AI intelligent regulation and control system based on infrared matrix thermal field sensing

Abstract

This invention discloses an AI intelligent control system based on infrared matrix thermal field perception, relating to the field of smart home and smart device control technology. The system includes: an image acquisition module; an infrared matrix sensing unit for acquiring thermal field distribution data of the human body and environment, and extracting body surface temperature, ambient temperature, temperature rise rate, and effective heat source pixel area; a business logic layer, including a cloud-based AI large-model analysis engine interface and multi-source data fusion decision logic, used to generate control commands by combining spatial positioning, weather data, and user habit baselines; and a dual-mode communication module with Wi-Fi / LAN and infrared transmission modules, supporting both cloud-based distribution and direct infrared control. This invention can proactively predict and turn on the air conditioner based on changes in human body temperature when the air conditioner is off, achieving whole-house thermal scheduling through a distributed panel, and possessing local infrared backup control capabilities in the event of a network outage. It can be applied to various scenarios such as homes, hotels, offices, and elderly care facilities.

Description

Technical Field

[0001] This invention relates to the field of smart home control technology, specifically to a smart control system, method, and storage medium for air conditioning and environmental equipment based on the fusion of infrared matrix thermal field sensing and artificial intelligence (AI) algorithms. Background Technology

[0002] With the development of IoT and AI technologies, smart air conditioners have gradually become widespread. The control logic of existing smart air conditioners mainly falls into two categories: The first type is feedback control based on ambient temperature sensors, which can only sense air temperature and cannot sense changes in the user's body surface temperature, resulting in lag and the risk of misjudgment. The second category is AI adaptive control, which has emerged in recent years, such as the solution disclosed in Chinese patent application CN121048247A (Beijing University of Science and Technology). This type of solution collects the user's body surface temperature and ambient temperature, and combines this with historical data from the user's manual adjustments of the remote control to train a model to optimize the operating parameters of the already turned-on air conditioner. However, this type of technology has the following significant drawbacks: Delayed decision-making starting point: Its control logic begins with the "air conditioner is on" state, and it cannot solve the problem of how to actively decide to turn on the air conditioner based on the human body's physiological thermal state when the air conditioner is off. Single perception dimension: It only relies on local sensor data of the device and does not integrate multi-source heterogeneous data such as user spatial location and external weather; Lack of contextualized care: It does not cover the fine-grained anti-misjudgment processing for special scenarios such as the user's sleep state, getting out of bed / turning over; To address the aforementioned shortcomings, this invention provides a novel solution. Summary of the Invention

[0003] I. Technical problems to be solved The present invention aims to solve the technical problems of existing smart air conditioners being unable to proactively predict and turn on based on the trend of human physiological thermal comfort when the air conditioner is off, and to autonomously adjust the operating mode, as well as being unable to achieve multi-room temperature data collection, centralized analysis and processing, and unified centralized management and control. Technical solution

[0004] To achieve the above objectives, the present invention provides the following technical solution: An AI-powered intelligent control system based on infrared matrix thermal field sensing, characterized in that it comprises: 1. System Architecture At least one infrared matrix sensing unit: deployed on a fixed device (such as the companion robot body or a deployment panel) to collect two-dimensional thermal field distribution data of the human body and environment in the target area in a non-contact manner, and extract body surface temperature T_body, ambient temperature T_env, temperature rise rate ΔT / Δt, and effective heat source pixel area S.

[0005] Environmental sensing unit: The temperature and humidity sensor and light sensor integrated on the motherboard 0001 are used to collect environmental temperature, humidity and light data as auxiliary input for multi-source fusion decision-making.

[0006] AI Decision Control Unit: (i) Cloud-based analytics layer: Includes cloud-based AI large model analysis engine interface, used to access third-party AI large model service platforms (users can choose the large model service specified by the system, or configure a third-party large model API that conforms to the interface standard). The AI ​​big model is used to perform trend analysis and feature extraction on the temperature data uploaded by the infrared matrix and the environmental sensing unit, and output the body surface temperature change trend and prediction results. (ii) Business logic layer: Pre-set multi-source data fusion decision logic. Based on the analysis results output by the large model, combined with spatial positioning data, weather data and user habit baseline database, it performs comprehensive judgment and generates control commands, which are then sent out through the dual-mode communication module.

[0007] (III) Local emergency decision-making layer: Each panel has a built-in lightweight decision engine that is decoupled from the business logic layer and is used for emergency judgment and direct infrared control when cloud services are unavailable.

[0008] 2. Equipment Hardware Architecture In a preferred embodiment, the mechanical structure and electronic component layout of the companion robot are as follows: The fuselage structure includes a head shell 1004, a body shell 1002, and a chassis 1006 connected to the body. The chassis gears are coupled with the bottom servo motor 0002 to rotate, enabling the body and head to rotate over a wide range.

[0009] Internal support frame 1001: Located inside the body, it is used to support neck rotation and stabilize the internal structure.

[0010] Neck 1003: The node connecting the head and the body. The support part on it is used in conjunction with the screen servo 0004 to drive the head to achieve up and down tilting (nodding) movements. The neck is also equipped with a gear 1005 that works with the head servo 0003 to achieve left and right rotation of the head through gear transmission.

[0011] Multi-axis pan-tilt system, including: • Bottom Servo 0002: Installed at the bottom of the inner shell, with a rotation range of 0-270°, used to control the robot body to rotate in a wide range in the horizontal direction.

[0012] • Head servo 0003: Installed at the connection between the body and the head (neck position), it is coupled to the neck through gear 1005 and is used to control the independent left and right rotation of the head relative to the body.

[0013] • Screen Servo 0004: Installed on the back of the screen, used to control the screen's tilt angle.

[0014] Core sensor group: All integrated in the chest area of ​​the device, including: • Infrared matrix sensing unit 0007: Used to collect human body thermal field data.

[0015] • Camera 0008: Used to capture images of the environment and human body to assist in positioning and tracking; when a person is missing from the lens, it works with the bottom servo motor 0002 to turn and search for the person, so that the infrared matrix sensing unit 0007 can continue to monitor the user's body temperature.

[0016] • Infrared transmitter 0006: Used to send infrared control signals directly to air conditioners or smart home devices when cloud-based smart control is not available (provided the robot is in the room where the device is located).

[0017] Mainboard 0001: Integrated inside the body, it serves as the core for controlling output signals, electrically connecting and controlling all the aforementioned servos, sensors, and communication modules.

[0018] Screen 0005: Installed in front of the head, it works in conjunction with screen servo 0004 and head servo 0003 to display the interactive interface, robot image and system status information.

[0019] 3. Core control logic (parallel multi-source fusion decision-making) The AI ​​decision-making control unit executes the following parallel processing threads: First thread (physiological and environmental monitoring): Control the infrared matrix to continuously collect body surface temperature T_body and ambient temperature T_env in an adaptive frequency conversion scanning mode, and extract the temperature rise rate ΔT / Δt and the effective heat source pixel area S.

[0020] The second thread (spatial and meteorological positioning): The spatial positioning algorithm and weather query thread run continuously in the background. Location information comes from user-authorized location access within the software or from user-initiated voice / APP input; more refined area positioning comes from camera visual feature matching to pinpoint the location within the current area (based on a pre-stored 2D overhead view of the home, marked with boundaries of doors, windows, beds, etc.); weather data comes from local positioning and authorized access to local weather information. The third thread (integrated decision-making and execution): When the user-authorized system makes autonomous decisions regarding air conditioning or smart home control, and when it detects that T_body or ΔT / Δt meets the preset abnormality triggering conditions, it immediately calls the latest cached spatial location and weather data from the second thread, combines it with the pre-stored user thermal comfort baseline database, and performs weighted fusion decision-making; it generates control commands and issues them through the dual-mode communication module: the primary path prioritizes cloud control of smart devices via Wi-Fi / LAN; the backup path calls infrared control of smart devices when the primary path is unavailable.

[0021] 4. Key Sub-algorithm 1. Self-learning thermal comfort baseline: Records and clusters data to analyze the body surface temperature ranges and corresponding air conditioning settings (temperature, fan speed, sleep mode, etc.) at which users feel comfortable in different times and seasons.

[0022] 2. Device Compatibility and Adaptive Expansion: The self-learning mechanism is not limited to temperature control; it can be extended to various smart home devices that support cloud control or infrared remote control throughout the house. (1) Device entry and registration: Users can actively enter devices in their home that support smart control or infrared control (including but not limited to: air conditioners, lighting systems, electric curtains, humidifiers, air purifiers, fans, electric heaters, projectors, etc.) through the software interface. The system will automatically identify and access the control protocol of the device and incorporate it into the unified scheduling system.

[0023] (2) User habit recording and learning: The system continuously monitors and records user behavior on each device, including but not limited to: device on / off time, operating mode preference, parameter setting preference (such as light brightness, curtain opening ratio, wind speed, etc.), usage differences in different seasons or time periods, and linkage operation modes between devices. The recorded data is clustered to form a personalized user habit baseline library for each device.

[0024] (3) Adaptive Execution and Reproduction: With user authorization, the system automatically predicts the user's intention to use each device based on the current time, season, environmental parameters, and user habit baseline library, and generates control suggestions or directly executes control commands. For example, if the system learns and records that the user is accustomed to dimming the lights to 30% and closing the curtains to 80% when watching a movie at night, it will automatically execute this combination configuration when the system detects that the user is watching a movie; if the system learns and records that the user is accustomed to setting the air conditioner to 26℃ and running the fan during the day in summer, it will automatically reproduce this linkage strategy in this scenario.

[0025] (4) The adaptive expansion mechanism is based on a unified software architecture. New devices do not need to modify the core decision-making logic and can be included in the system management simply by accessing the protocol through a standardized interface. The system learns from the historical control data of each device and user feedback, and automatically generates personalized control strategies for different devices and incorporates them into the unified scheduling system.

[0026] 3. Anti-false alarm tracking mechanism: When the infrared pixel area S shrinks by more than 50% in a short period of time (e.g., within 2 seconds), it is determined that the user's position has changed. At this time, combined with environmental semantic judgment: if the environmental image captured by camera 0008 confirms after feature matching that the user's movement direction is pointing towards the door area, it is determined that the user has left the current room, the system does not start active tracking, only maintains the current state of the air conditioner, and suspends decision output; when the infrared matrix sensing unit 0007 detects the heat source entering the room area again, it triggers the visual follow mode, the bottom servo motor 0002 drives the body to rotate according to the direction of the heat source, and the camera 0008 continues to capture the human body outline until S returns to stability (lying down again), and then decision output resumes.

[0027] 4. Adaptive Scanning Frequency: The system dynamically adjusts the scanning interval of the infrared matrix according to the monitoring status. In standby mode (no one or no active heat source), it scans at a low frequency T1 (e.g., 60s) to detect human entry and reduce standby power consumption. When an effective heat source is detected and the user is determined to be asleep or sitting still, it switches to a medium frequency T2 (e.g., 10s) to balance power consumption and data accuracy. When the temperature rise rate ΔT / Δt or the body surface temperature T_body triggers the abnormal threshold, it immediately switches to a high frequency T3 (e.g., 1s) to support accurate decision-making with dense sampling. After the body temperature returns to a stable state, the frequency gradually decreases and falls back. 5. System Self-Check and Zeroing: Upon system startup, motherboard 0001 performs a zeroing reset operation on each servo: bottom servo 0002 rotates to the preset 0° reference direction (e.g., facing the center of the bed); head servo 0003 and screen servo 0004 reset to the default elevation angle; camera 0008 acquires a reference image and performs a preliminary match with a pre-stored 2D top view to confirm the room location and orientation of the equipment. This self-check procedure verifies whether the servo operation is sluggish and whether the sensors are functioning properly. If the self-check fails, the user will be prompted for repair via screen 0005 or voice prompts.

[0028] (III) Beneficial Effects 1. It has enabled air conditioners to move from "passively responding to remote controls" to "actively predicting human comfort," filling the decision-making gap when the unit is off.

[0029] 2. By introducing spatial positioning and weather data, the system effectively avoids accidental power-on caused by sudden changes in the external environment (such as a sudden drop in temperature), thus improving the stability and rationality of decision-making.

[0030] 3. The infrared matrix sensing unit, combined with the anti-misjudgment tracking mechanism, can accurately distinguish between "normal sleep turning over" and "heat-induced discomfort", significantly improving the user's nighttime experience and avoiding arbitrary judgments and adjustments.

[0031] 4. The dual-mode communication (Wi-Fi + infrared) design ensures high availability when devices are not accessed via the cloud. In the absence of cloud access, basic control of smart devices in the current room and other rooms can be performed via the infrared signal in the LAN / WiFi control panel. Alternatively, if the LAN / WiFi is interrupted, the relevant devices can still be controlled directly via the infrared signal on the panel. The system has complete disaster recovery capabilities. Attached Figure Description

[0032] Figure 1 This is a block diagram of the main architecture of the key system of the present invention; Figure 2 This is a schematic diagram showing the cross-sectional positions of the infrared matrix sensing unit, infrared transmitter, camera, and servo motor inside the robot according to the present invention. Figure 3 This is a schematic diagram of the head side structure of the present invention, mainly showing the cooperation relationship between the head servo, screen servo, neck and gears; Figure 4 This is a topology diagram for whole-house distributed node deployment, representing a future expansion direction of the present invention. Detailed Implementation

[0033] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Example

[0034] Application of stationary companion robots in bedroom sleep scenarios (combined with)Figure 1 , 2 3) In this embodiment, the hardware carrier is a companion robot that is permanently placed in a bedroom or other setting. Figure 2 As shown, its structure includes a head shell 1004, a body shell 1002, and a chassis 1006 connected to the body, with a total height of approximately 20cm to 30cm. (It should be noted that the following embodiments describe a companion robot as a preferred form of the carrier, but the system architecture and core control logic are also applicable to wall panels, fixed monitoring equipment, smart terminals, or other fixed carriers. Those skilled in the art should understand that the anti-misjudgment tracking mechanism and multi-source data fusion decision logic are unrelated to whether the carrier has the ability to move; their core lies in judging based on the changing trend of infrared matrix thermal field data. Carriers with rotation functions can achieve visual tracking through servo motors, while fixed carriers without rotation functions (such as wall panels and monitoring equipment) can achieve anti-misjudgment decisions solely through logical judgment. All carrier forms that adopt the core logic described in this invention fall within the protection scope of this invention.) The layout of the core components of the device is as follows: • Chest position: An infrared matrix sensing unit 0007 (e.g., but not limited to model MLX90640; those skilled in the art can select other equivalent models and parameters according to actual design requirements), a camera 0008, and an infrared transmitter 0006 are integrated side by side.

[0035] • Internal components: The integrated motherboard 0001 is equipped with an AI decision-making main control chip and a dual-mode communication module. The infrared matrix sensing unit 0007 transmits and collects detection data by connecting to the main control chip.

[0036] • Motion mechanism: Bottom servo 0002 is installed at the bottom of the inner shell of the body and is coupled to the chassis 1006 connected to the body through chassis gears. The rotation range is 0-270° and it is responsible for the overall horizontal steering of the device. Head servo 0003 is installed at the connection between the body and the head (neck position) and is coupled to the neck 1003 through gear 1005. It is responsible for the left and right rotation of the head. Screen servo 0004 is installed on the back of the screen 0005 and is used in conjunction with the bracket part on the neck 1003. It is responsible for the vertical tilt angle of the screen.

[0037] (1) Initial configuration and authorization Before first use, users can upload floor plans or modeling diagrams of their homes via the app to create a 2D top-down view of their home, marking the boundaries of each room (such as the locations of doors, windows, and beds) for AI to train on the environment and understand spatial semantics. Simultaneously, users can grant the system long-term or one-time authorization to "actively detect and control temperature during nighttime sleep periods (e.g., 11:00 PM to 7:00 AM) and make autonomous decisions" (i.e., users can set the authorization to be in effect indefinitely, or the system can proactively ask the user for authorization each time it starts).

[0038] System power-on self-test and initialization: Each time the system starts up, motherboard 0001 performs a zero-reset operation on each servo motor: • Rotate the bottom servo motor 0002 to the preset 0° reference direction (e.g., facing the center of the bed).

[0039] • Reset head servo 0003 and screen servo 0004 to default elevation angle.

[0040] • Camera 0008 captures a reference image and performs preliminary feature matching with a pre-stored 2D top view or a long-term recorded image of the room (which can be entered by the user or authorized by the system) to confirm the room and orientation of the device.

[0041] This self-test program is used to verify whether the servo motor is stuck or whether the sensors are working properly. If the self-test fails, the user will be prompted to repair via screen 0005 or voice prompt.

[0042] (2) Parallel sensing and localization Background Thread A (Thermal Field Monitoring): After the robot starts, the infrared matrix sensing unit 0007 begins scanning the monitoring area (initially, the scanning cycle is, for example, 1 second / scan). When the effective human heat source area S is detected to be greater than the preset threshold (e.g., >500 pixels) and maintained for more than the preset duration (e.g., 30 seconds), it is determined that the user has taken their place (e.g., in bed), and the system automatically switches to "sleep monitoring mode," with the scanning cycle extended accordingly to reduce power consumption (e.g., adjusted to once every 10 seconds).

[0043] Background Thread B (Spatial and Weather): The robot uses camera 0008 to capture images of its surrounding environment. By performing feature matching (e.g., image matching algorithm based on ORB feature points) with pre-stored 2D top-down views or user-recorded environmental images, it determines its spatial location (e.g., "master bedroom" or "study"). Simultaneously, the system requests outdoor temperature, humidity, and perceived temperature forecast data from the weather service interface via Wi-Fi at preset time intervals (e.g., every 30 minutes) and stores it in a local cache.

[0044] (3) Anomaly Triggering and Comprehensive Decision-Making Assuming it is currently nighttime sleep and the air conditioner is off, the infrared matrix sensor unit 0007 detects that the surface temperature T_body of the user's torso area continuously rises within a preset time period (e.g., from 36.2℃ to 36.9℃ within 5 consecutive minutes), and the rate of temperature rise ΔT / Δt exceeds a preset threshold (e.g., 0.1℃ / min), then it is determined to be a "thermal anomaly" state.

[0045] At this moment, the AI ​​master controller immediately retrieves the following in parallel: Latest spatial positioning data; Latest weather cache data; User habit baseline data (such as user preferred temperature settings, operating modes, wind speed, etc. during this period).

[0046] (4) Instruction execution and closed-loop feedback The AI ​​controller determined that the user was in a state of "early signs of fever discomfort" and then issued control commands through the dual-mode communication module: Main path: Prioritize sending control commands (such as powering on, setting the temperature, activating sleep mode, etc.) to the target air conditioner's cloud interface via Wi-Fi / LAN. Backup path: If the main path fails to be sent, infrared transmitter 0006 will be activated immediately to simulate the remote control code of the corresponding brand of air conditioner and directly control the air conditioner through infrared signals to ensure reliability.

[0047] (5) False judgment prevention mechanism and environmental semantic tracking During air conditioner operation, if the infrared matrix sensing unit 0007 detects that the effective heat source pixel area S shrinks by more than a preset ratio (e.g., 60%) within a short period of time (e.g., within 2 seconds), the AI ​​main controller determines it as "non-thermal displacement" (e.g., the user gets up at night, leaves the room, etc.), and after comprehensive judgment, pauses the current temperature judgment acquisition, maintains the current air conditioner data without change, and avoids erroneous judgment operations.

[0048] At the same time, the environmental semantic auxiliary tracking program is triggered: Record of heat source disappearance direction: Record the azimuth angle of the heat source last captured by the infrared matrix sensing unit 0007; Camera visual judgment: Camera 0008 follows the user's departure path, captures images of the current environment, and performs feature matching with a pre-stored 2D top-down view. If the AI ​​determines that the user's movement direction is towards the exit area (such as a door), it determines that the user has left the current room, the system does not initiate active tracking of the user's actions, keeps the device stationary, and only maintains the current state of the air conditioner. User return tracking: When the infrared matrix sensor unit 0007 detects a heat source entering the current room again, it triggers the visual follow mode. The bottom servo motor 0002 drives the body to rotate slowly according to the location of the heat source, and the camera 0008 continuously captures the human body outline to achieve body following (the head servo motor and the bottom servo motor work in sync). This continues until the infrared matrix sensor unit 0007 confirms that the effective heat source pixel area S has recovered to a stable state (e.g., recovered to more than 90% of the original area). The system then automatically resumes temperature monitoring and decision-making logic. Example

[0049] Coordinated control of distributed panels throughout the house (combined with) Figure 4 ) This embodiment is an extended deployment scheme of Embodiment 1. Users install embedded wall panels in relevant rooms such as the living room, master bedroom, and secondary bedroom. Each panel has the following built-in features: Infrared matrix sensing unit sensor and AI edge node; Infrared transmitter: Used to directly control the air conditioner or other smart home devices in the corresponding room (can be controlled one-to-one or one-to-many), ensuring that the device can still be turned on and off and its parameters set via local infrared commands when cloud services are unavailable or the device does not support cloud control.

[0050] Each panel uploads the locally extracted "maximum body surface temperature T_max" and "effective pixel area S" to the cloud or the business logic layer of the current local area network software at preset time intervals (e.g., every 60 seconds). The business logic layer has built-in multi-source data fusion decision logic, connects to third-party AI large model interfaces for auxiliary analysis, and generates control commands after comprehensive judgment and sends them to the corresponding panel for execution.

[0051] When the business logic layer detects an increase in T_max reported by the living room panel and the user is located in the living room, it executes global thermal scheduling: If the total power consumption of the entire house is limited, the business logic layer determines whether there are people in each space based on the S value reported by each panel. Priority is given to ensuring the cooling needs of rooms where people are frequently present or tend to sleep (S value is stable and lasts for more than 2 hours), while the smart home systems in unoccupied rooms are shut down or put into energy-saving mode. Main path: After receiving instructions from the business logic layer, the system calls the cloud to control the relevant smart home devices. Secondary path: When the smart home devices in the controlled room lack cloud access, the infrared transmitters on each panel, after receiving instructions from the business logic layer, directly send control signals to the air conditioner or other devices in the room, forming a distributed control architecture of "cloud decision-making + panel execution".

[0052] Based on the daily integrated data, either daily or weekly (the user chooses the receiving time period), the business logic layer generates a thermal comfort log based on the data summarized from all panels throughout the day, and pushes it to the user via the APP. This log includes "average temperature curves for each room", "statistics on user body surface temperature fluctuations", and "energy consumption assessment", etc. Example

[0053] High availability and disaster recovery in the event of cloud service unavailability (combined with Figure 4 ) When smart home cloud services are unavailable (e.g., cloud platform services are interrupted, devices are not bound to cloud accounts, or devices themselves do not support cloud control), the system automatically downgrades to "local infrared direct control mode".

[0054] In this mode, local area network communication (Wi-Fi) is usually still normal, data transmission between each panel and the business logic layer and collaborative communication between panels continue to be effective, and the business logic layer can continue to receive and analyze the body temperature and environmental data reported by each panel. However, if smart home devices cannot be controlled through the cloud interface, it will be downgraded to "local infrared direct control mode", which will not affect other data transmission and detection.

[0055] At this point, the system switches to local baseline decision mode (i.e., a simplified backup strategy at the business logic layer, pre-deployed in the AI ​​edge nodes of each panel): The panel compares the real-time collected infrared body temperature data T_body with the historical body temperature baseline T_base stored locally. When T_body scans N times consecutively and exceeds T_base by 1°C, the air conditioner is directly turned on via the panel's built-in infrared transmitter (0006), and the set temperature is fixed to the user's most frequently used temperature in history (e.g., 25°C). After the cloud service is restored, all locally cached data is automatically synchronized to the cloud, updating the user habit model.

[0056] As a further optimization, when the local area network (Wi-Fi) is completely interrupted (communication between the panel and the business logic layer is disconnected), the user can still operate the panel directly and independently execute the backup control command (such as turning on the machine at the preset temperature) through the built-in infrared transmitter. However, the cross-room collaborative scheduling function will be temporarily unavailable. Once the network is restored, the whole-house collaboration and data monitoring and uploading will be automatically rebuilt. Industrial applicability

[0057] The system and method described in this invention can be widely applied to scenarios such as home companion robots, whole-house distributed intelligent temperature control systems, and intelligent management of hotel rooms. They can also be used in places such as nursing homes and hospital wards that require precise nighttime temperature monitoring, and have good prospects for industrial application.

Claims

1. An AI intelligent control system based on infrared matrix thermal field sensing, characterized in that, include: The image acquisition module (including a camera) is used to acquire environmental and human images of the target area to assist in positioning and tracking. The infrared matrix sensing unit 0007 is deployed on a fixed device or wall panel to collect two-dimensional thermal field distribution data of the human body and environment in the target area in a non-contact manner, and extract the body surface temperature T_body, ambient temperature T_env, temperature rise rate ΔT / Δt and effective heat source pixel area S. The business logic layer includes a cloud-based AI large-scale model analysis engine interface and multi-source data fusion decision logic. The AI ​​large-scale model analysis engine interface is used to access a third-party AI large-scale model service platform and call the inference capabilities of the large model to perform trend analysis and feature extraction on the temperature data uploaded by the infrared matrix sensing unit. The multi-source data fusion decision logic performs comprehensive judgment and generates control commands based on the analysis results output by the large model, combined with spatial positioning data, weather data, and user habit baseline database. The dual-mode communication module includes both a Wi-Fi / LAN communication module and an infrared transmitting module, used to send the control commands to the target device. The sending includes a primary path and a backup path. The primary path is sent via Wi-Fi / LAN through the cloud, and the backup path is sent directly to the target device via Wi-Fi / LAN by controlling the infrared transmitting module.

2. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 1, characterized in that, The control commands include: when the target air conditioner is in the off state, when the change trend of the body surface temperature T_body and the temperature rise rate ΔT / Δt meets the preset abnormality triggering conditions, actively generating a start command and setting the operating temperature and operating mode of the air conditioner.

3. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 2, characterized in that, The system further includes: when the body surface temperature T_body is detected to rise continuously within a preset time period and the rate of temperature rise ΔT / Δt exceeds a preset threshold, it is determined to be a thermal anomaly state, and the current spatial positioning data, weather cache data and user habit baseline data are retrieved in parallel for weighted fusion decision-making.

4. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 1, characterized in that, The system also includes an intelligent device adaptive expansion module: The thermal field data collected by the infrared matrix sensing unit and the analysis results of the business logic layer are used as a unified data source to provide services to various smart devices. Users can actively input devices that support intelligent control or infrared control through the software interface. The system will automatically identify and access the control protocol of the device and incorporate it into the unified scheduling system. The system continuously monitors and records user behavior on each device, including device on / off time, operating mode preference, parameter setting preference, usage differences in different seasons or time periods, and the linkage operation mode between devices. After cluster analysis, a personalized user habit baseline library is formed for each device. With user authorization, the system automatically predicts the user's intention to use each device based on the current time, season, environmental parameters, and user habit baseline library, and generates control suggestions or directly executes control commands.

5. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 1, characterized in that, The infrared matrix sensing unit is disposed on the chest area of ​​a mounting carrier; the mounting carrier includes, but is not limited to, a companion robot, a wall panel, or other fixed devices; the mounting carrier also includes: Camera 0008 is located on the chest and is used to capture images of the environment and the human body to assist in positioning and tracking. Bottom servo motor 0002 is installed at the bottom of the carrier and is used to control the overall rotation of the carrier in the horizontal direction; Head servo motor 0003 is mounted on the upper part of the carrier and is used to control the left and right rotation of the head structure; Screen servo 0004 is mounted on the back of screen 0005 and is used to control the screen's tilt angle. Infrared transmitter 0006 is located on the chest and is used to send infrared control signals directly to the device when cloud-based intelligent control is not available. The motherboard 0001 is integrated inside the carrier and electrically connects to and controls all the aforementioned servos, sensors, and communication modules.

6. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 1, characterized in that, The system also includes a tracking mechanism to prevent false positives: When the effective heat source pixel area S shrinks by more than a preset ratio within a preset time, it is determined that the user's position status has changed, the decision output is paused, and the image acquisition module is triggered to acquire the current environmental image. The acquired environmental images are matched with pre-stored 2D top-down views for feature matching: If the user moves towards the exit area, active tracking will not be initiated, and the current device status will be maintained. The decision output will resume after the infrared matrix sensing unit detects the return of the heat source and the effective heat source pixel area S stabilizes. If the user's movement direction does not point towards the exit area, monitoring continues until the infrared matrix sensing unit confirms that the effective heat source pixel area S has stabilized, and then the decision output is resumed.

7. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 1, characterized in that, The system also includes multiple distributed wall panels, each with a built-in infrared matrix sensing unit and infrared transmitter. Each panel uploads the collected data to the business logic layer via a local area network. The business logic layer performs whole-house thermal scheduling based on the data reported by each panel: when multiple spaces have demand at the same time and the total power is limited, the system determines the user status of each space based on the stability and duration of the effective heat source pixel area S reported by each panel, and prioritizes the needs of spaces where users are sleeping or staying.

8. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 7, characterized in that, The business logic layer generates a thermal comfort log based on the data summarized from each panel at a preset time period. The thermal comfort log includes temperature curves for each room, statistics on fluctuations in user body surface temperature, and energy consumption assessment data.

9. The AI ​​intelligent control system based on infrared matrix thermal field sensing according to claim 1, characterized in that, The system also includes multiple control modes: Cloud-based control mode: The business logic layer controls the target device through a cloud interface; Local Area Network Infrared Control Mode: The business logic layer controls the infrared transmitters of each panel to execute commands via the local area network; Local independent control mode: Each panel has a built-in infrared transmitter, allowing users to send infrared control signals directly to the target device through the panel; The various control modes can be switched automatically or manually selected by the user.

Images