A star flash near field communication group networking method

Abstract

This invention belongs to the field of wireless communication technology, specifically a group networking method for star-flash near-field communication. It includes a management node dividing nodes into logical groups and establishing a resource index mapping table to ensure orthogonal isolation of scheduling windows; achieving bit-level directional wake-up through a scheduling mask bitmap, enabling member nodes to adaptively control RF switches based on mask states; introducing an artificial intelligence model to predict traffic trends; pre-allocating elastic buffer resource blocks; and performing access coordination control based on backoff weight factors. By introducing a scheduling mask bitmap mechanism based on bit-level offsets, the energy efficiency limitations of traditional full-broadcast scheduling modes on member nodes are changed, allowing member nodes to achieve millisecond-level rapid sleep switching of the RF front-end through extremely short header field determination.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, specifically a group networking method for star-flash near-field communication. Background Technology

[0002] With the profound evolution of industrial automation transformation and intelligent connectivity, short-range wireless communication technology has penetrated from the consumer electronics field to core industrial control scenarios. Industrial applications have stringent requirements for deterministic latency, nanosecond-level time synchronization accuracy, and ultra-high connection robustness of communication links. As a new generation of short-range wireless communication standard, StarFlash technology has built a technical foundation for extremely low latency and microsecond-level synchronization. Existing communication protocols use a logical grouping management mode to divide member nodes, which facilitates the management node to schedule air interface resources.

[0003] The existing StarSpark technology group networking relies on a logical partitioning management mode. The management node adopts a full broadcast scheduling logic, which issues scheduling instructions covering all logical groups in each scheduling cycle. This mode can ensure that all active member nodes synchronously obtain the global view of the system during the initial operation phase of the system, and maintain the consistency of network status.

[0004] The full broadcast mechanism of existing networking technologies leads to redundancy in physical layer resource utilization, invalid node wake-ups, and increased energy consumption, shortening the lifespan of energy-constrained equipment. When multiple groups access concurrently, air interface conflicts are prone to occur, resulting in communication latency jitter and control loop instability. Furthermore, there is a lack of a fine-grained scheduling scheme that deeply couples logical groups with physical resources, which cannot meet the industrial-grade deterministic communication requirements.

[0005] Therefore, the present invention provides a group networking method for star-flash near-field communication. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is: a group networking method for star-flash near-field communication, comprising the following steps: The management node divides the member nodes into several logical groups based on the service flow characteristics and latency sensitivity threshold of the access member nodes, assigns a unique group identifier to each logical group, and constructs a resource index mapping table that establishes the correspondence between the logical groups and physical layer resource blocks. The management node generates a scheduling mask bitmap based on the interaction requirements of each logical group. The scheduling mask bitmap contains multiple bits, and the index position of the bits corresponds to the group identifier. The management node encapsulates the scheduling mask bitmap in downlink broadcast signaling and sends it out. After receiving the downlink broadcast signaling, the member node extracts the bit state corresponding to the group identifier of the logical group to which it belongs from the scheduling mask bitmap; If the bit state is the first preset state, the member node accurately locates the physical resource block according to the resource index mapping table to perform service interaction; if the bit state is the second preset state, the member node shuts down the radio frequency transceiver link and enters standby mode.

[0008] Preferably, the step of the management node performing pre-configuration and logical mapping of grouped resource spaces includes: The management node extracts the service flow characteristics of each accessing member node through the service feature perception module. The service flow characteristics include data packet arrival interval distribution, average message length, service burst probability, and historical traffic envelope. The management node obtains the latency sensitivity threshold of the member node, which is defined as the maximum allowable latency from the time the physical layer service data unit enters the transmission buffer until it is correctly received. The management node, by combining the peak throughput requirements and average throughput requirements of the member nodes, divides the member nodes across the entire network into logical groups. The group identifier is configured as a 16-bit binary code; Preferably, the step of establishing a deterministic correspondence between the logical group and the physical layer resource block using the resource index mapping table includes: The management node defines the subcarrier index set, symbol index range, and power mask parameters for each logical group through the physical layer resource block; The management node divides the communication superframe into multiple independent scheduling windows that do not overlap in the time dimension in the time domain, and sets a physical isolation interval of a preset length between each independent scheduling window. The management node ensures the orthogonality of different logical groups on air interface resources by setting the physical isolation interval. The orthogonality is configured to suppress intermodulation interference between groups below the noise floor through time offset alignment. The management node performs pre-configuration and logical mapping of grouped resource spaces. In this process, the management node divides all member nodes in the network into several logical groups with specific functional attributes based on the service flow characteristics, latency sensitivity thresholds and throughput requirements of the access member nodes. The management node assigns a globally unique group identifier to each logical group and simultaneously builds a system-level resource index mapping table. The resource index mapping table establishes a deterministic correspondence between logical groups and physical layer resource blocks. The management node divides communication superframes into multiple independent scheduling windows in the time domain and ensures the orthogonality of different logical groups on air interface resources by setting physical isolation intervals. This deep coupling between the logical layer and the physical layer provides a basic data foundation for subsequent fine-grained resource scheduling.

[0009] The management node executes a targeted and precise wake-up mechanism based on the scheduling mask bitmap. At the beginning of each scheduling cycle, the management node dynamically generates a scheduling mask bitmap based on the downlink data backlog status and uplink authorization requirements of each logical group. The scheduling mask bitmap consists of multiple consecutive bits, where the index position of each bit is uniquely bound to a specific group identifier. When the management node determines that a specific logical group has valid business interaction within the current scheduling window, it sets the corresponding bit to the first preset state; if there is no business interaction requirement, it sets it to the second preset state. The management node then encapsulates the generated scheduling mask bitmap in the header field of the downlink broadcast signaling for distribution.

[0010] Preferably, the step of the member node executing the RF front-end adaptive control logic based on the scheduling mask bitmap includes: The member node controls the wake-up time through an internal synchronization timer. After entering the preset synchronization capture window, it activates the narrowband receiving link in the radio frequency transceiver link. The narrowband receiving link includes a low-noise amplifier, a mixer, and an analog-to-digital converter. In the narrowband receive link active state, the member node performs baseband demodulation on the header field of the downlink broadcast signaling; The logic determination module inside the member node extracts and determines the bit state corresponding to the group identifier in the scheduling mask bitmap; If the determination result is the second preset state, the medium access control layer of the member node triggers a radio frequency shutdown command to the power management unit to shut down the active devices in the radio frequency transceiver link and skip the reception, sampling and demodulation of the subsequent full broadcast payload of the current communication superframe. Member nodes execute RF front-end adaptive control logic based on the scheduling mask bitmap. After entering the preset synchronization acquisition window, member nodes only activate the narrowband receive link of the RF front-end and perform baseband demodulation on the header field of the downlink broadcast signaling. Member nodes extract and determine the bit status corresponding to their group identifier in the scheduling mask bitmap. If the determination result is the second preset state, the member node immediately triggers an RF shutdown command, shuts down the RF transceiver link and enters a low-power standby mode, skipping the parsing of the subsequent full broadcast payload of the superframe. If the determination result is the first preset state, the member node maintains the active state of the RF link and accurately locates the physical resource block of its group according to the resource index mapping table. Through this bit-level mask-driven targeted wake-up mechanism, the on-demand allocation of member node power consumption is realized, eliminating the invalid demodulation overhead of non-target groups.

[0011] Preferably, the step of the management node performing dynamic resource elastic reconfiguration includes: The management node integrates a resource prediction model, which is configured to determine the traffic evolution trend of the logical group. The resource prediction model includes an input layer, a feature extraction layer, a temporal correlation layer, and an output layer. The input layer receives the historical traffic fingerprint sequence, current queue depth offset, node activity distribution data, and air interface channel quality parameters of each logical group. The historical traffic fingerprint sequence is configured to include a preset number of throughput distribution data within the communication superframe; To achieve dynamic resource elastic reconfiguration with extremely low signaling overhead, this invention introduces an artificial intelligence-based resource prediction model in the management node. The resource prediction model is configured to make forward-looking judgments on the traffic evolution trend of logical groups. The construction of the resource prediction model includes an input layer, a feature extraction layer, a temporal correlation layer, and an output layer. The input layer is responsible for receiving the historical traffic fingerprint sequence, current queue depth offset, node activity distribution data, and air interface channel quality parameters of each logical group.

[0012] Preferably, the internal processing logic of the resource prediction model includes: The feature extraction layer uses a convolutional neural network unit composed of multiple nonlinear activation functions to perform multi-scale convolution processing on the spatial tensor of the traffic data to extract the spatial features of the traffic data. The temporal correlation layer accumulates the extracted spatial features through a cyclic processing structure to capture the temporal periodicity and random burst probability of the service flow. The output layer generates a predicted resource demand matrix, which is used to indicate the probability of resource surplus or resource shortage for each of the logical groups within a future preset number of communication superframes. Preferably, the training and resource injection steps of the resource prediction model include: The management node collects historical operation logs, including interference intensity, node size, and business topology, as a training sample set. During training, mean squared error is used as the target loss function to quantify the residual between the predicted resource proportion and the actual demand, and the weight parameters between neurons inside the model are corrected through the backpropagation algorithm. When the prediction confidence level output online by the resource prediction model exceeds a preset threshold, the management node adopts the predicted resource demand matrix and pre-defines elastic buffer resource blocks at the physical layer. When a burst of traffic occurs within the burst probability in the logical group, the management node injects the elastic buffer resource block into the target group by updating the resource index mapping table; In the internal connection logic of the resource prediction model, the feature extraction layer extracts the spatial features of traffic data through convolution processing of multiple nonlinear activation functions; the temporal correlation layer uses a cyclic processing structure to capture the temporal periodicity and random burst probability of the business flow; and the output layer finally generates a predicted resource demand matrix. The training process of the resource prediction model includes: the management node collects historical operation logs containing different interference intensities and node sizes as a training sample set; the mean squared error is used as the target loss function; the weight parameters between neurons inside the model are continuously corrected through the backpropagation algorithm; when the confidence of the model output exceeds a preset threshold, the management node pre-defines elastic buffer resource blocks at the physical layer according to the predicted resource demand matrix, thereby achieving smooth absorption of burst traffic without relying on high-frequency signaling interaction.

[0013] Preferably, the step of the management node executing the access coordination control logic includes: The management node calculates a specific backoff weight factor for each logical group based on the real-time perceived air interface channel quality and the service urgency of each logical group using a weighted equalization algorithm. The management node periodically publishes an access control list in the downlink broadcast signaling, and the access control list contains the mapping relationship of the backoff weight factor; When the service triggering module of the member node generates a random access request, the media access control unit of the member node extracts the backoff weight factor of the group to which it belongs, and adjusts the initial contention window size of the conflict backoff algorithm accordingly. The size of the initial competition window is positively correlated with the backoff weight factor configuration; The management node executes access coordination control logic based on backoff weight factors. To avoid signal collisions when multiple groups access concurrently, the management node calculates a specific backoff weight factor for each logical group based on the real-time perceived air interface channel quality and the service urgency of each logical group. The management node publishes an admission control list in the downlink signaling, which contains a controlled weight mapping relationship. When a member node initiates a random access request, its internal media access control unit extracts the backoff weight factor of its group and adjusts the initial contention window size of the conflict backoff algorithm accordingly. By allocating a smaller backoff weight factor to high-priority groups, fine-grained intervention in access probability at the physical layer is achieved, ensuring the access determinism of key control commands.

[0014] Preferably, the networking method further includes closed-loop feedback and link compensation steps: The physical layer measurement unit of the member node generates a link status report after each service interaction cycle. The link status report includes received signal strength indication, signal-to-noise ratio, and cyclic redundancy check result. The management node uses the collected link status reports from each of the logical groups to calculate the current link's channel capacity margin through a link quality assessment algorithm, and dynamically adjusts the modulation and coding strategy and transmit power offset of the logical group in the next cycle. If the management node determines that the communication quality of a specific logical group is lower than a preset reliability threshold, it triggers a group reorganization process and controls the affected member nodes to migrate to a backup clean frequency band via downlink commands. After each service interaction cycle, member nodes send a link status report to the management node, which includes received signal strength indication, signal-to-noise ratio, and cyclic redundancy check results. The management node uses the link status report and a built-in link quality assessment algorithm to dynamically adjust the modulation and coding strategy and transmit power offset of the group in the next cycle. If the management node determines that the communication quality of a specific group is continuously lower than a preset threshold, it triggers the group reorganization process and migrates the affected member nodes to a spare clean frequency band via downlink commands, thus ensuring the self-healing capability of the network system.

[0015] Preferably, the steps for the management node to execute the synchronization maintenance logic and energy efficiency management logic include: The management node transmits a synchronization fingerprint sequence at the beginning of each communication superframe. The synchronization fingerprint sequence consists of a pair of complementary Gole codes, and the autocorrelation function of the synchronization fingerprint sequence in the time domain is configured to have spike and sidelobe cancellation characteristics. The member node uses an internally integrated hardware digital phase-locked loop circuit to perform oversampled correlation peak detection on the synchronization fingerprint sequence, and compensates for the frequency deviation and phase shift of the local crystal oscillator based on the detected peak time. The switching of physical layer parameters configured in the management node is all completed within the cyclic prefix guard interval of the orthogonal frequency division multiplexing symbol; The management node determines the inactive duration of each logical group based on the output of the artificial intelligence resource prediction model, generates discontinuous reception cycle parameters, and the member node enters a hardware sleep state based on the discontinuous reception cycle parameters until the next predicted activation time corresponding to the scheduling mask bitmap. The management node executes high-precision synchronization maintenance logic based on Gory code sequences. At the beginning of the superframe, the management node transmits a synchronization fingerprint sequence containing high-intensity autocorrelation characteristics. Member nodes use internal digital phase-locked loop circuits to perform correlation peak detection on the synchronization fingerprint sequence and compensate for the frequency deviation and phase shift of the local crystal oscillator in real time. The management node ensures that when performing dynamic adjustment of resource boundaries, all parameter switching is completed within the cyclic prefix guard interval of the orthogonal frequency division multiplexing symbol. Through this physical layer phase continuity control, it can be ensured that the resource elastic adjustment process does not cause baseband signal parsing errors, maintaining nanosecond-level time synchronization accuracy. The management node also executes targeted dynamic alignment logic for group sleep cycles. Based on the output of the artificial intelligence resource prediction model, the management node determines the inactivity duration of each group and generates discontinuous reception cycle parameters accordingly. After receiving the parameters, member nodes enter a deep hardware sleep state until the predicted activation time corresponding to the next scheduling mask bitmap. This long-cycle energy efficiency management strategy, in conjunction with the short-cycle scheduling mask bitmap, constructs a multi-level energy-saving architecture.

[0016] The beneficial effects of this invention are as follows: 1. The star-flash near-field communication group networking method described in this invention introduces a scheduling mask bitmap mechanism based on bit-level offset, which changes the energy efficiency limitation of member nodes in the traditional full broadcast scheduling mode. Member nodes can realize rapid sleep switching of the radio frequency front-end through extremely short header field determination. In the scenario of large-scale access of multiple groups, the invalid power consumption of non-target group nodes is significantly reduced, which is conducive to extending the service life of terminal equipment.

[0017] 2. The group networking method for Star Flash near-field communication described in this invention can realize a paradigm shift from static reservation to prediction-driven dynamic reconstruction of physical layer resource slicing by constructing a resource prediction model based on deep learning. The model can complete the elastic expansion of physical resource blocks in advance before the arrival of peak business by extracting deep features from historical traffic fingerprints, which can avoid scheduling delay jitter caused by signaling reconstruction and ensure the deterministic index of communication links.

[0018] 3. The star-flash near-field communication group networking method described in this invention can suppress air interface contention conflicts in multi-group concurrent scenarios by implementing admission control logic based on backoff weight factors. The management node reduces the retransmission probability of data payload by dynamically adjusting the contention window weight, thereby significantly improving the deterministic throughput of the system.

[0019] 4. The star-flash near-field communication group networking method described in this invention, through the closed-loop feedback compensation and smooth synchronization switching mechanism established by this invention, can ensure that the time base of all network nodes remains highly consistent in complex dynamic networking environments. By performing resource boundary adjustment within the guard interval of orthogonal frequency division multiplexing symbols, the performance degradation caused by phase abrupt changes is avoided, thus providing a guarantee for the high-precision synchronization application of star-flash near-field communication. Attached Figure Description

[0020] The invention will now be further described with reference to the accompanying drawings.

[0021] Figure 1 This is a flowchart of a group networking method for star-flash near-field communication in this invention. Detailed Implementation

[0022] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0023] like Figure 1 As shown in the embodiment of the present invention, a group networking method for star-flash near-field communication includes the following steps: The management node performs pre-configuration and logical mapping of the grouped resource space. In this initial stage, the management node extracts the service flow characteristics of each access member node in multiple dimensions through its internally integrated service feature perception module. The service flow characteristics include, but are not limited to, the data packet arrival interval distribution, average message length, service burst probability and historical traffic envelope reported by the member node. The management node synchronously obtains the latency sensitivity threshold of each member node. This threshold defines the maximum allowable latency from the physical layer service data unit entering the transmission buffer to being correctly received. At the same time, based on the peak and average throughput requirements reported by member nodes, member nodes across the entire network are precisely divided into several logical groups with specific functional attributes. During this process, the management node will assign a globally unique group identifier to each logical group. This identifier is a 16-bit binary code that is exclusive throughout the entire network lifecycle. To ensure deep alignment between the logical layer and the physical layer, the management node synchronously constructs a system-level resource index mapping table. This resource index mapping table is stored in the high-speed static random access memory of the management node. It explicitly establishes a deterministic correspondence between each logical group and the physical layer resource block. Each physical layer resource block is defined by a specific set of subcarrier indexes, symbol index ranges, and power mask parameters. Furthermore, the management node divides the communication superframe into multiple independent scheduling windows that do not overlap in the time dimension in the time domain. By setting a preset physical isolation interval between each independent scheduling window, the orthogonality of different logical groups on air interface resources is achieved. This orthogonality, through strict time offset alignment, can ensure that intermodulation interference between groups is suppressed below the noise floor, providing a deterministic basic data foundation for subsequent fine-grained resource scheduling.

[0024] Furthermore, in order to solve the problem of excessive redundant wake-up energy consumption caused by traditional full broadcast, in this embodiment of the invention, the management node executes a targeted and precise wake-up mechanism based on a scheduling mask bitmap. At the beginning of each scheduling cycle, the central scheduler of the management node polls the internal group queue management modules. Based on the downlink data backlog status of each logical group in the downlink data buffer and the uplink authorization requirements parsed from the uplink resource scheduling request, the activation requirements of the current superframe are calculated in real time. The management node dynamically generates a scheduling mask bitmap, which is represented at the physical layer as a binary vector composed of multiple consecutive bits. The index position of each bit is uniquely bound to the aforementioned globally unique group identifier through a preset modulo algorithm. Specifically, when the management node determines that a specific logical group has valid service interactions such as payload data that needs to be sent or uplink feedback that needs to be received within the current scheduling window, the management node sets the corresponding bit in the scheduling mask bitmap to the first preset state. In some implementations, the first preset state is defined as logic high level 1; Conversely, if the management node determines that the logical group has no business interaction needs in the current superframe, it will set the corresponding bit to the second preset state, i.e., logic low level 0. The management node encapsulates the generated scheduling mask bitmap in the header field of the downlink broadcast signaling. To ensure that this field can be parsed quickly with extremely low power consumption, the header field adopts the most robust modulation and coding scheme. For example, binary phase shift keying ensures that member nodes can achieve zero-error extraction in edge coverage scenarios.

[0025] Furthermore, on the member node side, the member node executes RF front-end adaptive control logic based on the scheduling mask bitmap. This logic, as an important component of the member node's physical layer processor, aims to minimize the RF link activity time during non-service periods. The synchronization timer inside the member node precisely controls the wake-up timing. When a member node enters the preset synchronization capture window, its hardware circuitry activates the narrowband receive link of the RF front end, which includes the necessary low-noise amplifier, mixer and low-bit analog-to-digital converter, thereby reducing instantaneous power consumption. In this narrowband activation state, member nodes only perform baseband demodulation on the header field of downlink broadcast signaling. The logic determination module inside the member node will immediately extract and determine the bit status corresponding to its group identifier in the scheduling mask bitmap. If the determination result is the second preset state, it means that there is no service data of this group in the current superframe. The medium access control layer of the member node will immediately trigger an RF shutdown command to the power management unit. This command will instantly shut down all active devices in the RF transceiver link and put the member node into a deep low-power standby mode, thereby directly skipping the reception, sampling and complex demodulation of the subsequent full broadcast payload of the superframe, eliminating the invalid power consumption of non-target groups in the long message reception process. If the determination result is the first preset state, the member node will maintain the active state of the RF link and immediately call the resource index mapping table stored in local memory to accurately locate the specific physical resource block occupied by this group in the current superframe, and prepare for subsequent data payload transmission and reception. Through this bit-level mask-driven directional wake-up mechanism, the dual on-demand allocation of member node power consumption in the spatial domain and time domain is realized.

[0026] Furthermore, in order to achieve dynamic resource elastic reconfiguration with extremely low signaling overhead and avoid the failure of static resource reservation due to drastic changes in the channel environment, an artificial intelligence-based resource prediction model was introduced into the management node. The resource prediction model is configured to make a forward-looking judgment on the traffic evolution trend of logical groups. The construction logic of the resource prediction model has a hierarchical structure, which includes an input layer, a feature extraction layer, a temporal correlation layer and an output layer. The input layer serves as the data entry point of the model and is responsible for receiving the historical traffic fingerprint sequence of each logical group. This sequence contains the throughput distribution within the past one hundred superframes. The current queue depth offset, node activity distribution data and air interface channel quality parameters are also input synchronously. The feature extraction layer consists of convolutional neural network units composed of multiple nonlinear activation functions. It extracts the spatial features of the traffic data by performing multi-scale convolution processing on the spatial tensor of the traffic data. The temporal correlation layer utilizes a recurrent processing structure, such as a gated recurrent unit or a long short-term memory network, to accumulate the extracted features over time, thereby capturing the temporal periodicity and random burst probability of the business flow at the microsecond scale. The output layer finally generates a predicted resource demand matrix, which indicates the probability of resource surplus or shortage for each group in the next ten consecutive superframes.

[0027] Furthermore, in the training process and internal connection logic of the resource prediction model, the management node ensures the generalization ability of the model by combining offline training and online fine-tuning. The management node pre-collects historical operation logs containing different interference intensities, different node sizes and different business topologies as training sample sets, with a total number of records not less than 100,000. During training, mean squared error is used as the target loss function to quantify the residual between the predicted resource proportion and the actual demand. Through the backpropagation algorithm, the model continuously corrects the weight parameters between internal neurons. When the prediction confidence of the model output online exceeds the preset 95% threshold, the management node will adopt the predicted resource demand matrix and pre-define a portion of elastic buffer resource blocks at the physical layer. These elastic buffer resource blocks serve as redundant reserves for the resource pool. When a group experiences a predicted burst of traffic, the management node can directly inject resources by quickly updating the resource index mapping table, without having to perform complex multi-round signaling handshakes over the air interface. This prediction-driven dynamic reconstruction enables smooth absorption of burst traffic and ensures that the system's response latency remains constant at the microsecond level.

[0028] Furthermore, in order to avoid signal collisions when multiple groups access concurrently, especially interference when multiple groups initiate requests in the same random access time slot, the management node executes access coordination control logic based on backoff weight factors. Under this logic framework, the management node does not adopt the traditional fixed window backoff mechanism, but calculates a specific backoff weight factor for each logical group based on the real-time perceived air interface channel quality (such as noise floor level and interference signal duty cycle) and the service urgency of each logical group using a weighted equalization algorithm. The management node periodically publishes an access control list in downlink broadcast signaling, which contains controlled weight mapping relationships in a specific mapping structure; When a random access request is generated by the service triggering module inside a member node, its internal media access control unit will automatically retrieve and extract the backoff weight factor of the group to which it belongs. The member node will then dynamically adjust the initial contention window size of the conflict backoff algorithm accordingly. Specifically, the size of the initial contention window is proportional to the backoff weight factor. By allocating a very small backoff weight factor to the high-priority group, its initial contention window is kept at a minimum, thereby increasing the access probability of the group's nodes at the physical layer and realizing fine-grained intervention in access determinism.

[0029] Furthermore, the Star Flash Near Field Communication Group Networking Method of the present invention also includes a closed-loop feedback and link compensation mechanism to cope with complex dynamic channel environments. After each service interaction cycle, that is, after each complete data payload reception or transmission, the member node will generate a link status report including received signal strength indication, signal-to-noise ratio and cyclic redundancy check result by the physical layer measurement unit. This report is fed back to the management node through the uplink control channel. The management node uses the collected link status reports of each group. The built-in link quality assessment algorithm calculates the current link's channel capacity margin and dynamically adjusts the modulation and coding strategy of the group in the next cycle. For example, if the 64th-order orthogonal amplitude modulation is reverted to the 16th order to enhance robustness, and the transmit power offset is adjusted, if the management node finds that the communication quality of a specific group is consistently lower than the preset reliability threshold through statistical judgment of multiple consecutive superframes, the group reassembly process is automatically triggered. The management node then uses downlink commands to direct the affected member nodes to migrate as a whole to the pre-detected spare clean frequency band. This dynamic migration based on a global perspective ensures the self-healing capability of the networking system when encountering narrowband interference.

[0030] Furthermore, the management node executes high-precision synchronization maintenance logic based on the Gory code sequence. At the beginning of each superframe, i.e., the synchronization start time slot, the management node transmits a synchronization fingerprint sequence containing high-intensity autocorrelation characteristics. This sequence consists of a pair of complementary Gory codes, and its autocorrelation function in the time domain exhibits extremely high peaks and extremely low sidelobes, which can effectively resist inter-symbol interference caused by multipath delay spread. On the member node side, the internally integrated hardware digital phase-locked loop circuit performs high-rate oversampling correlation peak detection on the synchronization fingerprint sequence. Based on the detected peak time, member nodes calculate the deviation between their local clock and the management node's system time, and compensate for the frequency deviation and phase shift of their local crystal oscillator in real time. When the management node performs the aforementioned dynamic adjustment of AI resource boundaries, it strictly locks the switching timing to ensure that all physical layer parameter switching is completed within the cyclic prefix protection interval of the orthogonal frequency division multiplexing symbol. Due to the existence of the cyclic prefix, transient phase changes caused by parameter switching will not propagate into the data sampling window. Through this physical layer phase continuity control, it can be ensured that the resource elastic adjustment process does not cause baseband signal parsing errors.

[0031] Furthermore, for IoT sensor nodes with extremely limited energy, the management node also executes targeted dynamic alignment logic for group sleep cycles. The management node combines the long-term evolution output of the artificial intelligence resource prediction model to determine the inactive duration of each logical group on a future minute or hour scale. Based on this, the management node generates targeted discontinuous reception cycle parameters. These parameters define the long sleep cycles that member nodes can skip when monitoring the scheduling mask bitmap. After receiving these parameters, member nodes enter a deep hardware sleep state, shutting down most logic circuits, including the digital baseband, until the predicted activation time corresponding to the next scheduling mask bitmap is reached. Only then will the member node restart the synchronization process. This long-cycle energy efficiency management strategy, combined with the short-cycle scheduling mask bitmap fast sleep mechanism, constructs a multi-level energy-saving architecture ranging from microseconds to minutes.

[0032] Example: A StarFlash network consisting of one management node and one thousand member nodes was deployed. The member nodes were divided into four logical groups: Group A is a high-frequency motion control group (latency requirement less than 100 microseconds), Group B is an ultra-high-definition video backhaul group (bandwidth requirement greater than 50Mbps), Group C is an environmental sensor data group (low-frequency triggering), and Group D is an emergency alarm group (extremely high priority). The management node operates in the 5.8GHz frequency band, and the superframe length is set to 10 milliseconds.

[0033] During the operation of the embodiment, the management node discovered through the artificial intelligence resource prediction model that group B has significant traffic peaks at every hour, while group C is in a state of extremely low activity at night. During the nighttime period, the management node reclaims most of the physical resource blocks of group C to the elastic buffer pool and sets the corresponding positions of its scheduling mask bitmap to 0 for a long time. When the video data volume of group B suddenly increases, the management node does not need to perform multiple rounds of RRC signaling interaction. It can directly allocate elastic buffer blocks in the superframe based on the prediction results, which instantly increases the air interface throughput of group B by three times. Member node C hardly turns on its radio frequency throughout the night, only performing a very short synchronization calibration within a nanosecond-level synchronization window.

[0034] Comparative example: It adopts the existing standard star-flash full-broadcast networking mode, which does not have a scheduling mask bitmap mechanism, and the resource allocation is completely statically reserved; Under the same network load, member node C in the comparison must be forced to receive and parse the complete downlink broadcast message in each superframe to determine whether it contains instructions targeting itself. When a sudden surge in traffic occurs in group B, the buffer of group B overflows rapidly because the resource reservation is static, resulting in a large number of payload retransmissions and a significant increase in latency.

[0035] Table 1: Comparison of Embodiments of the Invention with Comparative Examples Performance evaluation indicators Embodiment of the Invention (Mask Scheduling + Intelligent Prediction) Comparative Example (Full Broadcast + Static Reservation) Performance improvement ratio Average power consumption of non-target group nodes (mW) 0.45 12.80 96.48% (decreased) System average access latency (μs) 85 420 79.76% (decreased) Retransmission rate (%) under sudden traffic 0.12 5.45 97.80% (decreased) Collision probability (%) of multiple groups accessing concurrently 0.8 15.2 94.74% (decreased) Network-wide synchronization accuracy error (ns) 12 85 85.88% (Increase) Resource utilization rate (spectral efficiency Bits / Hz) 8.4 2.1 300.00% (Increase) As can be seen from the quantitative results in the comparison data table above, the star-flash near-field communication group networking method provided by this invention has generational advantages in reducing power consumption, reducing latency, and improving resource utilization. The reduction in power consumption of non-target group nodes proves the energy efficiency of the scheduling mask bitmap mechanism in realizing on-demand wake-up, while the significant reduction in retransmission rate strongly supports the role of the artificial intelligence resource prediction model in smooth traffic absorption.

[0036] In the deeper implementation details of the artificial intelligence resource prediction model, the embodiments of the present invention have made special optimizations to the convolution kernel size of the feature extraction layer. The feature extraction layer adopts a three-layer one-dimensional convolution structure with convolution kernel sizes of 3, 5, and 7 respectively. This multi-scale design can simultaneously capture the short-term fluctuations and medium-term trends of traffic data. Each convolution layer is connected to a modified linear unit (ReLU) activation function to introduce non-linear expressive power. The temporal correlation layer uses two layers of gated recurrent units, each containing 128 hidden neurons during training. The management node uses the Adam optimizer with a learning rate of 0.001 and introduces a dropout method to prevent the model from overfitting to the traffic characteristics of specific nodes. The predicted resource demand matrix output by training is a 10x10 tensor composed of probability values. The management node converts this predicted probability into specific physical layer resource block request instructions by setting a decision threshold of 0.85.

[0037] In the underlying details of the access cooperative control logic, the calculation formula of the backoff weight factor in this embodiment of the invention is designed to be nonlinear. The calculation of the weight factor takes into account both the static priority weight of the logical group and the dynamic channel competition pressure. Specifically, when the monitored air interface conflict rate increases by 10%, the backoff weight factor of the emergency group increases by only 2%, while the weight factor of the ordinary service group increases by 15%. This differentiated dynamic adjustment can ensure that even in the case of extremely congested channels, critical control commands (such as robot emergency stop commands) can still be successfully preempted within the first backoff window with a very high probability.

[0038] At the hardware connection level of the synchronization maintenance logic, the digital phase-locked loop circuit inside the member node adopts a high-order loop filter. The parameters of the filter are dynamically issued by the management node based on the current mobility prediction. When a member node is in a high-speed moving state, the management node instructs the member node to increase the loop bandwidth in order to quickly track the phase change caused by the Doppler frequency shift; When member nodes are in a static state, the loop bandwidth is reduced to obtain a higher phase noise suppression capability. This hardware and software coordinated synchronization logic, combined with the strong autocorrelation characteristics of the Gory code sequence, can ensure that the clock drift of all network nodes is always controlled within a preset 15 nanoseconds under extremely poor fading channel conditions.

[0039] The modulation and coding strategy table in the closed-loop feedback compensation mechanism is stored in the read-only memory of the management node, containing sixteen levels from QPSK to 1024QAM. After receiving the link status report, the management node uses a moving average window to calculate the average signal-to-noise ratio over the past second. If the downward slope of the average signal-to-noise ratio exceeds a preset threshold, the management node will immediately issue an MCS handover command to the corresponding member node and inform all members of the group through a specific cyclic shift code in the downlink control channel. This fast link adaptation and group reorganization mechanism can ensure the robustness of the system in a dynamic networking environment.

[0040] During the maintenance of logical mapping, when a member node needs to migrate from group C to group B due to business changes, the management node will set a transition period for the node in the resource index mapping table. During the transition period, the management node will simultaneously activate the corresponding bits of group B and group C in the scheduling mask bitmap. After the node demodulates the two masks at the same time, it will complete the smooth state transfer of the internal protocol stack. This design avoids business interruption during group changes and further enhances the flexibility of the network architecture.

[0041] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A group networking method for star-flash near-field communication, characterized in that, Includes the following steps: The management node divides the member nodes into several logical groups based on the service flow characteristics and latency sensitivity threshold of the access member nodes, assigns a unique group identifier to each logical group, and constructs a resource index mapping table that establishes the correspondence between the logical groups and physical layer resource blocks. The management node generates a scheduling mask bitmap based on the interaction requirements of each logical group. The scheduling mask bitmap contains multiple bits, and the index position of the bits corresponds to the group identifier. The management node encapsulates the scheduling mask bitmap in downlink broadcast signaling and sends it out. After receiving the downlink broadcast signaling, the member node extracts the bit state corresponding to the group identifier of the logical group to which it belongs from the scheduling mask bitmap; If the bit state is the first preset state, the member node accurately locates the physical resource block according to the resource index mapping table to perform service interaction; if the bit state is the second preset state, the member node shuts down the radio frequency transceiver link and enters standby mode.

2. The group networking method for star-flash near-field communication according to claim 1, characterized in that, The steps for the management node to perform pre-configuration and logical mapping of grouped resource spaces include: The management node extracts the service flow characteristics of each accessing member node through the service feature perception module. The service flow characteristics include data packet arrival interval distribution, average message length, service burst probability, and historical traffic envelope. The management node obtains the latency sensitivity threshold of the member node, which is defined as the maximum allowable latency from the time the physical layer service data unit enters the transmission buffer until it is correctly received. The management node, by combining the peak throughput requirements and average throughput requirements of the member nodes, divides the member nodes across the entire network into logical groups. The group identifier is configured as a 16-bit binary code.

3. The group networking method for star-flash near-field communication according to claim 1, characterized in that, The steps for establishing a deterministic correspondence between the logical group and the physical layer resource block using the resource index mapping table include: The management node defines the subcarrier index set, symbol index range, and power mask parameters for each logical group through the physical layer resource block; The management node divides the communication superframe into multiple independent scheduling windows that do not overlap in the time dimension in the time domain, and sets a physical isolation interval of a preset length between each independent scheduling window. The management node ensures the orthogonality of different logical groups on air interface resources by setting the physical isolation interval. The orthogonality is configured to suppress intermodulation interference between groups below the noise floor through time offset alignment.

4. The group networking method for star-flash near-field communication according to claim 3, characterized in that, The steps by which the member node executes the RF front-end adaptive control logic based on the scheduling mask bitmap include: The member node controls the wake-up time through an internal synchronization timer. After entering the preset synchronization capture window, it activates the narrowband receiving link in the radio frequency transceiver link. The narrowband receiving link includes a low-noise amplifier, a mixer, and an analog-to-digital converter. In the narrowband receive link active state, the member node performs baseband demodulation on the header field of the downlink broadcast signaling; The logic determination module inside the member node extracts and determines the bit state corresponding to the group identifier in the scheduling mask bitmap; If the determination result is the second preset state, the media access control layer of the member node triggers a radio frequency shutdown command to the power management unit to shut down the active devices in the radio frequency transceiver link and skip the reception, sampling and demodulation parsing of the subsequent full broadcast payload of the current communication superframe.

5. The group networking method for star-flash near-field communication according to claim 4, characterized in that, The steps for the management node to perform dynamic resource elastic reconstruction include: The management node integrates a resource prediction model, which is configured to determine the traffic evolution trend of the logical group. The resource prediction model includes an input layer, a feature extraction layer, a temporal correlation layer, and an output layer. The input layer receives the historical traffic fingerprint sequence, current queue depth offset, node activity distribution data, and air interface channel quality parameters of each logical group. The historical traffic fingerprint sequence is configured to contain a preset number of throughput distribution data within the communication superframe.

6. The group networking method for star-flash near-field communication according to claim 5, characterized in that, The internal processing logic of the resource prediction model includes: The feature extraction layer uses a convolutional neural network unit composed of multiple nonlinear activation functions to perform multi-scale convolution processing on the spatial tensor of the traffic data to extract the spatial features of the traffic data. The temporal correlation layer accumulates the extracted spatial features through a cyclic processing structure to capture the temporal periodicity and random burst probability of the service flow. The output layer generates a predicted resource demand matrix, which is used to indicate the probability of resource surplus or resource shortage for each of the logical groups within a future preset number of communication superframes.

7. A group networking method for star-flash near-field communication according to claim 6, characterized in that, The training and resource injection steps of the resource prediction model include: The management node collects historical operation logs, including interference intensity, node size, and business topology, as a training sample set. During training, mean squared error is used as the target loss function to quantify the residual between the predicted resource proportion and the actual demand, and the weight parameters between neurons inside the model are corrected through the backpropagation algorithm. When the prediction confidence level output online by the resource prediction model exceeds a preset threshold, the management node adopts the predicted resource demand matrix and pre-defines elastic buffer resource blocks at the physical layer. When a burst of traffic occurs within the burst probability in the logical group, the management node injects the elastic buffer resource block into the target group by updating the resource index mapping table.

8. A group networking method for star-flash near-field communication according to claim 1, characterized in that, The steps by which the management node executes the access coordination control logic include: The management node calculates a specific backoff weight factor for each logical group based on the real-time perceived air interface channel quality and the service urgency of each logical group using a weighted equalization algorithm. The management node periodically publishes an access control list in the downlink broadcast signaling, and the access control list contains the mapping relationship of the backoff weight factor; When the service triggering module of the member node generates a random access request, the media access control unit of the member node extracts the backoff weight factor of the group to which it belongs, and adjusts the initial contention window size of the conflict backoff algorithm accordingly. The size of the initial competition window is positively correlated with the backoff weight factor configuration.

9. A group networking method for star-flash near-field communication according to claim 1, characterized in that, The networking method also includes closed-loop feedback and link compensation steps: The physical layer measurement unit of the member node generates a link status report after each service interaction cycle. The link status report includes received signal strength indication, signal-to-noise ratio, and cyclic redundancy check result. The management node uses the collected link status reports from each of the logical groups to calculate the current link's channel capacity margin through a link quality assessment algorithm, and dynamically adjusts the modulation and coding strategy and transmit power offset of the logical group in the next cycle. If the management node determines that the communication quality of a specific logical group is lower than a preset reliability threshold, it triggers a group reorganization process and controls the affected member nodes to migrate to a backup clean frequency band via downlink commands.

10. A group networking method for star-flash near-field communication according to claim 6, characterized in that, The steps for the management node to execute the synchronization maintenance logic and energy efficiency management logic include: The management node transmits a synchronization fingerprint sequence at the beginning of each communication superframe. The synchronization fingerprint sequence consists of a pair of complementary Gole codes, and the autocorrelation function of the synchronization fingerprint sequence in the time domain is configured to have spike and sidelobe cancellation characteristics. The member node uses an internally integrated hardware digital phase-locked loop circuit to perform oversampled correlation peak detection on the synchronization fingerprint sequence, and compensates for the frequency deviation and phase shift of the local crystal oscillator based on the detected peak time. The switching of physical layer parameters configured in the management node is all completed within the cyclic prefix guard interval of the orthogonal frequency division multiplexing symbol; The management node determines the inactivity duration of each logical group based on the output of the artificial intelligence resource prediction model, generates discontinuous reception cycle parameters, and the member nodes enter a hardware sleep state based on the discontinuous reception cycle parameters until the predicted activation time corresponding to the next scheduling mask bitmap.

Images