A method and system for adaptive regulation of light intensity for outdoor lights

Abstract

The application relates to the technical field of intelligent lighting control, and discloses a light brightness self-adaptive regulation method and system for an outdoor lamp. The method comprises the following steps: fusing multi-source heterogeneous sensor information to construct a dynamic environment perception scene atlas, and generating a phased light demand blueprint; carrying out noise filtering and confidence fusion on real-time monitoring data to form a high-confidence environment state snapshot; performing multi-target collaborative optimization calculation in a virtual light control space based on the demand blueprint and the environment snapshot to generate preliminary regulation and control instructions; and finally outputting a final brightness control signal after physical constraint checking and smoothing processing. The method realizes active perception and forward-looking regulation of a complex outdoor environment, and optimizes energy-saving effect and visual comfort under the premise of meeting safety lighting.

Description

Technical Field

[0001] This invention relates to the field of intelligent lighting control technology, specifically to a method and system for adaptive adjustment of the brightness of outdoor lights. Background Technology

[0002] Currently, the brightness adjustment of outdoor lighting systems largely relies on a single type of environmental sensor, such as a light sensor. This type of technology typically presets a fixed threshold; when the ambient light intensity is below or above a certain value, it triggers the lights to turn on, off, or perform simple linear brightness adjustments. This control mode, based on single-point, instantaneous data, treats the environment as an isolated, static set of parameters, lacking the ability to perceive and respond to complex dynamic factors such as weather changes, pedestrian and vehicular activity, and seasonal changes. Its decision-making is one-sided and delayed.

[0003] Most existing solutions employ isolated control rules, such as excessively reducing brightness in pursuit of energy conservation or maintaining high brightness continuously for safety. These rules often fail to address multiple interrelated and even conflicting objectives, such as energy saving, safety, visual comfort, and equipment wear and tear. The system lacks a decision-making framework capable of comprehensively and quantitatively evaluating multiple objectives and coordinating and optimizing them at a global level. This results in control outcomes often falling into local optima for a single objective, failing to meet the increasingly refined, human-centered, and systematic demands of intelligent lighting in outdoor scenarios. A method for adaptive lighting adjustment that deeply understands the environmental context and enables multi-objective collaborative decision-making is needed. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for adaptive adjustment of the brightness of outdoor lights, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for adaptive adjustment of the brightness of outdoor lights, the method comprising:

[0006] Receive initial environmental monitoring information from multi-source heterogeneous sensors, perform spatiotemporal calibration and format unification operations on the initial environmental monitoring information, and generate a structured environmental data stream;

[0007] Multidimensional feature mining and context association analysis are performed on the structured environment data stream to construct a dynamic environment perception scene map. Based on the dynamic environment perception scene map, dynamic reasoning of lighting requirements is performed to generate a phased lighting requirement blueprint.

[0008] The raw monitoring stream reported by the real-time heterogeneous sensor network is acquired, and noise filtering, anomaly repair and confidence weighted fusion are performed on the raw monitoring stream to form a high-confidence environmental state snapshot;

[0009] Based on the phased lighting demand blueprint and the high-confidence environmental state snapshot, multi-objective collaborative optimization calculations are performed in the virtual lighting control space to generate a preliminary brightness adjustment instruction set.

[0010] The initial brightness control instruction set is input into an actuator drive pipeline containing a feedback correction mechanism. After physical constraint verification and smooth transition processing, the final light brightness control signal is output.

[0011] Preferably, the step of performing multi-dimensional feature mining and contextual analysis on the structured environmental data stream to construct a dynamic environmental perception scene map, and performing dynamic reasoning of lighting requirements based on the dynamic environmental perception scene map to generate a phased lighting requirement blueprint includes:

[0012] The structured environmental data stream is segmented by a sliding window to obtain continuous time segment data. For each continuous time segment data, features such as light intensity trend, moving object trajectory density, and correlation features of natural astronomical events are extracted to form a multi-dimensional feature set.

[0013] The multidimensional feature set is input into a graph-based feature relationship modeling network. The feature relationship modeling network treats each feature as a node and calculates the statistical dependencies between features as edge weights, thereby constructing a dynamic environmental perception scene map that describes the intrinsic relationship between environmental states.

[0014] Load a baseline lighting pattern that matches the current geographic region and season from the preset environmental knowledge base, perform pattern matching and difference calculation between the baseline lighting pattern and the dynamic environmental perception scene map, and identify the current environment's twilight transition period, nighttime active period, late night silent period and dawn preparation period.

[0015] For each identified period, the expected range of illumination intensity, allowable slope of change, and special triggering conditions are inferred by combining the weights of nodes and edges in the dynamic environment perception scene map, and integrated to form a phased illumination demand blueprint containing multiple periods and their corresponding illumination parameters.

[0016] Preferably, the step of loading a baseline lighting pattern matching the current geographical area and season from a preset environmental knowledge base, performing pattern matching and difference calculation between the baseline lighting pattern and the dynamic environmental perception scene map, and identifying the current environment's twilight transition period, nocturnal active period, late night silent period, and dawn preparation period includes:

[0017] Based on the latitude and longitude coordinates of the outdoor light installation point and the current date, the theoretical sunset time, theoretical sunrise time, and historical light intensity curves for the same period are retrieved from the environmental knowledge base and used as the reference lighting pattern.

[0018] Extract the actual light intensity node value at the current time point, the edge weight of the light intensity change trend over a period of time, and the node value of the trajectory density of moving objects from the dynamic environment perception scene map to form the current actual state vector.

[0019] Calculate the Euclidean distance between the current actual state vector and the reference illumination pattern within the corresponding time window to obtain the pattern difference degree. When the pattern difference degree is lower than the quiet threshold and the illumination intensity is in a decreasing channel, the current environment is determined to be the twilight transition period.

[0020] When the pattern difference is within the active threshold range and the moving object trajectory density node value is higher than the preset density threshold, the current environment is determined to be the nighttime active period; when the pattern difference is lower than the quiet threshold and the moving object trajectory density node value is lower than the preset density threshold, the current environment is determined to be the late-night silent period.

[0021] When the pattern difference is below the quiet threshold and the light intensity is on the rise, the current environment is determined to be in the pre-dawn period.

[0022] Preferably, the step of acquiring the raw monitoring stream reported by the real-time heterogeneous sensor network, and performing noise filtering, outlier repair, and confidence-weighted fusion on the raw monitoring stream to form a high-confidence environmental state snapshot includes:

[0023] Simultaneously receive raw monitoring streams with timestamps reported by light sensors, pyroelectric infrared sensors, and meteorological microsensors;

[0024] A combined filtering algorithm based on median filtering and moving average is applied to the light sensor data in the original monitoring stream to filter out impulse noise and high-frequency disturbances, resulting in a denoised light data sequence.

[0025] Anomaly detection based on logic judgment is performed on the motion signal sequence reported by the pyroelectric infrared sensor. Isolated signal points with too short a duration or too low intensity are interpolated and repaired to obtain a continuous motion state sequence.

[0026] Based on the factory calibration parameters and recent self-diagnostic reports of each sensor, dynamic confidence weights are assigned to the denoised illumination data sequence, continuous motion state sequence, and meteorological micro-sensor data, respectively.

[0027] The dynamic confidence weights are used to perform a weighted average fusion of the denoised illumination data sequence, the continuous motion state sequence, and the meteorological microsensor data to generate a comprehensive, highly reliable environmental state snapshot that includes illumination level, motion presence, and weather conditions.

[0028] Preferably, the step of performing multi-objective collaborative optimization calculations within the virtual lighting control space based on the phased lighting demand blueprint and the high-confidence environmental state snapshot to generate a preliminary brightness adjustment instruction set includes:

[0029] In the virtual lighting control space, an objective function is established for the expected lighting intensity range corresponding to the current period, a constraint condition is established for the allowable value of the slope of change, and an optimization boundary is established for special triggering conditions;

[0030] The illumination level and motion presence values ​​in the high-confidence environmental state snapshot are mapped to the initial state points in the virtual lighting control space;

[0031] With the dual optimization objectives of minimizing energy consumption and maximizing lighting comfort, a gradient descent search algorithm is run within the solution space defined by the constraints and optimization boundaries to find the optimal brightness setpoint trajectory that satisfies the objective function.

[0032] From the trajectory of the optimal brightness setpoint, a series of discrete time-brightness pairs are obtained by sampling at fixed time intervals;

[0033] Each time-brightness pair is encapsulated into a control instruction containing an execution timestamp and a target brightness value. All control instructions are sorted in chronological order to form a preliminary brightness control instruction set.

[0034] Preferably, the step of using minimizing energy consumption and maximizing lighting comfort as dual optimization objectives, and running a gradient descent search algorithm within the solution space defined by the constraints and optimization boundaries to find the optimal brightness setpoint trajectory that satisfies the objective function, includes:

[0035] Initialize a trial brightness variation curve from the current brightness to a point within the range of the expected illumination intensity;

[0036] The instantaneous power consumption of the tentative brightness change curve at each sampling point is calculated, and the integral is used to obtain the total energy consumption estimate. At the same time, the integral of the deviation between the tentative brightness change curve and the ideal comfort curve is calculated to obtain the discomfort estimate.

[0037] The total energy consumption estimate and the discomfort estimate are weighted and summed to obtain the loss function value of the current tentative brightness change curve;

[0038] Calculate the gradient of the loss function with respect to each control point of the tentative brightness variation curve;

[0039] Based on the gradient direction and under the constraints, the coordinates of the control points of the tentative brightness variation curve are fine-tuned to generate a new tentative brightness variation curve.

[0040] The process of iteratively calculating the loss function value, calculating the gradient, and fine-tuning the control point continues until the change in the loss function value is less than the convergence threshold. The final tentative brightness change curve is then used as the optimal brightness setpoint trajectory. The calculation of the instantaneous power consumption of the tentative brightness change curve at each sampling point, and the integration of this value to obtain the total energy consumption estimate, includes:

[0041] Obtain a lookup table of the correspondence between brightness and power consumption of outdoor lighting fixtures. The lookup table stores the instantaneous power values ​​corresponding to different brightness values.

[0042] For each sampling point on the tentative brightness change curve, the instantaneous power value of the sampling point is obtained by querying the brightness power consumption correspondence lookup table according to the brightness value of the sampling point.

[0043] Connect the instantaneous power values ​​of all sampling points to form a curve showing the change of instantaneous power over time;

[0044] Numerical integration is performed on the curve of instantaneous power changing with time over the total time interval covered by the tentative brightness change curve;

[0045] The result of numerical integration is the estimated total energy consumption corresponding to the tentative brightness variation curve.

[0046] Preferably, the step of inputting the preliminary brightness control instruction set into an actuator drive pipeline containing a feedback correction mechanism, and outputting the final light brightness control signal after physical constraint verification and smooth transition processing, includes:

[0047] Read the real-time operating parameters of the current lamp from the outdoor light driver circuit, including the current brightness, temperature and voltage;

[0048] Compare the target brightness value of the first control instruction in the preliminary brightness control instruction set with the current brightness value in the real-time working parameters, and calculate the step change.

[0049] If the step change exceeds the maximum instantaneous change threshold allowed by the lighting hardware, the initial brightness control instruction set is reprogrammed based on the maximum instantaneous change threshold, and intermediate transition instructions are inserted between the instructions to generate a revised brightness control instruction set that conforms to physical constraints.

[0050] The revised brightness control instruction set or the preliminary brightness control instruction set is sent into a linear execution queue, and the target brightness value of each instruction is converted into the corresponding pulse width modulation duty cycle value according to the execution timestamp order.

[0051] In each control cycle, the current pulse width modulation duty cycle value is output as the final light brightness control signal to drive the outdoor light-emitting device.

[0052] Preferably, the step of assigning dynamic confidence weights to the denoised illumination data sequence, continuous motion state sequence, and meteorological microsensor data based on the factory calibration parameters and recent self-diagnostic reports of each sensor includes:

[0053] The factory calibration parameters of the light sensor are queried to obtain its accuracy level and linearity error. The basic confidence coefficient is calculated based on the accuracy level and linearity error.

[0054] The false alarm rate and false negative rate statistics recorded in the recent self-diagnosis report of the pyroelectric infrared sensor are retrieved, and the reliability attenuation factor is calculated based on the false alarm rate and false negative rate.

[0055] The signal strength value and the most recent calibration time of the meteorological microsensor are obtained, and the health status index is calculated based on the signal strength value and the time elapsed since calibration.

[0056] The basic confidence coefficient, reliability decay factor, and state health index are normalized and then multiplied to obtain the dynamic confidence weight corresponding to each sensor data. The dynamic confidence weight corresponding to the illumination sensor data is used for the denoised illumination data sequence, the dynamic confidence weight corresponding to the pyroelectric infrared sensor data is used for the continuous motion state sequence, and the dynamic confidence weight corresponding to the meteorological micro-sensor data is used for the meteorological micro-sensor data.

[0057] Preferably, the step of calculating the Euclidean distance between the current actual state vector and the reference illumination mode within the corresponding time window to obtain the mode difference includes:

[0058] Extract time series data of length N from the current actual state vector, where N is determined by the size of the time window;

[0059] Extract theoretical time series data of the same time window and length as the current moment from the reference illumination pattern;

[0060] The time series data and the theoretical time series data are considered as two points in an N-dimensional space;

[0061] Calculate the difference between the value of each dimension of the time series data and the value of the corresponding dimension of the theoretical time series data;

[0062] Squaring the difference in each dimension;

[0063] Sum the squared differences of all dimensions;

[0064] The scalar value obtained by taking the square root of the summation result is the pattern difference degree.

[0065] Preferably, a light brightness adaptive adjustment system for outdoor lights includes a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, when the processor executes the computer program, it implements the steps of the light brightness adaptive adjustment method for outdoor lights.

[0066] Compared with the prior art, the beneficial effects of the present invention are:

[0067] By performing spatiotemporal calibration and format unification on information from multi-source heterogeneous sensors, a structured environmental data stream is formed. Multidimensional feature mining and contextual analysis are then performed on this data stream to construct a dynamically evolving environmental perception scene map. This map not only integrates real-time data but also reveals the intrinsic connections and evolutionary trends among environmental elements. Based on this map, dynamic reasoning of lighting requirements can generate a forward-looking, phased demand blueprint. This transforms the lighting system from a mechanical response to instantaneous environmental changes into a proactive control mode with scene understanding and trend prediction capabilities. Based on deep identification of different scene patterns, the system can pre-plan and implement lighting strategies that best suit the essential needs of each scene, achieving a leap from passive threshold response to proactive contextual adaptation.

[0068] A virtual lighting control space was created, coupling and mapping a phased blueprint reflecting future needs with a highly reliable fused real-time environmental snapshot within this space. Within this space, multiple control objectives, such as energy saving, safety, comfort, and glare control, are mathematically modeled and used as common constraints and objective functions for collaborative optimization. Solving this through a multi-objective optimization algorithm outputs a preliminary set of brightness control instructions that balances multiple objectives and achieves optimal overall system performance under given real-time conditions and future expectations. Previously scattered, isolated, and even contradictory control rules are unified under a globally optimized mathematical framework for decision-making. The resulting brightness control signal is no longer the result of a simple superposition or game of multiple single rules, but a comprehensive optimal solution calculated precisely at the system level, achieving improved energy efficiency and user experience while ensuring basic lighting needs are met. Attached Figure Description

[0069] Figure 1 This is a schematic diagram illustrating the working principle of the adaptive brightness adjustment method for outdoor lights according to the present invention.

[0070] Figure 2 A flowchart for constructing a scene map and generating a requirements blueprint;

[0071] Figure 3 A bar chart comparing regional thresholds for outdoor lighting environment assessment;

[0072] Figure 4 Energy consumption-comfort scatter plot for optimizing outdoor lighting brightness;

[0073] Figure 5 A flowchart for generating a set of control instructions for multi-objective collaborative optimization. Detailed Implementation

[0074] The technical solutions of 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 within the scope of protection of the present invention.

[0075] Please see Figure 1 This invention provides a method for adaptive brightness adjustment of outdoor lights. The method includes: receiving initial environmental monitoring information from multi-source heterogeneous sensors; performing spatiotemporal calibration and format unification operations on the initial environmental monitoring information to generate a structured environmental data stream; performing multi-dimensional feature mining and contextual correlation analysis on the structured environmental data stream to construct a dynamic environmental perception scene map; performing dynamic inference of illumination demand based on the dynamic environmental perception scene map to generate a phased illumination demand blueprint; acquiring the raw monitoring stream reported by the real-time heterogeneous sensor network; performing noise filtering, outlier repair, and confidence-weighted fusion on the raw monitoring stream to form a high-confidence environmental state snapshot; performing multi-objective collaborative optimization calculations in a virtual illumination control space based on the phased illumination demand blueprint and the high-confidence environmental state snapshot to generate a preliminary brightness control instruction set; inputting the preliminary brightness control instruction set into an actuator drive pipeline containing a feedback correction mechanism; and outputting the final light brightness control signal after physical constraint verification and smooth transition processing.

[0076] Example 1: See Figure 2The structured environmental data stream is segmented using a sliding window to obtain continuous time-segment data. For each continuous time-segment, features such as light intensity trend, moving object trajectory density, and correlation with natural astronomical events are extracted to form a multi-dimensional feature set. This multi-dimensional feature set is input into a graph-based feature relationship modeling network. This network treats each feature as a node and calculates the statistical dependencies between features as edge weights, thereby constructing a dynamic environmental perception scene map describing the intrinsic relationships of the environmental state. A baseline lighting pattern matching the current geographic region and season is loaded from a pre-set environmental knowledge base. This baseline lighting pattern is then matched with the dynamic environmental perception scene map using pattern matching and difference calculations to identify the current environmental state as the twilight transition period, nighttime active period, late-night quiet period, and dawn preparation period. For each identified period, the expected light intensity range, allowable slope of change, and special triggering conditions are inferred by combining the weights of nodes and edges in the dynamic environmental perception scene map. This is then integrated to form a phased lighting demand blueprint containing multiple periods and their corresponding lighting parameters.

[0077] In a specific implementation, when performing sliding window segmentation on the structured environmental data stream to obtain continuous time-segment data, the width of the sliding window is set to five minutes and slides forward in one-minute increments. Each continuous time-segment data contains three hundred sampling points arranged in chronological order. Illumination intensity trend features are extracted for each continuous time-segment data. This involves calculating the first derivative sequence of the illumination intensity readings within the segment and statistically analyzing the distribution ratio of positive and negative values. Moving object trajectory density features are extracted for each continuous time-segment data. This involves statistically analyzing the number of grid points and trigger frequency of pyroelectric infrared sensors within the segment. Natural astronomical event correlation features are extracted for each continuous time-segment data. This involves performing correlation analysis between the illumination intensity curve within the segment and the theoretical solar altitude angle curve calculated based on latitude, longitude, and time. All extracted features are collected and organized by time segment to form a multi-dimensional feature set containing illumination intensity trend values, moving object trajectory density values, and natural astronomical event correlation coefficients.

[0078] In some embodiments, when a multidimensional feature set is input into a graph-based feature relationship modeling network to construct a dynamic environment-aware scene atlas, the feature relationship modeling network maps three features—light intensity trend, moving object trajectory density, and correlation of natural and astronomical events—to three independent nodes in the atlas. The feature relationship modeling network calculates the statistical dependencies between features as edge weights. The method for calculating the edge weight between the light intensity trend node and the moving object trajectory density node is to calculate the absolute value of the Pearson correlation coefficient between the two feature sequences over the past hour. The method for calculating the edge weight between the light intensity trend node and the natural and astronomical event correlation node is to calculate the mutual information value of the two feature sequences. The method for calculating the edge weight between the moving object trajectory density node and the natural and astronomical event correlation node is to use distance-based covariance calculation. All calculated edge weights and corresponding nodes together constitute a dynamic environment-aware scene atlas describing the intrinsic relationships of the environmental state.

[0079] In a specific implementation, when loading a baseline illumination pattern matching the current geographic region and season from a preset environmental knowledge base, the environmental knowledge base stores historical illumination statistics indexed by latitude and longitude grids and months. The loading process involves retrieving the average illumination intensity curve of the corresponding grid point from dusk to dawn each day of that month, based on the latitude and longitude coordinates of the outdoor light installation point and the month of the current date, as the baseline illumination pattern. When performing pattern matching and difference calculation between the baseline illumination pattern and the dynamic environmental perception scene map, pattern matching aligns the actual illumination intensity sequence within each 15-minute window before and after the current time with the curves of the same time period in the baseline illumination pattern. The difference calculation uses Euclidean distance as a metric, and its formula is:

[0080]

[0081] Where: symbol Represents the calculated scalar value of pattern dissimilarity, with the symbol... Represents the total number of sampling points within the time window, with the sign... The actual illumination intensity sequence extracted from the dynamic environment-aware scene map is in the first... The value of each sampling point, sign This represents the theoretical illumination intensity sequence extracted from the reference illumination mode in the 1st... The values ​​of each sampling point.

[0082] It is understandable that identifying the current environmental phase—dusk transition, nocturnal activity, late-night silence, and dawn preparation—is based on a logical judgment made through continuous pattern dissimilarity calculations and the states of map nodes. When the calculated pattern dissimilarity... When the light intensity remains below the preset quiet threshold and the light intensity trend node shows a predominantly negative slope, the system determines that the current environment is in the twilight transition period. (When the mode difference...) When the value of the moving object trajectory density node rises and enters the preset activity threshold range, and the value of the moving object trajectory density node is continuously higher than the preset density threshold, the system determines that the current environment is a nighttime active period. When the pattern difference... When the temperature drops below the quiet threshold and the value of the moving object trajectory density nodes is consistently below the preset density threshold, the system determines that the current environment is a quiet period at night. (When the mode difference...) When the light intensity trend node remains below the quiet threshold and the positive slope begins to dominate, the system determines that the current environment is in the pre-dawn period.

[0083] Optionally, when inferring the corresponding illumination parameters for each identified period and integrating them to form a phased illumination demand blueprint, the inference process incorporates the weights of nodes and edges in the dynamic environment-aware scene graph. For nighttime active periods, the lower limit of the expected illumination intensity range is determined by a linear mapping of the values ​​of the moving object trajectory density nodes, while the upper limit is determined by a fixed value of the safety lighting standard. The allowable slope of change is dynamically adjusted based on the edge weights between the illumination intensity trend nodes and the moving object trajectory density nodes; higher edge weights allow for a smoother slope of brightness change. A special trigger condition is defined as when the value of the natural astronomical event correlation node falls below a certain threshold, indicating a sudden change in weather, requiring immediate activation of the backup lighting mode.

[0084] Example 2: Based on the latitude and longitude coordinates of the outdoor light installation point and the current date, the theoretical sunset time, theoretical sunrise time, and historical light intensity curves for the same time period are retrieved from the environmental knowledge base as the baseline lighting pattern. The actual light intensity node value at the current time point, the edge weight of the light intensity change trend over a past period, and the moving object trajectory density node value are extracted from the dynamic environment perception scene map to form the current actual state vector. The Euclidean distance between the current actual state vector and the baseline lighting pattern within the corresponding time window is calculated to obtain the pattern difference degree. When the pattern difference degree is below the quiet threshold and the light intensity is in a decreasing channel, the current environment is determined to be in the twilight transition period. When the pattern difference degree is in the active threshold range and the moving object trajectory density node value is above the preset density threshold, the current environment is determined to be in the nighttime active period; when the pattern difference degree is below the quiet threshold and the moving object trajectory density node value is below the preset density threshold, the current environment is determined to be in the late night silent period. When the pattern difference degree is below the quiet threshold and the light intensity is in an increasing channel, the current environment is determined to be in the dawn preparation period. To calculate the Euclidean distance, time-series data of length N is extracted from the current actual state vector, where N is determined by the time window size. Theoretical time-series data of the same length and time window as the current moment is extracted from the reference illumination pattern. The time-series data and the theoretical time-series data are considered as two points in an N-dimensional space. The difference between the value of each dimension of the time-series data and the corresponding dimension value of the theoretical time-series data is calculated. The difference in each dimension is squared. The squared differences of all dimensions are summed. The square root of the summation result is taken, and the resulting scalar value is the mode dissimilarity.

[0085] In a specific implementation, when retrieving a baseline illumination pattern from the environmental knowledge base based on the latitude and longitude coordinates of the outdoor light installation point and the current date, the latitude and longitude coordinates of the outdoor light installation point are stored in degrees, for example, 116.407 degrees east longitude, 39.904 degrees north latitude, and the current date is October 15, 2023. The environmental knowledge base is a database containing theoretical astronomical data and historical statistical information. The retrieval process calculates the theoretical sunset time of 19:12 and the theoretical sunrise time of 06:28 for the current location based on the latitude and longitude coordinates and the date. At the same time, it retrieves the average hourly illumination intensity readings for the coordinate point on the night of October 15th of each year over the past five years and fits them into a smooth curve. The theoretical sunset time, theoretical sunrise time, and the historical illumination intensity curves for the same time period together constitute the baseline illumination pattern for comparison. When extracting information from the dynamic environment perception scene map to form the current actual state vector, the extraction operation is executed at 20:00 system time. The actual light intensity node value at the current time point in the extracted dynamic environment perception scene map is 15 lux, the edge weight of the light intensity change trend in the past 30 minutes in the extracted dynamic environment perception scene map is -0.8, and the node value of the trajectory density of moving objects in the extracted dynamic environment perception scene map is 0.3. These three values ​​are organized into a three-dimensional vector [15, -0.8, 0.3] as the current actual state vector.

[0086] In some embodiments, the mode difference is obtained by calculating the Euclidean distance between the current actual state vector and the reference illumination mode within a corresponding time window, where the corresponding time window is set to ten minutes before and after the current moment. Twenty historical readings of actual illumination intensity node values ​​at minute intervals over the past twenty minutes are extracted from the current actual state vector to form actual illumination intensity time series data. Twenty theoretical illumination intensity values ​​within the same time window (20:00 AM) are extracted from the historical illumination intensity curves of the reference illumination mode to form theoretical time series data. The mode difference is obtained by calculating the Euclidean distance between the two sequences in twenty-dimensional space, using the following formula:

[0087]

[0088] Where: symbol Represents the final calculated scalar value of pattern dissimilarity, with the symbol... Represents the total number of sampling points within the time window, with a value of twenty, symbol The actual light intensity time series data extracted from the dynamic environment-aware scene map is represented in the first... The value of each sampling point, sign This represents the theoretical time series data extracted from the baseline illumination model in the 1st... The values ​​of each sampling point.

[0089] It is understandable that the logic for determining the current environmental period relies on the pattern difference degree obtained through continuous calculation. And the comparison results with the reference lighting mode. The quiet threshold was set to a fixed value of 5, when the mode difference... The calculated result is 3.2, which is lower than the quiet threshold of 5. Simultaneously, based on the light intensity change trend edge weight of -0.8 in the current actual state vector, the light intensity is determined to be in a stable decreasing channel. Therefore, the system identifies the current environment as a twilight transition period. The active threshold range is set as a closed interval [6,15]. When the mode difference... When the calculated value is 8.5 and this value is within the active threshold range [6, 15], and the current actual state vector's moving object trajectory density node value of 0.7 is higher than the preset density threshold of 0.5, the system determines that the current environment is a nighttime active period. When the mode difference... When the calculated value is 2.1 and this value is lower than the quiet threshold of 5, and the value of the moving object trajectory density node in the current actual state vector is 0.1 and lower than the preset density threshold of 0.5, the system determines that the current environment is a late-night quiet period. When the mode difference degree... The calculated result is 3.8, which is lower than the quiet threshold of 5. At the same time, judging from the light intensity change trend edge weight of 0.6 in the current actual state vector, when the light intensity is in the upward channel, the system determines that the current environment is the dawn preparation period.

[0090] Optionally, the specific values ​​for the quiet and active threshold ranges can be adjusted based on geographical characteristics. For urban centers with heavy light pollution, the quiet threshold can be increased to 8 to accommodate higher background light fluctuations. The preset density threshold is related to the pedestrian and vehicle flow statistics of the specific monitoring area. The preset density threshold for residential entrance areas can be set to 0.3, and for urban park walkways, it can be set to 0.6. The criterion for determining whether the light intensity is in a decreasing or increasing channel is the sign of the edge weights representing the light intensity change trend in the current actual state vector. Negative values ​​represent a decreasing channel, and positive values ​​represent an increasing channel. The absolute value of the edge weights reflects the intensity of the change trend.

[0091] See Figure 3This is a bar chart comparing regional thresholds for outdoor lighting environment determination. It is used to show the differences in the configuration of "quiet threshold" and "density threshold" under different regional types. It is one of the core reference data in the environmental state inference stage of the outdoor lighting brightness adaptive adjustment system. City center (8) > city park (6) > residential area (5) > suburban road (4), which shows that the light pollution in the city center is more severe and a higher threshold is needed to accommodate background fluctuations. City park (approximately 0.6) > city center (approximately 0.5) > residential area (approximately 0.3) > suburban road (approximately 0.2), which matches the characteristics of pedestrian and vehicle traffic density in the region. It is used for regional parameter configuration of outdoor lighting system to help the system set reasonable environmental determination thresholds for different scenarios and ensure the accuracy and adaptability of lighting brightness adjustment.

[0092] Example 3: Synchronously receive raw monitoring streams with timestamps reported by a light sensor, a pyroelectric infrared sensor, and a meteorological microsensor. Apply a combined filtering algorithm based on median filtering and moving average to the light sensor data in the raw monitoring stream to filter out impulse noise and high-frequency disturbances, obtaining a denoised light data sequence. Perform anomaly detection based on logic judgment on the motion signal sequence reported by the pyroelectric infrared sensor, interpolating and repairing isolated signal points with excessively short durations or low intensities to obtain a continuous motion state sequence. Based on the factory calibration parameters and recent self-diagnostic reports of each sensor, assign dynamic confidence weights to the denoised light data sequence, continuous motion state sequence, and meteorological microsensor data respectively. Use the dynamic confidence weights to perform a weighted average fusion of the denoised light data sequence, continuous motion state sequence, and meteorological microsensor data to generate a comprehensive, high-confidence environmental state snapshot that includes light level, motion presence, and weather conditions. When assigning dynamic confidence weights, query the factory calibration parameters of the light sensor to obtain its accuracy level and linearity error, and calculate the basic confidence coefficient based on the accuracy level and linearity error. The false alarm rate and false negative rate statistics recorded in the recent self-diagnostic reports of the pyroelectric infrared sensor were retrieved, and the reliability decay factor was calculated based on the false alarm rate and false negative rate. The signal strength value and the most recent calibration time of the meteorological microsensor were obtained, and the state health index was calculated based on the signal strength value and the time elapsed since calibration. The basic confidence coefficient, reliability decay factor, and state health index were normalized and then multiplied to obtain the dynamic confidence weight corresponding to each sensor data. The dynamic confidence weight corresponding to the illumination sensor data was used for denoising the illumination data sequence, the dynamic confidence weight corresponding to the pyroelectric infrared sensor data was used for the continuous motion state sequence, and the dynamic confidence weight corresponding to the meteorological microsensor data was used for the meteorological microsensor data.

[0093] In a specific implementation, raw monitoring streams with timestamps are simultaneously received from a light sensor, a pyroelectric infrared sensor, and a meteorological microsensor. These raw monitoring streams are updated once per second, with timestamps accurate to milliseconds. For example, at 20:00:00.000, the light sensor reports a raw light intensity reading of 12.5 lux, the pyroelectric infrared sensor reports a motion signal of binary 0, and the meteorological microsensor reports a current ambient temperature reading of 18 degrees Celsius. A combined filtering algorithm based on median filtering and moving average is applied to the light sensor data in the raw monitoring stream. The combined filtering algorithm first sorts the light intensity values ​​of the five most recent sampling points by median and takes the median value. Then, a moving average with a window size of three is calculated on the median-filtered sequence. After filtering out impulse noise and high-frequency disturbances, the resulting light sensor data is 12.3 lux. This data sequence constitutes the denoised light data sequence. Anomaly detection based on logic judgment is performed on the motion signal sequence reported by the pyroelectric infrared sensor. The logic judgment rule is to check whether the duration of a single trigger signal is less than 100 milliseconds or whether the signal strength is less than 1.0 volt. An isolated signal point with a duration of 50 milliseconds and an intensity of 0.8 volts is detected. The intensity values ​​of the two normal signal points before and after this isolated signal point are linearly interpolated, and the interpolated 0.2 volts are used to replace the original value. After repair, a continuous motion state sequence without abrupt spikes is obtained.

[0094] In some embodiments, dynamic confidence weights are assigned based on the factory calibration parameters and recent self-diagnostic reports of each sensor. The factory calibration parameters of the illumination sensor are queried, and its accuracy level is found to be 0.5 and its linearity error to be ±0.1%. The basic confidence coefficient is calculated to be 0.95 based on the accuracy level and linearity error. The false alarm rate of the pyroelectric infrared sensor is retrieved from the recent self-diagnostic report, showing a false alarm rate of 2% and a false alarm rate of 1%. The reliability attenuation factor is calculated to be 0.97 based on the false alarm and false alarm rates. The signal strength value of the meteorological microsensor is obtained as 80% of its full scale, and the most recent calibration time is 30 days ago. The health status index is calculated to be 0.90 based on the signal strength value and the time since calibration. The basic confidence coefficient, reliability attenuation factor, and health status index are normalized and then multiplied. The formula for calculating the dynamic confidence weight corresponding to the illumination sensor data is as follows:

[0095]

[0096] Where: symbol The dynamic confidence weights corresponding to the illumination sensor data are represented by the symbol. The baseline confidence coefficient for the light sensor is 0.95, and the symbol is... The reliability degradation factor of 1.00 represents the light sensor's reliability. The health index of the light sensor is 1.00. Similarly, the base confidence coefficient, reliability decay factor, and health index of the pyroelectric infrared sensor are 0.90, 0.97, and 1.00, respectively. Multiplying these together yields a dynamic confidence weight of 0.873 for the pyroelectric infrared sensor data. The base confidence coefficient, reliability decay factor, and health index of the meteorological microsensor are 0.85, 1.00, and 0.90, respectively. Multiplying these together yields a dynamic confidence weight of 0.765 for the meteorological microsensor data.

[0097] It is understandable that dynamic confidence weights are used to perform a weighted average fusion of the denoised illumination data sequence, the continuous motion state sequence, and the meteorological microsensor data. The representative value of the denoised illumination data sequence (12.3 lux), the representative value of the continuous motion state sequence (0.2 volts), and the representative value of the meteorological microsensor data (18 degrees Celsius) are all converted into dimensionless normalized scale values. The formula for weighted average fusion is that the comprehensive value equals the sum of the products of each normalized data point and its corresponding dynamic confidence weight, divided by the sum of all weights. After weighted average fusion, a comprehensive high-confidence environmental state snapshot is generated, which includes illumination level, motion presence, and weather conditions. The high-confidence environmental state snapshot is encapsulated in vector form, for example, the illumination level component is 0.45, the motion presence component is 0.12, and the weather condition component is 0.70.

[0098] Optionally, the window size of the moving average in the combined filtering algorithm can be dynamically adjusted according to the environmental noise level. When electrical noise interference is enhanced due to thunderstorms, the moving average window size can be adjusted from three to five. The logical judgment threshold in the abnormal point repair of the pyroelectric infrared sensor can also be fine-tuned according to the installation environment. In areas where small animals frequently appear, the duration threshold can be relaxed from 100 milliseconds to 200 milliseconds to reduce false repairs. The update frequency of the reliability decay factor in the confidence weight calculation is consistent with the generation cycle of the self-diagnostic report. The self-diagnostic report is generated once every 24 hours, and the reliability decay factor is updated accordingly once a day to ensure that the dynamic confidence weight can reflect the latest working status of the sensor.

[0099] See Figure 4This is a scatter plot of energy consumption and comfort for outdoor lighting brightness optimization. It illustrates the relationship between "energy consumption" and "comfort" during brightness adjustment and is one of the core analytical tools in the brightness control optimization stage of an adaptive lighting system. The relationship between energy consumption and comfort is non-monotonic: initially (dark dots), comfort fluctuates greatly when energy consumption is low; with iteration (light dots), comfort is generally higher (70%-90%) in the medium energy consumption range (30-60W); the high comfort concentration area: when energy consumption is 30-50W, comfort is mostly maintained above 80%, representing the optimal range of "low energy consumption - high comfort"; iterative trend: in the later stages of optimization (light yellow dots), the points are more concentrated in the high comfort area, indicating that the optimization process has effectively converged. This is used for multi-objective optimization analysis of brightness adjustment strategies, helping to screen the optimal brightness parameters that achieve "low energy consumption and high comfort," while tracking the convergence effect of optimization iterations, providing a basis for the final output of the lighting control signal.

[0100] Example 4: See Figure 5 In the virtual lighting control space, an objective function is established for the expected light intensity range corresponding to the current period, constraints are established for the allowable slope of change, and optimization boundaries are established for special triggering conditions. The light level and motion presence values ​​in the high-confidence environmental state snapshot are mapped to initial state points in the virtual lighting control space. With the dual optimization objectives of minimizing energy consumption and maximizing lighting comfort, a gradient descent search algorithm is run within the solution space defined by the constraints and optimization boundaries to find the optimal brightness setpoint trajectory that satisfies the objective function. From the optimal brightness setpoint trajectory, samples are taken at fixed time intervals to obtain a series of discrete time-brightness pairs. Each time-brightness pair is encapsulated as a control instruction containing an execution timestamp and a target brightness value. All control instructions are sorted chronologically to form a preliminary brightness control instruction set. When running the gradient descent search algorithm, an exploratory brightness change curve from the current brightness to a point within the expected light intensity range is initialized.

[0101] The instantaneous power consumption of the tentative brightness variation curve at each sampling point is calculated, and its integration yields the total energy consumption estimate. Simultaneously, the integral of the deviation between the tentative brightness variation curve and the ideal comfort curve is calculated to obtain the discomfort estimate. The total energy consumption estimate and the discomfort estimate are weighted and summed to obtain the loss function value of the current tentative brightness variation curve. The gradient of the loss function relative to each control point of the tentative brightness variation curve is calculated. Based on the gradient direction, and under constraints, the coordinates of the control points of the tentative brightness variation curve are fine-tuned to generate a new tentative brightness variation curve. The process of calculating the loss function value, calculating the gradient, and fine-tuning the control points is iteratively executed until the change in the loss function value is less than the convergence threshold. The final tentative brightness variation curve is then used as the optimal brightness setpoint trajectory. When calculating the total energy consumption estimate, a lookup table corresponding to the brightness and power consumption of outdoor lighting fixtures is obtained. This lookup table stores the instantaneous power values ​​corresponding to different brightness values. For each sampling point on the tentative brightness variation curve, the instantaneous power value of that sampling point is obtained by querying the lookup table based on the brightness value of that sampling point. Connect the instantaneous power values ​​of all sampling points to form a curve showing the instantaneous power change over time. Perform numerical integration on the curve over the total time interval covered by the tentative brightness variation curve. The result of the numerical integration is the estimated total energy consumption corresponding to the tentative brightness variation curve.

[0102] In a specific implementation, multi-objective collaborative optimization calculations are performed within a virtual lighting control space based on a phased lighting demand blueprint and a high-confidence environmental state snapshot. The virtual lighting control space is a two-dimensional planning space with time as the horizontal axis and lighting intensity as the vertical axis. An objective function is established for the expected lighting intensity range corresponding to the current period, set to make the actual brightness trajectory as close as possible to the median of the expected lighting intensity range. Constraints are established for the allowable slope of change, stipulating that the absolute value of the rate of change of brightness between any adjacent sampling points must not exceed 5 lux per second. An optimization boundary is established for special triggering conditions; when the motion presence value in the high-confidence environmental state snapshot exceeds 0.8, the optimization boundary temporarily raises the lower limit of the expected lighting intensity range by 20%. The illumination level value of 0.45 in the high-confidence environment state snapshot is mapped to a point on the vertical axis of the virtual lighting control space, and the motion existence value of 0.12 in the high-confidence environment state snapshot is mapped to a time offset on the horizontal axis of the virtual lighting control space. Together, these coordinates determine the initial state point coordinates in the virtual lighting control space as (current time + 0.12 seconds, 0.45 seconds). (Maximum brightness).

[0103] In some embodiments, a gradient descent search algorithm is run with the dual optimization objectives of minimizing energy consumption and maximizing lighting comfort. The algorithm initializes a tentative brightness variation curve from the current brightness to a point within the expected lighting intensity range. The tentative brightness variation curve is represented by a spline curve defined by ten control points. The instantaneous power consumption of the tentative brightness variation curve at each sampling point is calculated. For each sampling point on the tentative brightness variation curve, the instantaneous power value is obtained by looking up the brightness-power consumption correspondence table of the outdoor lighting fixture based on the brightness value of the sampling point, as shown in Table 1.

[0104] Table 1. Lookup Table for the Correspondence between Brightness and Power Consumption of Outdoor Lighting Fixtures

[0105]

[0106] Connecting the instantaneous power values ​​of all sampling points forms a curve showing the instantaneous power change over time. Numerical integration is performed on this curve over the total time interval covered by the tentative brightness change curve; the result of this integration is the estimated total energy consumption corresponding to the tentative brightness change curve. Simultaneously, the integral of the deviation between the tentative brightness change curve and the ideal comfort curve is calculated to obtain an estimated discomfort level. The ideal comfort curve is a reference line where brightness changes linearly over time. A weighted sum of the estimated total energy consumption and the estimated discomfort level is then used as the loss function value for the current tentative brightness change curve. The formula for calculating the loss function is:

[0107]

[0108] Where: symbol Represents the loss function value, symbol Represents the estimated total energy consumption, symbol Represents the estimated discomfort level, symbol The weighting factor for the energy consumption item is 0.7, and the symbol is... The weighting factor for the comfort item is 0.3.

[0109] It is understandable that the gradient of the loss function relative to each control point of the tentative brightness change curve is calculated using the numerical difference method, and the gradient direction indicates the direction in which the loss function value decreases the fastest. In a specific implementation, the gradient of the loss function relative to each control point of the tentative brightness change curve is calculated using the numerical difference method. This method applies a small brightness value perturbation to each control point on the curve, for example, increasing the brightness value of the control point by a small increment such as 0.1 lux, and then recalculates the loss function value. The difference between the perturbed loss function value and the original loss function value is obtained, and this difference is divided by the perturbation increment to obtain an approximate estimate of the gradient at that control point. The gradient direction is determined by the sign and magnitude of the gradient value at each control point, indicating the direction in which the loss function value decreases the fastest. The system fine-tunes the coordinates of the control points based on this direction to gradually approach the optimal solution. Samples are taken from the trajectory of the optimal brightness setpoint at fixed time intervals, with the fixed time interval set to ten seconds, to obtain a series of discrete time-brightness pairs, such as (20:01:00, 50 lux) and (20:01:10, 55 lux). Each time-brightness pair is encapsulated into a control instruction containing an execution timestamp and a target brightness value. All control instructions are sorted in chronological order to form a preliminary brightness control instruction set.

[0110] Optionally, a lookup table for the correspondence between brightness and power consumption of outdoor lighting fixtures can be obtained through experimental calibration and loaded during system initialization. The lookup table supports linear interpolation to obtain the power corresponding to brightness values ​​not directly defined in the table. The weight coefficients of the energy consumption item and the comfort item can be dynamically configured according to management strategies. During periods emphasizing energy conservation, the weight coefficient of the energy consumption item can be set to 0.9, while during periods emphasizing pedestrian safety, the weight coefficient of the comfort item can be set to 0.7. The setting of the convergence threshold affects the balance between optimization accuracy and computation time. Setting the convergence threshold to 0.05 will result in more iterations to seek a better solution, while setting it to 0.5 will produce results faster but with reduced accuracy. During the gradient descent search algorithm, the constraints are always enforced. Any control point adjustment that causes a rate of change exceeding 5 lux per second will be rejected and projected along the constraint boundary.

[0111] Example 5: Read the real-time operating parameters of the current luminaire from the outdoor light driver circuit, including the current brightness, temperature, and voltage. Compare the target brightness value of the first control instruction in the preliminary brightness control instruction set with the current brightness value in the real-time operating parameters to calculate the step change. If the step change exceeds the maximum instantaneous change threshold allowed by the luminaire hardware, the preliminary brightness control instruction set is reprogrammed according to the maximum instantaneous change threshold, inserting intermediate transition instructions between instructions to generate a revised brightness control instruction set that conforms to physical constraints. Send the revised brightness control instruction set or the preliminary brightness control instruction set into a linear execution queue, and convert the target brightness value of each instruction into the corresponding pulse width modulation duty cycle value according to the execution timestamp order. In each control cycle, output the current pulse width modulation duty cycle value as the final light brightness control signal driving the outdoor light-emitting device.

[0112] In a specific implementation, a preliminary brightness control command set is input to an actuator drive pipeline containing a feedback correction mechanism. At the pipeline inlet, real-time operating parameters of the current luminaire are read from the outdoor light driver circuit. These real-time operating parameters include a current brightness value of 60 lux obtained through a current sensor and brightness feedback circuit, a luminaire heatsink temperature of 45 degrees Celsius obtained through a temperature sensor, and an input voltage value of 220 volts obtained through a voltage sampling circuit. The target brightness value of 100 lux for the first control command in the preliminary brightness control command set is compared with the current brightness value of 60 lux in the real-time operating parameters. A step change is calculated by subtracting the current brightness value from the target brightness value, resulting in 40 lux. The maximum allowable instantaneous change threshold for the outdoor light driver circuit is pre-stored in a configuration file and is set to 30 lux per second. Since the calculated step change of 40 lux exceeds the maximum instantaneous change threshold of 30 lux per second, the actuator drive pipeline reprograms the preliminary brightness control command set based on the maximum instantaneous change threshold.

[0113] In some embodiments, replanning is performed based on a maximum instantaneous change threshold, and intermediate transition instructions are inserted between commands. The replanning algorithm calculates the maximum allowable brightness change of 3 lux per control cycle based on the maximum instantaneous change threshold of 30 lux per second and a system control cycle of 100 milliseconds. The initial command requires a change of 40 lux to go from 60 lux to 100 lux, therefore at least thirteen intermediate transition brightness points need to be inserted to meet the rate of change constraint. The inserted intermediate transition instructions are evenly distributed between the initial brightness and the target brightness. For example, the first inserted intermediate transition instruction has a target brightness of 63 lux and an execution timestamp delayed by 100 milliseconds from the original timestamp; the second intermediate transition instruction has a target brightness of 66 lux and an execution timestamp delayed by another 100 milliseconds, and so on. All original commands and newly added intermediate transition instructions are reordered chronologically to generate a revised brightness control instruction set that conforms to physical constraints. The revised brightness control instruction set contains a series of closely arranged time-brightness pairs.

[0114] It is understandable that the revised brightness control instruction set is sent to a linear execution queue for signal conversion. The linear execution queue is a first-in, first-out buffer that stores control instructions in the order of their execution timestamps. The mapping relationship between the target brightness value of each control instruction and the corresponding pulse width modulation duty cycle value is determined by the characteristics of the driver chip, and the conversion formula is as follows:

[0115]

[0116] Where: symbol This represents the required pulse width modulation duty cycle value, with the sign... Represents the target brightness value in the control command, symbol This represents the maximum luminance value achievable by the outdoor lighting fixture's light-emitting device, 150 lux. For example, for a control command with a target luminance value of 75 lux, the calculated pulse width modulation duty cycle value... It is 50%.

[0117] In each control cycle, the system clock trigger signal is output, and the linear execution queue checks whether the execution timestamp of the instruction at the head of the queue has reached or exceeded the current time. When the condition is met, the queue pops the instruction and calculates its target brightness value into a pulse width modulation duty cycle value using a conversion formula. The hardware driver circuit applies this pulse width modulation duty cycle value to the control terminal of the light-emitting device. This applied value is the final light brightness control signal that drives the outdoor light-emitting device.

[0118] Optionally, the maximum instantaneous change threshold can be dynamically adjusted based on the lamp model and its age. For aging lamps, the maximum instantaneous change threshold can be lowered to 20 lux per second to protect the circuit. The depth setting of the linear execution queue matches the number of instructions after replanning, ensuring that all instructions to be executed can be accommodated before the next planning cycle arrives. The maximum brightness value in the pulse width modulation duty cycle conversion formula. The light intensity meter can be periodically calibrated and updated to compensate for the light decay of the light-emitting device. In scenarios where the maximum instantaneous change threshold is not exceeded, the initial brightness control instruction set does not need to be reprogrammed and can directly enter the linear execution queue for conversion and output.

[0119] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0120] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for adaptive regulation of light intensity of an outdoor light, characterized in that, The method comprises: receiving initial environment monitoring information from a plurality of heterogeneous sensors, performing spatio-temporal calibration and format unification operations on the initial environment monitoring information to generate a structured environment data stream; performing multi-dimensional feature mining and context association analysis on the structured environment data stream to construct a dynamic environment perception scene graph, and performing dynamic reasoning of lighting requirements based on the dynamic environment perception scene graph to generate a phased lighting requirement blueprint; obtaining original monitoring streams reported by a real-time heterogeneous sensor network, performing noise filtering, abnormal point repair and confidence weighted fusion on the original monitoring streams to form a high-confidence environment state snapshot; based on the phased lighting requirement blueprint and the high-confidence environment state snapshot, performing multi-objective collaborative optimization calculation in a virtual lighting control space to generate a preliminary brightness control instruction set; inputting the preliminary brightness control instruction set into an executor driving pipeline comprising a feedback correction mechanism, and outputting a final light brightness control signal after physical constraint checking and smooth transition processing; the multi-dimensional feature mining and context association analysis on the structured environment data stream to construct a dynamic environment perception scene graph, and performing dynamic reasoning of lighting requirements based on the dynamic environment perception scene graph to generate a phased lighting requirement blueprint, comprising: performing sliding window segmentation on the structured environment data stream to obtain continuous time segment data, and performing lighting intensity trend, moving object trajectory density and natural astronomical event correlation feature extraction on each continuous time segment data to form a multi-dimensional feature set; inputting the multi-dimensional feature set into a feature relationship modeling network based on a graph structure, taking each feature as a node, calculating the statistical dependence relationship between features as edge weights, and thus constructing a dynamic environment perception scene graph describing the internal relationship of the environment state; loading a reference lighting mode matching the current geographical area and season from a preset environment knowledge base, performing mode matching and difference calculation on the reference lighting mode and the dynamic environment perception scene graph, and identifying the current environment in the twilight transition period, the night active period, the deep night silent period and the dawn preparation period; for each identified period, combining the weights of nodes and edges in the dynamic environment perception scene graph, reasoning the expected lighting intensity range, change slope tolerance value and special trigger condition corresponding to the period, and integrating to form a phased lighting requirement blueprint containing a plurality of periods and corresponding lighting parameters; the multi-objective collaborative optimization calculation in the virtual lighting control space based on the phased lighting requirement blueprint and the high-confidence environment state snapshot to generate a preliminary brightness control instruction set, comprising: in the virtual lighting control space, establishing a target function for the expected lighting intensity range corresponding to the current period, establishing a constraint condition for the change slope tolerance value, and establishing an optimization boundary for the special trigger condition; mapping the lighting level and motion existence value in the high-confidence environment state snapshot to the initial state point in the virtual lighting control space; In the solution space defined by the constraint conditions and the optimization boundary, a gradient descent search algorithm is run to find an optimal luminance setpoint trajectory that satisfies the target function, with the dual optimization goals of minimizing energy consumption and maximizing lighting comfort; A series of discrete time-luminance pairs are sampled from the optimal luminance setpoint trajectory at fixed time intervals; Each time-luminance pair is encapsulated into a control instruction containing an execution timestamp and a target luminance value, and all control instructions are sorted in chronological order to form a preliminary luminance control instruction set.

2. The light brightness self-adaptive adjusting method for outdoor lamps according to claim 1, characterized in that, The reference lighting pattern matching the current geographical region and season is loaded from the preset environmental knowledge base, and the reference lighting pattern is matched with the dynamic environment perception scenario graph for pattern matching and difference calculation to identify the current environment in the dusk transition period, the night active period, the deep night silent period, and the dawn preparation period, including: The theoretical sunset time, the theoretical sunrise time, and the historical lighting intensity curve in the same period are retrieved from the environmental knowledge base as the reference lighting pattern according to the latitude and longitude coordinates of the outdoor lamp installation point and the current date; The actual lighting intensity node value at the current time point, the lighting intensity change trend edge weight in the past period, and the moving object trajectory density node value are extracted from the dynamic environment perception scenario graph to form a current actual state vector; The Euclidean distance between the current actual state vector and the reference lighting pattern in the corresponding time window is calculated to obtain a pattern difference degree, and when the pattern difference degree is lower than a quiet threshold and the lighting intensity is in a descending channel, it is determined that the current environment is in the dusk transition period; When the pattern difference degree is in an active threshold interval and the moving object trajectory density node value is higher than a preset density threshold, it is determined that the current environment is in the night active period; when the pattern difference degree is lower than the quiet threshold and the moving object trajectory density node value is lower than the preset density threshold, it is determined that the current environment is in the deep night silent period; When the pattern difference degree is lower than the quiet threshold and the lighting intensity is in an ascending channel, it is determined that the current environment is in the dawn preparation period.

3. The light intensity self-adaptive adjusting method for outdoor lamps according to claim 1, characterized in that, The original monitoring stream reported by the real-time heterogeneous sensor network is obtained, noise filtering, abnormal point repair, and confidence weighting fusion are performed on the original monitoring stream to form a high-confidence environmental state snapshot, including: Synchronously receiving the original monitoring stream with a timestamp reported by the light sensor, the pyroelectric infrared sensor, and the meteorological microsensor; Applying a combined filtering algorithm based on median filtering and moving average to the light sensor data in the original monitoring stream to filter out impulse noise and high-frequency disturbances to obtain a denoised light data sequence; Performing abnormal point detection based on logical judgment on the motion signal sequence reported by the pyroelectric infrared sensor, and performing interpolation repair on isolated signal points with too short duration or too low intensity to obtain a continuous motion state sequence; According to the factory calibration parameters of each sensor and the recent self-diagnosis report, dynamic confidence weights are assigned to the denoised light data sequence, the continuous motion state sequence, and the meteorological microsensor data, respectively; The dynamic confidence weight is used to perform weighted average fusion on the denoised light data sequence, the continuous motion state sequence, and the meteorological micro-sensor data, to generate a comprehensive high-confidence environmental state snapshot containing light levels, motion existence, and weather conditions.

4. The light brightness self-adaptive adjusting method for outdoor lamps according to claim 1, characterized in that, The gradient descent search algorithm is run in the solution space defined by the constraint conditions and the optimization boundary to find an optimal luminance setpoint trajectory satisfying the target function, including: initializing a tentative luminance change curve from the current luminance to a certain point in the expected light intensity range; calculating the instantaneous power consumption of the tentative luminance change curve at each sampling point, integrating to obtain a total energy consumption estimate, and simultaneously calculating the deviation integral of the tentative luminance change curve from the ideal comfort curve to obtain an estimate of discomfort; performing a weighted sum of the total energy consumption estimate and the discomfort estimate as the loss function value of the current tentative luminance change curve; calculating the gradient of the loss function with respect to each control point of the tentative luminance change curve; according to the gradient direction, fine-tune the control point coordinates of the tentative luminance change curve under the constraint conditions to generate a new tentative luminance change curve; iteratively performing the processes of calculating the loss function value, calculating the gradient, and fine-tuning the control points until the change in the loss function value is less than the convergence threshold, and the final tentative luminance change curve is obtained as the optimal luminance setpoint trajectory; the calculation of the instantaneous power consumption of the tentative luminance change curve at each sampling point, including: obtaining a luminance-power correspondence lookup table of an outdoor lamp, which stores the corresponding instantaneous power values at different luminance values; for each sampling point on the tentative luminance change curve, querying the luminance-power correspondence lookup table according to the luminance value of the sampling point to obtain the instantaneous power value of the sampling point; connecting the instantaneous power values of all sampling points to form a curve of instantaneous power change over time; numerically integrating the curve of instantaneous power change over time in the total time interval covered by the tentative luminance change curve; the result of numerical integration is the total energy consumption estimate corresponding to the tentative luminance change curve.

5. The light intensity self-adaptive adjusting method for outdoor light according to claim 1, wherein, the preliminary luminance control instruction set is input into an executor driving pipeline containing a feedback correction mechanism, and after physical constraint checking and smooth transition processing, the final light luminance control signal is output, including: reading the real-time working parameters of the current lamp from the outdoor lamp driving circuit, including the current luminance, temperature, and voltage; comparing the target luminance value of the first control instruction in the preliminary luminance control instruction set with the current luminance value in the real-time working parameters to calculate the step change; if the step change exceeds the maximum instantaneous change threshold allowed by the lamp hardware, re-planning the preliminary luminance control instruction set according to the maximum instantaneous change threshold, inserting intermediate transition instructions between the instructions to generate a revised luminance control instruction set that meets the physical constraints; The revised light regulation instruction set or the preliminary light regulation instruction set is sent into a linear execution queue, and a target light value of each instruction is converted into a corresponding pulse width modulation duty cycle value in order of execution time stamp; In each control cycle, a current pulse width modulation duty cycle value is output as a final light brightness control signal for driving an outdoor light emitting device.

6. The light intensity self-adaptive adjusting method for outdoor light according to claim 3, wherein, The dynamic confidence weight of the denoised light data sequence, the continuous motion state sequence and the meteorological micro sensor data is respectively assigned according to the factory calibration parameter of each sensor and the recent self-diagnosis report, including: The factory calibration parameter of the light sensor is inquired to obtain the accuracy level and linearity error, and a basic confidence coefficient is calculated according to the accuracy level and linearity error; The false alarm rate and the missed alarm rate statistics recorded in the recent self-diagnosis report of the pyroelectric infrared sensor are called, and a reliability attenuation factor is calculated according to the false alarm rate and the missed alarm rate; The signal strength value and the latest calibration time of the meteorological micro sensor are obtained, and a state health index is calculated according to the signal strength value and the length of time from the calibration time; The basic confidence coefficient, the reliability attenuation factor and the state health index are normalized and multiplied to obtain the dynamic confidence weight corresponding to each sensor data, wherein the dynamic confidence weight corresponding to the light sensor data is used for the denoised light data sequence, the dynamic confidence weight corresponding to the pyroelectric infrared sensor data is used for the continuous motion state sequence, and the dynamic confidence weight corresponding to the meteorological micro sensor data is used for the meteorological micro sensor data.

7. The light intensity self-adaptive adjusting method for outdoor lamps according to claim 2, wherein, The Euclidean distance between the current actual state vector and the reference light mode in the corresponding time window is calculated to obtain the mode difference degree, including: The time series data with a length of N is extracted from the current actual state vector, wherein N is determined by the time window size; The theoretical time series data with the same time window and the same length as the current time is intercepted from the reference light mode; The time series data and the theoretical time series data are regarded as two points in an N-dimensional space; The difference between the value of each dimension of the time series data and the corresponding dimension value of the theoretical time series data is calculated; Each dimension difference is squared; The squared differences of all dimensions are summed up; The square root of the sum is calculated to obtain a scalar value, which is the mode difference degree.

8. A light brightness self-adaptive adjusting system for an outdoor light, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor implements the steps of the light brightness adaptive adjustment method for outdoor lamps according to any one of claims 1 to 7 when executing the computer program.

Images