Indoor space lighting fast response system based on edge computing

Abstract

The application discloses an indoor space lighting fast response system based on edge computing, which comprises a space state acquisition module, an edge computing processing module, a mode determination module, a mode consistency constraint module, a parameter updating module and a lighting control output module.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology for intelligent lighting, and more particularly to a fast response system for indoor space lighting based on edge computing. Background Technology

[0002] With the improvement of building intelligence, indoor lighting systems are gradually evolving from traditional manual control to automation and intelligence. Existing indoor lighting control systems typically manage lighting equipment through centralized controllers or cloud platforms, adjusting brightness and color temperature based on preset rules or simple sensor data. Structurally, these systems rely on a central node to complete data processing and decision generation. Data needs to be transmitted via network to a remote location for processing before returning to the execution end, resulting in a long control chain.

[0003] In application scenarios with frequent personnel movement and rapid changes in spatial conditions, the aforementioned centralized or cloud-based control methods are prone to response delays, making it difficult to reflect real-time changes in spatial conditions in a timely manner. Furthermore, existing lighting controls often employ fixed thresholds or single models for state determination, which have limited adaptability to complex changes in spatial conditions and are prone to frequent switching or control instability near state boundaries.

[0004] Furthermore, while some lighting systems incorporate local control units, their control models typically employ simple linear relationships or empirical rules, lacking a systematic description of the constraints on control parameters under different operating conditions. This makes it difficult to balance real-time performance with constraint consistency during parameter updates. When a lighting system needs to switch between multiple operating modes, existing technologies often lack a clear mode-consistency control mechanism, easily leading to abrupt changes in control parameters and affecting the continuity and stability of the lighting effect.

[0005] Therefore, existing indoor space lighting control technologies still have room for improvement in terms of real-time response capability, mode switching stability, and parameter constraint handling, and need to be further refined. Summary of the Invention

[0006] One objective of this invention is to propose a rapid response system for indoor space lighting based on edge computing. This invention fully integrates piecewise affine modeling, mode consistency constraint mechanism, projection gradient update method under the feasible domain of polyhedral parameters, and real-time computing architecture on the edge side. It systematically describes the entire process of indoor space lighting control from spatial state acquisition, state analysis, mode determination, mode switching constraints, online parameter updates to lighting control output. By completing mode determination and parameter update decisions within the edge computing node, the control link is effectively shortened. Under the condition of introducing a switching hysteresis domain and a switching cost potential function, frequent mode switching is suppressed. Furthermore, the polyhedral decomposition projection mechanism ensures that the control parameters meet the constraint consistency requirements under different operating modes. Thus, in indoor scenes with frequent personnel flow and complex changes in ambient lighting, a timely response, continuous control, and stable parameter lighting control effect is achieved.

[0007] According to an embodiment of the present invention, an indoor space lighting fast response system based on edge computing includes:

[0008] The spatial state acquisition module is used to acquire indoor spatial state data and form a spatial state vector;

[0009] The edge computing processing module is used to perform analytical operations on the spatial state vector on the local side of the indoor space and generate spatial state analysis results.

[0010] The pattern determination module is used to map the spatial state analysis results to multiple preset state intervals to form a candidate pattern set.

[0011] The mode consistency constraint module is used to introduce a switching hysteresis domain into the candidate mode set and calculate the mode switching cost potential function. When the spatial state change is within the switching hysteresis domain, the current mode identifier remains unchanged. When the spatial state change exceeds the switching hysteresis domain, the module determines whether to update the mode identifier based on the comparison result between the switching cost potential function and the preset cost threshold.

[0012] The parameter update module is used to perform gradient update operations on the corresponding parameter vector under the current mode identifier constraint, and project the updated parameter vector into the polyhedral parameter feasible region associated with the current mode identifier. The polyhedral parameter feasible region consists of several constraint subdomains. The parameter projection process performs decomposition projection operations according to the constraint subdomains. When the mode identifier changes, the parameter vector is mapped across the feasible region.

[0013] The lighting control output module is used to generate lighting control parameters based on the current mode identifier and the updated parameter vector, and output them to the lighting equipment.

[0014] Optionally, the spatial state acquisition module specifically includes:

[0015] Within the indoor space, spatial area identifiers are read according to a preset data acquisition configuration table. The sensor nodes corresponding to each spatial area identifier are polled and accessed to collect personnel presence status data, spatial illuminance data, and ambient light data with consistent time indices. A unified time index is added to each type of collected data to form raw status data items. A time index alignment verification operation is performed on the raw status data items, eliminating data items with inconsistent time indices and retaining those that meet the time index consistency condition to form an aligned status data set. Discrete state encoding is performed on the personnel presence status data within the aligned status data set, and numerical normalization is performed on the spatial illuminance data and ambient light data to generate corresponding coded state components and normalized state components. Vector concatenation is performed on the coded state components and normalized state components according to a preset state component arrangement order to form a spatial state vector. The spatial state vector is written to the state buffer of the edge computing node, stored in time index order, and output to the edge computing processing unit.

[0016] Optionally, the edge computing processing module specifically includes:

[0017] The system reads spatial state vectors sorted by time index from the state buffer, performs integrity verification on the spatial state vectors, filters out spatial state vectors with missing state components, and retains spatial state vectors that meet the integrity conditions to form a valid state sequence; performs difference operation on spatial state vectors corresponding to adjacent time indices in the valid state sequence to calculate the spatial state change vector, and writes the spatial state change vector into the change buffer; performs amplitude determination operation on the spatial state change vector in the change buffer to generate a change determination mark set; performs joint analysis operation on the spatial state vector and the change determination mark set according to preset analysis rules to generate spatial state analysis results; writes the spatial state analysis results into the analysis result buffer and outputs them to the mode determination unit in time index order.

[0018] Optionally, the pattern determination module specifically includes:

[0019] Read the spatial state parsing results from the parsing result cache in time index order, perform interval mapping operation on the state components in the spatial state parsing results, map each state component to a preset state interval identifier, and form a state interval identifier vector.

[0020] Perform a combination mapping operation on the state interval identifier vector to generate a candidate pattern set according to the preset interval combination rules; perform a consistency filtering operation on the candidate pattern set to remove pattern identifiers that do not meet the continuity condition with the historical pattern identifiers corresponding to the current time index, and retain the pattern identifiers that meet the continuity condition to form a filtered pattern set; perform a priority sorting operation on the filtered pattern set, and select the pattern identifier at the top of the sorting result as the current pattern identifier; write the current pattern identifier into the pattern identifier buffer, and output it to the pattern consistency constraint module in the order of time index.

[0021] Optionally, the process of introducing the switching hysteresis domain and calculating the mode switching cost potential function in the mode consistency constraint module includes the following steps:

[0022] Read the historical pattern identifier corresponding to the previous time index from the pattern identifier cache, and read the spatial state parsing result corresponding to the current time index from the parsing result cache; perform hysteresis interval mapping operation on the state components in the spatial state parsing result according to the preset hysteresis parameter table to generate corresponding hysteresis interval identifiers for each state component; perform joint judgment operation on the hysteresis interval identifiers, generate a hysteresis hold flag when all state components fall into the range corresponding to the hysteresis interval identifier, and generate a hysteresis release flag when any state component exceeds the range corresponding to the hysteresis interval identifier.

[0023] When generating the hysteresis release flag, for each mode identifier in the candidate mode set, a mode index parsing operation is performed on the current parameter vector to form a mode parameter set that corresponds one-to-one with each mode identifier; a difference calculation operation is performed on the spatial state parsing result and each mode parameter set to generate a mode deviation vector; a weighted mapping operation is performed on the mode deviation vector to generate a cost component vector that corresponds one-to-one with each candidate mode identifier; an aggregation operation is performed on the cost component vector to calculate the mode switching cost potential function value corresponding to each mode identifier in the candidate mode set; the mode switching cost potential function value and the corresponding mode identifier are written into the cost buffer, stored in time index order and output.

[0024] Optionally, the process of determining whether to update the mode identifier based on the mode switching cost potential function in the mode consistency constraint module includes the following steps:

[0025] The system reads the set of mode switching cost potential function values ​​corresponding to the current time index from the cost buffer, and reads the historical mode identifier corresponding to the previous time index from the mode identifier buffer. When generating a hysteresis hold flag, it directly selects the historical mode identifier as the current mode identifier and writes the current mode identifier into the mode identifier buffer. When generating a hysteresis release flag, it reads the hold cost value corresponding to the historical mode identifier from the set of mode switching cost potential function values, and reads the switching cost value corresponding to the other candidate mode identifiers. It performs a difference comparison operation on the switching cost value and the hold cost value to generate a cost comparison result vector. It performs an item-by-item comparison operation on the cost comparison result vector and a preset cost threshold to filter candidate mode identifiers that meet the cost threshold constraint to form an allowed update mode set. When the allowed update mode set is not empty, it performs a minimum cost selection operation and selects the candidate mode identifier with the smallest cost value as the current mode identifier. When the allowed update mode set is empty, it maintains the historical mode identifier as the current mode identifier. It writes the current mode identifier into the mode identifier buffer and outputs it to the parameter update unit in time index order.

[0026] Optionally, the process of performing gradient update operations under the current mode identifier constraint in the parameter update module includes the following steps:

[0027] The parameter update module reads the parameter vector associated with the current mode identifier from the parameter buffer, and parses the parameter vector according to a preset parameter arrangement rule to generate a parameter component sequence. It then reads the spatial state parsing result corresponding to the current time index from the parsing result buffer, performs an association mapping operation on the spatial state parsing result and the parameter component sequence to generate a target evaluation value. The parsing result buffer refers to the spatial state parsing result. Gradient calculation is performed on the target evaluation value along the direction of the parameter component sequence to generate a gradient vector. A scaling operation is performed on the gradient vector according to a preset step size parameter to generate an update step size vector. Vector subtraction is performed on the parameter component sequence and the update step size vector to generate an intermediate gradient update parameter vector. This intermediate gradient update parameter vector is written into the intermediate buffer of the parameter update module, stored in parameter component order, and output to the decomposition and projection operation process.

[0028] Optionally, the process of performing decomposition projection operations and cross-feasible region mapping on the intermediate parameter vector for gradient update in the parameter update module includes the following steps:

[0029] The gradient update intermediate parameter vector is read from the intermediate buffer, and the description information of the polyhedral parameter feasible region associated with the current mode identifier is read. The polyhedral parameter feasible region is parsed according to a preset constraint partitioning rule to generate several constraint subdomains, and a constraint index identifier is assigned to each constraint subdomain. According to the constraint index identifier order, the gradient update intermediate parameter vector is sequentially input into the projection operator corresponding to each constraint subdomain, and parameter clipping operations are performed within the constraint subdomain to generate intermediate projection results for the subdomains. Parameter overlay update operations are performed sequentially on the intermediate projection results for each subdomain to form the updated parameter vector after decomposition and projection. When a change in the mode identifier is detected, the polyhedral parameter feasible region associated with the mode identifier before the change and the polyhedral parameter feasible region associated with the mode identifier after the change are read. A cross-feasible region mapping operation is performed on the updated parameter vector to map the updated parameter vector into the polyhedral parameter feasible region corresponding to the mode identifier after the change. The final parameter vector is written to the parameter buffer and output to the lighting control output unit in time index order.

[0030] Optionally, the lighting control output module specifically includes:

[0031] Read the updated parameter vector from the parameter buffer in time index order; perform position parsing operation on the parameter vector according to the preset control parameter arrangement rules to generate brightness control component and color temperature control component; perform numerical range verification operation on brightness control component and color temperature control component to generate control parameter vector that meets the execution range constraint; write the control parameter vector into the control output buffer and send it to the lighting equipment in time index order to perform lighting control operation.

[0032] The beneficial effects of this invention are:

[0033] (1) This invention completes the spatial state analysis, mode determination and parameter update process in the edge computing node, reduces the dependence of lighting control decision on the central node or remote platform, shortens the data transmission and decision link, and enables the lighting control response to be updated synchronously with the changes in indoor spatial state. It is suitable for indoor scenarios with frequent human activities and rapid changes in ambient light.

[0034] (2) The present invention introduces a switching hysteresis domain and a switching cost potential function into the piecewise affine model to impose consistency constraints on the switching process of the lighting operation mode. When the spatial state change is in the critical range, the mode is kept stable. The mode identifier is updated when the switching conditions are met, thereby reducing the risk of sudden changes in control parameters caused by frequent mode switching and maintaining the continuity of lighting output.

[0035] (3) In the parameter update process, the present invention introduces the feasible domain of polyhedral parameters and its decomposition projection mechanism, so that the control parameters are always subject to constraints under different operating modes, and maintain the continuity of parameter evolution through cross-feasible domain mapping when the mode changes, which is conducive to achieving stable and controllable lighting parameter output under complex indoor environment conditions. Attached Figure Description

[0036] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0037] Figure 1 This is a flowchart of a fast response system for indoor space lighting based on edge computing proposed in this invention.

[0038] Figure 2 This is a schematic diagram illustrating the pattern consistency constraint implementation process of an edge computing-based rapid response system for indoor space lighting proposed in this invention.

[0039] Figure 3 This is a schematic diagram illustrating the process of updating and decomposing projection of mode constraint parameters in an indoor space lighting fast response system based on edge computing proposed in this invention. Detailed Implementation

[0040] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0041] refer to Figures 1-3 An edge computing-based rapid response system for indoor space lighting includes:

[0042] The spatial state acquisition module is used to acquire indoor spatial state data and form a spatial state vector;

[0043] The edge computing processing module is used to perform analytical operations on the spatial state vector on the local side of the indoor space and generate spatial state analysis results.

[0044] The pattern determination module is used to map the spatial state analysis results to multiple preset state intervals to form a candidate pattern set.

[0045] The mode consistency constraint module is used to introduce a switching hysteresis domain into the candidate mode set and calculate the mode switching cost potential function. When the spatial state change is within the switching hysteresis domain, the current mode identifier remains unchanged. When the spatial state change exceeds the switching hysteresis domain, the module determines whether to update the mode identifier based on the comparison result between the switching cost potential function and the preset cost threshold.

[0046] The parameter update module is used to perform gradient update operations on the corresponding parameter vector under the current mode identifier constraint, and project the updated parameter vector into the polyhedral parameter feasible region associated with the current mode identifier. The polyhedral parameter feasible region consists of several constraint subdomains. The parameter projection process performs decomposition projection operations according to the constraint subdomains. When the mode identifier changes, the parameter vector is mapped across the feasible region.

[0047] The lighting control output module is used to generate lighting control parameters based on the current mode identifier and the updated parameter vector, and output them to the lighting equipment.

[0048] In this embodiment, the spatial state acquisition module specifically includes:

[0049] Within the indoor space, spatial area identifiers are read according to a preset data acquisition configuration table. The sensor nodes corresponding to each spatial area identifier are polled and accessed to collect personnel presence status data, spatial illuminance data, and ambient light data with consistent time indices. A unified time index is added to each type of collected data to form raw status data items. A time index alignment verification operation is performed on the raw status data items, eliminating data items with inconsistent time indices and retaining those that meet the time index consistency condition to form an aligned status data set. Discrete state encoding is performed on the personnel presence status data within the aligned status data set, and numerical normalization is performed on the spatial illuminance data and ambient light data to generate corresponding coded state components and normalized state components. Vector concatenation is performed on the coded state components and normalized state components according to a preset state component arrangement order to form a spatial state vector. The spatial state vector is written to the state buffer of the edge computing node, stored in time index order, and output to the edge computing processing unit.

[0050] In this embodiment, performing polling access specifically includes:

[0051] After reading the spatial region identifier, data reading instructions are sent to each sensor node in sequence according to the sensor node address order defined in the preset acquisition configuration table. After each sensor node completes the data response, the returned data is written to the data cache unit of the corresponding region, and the current time index is recorded. After completing the data reading of all sensor nodes in the current region, the sequence of sensor nodes corresponding to the next spatial region identifier is entered according to the preset polling cycle, and the above access process is executed cyclically.

[0052] In this embodiment, performing a time index alignment check operation on the original state data item specifically includes:

[0053] When performing time index alignment verification on the original state data items, the time index value attached to each original state data item is first read and sorted according to the time index value. Then, the time index of each type of data in the same acquisition period is compared with the preset standard time index item by item to determine whether the time index difference is within the preset time deviation range. When the time index difference of a certain data item exceeds the preset time deviation range, the data item is marked as invalid data and removed from the original state data set. Only the data items that meet the time index consistency condition are retained to form the aligned state data set.

[0054] In this embodiment, performing discrete state encoding operations specifically includes:

[0055] Read the personnel presence status data from the aligned status data set and determine the encoding rules corresponding to the personnel presence status according to the preset status encoding table; perform status judgment operation on the personnel presence status data, assign the first encoding value when the detection value is equal to the presence flag, assign the second encoding value when the detection value is equal to the departure flag; assign the third encoding value when the detection value is within the preset abnormal range, and write the corresponding encoding result into the status encoding buffer; output the encoded status components according to the preset encoding arrangement order, as a component of the spatial status vector.

[0056] In this embodiment, the edge computing processing module specifically includes:

[0057] The system reads spatial state vectors sorted by time index from the state buffer. It performs an integrity check on these vectors, filtering out those with missing state components and retaining only those that meet the integrity criteria to form a valid state sequence. It then performs a difference operation on the spatial state vectors corresponding to adjacent time indices within the valid state sequence to calculate the spatial state change vector, which is written into the change buffer. Within the change buffer, it performs an amplitude determination operation on the spatial state change vector, generating a change determination tag set. Specifically, the amplitude determination operation calculates the absolute value of each component in the spatial state change vector and compares it component-by-component with a preset amplitude threshold. Finally, it selects the spatial state vectors that meet the comparison criteria. The components of a state vector are marked as valid change markers, and the remaining components are marked as invalid change markers, forming a change judgment marker set. Following the component index order of the spatial state vector, each state component and its corresponding change judgment marker at its index position in the change judgment marker set are read, forming an index-aligned joint input sequence. For each index position in the joint input sequence, the state component value and the change judgment marker value form a joint value pair. Based on a preset parsing rule table, the parsing component identifier corresponding to this joint value pair is determined. The parsing rule table maps the combination relationships between different state component values ​​and change judgment marker values ​​one-to-one. The parsing component identifiers are then combined according to a preset arrangement order to form the spatial state parsing result.

[0058] The spatial state resolution results are written to the resolution result buffer and output to the mode determination unit in time index order.

[0059] In this embodiment, the pattern determination module specifically includes:

[0060] Read the spatial state parsing results from the parsing result cache in time index order, perform interval mapping operation on the state components in the spatial state parsing results, map each state component to a preset state interval identifier, and form a state interval identifier vector.

[0061] A combination mapping operation is performed on the state interval identifier vector to generate a candidate pattern set according to preset interval combination rules. Specifically, the combination mapping operation includes reading the state interval identifiers corresponding to each state component according to the component order of the state interval identifier vector, forming an interval identifier combination input; performing a lookup mapping operation on the interval identifier combination input according to the preset interval combination rule table to generate a pattern identifier that corresponds one-to-one with the interval identifier combination; collecting the mapped pattern identifiers to form a candidate pattern set and outputting it. A consistency filtering operation is performed on the candidate pattern set to remove pattern identifiers that do not meet the continuity condition with the historical pattern identifiers corresponding to the current time index, retaining the pattern identifiers that meet the continuity condition to form a filtered pattern set; performing a priority sorting operation on the filtered pattern set, selecting the pattern identifier at the top of the sorting result as the current pattern identifier; writing the current pattern identifier into the pattern identifier buffer and outputting it to the pattern consistency constraint module in time index order.

[0062] In this embodiment, the process of introducing a switching hysteresis domain and calculating the mode switching cost potential function in the mode consistency constraint module includes the following steps:

[0063] The system reads the historical pattern identifier corresponding to the previous time index from the pattern identifier cache and the spatial state parsing result corresponding to the current time index from the parsing result cache. It then performs a hysteresis interval mapping operation on the state components in the spatial state parsing result according to a preset hysteresis parameter table, generating a corresponding hysteresis interval identifier for each state component. Specifically, the hysteresis interval mapping operation includes reading the values ​​of the state components at the current and previous time indices according to the hysteresis interval boundaries configured for each state component in the preset hysteresis parameter table; comparing the current and previous values ​​of the state component with the entry and release boundaries of the hysteresis interval, respectively, to determine whether the state component is in a hysteresis hold interval or a hysteresis release interval; and generating and outputting the corresponding hysteresis interval identifier for each state component based on the comparison results. A joint decision operation is performed on the hysteresis interval identifiers. A hysteresis hold flag is generated when all state components fall within the range corresponding to the hysteresis interval identifiers, and a hysteresis release flag is generated when any state component exceeds the range corresponding to the hysteresis interval identifiers. The joint decision operation reads the hysteresis interval identifiers corresponding to each state component one by one according to the component order of the hysteresis interval identifiers to form a hysteresis decision input sequence. A full consistency check operation is performed on the hysteresis decision input sequence. A hysteresis hold flag is generated when all hysteresis interval identifiers point to the hysteresis hold interval, and a hysteresis release flag is generated when any hysteresis interval identifier points to the hysteresis release interval.

[0064] When generating hysteresis release markers, for each mode identifier in the candidate mode set, a mode index parsing operation is performed on the current parameter vector to form a mode parameter set corresponding to each mode identifier. The mode index parsing operation specifically includes reading the current parameter vector from the parameter buffer and reading the mode parameter arrangement rules corresponding to the candidate mode set; according to the parameter index positions predefined for each mode identifier in the mode parameter arrangement rules, a position parsing operation is performed on the parameter vector, reading the parameter components at the specified index positions one by one; the read parameter components are reassembled in index order to form a mode parameter set corresponding to each mode identifier and output. Position parsing is the process of reading and reassembling parameter components from the parameter vector according to the index positions based on the preset parameter arrangement rules. A difference calculation operation is performed between the spatial state parsing result and each mode parameter set to generate a mode deviation vector; a weighted mapping operation is performed on the mode deviation vector to generate a cost component vector corresponding to each candidate mode identifier; an aggregation operation is performed on the cost component vector to calculate the mode switching cost potential function value corresponding to each mode identifier in the candidate mode set; the mode switching cost potential function value and the corresponding mode identifier are written into the cost buffer, stored in time index order, and output.

[0065] In this embodiment, calculating the mode switching cost potential function value corresponding to each mode identifier in the candidate mode set specifically includes:

[0066] When performing aggregation operations on the cost component vector, the values ​​of each cost component in the cost component vector are read sequentially according to the preset cost component arrangement order; the read cost component values ​​are combined according to the preset aggregation rules, which include performing a weighted summation operation on each cost component value, wherein the weight parameters corresponding to each cost component are given by a preset weight configuration table; the value obtained by the weighted summation operation is used as the mode switching cost potential function value of the corresponding mode identifier, and after establishing an association relationship with the mode identifier, it is written into the cost buffer.

[0067] In this embodiment, the process of determining whether to update the mode identifier based on the mode switching cost potential function in the mode consistency constraint module includes the following steps:

[0068] The system reads the set of mode switching cost potential function values ​​corresponding to the current time index from the cost buffer, and reads the historical mode identifier corresponding to the previous time index from the mode identifier buffer. When generating a hysteresis hold flag, it directly selects the historical mode identifier as the current mode identifier and writes the current mode identifier into the mode identifier buffer. When generating a hysteresis release flag, it reads the hold cost value corresponding to the historical mode identifier from the set of mode switching cost potential function values, and reads the switching cost value corresponding to the other candidate mode identifiers. It performs a difference comparison operation on the switching cost value and the hold cost value to generate a cost comparison result vector. It performs an item-by-item comparison operation on the cost comparison result vector and a preset cost threshold to filter candidate mode identifiers that meet the cost threshold constraint to form an allowed update mode set. When the allowed update mode set is not empty, it performs a minimum cost selection operation and selects the candidate mode identifier with the smallest cost value as the current mode identifier. When the allowed update mode set is empty, it maintains the historical mode identifier as the current mode identifier. It writes the current mode identifier into the mode identifier buffer and outputs it to the parameter update unit in time index order.

[0069] In this embodiment, the process of performing gradient update operation under the current mode identifier constraint in the parameter update module includes the following steps:

[0070] The parameter update module reads the parameter vector associated with the current mode identifier from the parameter cache, and parses the parameter vector according to a preset parameter arrangement rule to generate a parameter component sequence. It then reads the spatial state analysis result corresponding to the current time index from the analysis result cache, performs an association mapping operation on the spatial state analysis result and the parameter component sequence, and generates a target evaluation value. The analysis result cache refers to the spatial state analysis result. The association mapping operation specifically includes reading the analysis component identifiers or values ​​that are pre-established with corresponding parameters from the spatial state analysis result according to the component order of the parameter component sequence, forming a state-parameter aligned input sequence. Based on the preset association mapping rule, it performs a numerical mapping operation on each group of parameter components in the aligned input sequence and the corresponding spatial state analysis component, generating a corresponding intermediate evaluation component. Finally, it performs a numerical combination operation on each intermediate evaluation component according to a preset aggregation rule, forming a single target evaluation value and outputting it.

[0071] Gradient calculation is performed on the target evaluation value along the direction of the parameter component sequence to generate a gradient vector; scaling operation is performed on the gradient vector according to the preset step size parameter to generate an update step size vector; vector subtraction operation is performed on the parameter component sequence and the update step size vector to generate a gradient update intermediate parameter vector; the gradient update intermediate parameter vector is written into the intermediate buffer of the parameter update module, stored in the order of parameter components, and output to the decomposition projection operation process.

[0072] In this embodiment, the process of performing decomposition projection operation and cross-feasible region mapping on the intermediate parameter vector of gradient update in the parameter update module includes the following steps:

[0073] The gradient update intermediate parameter vector is read from the intermediate buffer, and the description information of the polyhedral parameter feasible region associated with the current mode identifier is read. The polyhedral parameter feasible region is parsed according to the preset constraint partitioning rules to generate several constraint subdomains, and a constraint index identifier is assigned to each constraint subdomain. According to the constraint index identifier order, the gradient update intermediate parameter vector is sequentially input into the projection operator corresponding to each constraint subdomain, and the parameter clipping operation within the constraint subdomain is performed to generate the subdomain projection intermediate result. The parameter overwrite update operation is performed sequentially on each subdomain projection intermediate result to form the updated parameter vector after decomposition and projection. The parameter overwrite update operation refers to reading the original parameter value at the same index position in the gradient update intermediate parameter vector according to the parameter index set corresponding to the current constraint subdomain, and writing the corresponding parameter value in the subdomain projection intermediate result into the index position, thereby replacing the original parameter value. When a change in the mode identifier is detected, the feasible domain of the polyhedral parameters associated with the mode identifier before the change and the feasible domain of the polyhedral parameters associated with the mode identifier after the change are read. A cross-feasibility domain mapping operation is performed on the updated parameter vector to map the updated parameter vector into the feasible domain of the polyhedral parameters corresponding to the changed mode identifier. The final parameter vector is written into the parameter buffer and output to the lighting control output unit in time index order.

[0074] In this embodiment, the lighting control output module specifically includes:

[0075] The updated parameter vector is read from the parameter buffer in chronological order. Position parsing is performed on the parameter vector according to a preset control parameter arrangement rule to generate brightness control components and color temperature control components. Numerical range verification is performed on the brightness and color temperature control components to generate a control parameter vector that meets the execution range constraints. The numerical range verification compares the brightness and color temperature control components with preset minimum and maximum execution values, respectively. If a value exceeds the execution range, it is adjusted to the corresponding boundary value, thus generating a control parameter vector that meets the execution range constraints. The control parameter vector is written to the control output buffer and sent to the lighting equipment in chronological order to perform lighting control operations.

[0076] Example 1:

[0077] To verify the feasibility of this invention in practice, it was applied to an indoor open office area of ​​a large public building. This area, approximately 3200 square meters, includes 168 sets of zoned lighting fixtures and is equipped with personnel presence sensors, illuminance sensors, and ambient light acquisition units. Typical problems in this scenario include: frequent but uneven personnel activity, significant fluctuations in external natural light over time, lag in response to traditional centralized control systems, frequent flickering or excessive switching of lights affecting comfort, and high energy consumption. In actual operation, the spatial state acquisition module continuously collects personnel presence data, spatial illuminance data, and ambient light data, forming a spatial state vector. The edge computing processing module performs differential analysis and change determination on the spatial state vector locally, generating a spatial state analysis result. The mode determination module forms a candidate mode set based on state interval mapping rules. Subsequently, the mode consistency constraint module introduces a switching hysteresis domain, maintaining the current mode unchanged when the state fluctuation amplitude is small, and calculating the mode switching cost potential function only when the change exceeds the hysteresis interval. The parameter update module performs gradient updates under the current mode constraints and projects the update results to the corresponding polyhedral parameter feasible region. The lighting control output module generates brightness and color temperature control values ​​based on the updated parameter vector and sends them to the lighting equipment for adjustment. Table 1: Comparison of Indoor Lighting Control Effects

[0078] Daily mode switching frequency 82 times 37 times Brightness fluctuation range 18.6% 7.2% Average illuminance stability 0.78 0.92 Daily power consumption per unit area 3.84kWh 3.15kWh Average response time 480ms 135ms Standard deviation of illuminance on sunny days 12.4lx 5.6lx Standard deviation of illuminance on cloudy days 18.7lx 8.9lx Personnel misjudgment rate 14.2% 3.8% Subjective satisfaction rating 3.6 4.5

[0079] As shown in Table 1, the system of this invention exhibits significant numerical differences in several key indicators. Firstly, regarding the average daily number of mode switching times, the traditional control system switches between 82 and 37 times, a reduction of 45 times, or 54.9%. This indicates that the system effectively reduces unnecessary mode changes under conditions of frequent personnel movement and fluctuating natural light. The brightness fluctuation range decreased from 18.6% to 7.2%, a reduction of 11.4 percentage points, indicating more stable lamp output and avoiding frequent changes in brightness within a short period.

[0080] Regarding the illuminance stability index, the traditional system has an index of 0.78, while the system of this invention reaches 0.92, an improvement of 0.14, representing a stability improvement of approximately 17.9%. This reflects a stronger ability to maintain illuminance within the set range. The average daily energy consumption per unit area decreased from 3.84 kWh to 3.15 kWh, saving 0.69 kWh per day. Over 30 days, this translates to a cumulative saving of 20.7 kWh, representing an energy saving rate of 17.9%.

[0081] In terms of response performance, the average response latency decreased from 480ms to 135ms, a reduction of 345ms, or 71.9%, indicating that the edge-side processing structure completes parsing and judgment locally, significantly shortening the control link time.

[0082] Under different meteorological conditions, the standard deviation of illuminance on sunny days decreased from 12.4 lx to 5.6 lx, a reduction of 6.8 lx; and on cloudy days it decreased from 18.7 lx to 8.9 lx, a reduction of 9.8 lx. This indicates that the system can maintain a smaller degree of illuminance dispersion under drastic changes in light intensity.

[0083] The human error rate decreased from 14.2% to 3.8%, a reduction of 10.4 percentage points, or approximately 73.2%. Subjective satisfaction rating improved from 3.6 to 4.5, an increase of 0.9 points, indicating more natural lighting changes. Overall, the data shows that this invention demonstrates significant numerical improvements in mode stability, energy consumption control, response speed, and illuminance consistency.

[0084] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An edge computing based indoor space lighting fast response system, characterized in that, include: The spatial state acquisition module is used to acquire indoor spatial state data and form a spatial state vector; The edge computing processing module is used to perform analytical operations on the spatial state vector on the local side of the indoor space and generate spatial state analysis results. The pattern determination module is used to map the spatial state analysis results to multiple preset state intervals to form a candidate pattern set. The mode consistency constraint module is used to introduce a switching hysteresis domain into the candidate mode set and calculate the mode switching cost potential function. When the spatial state change is within the switching hysteresis domain, the current mode identifier remains unchanged. When the spatial state change exceeds the switching hysteresis domain, the module determines whether to update the mode identifier based on the comparison result between the switching cost potential function and the preset cost threshold. The parameter update module is used to perform gradient update operations on the corresponding parameter vector under the current mode identifier constraint, and project the updated parameter vector into the polyhedral parameter feasible region associated with the current mode identifier. The polyhedral parameter feasible region consists of several constraint subdomains. The parameter projection process performs decomposition projection operations according to the constraint subdomains. When the mode identifier changes, the parameter vector is mapped across the feasible region. The lighting control output module is used to generate lighting control parameters based on the current mode identifier and the updated parameter vector, and output them to the lighting equipment.

2. The fast response system for indoor space lighting based on edge computing according to claim 1, characterized in that, The spatial status acquisition module specifically includes: Within the indoor space, spatial area identifiers are read according to a preset data acquisition configuration table. The sensor nodes corresponding to each spatial area identifier are polled and accessed to collect personnel presence status data, spatial illuminance data, and ambient light data with consistent time indices. A unified time index is added to each type of collected data to form raw status data items. A time index alignment verification operation is performed on the raw status data items, eliminating data items with inconsistent time indices and retaining those that meet the time index consistency condition to form an aligned status data set. Discrete state encoding is performed on the personnel presence status data within the aligned status data set, and numerical normalization is performed on the spatial illuminance data and ambient light data to generate corresponding coded state components and normalized state components. Vector concatenation is performed on the coded state components and normalized state components according to a preset state component arrangement order to form a spatial state vector. The spatial state vector is written to the state buffer of the edge computing node, stored in time index order, and output to the edge computing processing unit.

3. The fast response system for indoor space lighting based on edge computing according to claim 2, characterized in that, The edge computing processing module specifically includes: The system reads spatial state vectors sorted by time index from the state buffer, performs integrity verification on the spatial state vectors, filters out spatial state vectors with missing state components, and retains spatial state vectors that meet the integrity conditions to form a valid state sequence; performs difference operation on spatial state vectors corresponding to adjacent time indices in the valid state sequence to calculate the spatial state change vector, and writes the spatial state change vector into the change buffer; performs amplitude determination operation on the spatial state change vector in the change buffer to generate a change determination mark set; performs joint analysis operation on the spatial state vector and the change determination mark set according to preset analysis rules to generate spatial state analysis results; writes the spatial state analysis results into the analysis result buffer and outputs them to the mode determination unit in time index order.

4. The fast response system for indoor space lighting based on edge computing according to claim 3, characterized in that, The pattern determination module specifically includes: Read the spatial state parsing results from the parsing result cache in chronological order, perform interval mapping operation on the state components in the spatial state parsing results, map each state component to a preset state interval identifier, and form a state interval identifier vector. Perform a combination mapping operation on the state interval identifier vector to generate a candidate pattern set according to the preset interval combination rules; perform a consistency filtering operation on the candidate pattern set to remove pattern identifiers that do not meet the continuity condition with the historical pattern identifiers corresponding to the current time index, and retain the pattern identifiers that meet the continuity condition to form a filtered pattern set; perform a priority sorting operation on the filtered pattern set, and select the pattern identifier at the top of the sorting result as the current pattern identifier; write the current pattern identifier into the pattern identifier buffer, and output it to the pattern consistency constraint module in the order of time index.

5. The edge computing based indoor space lighting fast response system according to claim 4, wherein, The process of introducing the switching hysteresis domain and calculating the mode switching cost potential function in the mode consistency constraint module includes the following steps: Read the historical pattern identifier corresponding to the previous time index from the pattern identifier cache, and read the spatial state parsing result corresponding to the current time index from the parsing result cache; Perform hysteresis interval mapping operation on the state components in the spatial state analysis result according to the preset hysteresis parameter table to generate corresponding hysteresis interval identifiers for each state component. Perform a joint decision operation on the hysteresis interval identifier. Generate a hysteresis hold flag when all state components fall within the range corresponding to the hysteresis interval identifier, and generate a hysteresis release flag when any state component exceeds the range corresponding to the hysteresis interval identifier. When generating the hysteresis release flag, for each mode identifier in the candidate mode set, a mode index parsing operation is performed on the current parameter vector to form a mode parameter set that corresponds one-to-one with each mode identifier; a difference calculation operation is performed on the spatial state parsing result and each mode parameter set to generate a mode deviation vector; a weighted mapping operation is performed on the mode deviation vector to generate a cost component vector that corresponds one-to-one with each candidate mode identifier; an aggregation operation is performed on the cost component vector to calculate the mode switching cost potential function value corresponding to each mode identifier in the candidate mode set; the mode switching cost potential function value and the corresponding mode identifier are written into the cost buffer, stored in time index order and output.

6. The edge computing based indoor space lighting fast response system according to claim 5, wherein, The process of determining whether to update the mode identifier based on the mode switching cost potential function in the mode consistency constraint module includes the following steps: The system reads the set of mode switching cost potential function values ​​corresponding to the current time index from the cost buffer, and reads the historical mode identifier corresponding to the previous time index from the mode identifier buffer. When generating a hysteresis hold flag, the system directly selects the historical mode identifier as the current mode identifier and writes the current mode identifier into the mode identifier buffer. When generating a hysteresis release flag, the system reads the hold cost value corresponding to the historical mode identifier from the set of mode switching cost potential function values, and reads the switching cost value corresponding to the other candidate mode identifiers. The system performs a difference comparison operation on the switching cost value and the hold cost value to generate a cost comparison result vector. The system performs an item-by-item comparison operation on the cost comparison result vector and a preset cost threshold to filter candidate mode identifiers that meet the cost threshold constraint to form a set of allowed update modes. When the set of update patterns is not empty, perform the minimum cost selection operation and select the candidate pattern identifier with the smallest cost value as the current pattern identifier. When the set of allowed update patterns is empty, the historical pattern identifier is maintained as the current pattern identifier; Write the current pattern identifier to the pattern identifier buffer and output it to the parameter update unit in time index order.

7. The edge computing based indoor space lighting fast response system according to claim 6, wherein, The process of performing gradient update operations under the current mode identifier constraint in the parameter update module includes the following steps: The parameter vector associated with the current mode identifier is read from the parameter cache of the parameter update module, and the parameter vector is parsed according to the preset parameter arrangement rules to generate a parameter component sequence; the spatial state parsing result corresponding to the current time index is read from the parsing result cache, and the spatial state parsing result and the parameter component sequence are associated with a mapping operation to generate the target evaluation value; the parsing result cache refers to the spatial state parsing result. Gradient calculation is performed on the target evaluation value along the direction of the parameter component sequence to generate a gradient vector; The gradient vector is scaled according to a preset step size parameter to generate an updated step size vector; the parameter component sequence and the updated step size vector are subtracted to generate an intermediate gradient update parameter vector; the intermediate gradient update parameter vector is written into the intermediate buffer of the parameter update module, stored in the order of parameter components, and output to the decomposition projection operation process.

8. The edge computing based indoor space lighting fast response system according to claim 7, wherein, The parameter update module performs decomposition projection operations and cross-feasible region mapping on the intermediate parameter vector for gradient update, including the following steps: The gradient update intermediate parameter vector is read from the intermediate buffer, and the description information of the polyhedral parameter feasible region associated with the current mode identifier is read. The polyhedral parameter feasible region is parsed according to a preset constraint partitioning rule to generate several constraint subdomains, and a constraint index identifier is assigned to each constraint subdomain. According to the constraint index identifier order, the gradient update intermediate parameter vector is sequentially input into the projection operator corresponding to each constraint subdomain, and parameter clipping operations are performed within the constraint subdomain to generate intermediate projection results for the subdomains. Parameter overlay update operations are performed sequentially on the intermediate projection results for each subdomain to form the updated parameter vector after decomposition and projection. When a change in the mode identifier is detected, the polyhedral parameter feasible region associated with the mode identifier before the change and the polyhedral parameter feasible region associated with the mode identifier after the change are read. A cross-feasible region mapping operation is performed on the updated parameter vector to map the updated parameter vector into the polyhedral parameter feasible region corresponding to the mode identifier after the change. The final parameter vector is written to the parameter buffer and output to the lighting control output unit in time index order.

9. The edge computing based indoor space lighting fast response system according to claim 8, wherein, The lighting control output module specifically includes: Read the updated parameter vector from the parameter buffer in time index order; perform position parsing operation on the parameter vector according to the preset control parameter arrangement rules to generate brightness control component and color temperature control component; perform numerical range verification operation on brightness control component and color temperature control component to generate control parameter vector that meets the execution range constraint; write the control parameter vector into the control output buffer and send it to the lighting equipment in time index order to perform lighting control operation.

Images