Communication resource allocation method, device, system, equipment and medium

Abstract

The embodiment of the invention discloses a communication resource allocation method, device, system, equipment and medium, and the method comprises the steps: receiving equipment parameters of smart home equipment reported by a Bluetooth module, and training a local scheduling model according to the equipment parameters; uploading a first model parameter of the local scheduling model obtained by training to a central server, so that the central server performs weighted aggregation on the received first model parameter to generate an optimized global scheduling model; receiving a second model parameter of the global scheduling model sent by the central server, and updating the local scheduling model according to the second model parameter; and generating a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model. Each edge node generates a communication resource allocation strategy for the connected Bluetooth module based on a locally trained scheduling model, and only model parameters need to be exchanged between the edge nodes and the central server, so that the data processing pressure is greatly reduced, and a high-concurrency scene can be dealt with more easily.

Description

A method, apparatus, system, device, and medium for allocating communication resources. Technical Field

[0001] This invention relates to the field of communication resource technology, and in particular to a communication resource allocation method, a communication resource allocation device, a communication resource allocation system, an electronic device, and a computer-readable storage medium. Background Technology

[0002] In smart home systems, a large number of devices often need to be connected. Currently, one of the main solutions to address signal collisions caused by a large number of devices in densely populated environments is to set up a central server to manage all terminal access and allocate communication time slots and channels. However, all terminal access requests and status reports need to be processed through the central server. In high-concurrency scenarios, the server can easily become a performance bottleneck, causing processing delays and affecting the system's real-time response capabilities. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention are proposed to provide a communication resource allocation method, a communication resource allocation device, a communication resource allocation system, an electronic device, and a computer-readable storage medium that overcome or at least partially solve the above problems.

[0004] To address the aforementioned problems, a first aspect of this invention provides a communication resource allocation method applied to at least one edge node, wherein the edge node is communicatively connected to Bluetooth modules and a central server respectively installed on multiple smart home devices; the method includes: receiving device parameters of the smart home devices reported by the Bluetooth modules, and training a local scheduling model based on the device parameters; the device parameters include local environmental data and business requirement data; uploading first model parameters of the trained local scheduling model to the central server, so that the central server performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model; receiving second model parameters of the global scheduling model sent from the central server, and updating the local scheduling model based on the second model parameters; and generating a communication resource allocation strategy for the Bluetooth modules based on the updated local scheduling model; the communication resource allocation strategy includes communication time slots and communication channels of the Bluetooth modules.

[0005] Optionally, generating a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model includes: receiving current network status information reported by the Bluetooth module, and generating multiple candidate communication resource allocation strategies based on the current network status information through the local scheduling model; and determining the target communication resource allocation strategy for the Bluetooth module from the multiple candidate communication resource allocation strategies.

[0006] Optionally, determining the target communication resource allocation strategy for the Bluetooth module from the plurality of candidate communication resource allocation strategies includes: generating a comprehensive utility score corresponding to the plurality of candidate communication resource allocation strategies according to a preset weighted utility function; predicting each candidate communication resource allocation strategy using the local scheduling model to obtain a prediction result; the prediction result includes the benefit of each candidate communication resource allocation strategy in a future finite time slot and the potential resource allocation conflicts identified in the prediction process; and determining the target communication resource allocation strategy for the Bluetooth module based on the prediction result and the comprehensive utility score corresponding to each candidate communication resource allocation strategy.

[0007] Optionally, training the local scheduling model based on the device parameters includes: preprocessing the device parameters reported by the Bluetooth module; training the local scheduling model with the device parameters as input, with the optimization objectives of minimizing the signal collision rate and maximizing network throughput, and the output of the local scheduling model being the optimal communication resource allocation strategy for each Bluetooth module.

[0008] Optionally, the global scheduling model is generated by the central server aggregating the parameters of the local scheduling model using a federated averaging algorithm, and the aggregation weight is determined based on the number of valid Bluetooth modules connected to each edge node.

[0009] Optionally, the edge node includes a logical login channel for Bluetooth module registration; before receiving the device parameters of the smart home device reported by the Bluetooth module, the method further includes: receiving a login request message broadcast by the Bluetooth module on the logical login channel; after receiving the login request message, allocating a temporary time slot for the Bluetooth terminal to perform initial signaling interaction; and after successfully performing the initial signaling interaction with the Bluetooth module, performing a registration operation on the Bluetooth module.

[0010] Optionally, determining the target communication resource allocation strategy for the Bluetooth module based on the prediction results and the comprehensive utility score corresponding to each candidate communication resource allocation strategy includes: when the local scheduling model predicts a resource allocation conflict, sending a query request to the Bluetooth module to query the urgency of the current pending service of the Bluetooth module; or, parsing the service request messages reported and cached by each Bluetooth module to determine the urgency of the current pending service of each Bluetooth module; and allocating communication resources to Bluetooth modules with higher urgency based on the urgency of the current pending service of the Bluetooth module.

[0011] Optionally, the method further includes: for Bluetooth modules without allocated communication resources, allocating compensation communication resources to the Bluetooth modules according to a compensation communication resource strategy; the compensation communication resource strategy is to meet the service delay limit of the Bluetooth modules, and to preferentially select communication time slots and communication channels that are closest in time to the communication resources of the original conflicting Bluetooth modules and whose channel quality meets the transmission requirements.

[0012] Optionally, the method further includes: when the edge node fails, notifying the edge node adjacent to the failed edge node to take over the Bluetooth module originally belonging to the failed edge node through the central server.

[0013] According to a second aspect of the present invention, a communication resource allocation apparatus is provided, the apparatus comprising: a scheduling model training module, configured to receive device parameters of a smart home device reported by a Bluetooth module, and train a local scheduling model based on the device parameters; the device parameters include local environment data and service requirement data; a model parameter aggregation module, configured to upload first model parameters of the trained local scheduling model to a central server, so that the central server performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model; a scheduling model updating module, configured to receive second model parameters of the global scheduling model sent from the central server, and update the local scheduling model based on the second model parameters; and a resource allocation strategy generation module, configured to generate a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

[0014] According to a third aspect of the present invention, a communication resource allocation system is provided. The system includes at least one edge node, a Bluetooth module and a central server respectively communicatively connected to the edge node on multiple smart home devices. The Bluetooth module is configured to acquire device parameters of the smart home devices and report the device parameters to the edge node. The device parameters include local environment data and service requirement data. The edge node is configured to receive the device parameters of the smart home devices reported by the Bluetooth module and train a local scheduling model based on the device parameters. The edge node uploads first model parameters of the trained local scheduling model to the central server. The central server receives the first model parameters of the local scheduling model uploaded by the edge node and performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model. The central server sends second model parameters of the global scheduling model to the edge node. The edge node is further configured to receive the second model parameters of the global scheduling model sent by the central server and update the local scheduling model based on the second model parameters. Based on the updated local scheduling model, the edge node generates a communication resource allocation strategy for the Bluetooth module. The communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

[0015] According to a fourth aspect of the present invention, an electronic device is provided, comprising: a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the communication resource allocation method as described in any of the preceding embodiments.

[0016] According to a fifth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps of the communication resource allocation method as described in any of the preceding embodiments.

[0017] The technical solutions provided by the embodiments of the present invention can include the following beneficial effects: The embodiments of the present invention disclose a communication resource allocation method, apparatus, system, device, and medium. The method includes: applying to at least one edge node, the edge node being communicatively connected to Bluetooth modules and a central server respectively installed on multiple smart home devices; the method includes: receiving device parameters of the smart home devices reported by the Bluetooth modules, and training a local scheduling model based on the device parameters; the device parameters include local environmental data and business requirement data; uploading the first model parameters of the trained local scheduling model to the central server, so that the central server performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model; receiving the second model parameters of the global scheduling model sent from the central server, and updating the local scheduling model based on the second model parameters; generating a communication resource allocation strategy for the Bluetooth modules based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth modules. Each edge node, based on the locally trained scheduling model, can autonomously and in real time generate a communication resource allocation strategy for the connected Bluetooth modules, continuously ensuring the normal communication of smart home devices. The edge nodes and the central server only need to exchange model parameters, which greatly reduces the transmission load and the data processing pressure on the central server. By generating a communication resource allocation strategy, appropriate communication time slots and communication channels are allocated to each Bluetooth module, enabling the system to better cope with high-concurrency scenarios. Attached Figure Description

[0018] Figure 1 is a flowchart of a communication resource allocation method according to an embodiment of the present invention; Figure 2 is a flowchart of another communication resource allocation method according to an embodiment of the present invention; Figure 3 is a schematic diagram of the communication resource allocation process of a communication resource allocation method according to an embodiment of the present invention; Figure 4 is a structural block diagram of a communication resource allocation device according to an embodiment of the present invention; Figure 5 is a structural block diagram of a communication resource allocation system according to an embodiment of the present invention. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] In smart home systems, a large number of devices often need to be connected. However, all access requests and status reports from these terminals must be processed by a central server. In high-concurrency scenarios, the server can easily become a performance bottleneck, causing processing delays and affecting the system's real-time responsiveness. One of the core concepts of this invention is that each edge node, based on a locally trained scheduling model, can autonomously and in real-time generate communication resource allocation strategies for its connected Bluetooth modules, continuously ensuring normal communication of smart home devices. The edge nodes and the central server only need to exchange model parameters, greatly reducing the transmission load and the data processing pressure on the central server. By generating communication resource allocation strategies, appropriate communication time slots and channels are allocated to each Bluetooth module, enabling the system to better handle high-concurrency scenarios.

[0021] Referring to Figure 1, a flowchart of a communication resource allocation method according to an embodiment of the present invention is shown. The method is applied to at least one edge node, which is communicatively connected to Bluetooth modules on multiple smart home devices and a central server. The method specifically includes the following steps: Step 101, receiving device parameters of the smart home devices reported by the Bluetooth modules, and training a local scheduling model based on the device parameters; the device parameters include local environmental data and business requirement data. A smart home system contains a large number of smart home devices, typically managed by a central server. However, in high-concurrency scenarios, this can cause processing delays, affecting the system's real-time response capability. This embodiment of the present invention generates a communication resource allocation strategy based on a locally trained scheduling model at the edge node. The edge node and the central server only need to exchange model parameters, greatly reducing the transmission load and the data processing pressure on the central server, enabling the system to better handle high-concurrency scenarios.

[0022] The Bluetooth module refers to each smart home device. Each Bluetooth module uses a MAC address as a unique ID. It is recommended to support the Bluetooth Low Energy protocol and integrate environmental sensors (such as an RSSI detection module) to collect local signal strength and interference data. Edge nodes are deployed within the home smart hub / router in the smart home scenario. Edge node devices deploy edge computing capabilities, including a login channel (for terminal registration) and a data channel (for data transmission). The channels use frequency division multiplexing to avoid interference. A lightweight machine learning model is built-in to train local scheduling strategies based on data reported by the terminals, and the model parameters are uploaded to the central server.

[0023] The central server is typically deployed using cloud servers. To ensure the quality of the network connection between the cloud server and edge nodes, high-performance leased lines are primarily used to guarantee low latency and stability. For reliability, built-in disconnection caching mechanisms are employed to ensure service continuity during network jitter or when terminals are not actively accessing the network. As the coordination center, the central server uses a federated averaging algorithm to fuse model parameters uploaded from each base station to generate a global scheduling strategy. By analyzing global parameters, resource allocation conflicts (such as overlapping time slots) are identified, and reconciliation instructions are issued. The central server is responsible for formulating macro-level strategies or complex long-term forecasts, while edge nodes, within the framework of rules defined by the server, possess autonomy to handle local real-time scheduling or unexpected situations, thus achieving a balance between the high efficiency of centralized control and the low latency of edge autonomy.

[0024] Equipment parameters include local environment data and service requirement data. Local environment data includes signal quality indicators (such as RSSI received signal strength and channel interference intensity), while service requirement data includes data throughput and latency sensitivity.

[0025] In this embodiment of the invention, the edge node first continuously receives multi-dimensional real-time data proactively or on-demand reported by the Bluetooth modules of various smart home devices within its jurisdiction via a logical data channel. This data includes local environmental data reflecting the wireless channel conditions (such as received signal strength, channel interference noise, and bit error rate), as well as service requirement data characterizing the device's service characteristics (such as data type, throughput requirements, and latency sensitivity). After receiving the data, the edge node preprocesses the raw data, for example, by using a sliding window averaging method to smooth signal fluctuations and normalizing heterogeneous data, to construct a regularized dataset for model training.

[0026] Building upon this, the edge node initiates the local training process of its built-in lightweight machine learning model (such as a small neural network or gradient boosting tree). This model takes preprocessed device parameters and real-time channel state information as input, aiming to minimize the signal collision rate and maximize network throughput. Through iterative learning, it outputs optimal communication resource allocation suggestions for each Bluetooth module. Training is not mechanically triggered at set intervals, but rather employs a combination of event-driven and periodic-driven mechanisms: it can be executed at fixed intervals (e.g., every 5 minutes) or initiated immediately when enough new samples representing network changes are accumulated. The training dataset uses a sliding time window mechanism to ensure that the model always learns and evolves based on the latest and most relevant local network situation, thus laying an adaptive decision-making foundation for subsequent real-time and accurate resource scheduling.

[0027] Step 102: The first model parameters of the trained local scheduling model are uploaded to the central server, allowing the central server to perform weighted aggregation on the received first model parameters to generate an optimized global scheduling model. Edge nodes encrypt and upload their local model parameters to the central server, which then performs global aggregation using a weighted average method. Aggregation weights are assigned based on the number of devices connected to each node, with edge nodes having a larger number of devices receiving higher weights. The aggregation is updated every 30 minutes to balance real-time performance and communication overhead. The global aggregation model is distributed to each edge node to update local policies and ensure full network coordination. In addition to the regular aggregation every 30 minutes, the system continuously monitors key indicators, such as drastic changes in device connection status (e.g., a large number of devices going offline or switching base stations due to mobility issues). Once these indicators exceed preset thresholds, an emergency global aggregation is immediately initiated.

[0028] In this embodiment of the invention, after the edge node completes local training, the first model parameters (such as the weights and biases of the neural network) of the trained local scheduling model are encrypted and then uploaded to the central server through a secure connection.

[0029] After aggregating the initial model parameters from multiple edge nodes across the network, the central server initiates a federated learning global aggregation process. The server employs a weighted average algorithm, dynamically allocating aggregation weights based on factors such as the number of active Bluetooth terminals connected to each edge node, thereby integrating local scheduling experience from different regions and scenarios. This aggregation process periodically (e.g., every 30 minutes) generates a more comprehensive and robust global scheduling model, whose parameters become the second model parameters. Furthermore, the system incorporates an emergency trigger mechanism: when a drastic change in network status is detected (such as large-scale device movement or access), an emergency aggregation can be immediately initiated to ensure the global model can quickly respond to sudden network dynamics.

[0030] Step 103: Receive the second model parameters of the global scheduling model sent by the central server, and update the local scheduling model according to the second model parameters. In this embodiment of the invention, the edge node receives the second model parameters, optimized by federated aggregation, from the central server via a secure channel. These parameters contain the scheduling experience of multiple edge nodes and represent the optimal or better solutions obtained through training in a wider range of data distributions and scenarios.

[0031] After receiving the second model parameters, edge nodes integrate this global knowledge into their local scheduling models through model fusion or incremental updates. This update process enables each edge node to not only make rapid decisions based on local real-time data but also receive optimization guidance from a global perspective. For example, the model may learn more efficient resource allocation patterns during specific time periods or under interference modes, or adjust the weight parameters in its multi-objective optimization function to better balance conflict, throughput, and energy consumption. By executing this update periodically (such as during the delivery cycle) or through event-triggered events (such as emergency updates), all edge nodes can evolve synchronously, ensuring that the entire distributed system maintains low-latency local autonomy while collaboratively approaching globally optimal scheduling performance, effectively avoiding bias or overfitting of local models.

[0032] Step 104: Based on the updated local scheduling model, generate a communication resource allocation strategy for the Bluetooth module; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

[0033] In this embodiment of the invention, the edge node utilizes the updated local scheduling model, combined with real-time collected network-aware information (including terminal service profiles, channel quality maps, and network load status), to calculate and generate a customized communication resource allocation strategy for each connected Bluetooth module. The generation of this strategy is a dynamic multi-objective optimization process: the model uses finite-step random sampling simulations to prospectively evaluate the long-term benefits and conflict risks of different "time slot-channel" combinations in the near future, and uses a weighted utility function dynamically optimized by the cloud for comprehensive scoring, ultimately selecting the scheme with the highest comprehensive utility as the output.

[0034] Communication time slot: A specific time segment allocated to a particular Bluetooth module for sending or receiving data. It clarifies "when" a device can communicate, avoiding collisions caused by multiple devices transmitting signals simultaneously through time-division multiplexing (TDM). Time slot allocation must accurately consider service latency requirements, periodicity characteristics, and queue status. Communication channel: A specific carrier frequency range allocated to a device within a specified time slot within the available frequency bands of the Bluetooth protocol. By allocating different frequencies to different devices or different time periods of the same device (Frequency Division Multiplexing), continuous interference from fixed frequency bands can be avoided, and spectrum resources can be fully utilized.

[0035] By dynamically allocating time slots and channels simultaneously, communication resources are optimized collaboratively in both time and frequency dimensions, thereby maximizing spectral efficiency, minimizing signal conflicts, and meeting the differentiated quality of service requirements of various smart home devices in dense device environments.

[0036] Referring to Figure 2, a flowchart of another communication resource allocation method provided by an embodiment of the present invention is shown, which is applied to at least one edge node. The edge node is communicatively connected to Bluetooth modules and a central server respectively set on multiple smart home devices. The edge node includes a logical login channel for Bluetooth module registration. The method specifically includes the following steps: Step 201, receiving a login request message broadcast by the Bluetooth module on the logical login channel; after the Bluetooth module is powered on, it broadcasts a login request message through the login channel of the edge node, which includes terminal ID and initial location information.

[0037] In this embodiment of the invention, the edge node continuously monitors the logical login channel it manages, which is dedicated to device discovery and registration. When a new smart home device is powered on or reset, its Bluetooth module actively broadcasts a login request message containing its unique device identifier (such as a MAC address) and initial location on this public channel using a random backoff mechanism. This design avoids continuous message collisions that may occur when multiple devices attempt to connect simultaneously, ensuring the reliability of the registration process.

[0038] Upon successfully receiving the broadcast message, the edge node identifies a new device attempting to join the network. The edge node first parses the message content, verifies its format validity, and uses the information in the message to initially understand the device's identity and location. Subsequently, the edge node initiates the subsequent access process for the device. This typically includes allocating a temporary dedicated signaling interaction slot to transition to controlled point-to-point communication after the initial public broadcast, completing secure registration operations such as key negotiation and capability exchange, and finally formally adding the device to its local device management queue, assigning it a logical synchronization base station identifier, thus laying the foundation for potential cross-node roaming in the future.

[0039] Step 202: After receiving the login request message, a temporary time slot is allocated to the Bluetooth terminal for initial signaling interaction. In this embodiment of the invention, after receiving the login request message broadcast by the device, the edge node needs to perform a series of subsequent, non-interference-prone point-to-point secure signaling interactions with the device (e.g., authentication, key negotiation, capability reporting, etc.). To avoid these critical signaling messages colliding with access requests or data packets from other devices on the public login channel, the edge node will immediately dynamically allocate and assign a temporary, exclusive communication time slot from its managed time slot resource pool for the initial interaction process of the device.

[0040] Edge nodes notify devices of their allocated temporary time slot information via downlink control signaling (either on a common channel or within an extremely small time slot negotiated in a previous broadcast). Upon receiving the instruction, the device will strictly exchange subsequent registration signaling with the edge node within the specified time slot. This mechanism effectively isolates the initialization process of different new devices, ensures reliable transmission of critical signaling, and provides edge nodes with initial control over devices before formal registration is completed and long-term stable data channels and scheduling time slots are allocated to them.

[0041] Step 203: After successfully completing the initial signaling interaction with the Bluetooth module, register the Bluetooth module.

[0042] After receiving the request, the edge node allocates a temporary time slot for the Bluetooth module and registers the Bluetooth module ID to its local queue. Simultaneously, the edge node assigns a synchronization base station ID to the Bluetooth module, ensuring connectivity during base station switching while the module is moving. Once the Bluetooth module powers on, it randomly sends a login request broadcast packet on the public channel. After parsing the request, the server or edge node allocates a temporary, time-window-based dedicated time slot for the terminal for subsequent initial signaling interactions. Before allocating the final data channel and stable time slot, a random backoff mechanism can be set to avoid continuous collisions between multiple terminals on the login channel. If a terminal does not receive an allocation instruction within a specific time, it will randomly delay for a period and then resend the request, thus distributing concurrency pressure and ensuring the registration process proceeds in an orderly manner. The synchronization base station ID allows the terminal to switch to the new room's base station ID (node) even after moving from one room to another.

[0043] In this embodiment of the invention, after the edge node and the Bluetooth module successfully complete a series of secure point-to-point initial signaling interactions (including two-way identity authentication, session key negotiation, device capability and service characteristic reporting, etc.) within the allocated temporary time slot, the edge node obtains the module's complete and reliable network access credentials and context information.

[0044] The edge node persistently stores the verified device's unique identifier (such as MAC address) and its context information (including capabilities, initial location, initial service profile values, etc.) in the local device registry and marks its status as "online". Simultaneously, it assigns a logical synchronization base station identifier to the device, which will be used to support smooth roaming during subsequent movement. Finally, the edge node incorporates the device into its local scheduling model, preparing it for generating resource allocation strategies containing formal communication time slots and channels in subsequent periodic scheduling.

[0045] Step 204: Receive device parameters of the smart home devices reported by the Bluetooth module, and train a local scheduling model based on the device parameters. The device parameters include local environmental data and service requirement data. In this embodiment of the invention, the edge node first continuously receives multi-dimensional real-time data actively or on demand reported by the Bluetooth modules of each smart home device within its jurisdiction through a logical data channel. This data includes local environmental data reflecting the wireless channel conditions (such as received signal strength, channel interference noise, and bit error rate), and service requirement data characterizing the device's service characteristics (such as data type, throughput requirements, and latency sensitivity). After receiving the data, the edge node preprocesses the raw data, for example, by using a sliding window averaging method to smooth signal fluctuations and normalizing heterogeneous data, to construct a regular dataset for model training.

[0046] Building upon this, the edge node initiates a local training process for its built-in lightweight machine learning model (such as a small neural network or gradient boosting tree). This model takes preprocessed device parameters and real-time channel state information as input, and aims to minimize the signal collision rate and maximize network throughput. Through iterative learning, it outputs optimal communication resource allocation recommendations for each Bluetooth module.

[0047] In some embodiments, step 204 may include the following sub-steps: sub-step S11, preprocessing the device parameters reported by the Bluetooth module; sub-step S12, training the local scheduling model with the device parameters as input, with the optimization objectives of minimizing the signal collision rate and maximizing the network throughput, wherein the output of the local scheduling model is the optimal communication resource allocation strategy for each Bluetooth module.

[0048] The Bluetooth module continuously collects local environmental data, including signal quality metrics (such as RSSI received signal strength and channel interference intensity) and service requirement data (such as data throughput and latency sensitivity). Edge nodes preprocess the data, using a sliding window averaging method to smooth RSSI fluctuations and normalizing the data format to reduce noise impact. Local environmental data is typically reported proactively to reduce the burden on base station nodes, while service data is usually polled from the terminal when the edge node detects high load. The timing varies depending on the data type; for example, signal strength data might be between 100ms and several seconds. Too short a timeframe results in significant network overhead, while too long a time reduces tracking effectiveness. Service data is primarily reported when changes occur.

[0049] Each edge node maintains a local scheduling model, taking terminal data and channel status as input and outputting optimal time slot and channel allocation suggestions. The model aims to minimize the signal collision rate and maximize throughput, performing a local update every 5 minutes to ensure the policy adapts to network changes and continuously optimizes the strategy. Edge nodes intelligently allocate appropriate communication times and channels to devices within their management area by observing and learning the communication status of Bluetooth modules within their range (e.g., which time periods are prone to collisions, which channels have good quality). Updates are also triggered when edge nodes collect a sufficient number of new or high-value data samples, such as new network conditions. The dataset uses a sliding window approach.

[0050] In this embodiment of the invention, edge nodes perform real-time cleaning of the received raw device parameter stream. For time-series data such as signal strength, a sliding window weighted average method is used to smooth instantaneous fluctuations and impulse noise; heterogeneous data (such as strength values, bit error rate, and throughput requirements) are subjected to max-min normalization or Z-score standardization to bring them into a uniform numerical range, thereby accelerating model convergence and improving stability; simultaneously, feature combination (such as generating signal-to-interference-plus-noise ratio derived features) and outlier filtering are performed to construct a high-quality training sample set that can be directly digested by the model.

[0051] Edge nodes are primarily trained using their built-in lightweight machine learning models, such as a small fully connected neural network or a gradient boosting decision tree. Training uses minimizing the signal collision rate and maximizing network throughput as the core components of a dual-objective loss function, typically combined using weighted and / or Pareto optimization methods.

[0052] The training process involves inputting preprocessed device parameter samples (including environmental data, service data, and corresponding historical channel states) into the model. The model then propagates forward and outputs a set of resource allocation policy predictions. Subsequently, the error gradient between the predicted policy and the actual optimal policy (or the reward signal obtained through simulation) is calculated using the backpropagation algorithm. Finally, stochastic gradient descent or its variants (such as the Adam optimizer) are used to update the model's internal parameters (weights and biases), enabling the model's output policy to continuously approach the goal of simultaneously reducing conflicts and increasing throughput in the simulated environment. Training can be periodic batch learning or online learning based on new samples, ensuring that the model continuously adapts to network dynamics. By fully decentralizing model training and optimization to edge nodes, resource allocation policies can be generated in real time based on the latest local data, eliminating the need to upload massive amounts of raw data to a central server and wait for its computation, thus greatly improving scheduling real-time performance.

[0053] Step 205: Upload the first model parameters of the trained local scheduling model to the central server so that the central server can perform weighted aggregation on the received first model parameters to generate an optimized global scheduling model. In this embodiment of the invention, after the edge node completes local training, the first model parameters of the trained local scheduling model (such as the weights and biases of the neural network) are encrypted and uploaded to the central server through a secure connection.

[0054] After aggregating the initial model parameters from multiple edge nodes across the network, the central server initiates a federated learning global aggregation process. The server employs a weighted average algorithm, dynamically allocating aggregation weights based on factors such as the number of active Bluetooth terminals connected to each edge node, thereby integrating local scheduling experience from different regions and scenarios. This aggregation process periodically (e.g., every 30 minutes) generates a more comprehensive and robust global scheduling model, whose parameters become the second model parameters. Furthermore, the system incorporates an emergency trigger mechanism: when a drastic change in network status is detected (such as large-scale device movement or access), an emergency aggregation can be immediately initiated to ensure the global model can quickly respond to sudden network dynamics.

[0055] In some embodiments, the global scheduling model is generated by aggregating the parameters of the local scheduling model using a federated averaging algorithm by the central server, and the aggregation weight is determined based on the number of valid Bluetooth modules connected to each edge node.

[0056] In this embodiment of the invention, after receiving encrypted uploads of local scheduling model parameters (i.e., first model parameters) from multiple edge nodes, the central server employs a federated averaging algorithm as the aggregation engine. The federated averaging algorithm performs a weighted average: the aggregation weight is dynamically calculated and determined based on the number of valid Bluetooth modules connected to each edge node. Edge nodes with more connected devices and more complex management scenarios have a higher proportion of experiential knowledge carried by their local models in the global model, thus making the generated global model more representative of the mainstream and critical operating conditions in the network.

[0057] Based on these weighted and aggregated parameters, the central server generates a new generation of global scheduling model (whose parameters are the parameters of the second model). This model integrates scheduling intelligence from different regions and scenarios across the entire network, and its performance and robustness surpass any single local model. By distributing these global model parameters to each edge node, the optimization strategy, processed by data from the entire network, is seamlessly and synchronously empowered to each local node. This allows each edge node to have a global perspective in subsequent decisions, effectively avoiding local overfitting or decision bias, and driving the entire distributed system to converge towards the global optimal solution collaboratively and efficiently.

[0058] Step 206: Receive the second model parameters of the global scheduling model sent by the central server, and update the local scheduling model according to the second model parameters. In this embodiment of the invention, the edge node receives the second model parameters, optimized by federated aggregation, from the central server via a secure channel. These parameters contain the scheduling experience of multiple edge nodes, representing the optimal or better solutions trained under a wider data distribution and scenario. After receiving the second model parameters, the edge node integrates this global knowledge into its own local scheduling model through model fusion or incremental updates. This update process enables each edge node not only to make rapid decisions based on local real-time data, but also to obtain optimization guidance from a global perspective.

[0059] Step 207: Based on the updated local scheduling model, generate a communication resource allocation strategy for the Bluetooth module; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

[0060] In this embodiment of the invention, the edge node utilizes the updated local scheduling model, combined with real-time collected network-aware information (including terminal service profiles, channel quality maps, and network load status), to calculate and generate a customized communication resource allocation strategy for each connected Bluetooth module. The generation of this strategy is a dynamic multi-objective optimization process: the model uses finite-step random sampling simulations to prospectively evaluate the long-term benefits and conflict risks of different "time slot-channel" combinations in the near future, and uses a weighted utility function dynamically optimized by the cloud for comprehensive scoring, ultimately selecting the scheme with the highest comprehensive utility as the output.

[0061] By dynamically allocating time slots and channels simultaneously, communication resources are optimized collaboratively in both time and frequency dimensions, thereby maximizing spectral efficiency, minimizing signal conflicts, and meeting the differentiated quality of service requirements of various smart home devices in dense device environments.

[0062] In some embodiments, step 207 may include the following sub-steps: sub-step S21, receiving the current network status information reported by the Bluetooth module, and generating multiple candidate communication resource allocation strategies through the local scheduling model based on the current network status information; sub-step S22, determining the target communication resource allocation strategy for the Bluetooth module from the multiple candidate communication resource allocation strategies.

[0063] Edge nodes allocate resources to terminals based on a scheduling model. The local scheduling model takes into account multi-dimensional real-time environmental characteristics reported by terminal devices, including terminal service profiles: learning the behavioral patterns of each type of terminal based on historical data (e.g., sensor terminals report intermittently with low power consumption, while audio terminals require high-bandwidth continuous streaming), quantifying their priority, latency sensitivity, and bandwidth requirements. Real-time channel quality mapping: dynamically constructing interference heatmaps for each frequency band using RSSI (Received Signal Strength Indicator) and bit error rate reported by terminals, and predicting future time slot quality trends (e.g., using time-series prediction algorithms). Network load status: real-time monitoring of the number of active terminals and data queue length on each channel, calculating instantaneous load rates, and avoiding allocation to congested channels.

[0064] The local scheduling model generates a set of candidate "time slot-channel" combinations, quickly simulating the long-term benefits of different allocation strategies (such as the probability of conflict in the next 5 time slots), and retaining only the few candidate schemes with the highest benefits. Within a limited extrapolation depth (e.g., the next 5 time slots), multiple random sampling simulations are performed rapidly. Each simulation is like a "fast-forward preview," exploring the next few steps along a possible allocation path and accumulating the "benefits" gained along that path. In this embodiment, the edge node first receives and integrates real-time network status information reported by the Bluetooth module, including instantaneous channel quality, queue lengths of each device, and service urgency flags. The local scheduling model uses this as input, utilizing its internally learned scheduling rules, to generate a set of different candidate communication resource allocation strategies through rapid extrapolation or sampling based on policy gradients. Each candidate strategy corresponds to a specific "time slot-channel" allocation scheme, aiming to balance objectives such as conflict, throughput, and energy consumption from different perspectives, providing a rich selection space for subsequent optimal selection.

[0065] Instead of randomly selecting candidate strategies, edge nodes perform rapid, forward-looking simulations to evaluate each strategy, predicting its overall performance (such as collision probability and spectral efficiency) over several future time slots. This prediction, combined with dynamic weighting parameters from the cloud, calculates a quantified overall utility score. Finally, the system selects the candidate strategy with the highest utility score, designating it as the target communication resource allocation strategy for the current scheduling cycle, and encapsulates it into an instruction for issuance. This process ensures that each scheduling operation is the result of intelligent comparison and optimization, rather than a simple, single prediction.

[0066] In some embodiments, step S22 may include the following sub-steps: sub-step S221, generating a comprehensive utility score corresponding to the plurality of candidate communication resource allocation strategies according to a preset weighted utility function; sub-step S222, predicting each candidate communication resource allocation strategy using the local scheduling model to obtain a prediction result; the prediction result includes the benefit of each candidate communication resource allocation strategy in a limited time slot in the future and the potential resource allocation conflicts identified in the prediction process; sub-step S223, determining the target communication resource allocation strategy of the Bluetooth module based on the prediction result and the comprehensive utility score corresponding to each candidate communication resource allocation strategy.

[0067] For each candidate solution, the local scheduling model calculates a comprehensive utility score, determined by a multi-objective weighted function: Utility Score = A * (1 - Predicted Conflict Probability) + B * Spectrum Utilization + C * (1 - Estimated Terminal Energy Consumption). The weights A, B, and C are not fixed values ​​but are dynamically optimized by the cloud based on global historical performance and periodically updated to edge nodes to adapt to different scenarios (e.g., meeting mode prioritizes low network conflict, while energy-saving mode prioritizes low device energy consumption). The cloud continuously monitors the historical trends of each objective (conflict probability, spectrum utilization, terminal energy consumption, etc.). For objectives with slow convergence and high optimization difficulty, the system dynamically assigns them higher weights to allocate more training resources, thereby guiding the model towards a more balanced and optimal Pareto front.

[0068] In this embodiment of the invention, each candidate strategy is statically scored according to a preset weighted utility function. This function maps the strategy to a comprehensive utility score, the core of which is to quantify the real-time trade-offs of the strategy on multiple key dimensions such as conflict avoidance, spectral efficiency, and energy consumption control. The weight parameters in the function are not fixed, but are dynamically distributed by the cloud according to the global optimization objective, so that the evaluation criteria can adapt to different scenarios (such as meeting mode focusing on low conflict, and night mode focusing on low energy consumption).

[0069] The local scheduling model extrapolates and predicts the network state evolution for each candidate strategy within a finite number of future time slots (e.g., the next 5 time slots). This process typically employs finite-step random sampling simulations, repeatedly evaluating the long-term benefits of the strategy (e.g., cumulative throughput, average latency) through random evaluations, and proactively identifying potential resource allocation conflicts during the simulation (e.g., predicting multiple devices competing for the same channel in the same time slot). This upgrades the decision-making basis from single-step optimization to forward-looking multi-step planning.

[0070] Real-time quantitative scoring and forward-looking simulation results not only select strategies with high overall utility scores, but also prioritize the exclusion or downweighting of strategies that expose high-risk conflicts in the simulation. Through this dual filtering mechanism of "real-time assessment and risk warning," it is possible to select target communication resource allocation strategies that not only perform well in the present but also maintain stability and low conflict in the future, thereby significantly improving the intelligence level of scheduling decisions and the long-term operational efficiency of the network.

[0071] In some embodiments, step S223 may include the following sub-steps: Sub-step S2231, when the local scheduling model predicts that there is a resource allocation conflict, a query request is sent to the Bluetooth module to query the urgency of the current pending service of the Bluetooth module; or, the service request messages reported and cached by each Bluetooth module are parsed to determine the urgency of the current pending service of each Bluetooth module; Sub-step S2232, according to the urgency of the current pending service of the Bluetooth module, communication resources are allocated preferentially to the Bluetooth module with a higher urgency.

[0072] When it is predicted that multiple Bluetooth modules may transmit in the same time slot, a dynamic reconfiguration strategy is activated. Based on the urgency requests of the services reported by the terminals, the local scheduling model temporarily allocates resources to devices with high urgency and immediately allocates compensation time slots to terminals whose requests are not yet fulfilled. The service requests will include a type flag, and the urgency level will be determined based on the service type.

[0073] In this embodiment of the invention, when the local scheduling model identifies a potential resource allocation conflict in a forward-looking simulation (i.e., multiple Bluetooth modules are predicted to compete for the same channel in the same time slot), a conflict resolution process is immediately initiated. Edge nodes need to accurately determine the urgency of the services carried by the relevant modules. This is achieved in two ways: First, proactive querying, which sends a lightweight query request to the module involved in the conflict, instructing it to immediately report the urgency flag of the currently pending transmission service (usually set by the application layer based on the service type, such as "extremely high" for security alarms and "low" for file downloads). Second, passive parsing, which directly parses and extracts the service type field from the service request messages recently reported by the relevant modules that are cached locally on the node, and determines its urgency according to a preset mapping rule (e.g., audio streams are greater than sensor data). The latter has lower latency, but the former provides more real-time and accurate information.

[0074] After obtaining a clear priority ranking, the decision engine performs preemptive or negotiated resource reallocation. Scarce communication resources (time slots / channels) are preferentially allocated to the Bluetooth modules with the highest urgency to ensure that critical services (such as real-time control commands and emergency alarms) are not delayed. For lower-urgency modules that fail to obtain resources due to this reallocation, a compensation mechanism is immediately activated to arrange new transmission opportunities for them in the next available time slot or through other available channels. This ensures the quality of service for high-priority services while maintaining overall system fairness and throughput as much as possible.

[0075] In some embodiments, the method further includes: for Bluetooth modules without allocated communication resources, allocating compensation communication resources to the Bluetooth modules according to a compensation communication resource strategy; the compensation communication resource strategy is to meet the service delay limit of the Bluetooth modules, and to preferentially select communication time slots and communication channels that are closest in time to the communication resources of the original conflicting Bluetooth modules, and whose channel quality meets the transmission requirements.

[0076] In this embodiment of the invention, after the edge node completes priority scheduling based on urgency, it immediately initiates a compensation process for the relevant modules that have not obtained resources. According to the compensation communication resource strategy, the service latency limit of the module's service is first determined, serving as a hard time constraint for compensation scheduling. Subsequently, under the premise of meeting this latency limit, the edge node dynamically searches for available idle resources in the candidate resource combinations. Following the dual principles of closest proximity in time and best quality, it prioritizes finding the available time slot closest to the original conflicting time slot on the time axis to minimize service interruption duration.

[0077] Simultaneously, within this time slot, a communication channel whose channel quality (based on real-time interference maps) meets the module's transmission requirements is evaluated and selected. Finally, the determined compensation time slot and channel are encapsulated into a new scheduling instruction and sent to the module. This mechanism not only effectively avoids indefinite waiting for low-priority services but also maximizes overall network throughput efficiency and user experience by intelligently reallocating resources, resolving conflicts.

[0078] In some embodiments, the method further includes: when the edge node fails, notifying the edge node adjacent to the failed edge node to take over the Bluetooth module originally belonging to the failed edge node through the central server.

[0079] If an edge node goes down, the cloud will be notified that the edge node is offline. Neighboring nodes can quickly reconnect by subscribing to cluster status messages in the cloud to obtain context information such as the list of terminals to be taken over, terminal session keys, and task status. If the central server temporarily loses connection with the edge node, the edge node continues to operate independently using its local model.

[0080] If a terminal fails to report or experiences an anomaly for an extended period, reaching a certain threshold (e.g., more than 3 minutes without reporting or reporting content that does not conform to the rules), the edge node removes the terminal from the scheduling queue. The connection is then re-initialized after the device re-checks the problem and resets.

[0081] When the number of terminals under an edge node surges, the node's limited computing power, memory, and I / O bandwidth must cope with the sudden increase in computing tasks and data traffic, leading to request queuing, resource contention, and consequently, slower response times. To address this issue, solutions include lightweighting local AI models, implementing dynamic resource scheduling strategies on edge nodes, and building collaborative mechanisms between edge nodes.

[0082] In this embodiment of the invention, when an edge node goes offline due to hardware failure, software crash, or power outage, the central server, acting as the coordinator, will be the first to detect the node's disconnection through a heartbeat detection mechanism. The server activates a preset failover process: First, based on pre-set or dynamically learned network topology information, it quickly identifies one or more adjacent and normally functioning edge nodes as takeover nodes. Subsequently, the central server notifies these takeover nodes through a secure channel and sends them the device context information of the original faulty node (including a list of registered Bluetooth module IDs, session keys, service profiles, and the status of incomplete scheduling tasks). After receiving the instructions and data, the takeover node immediately broadcasts synchronization signaling within its wireless coverage area, guiding the Bluetooth modules of the original faulty node to perform rapid re-association. After detecting the disappearance of the original master node's signal, the module switches to the new takeover node according to the received synchronization signaling and quickly rebuilds the secure connection and service session using the sent context information, thereby achieving seamless service hot migration for users and ensuring the continuity and stability of smart home network services.

[0083] Referring to Figure 3, a schematic diagram of the communication resource allocation process of a communication resource allocation method provided in an embodiment of the present invention is shown. Figure 3 illustrates the core process of the communication resource allocation system of the embodiment of the present invention from device network access to resource allocation: After completing the security signaling interaction with the Bluetooth module, it is formally registered and included in the local management queue. Device environment and service parameters reported from registered modules are continuously received, and a dedicated scheduling model is trained locally on the edge nodes based on this data. The central server aggregates the local model parameters of each node, generates a better global scheduling model through federated learning, and distributes it. Based on the updated local model, the edge nodes generate communication resource allocation strategies that integrate global intelligence and local situational awareness for each Bluetooth module in real time, thereby completing the entire process from device access, data-driven learning, global knowledge fusion to intelligent scheduling decision-making.

[0084] This invention discloses a communication resource allocation method, apparatus, system, device, and medium. The method includes: applying to at least one edge node, the edge node being communicatively connected to Bluetooth modules and a central server respectively installed on multiple smart home devices; the method includes: receiving device parameters of the smart home devices reported by the Bluetooth modules, and training a local scheduling model based on the device parameters; the device parameters include local environmental data and service requirement data; uploading the first model parameters of the trained local scheduling model to the central server, so that the central server performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model; receiving the second model parameters of the global scheduling model sent from the central server, and updating the local scheduling model based on the second model parameters; and generating a communication resource allocation strategy for the Bluetooth modules based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth modules. Each edge node, based on the locally trained scheduling model, can autonomously and in real time generate a communication resource allocation strategy for the connected Bluetooth modules, continuously ensuring the normal communication of smart home devices. The edge nodes and the central server only need to exchange model parameters, which greatly reduces the transmission load and the data processing pressure on the central server. By generating a communication resource allocation strategy, appropriate communication time slots and communication channels are allocated to each Bluetooth module, enabling the system to better cope with high-concurrency scenarios.

[0085] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0086] Referring to Figure 4, a structural block diagram of a communication resource allocation device provided in an embodiment of the present invention is shown. The device specifically includes the following modules: a scheduling model training module 301, used to receive device parameters of smart home devices reported by a Bluetooth module, and train a local scheduling model based on the device parameters; the device parameters include local environment data and service requirement data; a model parameter aggregation module 302, used to upload the first model parameters of the trained local scheduling model to a central server, so that the central server performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model; a scheduling model updating module 303, used to receive the second model parameters of the global scheduling model sent from the central server, and update the local scheduling model based on the second model parameters; and a resource allocation strategy generation module 304, used to generate a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

[0087] In some embodiments, the resource allocation strategy generation module 304 includes: a network status information receiving submodule, configured to receive current network status information reported by the Bluetooth module, and generate multiple candidate communication resource allocation strategies based on the current network status information through the local scheduling model; and a resource allocation strategy filtering submodule, configured to determine the target communication resource allocation strategy for the Bluetooth module from the multiple candidate communication resource allocation strategies.

[0088] In some embodiments, the resource allocation strategy screening submodule includes: a comprehensive utility score generation unit, configured to generate comprehensive utility scores corresponding to the plurality of candidate communication resource allocation strategies according to a preset weighted utility function; a resource allocation strategy prediction unit, configured to predict each candidate communication resource allocation strategy using the local scheduling model to obtain a prediction result; the prediction result includes the benefits of each candidate communication resource allocation strategy in a future finite time slot and potential resource allocation conflicts identified in the prediction process; and a target allocation strategy determination unit, configured to determine the target communication resource allocation strategy of the Bluetooth module based on the prediction result and the comprehensive utility score corresponding to each candidate communication resource allocation strategy.

[0089] In some embodiments, the scheduling model training module 301 includes: a parameter preprocessing submodule, used to preprocess the device parameters reported by the Bluetooth module; and a model training submodule, used to train the local scheduling model with the device parameters as input, with the optimization objectives of minimizing the signal collision rate and maximizing the network throughput, wherein the output of the local scheduling model is the optimal communication resource allocation strategy for each Bluetooth module.

[0090] In some embodiments, the global scheduling model is generated by aggregating the parameters of the local scheduling model using a federated averaging algorithm by the central server, and the aggregation weight is determined based on the number of valid Bluetooth modules connected to each edge node.

[0091] In some embodiments, the edge node includes a logical login channel for Bluetooth module registration; it also includes: a request message receiving module for receiving a login request message broadcast by the Bluetooth module on the logical login channel; a temporary time slot allocation module for allocating a temporary time slot for the Bluetooth terminal after receiving the login request message, to perform initial signaling interaction; and a registration operation module for performing a registration operation on the Bluetooth module after successfully performing the initial signaling interaction with the Bluetooth module.

[0092] In some embodiments, the target allocation strategy determination unit includes: an urgency determination subunit, configured to send a query request to the Bluetooth module to query the urgency of the current pending service of the Bluetooth module when the local scheduling model predicts that there is a resource allocation conflict; or, to parse the service request messages reported and cached by each Bluetooth module to determine the urgency of the current pending service of each Bluetooth module; and a communication resource allocation subunit, configured to allocate communication resources to Bluetooth modules with higher urgency based on the urgency of the current pending service of the Bluetooth module.

[0093] In some embodiments, the apparatus further includes: a compensation communication resource allocation module, configured to allocate compensation communication resources to a Bluetooth module for which no communication resources have been allocated, according to a compensation communication resource strategy; the compensation communication resource strategy is to meet the service delay limit of the Bluetooth module, and to preferentially select communication time slots and communication channels that are closest in time to the communication resources of the original conflicting Bluetooth module and whose channel quality meets the transmission requirements.

[0094] In some embodiments, the apparatus further includes an edge node fault handling module, configured to, when the edge node fails, notify the adjacent edge nodes of the failed edge node to take over the Bluetooth module originally belonging to the failed edge node through the central server.

[0095] As the device embodiment is basically similar to the system embodiment, it is described in a relatively simple way. For relevant details, please refer to the description of the method embodiment.

[0096] Referring to Figure 5, a structural block diagram of a communication resource allocation system provided in an embodiment of the present invention is shown. The system 40 includes at least one edge node, and Bluetooth modules and a central server respectively connected to the edge node on multiple smart home devices. The Bluetooth module 401 is used to acquire device parameters of the smart home devices and report the device parameters to the edge node. The device parameters include local environment data and service requirement data. The edge node 402 is used to receive the device parameters of the smart home devices reported by the Bluetooth module, and train a local scheduling model based on the device parameters. The first model parameters of the trained local scheduling model are uploaded to the central server. The central server 403 is configured to receive first model parameters of the local scheduling model uploaded by the edge nodes, and perform weighted aggregation on the received first model parameters to generate an optimized global scheduling model; and send second model parameters of the global scheduling model to the edge nodes; the edge nodes 402 are further configured to receive the second model parameters of the global scheduling model sent by the central server, and update the local scheduling model according to the second model parameters; and generate a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

[0097] In some embodiments, the edge node 402 is configured to receive current network status information reported by the Bluetooth module, and generate multiple candidate communication resource allocation strategies based on the current network status information through the local scheduling model; and determine the target communication resource allocation strategy for the Bluetooth module from the multiple candidate communication resource allocation strategies.

[0098] In some embodiments, the edge node 402 is configured to generate a comprehensive utility score corresponding to the plurality of candidate communication resource allocation strategies according to a preset weighted utility function; predict each candidate communication resource allocation strategy through the local scheduling model to obtain a prediction result; the prediction result includes the benefit of each candidate communication resource allocation strategy in a limited time slot in the future and the potential resource allocation conflicts identified in the prediction process; and determine the target communication resource allocation strategy of the Bluetooth module based on the prediction result and the comprehensive utility score corresponding to each candidate communication resource allocation strategy.

[0099] In some embodiments, the edge node 402 is used to preprocess the device parameters reported by the Bluetooth module; with the optimization objectives of minimizing the signal collision rate and maximizing the network throughput, the device parameters are used as input to train the local scheduling model, and the output of the local scheduling model is the optimal communication resource allocation strategy for each Bluetooth module.

[0100] In some embodiments, the central server 403 is used to aggregate the parameters of the local scheduling model using a federated averaging algorithm, and generate the global scheduling model based on the aggregated parameters; the aggregation weight is determined according to the number of valid Bluetooth modules connected to each edge node.

[0101] In some embodiments, the edge node includes a logical login channel for Bluetooth module registration; the Bluetooth module 401 is configured to broadcast a login request message to the edge node through the logical login channel; the edge node 402 is configured to receive the login request message broadcast by the Bluetooth module on the logical login channel; after receiving the login request message, it allocates a temporary time slot for the Bluetooth terminal to perform initial signaling interaction; after successfully performing the initial signaling interaction with the Bluetooth module, it performs a registration operation on the Bluetooth module.

[0102] In some embodiments, the edge node 402 is configured to send a query request to the Bluetooth module when the local scheduling model predicts a resource allocation conflict, in order to query the urgency of the current pending service of the Bluetooth module; or, parse the service request messages reported and cached by each Bluetooth module to determine the urgency of the current pending service of each Bluetooth module; and allocate communication resources to the Bluetooth module with higher urgency based on the urgency of the current pending service of the Bluetooth module.

[0103] In some embodiments, the edge node 402 is configured to allocate compensation communication resources to Bluetooth modules that have not been allocated communication resources, according to a compensation communication resource strategy. The compensation communication resource strategy is to meet the service latency limit of the Bluetooth module and to preferentially select communication time slots and communication channels that are closest in time to the communication resources of the original conflicting Bluetooth module and whose channel quality meets the transmission requirements.

[0104] In some embodiments, the edge node 402 is configured to, when the edge node fails, notify the adjacent edge nodes to take over the Bluetooth module originally belonging to the failed edge node through the central server.

[0105] As the system implementation is basically similar to the system implementation, it is described in a relatively simple way. For relevant details, please refer to the description of the method implementation.

[0106] This invention also provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the various processes of the above-described communication resource allocation method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here.

[0107] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described communication resource allocation method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0108] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0109] Furthermore, it should be noted that the scope of the methods and apparatus in the embodiments of the present invention is not limited to performing functions in the order shown or discussed. It may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0111] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of the present invention.

Claims

1. A method for allocating communication resources, characterized in that, The method is applied to at least one edge node, which is communicatively connected to Bluetooth modules and a central server on multiple smart home devices. The method includes: receiving device parameters of the smart home devices reported by the Bluetooth modules, and training a local scheduling model based on the device parameters; the device parameters include local environmental data and business requirement data; uploading first model parameters of the trained local scheduling model to the central server, so that the central server performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model; receiving second model parameters of the global scheduling model sent from the central server, and updating the local scheduling model based on the second model parameters; and generating a communication resource allocation strategy for the Bluetooth modules based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth modules.

2. The communication resource allocation method according to claim 1, characterized in that, The step of generating a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model includes: receiving current network status information reported by the Bluetooth module, and generating multiple candidate communication resource allocation strategies based on the current network status information through the local scheduling model; and determining the target communication resource allocation strategy for the Bluetooth module from the multiple candidate communication resource allocation strategies.

3. The communication resource allocation method according to claim 2, characterized in that, The step of determining the target communication resource allocation strategy for the Bluetooth module from the plurality of candidate communication resource allocation strategies includes: generating a comprehensive utility score corresponding to the plurality of candidate communication resource allocation strategies according to a preset weighted utility function; predicting each candidate communication resource allocation strategy using the local scheduling model to obtain a prediction result; the prediction result includes the benefit of each candidate communication resource allocation strategy in a future finite time slot and the potential resource allocation conflicts identified in the prediction process; and determining the target communication resource allocation strategy for the Bluetooth module based on the prediction result and the comprehensive utility score corresponding to each candidate communication resource allocation strategy.

4. The communication resource allocation method according to claim 1, characterized in that, The step of training the local scheduling model based on the device parameters includes: preprocessing the device parameters reported by the Bluetooth module; training the local scheduling model with the device parameters as input, with the optimization objectives of minimizing the signal collision rate and maximizing network throughput, and the output of the local scheduling model being the optimal communication resource allocation strategy for each Bluetooth module.

5. The communication resource allocation method according to claim 1, characterized in that, The global scheduling model is generated by aggregating the parameters of the local scheduling model using a federated averaging algorithm on the central server; the aggregation weight is determined based on the number of valid Bluetooth modules connected to each edge node.

6. The communication resource allocation method according to claim 1, characterized in that, The edge node includes a logical login channel for Bluetooth module registration; Before receiving the device parameters of the smart home device reported by the Bluetooth module, the method further includes: receiving a login request message broadcast by the Bluetooth module on the logical login channel; after receiving the login request message, allocating a temporary time slot for the Bluetooth terminal to perform initial signaling interaction; and after successfully performing the initial signaling interaction with the Bluetooth module, performing a registration operation on the Bluetooth module.

7. The communication resource allocation method according to claim 3, characterized in that, The step of determining the target communication resource allocation strategy for the Bluetooth module based on the prediction results and the comprehensive utility score corresponding to each candidate communication resource allocation strategy includes: when the local scheduling model predicts that there is a resource allocation conflict, sending a query request to the Bluetooth module to query the urgency of the current pending service of the Bluetooth module; or, parsing the service request messages reported and cached by each Bluetooth module to determine the urgency of the current pending service of each Bluetooth module; and allocating communication resources to the Bluetooth module with higher urgency based on the urgency of the current pending service of the Bluetooth module.

8. The communication resource allocation method according to claim 7, characterized in that, The method further includes: for Bluetooth modules without allocated communication resources, allocating compensation communication resources to the Bluetooth modules according to a compensation communication resource strategy; the compensation communication resource strategy is to meet the service delay limit of the Bluetooth modules, and to preferentially select communication time slots and communication channels that are closest in time to the communication resources of the original conflicting Bluetooth modules, and whose channel quality meets the transmission requirements.

9. The communication resource allocation method according to claim 1, characterized in that, The method further includes: when the edge node fails, the central server notifies the adjacent edge nodes to take over the Bluetooth module originally belonging to the failed edge node.

10. A communication resource allocation device, characterized in that, The device includes: a scheduling model training module, used to receive device parameters of smart home devices reported by a Bluetooth module, and train a local scheduling model based on the device parameters; the device parameters include local environment data and business requirement data; a model parameter aggregation module, used to upload the first model parameters of the trained local scheduling model to a central server, so that the central server performs weighted aggregation on the received first model parameters to generate an optimized global scheduling model; a scheduling model updating module, used to receive the second model parameters of the global scheduling model sent from the central server, and update the local scheduling model based on the second model parameters; and a resource allocation strategy generation module, used to generate a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

11. A communication resource allocation system, characterized in that, The system includes at least one edge node, and Bluetooth modules and a central server configured on multiple smart home devices, which are respectively communicatively connected to the edge node. The Bluetooth module is used to acquire device parameters of the smart home devices and report the device parameters to the edge node. The device parameters include local environment data and business requirement data. The edge node is used to receive the device parameters of the smart home devices reported by the Bluetooth module, and train a local scheduling model based on the device parameters. The first model parameters of the trained local scheduling model are uploaded to the central server. The central server is configured to receive first model parameters of the local scheduling model uploaded from the edge nodes, and perform weighted aggregation on the received first model parameters to generate an optimized global scheduling model; and send second model parameters of the global scheduling model to the edge nodes; the edge nodes are further configured to receive the second model parameters of the global scheduling model sent from the central server, and update the local scheduling model according to the second model parameters; and generate a communication resource allocation strategy for the Bluetooth module based on the updated local scheduling model; the communication resource allocation strategy includes the communication time slots and communication channels of the Bluetooth module.

12. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the communication resource allocation method as described in any one of claims 1-9.

13. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the communication resource allocation method as described in any one of claims 1-9.

Images