Home network intelligent optimization system and method based on AI algorithm

Abstract

The invention discloses a home network intelligent optimization system and method based on an AI algorithm, and the system comprises a full-link monitoring module, an AI intelligent analysis module, a dynamic resource scheduling module, a fault self-repairing module and a user interaction module, all modules cooperate to form a complete optimization link, and achieve the combination of real-time performance and predictability, thereby quantitatively driving precise scheduling. According to the full-link monitoring module, a flow sensor and a time delay monitoring unit are arranged in a home gateway, and RSSI detectors are embedded in sub-routers and terminal equipment; the AI intelligent analysis module internally comprises an LSTM (Long Short Term Memory) load prediction model, a K-means equipment clustering algorithm and bottleneck identification logic; and the dynamic resource scheduling module is used for allocating the 2.4 GHz frequency band to the IoT equipment and allocating the 5GHz frequency band to the high-bandwidth equipment through channel and frequency band optimization, and switching to an idle channel when the number of the same-channel equipment is more than or equal to 3. According to the invention, dynamic scheduling of network resources, automatic fault repair and scenarized adaptation can be realized, and the home network stability and the user experience are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of home network intelligent optimization, and in particular to a home network intelligent optimization system and method based on an AI algorithm. Background Art

[0002] With the popularization of smart home devices such as smart TVs, security cameras, and sweeping robots, the number of home network access devices has increased sharply, with the average number of devices in a household reaching 8-15. The network load fluctuates significantly, such as during the morning rush hour for video conferencing on weekdays and the evening entertainment traffic peak.

[0003] Existing home networks have the following problems: Imbalanced resource allocation, with high-priority devices (such as office computers) competing for bandwidth with low-priority devices (such as idle IoT devices), causing critical applications to stall; Delayed fault response: Problems such as network congestion and signal interference require manual troubleshooting, making it difficult for non-professional users to quickly resolve them. Insufficient scenario adaptation: Network policies cannot be dynamically adjusted based on user behavior (such as switching between work and entertainment), resulting in resource waste or insufficient performance. With the increasing requirements for network usage and the development of intelligent technology, it is of great practical significance to develop a home network intelligent optimization system and method based on AI algorithm. Summary of the Invention

[0004] The purpose of the present invention is to provide a home network intelligent optimization system and method based on AI algorithm to solve the problems of large load fluctuations, unreasonable resource allocation, and low fault handling efficiency in existing home networks, realize dynamic scheduling of network resources, automatic fault repair and scenario adaptation, and improve home network stability and user experience.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: Home network intelligent optimization system based on AI algorithm, including: The full-link monitoring module, AI intelligent analysis module, dynamic resource scheduling module, fault self-repair module, and user interaction module work together to form a complete optimization link, combining real-time and predictive capabilities to quantitatively drive precise scheduling. The full-link monitoring module has a built-in traffic sensor (accuracy 1Mbps) and a latency monitoring unit (sampling frequency 10 times / second) in the home gateway (main router), and an RSSI (signal strength) detector (monitoring range -100dBm to -30dBm) in the sub-routers and terminal devices (computers / TVs). AI intelligent analysis module, which includes: LSTM load prediction model, K-means device clustering algorithm, and bottleneck identification logic; The dynamic resource scheduling module optimizes channels and frequency bands, allocating the 2.4 GHz band to IoT devices (prioritizing coverage) and the 5 GHz band to high-bandwidth devices (prioritizing speed). When there are three or more devices on the same channel, the module switches to an idle channel (spectrum scanning identifies channels with an idleness of 80% or more).

[0006] As a further improvement of the present technical solution: the fault diagnosis and repair strategy of the fault self-repair module is: when the fault phenomenon is a single device packet loss rate ≥5%, its repair strategy is to switch the frequency band (2.4G / 5G). If it is still ineffective, it will prompt to check the network card, and the repair time is ≤10 seconds. When the fault phenomenon is a gateway delay ≥200ms, its repair strategy is to restart the gateway (retain key connections). If it is still ineffective, the cache is cleared, and the repair time is ≤30 seconds. When the fault phenomenon is multiple devices disconnected, its repair strategy is to switch to the backup 4G route. If it is still ineffective, it will report to the operator, and the repair time is ≤1 minute. When the fault phenomenon is the access of an unfamiliar device, its repair strategy is to automatically isolate the unfamiliar device. If it is still ineffective, it will push an early warning to the user APP, and the repair time is ≤5 seconds.

[0007] As a further improvement to this technical solution: the user interaction module includes a real-time dashboard, manual intervention, optimization log, and personalized settings. The real-time dashboard contains device bandwidth usage (histogram), signal heat map, and network health score. The manual intervention content includes: adjusting priority and bandwidth locking. The optimization log records the execution time, content, and effect of the policy. The personalized settings include custom do not disturb periods and a high-priority device whitelist.

[0008] As a further improvement of this technical solution: in the full-link monitoring module, when the monitoring object is the device status, the main monitored key indicators are: online / offline, MAC address, IP mapping, the collection frequency is 1 time / 30 seconds, and there is no abnormal threshold; when the monitoring object is the network load, the main monitored key indicators are upload / download rate and bandwidth occupancy, the collection frequency is 1 time / 5 seconds, and the abnormal threshold is: single device occupancy ≥ 50% (triggering attention); when the monitoring object is the transmission quality, the main monitored key indicators are: delay (RTT), packet loss rate, and number of retransmissions, the collection frequency is 1 time / 2 seconds, and the abnormal threshold is: delay ≥ 100ms or packet loss rate ≥ 3% (triggering warning); when the monitoring object is the signal environment, the main monitored key indicators are RSSI value and the number of co-channel devices, the collection frequency is 1 time / 10 seconds, and the abnormal threshold is: RSSI ≤ -85dBm (weak signal). The full-link monitoring module uses 5G edge computing nodes to locally process raw data and only uploads characteristic values ​​(such as load peaks, abnormal events) to reduce transmission delay.

[0009] As a further improvement to this technical solution, the LSTM load prediction model takes as input the bandwidth usage rate of the past hour, the number of online devices, and a timestamp (weekday / weekend mark) and outputs a load prediction curve for the next 15 minutes. Its training logic is as follows: initialization imports 30 days of historical data, iterative training with newly added data daily, and the Adam optimizer is used.

[0010] As a further improvement of this technical solution: the K-means device clustering algorithm, clustering features: device type (work / entertainment / IoT), average bandwidth demand, usage time, output priority label (level 1-5), (where level 1 is the highest, such as a video conferencing computer, and level 5 is the lowest, such as an idle smart socket).

[0011] As a further improvement of this technical solution: the bottleneck identification logic determines the cause of congestion through a decision tree model: when the bandwidth occupancy of a single device is ≥70% and lasts for 5 minutes, a "device overload" prompt is given; when there are ≥4 devices on the same channel and the RSSI fluctuation is ≥15dBm, a "signal interference" prompt is given; when the gateway CPU occupancy is ≥90% and the delay is ≥200ms, a "routing performance is insufficient" prompt is given.

[0012] As a further improvement of this technical solution: the bandwidth allocation formula in the dynamic resource scheduling module is:

[0013] in: : allocated bandwidth of device i; : The real-time requirements of device i (e.g., 4K video requires 20Mbps); : total household bandwidth; : Elastic bandwidth pool (20% of total bandwidth); (base weight), (priority weight), optimized through A / B testing.

[0014] As a further improvement to this technical solution: the QoS adaptation in the dynamic resource scheduling module is as follows: low-latency applications (games / video conferencing) are marked as EF (accelerated forwarding), non-real-time applications (downloads / cloud synchronization) are marked as BE (best effort forwarding), and the speed limit is ≤ 30% of the total bandwidth.

[0015] The present invention also proposes a method for a home network intelligent optimization system based on an AI algorithm, comprising the following steps: Step 1: System initialization (takes about 1 hour); Step 2: Real-time monitoring and data preprocessing; Step 3: AI load prediction and bottleneck identification; Step 4: Dynamic scheduling and execution (response ≤ 2 seconds); Step 5: Feedback and model iteration (performed every morning); Step 6: Safety and redundancy design.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention combines real-time and predictive capabilities. A full-link monitoring module collects data such as device status, network load, and signal quality in real time, ensuring a rapid response to traffic bursts (such as video conferencing and 4K streaming). It uses an LSTM model to predict future load trends, adjust resource allocation in advance, and avoid congestion. 2. This invention uses quantization-driven precise scheduling and spectrum scanning to automatically switch to idle channels, solving the 2.4GHz band interference problem. Traditional QoS policies rely solely on static rules, such as prioritizing video traffic. This invention adapts to complex home scenarios through formulaic allocation and dynamic adjustment. 3. The present invention fully automated fault repair, fault diagnosis and repair strategy covers scenarios such as single device packet loss and insufficient gateway performance, with a repair time of ≤ 1 minute. When the main gateway fails, the sub-router automatically switches to a temporary gateway to ensure the core equipment is connected to the network; 4. The present invention has scenario-based intelligent adaptation, automatically switches office / entertainment / sleep modes based on time and device type, limits the bandwidth of video devices, and ensures smooth operation of the computer.

[0017] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and to implement it according to the contents of the description, the following preferred embodiments of the present invention are described in detail with reference to the accompanying drawings. The specific implementation methods of the present invention are given in detail by the following embodiments and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 This is a schematic diagram of the home network intelligent optimization system based on AI algorithm proposed by the present invention; Figure 2 This is a schematic diagram of the method flow of the home network intelligent optimization system based on AI algorithm proposed by the present invention. DETAILED DESCRIPTION

[0019] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples provided are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. The following paragraphs describe the present invention in more detail by way of example with reference to the accompanying drawings. It should be noted that the drawings are all in a very simplified form and are not to exact scale, and are only used for the purpose of conveniently and clearly illustrating the embodiments of the present invention.

[0020] See also Figures 1-2 In an embodiment of the present invention, a home network intelligent optimization system based on an AI algorithm comprises a full-link monitoring module, an AI intelligent analysis module, a dynamic resource scheduling module, a fault self-repair module, and a user interaction module. Each module collaborates to form a complete optimization link, achieving a combination of real-time and predictive performance, thereby quantitatively driving precise scheduling. The full-link monitoring module has a built-in traffic sensor (accuracy 1Mbps) and a latency monitoring unit (sampling frequency 10 times / second) in the home gateway (main router), and an RSSI (signal strength) detector (monitoring range -100dBm to -30dBm) in the sub-routers and terminal devices (computers / TVs). When the gateway and sub-routers are powered on and connected to the network, the terminal SDK is activated, and the device automatically completes the network topology drawing (such as the connection relationship of "main router-living room sub-router-bedroom computer") and initializes the model.

[0021] The AI ​​intelligent analysis module includes: LSTM load prediction model, K-means device clustering algorithm, and bottleneck identification logic; the AI ​​module imports the past 30 days of historical network data (if the network is new, the default model of the same unit type is loaded), and determines the initial LSTM parameters (such as 32 hidden layer neurons and a learning rate of 0.001) through cross-validation. Users mark "critical equipment" (such as office computers) and "sensitive time periods" (such as 9:00-18:00) through the interactive app, and the system stores the configuration information. The dynamic resource scheduling module allocates the 2.4GHz band to IoT devices (coverage priority) through channel and frequency band optimization; allocates the 5GHz band to high-bandwidth devices (rate priority), and switches to idle channels (spectrum scanning identifies channels with idleness ≥80%) when there are ≥3 devices on the same channel. The dynamic scheduling and execution phase involves devices and modules including: hardware of the dynamic resource scheduling module: home gateway (supports SDN software-defined network), sub-router (switchable 2.4G / 5G band), algorithm of the dynamic resource scheduling module, through the bandwidth allocation formula , channel switching logic, real-time data of the full-link monitoring module: device priority label, current bandwidth occupancy, and signal environment data.

[0022] Specifically, the fault diagnosis and repair strategy of the fault self-repair module is as follows: when the fault phenomenon is a single device packet loss rate ≥5%, the repair strategy is to switch the frequency band (2.4G / 5G). If it is still ineffective, it will prompt to check the network card, and the repair time is ≤10 seconds. When the fault phenomenon is a gateway delay ≥200ms, the repair strategy is to restart the gateway (retain key connections). If it is still ineffective, the cache is cleared, and the repair time is ≤30 seconds. When the fault phenomenon is multiple devices disconnected, the repair strategy is to switch to the backup 4G route. If it is still ineffective, it will report to the operator, and the repair time is ≤1 minute. When the fault phenomenon is the access of an unfamiliar device, the repair strategy is to automatically isolate the unfamiliar device. If it is still ineffective, an early warning will be pushed to the user APP, and the repair time is ≤5 seconds.

[0023] Specifically, the user interaction module includes real-time dashboard, manual intervention, optimized log, and personalized settings. The real-time dashboard includes device bandwidth usage (histogram), signal heat map, and network health score. Manual intervention includes: adjusting priority, bandwidth locking, optimizing log recording policy execution time, content and effect. Personalized settings include customized do not disturb periods and high-priority device whitelist.

[0024] Specifically, in the full-link monitoring module, when the monitoring object is the device status, the main monitored key indicators are: online / offline, MAC address, IP mapping, the collection frequency is 1 time / 30 seconds, and there is no abnormal threshold; when the monitoring object is network load, the main monitored key indicators are upload / download rate and bandwidth occupancy, the collection frequency is 1 time / 5 seconds, and the abnormal threshold is: single device occupancy ≥ 50% (triggering attention); when the monitoring object is transmission quality, the main monitored key indicators are: delay (RTT), packet loss rate, and number of retransmissions. The collection frequency is 1 time / 2 seconds, and the abnormal threshold is: delay ≥ 100ms or packet loss rate ≥ 3% (triggering warning); when the monitoring object is the signal environment, the main monitored key indicators are RSSI value and the number of co-channel devices. The collection frequency is 1 time / 10 seconds, and the abnormal threshold is: RSSI ≤ -85dBm (weak signal). The full-link monitoring module uses 5G edge computing nodes to locally process raw data and only uploads characteristic values ​​(such as load peaks and abnormal events) to reduce transmission delay.

[0025] Specifically, the LSTM load prediction model takes as input the bandwidth usage rate of the past hour, the number of online devices, and a timestamp (weekday / weekend mark) and outputs a load forecast curve for the next 15 minutes. Its training logic is as follows: Initially, it imports 30 days of historical data, iterates training with new data daily, and uses the Adam optimizer.

[0026] Specifically, the K-means device clustering algorithm uses clustering features: device type (work / entertainment / IoT), average bandwidth demand, usage time, and outputs a priority label (level 1-5), (where level 1 is the highest, such as a video conferencing computer, and level 5 is the lowest, such as an idle smart socket).

[0027] Specifically, the bottleneck identification logic uses a decision tree model to determine the cause of congestion: if a single device's bandwidth usage is ≥70% and persists for 5 minutes, a "device overload" prompt is displayed; when there are ≥4 devices on the same channel and the RSSI fluctuation is ≥15dBm, a "signal interference" prompt is displayed; when the gateway CPU usage is ≥90% and the delay is ≥200ms, a "routing performance is insufficient" prompt is displayed.

[0028] Specifically, the bandwidth allocation formula in the dynamic resource scheduling module is:

[0029] in: : allocated bandwidth of device i; : The real-time requirements of device i (e.g., 4K video requires 20Mbps); : total household bandwidth; : Elastic bandwidth pool (20% of total bandwidth); (base weight), (priority weight), optimized through A / B testing.

[0030] Specifically, the QoS adaptation in the dynamic resource scheduling module is as follows: low-latency applications (games / video conferencing) are marked as EF (accelerated forwarding), non-real-time applications (downloads / cloud synchronization) are marked as BE (best effort forwarding), and the speed limit is ≤ 30% of the total bandwidth.

[0031] The method of the home network intelligent optimization system based on AI algorithm includes the following steps: Step 1: System initialization (takes about 1 hour); Step 2: Real-time monitoring and data preprocessing; Step 3: AI load prediction and bottleneck identification; Step 4: Dynamic scheduling and execution (response ≤ 2 seconds); Step 5: Feedback and model iteration (performed every morning); Step 6: Safety and redundancy design.

[0032] The working principle and process of the present invention are: Step 1: System initialization phase. The devices and modules involved include: hardware for the full-link monitoring module: a home gateway (with built-in traffic sensors and latency monitoring units), a sub-router (including an RSSI detector), and a terminal device SDK (installed on a computer, TV, or other terminal); algorithm models for the AI ​​intelligent analysis module: an LSTM load prediction model and a K-means clustering model; and an app for the user interaction module: installed on the user's phone or tablet. Power on the gateway and sub-router and connect them to the Internet. Activate the terminal SDK and the device will automatically draw the network topology (such as the connection relationship between "main router-living room sub-router-bedroom computer"). Initialize the model: The AI ​​module imports the past 30 days of historical network data (if it is a new network, it loads the default model of the same apartment type). The initial LSTM parameters are determined through cross-validation (such as 32 hidden layer neurons and a learning rate of 0.001). Users mark "key devices" (such as office computers) and "sensitive time periods" (such as 9:00-18:00) through the interactive APP, and the system stores the configuration information.

[0033] Step 2: Real-time monitoring and data preprocessing. This involves devices and modules such as the full-link monitoring module's sensing hardware: traffic sensors (built into the gateway), RSSI detectors (sub-routers / terminals), and latency monitoring units (built into the gateway); and the full-link monitoring module's software layer: data cleaning engines and feature extraction algorithms. Data collection: Traffic sensors collect device upload / download rates every 5 seconds, RSSI detectors record signal strength every 10 seconds (for example, RSSI of a mobile phone in the bedroom = -65dBm), and latency monitoring units measure RTT (round-trip time) every 2 seconds. Data preprocessing extracts features such as "peak load period," "number of concurrent devices," and "signal attenuation trend" from the raw data.

[0034] Step 3: AI load prediction and bottleneck identification. This involves devices and modules such as the core algorithms of the AI ​​intelligent analysis module: the LSTM load prediction model, the K-means clustering model, and the decision tree bottleneck identification model; and the feature data of the full-link monitoring module: pre-processed features such as "bandwidth utilization," "number of online devices," and "signal strength." Through load prediction: The LSTM model uses the feature data of the past hour as input and outputs the load curve for the next 15 minutes. The K-means algorithm classifies devices into five levels based on: device type, bandwidth requirement, and usage time. The decision tree model combines the prediction results with real-time data to determine the cause of congestion.

[0035] Step 4: Dynamic scheduling and execution phase, involving devices and modules including: hardware of dynamic resource scheduling module: home gateway (supporting SDN software defined network), sub-router (switchable 2.4G / 5G frequency band), algorithm of dynamic resource scheduling module, through bandwidth allocation formula , channel switching logic, real-time data of the full-link monitoring module: device priority label, current bandwidth occupancy, signal environment data; Resources are allocated to devices according to the formula. If there are ≥3 devices on the same channel, the sub-router automatically switches to an idle channel. High-bandwidth devices (such as TVs) prioritize connecting to the 5GHz band (rate priority), and IoT devices (such as smart sockets) connect to the 2.4GHz band (coverage priority). When the system identifies the "office scene" (9:00-18:00), it limits the bandwidth of video devices. When identifying the "entertainment scene" (20:00-22:00), it locks the 5GHz band for TV / gaming devices.

[0036] Step 5: Feedback and model iteration phase. This involves devices and modules such as the AI ​​intelligent analysis module's iteration engine, a reinforcement learning optimizer; the user interaction module's logging system, which records policy execution results; and the full-link monitoring module's effectiveness evaluation data, including bandwidth utilization, latency, and packet loss rate after policy execution. Every morning, the system calculates the "strategy improvement rate" (such as "the ratio of target device latency reduction after bandwidth allocation") and optimizes the model: if the improvement rate is less than 30%, the reinforcement learning optimizer adjusts the LSTM weights (such as increasing the influencing factor of the "device priority" feature); if the success rate of repairing a certain type of fault is less than 70% (such as channel switching failure), the fault diagnosis matrix is ​​updated (a new repair strategy of "restarting sub-routes" is added), and the "manual adjustment" and "complaint" records in the app are used as negative samples (for example, if a user manually increases the priority of a device multiple times, it will be automatically added to the whitelist).

[0037] Step 6: Security and redundancy design phase, involving devices and modules including: security module encryption chip: integrated into the gateway, supporting AES-128 encryption, redundant module backup device: 4G backup router (optional, connected to the main router), user interaction module privacy configuration page: the "Data Management" entrance in the app; Monitoring data (such as device usage habits) is processed by an encryption chip and then transmitted. Only the system hub can decrypt it. When the main gateway fails, the backup 4G router is automatically activated to keep core devices (such as computers) connected to the Internet. Users can view / delete collected device data through the APP, which complies with the "Personal Information Protection Law of the People's Republic of China".

[0038] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any ordinary technician in this industry can smoothly implement the present invention as shown in the drawings and described above. However, any equivalent changes, modifications and evolutions made by technicians familiar with this profession without departing from the scope of the technical solution of the present invention using the technical content disclosed above are all equivalent embodiments of the present invention. At the same time, any equivalent changes, modifications and evolutions made to the above embodiments based on the essential technology of the present invention are still within the scope of protection of the technical solution of the present invention.

Claims

1. Home network intelligent optimization system based on AI algorithm, characterized by: include: The full-link monitoring module, AI intelligent analysis module, dynamic resource scheduling module, fault self-repair module, and user interaction module work together to form a complete optimization link, combining real-time and predictive capabilities to quantitatively drive precise scheduling. The full-link monitoring module has a built-in traffic sensor (accuracy 1Mbps) and a latency monitoring unit (sampling frequency 10 times / second) in the home gateway (main router), and an RSSI (signal strength) detector (monitoring range -100dBm to -30dBm) in the sub-routers and terminal devices (computers / TVs). AI intelligent analysis module, which includes: LSTM load prediction model, K-means device clustering algorithm, and bottleneck identification logic; The dynamic resource scheduling module optimizes channels and frequency bands, allocating the 2.4 GHz band to IoT devices (prioritizing coverage) and the 5 GHz band to high-bandwidth devices (prioritizing speed). When there are three or more devices on the same channel, the module switches to an idle channel (spectrum scanning identifies channels with an idleness of 80% or more).

2. The home network intelligent optimization system based on AI algorithm according to claim 1 is characterized in that: The fault diagnosis and repair strategy of the fault self-repair module is as follows: when the fault phenomenon is a single device packet loss rate ≥5%, the repair strategy is to switch the frequency band (2.4G / 5G). If it is still ineffective, it will prompt to check the network card, and the repair time is ≤10 seconds. When the fault phenomenon is a gateway delay ≥200ms, the repair strategy is to restart the gateway (retain key connections). If it is still ineffective, the cache is cleared, and the repair time is ≤30 seconds. When the fault phenomenon is multiple devices disconnected, the repair strategy is to switch to the backup 4G route. If it is still ineffective, it will report to the operator, and the repair time is ≤1 minute. When the fault phenomenon is the access of an unfamiliar device, the repair strategy is to automatically isolate the unfamiliar device. If it is still ineffective, an early warning will be pushed to the user APP, and the repair time is ≤5 seconds.

3. The home network intelligent optimization system based on AI algorithm according to claim 1, characterized in that: The user interaction module includes a real-time dashboard, manual intervention, optimization log, and personalized settings. The real-time dashboard includes device bandwidth usage (histogram), signal heat map, and network health score. The manual intervention includes: adjusting priority and bandwidth locking. The optimization log records the execution time, content, and effect of the policy. The personalized settings include custom do not disturb periods and a high-priority device whitelist.

4. The home network intelligent optimization system based on AI algorithm according to claim 1, characterized in that: In the full-link monitoring module, when the monitoring object is the device status, the main monitored key indicators are: online / offline, MAC address, IP mapping, the collection frequency is 1 time / 30 seconds, and there is no abnormal threshold; when the monitoring object is the network load, the main monitored key indicators are upload / download rate and bandwidth occupancy rate, the collection frequency is 1 time / 5 seconds, and the abnormal threshold is: single device occupancy rate ≥50% (triggering attention); when the monitoring object is the transmission quality, the main monitored key indicators are: delay (RTT), packet loss rate, and number of retransmissions. The collection frequency is 1 time / 2 seconds, and the abnormal threshold is: delay ≥100ms or packet loss rate ≥3% (triggering warning); when the monitoring object is the signal environment, the main monitored key indicators are RSSI value and the number of co-channel devices. The collection frequency is 1 time / 10 seconds, and the abnormal threshold is: RSSI ≤-85dBm (weak signal). The full-link monitoring module uses 5G edge computing nodes to locally process raw data and only uploads characteristic values ​​(such as load peaks, abnormal events) to reduce transmission delay.

5. The home network intelligent optimization system based on AI algorithm according to claim 1, characterized in that: The LSTM load prediction model takes as input the bandwidth usage rate of the past hour, the number of online devices, and a timestamp (weekday / weekend marker) and outputs a load forecast curve for the next 15 minutes. Its training logic is as follows: Initially, it imports 30 days of historical data, and iterates training with newly added data daily, using the Adam optimizer.

6. The home network intelligent optimization system based on AI algorithm according to claim 1, characterized in that: The K-means device clustering algorithm, clustering features: device Type (work / entertainment / IoT), average bandwidth demand, usage time, output priority label (1-5 levels), (where 1 is the highest, such as a video conferencing computer, and 5 is the lowest, such as an idle smart socket).

7. The AI ​​algorithm-based home network intelligent optimization system according to claim 1, characterized in that: The bottleneck identification logic uses a decision tree model to determine the cause of congestion: if a single device's bandwidth usage is ≥70% and persists for 5 minutes, a "device overload" prompt is displayed; if there are ≥4 devices on the same channel and the RSSI fluctuation is ≥15dBm, a "signal interference" prompt is displayed; and if the gateway CPU usage is ≥90% and the latency is ≥200ms, a "routing performance deficiency" prompt is displayed.

8. The AI ​​algorithm-based home network intelligent optimization system according to claim 1, characterized in that: The bandwidth allocation formula in the dynamic resource scheduling module is:

9. Among them: : allocated bandwidth of device i; : The real-time requirements of device i (e.g., 4K video requires 20Mbps); : total household bandwidth; : Elastic bandwidth pool (20% of total bandwidth); (base weight), (priority weight), optimized through A / B testing.

10. The home network intelligent optimization system based on AI algorithm according to claim 1, characterized in that: The QoS adaptation in the dynamic resource scheduling module is as follows: low-latency applications (games / video conferencing) are marked as EF (accelerated forwarding), non-real-time applications (downloads / cloud synchronization) are marked as BE (best effort forwarding), and the speed limit is ≤ 30% of the total bandwidth.

11. The method of the home network intelligent optimization system based on AI algorithm according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: System initialization (takes about 1 hour); Step 2: Real-time monitoring and data preprocessing; Step 3: AI load prediction and bottleneck identification; Step 4: Dynamic scheduling and execution (response ≤ 2 seconds); Step 5: Feedback and model iteration (performed every morning); Step 6: Safety and redundancy design.

Images