Intelligent classroom control system based on Internet of Things technology

Abstract

The invention discloses an intelligent classroom control system based on an Internet of Things technology, belongs to the technical field of classroom control, and remarkably improves the matching precision of a teaching environment and a learning activity by deeply integrating Internet of Things perception and a situation self-adaption technology. The system can analyze essential requirements of a teaching scene in real time, and based on an environment adaptive mechanism of teaching intention, not only eliminates operation burden of frequent manual adjustment of equipment in a traditional classroom, but also effectively improves student knowledge absorption efficiency and classroom participation degree by optimizing a combination relationship of environment elements such as sound, light and heat, and improves teaching experience. Equipment operation data and environment feedback are continuously collected through an Internet of Things sensor network, an energy consumption fluctuation rule is predicted in combination with a teaching schedule, the risk of abnormal shutdown of the equipment is greatly reduced, the life cycle of the teaching equipment is prolonged, and unnecessary energy consumption is reduced through an intelligent strategy. A teaching manager can intuitively master the use characteristics of different classrooms through the multi-dimensional data board, and a decision basis is provided for optimizing a course arrangement scheme and improving space design.

Description

Technical Field

[0001] This invention belongs to the field of classroom control technology, specifically an intelligent classroom control system based on Internet of Things (IoT) technology. Background Technology

[0002] A classroom control system is an integrated technological solution designed to intelligently manage and control various classroom equipment and the environment through a central control platform. This system typically integrates multiple devices such as lighting, air conditioning, projectors, screens, audio equipment, curtains, and security monitoring. Users can operate it uniformly via touchscreen, computer, mobile app, or voice commands. Its core functions include equipment on / off control, scene mode presets (such as class, projection, discussion, energy saving, etc.), environmental parameter monitoring and adjustment (such as temperature, humidity, and illuminance), equipment status monitoring and fault alarms, and automatic execution of timetable linkages. Classroom control systems not only improve the comfort, convenience, and energy efficiency of the teaching environment but also effectively simplify management and maintenance, providing fundamental support for the construction of modern smart classrooms.

[0003] However, existing teaching environment control systems mostly rely on fixed thresholds to trigger simple device linkages, which cannot deeply understand the dynamic relationship between teaching activities and environmental parameters. At the same time, they lack the ability to learn from long-term operating data, resulting in a mismatch between device response and actual teaching needs, making it difficult to achieve a balance between teaching effectiveness and resource consumption. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent classroom control system based on Internet of Things (IoT) technology in order to solve the problems mentioned above.

[0005] The technical solution adopted in this invention is as follows: an intelligent classroom control system based on Internet of Things (IoT) technology, the system comprising: a perception layer module, a network layer module, a context-aware and adaptive learning environment engine, a control layer module, and an application layer module;

[0006] The context-aware and adaptive learning environment engine is internally configured with a data aggregation and preprocessing submodule, a context understanding and state recognition submodule, a policy generation and optimization submodule, an instruction conversion and coordination submodule, and a feedback learning and system adaptation submodule.

[0007] The environmental sensor output of the perception layer module is connected to the IoT gateway data acquisition port of the network layer module via the Zigbee wireless protocol.

[0008] The protocol conversion gateway command output terminal of the network layer module is connected to the gigabit Ethernet controller of the edge decision unit of the control layer module via an Ethernet physical interface.

[0009] The policy output interface of the context-aware and adaptive learning environment engine is bound to the MQTT topic subscription channel of the control layer execution mechanism drive cluster to receive control command streams in JSON format in real time.

[0010] The PWM dimming signal output terminal of the control layer module directly drives the constant current drive board of the lighting system;

[0011] The WebSocket server of the application layer module's visual management interface continuously monitors the real-time device status data stream forwarded by the network layer traffic scheduling controller, and simultaneously establishes a secure data pipeline with the campus academic affairs system's timetable interface through the OAuth2.0 authorization channel.

[0012] In a preferred embodiment, the perception layer module internally comprises three core components: an environmental sensor network, an equipment status monitoring unit, and a human-computer interaction terminal. The environmental sensor network deploys light sensors, temperature and humidity sensors, carbon dioxide concentration sensors, noise sensors, smoke sensors, and infrared human body detectors, forming a spatial three-dimensional monitoring matrix to capture real-time classroom light intensity distribution, air quality, and thermal maps of human activity. The equipment status monitoring unit integrates a power parameter acquisition chip and an equipment communication protocol parser, continuously tracking the projector's operating power curve, the air conditioner compressor's operating cycle, the opening and closing status of doors and windows, and equipment operating parameters, accurately recording energy consumption data through pulse counting technology.

[0013] In a preferred embodiment, the network layer module comprises a core architecture consisting of a heterogeneous protocol conversion gateway, a wireless mesh networking node, and a traffic scheduling controller. The heterogeneous protocol conversion gateway is equipped with a Zigbee-WiFi dual-mode communication chip and a Modbus-TCP protocol stack, enabling seamless integration between RS485 bus devices and IP networks. The wireless mesh networking node employs a three-band dynamic switching technology: it carries device control command streams in the 2.4GHz band, transmits high-definition camera video streams in the 5.8GHz band, and maintains sensor data backhaul in the sub-1GHz band, avoiding channel congestion through a time slot allocation algorithm.

[0014] In a preferred embodiment, the data aggregation and preprocessing submodule includes a distributed data acquisition interface, a spatiotemporal alignment processor, a multimodal data fusion unit, an anomaly detection filter, and a standardized output channel. A unified data buffer is established to receive raw signal streams from ambient light sensors, temperature sensors, humidity sensors, CO2 concentration sensors, noise sensors, infrared human body detection modules, and power monitoring equipment sensors. A timestamp calibration mechanism is used to eliminate timing misalignment problems caused by differences in sampling frequencies of different devices, and a sliding window technique is used to process discrete data continuously.

[0015] In a preferred embodiment, the context understanding and state recognition submodule first establishes a multi-source feature extraction network to extract temporal features, spatial distribution features, and cross-modal association features from a standardized dataset. The multi-dimensional vector output from the feature engineering layer is input into a two-layer hybrid inference model. The bottom layer employs a rule-based inference engine based on a knowledge graph; the top layer deploys a lightweight deep neural network, using historical labeled data to train an end-to-end context classification model. An attention mechanism dynamically adjusts the contribution weights of different sensor features to achieve joint probability prediction of environmental states and teaching activity types. To improve real-time performance, the module embeds an incremental learning framework. When a new device is detected or the teaching process changes, local model fine-tuning is automatically triggered to ensure the continuous adaptability of state recognition.

[0016] The formula for calculating context fusion weights is defined as follows:

[0017] W_ij=σ×F_hi+τ×C_rt×S_dj;

[0018] Where W_ij represents the feature weight coefficient of the i-th type of sensor in the j-th teaching scenario, σ is the historical accuracy decay factor of the sensor, calculated exponentially based on the calibration error rate of the past 7 days; F_hi represents the average feature contribution of the sensor in the same historical scenario, obtained through sliding window statistics; C_rt is the real-time confidence score in the current environment, calculated inversely proportional to the device health status code and the standard deviation of data fluctuation; S_dj is the scenario adaptation parameter, dynamically assigned according to the preset teaching activity priority matrix.

[0019] The dynamic confidence adjustment formula is:

[0020] D_t=D_0×e^(-λt)+μ×A_c;

[0021] Where D_t represents the confidence level of the environmental state at time t, D_0 is the initial confidence benchmark value, λ is the time decay coefficient, which is dynamically adjusted according to the intensity of classroom use; A_c is the anomaly detection counter, which triggers the μ coefficient multiplication mechanism when more than 3 consecutive feature contradictions occur.

[0022] The attention allocation formula innovatively incorporates perceptual variables from different teaching stages, namely:

[0023] A_k = T_p × L_s + E_m × R_f,

[0024] Where A_k represents the attention weight of the k-th feature channel, T_p is the teaching progress parameter, which changes linearly according to the proportion of remaining time in the timetable; L_s is the learning effect correlation, which is obtained by regression analysis of historical data to obtain the correlation coefficient between each environmental parameter and student attention; E_m represents the equipment energy consumption sensitivity, which is determined by the product of the current electricity price period and the equipment's rated power; R_f is the real-time reliability score, which is calculated by combining the equipment's running time and the most recent maintenance cycle.

[0025] In a preferred embodiment, the strategy generation and optimization submodule comprises four main parts: a teaching rule knowledge base, a real-time environment optimization engine, a multi-objective trade-off matrix, and a strategy verification loop. The knowledge base contains a set of priority rules defined by educational experts; the real-time optimization engine employs an improved multi-agent collaborative algorithm, abstracting air conditioning, lighting, and projection equipment as independent decision-making units, and solves for the global optimal solution through a distributed game model, while introducing a time decay factor to dynamically adjust the balance weight between short-term comfort and long-term energy consumption; the multi-objective trade-off matrix establishes a conflict relationship graph of temperature deviation, illumination uniformity, noise suppression rate, and equipment energy consumption ratio, and uses the analytic hierarchy process (AHP) to calculate the dynamic priority coefficients of each indicator at different teaching stages.

[0026] In a preferred embodiment, the instruction conversion and coordination submodule consists of four parts: an instruction parsing engine, a multi-protocol adaptation matrix, a timing orchestration controller, and a conflict resolution unit. The instruction parsing engine uses semantic segmentation technology to decompose abstract strategies into atomic operation instructions that can be executed by the device.

[0027] In a preferred embodiment, the feedback learning and system adaptation submodule is internally configured with four core components: a multi-source feedback acquisition interface, an effect quantification evaluation model, a parameter dynamic adjustment engine, and a knowledge graph update network.

[0028] In a preferred embodiment, the control layer module comprises two main functional units: an edge decision unit and an actuator drive cluster. The edge decision unit is equipped with a dual-core real-time processor, runs an adaptive PID control algorithm library, and dynamically adjusts control parameters according to the rate of environmental change. The actuator drive cluster is divided into a lighting dimming group, an air conditioning group, and a teaching equipment group. The dimming group is equipped with a PWM dimming driver board and a DMX512 protocol conversion module to achieve precise control of the color temperature and brightness of the LED light strips.

[0029] In a preferred embodiment, the application layer module is configured with a three-dimensional system comprising a cloud-based decision engine, a visual management interface, and a cross-system interface. The cloud-based decision engine deploys an LSTM neural network model, analyzes three years of historical environmental data to establish a seasonal pattern prediction model, and combines real-time meteorological API data to pre-adjust the classroom microclimate two hours in advance.

[0030] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0031] 1. This invention significantly improves the matching accuracy between the teaching environment and learning activities by deeply integrating IoT sensing and contextual adaptation technologies. The system can analyze the essential needs of the teaching scenario in real time. For example, when the teacher starts the interactive discussion mode, it automatically adjusts the light color temperature to a cool white light range that promotes active thinking, simultaneously reduces the air conditioner fan speed to reduce noise interference, and switches the projection screen to a multi-window split-screen display state, so that environmental parameters and teaching activities are dynamically coordinated. This environment adaptation mechanism based on teaching intent not only eliminates the operational burden of frequently manually adjusting equipment in traditional classrooms, but also effectively improves students' knowledge absorption efficiency and classroom participation by optimizing the combination of environmental elements such as sound, light, and heat.

[0032] 2. In this invention, an IoT sensor network continuously collects equipment operation data and environmental feedback, and combines this with the teaching schedule to predict energy consumption fluctuations. For example, the ventilation system is pre-activated before experimental classes to maintain air cleanliness, and automatically enters a low-power inspection mode after class. This predictive maintenance capability significantly reduces the risk of abnormal equipment downtime, extends the lifespan of teaching equipment, and reduces unnecessary energy consumption through intelligent strategies. Teaching administrators can intuitively understand the usage characteristics of different classrooms through multi-dimensional data dashboards, providing a basis for decision-making to optimize scheduling plans and improve space design. Attached Figure Description

[0033] Figure 1 This is an overall system block diagram of the present invention;

[0034] Figure 2 This is a block diagram of the context-aware and adaptive learning environment engine system in this invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0036] Example:

[0037] Reference Figure 1-2 A smart classroom control system based on Internet of Things (IoT) technology includes: a perception layer module, a network layer module, a context-aware and adaptive learning environment engine, a control layer module, and an application layer module.

[0038] The context-aware and adaptive learning environment engine has internal sub-modules for data aggregation and preprocessing, context understanding and state recognition, policy generation and optimization, instruction conversion and coordination, and feedback learning and system adaptation.

[0039] The environmental sensor output of the perception layer module is connected to the IoT gateway data acquisition port of the network layer module via the Zigbee wireless protocol.

[0040] The protocol conversion gateway command output of the network layer module is connected to the gigabit Ethernet controller of the edge decision unit of the control layer module via the Ethernet physical interface.

[0041] The policy output interface of the context-aware and adaptive learning environment engine is bound to the MQTT topic subscription channel of the control layer execution mechanism driving cluster to receive control command streams in JSON format in real time.

[0042] The PWM dimming signal output of the control layer module directly drives the constant current driver board of the lighting system;

[0043] The WebSocket server in the application layer module's visual management interface continuously monitors the real-time device status data stream forwarded by the network layer traffic scheduling controller, and simultaneously establishes a secure data pipeline with the campus academic affairs system's timetable interface through the OAuth2.0 authorization channel.

[0044] The perception layer module comprises three core components: an environmental sensor network, an equipment status monitoring unit, and a human-computer interaction terminal. The environmental sensor network deploys light sensors, temperature and humidity sensors, carbon dioxide concentration sensors, noise sensors, smoke sensors, and infrared human detection devices, forming a three-dimensional spatial monitoring matrix to capture real-time classroom light intensity distribution, air quality, and thermal maps of human activity. The equipment status monitoring unit integrates a power parameter acquisition chip and an equipment communication protocol parser, continuously tracking equipment operating parameters such as the projector's power consumption curve, the air conditioner compressor's operating cycle, and the opening and closing status of doors and windows, accurately recording energy consumption data through pulse counting technology. The human-computer interaction terminal is equipped with a capacitive touchscreen and an array microphone.

[0045] The network layer module consists of a heterogeneous protocol conversion gateway, wireless mesh networking nodes, and a traffic scheduling controller. The heterogeneous protocol conversion gateway is equipped with a Zigbee-WiFi dual-mode communication chip and a Modbus-TCP protocol stack, enabling seamless integration between RS485 bus devices and IP networks. The wireless mesh networking nodes employ three-band dynamic switching technology: the 2.4GHz band carries device control command streams, the 5.8GHz band transmits high-definition camera video streams, and the sub-1GHz band maintains sensor data backhaul, avoiding channel congestion through a time slot allocation algorithm. The traffic scheduling controller deploys a deep packet inspection engine, implementing differentiated QoS policies based on service type. It ensures millisecond-level low-latency transmission of projector HDMI-CEC control commands and allocates elastic bandwidth for batch uploading of energy consumption statistics.

[0046] The data aggregation and preprocessing submodule includes a distributed data acquisition interface, a spatiotemporal alignment processor, a multimodal data fusion unit, an anomaly detection filter, and a standardized output channel. A unified data buffer is established to receive raw signal streams from ambient light sensors, temperature sensors, humidity sensors, CO2 concentration sensors, noise sensors, infrared human detection modules, and power monitoring equipment sensors. A timestamp calibration mechanism eliminates timing misalignment issues caused by differences in sampling frequencies between different devices. A sliding window technique is used to process discrete data continuously. An adaptive filtering algorithm is deployed to remove spike noise caused by electromagnetic interference or equipment failure. To address potential data conflicts when multiple sensors measure the same physical quantity, a dynamic weighted fusion algorithm based on Bayesian inference is designed. This algorithm dynamically adjusts the fusion weights of each data source by combining historical reliability assessment values ​​of sensors, current environmental confidence factors, and equipment health status indices. Finally, a standardized multidimensional dataset with temporal consistency, spatial integrity, and quality assurance is generated, providing high-quality input for the upper-level context understanding module.

[0047] The core calculation formula of the three-dimensional weight calculation model in the dynamic weighted fusion algorithm is:

[0048] W_i=α×C_ri+β×R_hi+γ×S_ci;

[0049] Where W_i represents the comprehensive weighting coefficient of the i-th sensor, C_ri represents the real-time confidence factor, which is obtained by analyzing the sensor signal fluctuation range, data consistency with adjacent sensors, and device self-test status; R_hi is the historical reliability index, dynamically calculated based on the sensor's failure rate, calibration deviation records, and maintenance cycle compliance rate over the past 30 days; S_ci is the context-related parameter, obtained by mapping the currently identified teaching scenario type to the scenario importance weight table of each sensor; α, β, and γ are the real-time factor, stability factor, and scenario adaptation factor, respectively, which constitute a dynamically adjustable parameter group. The proportions are automatically adjusted according to the teaching time period, device network load status, and system operation mode to ensure that historical reliability assessment is emphasized during the morning's intensive device startup phase, real-time data accuracy is prioritized during classroom teaching periods, and scenario adaptation capabilities are strengthened under special activity modes.

[0050] The context understanding and state recognition submodule first establishes a multi-source feature extraction network to extract temporal features, spatial distribution features, and cross-modal correlation features from a standardized dataset. Spectral analysis is used to capture sudden fluctuations in noise signals, and optical flow algorithms are employed to analyze the density of human movement trajectories from infrared sensors. The multi-dimensional vectors output from the feature engineering layer are input into a two-layer hybrid inference model. The bottom layer uses a rule-based inference engine based on a knowledge graph; the top layer deploys a lightweight deep neural network, using historical labeled data to train an end-to-end context classification model. An attention mechanism dynamically adjusts the contribution weights of different sensor features to achieve joint probability prediction of environmental states and teaching activity types. To improve real-time performance, the module embeds an incremental learning framework that automatically triggers local model fine-tuning when new devices are detected or the teaching process changes, ensuring continuous adaptability of state recognition.

[0051] The formula for calculating context fusion weights is defined as follows:

[0052] W_ij=σ×F_hi+τ×C_rt×S_dj;

[0053] Where W_ij represents the feature weight coefficient of the i-th type of sensor in the j-th teaching scenario, σ is the historical accuracy decay factor of the sensor, calculated exponentially based on the calibration error rate of the past 7 days; F_hi represents the average feature contribution of the sensor in the same historical scenario, obtained through sliding window statistics; C_rt is the real-time confidence score in the current environment, calculated inversely proportional to the device health status code and the standard deviation of data fluctuation; S_dj is the scenario adaptation parameter, dynamically assigned according to the preset teaching activity priority matrix.

[0054] The dynamic confidence adjustment formula is:

[0055] D_t=D_0×e^(-λt)+μ×A_c;

[0056] Where D_t represents the confidence level of the environmental state at time t, D_0 is the initial confidence benchmark value, λ is the time decay coefficient, which is dynamically adjusted according to the intensity of classroom use, and A_c is the anomaly detection counter, which triggers the μ coefficient multiplication mechanism when more than 3 consecutive feature contradictions occur. The attention allocation formula innovatively introduces the perceptual variable of the teaching stage, namely A_k=T_p×L_s+E_m×R_f, where A_k represents the attention weight of the k-th feature channel, T_p is the teaching progress parameter, which changes linearly according to the proportion of remaining time in the timetable, L_s is the correlation degree of learning effect, which is obtained by regression analysis of historical data to obtain the correlation coefficient between each environmental parameter and student attention, E_m represents the equipment energy consumption sensitivity, which is determined by the product of the current electricity price period and the rated power of the equipment, and R_f is the real-time reliability score, which is calculated by combining the equipment running time and the most recent maintenance cycle.

[0057] The strategy generation and optimization submodule comprises four main parts: a teaching rule knowledge base, a real-time environment optimization engine, a multi-objective trade-off matrix, and a strategy verification loop. The knowledge base contains a set of priority rules defined by educational experts. The real-time optimization engine employs an improved multi-agent collaborative algorithm, abstracting equipment such as air conditioners, lighting, and projectors into independent decision-making units. It solves for the global optimum through a distributed game model, while introducing a time decay factor to dynamically adjust the balance between short-term comfort and long-term energy consumption. The multi-objective trade-off matrix establishes a conflict relationship graph for 12 quantitative indicators, including temperature deviation, illumination uniformity, noise suppression rate, and equipment energy consumption ratio. It uses the analytic hierarchy process (AHP) to calculate the dynamic priority coefficients of each indicator at different teaching stages. The strategy verification loop uses digital twin technology to pre-simulate the execution effect of the control strategy and iteratively optimizes parameters based on satisfaction feedback data from similar historical scenarios, ensuring that the generated strategy simultaneously meets the requirements for real-time response speed and long-term teaching effectiveness improvement.

[0058] The multimodal target fusion formula is:

[0059] Ψ = ω_t × ΔT_norm + ω_l × Φ_light + ω_e × E_ratio × ξ(t). Where Ψ represents the comprehensive optimization target value; ω_t is the temperature weighting coefficient, determined by the teaching scenario type mapping table and the current indoor-outdoor temperature gradient; ΔT_norm is the normalized temperature deviation, using a hyperbolic tangent function to compress the difference between the actual temperature and the set value; ω_l represents the dynamic weighting of illumination, adaptively adjusted according to the time slots and natural light intensity; Φ_light calculates the lighting uniformity index, fusing the spatial distribution variance of the illuminance sensor network with the color temperature coordination coefficient; E_ratio is the equipment energy efficiency ratio, dynamically updated by the ratio of real-time power monitoring values ​​to the equipment energy efficiency benchmark value; ξ(t) is the time influence factor, decaying exponentially according to the remaining teaching time, ensuring that the energy consumption weight is gradually reduced near the end of class to prioritize comfort. This formula creatively introduces a spatiotemporal coupling parameter system, achieving deep collaborative optimization of teaching needs, environmental parameters, and equipment status.

[0060] The instruction conversion and coordination submodule consists of four parts: an instruction parsing engine, a multi-protocol adaptation matrix, a timing orchestration controller, and a conflict resolution unit. The instruction parsing engine uses semantic segmentation technology to decompose abstract strategies into atomic operation instructions executable by the devices. The multi-protocol adaptation matrix internally incorporates conversion logic for over twenty industrial protocols, including Zigbee, Modbus, and KNX. It automatically matches the communication specifications of devices from different manufacturers through a dynamic protocol mapping table, achieving unified encoding and distribution of instructions from heterogeneous devices. The timing orchestration controller constructs a dependency graph based on the device response delay model and uses a priority queue and time window sliding algorithm to optimize the instruction execution sequence, ensuring that time-sensitive operations are triggered first while long-cycle tasks proceed in parallel. For example, when a projector is turned on, the curtain closing action is synchronously delayed to avoid instantaneous current overload. The conflict resolution unit monitors the device status feedback flow in real time, using rule-based reasoning and fuzzy logic to predict potential operational conflicts. When a parameter conflict is detected between an air conditioner temperature increase instruction and a fresh air unit energy-saving mode, a strategy renegotiation mechanism is automatically triggered. An energy consumption and comfort balance algorithm generates a compromise execution plan and feeds it back to the strategy generation layer for global optimization and calibration.

[0061] The internal settings of the feedback learning and system adaptation submodule include:

[0062] The system comprises four core components: a multi-source feedback acquisition interface, an effectiveness quantification evaluation model, a parameter dynamic adjustment engine, and a knowledge graph update network. The feedback acquisition interface is deployed in a bidirectional channel between the classroom physical layer and the application layer, capturing real-time environmental sensor data streams, device execution status codes, user interaction logs, and teaching effectiveness evaluation data. Time-series features are extracted using sliding window statistical techniques. The effectiveness quantification evaluation model employs a multi-dimensional weighted scoring system, nonlinearly mapping 12 technical indicators, such as temperature stability, light uniformity, and noise suppression rate, to teaching effectiveness parameters like student concentration improvement and teacher satisfaction scores, constructing a comprehensive effectiveness index based on the analytic hierarchy process (AHP). The parameter dynamic adjustment engine integrates Bayesian optimization algorithms and a meta-learning framework, automatically adjusting the feature weights of the context recognition model, the optimization coefficients of the strategy generation module, and the protocol adaptation parameters of the instruction conversion layer based on the effectiveness index's changing trends. It also introduces an anomaly detection mechanism to identify model shifts caused by equipment aging or sudden environmental changes, triggering a local model retraining process. The knowledge graph update network establishes a database of equipment operation experience and a database of teaching scenarios. Through event-result association mining technology, it extracts high-frequency success strategy patterns and solidifies the verified control logic into the rule knowledge base, forming a self-iterable cognitive loop for the system. This ensures that the smart classroom continuously optimizes its response accuracy and scenario adaptability during long-term operation.

[0063] The control layer module comprises two main functional units: an edge decision unit and an actuator drive cluster. The edge decision unit is equipped with a dual-core real-time processor, running an adaptive PID control algorithm library to dynamically adjust control parameters based on environmental change rates. For example, it automatically enhances the air conditioning cooling output response coefficient when the temperature sensor detects a sudden increase. The actuator drive cluster is divided into a lighting dimming group, an air conditioning group, and a teaching equipment group. The dimming group is equipped with a PWM dimming driver board and a DMX512 protocol conversion module to achieve precise control of LED light strip color temperature and brightness. The air conditioning group integrates a variable frequency compressor drive circuit and a fresh air damper stepper motor controller, supporting 0.1℃-level temperature adjustment and 15-level fan speed fine-tuning. The teaching equipment group deploys a relay matrix and a CEC command generator, providing projector power sequence management and intelligent switching capabilities for HDMI signal sources.

[0064] The application layer module features a three-dimensional architecture comprising a cloud-based decision engine, a visual management interface, and cross-system integration channels. The cloud-based decision engine deploys an LSTM neural network model, analyzes three years of historical environmental data to establish a seasonal pattern prediction model, and combines real-time meteorological API data to pre-adjust the classroom microclimate two hours in advance. The visual management interface uses a WebGL 3D engine to reconstruct a digital twin of the classroom, supporting 3D presentation of device status and drag-and-drop strategy arrangement, automatically triggering an amber alarm pulse effect in abnormal states. The cross-system integration channels utilize a suite of RESTful API interfaces to synchronize timetable data with the campus academic affairs system, exchange energy data with the building management (BA) system, and integrate and analyze teaching behavior data with third-party AI teaching assistant platforms, forming a complete intelligent education ecosystem.

[0065] This invention significantly improves the matching accuracy between the teaching environment and learning activities by deeply integrating IoT sensing and contextual adaptation technologies. The system can analyze the essential needs of the teaching scenario in real time. For example, when the teacher initiates an interactive discussion mode, it automatically adjusts the light color temperature to a cool white light range that promotes active thinking, simultaneously reduces the air conditioning fan speed to minimize noise interference, and switches the projection screen to a multi-window split-screen display state, enabling dynamic coordination between environmental parameters and teaching activities. This environment-adaptive mechanism based on teaching intent not only eliminates the burden of frequent manual equipment adjustments in traditional classrooms but also effectively improves students' knowledge absorption efficiency and classroom participation by optimizing the combination of environmental elements such as sound, light, and heat.

[0066] In this invention, an IoT sensor network continuously collects equipment operation data and environmental feedback, and combines this with the teaching schedule to predict energy consumption fluctuations. For example, the ventilation system is pre-activated before experimental classes to maintain air cleanliness, and automatically enters a low-power inspection mode after class. This predictive maintenance capability significantly reduces the risk of abnormal equipment downtime, extends the lifespan of teaching equipment, and reduces unnecessary energy consumption through intelligent strategies. Teaching administrators can intuitively understand the usage characteristics of different classrooms through multi-dimensional data dashboards, providing a basis for decision-making to optimize scheduling plans and improve space design.

[0067] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0068] The foregoing description enables those skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart classroom control system based on Internet of Things (IoT) technology, characterized in that: The system includes: a perception layer module, a network layer module, a context-aware and adaptive learning environment engine, a control layer module, and an application layer module; The context-aware and adaptive learning environment engine is internally configured with a data aggregation and preprocessing submodule, a context understanding and state recognition submodule, a policy generation and optimization submodule, an instruction conversion and coordination submodule, and a feedback learning and system adaptation submodule. The environmental sensor output of the perception layer module is connected to the IoT gateway data acquisition port of the network layer module via the Zigbee wireless protocol. The protocol conversion gateway command output terminal of the network layer module is connected to the gigabit Ethernet controller of the edge decision unit of the control layer module via an Ethernet physical interface. The policy output interface of the context-aware and adaptive learning environment engine is bound to the MQTT topic subscription channel of the control layer execution mechanism drive cluster to receive JSON format control command streams in real time. The PWM dimming signal output terminal of the control layer module directly drives the constant current drive board of the lighting system; The WebSocket server of the application layer module's visual management interface continuously monitors the real-time device status data stream forwarded by the network layer traffic scheduling controller, and simultaneously establishes a secure data pipeline with the campus academic affairs system's timetable interface through the OAuth2.0 authorization channel.

2. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The perception layer module is internally equipped with three core components: an environmental sensor network, an equipment status monitoring unit, and a human-computer interaction terminal. The environmental sensor network deploys light sensors, temperature and humidity sensors, carbon dioxide concentration sensors, noise sensors, smoke sensors, and infrared human body detectors to form a three-dimensional spatial monitoring matrix, capturing real-time classroom light intensity distribution, air quality, and thermal maps of human activity. The equipment status monitoring unit integrates a power parameter acquisition chip and an equipment communication protocol parser to continuously track the projector's operating power curve, the air conditioner compressor's working cycle, the opening and closing status of doors and windows, and equipment operating parameters, accurately recording energy consumption data through pulse counting technology.

3. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The network layer module consists of a core architecture comprising a heterogeneous protocol conversion gateway, a wireless mesh networking node, and a traffic scheduling controller. The heterogeneous protocol conversion gateway is equipped with a Zigbee-WiFi dual-mode communication chip and a Modbus-TCP protocol stack, enabling seamless integration between RS485 bus devices and IP networks. The wireless mesh networking node employs a three-band dynamic switching technology, carrying device control command streams in the 2.4GHz band, transmitting high-definition camera video streams in the 5.8GHz band, and maintaining sensor data backhaul in the sub-1GHz band, while avoiding channel congestion through a time slot allocation algorithm.

4. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The data aggregation and preprocessing submodule includes a distributed data acquisition interface, a spatiotemporal alignment processor, a multimodal data fusion unit, an anomaly detection filter, and a standardized output channel. It establishes a unified data buffer to receive raw signal streams from ambient light sensors, temperature sensors, humidity sensors, CO2 concentration sensors, noise sensors, infrared human body detection modules, and power monitoring equipment sensors. It eliminates timing misalignment problems caused by differences in sampling frequencies of different devices through a timestamp calibration mechanism and uses sliding window technology to process discrete data continuously.

5. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The context understanding and state recognition submodule first establishes a multi-source feature extraction network to extract temporal features, spatial distribution features, and cross-modal association features from a standardized dataset. The multi-dimensional vector output from the feature engineering layer is input into a two-layer hybrid inference model. The bottom layer uses a rule-based inference engine based on a knowledge graph. The top layer deploys a lightweight deep neural network to train an end-to-end context classification model using historical labeled data. The contribution weights of different sensor features are dynamically adjusted through an attention mechanism to achieve joint probability prediction of environmental state and teaching activity type. To improve real-time performance, the module embeds an incremental learning framework. When a new device is detected or the teaching process changes, local model fine-tuning is automatically triggered to ensure the continuous adaptability of state recognition. The formula for calculating context fusion weights is defined as follows: W_ij=σ×F_hi+τ×C_rt×S_dj; Where W_ij represents the feature weight coefficient of the i-th type of sensor in the j-th teaching scenario, σ is the historical accuracy decay factor of the sensor, calculated exponentially based on the calibration error rate of the past 7 days; F_hi represents the average feature contribution of the sensor in the same historical scenario, obtained through sliding window statistics; C_rt is the real-time confidence score in the current environment, calculated inversely proportional to the device health status code and the standard deviation of data fluctuation; S_dj is the scenario adaptation parameter, dynamically assigned according to the preset teaching activity priority matrix; The dynamic confidence adjustment formula is: D_t=D_0×e^(-λt)+μ×A_c; Where D_t represents the confidence level of the environmental state at time t, D_0 is the initial confidence benchmark value, λ is the time decay coefficient, which is dynamically adjusted according to the intensity of classroom use; A_c is the anomaly detection counter, which triggers the μ coefficient multiplication mechanism when more than 3 consecutive feature contradictions occur. The attention allocation formula innovatively incorporates perceptual variables from different teaching stages, namely: A_k = T_p × L_s + E_m × R_f, Where A_k represents the attention weight of the k-th feature channel, T_p is the teaching progress parameter, which changes linearly according to the proportion of remaining time in the timetable; L_s is the learning effect correlation, which is obtained by regression analysis of historical data to obtain the correlation coefficient between each environmental parameter and student attention; E_m represents the equipment energy consumption sensitivity, which is determined by the product of the current electricity price period and the equipment's rated power; R_f is the real-time reliability score, which is calculated by combining the equipment's running time and the most recent maintenance cycle.

6. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The strategy generation and optimization submodule comprises four main parts: a teaching rule knowledge base, a real-time environment optimization engine, a multi-objective trade-off matrix, and a strategy verification loop. The knowledge base contains a set of priority rules defined by educational experts. The real-time optimization engine employs an improved multi-agent collaborative algorithm, abstracting air conditioning, lighting, and projection equipment into independent decision-making units. It solves the global optimal solution through a distributed game model, while introducing a time decay factor to dynamically adjust the balance weight between short-term comfort and long-term energy consumption. The multi-objective trade-off matrix establishes a conflict relationship graph of indicators such as temperature deviation, illumination uniformity, noise suppression rate, and equipment energy consumption ratio, and uses the analytic hierarchy process (AHP) to calculate the dynamic priority coefficients of each indicator at different teaching stages.

7. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The instruction conversion and coordination submodule consists of four parts: an instruction parsing engine, a multi-protocol adaptation matrix, a timing orchestration controller, and a conflict resolution unit. The instruction parsing engine uses semantic segmentation technology to decompose abstract strategies into atomic operation instructions that can be executed by the device.

8. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The feedback learning and system adaptation submodule is internally configured with four core components: a multi-source feedback acquisition interface, an effect quantification evaluation model, a parameter dynamic adjustment engine, and a knowledge graph update network.

9. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The control layer module comprises two main functional units: an edge decision unit and an actuator drive cluster. The edge decision unit is equipped with a dual-real-time processor and runs an adaptive PID control algorithm library to dynamically adjust control parameters according to the rate of environmental change. The actuator drive cluster is divided into a lighting dimming group, an air conditioning group, and a teaching equipment group. The dimming group is equipped with a PWM dimming driver board and a DMX512 protocol conversion module to achieve precise control of the color temperature and brightness of LED light strips.

10. The intelligent classroom control system based on Internet of Things technology as described in claim 1, characterized in that: The application layer module is equipped with a three-dimensional system including a cloud-based decision engine, a visual management and control interface, and a cross-system interface channel. The cloud-based decision engine deploys an LSTM neural network model, analyzes three years of historical environmental data to establish a seasonal pattern prediction model, and combines real-time meteorological API data to pre-adjust the classroom microclimate two hours in advance.

Images