module, where W q (ω q ) represents the loss function of device q in local training, H q represents the number of training data points on device q, W(f(ω q ,x z ),y z ) represents the prediction error, f(ω q ,x z ) represents the device model at data point x z The predicted value on y z represents the true label, ω q represents the local model parameters of device q; S2.4: The furniture integration module calculates the model parameters of each smart home device through the weighted average method and aggregates them into a global model. Then, the global model is updated through mean iteration. After the global model is updated, the furniture integration module broadcasts the new global model to all smart home devices and updates the parameter information of the local model of each smart home device. S2.5: Each smart home device continues to train locally according to the new global model and performs the next round of mean iteration. When the change in the global model parameter update of the furniture integration module is less than the preset threshold, the change in the loss function of multiple consecutive rounds converges to the preset threshold, or the maximum number of training rounds is reached, the training ends; S2.6: The furniture integration module sends the final global model to all smart home devices. Each device adjusts its local configuration according to the global model. At the same time, each smart home device makes adaptive adjustments based on the user's personal habits and needs. After the automatic adjustment is completed, the actual performance of the device is monitored in real time, and feedback data is collected to evaluate whether the device has achieved the expected effect. If not, it will be readjusted.

5. A digital home design method based on artificial intelligence, used to implement the functions of a digital home design system based on artificial intelligence as described in any one of claims 1 to 4, characterized in that: The specific steps of this design method are as follows: Ⅰ. Collect users' home design needs, and divide space functions and design layouts based on the space data provided by users and their style preferences; Ⅱ. Generate multiple design solutions that suit the user's taste based on the user's needs, space characteristics and personal preferences; Ⅲ. Based on the space layout, style recommendation results, user's budget and space requirements, select furniture, home appliances and materials that meet user needs; IV. Optimize the lighting and ventilation effects in the space according to the actual layout of the room, the location of the windows, the lighting conditions of the surrounding environment and the air flow; V. Generate personalized smart home control solutions, update the control solutions in real time according to user habits, and dynamically adjust the design solutions according to the budget set by the user; Ⅵ. Present the design results to users in three-dimensional form, generate construction drawings and installation guidance plans, analyze user evaluations and demand changes, and adjust and improve the design plan; Ⅶ. After the home design is completed, the changes in the home environment are detected in real time through IoT sensors to provide suggestions for subsequent maintenance and upgrades.

6. The digital home design method based on artificial intelligence according to claim 5 is characterized in that: The specific steps of the dynamic adjustment design scheme described in step V are as follows: S3.1: Each design element in the design plan u As a design variable, each design element has a corresponding cost parameter. The design scheme is composed of multiple design elements to form a complete home design, generating multiple groups of optimization bodies, each of which represents a design scheme that meets the user's requirements S = (e1, e2, ... e J ) u∈J, then initialize the design scheme represented by each optimization body by random selection, and initialize the pheromone concentration of each design element; S3.2: Calculate the selection probability of each design element in the design scheme. The optimization body selects the next design element for adjustment according to the probability. After each optimization body completes a round of scheme selection, the cost of each design element in the corresponding design scheme of each optimization body is added up to obtain the total cost of the scheme. Where C(S t ) represents the solution S for optimizing volume t t The total cost, c g (e tg ) represents the selection of e on the g-th design element tg Costs; S3.3: Compare the calculated total cost of each optimization body with the user's budget. If the total cost exceeds the budget, it means that the design of the optimization body is an infeasible solution. The total pheromone of each design solution is updated according to the preset pheromone evaporation factor and the cost of the design solution. S3.4: After the design scheme is selected, the scheme is adjusted twice through local search, the adjusted cost is calculated, and it is evaluated whether the total cost can be reduced. The design scheme selection and local search are repeated until the predetermined maximum number of iterations is reached or the cost converges to the preset threshold. The scheme update is stopped, and the scheme with the lowest final cost and meeting the budget constraint is selected as the final design.

Images