Lamp adaptive control method and system based on multi-sensor fusion

Abstract

The present application relates to the field of data processing, more particularly, the present application relates to a lamp adaptive control method and system based on multi-sensor fusion, the method comprises: collecting multi-sensor illumination data and mobile device positioning information, preprocessing to obtain multi-source input containing equivalent illumination value, coordinate, persistent confidence factor and transient discount factor;Using the persistent confidence factor, the transient discount factor and the distance weight based on the space correlation, the multi-sensor data is weighted and fused, and the continuous fusion illumination surface is generated;Taking the user dimming intervention as an implicit supervision signal, the persistent confidence factor is corrected, so that the fusion surface converges to the personalized preference;According to the deviation of the fusion illumination and the user expected illumination, the lamp dimming control is executed.The present application solves the problem of low fusion precision caused by the time and space asynchronization of multi-source heterogeneous sensors, realizes the continuity and robustness of illumination estimation, and can continuously adapt to the user's personalized preference.

Description

Technical Field

[0001] This invention relates to the field of data processing. More specifically, this invention relates to a method and system for adaptive control of lighting fixtures based on multi-sensor fusion. Background Technology

[0002] With the rapid development of smart home and IoT technologies, lighting systems are evolving from simple on / off control to intelligent and personalized solutions. Intelligent lighting solutions that integrate multi-sensor sensing technology and adaptive control algorithms have become an important way to improve living comfort and achieve energy conservation and emission reduction. Against this backdrop, how to enable lighting fixtures to automatically adjust according to environmental changes and user needs has become a research hotspot in this field.

[0003] However, existing adaptive lighting control solutions generally suffer from the following technical problems: due to the difficulty in strictly synchronizing the data sampling time of multiple sensors, the time reference is difficult to align when the system fuses illuminance data from different sources such as fixed sensors and mobile terminals. This results in low accuracy of the fused illuminance estimation, which in turn affects the accuracy and reliability of subsequent lamp dimming control. Summary of the Invention

[0004] To address the technical problem that existing adaptive lighting control schemes suffer from difficulty in aligning time references and low accuracy of illuminance estimation after fusion when processing illuminance data from different sources such as fixed sensors and mobile terminals, which in turn affects the accuracy and reliability of subsequent lamp dimming control, this invention provides solutions in the following aspects.

[0005] In the first aspect, the luminaire adaptive control method based on multi-sensor fusion includes: Multi-sensor illuminance data and mobile device positioning information are collected, and the collected data is preprocessed to obtain spatiotemporally consistent multi-source sensing input with a reliable initial state. The multi-source sensing input includes the equivalent illuminance value of each sensor at the current fusion time, the planar coordinates of each sensor, the persistent confidence factor and transient discount factor of each sensor. By utilizing the persistent confidence factor, transient discount factor, and spatial distance weight calculated based on the principle of spatial correlation of indoor illuminance field of each sensor, the equivalent illuminance values ​​of the multi-sensor after spatiotemporal alignment are weighted and fused to obtain the fused illuminance surface. By using user dimming intervention as a monitoring signal, the persistent confidence factor of each sensor is corrected, so that the fused illuminance surface gradually converges to the user's personalized preference. The luminaire dimming control is executed based on the deviation between the fused illuminance surface and the user's desired illuminance.

[0006] Optionally, the preprocessing includes: mapping the sampling time of each sensor to a unified time reference; for sensors that have not obtained new sampling data at the current fusion time, calculating the equivalent illuminance value of the sensor at the current fusion time based on its historical sampling sequence.

[0007] Optionally, the calculation of the spatial distance weight includes: For the target location to be estimated, the spatial distance weight of each sensor to the target location is calculated based on the Euclidean distance between each sensor and the target location and the effective spatial scale of each sensor; wherein, the effective spatial scale is obtained by adding the positioning uncertainty of the sensor to the preset basic spatial correlation scale.

[0008] Optionally, the weighted fusion includes: The fusion weight of each sensor is obtained by multiplying the persistent confidence factor, transient discount factor and spatial distance weight of each sensor. The equivalent illuminance values ​​of each sensor are then weighted and averaged to obtain the fused illuminance value of the target location. By traversing multiple target locations in the plane, a spatially continuous fused illuminance surface is obtained.

[0009] Optionally, it also includes dynamically maintaining the transient discount factor for each sensor: For each sensor, multiply the persistent confidence factor, transient discount factor, and spatial distance weight of all other sensors except the sensor itself, then multiply by their respective equivalent illuminance values, and sum the products to obtain the numerator; sum the products of the persistent confidence factor, transient discount factor, and spatial distance weight of all other sensors except the sensor itself to obtain the denominator; divide the numerator by the denominator to obtain the expected illuminance value at the location of the sensor. The actual equivalent illuminance value of the sensor is compared with the expected illuminance value, and the transient discount factor of the sensor is adjusted according to the comparison result.

[0010] Optionally, when the deviation between the actual equivalent illuminance value and the expected illuminance value continues to exceed a preset threshold, the transient discount factor of the sensor is reduced; when the deviation continues to fall below the preset threshold, the transient discount factor of the sensor is restored.

[0011] Optionally, when correcting the persistent confidence factor, the correction of the persistent confidence factor of the sensor involved in the region is triggered only when multiple dimming interventions in the same direction occur consecutively in the same region within a preset time period.

[0012] Optionally, the correction includes: If the illuminance reading of a sensor is higher than the fused illuminance and the user dimming direction is to brighten, or if the illuminance reading of the sensor is lower than the fused illuminance and the user dimming direction is to darken, then the persistent confidence factor of the sensor is enhanced; otherwise, the persistent confidence factor of the sensor is weakened.

[0013] Secondly, a lighting adaptive control system based on multi-sensor fusion includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the lighting adaptive control method based on multi-sensor fusion described in any one of the claims is implemented.

[0014] The present invention has the following beneficial effects: 1. This invention solves the problem of time asynchrony caused by differences in reporting cycles and transmission delays by unifying the data from various sensors with inconsistent sampling times to the same time base. Based on this, it performs weighted fusion of three factors: long-term reliability of the sensors, short-term operating status, and spatial distance between the sensors and the target location. This ensures that readings from nearby, reliable, and normal sensors dominate the fusion process, while contributions from distant, questionable, or abnormal sensors are effectively suppressed. Thus, without requiring sensor hardware synchronization or centimeter-level positioning, it can still stably output a continuous spatial illuminance surface that closely matches the actual illuminance distribution. Simultaneously, it utilizes the user's daily dimming behavior as a monitoring signal to continuously optimize sensor reliability, allowing illuminance estimation to gradually adapt to the user's subjective preferences over time. This fundamentally avoids control inaccuracies caused by single false alarms or local anomalies, achieving long-term stable, accurate, and reliable adaptive operation of the lighting system in real-world living environments.

[0015] 2. This invention dynamically maintains the transient discount factor, utilizes cross-validation between sensors to detect and temporarily weaken the impact of abnormal readings in real time, effectively filters out occasional interference such as people blocking the light and curtains swaying, and ensures that the fused illuminance surface does not undergo severe distortion when local sensors are temporarily inaccurate, thereby improving the system's anti-interference capability in real living scenarios.

[0016] 3. This invention uses the user's dimming intervention behavior as a monitoring signal to correct the sensor's persistent confidence factor over a long time scale, so that the fused illuminance surface gradually converges to the user's subjective preference. Without the need for additional special equipment or active labeling, it realizes the personalized evolution of adaptive lighting and significantly improves user comfort.

[0017] 4. The transient discount factor independently responds to minute-level sudden anomalies and automatically recovers, while the persistent confidence factor accumulates intervention evidence and updates progressively over a scale of several hours to several days. The two do not interfere with each other, enabling the system to have both sensitive anomaly suppression capabilities and stable preference learning capabilities, avoiding control oscillations and misjudgments, and ensuring the reliability and stability of the system's long-term operation. Attached Figure Description

[0018] Figure 1 This is a flowchart of steps S1-S4 in the adaptive control method for lighting fixtures based on multi-sensor fusion according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0020] This invention is applied to the lighting environment of a family living room. Multiple illuminance sensors are deployed in the living room, including a main ambient light illuminance sensor on the ceiling, fixed auxiliary illuminance sensors in the activity area, and illuminance sensors built into the user's personal mobile device. When the user's mobile device does not have its location function enabled, the system can still receive its illuminance data, but will not incorporate it into spatial fusion until location is restored or the user's location is obtained through an alternative location solution.

[0021] Reference Figure 1 The adaptive control method for lighting fixtures based on multi-sensor fusion includes steps S1-S4, as detailed below: S1: Collect multi-sensor illuminance data and mobile device positioning information, and preprocess the collected data to obtain spatiotemporally consistent multi-source sensing input with a reliable initial state; the multi-source sensing input includes the equivalent illuminance value of each sensor at the current fusion time, the planar coordinates of each sensor, the persistent confidence factor and transient discount factor of each sensor.

[0022] The system receives raw illuminance measurements and their local timestamps from multiple illuminance sensors deployed in a home living room via wired or wireless communication networks. Simultaneously, for user-carried mobile devices, the system obtains real-time planar coordinates and positioning uncertainty through their built-in indoor positioning module. This positioning uncertainty is typically output directly by the positioning module or preset by the system based on the currently active positioning technology type, such as 3 meters for Wi-Fi positioning, 2 meters for Bluetooth positioning, 0.3 meters for UWB positioning, and 5 meters for GNSS indoor or unknown positioning. If positioning is not enabled, the sensor is marked as having an unknown spatial location, and its illuminance data is not included in spatial fusion. Fixed auxiliary illuminance sensors have their two-dimensional planar projection coordinates measured using a ruler during deployment and entered into the system database.

[0023] Since the actual sampling times of the various sensors are often not strictly synchronized, the system first performs spatiotemporal alignment on the multi-source data. A fixed-length sliding data storage window is maintained for each sensor participating in illuminance sensing, with a window length of 20 sampling points, retaining the raw illuminance measurements from the most recent sampling times and their corresponding sampling times. All sensors map their local timestamps to the unified system clock via the home wireless LAN's clock synchronization protocol.

[0024] The system initiates spatial illuminance estimation calculations at a fixed fusion period of 1 second. The moment the calculation is triggered is recorded as the current fusion moment. At the current fusion moment, if a sensor has not yet reported new data, the local drift rate is calculated based on the recent historical sampling sequence within the window: the ratio of the illuminance difference between adjacent sampling points to the time interval is extracted to form a first-order difference sequence; if the effective difference data is insufficient, the most recent effective sampling value is directly used as the equivalent illuminance value. When data is sufficient, a weighted average is performed using a preset persistence confidence factor to obtain the local drift rate. Subsequently, a linear extrapolation is performed based on the most recent effective illuminance value to obtain the equivalent illuminance value of the sensor at the current fusion moment. In the formula, For the first The equivalent illuminance value of each sensor at the current fusion moment is not the actual sampled value of the sensor at the current fusion moment, but an estimated value calculated by the system in order to bring the data of each asynchronous sensor to the same time reference point; This is the illuminance value most recently sampled by the sensor; This refers to the local drift rate obtained from historical data statistics. This is the system moment at which the fusion estimation is triggered. This refers to the system time of the most recent actual sampling. If the extrapolation time is... If the preset maximum allowable delay is exceeded, such as 5 seconds, the sensor is marked as temporarily disabled and will not participate in the current fusion calculation. If the linear extrapolation result is negative, the equivalent illuminance value is clamped to 0. Through this linear extrapolation, the system can compensate for temporary data loss caused by wireless transmission delays or sensor sleep, providing each sensor with a set of equivalent illuminance representative values ​​consistent with the time reference at the current fusion moment.

[0025] During the initialization phase, the system configures persistent confidence factors and transient discount factors for each sensor. The initial persistent confidence factor is a dimensionless scalar with a value of [value missing]. The confidence factor characterizes the system's prior confidence in the consistency of sensor measurements, with a lower limit of 0.1 to avoid completely excluding a particular sensor in extreme cases. Initial values ​​are set according to sensor type: 0.7 for ceiling main ambient light sensors (meaning acceptable repeatability but limited spatial representativeness); 0.9 for fixed auxiliary illuminance sensors (meaning factory-calibrated sensors accurately reflect the true illuminance of the area); and 0.4 for mobile device built-in sensors (meaning significant deviations due to factors such as grip posture and occlusion). The transient discount factor is also a dimensionless scalar with an initial value of 1.0, indicating no persistent anomalies detected. This factor is only used for short-term anomaly suppression, and its adjustment does not change the persistent confidence factor; the two are completely decoupled in terms of value and update mechanism.

[0026] In addition, the system integrates the fixed sensor coordinates and the mobile device positioning coordinates into the two-dimensional coordinate system of the living room plane. After registration, each sensor with defined coordinates corresponds to an equivalent illuminance value and a planar coordinate at the current fusion time. The persistent confidence factor and transient discount factor of each sensor are state variables independent of the fusion time.

[0027] The system thus obtains the spatiotemporally aligned illuminance data and location information of each sensor, and has initial state parameters to distinguish the reliability of the sensors.

[0028] S2: By utilizing the persistent confidence factor, transient discount factor, and spatial distance weight calculated based on the principle of spatial correlation of indoor illuminance field of each sensor, the equivalent illuminance values ​​of the multi-sensor after spatiotemporal alignment are weighted and fused to obtain a smooth fused illuminance surface that suppresses extreme disturbances.

[0029] After completing spatiotemporal alignment, the system obtains the equivalent illuminance value, planar coordinates, persistent confidence factor, and transient discount factor for each sensor at the current fusion time. Sensors from mobile devices with unknown spatial locations are not included in this fusion calculation. Within the living room space, due to the influence of the natural light incidence direction and the light distribution of the lighting fixtures, illuminance exhibits a gradient change on the horizontal plane. The readings of each sensor are related to local occlusion at their location, installation height, and their own photoelectric characteristics; therefore, the measurement values ​​from different sensors do not reflect the true illuminance to a consistent degree. If the sensor readings are directly averaged or interpolated using a single distance attenuation weight, the deviations of low-confidence sensors or locally anomalous readings will indiscriminately spread across the entire plane, resulting in an unsmooth or even inaccurate illuminance estimation surface. Therefore, this step employs a triple-weighted fusion scheme that includes persistent confidence, transient discount, and spatial distance attenuation to synthesize multi-point discrete illuminance values ​​into a spatially continuous fused illuminance.

[0030] In an indoor illuminance field, the correlation between illuminance values ​​between two points weakens with increasing spatial distance; the closer the points, the more similar the illuminance values; the farther the points, the lower the correlation. Therefore, for any planar target location to be estimated, the illuminance value of a sensor closer to that location is more representative of the true illuminance at that location and should be given a higher weight; the illuminance value of a sensor farther away has lower reference value and its weight should be reduced accordingly.

[0031] For the planar target location to be estimated, the location can be the user's current location, a system-predefined lighting control reference point, a grid node of the spatial illuminance surface discretization, or other coordinate points specified by the implementer according to specific application requirements.

[0032] First, calculate the Euclidean distance between the target location and the plane coordinates of each sensor.

[0033] Based on the aforementioned spatial correlation principle, a Gaussian spatial correlation kernel is used to convert distance into spatial weights. Considering the positioning uncertainty of the planar coordinates obtained by mobile devices through indoor wireless positioning, using the same distance attenuation scale for both mobile devices and fixed sensors would lead to excessive rejection of the mobile device's illuminance value due to position drift. Therefore, a spatial effective scale is defined for each sensor, which is obtained by adding the sensor's positioning uncertainty to a preset basic spatial correlation scale.

[0034] The above-mentioned basic spatial dimensions are based on an empirical value of 1.0 meter. This parameter is a preset parameter set for the standard living room width of a family. Implementers can adjust it according to the actual size of the living room.

[0035] Based on this, the spatial distance weights of each sensor to the aforementioned target location are calculated: In the formula, For the first Each sensor has a spatial distance weight for the target location. For the first The Euclidean distance between each sensor and the target location. For the first The effective spatial scale of each sensor It is an exponential function with the natural constant e as its base. The physical meaning of this Gaussian kernel function is: when... much smaller At that time, the exponent term Approaching 0, A value close to 1 indicates that the sensor contributes strongly to the illuminance at the target location, and its illuminance value is highly correlated with the actual illuminance at the target location; when... Gradually increase and exceed At that time, the exponent term rapidly changes towards negative infinity. The rapid decay to near zero indicates that the sensor's illuminance value has extremely low reference value for the target location, and its contribution is effectively suppressed. Through this Gaussian decay, the system achieves the spatial weighted fusion principle of giving higher weight to nearby values ​​and lower weight to distant values.

[0036] Different sensors exhibit varying reliability over long-term use. The persistent confidence factor characterizes the system's level of trust in the long-term measurement consistency of the sensors. Simultaneously, sensors may be affected by sudden factors such as brief pedestrian obstruction or swaying curtains, resulting in transiently higher or lower illuminance measurements. Directly incorporating such outliers into the fusion process with full weight would cause local distortion of the illuminance surface at the time of the anomaly. Therefore, a transient discount factor is introduced. Under normal conditions where sensors do not experience persistent anomalies, the transient discount factor is set to 1.0, indicating that its fusion contribution is not subject to additional discounting. When a sensor experiences persistent anomalies, the transient discount factor is temporarily reduced to a value less than 1.0 to temporarily weaken the sensor's influence in the fusion process. This adjustment does not modify the persistent confidence factor, thus maintaining the independence between short-term anomaly suppression and long-term trust assessment. At the current fusion moment, the transient discount factor of each sensor is its current actual value (initially all 1.0) and directly participates in the weighted calculation.

[0037] In summary, by combining the three factors mentioned above—the persistent confidence factor, the transient discount factor, and the spatial distance weight—and performing illuminance weighted fusion on the target location, the fused illuminance at that target location is obtained, satisfying the following relationship: In the formula, For the system at the target location The fused illuminance at the location, For the first The persistent confidence factor of each sensor. For the first The transient discount factor of each sensor, For the first Each sensor has a spatial distance weight for the target location. For the first The equivalent illuminance values ​​of each sensor at the current fusion moment. The total number of sensors. This fused illuminance incorporates illuminance information from the sensor that is spatially closest, has the highest long-term reliability, and has not currently exhibited any anomalies. It reflects the most likely true illuminance level of the target location at the current moment and serves as the fundamental physical quantity for subsequent illuminance control decisions and user intervention learning by the system.

[0038] By repeating the above calculations across multiple target locations within the living room plane, a continuous blended illuminance surface can be obtained. The illuminance value at any point on this surface represents the blended illuminance result at the current blending moment. It's important to note that the blended illuminance surface essentially reflects the spatial illuminance distribution within the living room plane at the current blending moment. That is, by performing point-by-point calculations on multiple discrete target locations within the plane, combined with spatial interpolation fitting, a continuously distributed illuminance value is obtained throughout the entire living room area. Each point on this surface corresponds to a spatial coordinate, and its illuminance value is the blended illuminance result at that location at the current blending moment. In this way, a limited number of discrete sampling points are expanded into a complete illuminance field covering the entire area, clearly presenting the distribution patterns and transition trends of light in the indoor space.

[0039] As a supplement to the above fusion method, the system can further perform short-term anomaly suppression on each sensor involved in the fusion to improve the stability of the fused illuminance surface under occasional interference. This supplementary mechanism does not change the above triple-weighted fusion framework, but simultaneously maintains the transient discount factor of each sensor during the fusion calculation. Its purpose is to identify and suppress short-term anomalies of individual sensors and prevent instantaneous measurement disturbances caused by occasional events such as people walking or curtains swaying from contaminating the fusion results.

[0040] The basic idea is as follows: For each sensor, a predicted value is calculated at the sensor's location using all other sensors except the sensor itself. This employs the same weighted fusion method as described above, excluding the sensor's own data during calculation. Specifically, the persistent confidence factor, transient discount factor, and spatial distance weight of all other sensors are multiplied together, then multiplied by their respective equivalent illuminance values. The sum of these products forms the numerator. The sum of the products of the persistent confidence factor, transient discount factor, and spatial distance weight of all other sensors forms the denominator. The numerator is then divided by the denominator to obtain the expected illuminance value at the sensor's location. This expected illuminance value represents the consistent assessment of the illuminance at that sensor's location by the other sensors. Then, the actual reading of the sensor is compared with the predicted value to calculate the residual. If the two continue to deviate too much, it indicates that the sensor may be blocked or malfunctioning. The system will reduce the transient discount factor from 1.0 to 0.9 to temporarily reduce the weight of the sensor in subsequent fusion. After its reading continues to fall back to the normal range, the transient discount factor will be restored to 1.0.

[0041] The normal fluctuation range of the residuals is quantified using historical standard deviations: a residual sliding window is maintained for each sensor, recording the residual history of the most recent 50 fusion cycles and calculating its standard deviation. An out-of-limit threshold is set to three standard deviations, and a recovery threshold is set to two standard deviations, thus forming a hysteresis interval to prevent frequent state switching. A down-adjustment is triggered when there are 5 consecutive out-of-limit occurrences, and a recovery is triggered when there are 3 consecutive down-adjustments.

[0042] Throughout the process, the persistent confidence factor remains constant, achieving decoupling between short-term anomaly suppression and long-term confidence assessment. The adjusted transient discount factor is directly applied to subsequent fusion calculations.

[0043] S3: Using user dimming intervention as a monitoring signal, the persistent confidence factor of each sensor is corrected, so that the fused illuminance surface gradually converges to the user's personalized preferences.

[0044] After completing spatial illuminance fusion, the system outputs a fused illuminance surface that reflects the physical illuminance estimates for various locations in the living room. However, this estimate still relies on the current persistent confidence factors of each sensor, which are set to initial values ​​based on sensor type during system initialization. There is a difference between the user's subjective perception of the lighting environment and the purely physical illuminance values, and the time-varying characteristics of sensors, such as slow dust accumulation and aging, can cause the initial confidence allocation to gradually deviate from reality. When a user actively adjusts the brightness of the lights using wall knobs, touchscreen sliders, voice commands, or remote controls, this action directly expresses the user's dissatisfaction with the current illuminance level, meaning that a perceptible deviation has occurred between the system's fused illuminance estimate at the user's location and the user's subjective expectation. While performing dimming actions, the system naturally learns the timing, direction, and magnitude of this intervention event. Therefore, this step utilizes the user's dimming intervention as an implicit monitoring signal to correct the persistent confidence factors of each sensor over a long time scale, gradually converging the illuminance estimation surface to a state that conforms to the user's personalized preferences.

[0045] It's important to note that this step corrects the persistent confidence factor, not the transient discount factor. The transient discount factor handles short-term anomalies in the sensor within minutes, such as people blocking the light or curtains swaying; it returns to its original value once the anomaly is resolved, representing an immediate response mechanism. The persistent confidence factor, on the other hand, reflects the sensor's long-term reliability, indicating whether the sensor is trustworthy over a timescale of months or even longer. When a user repeatedly adjusts the brightness, it indicates a systematic bias in the sensor's response to user preferences, requiring a fundamental correction of its long-term reliability so that the system can gradually learn the user's personalized lighting preferences.

[0046] User dimming intervention can be achieved through wall knobs, touchscreen sliders, voice commands, or remote controls. While performing the dimming action, the system automatically learns the time, direction, and magnitude of the intervention. When the user's mobile device location is not enabled, the dimming intervention is still recorded but not updated until location is restored or the user's location is obtained through alternative means such as a fixed infrared sensor or Bluetooth beacon.

[0047] When intervention occurs, the system pauses the automatic control loop to avoid negating the intervention effect. The intervention time, user location, and dimming direction are recorded. Based on the photometric data calibrated at the luminaire's factory, the system calculates the actual change in illuminance at the user's location caused by the dimming, then infers the user's desired illuminance and compares it with the system's current fusion estimate at that location to obtain a deviation value. The sign of this deviation reflects whether the user wants brighter or dimmer light.

[0048] To prevent a single user error from causing an incorrect shift in confidence levels, the system divides the living room floor plan into several grid areas, with each area independently tracking the direction of dimming interventions. Only when the same dimming intervention occurs twice consecutively in the same area within 600 seconds does the system determine that this represents the user's true preference and trigger a persistent confidence factor update for the sensors involved in that area.

[0049] During updates, the system calculates the contribution and direction of each sensor to the fusion result at that location based on the sensor readings at the time of intervention, their current weights, and their distance from the user. If a sensor's reading is higher than the fusion value, but the user actually wants brighter light, it indicates that the sensor's reading direction aligns with the user's preference, and its confidence level should be increased; conversely, it should be weakened. The impact of recent interventions is greater than that of long-term interventions, and the system automatically achieves this effect through a time decay mechanism. The updated confidence factor is limited to between 0.1 and 1.0 to prevent any sensor from being completely excluded or over-relied. After each update, the system records the update timestamp of that sensor for regression management during periods of no updates.

[0050] Through repeated dimming interventions and cumulative learning, the persistent confidence factor of each sensor gradually evolves from a general preset value to a personalized value that adapts to the current user's preferences. As a result, the fused illuminance surface output by the system becomes more and more in line with the user's real lighting needs in different areas and at different times.

[0051] As a supplement to the above update mechanism, the system scans the most recent update timestamp of all sensors every 600 seconds. If the inactivity period of a certain sensor exceeds 1 day, the persistence confidence factor corresponding to the sensor will be regressed once to the initial value of the persistence confidence factor assigned to the sensor during system initialization in a very small step of one-thousandth.

[0052] If the sensor subsequently receives an update due to new user intervention, the most recent update timestamp is refreshed, and the regression timer restarts. This regression operation applies not only to sensors currently participating in the fusion process but also to sensors that are not currently participating due to reasons such as location services not being enabled but may resume participation in the future.

[0053] By combining user-intervention-driven active learning with long-term uninterrupted passive regression, the system can continuously learn users' personalized illuminance preferences in different areas. At the same time, it prevents individual sensors from being underestimated due to isolated historical events and thus failing to provide effective data for illuminance fusion. This ensures that lighting control decisions are always based on comprehensive and accurate illuminance perception information.

[0054] S4: Based on the deviation between the fused illuminance surface and the user's desired illuminance, calculate the dimming control amount of each lamp and execute dimming to achieve adaptive control of living room lighting.

[0055] After completing sensor data acquisition and spatiotemporal alignment, spatial illuminance fusion, and user preference learning from S1 to S3, the system has the following conditions at the current fusion time: (1) a spatially continuous fusion illuminance surface that reflects the current illuminance estimate of each location in the living room; (2) through user intervention learning in S3, the expected illuminance corresponding to the user's most recent effective dimming intervention in each area is recorded. This value serves as the current control target for that area, i.e., the user's expected illuminance (if there is no effective intervention record for a certain area, active control will not be performed for the time being, and the system will maintain the current illuminance output). The system uses the user's location or the center coordinates of each preset control area as the control target location, queries the expected illuminance at that location, compares it with the blended illuminance, and calculates the illuminance error. If the error is greater than zero, it means the illuminance is lower than the user's expectation and needs to be brightened; if the error is less than zero, it means the illuminance is higher than the user's expectation and needs to be dimmed; if the error is close to zero, the current brightness is maintained.

[0056] For areas requiring adjustment, the system calculates the contribution weight of each luminaire to the illuminance of that area based on its light distribution characteristics and spatial location. The contribution weight is determined by the proportion of the illuminance generated by that luminaire at that location to the total illuminance at that location; luminaires with greater contributions undertake more dimming tasks. Based on the total illuminance error and the contribution weight of each luminaire, the system calculates the required illuminance adjustment for each luminaire, maps this adjustment to a PWM duty cycle adjustment using the luminaire's photometric calibration curve, and then sends dimming commands to each luminaire driver via the communication network.

[0057] After the dimming command is executed, the system re-perceives and fuses the illuminance in the next fusion cycle, generating a new illuminance error, and enters the next control loop until the illuminance error at each control location converges to the preset tolerance range. Through the above closed-loop control mechanism, the system realizes a complete control loop of perception → fusion → learning → decision → execution → re-perception, ensuring that the illuminance in each area of ​​the living room is continuously maintained at the user's desired level, completing adaptive lighting control based on multi-sensor fusion.

[0058] The system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the luminaire adaptive control method based on multi-sensor fusion according to the first aspect of the present invention.

[0059] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0060] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A lighting fixture adaptive control method based on multi-sensor fusion, characterized in that, include: Collect illumination data from multiple sensors and positioning information from mobile devices, and preprocess the collected data to obtain spatiotemporally consistent multi-source sensing input with a reliable initial state; The multi-source sensing input includes the equivalent illuminance value of each sensor at the current fusion time, the planar coordinates of each sensor, the persistent confidence factor and the transient discount factor of each sensor. By utilizing the persistent confidence factor, transient discount factor, and spatial distance weight calculated based on the principle of spatial correlation of indoor illuminance field of each sensor, the equivalent illuminance values ​​of the multi-sensor after spatiotemporal alignment are weighted and fused to obtain the fused illuminance surface. By using user dimming intervention as a monitoring signal, the persistent confidence factor of each sensor is corrected, so that the fused illuminance surface gradually converges to the user's personalized preference. The luminaire dimming control is executed based on the deviation between the fused illuminance surface and the user's desired illuminance.

2. The adaptive control method for lighting fixtures based on multi-sensor fusion according to claim 1, characterized in that, The preprocessing includes: mapping the sampling time of each sensor to a unified time reference; for sensors that have not obtained new sampling data at the current fusion time, calculating the equivalent illuminance value of the sensor at the current fusion time based on its historical sampling sequence.

3. The adaptive control method for lighting fixtures based on multi-sensor fusion according to claim 1, characterized in that, The calculation of the spatial distance weight includes: For the target location to be estimated, the spatial distance weight of each sensor to the target location is calculated based on the Euclidean distance between each sensor and the target location and the effective spatial scale of each sensor; wherein, the effective spatial scale is obtained by adding the positioning uncertainty of the sensor to the preset basic spatial correlation scale.

4. The adaptive control method for lighting fixtures based on multi-sensor fusion according to claim 1, characterized in that, The weighted fusion includes: The fusion weight of each sensor is obtained by multiplying the persistent confidence factor, transient discount factor and spatial distance weight of each sensor. The equivalent illuminance values ​​of each sensor are then weighted and averaged to obtain the fused illuminance value of the target location. By traversing multiple target locations in the plane, a spatially continuous fused illuminance surface is obtained.

5. The adaptive control method for lighting fixtures based on multi-sensor fusion according to claim 1, characterized in that, It also includes dynamic maintenance of the transient discount factor for each sensor: For each sensor, multiply the persistent confidence factor, transient discount factor, and spatial distance weight of all other sensors except the sensor itself, then multiply by their respective equivalent illuminance values, and sum the products to obtain the numerator; sum the products of the persistent confidence factor, transient discount factor, and spatial distance weight of all other sensors except the sensor itself to obtain the denominator; divide the numerator by the denominator to obtain the expected illuminance value at the location of the sensor. The actual equivalent illuminance value of the sensor is compared with the expected illuminance value, and the transient discount factor of the sensor is adjusted according to the comparison result.

6. The adaptive control method for lighting fixtures based on multi-sensor fusion according to claim 5, characterized in that, When the deviation between the actual equivalent illuminance value and the expected illuminance value continues to exceed a preset threshold, the transient discount factor of the sensor is reduced; when the deviation continues to fall below the preset threshold, the transient discount factor of the sensor is restored.

7. The adaptive control method for lighting fixtures based on multi-sensor fusion according to claim 1, characterized in that, When correcting the persistent confidence factor, the correction of the persistent confidence factor of the sensor involved in the region is triggered only when multiple dimming interventions in the same direction occur consecutively in the same region within a preset time period.

8. The adaptive control method for lighting fixtures based on multi-sensor fusion according to claim 7, characterized in that, The corrections include: If the illuminance reading of a sensor is higher than the fused illuminance and the user dimming direction is to brighten, or if the illuminance reading of the sensor is lower than the fused illuminance and the user dimming direction is to darken, then the persistent confidence factor of the sensor is enhanced; otherwise, the persistent confidence factor of the sensor is weakened.

9. A lighting adaptive control system based on multi-sensor fusion, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the adaptive lighting control method based on multi-sensor fusion according to any one of claims 1-8.

Images