Home AI weather lighting system and method

Abstract

The invention discloses a home AI weather lighting system and method, and relates to the field of weather lighting, the home AI weather lighting system comprises an acquisition processing module, a fusion decision module, an optimization control module, a joint control module and a cooperative control module, outdoor weather data is acquired and preprocessed to obtain a structured weather database; hypergraph fusion decision making is carried out based on the structured weather database and the user behavior log, and an environment dynamic code is obtained; constructing a multi-objective optimization function based on environment dynamic coding to solve an optimal control strategy, and generating a Pareto optimal solution set; spectral sound field collaborative optimization is carried out through a Pareto optimal solution set to obtain a photoacoustic combined control strategy, multi-mode control signal analysis and collaborative optimization are carried out based on the photoacoustic combined control strategy to obtain a final multi-device collaborative control instruction, a user can obtain immersive environment experience, and multi-device work collaboration can be achieved.

Description

Technical Field

[0001] This invention relates to the field of weather lighting, specifically to a home AI weather lighting system and method. Background Technology

[0002] As consumers increasingly demand higher quality of life and personalization, their requirements for lighting products are also rising, along with the growing demand for smart homes. However, existing home AI weather lighting systems using LEDs cannot continuously adjust to the environment, failing to consider factors such as body sensation, sound, color temperature, and brightness. They also cannot synchronize outdoor temperature, humidity, fog, wind, rain, thunder, and lightning, making it difficult to provide users with an immersive environmental experience. Furthermore, they lack deep integration of light, sound, and environmental parameters, and cannot match dynamic weather with user behavior. They also fail to adequately implement multi-objective constraints, failing to comprehensively consider users' actual feelings, comfort levels, and energy consumption. Additionally, they lack accurate scheduling of multi-device collaboration, leading to command conflicts and asynchronous responses, hindering multi-device collaborative operation and failing to meet the demands of smart homes for intelligent, accurate, and immersive experiences.

[0003] To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention

[0004] To address the technical problems raised in the background section, this invention is proposed. Embodiments of this invention provide a home AI weather lighting system and method.

[0005] The objective of this invention can be achieved through the following technical solution: a home AI weather lighting system, including a data acquisition and processing module, a fusion decision module, an optimization control module, a joint control module, and a collaborative control module. The data acquisition and processing module collects and preprocesses outdoor weather data to obtain a structured weather database. The fusion decision module performs hypergraph fusion decision-making based on the structured weather database and user behavior logs to obtain dynamic environmental codes; The optimization control module constructs a multi-objective optimization function based on dynamic environmental coding to solve for the optimal control strategy and generate a Pareto optimal solution set. The joint control module performs spectral acoustic field co-optimization through Pareto optimal solution set to obtain a photoacoustic joint control strategy. The collaborative control module performs multi-modal control signal analysis and collaborative optimization based on the photoacoustic joint control strategy to obtain the final multi-device collaborative control command.

[0006] Furthermore, the structured weather database is analyzed as follows: Real-time acquisition of outdoor spectral distribution, light intensity, temperature and humidity, fog, and environmental sound waveform data; synchronous acquisition, processing, and time-series matching of the above multi-source heterogeneous data to obtain a multimodal weather synchronous dataset; calibration, denoising, and data optimization of the multimodal weather synchronous dataset to obtain a standardized weather data stream; storage of the standardized weather data stream in a time-series database; and establishment of a multidimensional data index to form a structured weather database.

[0007] Furthermore, the environmental dynamic coding analysis is as follows: Based on a structured weather database, multidimensional low-level static features are obtained through standardized algorithms and norms. The supergraph fusion decision signal and multidimensional low-level static features are combined in chronological order to form a multivariate time series. The dynamic environmental evolution pattern is extracted by multivariate time series and noise interference is filtered out to form a dynamic environmental code, which includes time series feature values, spectral feature values, acoustic feature values ​​and prediction feature values.

[0008] Furthermore, the hypergraph fusion decision signal analysis is as follows: User-defined function parameters are extracted from user behavior logs. A weather configuration hypergraph is constructed based on these parameters and a structured weather database. The specific node set is defined as environmental measurement nodes, user configuration nodes, and derived feature nodes. The hyperedge set is generated using a hypergraph clustering algorithm. The importance weight of each node to each hyperedge is calculated. Based on the importance weight of each node to each hyperedge, the hyperedge information containing all nodes is aggregated to update the feature vector of each node. A learnable global query vector is introduced to interact with the updated feature vector of each node to calculate the importance weight of each node to the final global decision. The hypergraph fusion decision signal is obtained by weighted summation of the updated feature vector of each node and the importance weight of each node to the final global decision.

[0009] Furthermore, the Pareto optimal solution set analysis is as follows: Set a dynamic weight for the timestamp, create a time-varying visual comfort function based on the dynamic weight of the timestamp, obtain the visual comfort level by passing the time-varying visual comfort function based on the temporal feature value and spectral feature value of the environment dynamic coding, and calculate the comfort index by combining the thermal comfort index and the visual comfort index. Based on the predicted feature value energy consumption index of environmental dynamic coding, the energy consumption temperature difference of air conditioning is analyzed to obtain the energy consumption index. Based on the environmental realism index, comfort index and energy consumption index, a multi-objective optimization function is constructed, and equipment physical constraints and user and scenario constraints are applied to obtain the Pareto optimal solution set.

[0010] Furthermore, the environmental realism index and thermal comfort index are analyzed as follows: A spectral generation function is constructed based on the Gaussian function. The spectral realism is obtained by using the spectral feature value based on the dynamic environmental coding. The sound field realism is obtained by calculating the cosine similarity between the acoustic feature value of the generated indoor sound and the acoustic feature value of the dynamic environmental coding. The environmental realism index is obtained by calculating the spectral realism and the sound field realism. Based on the time-series and predicted feature values ​​of the environment dynamic coding, thermal comfort is output through the thermal comfort objective function. The thermal comfort index is obtained by normalizing the deviation between the predicted average voting of the set temperature and the target temperature using the predicted average voting prediction model.

[0011] Furthermore, the optical-acoustic joint control strategy is analyzed as follows: Based on the dynamic environmental coding, the acoustic feature values ​​are matched with the corresponding sound source amplitude and phase parameters from the pre-set sound source library according to the sound source category coding. The complex sound pressure value at each point is calculated by the sound field reconstruction algorithm on the discretized indoor space grid points, and the complex sound pressure amplitude values ​​on all grid points are formed into a three-dimensional sound field distribution matrix. The optimal spectral driving vector and the three-dimensional sound field distribution matrix are used to establish the spectral sound field regular multiple loss function. The optical-acoustic joint control strategy is obtained by solving the alternating direction multiplier method.

[0012] Furthermore, the optimal spectral driving vector analysis is as follows: A dataset is constructed by experimentally measuring the output spectra of different driving current combinations. Based on the dataset, current-to-spectrum mapping analysis is performed using a multivariate regression method to obtain the LED spectral response function. An ideal policy point is defined, and the standardized Euclidean distance from each policy in the Pareto optimal solution set to the ideal policy point is calculated. The policy is selected by taking the minimum value of the standardized Euclidean distance. Based on the policy selected by the decision rule and the LED spectral response function, the optimal spectral driving vector is obtained by optimizing the model through spectral reconstruction.

[0013] Furthermore, the final multi-device collaborative control command is analyzed as follows: Based on the lighting control component in the photoacoustic joint control strategy, a spectral mapping optimization function is constructed to minimize the difference between the actual output spectrum of the LED light source and the target spectrum. The driving current vector and modulation parameter vector of the spectral mapping optimization function are iteratively solved by the projection gradient descent method to obtain the adaptive spectral driving command. The adaptive spectrum drive command, the sound field control component of the photoacoustic joint control strategy, and the temperature and humidity setpoints of the decision rule selection strategy are coordinated and formatted to obtain a preliminary multi-device control signal set. Discrete event scheduling is performed on the initial set of multi-device control signals to obtain the final multi-device collaborative control commands.

[0014] As a preferred embodiment of the present invention, the home AI weather lighting method includes the following steps: Outdoor weather data is collected and preprocessed to obtain a structured weather database; Hypergraph fusion decision-making based on structured weather database and user behavior logs yields dynamic environmental coding; Based on dynamic environmental coding, a multi-objective optimization function is constructed to solve the optimal control strategy and generate a Pareto optimal solution set. The spectral acoustic field is co-optimized by Pareto optimal solution set to obtain a photoacoustic joint control strategy; Based on the photoacoustic joint control strategy, multi-modal control signal analysis and collaborative optimization are performed to obtain the final multi-device collaborative control command.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention collects and preprocesses outdoor weather data to obtain a structured weather database; performs hypergraph fusion decision-making based on the structured weather database and user behavior logs to obtain dynamic environmental coding; constructs a multi-objective optimization function based on the dynamic environmental coding to solve for the optimal control strategy and generate a Pareto optimal solution set; and performs spectral sound field co-optimization through the Pareto optimal solution set to obtain a photoacoustic joint control strategy. LED lighting can continuously adjust according to the environment, considering tactile sensation, sound, color temperature, and brightness; simultaneously, it can synchronize outdoor temperature, humidity, fog, wind, rain, thunder, and lightning, providing users with an immersive environmental experience. It enables deep integration of light, sound, and environmental parameters, matching dynamic weather with user behavior; and fully implements multi-objective constraints, comprehensively considering the user's actual feelings, comfort level, and energy consumption.

[0016] 2. This invention analyzes and optimizes multimodal control signals based on a photoacoustic joint control strategy to obtain the final multi-device collaborative control command. It can accurately schedule the collaboration of multiple devices, and is less prone to command conflicts and asynchronous responses. It can achieve multi-device collaborative operation and meet the needs of smart homes for intelligent, accurate and immersive experience. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The following drawings are not drawn to scale according to the actual size, but are intended to show the main idea of ​​the present invention.

[0018] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention; Figure 3 This is a flowchart of step S5 of the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are also within the scope of protection of the present invention.

[0020] like Figure 2 As shown, the home AI weather lighting method includes the following steps: Step S1: Collect and preprocess outdoor weather data to obtain a structured weather database; Step S2: Perform hypergraph fusion decision-making based on the structured weather database and user behavior logs to obtain dynamic environmental coding; Step S3: Construct a multi-objective optimization function based on dynamic environmental coding to solve for the optimal control strategy and generate a Pareto optimal solution set; Step S4: Perform spectral acoustic field co-optimization using the Pareto optimal solution set to obtain the photoacoustic joint control strategy; Step S5: Perform multi-modal control signal analysis and collaborative optimization based on the photoacoustic joint control strategy to obtain the final multi-device collaborative control command.

[0021] In this embodiment, the analysis steps of step S1 are as follows: Real-time acquisition of outdoor spectral distribution, light intensity, temperature and humidity, fog, and environmental sound waveform data; synchronous acquisition, processing, and time-series matching of the above multi-source heterogeneous data to obtain a multimodal weather synchronous dataset; calibration, denoising, and data optimization of the multimodal weather synchronous dataset to obtain a standardized weather data stream; storage of the standardized weather data stream in a time-series database; and establishment of a multidimensional data index to form a structured weather database.

[0022] In this embodiment, outdoor spectral distribution, light intensity, temperature and humidity, fog level, and ambient sound waveform data are collected in real time using outdoor spectral sensors, illuminometers, temperature and humidity sensors, fog level detectors, and microphone arrays. Multi-source data synchronous acquisition technology ensures the temporal consistency of data from each sensor, and a timestamp alignment algorithm is used to perform time-series matching of heterogeneous data to obtain a multimodal weather synchronous dataset. The multi-source data synchronous acquisition technology and timestamp alignment algorithm are existing technologies and will not be elaborated upon here. A multi-sensor fusion algorithm based on extended Kalman filtering is used to establish a system state-space model including light intensity, spectral distribution, temperature and humidity, fog level, and sound characteristics. The original data is iteratively optimized through a prediction-correction loop to suppress noise interference and statistically detect and remove outliers, resulting in a standardized weather data stream. Finally, the standardized weather data stream is stored in a time-series database. A time-partitioned storage strategy is adopted, and a multi-dimensional joint index structure is established according to spatial location, sensor type, and physical quantity dimensions. Specifically, a B+ tree index is built with the timestamp as the primary key for range queries. At the same time, inverted indexes and bitmap indexes are built for dimensions such as spectral bands, temperature ranges, humidity levels, and fog values. The storage and retrieval efficiency of time-series data is optimized through data sharding and load balancing strategies, resulting in a structured weather database that supports multiple dimensions and improves query and analysis speed.

[0023] In this embodiment, the analysis steps of step S2 are as follows: User-defined function parameters are extracted from user behavior logs. A weather configuration hypergraph structure is constructed based on these parameters and a structured weather database. The specific node set is defined as environmental measurement nodes, user configuration nodes, and derived feature nodes. The hyperedge set is generated using a hypergraph clustering algorithm. The importance weight of each node to each hyperedge is calculated. Based on the importance weight of each node to each hyperedge, the hyperedge information containing all nodes is aggregated to update the feature vector of each node. A learnable global query vector is introduced to interact with the updated feature vector of each node to calculate the importance weight of each node to the final global decision. The hypergraph fusion decision signal is obtained by weighted summation of the updated feature vector of each node and the importance weight of each node to the final global decision. Based on a structured weather database, multidimensional low-level static features are obtained through standardized algorithms and norms. The supergraph fusion decision signal and multidimensional low-level static features are combined in chronological order to form a multivariate time series. The dynamic environmental evolution pattern is extracted by multivariate time series and noise interference is filtered out to form a dynamic environmental code, which includes time series feature values, spectral feature values, acoustic feature values ​​and prediction feature values. In this embodiment, user-defined functional parameters are extracted from user behavior logs. Specifically, these parameters include the color temperature, brightness level, air conditioning temperature, and speaker volume level set by the user at time t. A timestamp alignment algorithm is used to match the structured weather database with these user-defined parameters over time. A weather configuration hypergraph structure is constructed based on a hypergraph attention network. The specific node set is defined as environmental measurement nodes, user configuration nodes, and derived feature nodes. Each node in the environmental measurement nodes represents a raw physical quantity collected at a specific time from the structured weather database. These are the K-dimensional spectral intensity vector, illuminance value, temperature value, humidity value, and haze value collected at time t. Each node in the user-configured nodes represents a user-defined functional parameter. The derived feature nodes are high-level semantic features derived from the original physical quantities and user-defined functional parameters, specifically time-segment nodes derived based on timestamps. The hyperedge set is used to capture the complex, unpaired relationships between these nodes. Using a hypergraph clustering algorithm, it can analyze all node data to identify which nodes frequently appear simultaneously or have strong correlations, and generate a hyperedge from the set of these nodes. For hyperedge a, the importance weights of all nodes within it are calculated: ,in Let b represent the importance weight of node c to hyperedge a, b represent the learnable attention vector, and W represent the learnable weight matrix. This represents the feature vector of node c. Let represent the initial features of hyperedge a, represent the average features of the nodes connected by the hyperedge, and σ represent the activation function. The expression represents the concatenation operation, T represents the transpose, exp represents the exponential function used to calculate the Softmax normalized numerator / denominator of the attention function, ensuring the weights satisfy the probability distribution, and u represents the aggregation of hyperedge information containing node v for all nodes u contained in hyperedge a to update node features. , Let E(c) represent the updated feature vector of node c, and let E(c) contain the set of all hyperedges of node c. The importance weight of hyperedge a to node c is represented by a separate attention mechanism. A learnable global query vector q is introduced, which interacts with the attention of all updated node features to calculate the importance weight of each node for the final global decision. ,in This represents the importance weight of node c to the global decision. Let V represent the learnable weight matrix, V represent the set of all nodes, and the final hypergraph fusion decision signal G is obtained by weighted summation of the features of all nodes: The hypergraph fusion decision signal G is a fixed-dimensional dense vector that integrates key information from three aspects: real-time environmental state, long-term user preferences, and current context. Based on a structured weather database, multi-dimensional low-level static features are calculated and obtained through standardized algorithms and specifications. These features include spectral characteristics, acoustic characteristics, and basic environmental parameters. Spectral characteristics include outdoor light-related color temperature (PAR), deviation from the blackbody radiation line, wavelength and intensity characteristics of major spectral peaks, and melanin ratio. Specifically, the spectral data from the structured weather database is processed using the CIE 1931 standard chromaticity calculation procedure to obtain the PAR. The deviation from the blackbody radiation line is calculated using the closest point algorithm to find the blackbody radiation trajectory. The spectral peak finding algorithm is used to analyze the spectral curves of the structured weather database, extracting the wavelength and intensity characteristics of major spectral peaks. This is done according to CIE S... The black-view luminous efficacy function, defined in standard 026:2018, is used to calculate the black-view ratio by weighted integration of spectral data. Acoustic features include sound category codes, Mel frequency cepstral coefficients, and equivalent sound levels. Audio signals from a structured weather database are analyzed using an acoustic event detection model to output sound category codes. A standard speech signal processing workflow is employed, including framing, windowing, short-time Fourier transform, Mel filter bank processing, and discrete cosine transform, to extract Mel frequency cepstral coefficients. Based on IEC 61672:2013, time-domain weighted and integral operations are performed on the sound signals to calculate the equivalent sound level. The specific standardized algorithms and specifications described above are existing technologies and will not be elaborated upon here. Basic environmental parameters include spectral intensity vectors, illuminance values, temperature values, humidity values, and haze values ​​from the structured weather database. The hypergraph fusion decision signal and multidimensional low-level static features are combined chronologically to form a multivariate time series. This multivariate time series is then input into a spatiotemporal convolutional memory network for deep temporal modeling. This network captures short-term local dependency patterns in sequences through one-dimensional causal convolution, learns long-term dynamic evolution patterns and controls noise using gated recurrent units, and focuses on key time steps using an attention mechanism. The network's final output is a dynamic environmental encoding, which is a dynamic vector that integrates recent historical state change trends with current time-specific features. Specifically, it includes: 1) temporal feature values, encoding the changing trends, periodicity, and volatility of environmental parameters, corresponding to the first / second derivatives of temperature. Periodic features extracted by Fourier transform Such as the intensity of diurnal rhythms and the volatility of environmental parameters The environmental parameters are spectral intensity vector, illuminance value, temperature value, humidity value, and haze value. 2) Spectral characteristic values, which encode the key attributes of the outdoor spectrum, correspond to the estimated correlated color temperature of outdoor light and the deviation of outdoor light from the blackbody radiation line. Wavelength and intensity of main spectral peaks And the melanopsin luminous efficacy ratio, which affects human circadian rhythms. 3) Acoustic feature values, encoding key attributes of outdoor sound fields, sound category encoding. Mel frequency cepstral coefficients Equivalent sound level in outdoor environment 4) Predicted eigenvalues: the network's prediction of the short-term future state, such as the predicted temperature in the next few minutes. And the predicted value of illuminance in the next few minutes .

[0024] In this embodiment, the analysis steps of step S3 are as follows: A spectral generation function is constructed based on the Gaussian function. The spectral realism is obtained by using the spectral feature value based on the dynamic environmental coding. The sound field realism is obtained by calculating the cosine similarity between the acoustic feature value of the generated indoor sound and the acoustic feature value of the dynamic environmental coding. The environmental realism index is obtained by calculating the spectral realism and the sound field realism. Based on the time-series and predicted feature values ​​of the environment dynamic coding, thermal comfort is output through the thermal comfort objective function. The thermal comfort index is obtained by normalizing the deviation between the predicted average voting of the set temperature and the target temperature through the predicted average voting prediction model. The timestamp dynamic weight is set, and a time-varying visual comfort function is created based on the timestamp dynamic weight. The visual comfort is obtained by passing the time-varying visual comfort function based on the time-series and spectral feature values ​​of the environment dynamic coding. The comfort index is obtained by calculating the thermal comfort index and the visual comfort index. Based on the predicted feature value energy consumption index of environmental dynamic coding, the energy consumption temperature difference of air conditioning is analyzed to obtain the energy consumption index. Based on the environmental realism index, comfort index and energy consumption index, a multi-objective optimization function is constructed, and equipment physical constraints and user and scenario constraints are applied to obtain the Pareto optimal solution set.

[0025] In this embodiment, the spectral accuracy is obtained from the spectral feature values ​​based on dynamic environmental coding through a spectral generation function. ,in This represents the spectral fidelity index; the closer the value is to 1, the more realistic the spectrum. indoor (λ) represents the spectral distribution of the generated spectrum indoors, and the energy distribution as a function of wavelength λ. It is obtained by directly measuring the light intensity data of indoor light sources at different wavelengths using a spectrometer. This represents the spectral generation function, an approximate model of the target outdoor spectrum, and a Gaussian mixture model that uses the superposition of multiple Gaussian functions to simulate complex spectral shapes. Where I represents the number of Gaussian components, This represents the set of all parameters of the model. This represents the amplitude of the i-th Gaussian component. Obtaining the corresponding peak intensity information This represents the center wavelength of the i-th Gaussian component. Extract from, This represents the standard deviation of the i-th Gaussian component. The standard deviation of B is the baseline offset, which is a very small constant. The Mel-frequency cepstral coefficients of the generated indoor sound are calculated and used in the environmental dynamic coding. The cosine similarity is used to obtain the sound field realism. The sound field realism and spectral realism are weighted and multiplied by the corresponding weight factor coefficients to obtain the environmental realism index. Based on the temporal feature values ​​and predicted feature values ​​of the dynamic environmental coding, the thermal comfort index is output through the thermal comfort objective function. Where f(...) represents a dynamic function, Where α is the predictability compensation coefficient and β is the inertia compensation coefficient. They are Predicted outdoor temperature The outdoor temperature change rate and the current indoor temperature are used to calculate the thermal comfort index by normalizing the average voting deviation between the predicted set temperature and the target temperature using a predicted average voting model. PMV(...) represents the predicted average voting model, which is existing technology and will not be elaborated upon here. The visual comfort level (CCT) is obtained by using a time-varying visual comfort function based on the set temperature, temporal and spectral feature values ​​of the environment dynamically encoded, and the time-varying visual comfort function. target =g( , g is the time-varying visual comfort function, CCT target =w×CC Toutdoor +(1-w)×CCT comfort CC Toutdoor express Outdoor light source correlated color temperature (CCT) estimate comfort This indicates the preset comfortable color temperature values ​​for day and night: 5000K for daytime (6:00-16:00) and 2700K for nighttime (16:00-6:00). Hour represents the current time, w represents the timestamp, and dynamic weighting calculates the visual comfort and thermal comfort indices by weighting them and multiplying them by the corresponding weighting factor coefficients to obtain the comfort index. Based on the predicted feature value energy consumption index of the environment dynamic coding, the air conditioning energy consumption temperature difference analysis is performed to obtain the energy consumption index. ,in These represent the driving current and voltage of the e-th LED channel, respectively. Indicates the operating time of the lighting system. This indicates the maximum temperature range that the air conditioner can handle. Indicates the rated power of the air conditioner. This refers to the operating time of the air conditioning system. Physical constraints on the equipment are hard limitations on the operating parameters of various devices. Specifically, the LED current must be between its minimum and maximum current, determined by hardware and unaffected by environmental dynamic coding (Et). The air conditioning temperature setting must be within the air conditioner's own minimum and maximum temperature range. Regarding the air conditioner's power change rate, the ratio of the difference between the set temperature and the real-time temperature to the adjustment time step Δt cannot exceed the air conditioner's maximum power change rate. The color temperature setting must fall within the physical color gamut determined by the LED chip and phosphor. It also includes user and scene constraints, which are limitations for dynamic adaptation requirements. The absolute boundary is the user-preset temperature range, from the user-preset minimum temperature to the user-preset maximum temperature, unaffected by environmental dynamic coding (Et). Dynamic preference boundary requirements... -δ≤ melanin ratio of indoor light≤ +δ, where δ is the set deviation to ensure synchronization between the biological effects of light and outdoor conditions; dynamic illuminance constraints stipulate that indoor illuminance should not be lower than the outdoor illuminance predicted by the dynamic coding of the environment. The η ratio, with η as the adjustment coefficient, is used to avoid excessive darkness indoors on cloudy days. A multi-objective optimization function is constructed based on environmental realism, comfort, and energy consumption indicators. Minimize: [-F1(X), -F2(X), F3(X)], where F1(X) and F2(X) are the environmental realism and comfort indicators, respectively. That is, maximize the environmental realism and comfort indicators and minimize the energy consumption indicator. Equipment physical constraints and user and scene constraints are applied to obtain the Pareto optimal solution set Ω={ω1,ω2,…,ωn}, where each solution ωn is a control parameter vector that drives the specific instructions of all actuators (LED lights, air conditioners, humidifiers, speakers, etc.).

[0026] In this embodiment, the analysis steps of step S4 are as follows: A dataset is constructed by experimentally measuring the output spectra of different driving current combinations. Based on the dataset, current-to-spectrum mapping analysis is performed using a multivariate regression method to obtain the LED spectral response function. An ideal policy point is defined, and the standardized Euclidean distance from each policy in the Pareto optimal solution set to the ideal policy point is calculated. The policy is selected by taking the minimum value of the standardized Euclidean distance. Based on the policy selected by the decision rule and the LED spectral response function, the optimal spectral driving vector is obtained by optimizing the model through spectral reconstruction. Based on the dynamic environmental coding, the acoustic feature values ​​are matched with the corresponding sound source amplitude and phase parameters from the pre-set sound source library according to the sound source category coding. The complex sound pressure value at each point is calculated by the sound field reconstruction algorithm on the discretized indoor space grid points, and the complex sound pressure amplitude values ​​on all grid points are formed into a three-dimensional sound field distribution matrix. The optimal spectral driving vector and the three-dimensional sound field distribution matrix are used to establish the spectral sound field regular multiple loss function. The optical-acoustic joint control strategy is obtained by solving the alternating direction multiplier method.

[0027] In this embodiment, the specific model of LED chip used in the system is measured under different driving current combinations through experiments. Output spectrum , constitute the dataset The mapping relationship from driving current to spectrum was obtained by fitting using a multiple regression method (such as Gaussian process regression). ,in The fundamental spectrum is represented by the spectral dataset obtained from the measurement. The first L principal components were extracted by principal component analysis (PCA). The regression coefficients and bias terms to be learned are used to obtain the LED spectral response function. Define an ideal policy point [F1] * (X),F2 * (X),F3 * [X] represents the theoretical maximum comfort, the level of realism, and the theoretical minimum energy consumption, respectively. Then, the standardized Euclidean distance from each strategy ω in Ω to the ideal point is calculated. ,in The standard deviation of each objective value is used for normalization. The strategy closest to the ideal point is selected from the Pareto optimal solution set to obtain the policy selected by the decision rule. Based on the policy selected by the decision rule and the LED spectral response function, the optimal spectral driving vector is obtained through the spectral reconstruction optimization model. The spectral reconstruction optimization model includes: ,in The decision rule selects a specific strategy. The specified target spectral vector, R(I) is the regularization term. The regularization coefficient is . Represents the L2 norm. This represents the optimal spectral driving vector, and `arg min` performs the minimum value operation. Acoustic feature values ​​based on dynamic environmental coding are matched with corresponding sound source amplitudes from a pre-set sound source library according to the sound source category encoding. and phase parameters Subsequently, the complex sound pressure value at each point (x, y, z) in the discretized indoor space grid is calculated using a sound field reconstruction algorithm. ,in Let p represent the distance from the sound source q to the spatial point (x, y, z), Q be the wave number, and h be the imaginary unit. The complex sound pressure amplitudes |P(x, y, z)| at all grid points are combined into a three-dimensional matrix, thus obtaining the three-dimensional sound field distribution matrix. This describes the energy distribution of sound waves in an indoor space. A spectral sound field regularized multiple loss function is established by combining the optimal spectral driving vector with the three-dimensional sound field distribution matrix. The photoacoustic joint control strategy is obtained by solving this function using the alternating direction multiplier method. The spectral sound field regularized multiple loss function includes: Where U is the joint control vector Includes lighting control components Harmony control components , This is the sound field response function, which calculates the loudspeaker array drive signal using a beamforming model. The corresponding spatial sound field distribution U is the Frobenius norm of the matrix, used to measure the sound field reconstruction error. u1 and u2 are weighting coefficients used to balance the priority of spectral matching and sound field reconstruction. R(U) is a joint regularization term used to constrain the smoothness and energy consumption of the control signal. The regularization coefficients are solved using the alternating direction multiplier method. The sound field control is fixed to optimize the lighting control, and then the lighting control is fixed to optimize the sound field control. This process is iterated until convergence, yielding the joint sound control strategy. At the same time, it ensures spectral fidelity and accurate sound field reconstruction.

[0028] like Figure 3 As shown, in this embodiment, the analysis steps of step S5 are as follows: Based on the lighting control component in the photoacoustic joint control strategy, a spectral mapping optimization function is constructed to minimize the difference between the actual output spectrum of the LED light source and the target spectrum. The driving current vector and modulation parameter vector of the spectral mapping optimization function are iteratively solved by the projection gradient descent method to obtain the adaptive spectral driving command. The adaptive spectrum drive command, the sound field control component of the photoacoustic joint control strategy, and the temperature and humidity setpoints of the decision rule selection strategy are coordinated and formatted to obtain a preliminary multi-device control signal set. Discrete event scheduling is performed on the initial set of multi-device control signals to obtain the final multi-device collaborative control commands.

[0029] In this embodiment, the illumination control component, i.e., the target spectral vector, is based on the photoacoustic joint control strategy. The driving parameters are calculated using a spectral mapping function F. The purpose of this function is to find a set of driving signals that minimizes the difference between the actual output spectrum Φ(I) of the light source and the target spectrum. ,in Ω(I,D) is the driving current vector of each LED chip channel, D is the modulation parameter vector (such as the PWM duty cycle of each channel), Ω(I,D) is the regularization term used to constrain the feasibility of the solution, and λ1 is the regularization coefficient, which is solved by the projected gradient descent method. The final output is the adaptive spectrum driving command, i.e., the optimal driving current vector. and modulation parameter vector The signal is sent to the PWM generator of the multi-channel LED constant current driver chip and microcontroller. By controlling the current intensity and switching duty cycle of each LED chip channel, the intensity and shape of the synthesized output spectrum are physically adjusted to approximate the target spectrum. The adaptive spectrum drive command, the sound field control component of the photoacoustic joint control strategy, and the temperature and humidity setpoints of the decision rule-selected strategy are passed through a lightweight multimodal fusion network, such as an attention-based MLP. The network's role is not to learn new strategies, but rather to coordinate and format these signals and resolve potential micro-conflicts, such as avoiding sudden playback of extremely high-decibel sounds during rapid cooling. The output is a preliminary control signal set customized for each actuator. ,in These represent the final LED drive signal, multi-channel audio signal, and setpoint instructions sent to the air conditioner and humidifier, respectively. T stands for transpose. MMFNet is a lightweight multimodal fusion network. The final LED drive signal, multi-channel audio signal, and setpoint instructions sent to the air conditioner and humidifier are combined to form the initial multi-device control signal set Dpre. To ensure that all device instructions are synchronized in time and logic and to avoid conflicts (e.g., the light has changed but the sound is delayed), a discrete event scheduler is used. This scheduler assigns a timestamp and transition duration to each control instruction to form an instruction sequence, resulting in the final multi-device collaborative control instruction Dfinal. Specifically, Dfinal = Scheduler(Dpre), where Scheduler represents the discrete event scheduler. The final multi-device collaborative control instruction is a time-stamped, conflict-free instruction sequence sent to each device execution unit, ultimately enhancing the multi-sensory experience.

[0030] like Figure 1 As shown, as another embodiment of the present invention, the home AI weather lighting system includes a data acquisition and processing module, a fusion decision module, an optimization control module, a joint control module, and a collaborative control module. The data acquisition and processing module collects and preprocesses outdoor weather data to obtain a structured weather database. The fusion decision module performs hypergraph fusion decision-making based on the structured weather database and user behavior logs to obtain dynamic environmental codes; The optimization control module constructs a multi-objective optimization function based on dynamic environmental coding to solve for the optimal control strategy and generate a Pareto optimal solution set. The joint control module performs spectral acoustic field co-optimization through Pareto optimal solution set to obtain a photoacoustic joint control strategy. The collaborative control module performs multi-modal control signal analysis and collaborative optimization based on the photoacoustic joint control strategy to obtain the final multi-device collaborative control command.

[0031] The foregoing description is illustrative of the invention and should not be construed as limiting it. Although several exemplary embodiments of the invention have been described, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the invention. Therefore, all such modifications are intended to be included within the scope of the invention as defined in the claims. It should be understood that the foregoing description is illustrative of the invention and should not be construed as limiting it to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The invention is defined by the claims and their equivalents.

Claims

1. A home AI weather lighting system, comprising a data acquisition and processing module, a fusion decision-making module, an optimization control module, a joint control module, and a collaborative control module, characterized in that: The data acquisition and processing module collects and preprocesses outdoor weather data to obtain a structured weather database. The fusion decision module performs hypergraph fusion decision-making based on the structured weather database and user behavior logs to obtain dynamic environmental codes; The optimization control module constructs a multi-objective optimization function based on dynamic environmental coding to solve for the optimal control strategy and generate a Pareto optimal solution set. The joint control module performs spectral acoustic field co-optimization through Pareto optimal solution set to obtain a photoacoustic joint control strategy. The collaborative control module performs multi-modal control signal analysis and collaborative optimization based on the photoacoustic joint control strategy to obtain the final multi-device collaborative control command.

2. The home AI weather lighting system according to claim 1, characterized in that, The structured weather database analysis is as follows: Real-time acquisition of outdoor spectral distribution, light intensity, temperature and humidity, fog, and environmental sound waveform data; synchronous acquisition, processing, and time-series matching of the above multi-source heterogeneous data to obtain a multimodal weather synchronous dataset; calibration, denoising, and data optimization of the multimodal weather synchronous dataset to obtain a standardized weather data stream; storage of the standardized weather data stream in a time-series database; and establishment of a multidimensional data index to form a structured weather database.

3. The home AI weather lighting system according to claim 1, characterized in that, The dynamic coding analysis of the environment is as follows: Based on a structured weather database, multidimensional low-level static features are obtained through standardized algorithms and norms. The supergraph fusion decision signal and multidimensional low-level static features are combined in chronological order to form a multivariate time series. The dynamic environmental evolution pattern is extracted by multivariate time series and noise interference is filtered out to form a dynamic environmental code, which includes time series feature values, spectral feature values, acoustic feature values ​​and prediction feature values.

4. The home AI weather lighting system according to claim 3, characterized in that, The hypergraph fusion decision signal analysis is as follows: User-defined function parameters are extracted from user behavior logs. A weather configuration hypergraph is constructed based on these parameters and a structured weather database. The specific node set is defined as environmental measurement nodes, user configuration nodes, and derived feature nodes. The hyperedge set is generated using a hypergraph clustering algorithm. The importance weight of each node to each hyperedge is calculated. Based on the importance weight of each node to each hyperedge, the hyperedge information containing all nodes is aggregated to update the feature vector of each node. A learnable global query vector is introduced to interact with the updated feature vector of each node to calculate the importance weight of each node to the final global decision. The hypergraph fusion decision signal is obtained by weighted summation of the updated feature vector of each node and the importance weight of each node to the final global decision.

5. The home AI weather lighting system according to claim 1, characterized in that, The Pareto optimal solution set analysis is as follows: Set a dynamic weight for the timestamp, create a time-varying visual comfort function based on the dynamic weight of the timestamp, obtain the visual comfort level by passing the time-varying visual comfort function based on the temporal feature value and spectral feature value of the environment dynamic coding, and calculate the comfort index by combining the thermal comfort index and the visual comfort index. Based on the predicted feature value energy consumption index of environmental dynamic coding, the energy consumption temperature difference of air conditioning is analyzed to obtain the energy consumption index. Based on the environmental realism index, comfort index and energy consumption index, a multi-objective optimization function is constructed, and equipment physical constraints and user and scenario constraints are applied to obtain the Pareto optimal solution set.

6. The home AI weather lighting system according to claim 5, characterized in that, The environmental realism index and thermal comfort index are analyzed as follows: A spectral generation function is constructed based on the Gaussian function. The spectral realism is obtained by using the spectral feature value based on the dynamic environmental coding. The sound field realism is obtained by calculating the cosine similarity between the acoustic feature value of the generated indoor sound and the acoustic feature value of the dynamic environmental coding. The environmental realism index is obtained by calculating the spectral realism and the sound field realism. Based on the time-series and predicted feature values ​​of the environment dynamic coding, thermal comfort is output through the thermal comfort objective function. The thermal comfort index is obtained by normalizing the deviation between the predicted average voting of the set temperature and the target temperature using the predicted average voting prediction model.

7. The home AI weather lighting system according to claim 1, characterized in that, The optical-acoustic joint control strategy is analyzed as follows: Based on the dynamic environmental coding, the acoustic feature values ​​are matched with the corresponding sound source amplitude and phase parameters from the pre-set sound source library according to the sound source category coding. The complex sound pressure value at each point is calculated by the sound field reconstruction algorithm on the discretized indoor space grid points, and the complex sound pressure amplitude values ​​on all grid points are formed into a three-dimensional sound field distribution matrix. The optimal spectral driving vector and the three-dimensional sound field distribution matrix are used to establish the spectral sound field regular multiple loss function. The optical-acoustic joint control strategy is obtained by solving the alternating direction multiplier method.

8. The home AI weather lighting system according to claim 7, characterized in that, The optimal spectral driving vector analysis is as follows: A dataset is constructed by experimentally measuring the output spectra of different driving current combinations. Based on the dataset, a current-to-spectrum mapping analysis is performed using a multivariate regression method to obtain the LED spectral response function. An ideal policy point is defined, and the standardized Euclidean distance from each policy in the Pareto optimal solution set to the ideal policy point is calculated. The policy is selected by taking the minimum value of the standardized Euclidean distance. Based on the policy selected by the decision rule and the LED spectral response function, the optimal spectral driving vector is obtained by optimizing the model through spectral reconstruction.

9. The home AI weather lighting system according to claim 1, characterized in that, The final multi-device collaborative control command is analyzed as follows: Based on the lighting control component in the photoacoustic joint control strategy, a spectral mapping optimization function is constructed to minimize the difference between the actual output spectrum of the LED light source and the target spectrum. The driving current vector and modulation parameter vector of the spectral mapping optimization function are iteratively solved by the projection gradient descent method to obtain the adaptive spectral driving command. The adaptive spectrum drive command, the sound field control component of the photoacoustic joint control strategy, and the temperature and humidity setpoints of the decision rule selection strategy are coordinated and formatted to obtain a preliminary multi-device control signal set. Discrete event scheduling is performed on the initial set of multi-device control signals to obtain the final multi-device collaborative control commands.

10. A home AI weather lighting method, characterized in that, The system is applied to a home AI weather lighting system as described in any one of claims 1-9, comprising the following steps: Outdoor weather data is collected and preprocessed to obtain a structured weather database; Hypergraph fusion decision-making based on structured weather database and user behavior logs yields dynamic environmental coding; Based on dynamic environmental coding, a multi-objective optimization function is constructed to solve the optimal control strategy and generate a Pareto optimal solution set. The spectral acoustic field is co-optimized by Pareto optimal solution set to obtain a photoacoustic joint control strategy; Based on the photoacoustic joint control strategy, multi-modal control signal analysis and collaborative optimization are performed to obtain the final multi-device collaborative control command.

Images