Ultra-dense networking data transmission method for smart home communication

By constructing a three-dimensional interference map and multi-agent reinforcement learning decision-making at the core node of the home, and combining it with cross-technology direct communication, the problems of cross-standard interference and heterogeneous service latency in ultra-dense smart home networking are solved, thereby improving spectrum utilization and optimizing service latency.

CN123123476APending Publication Date: 2026-11-17上海嘉晨兴信息科技有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611350731.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the context of ultra-dense heterogeneous networking in smart homes, devices with different underlying communication protocols such as Wi-Fi, Bluetooth, and Zigbee coexist in the same unlicensed frequency band, leading to intense channel contention and severe cross-standard interference. At the same time, multiple services such as home security video, voice control commands, and environmental sensor data are concurrently running with highly heterogeneous latency requirements. Existing congestion control schemes based on a single standard and interconnection standards that rely on gateway relays cannot achieve native-level channel coordination and differentiated latency guarantees between heterogeneous protocols, resulting in low overall spectrum utilization and large latency jitter for critical services.

Method used

At the core home node, the 2.4GHz baseband signal is periodically collected and preamble feature matched to construct a three-dimensional interference map of time-frequency-protocol. Combined with multi-agent reinforcement learning decision-making, microsecond-level dynamic slicing of channels and time slots is achieved. The channel reservation information is broadcast to heterogeneous devices through a cross-technology direct communication mechanism, so that they can directly avoid channels without going through gateway protocol conversion.

Benefits of technology

It significantly reduces the probability of collisions across different standards, improves overall spectrum utilization, ensures differentiated quality of service, reduces end-to-end scheduling latency, and enhances the native-level channel coordination and avoidance effect between devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123476A_ABST
    Figure CN123123476A_ABST
Patent Text Reader

Abstract

This invention discloses a method for ultra-dense network data transmission in smart home communication. This invention relates to the field of IoT communication technology. The method constructs a three-dimensional heterogeneous interference map (time-space-frequency) at the core home node through preamble sniffing and energy detection. Security video, voice commands, and sensor data services are modeled as independent intelligent agents. Based on the interference map, multi-agent reinforcement learning is used for differentiated channel time slot slicing scheduling. Channel reservation broadcast signals are generated through payload length modulation or transmission interval modulation, enabling heterogeneous devices to demodulate channel reservation information and actively avoid interference without gateway conversion. This invention can improve the spectrum utilization of ultra-dense heterogeneous networks, achieve differentiated latency protection for multiple services, and is suitable for smart home communication scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) communication technology, specifically to a method for ultra-dense network data transmission in smart home communication. Background Technology

[0002] With the popularization of smart home concepts, the home environment has evolved into a natural ultra-dense heterogeneous wireless network. In the limited indoor space, a large number of heterogeneous wireless devices, such as Wi-Fi routers, smart speakers, Bluetooth lights, Zigbee temperature and humidity sensors, and Thread security devices, coexist on unlicensed frequency bands such as 2.4GHz. On the one hand, ultra-dense networking brings serious co-frequency competition and cross-standard interference. Due to the different underlying communication protocols, the devices lack the ability to recognize each other's protocols, which not only prevents them from effectively sharing channels, but also causes vicious collisions, resulting in a sharp deterioration in the overall spectrum efficiency. On the other hand, the business needs in smart homes are highly heterogeneous. Security cameras require stable high bandwidth and low latency, voice assistant control commands require extremely low burst latency, and environmental sensors only need periodic small packet transmission. The concurrent operation of multiple types of services and highly heterogeneous latency requirements further exacerbate the difficulty of resource allocation.

[0003] At the level of known technologies, those skilled in the art know that mainstream commercial wireless chips generally support the acquisition of channel state information and received signal strength indication, which can be used for channel quality assessment. Wi-Fi, Bluetooth, and Zigbee have significant differences in modulation methods, frequency hopping rules, and preamble structures at the physical layer, which can theoretically be distinguished by signal feature classification. The academic community also has a research foundation for cross-protocol direct communication based on commercial off-the-shelf devices, proving the feasibility of transmitting information between devices of different protocols through specific energy pulse patterns without hardware modifications.

[0004] The closest existing technology to this invention is an upper-layer interconnection scheme based on the Matter standard. Its technical solution is as follows: smart home devices from different manufacturers and with different underlying protocols are uniformly connected to a home gateway. The gateway completes protocol conversion, session management, and unified encapsulation and forwarding of application layer data. The business data of each device must be processed by the gateway through multiple protocol stacks before they can communicate. In addition, existing traditional network congestion control schemes adjust the transmission rate by monitoring end-to-end round-trip latency and packet loss rate, while carrier-grade ultra-dense network interference suppression schemes alleviate inter-cell interference in homogeneous cellular networks through inter-base station coordinated scheduling, power control, and other means.

[0005] The existing solutions have the following drawbacks: First, relying on the gateway as the central hub for heterogeneous protocol conversion and service scheduling, the data from each terminal device must be forwarded through multiple protocol stacks. The problem of different wireless access technologies operating independently and interfering with each other at the physical and link levels still exists, making it impossible to detect and suppress cross-protocol interference at the source. Second, the existing resource scheduling strategies fail to fully consider multiple constraints such as heterogeneous indoor protocols, small-scale mobility, and limited computing power of smart home devices, and cannot provide differentiated latency guarantees for concurrent services such as security video, voice commands, and sensor data. Third, the existing solutions lack a low-latency direct scheduling method for mixed concurrent indoor services, and relying on multiple protocol conversions through the gateway introduces additional latency, making it impossible to achieve native-level channel avoidance between devices. Therefore, we propose an ultra-dense network data transmission method for smart home communication. Summary of the Invention

[0006] The purpose of this invention is to provide a method for ultra-dense network data transmission in smart home communication.

[0007] The technical problem this invention aims to solve is as follows: In the ultra-dense heterogeneous networking environment of smart homes, devices with different underlying communication protocols such as Wi-Fi, Bluetooth, and Zigbee coexist in the same unlicensed frequency band. These devices lack the ability to recognize each other's protocols, leading to intense channel competition and severe cross-standard interference. At the same time, multiple services such as home security video, voice control commands, and environmental sensor data are concurrently running with highly heterogeneous latency requirements. Existing congestion control schemes based on a single standard and interconnection standards that rely on gateway relays cannot achieve native-level channel coordination and differentiated latency guarantees between heterogeneous protocols at the physical and link layers. This results in low overall spectrum utilization and large latency jitter for critical services.

[0008] To address the aforementioned technical problems, this invention provides a method for ultra-dense network data transmission in smart home communication, comprising three mutually cooperating and closed-loop technical steps: First, at the core home node, periodic sampling and preamble feature matching of the 2.4GHz baseband signal are performed to determine the communication protocol type of the signal. Combined with the received signal strength indication value, a time-frequency-protocol ternary occupancy record is formed. This record is then summarized according to time, frequency, and spatial dimensions to construct a three-dimensional heterogeneous interference spectrum tensor covering time, frequency channel, and spatial nodes. This interference spectrum tensor is periodically updated using a sliding window method. Second, security video services, voice command services, and sensor data services are modeled as independent intelligent agents, with each agent sharing the three-dimensional heterogeneous interference spectrum tensor. The system takes environmental state input as input, combines the queue length and latency sensitivity level of each service flow, and makes microsecond-level dynamic slicing decisions for channels and time slots. After resource conflict arbitration, it outputs a conflict-free channel time slot allocation scheme and iteratively updates the value function parameters of each agent based on the execution results. Thirdly, in response to the output of the channel time slot allocation scheme, it triggers the generation of channel reservation time schedule information in real time. After binary encoding, it is mapped into a Wi-Fi data packet sequence with a specific energy pulse pattern using payload length modulation or transmission interval modulation and broadcasts it. This allows heterogeneous devices such as Zigbee and Bluetooth within the coverage area to directly demodulate the channel reservation information through their existing idle channel assessment or received signal strength sampling mechanism without going through gateway protocol conversion, and actively avoid the channel during the corresponding reservation period.

[0009] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0010] 1. This invention constructs a three-dimensional heterogeneous interference map of time, space, and frequency at the physical layer of the core node in the home, enabling real-time joint sensing of heterogeneous protocol signals such as Wi-Fi, Bluetooth, and Zigbee. This makes cross-standard interference visible, measurable, and predictable, thereby significantly reducing the probability of collisions caused by blind competition and improving the overall spectrum utilization.

[0011] 2. This invention models the allocation of multi-service resources as a multi-agent reinforcement learning decision-making process, runs a lightweight learning model on the core node, and realizes microsecond-level dynamic slicing of channels and time slots, so that services of different priorities can obtain differentiated quality of service guarantees that match their latency sensitivity.

[0012] 3. This invention utilizes a cross-technology direct communication mechanism, enabling core nodes to directly broadcast channel reservation information to heterogeneous devices without gateway protocol conversion. Without modifying the existing hardware of heterogeneous IoT devices, it suppresses sudden congestion competition from the source, achieves native-level channel avoidance between devices, and reduces end-to-end scheduling latency. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the overall closed-loop process in an embodiment of the present invention;

[0014] Figure 2 This is a schematic diagram of the sub-process for constructing a three-dimensional interference map in an embodiment of the present invention;

[0015] Figure 3 This is a schematic diagram of the multi-agent reinforcement learning scheduling sub-process in an embodiment of the present invention. Detailed Implementation

[0016] The following is in conjunction with the appendix Figure 1-3 The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0017] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0018] The technical solution of the present invention will be further described in detail below with reference to specific embodiments. The smart home communication ultra-dense networking data transmission method provided in this embodiment includes three technical links that cooperate with each other and operate in a closed loop: heterogeneous protocol cross-layer perception and interference map construction, differentiated service quality slice scheduling based on multi-agent reinforcement learning, and lightweight collaborative feedback and congestion intervention based on cross-technology direct communication. All of them are deployed and run on the core node of the home without any hardware modification to the existing heterogeneous terminal devices.

[0019] The core node of the home integrates a preamble sniffing and protocol recognition module, a time-space-frequency three-dimensional interference map construction module, a multi-agent reinforcement learning scheduling engine based on a neural network processor, and a cross-technology direct connection communication and collaborative broadcasting module. The core node continuously monitors all wireless signals in the 2.4GHz band through its radio frequency front-end. Within its coverage area, there are heterogeneous terminals such as security cameras, smart speakers, Zigbee temperature and humidity sensors, Bluetooth lights, and Thread security sensors. The core node maintains a regular communication link with each Wi-Fi device. At the same time, it broadcasts channel reservation information to unrelated heterogeneous devices such as Zigbee and Bluetooth through a cross-technology direct connection communication mechanism. This link can directly act on the physical layer of heterogeneous devices without any gateway protocol conversion, which reflects the technical feature of this embodiment without gateway relay.

[0020] In the heterogeneous protocol cross-layer sensing and spatiotemporal-frequency three-dimensional interference map construction stage, the core node performs radio frequency signal acquisition on multiple sub-channels in the 2.4GHz band at a fixed period to obtain baseband in-phase and quadrature component sampling data. Preamble features are extracted from the acquired baseband data. These extracted features are then compared one by one with pre-stored Wi-Fi orthogonal frequency division multiplexing training sequence features, Bluetooth Gaussian frequency shift keying frequency hopping preamble features, and Zigbee offset orthogonal phase shift keying chirp preamble features to determine the protocol type of the signal segment. This is combined with the received signal strength index obtained from energy detection. The system records the timestamp of the signal's occurrence, the frequency channel number it occupies, and its duration, forming a time-frequency-protocol ternary occupancy record. All occupancy records collected within a preset observation period are summarized according to the time, frequency, and spatial dimensions to construct a three-dimensional heterogeneous interference spectrum tensor. The interference spectrum tensor is periodically updated using a sliding window method, removing historical data that exceeds the window duration and incorporating the latest collected occupancy records. This ensures that the interference spectrum reflects the current channel occupancy status in real time, and the spectrum is passed to the subsequent reinforcement learning scheduling engine as a shared environment state.

[0021] The construction of the above interference map tensor can be expressed by the following formula:

[0022] ;

[0023] In the formula, Indicates time slice Frequency Channel Spatial nodes The overall interference intensity value at the location, This indicates the total number of heterogeneous protocol types involved in the identification. As an indicator function, when in time slice Frequency Channel Spatial nodes Protocol detected The value is 1 when the signal is occupied, and 0 otherwise. For the agreement The severity of interference to the main Wi-Fi network in this solution is weighted by a coefficient that is adaptively adjusted based on historical collision statistics.

[0024] In the multi-service differentiated service quality slicing scheduling process based on multi-agent reinforcement learning, the input of the reinforcement learning scheduling engine is the real-time interference map output by the interference map construction module and the priority and latency requirements parameters of each service flow. The security video service, voice command service, and sensor data service are abstracted into Agent 1, Agent 2, and Agent 3, respectively. The three agents share the environmental state input provided by the interference map module. Each agent contains a state observation unit, an action selection unit, and a value function update unit. After the actions of each agent are summarized, the resource conflict arbitration unit performs unified channel time slot allocation. After the allocation result is executed, information such as actual throughput, latency, and number of collisions is fed back. The reward calculation unit generates a reward signal and returns it to each agent to drive the iterative optimization of the value function.

[0025] Specifically, the value function parameters of each agent corresponding to each service flow are first initialized, and the learning rate, discount factor, and exploration probability initial values ​​are set. At the beginning of each scheduling cycle, each agent obtains the current environmental state, which includes the time-space-frequency interference map, the length of the data queue to be sent for this service flow, and the latency sensitivity level of this service flow. Based on the current state, each agent selects the channel and time slot allocation action for this cycle from the candidate channel time slot set according to the strategy of combining exploration and utilization. The resource conflict arbitration unit performs conflict detection and coordination on the actions selected by each agent, outputs the final channel time slot microsecond-level slice allocation scheme without conflict, and executes data transmission according to the scheme. Then, the execution results of this cycle are collected, including the actual throughput, whether the latency exceeds the latency threshold of the corresponding service, and the number of collisions caused by interference. Based on this, the instant reward value of each agent is calculated. Based on the instant reward value and the value estimate of the next state, each agent iteratively updates its own value function parameters and determines whether the preset iteration termination condition is met. If not, it returns to the process of obtaining the environmental state and enters the next scheduling cycle. If it is met, it outputs the current converged scheduling strategy for direct use in subsequent cycles.

[0026] The update of the value function can be expressed by the following formula:

[0027] ;

[0028] In the formula, Represents intelligent agents In state Take action below The value estimate, The learning rate controls the step size for updating the value function. For instant reward value, This is a discount factor used to measure the weight of future rewards on current decisions. To perform the action The next state to enter These are candidate actions for the next state. Represents intelligent agents The estimate with the highest value among multiple candidate actions in the next state;

[0029] The instant reward value can be calculated using the following formula:

[0030] ;

[0031] In the formula, Indicates the first The actual throughput achieved by this service flow within the scheduling period. Indicates the first The penalty amount for the latency of this service flow exceeding its preset threshold during the scheduling period. Indicates the first The number of collisions caused by cross-protocol interference within the scheduling period. , , These are the weighting coefficients for throughput gain, latency penalty, and collision penalty, respectively. These can be preset according to business priority; for example, security video services should have higher weighting coefficients. Voice command services correspond to higher ;

[0032] In another implementation of this embodiment, the multi-agent reinforcement learning scheduling engine can also adopt a multi-agent deep deterministic policy gradient algorithm based on policy gradient. Each agent maintains an independent behavior network and evaluation network. Channel time slot allocation decisions are completed through centralized training and distributed execution, which can also achieve the technical effect of differentiated quality of service slice scheduling.

[0033] In the lightweight collaborative feedback and congestion intervention stage based on cross-technology direct communication, the purpose of the cross-technology direct communication collaborative broadcast module is to directly transmit the channel reservation information involved in the channel time slot allocation result generated by the scheduling engine to heterogeneous devices such as Zigbee and Bluetooth that are not associated with the master node. After the master node generates the channel reservation time schedule information according to the scheduling result, it enters the binary encoding unit to complete the bit sequence conversion. The encoded bit sequence enters the payload length or transmission interval modulation unit and is mapped into a series of Wi-Fi data packets with a specific length or interval combination. This data packet sequence is broadcast to the surrounding space. After the idle channel assessment or received signal strength sampling unit of the heterogeneous device captures the energy pulse sequence, it enters the local demodulation unit to restore the original channel reservation time schedule bit sequence. The channel avoidance decision unit of the heterogeneous device actively stops channel contention in the corresponding time period according to the reservation time period information obtained by demodulation, thereby completing a complete cross-protocol collaborative avoidance action. This process is executed cyclically with the scheduling cycle.

[0034] Specifically, the master node generates a channel reservation schedule based on the channel time slot allocation results of the current scheduling cycle. This information includes the reservation start time, reservation duration, and the channel number involved. The generated channel reservation schedule is encoded into a binary bit sequence. Using payload length modulation or transmission interval modulation, the encoded bit sequence is mapped bit by bit into a series of Wi-Fi data packets with specific length combinations or specific transmission interval combinations, forming an energy pulse pattern that can be directly perceived by the physical layer of surrounding heterogeneous devices. The master node continuously broadcasts the modulated data packet sequence. Heterogeneous devices such as Zigbee and Bluetooth within the coverage area capture the broadcast energy pulse sequence through their existing idle channel assessment mechanism or received signal strength sampling mechanism, and demodulate it locally according to the mapping rule opposite to the modulation process to restore the channel reservation schedule information. Based on the restored reservation schedule, the heterogeneous devices actively back off and suspend channel contention behavior during the corresponding reservation period, thereby achieving native-level channel avoidance with the Wi-Fi network where the master node is located.

[0035] The mapping relationship of the net charge length modulation can be expressed by the following formula:

[0036] ;

[0037] In the formula, Indicates the first The payload length of each data packet to be sent. This indicates the pre-agreed baseline net load length. Indicates the number to be transmitted Bit information, with a value of 0 or 1. This represents the payload length modulation step size used to distinguish between bit 0 and bit 1. Its value must be greater than the granularity threshold that can be distinguished by the existing idle channel evaluation or received signal strength sampling accuracy of surrounding heterogeneous devices.

[0038] In another implementation of this embodiment, in addition to using payload length modulation or transmission interval modulation, the modulation method for cross-technology direct communication collaborative feedback can also use a modulation method based on the jitter mode of data packet transmission time. That is, channel reservation information is encoded by controlling the combination of small offsets between the transmission times of multiple consecutive data packets. Heterogeneous devices complete demodulation by detecting the statistical regularity of the start and end times of channel occupancy. Similarly, the cross-protocol transmission of channel reservation information can be achieved without changing the hardware of heterogeneous devices.

[0039] The overall execution flow of the method in this embodiment is as follows: The process begins in the real-time interference map perception stage to obtain the current time-space-frequency interference map. Then, it enters the service priority identification stage to obtain the type and latency requirements of each service flow to be transmitted. Next, it enters the reinforcement learning dynamic resource allocation stage to obtain the channel time slot slice allocation scheme for this period. Then, it enters the cross-technology direct communication cooperative broadcast stage to broadcast the channel reservation information in the allocation result to heterogeneous devices. Then, it enters the data transmission execution stage, where each service flow actually sends data according to the allocation scheme and the heterogeneous device avoidance results. Finally, it enters the effect feedback update stage to collect the throughput, latency, and collision statistics for this period. On the one hand, it feeds back to the reinforcement learning scheduling engine for value function updates, and on the other hand, it feeds back to the interference map construction module for correcting the protocol interference weight coefficients. The process then returns to the real-time interference map perception stage and enters the next scheduling cycle, forming a complete closed loop.

[0040] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for ultra-dense network data transmission in smart home communication, characterized in that, include: The steps for constructing a three-dimensional heterogeneous interference map of time, space, and frequency specifically include: The home core node collects radio frequency signals on multiple sub-channels in the 2.4 GHz band at a pre-set collection period, acquiring baseband in-phase and quadrature component sampling data; preamble feature extraction is performed on the sampled data; the extracted features are compared one by one with pre-stored Wi-Fi orthogonal frequency division multiplexing training sequence features, Bluetooth Gaussian frequency shift keying frequency hopping preamble features, and Zigbee offset orthogonal phase shift keying chirp preamble features to determine the communication protocol type of the corresponding signal; combined with the received signal strength indication value obtained from energy detection, the timestamp of the signal occurrence, the frequency channel number occupied, and the duration are recorded to form a time-frequency-protocol ternary occupancy record; multiple occupancy records collected within a preset observation period are summarized according to the time dimension, frequency dimension, and spatial dimension to construct a three-dimensional heterogeneous interference map tensor; the three-dimensional heterogeneous interference map tensor is periodically updated in a sliding window manner, historical data exceeding the window duration is removed, and the latest collected occupancy records are included; The steps for differentiated quality of service (QoS) slicing scheduling based on multi-agent reinforcement learning specifically include: classifying security video services, voice command services, and sensor data services as independent agents; each agent acquiring the current environmental state in each scheduling cycle, the environmental state including the three-dimensional heterogeneous interference spectrum tensor, the length of the data queue to be sent for the corresponding service flow, and the latency sensitivity level of the corresponding service flow; each agent selecting the channel and time slot allocation action for the current cycle from the candidate channel and time slot set based on the current environmental state; performing conflict detection and coordination on the channel and time slot allocation actions selected by multiple agents; outputting a conflict-free channel and time slot allocation scheme; executing data transmission according to the channel and time slot allocation scheme; collecting the execution results for the current cycle, the execution results including actual throughput, whether the latency exceeds the latency threshold of the corresponding service, and the number of collisions caused by cross-protocol interference; calculating the instant reward value of each agent based on the execution results; and iteratively updating the value function parameters of each agent based on the instant reward value and the value estimate of the next state. The steps for channel reservation broadcasting based on cross-technology direct communication specifically include: generating channel reservation schedule information in real time in response to the output of the channel time slot allocation scheme, realizing direct mapping from scheduling decisions to cross-protocol broadcasting; the channel reservation schedule information includes the reservation start time, reservation duration, and the channel number involved; encoding the channel reservation schedule information into a binary bit sequence; mapping the encoded binary bit sequence bit by bit into a Wi-Fi data packet sequence with a specific length combination and a specific transmission interval combination using a preset modulation method; continuously broadcasting the modulated Wi-Fi data packet sequence; enabling Zigbee devices and Bluetooth devices within the coverage area to capture the Wi-Fi data packet sequence through one of their own idle channel evaluation mechanism and received signal strength sampling mechanism; and completing local demodulation according to the mapping rule opposite to the preset modulation method to restore the channel reservation schedule information; and actively suspending channel contention behavior during the corresponding reservation period.

2. The ultra-dense networking data transmission method for smart home communication according to claim 1, characterized in that: The comprehensive interference intensity value at each time slice, frequency channel, and spatial node in the three-dimensional heterogeneous interference spectrum tensor is obtained by multiplying the signal occupancy indication results corresponding to various communication protocols with their respective interference severity weighting coefficients and then summing them up. The signal occupancy indication result is 1 when the corresponding communication protocol signal is detected to occupy the corresponding time slice, frequency channel, and spatial node, and 0 when it is not detected to occupy the corresponding time slice, frequency channel, and spatial node. The interference severity weighting coefficient is adaptively adjusted based on historical collision statistics.

3. The ultra-dense networking data transmission method for smart home communication according to claim 1, characterized in that: The instant reward value of each intelligent agent is obtained by weighting and summing the actual throughput achieved by the current service flow in this period, the penalty amount for the latency of the current service flow exceeding the preset latency threshold in this period, and the number of collisions caused by cross-protocol interference in this period, according to a preset throughput benefit weighting coefficient, latency penalty weighting coefficient, and collision penalty weighting coefficient. Among them, the throughput benefit weighting coefficient corresponding to the security video service is higher than the throughput benefit weighting coefficient corresponding to the voice command service, and the latency penalty weighting coefficient corresponding to the voice command service is higher than the latency penalty weighting coefficient corresponding to the security video service.

4. The ultra-dense networking data transmission method for smart home communication according to claim 1, characterized in that: When iteratively updating the value function parameters of each agent, the target value of the value function is calculated based on the immediate reward value of the current period, the pre-set discount factor, and the value estimate that ranks first among the value estimates of multiple candidate actions in the next state. The current value function parameters are then adjusted to the target value according to the pre-set learning rate.

5. The ultra-dense networking data transmission method for smart home communication according to claim 1, characterized in that: The preset modulation method is payload length modulation, which maps each bit of encoded information to the payload length of the corresponding data packet to be sent. The payload length is equal to the sum of the products of the pre-agreed reference payload length, the payload length modulation step size, and the corresponding bit value. The value of the payload length modulation step size is greater than the distinguishable granularity threshold of one of the existing idle channel evaluation mechanism and the received signal strength sampling mechanism of the corresponding heterogeneous device.

6. The ultra-dense networking data transmission method for smart home communication according to claim 1, characterized in that: The preset modulation method is a transmission interval modulation method, which maps each bit of the encoded information to the transmission interval between two adjacent data packets to be transmitted, and different bit values ​​correspond to different transmission interval values.

7. The ultra-dense networking data transmission method for smart home communication according to claim 1, characterized in that: Each agent maintains an independent behavior network and an independent evaluation network. It adopts a centralized training and distributed execution approach, and iteratively updates the network parameters of the behavior network based on the output of the evaluation network to obtain the channel and time slot allocation actions for the current period.

8. A super-dense network data transmission system for smart home communication, characterized in that, This includes a preamble sniffing and protocol identification module set up at the core node of the home, a time-space-frequency three-dimensional interference map construction module, a multi-agent reinforcement learning scheduling engine based on a neural network processor, and a cross-technology direct communication and collaborative broadcasting module; The preamble sniffing and protocol identification module is used to collect radio frequency signals on multiple sub-channels in the 2.4GHz band at a fixed period, extract preamble features from the collected baseband sampling data, and determine the communication protocol type to which the corresponding signal belongs. The time-space-frequency three-dimensional interference map construction module is used to combine the received signal strength indication value obtained by energy detection to form a time-frequency-protocol ternary occupancy record, and to summarize multiple occupancy records within a preset observation period according to the time dimension, frequency dimension, and spatial dimension to construct a three-dimensional heterogeneous interference map tensor, and to periodically update the three-dimensional heterogeneous interference map tensor in a sliding window manner. The multi-agent reinforcement learning scheduling engine is used to correspond security video service, voice command service and sensor data service to an independent agent. Based on the three-dimensional heterogeneous interference spectrum tensor and the queue length and latency sensitivity level of each service stream, it performs channel and time slot allocation actions, conflict coordination and iterative update of value function parameters, and outputs channel and time slot allocation scheme. The cross-technology direct communication cooperative broadcast module is used to generate channel reservation schedule information in real time in response to the output of the channel time slot allocation scheme, encode and modulate the channel reservation schedule information into a Wi-Fi data packet sequence, and broadcast it to heterogeneous devices within the coverage area, so that the heterogeneous devices can actively suspend channel contention behavior after demodulating the channel reservation schedule information.

9. The ultra-dense network data transmission system for smart home communication according to claim 8, characterized in that: The cross-technology direct communication cooperative broadcasting module uses either payload length modulation or transmission interval modulation to modulate the binary bit sequence corresponding to the encoded channel reservation time schedule information.

10. The ultra-dense network data transmission system for smart home communication according to claim 8, characterized in that: The home core node is one of the main routing device and the gateway device. The preamble sniffing and protocol identification module, the time-space-frequency three-dimensional interference map construction module, the multi-agent reinforcement learning scheduling engine, and the cross-technology direct communication cooperative broadcasting module are connected in sequence to form a closed-loop data flow relationship in which interference perception, scheduling decision, cooperative broadcasting, and effect feedback are executed sequentially and cyclically.