A multi-mode fusion communication optimization method, medium and device based on 2.4G and Bluetooth

Abstract

The application relates to a 2.4G and Bluetooth-based multi-mode fusion communication optimization method, medium and equipment, and belongs to the technical field of communication, wherein the method comprises the following steps: obtaining historical working logs of all target devices, and obtaining an authorization probability distribution corresponding to each target device according to the historical working logs; if there is a predicted task in the next period, starting an active pre-authorization mechanism, distributing a virtual authorization resource pool based on the authorization probability distribution and historical channel utilization; and performing the predicted task according to the virtual authorization resource pool. The application breaks the passive and fixed rule switching mode of traditional communication technology, predicts the task demand in advance through a space-time convolution network, pre-distributes a virtual authorization resource pool, realizes active intelligent scheduling of communication resources, and effectively improves the resource utilization rate.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a multi-mode converged communication optimization method, medium and device based on 2.4G and Bluetooth. Background Technology

[0002] In the context of the booming development of the Internet of Things (IoT) and industry, multimode converged communication technology, as a key link connecting various smart devices, bears the important mission of achieving efficient and stable data interaction between devices. Multimode converged communication refers to the integration and application of multiple wireless communication technologies, working collaboratively to meet the needs of different scenarios. In the field of multimode converged communication based on the 2.4GHz band and Bluetooth technology, 2.4GHz technology, with its high transmission rate and long transmission distance, is often used for tasks with high bandwidth requirements such as video surveillance and file transfer; Bluetooth technology, with its advantages of low power consumption and low cost, is suitable for scenarios such as sensor data acquisition and equipment status monitoring. This convergence greatly expands the application scope of communication systems, effectively improves device compatibility and resource utilization, and provides flexible communication solutions for fields such as smart homes and industrial monitoring.

[0003] However, most current traditional multi-mode converged communication methods rely on fixed rules or simple threshold judgments for mode selection and resource allocation. In terms of mode selection, they usually follow fixed rules such as "select 2.4G for high bandwidth requirements and Bluetooth for low power consumption", or simple threshold logic of switching modes when the Received Signal Strength Indicator (RSSI) is less than -90dBm. From the perspective of resource allocation, traditional communication methods often adopt a passive mode of "real-time request-instant response", which only analyzes channel utilization based on physical layer parameters such as RSSI, and lacks in-depth mining of the device's historical communication behavior and prediction of future task requirements.

[0004] First, the above methods are poorly adaptable to complex and ever-changing communication environments and cannot comprehensively consider multiple factors such as device status, channel environment, and task characteristics. For example, the 2.4G frequency band is easily affected by interference sources such as microwave ovens and Wi-Fi, and traditional methods are difficult to dynamically adjust the mode selection probability based on the real-time interference level. Bluetooth is prone to frequency hopping conflicts in high-density device scenarios, but cannot optimize channel allocation by combining historical conflict data.

[0005] Secondly, the above methods are inefficient in resource allocation. Because they do not utilize historical work logs to mine communication behavior patterns, they cannot predict resource demand in advance, resulting in high real-time scheduling delays. Furthermore, they often lead to contradictions such as ineffective utilization of idle frequency bands and excessive contention of busy frequency bands. Summary of the Invention

[0006] To address the problems of poor adaptability to communication environments and low resource allocation efficiency in traditional communication methods, this application provides a multi-mode fusion communication optimization method based on 2.4G and Bluetooth.

[0007] Firstly, this application provides a multi-mode converged communication optimization method based on 2.4G and Bluetooth, employing the following technical solution:

[0008] A multi-mode fusion communication optimization method based on 2.4G and Bluetooth includes:

[0009] Obtain the historical work logs of all target devices, and obtain the authorization probability distribution corresponding to each target device based on the historical work logs. The authorization probability distribution is a selection probability model of 2.4G and Bluetooth modes in different scenarios, dynamically associated with device status, channel quality and task type.

[0010] If a prediction task exists in the next time period, the active pre-authorization mechanism is activated, and a virtual authorization resource pool is allocated based on the authorization probability distribution and historical channel utilization.

[0011] The prediction task is executed based on the virtual authorized resource pool.

[0012] By adopting the above technical solutions, this application achieves a leap from traditional fixed rules to data-driven intelligent decision-making: the authorization probability distribution generated based on historical logs can dynamically associate multi-dimensional parameters such as device power and channel interference level, enabling the selection probability of 2.4G and Bluetooth modes to adaptively adjust with the real-time environment, greatly improving the communication success rate compared with the traditional threshold switching method; the active pre-authorization mechanism predicts task requirements and pre-allocates virtual authorization resource pools through spatiotemporal convolutional networks, shortening the communication resource scheduling delay, especially in industrial real-time control scenarios, it can significantly reduce command transmission delay; the synergistic optimization of virtual authorization resource pool and authorization probability distribution achieves a multi-objective balance of success rate, delay, and power consumption, which helps to reduce the overall power consumption of the system, improve resource utilization, and effectively solve the technical problems of poor environmental adaptability, high scheduling delay, and serious resource waste in traditional multi-mode communication.

[0013] In one specific implementation, obtaining the authorization probability distribution corresponding to each target device based on the historical work log includes:

[0014] A multi-dimensional data feature system is constructed to parse and preprocess historical logs. The feature system covers device status, communication environment and task characteristics.

[0015] A state-space model of communication behavior is constructed, and a state transition function and a reward function are defined based on a Markov decision process. A pattern selection probability vector is generated through reinforcement learning.

[0016] Construct a multi-objective probabilistic optimization model and dynamically adjust the mode selection weights by combining Bayesian update and conflict resolution algorithms;

[0017] A spatiotemporal consistency verification mechanism is constructed, and the spatiotemporal adaptability of the probability distribution is ensured through cluster analysis and interpolation models;

[0018] The authorization probability distribution is obtained based on the communication behavior state space model, the multi-objective probability optimization model, and the spatiotemporal consistency verification mechanism.

[0019] By adopting the above technical solutions, this application constructs a full-link intelligent decision-making system from data preprocessing to model verification: the multi-dimensional feature system, through adaptive sliding window cleaning and spatiotemporal correlation modeling, removes abnormal data (such as RSSI mutation exceeding the threshold) from historical logs, improving the accuracy of the feature vectors input to the model and laying a data foundation for probability distribution; based on the state-space model of Markov decision process, through reward function design, dynamic modeling of communication mode selection is realized, which reduces the prediction error of state transition probability compared with traditional finite state machine model; the spatiotemporal consistency verification mechanism shortens the model adaptation time when deploying across regions through DBSCAN spatial clustering and Kriging interpolation;

[0020] This complete technology chain enables the authorized probability distribution to have three-dimensional decision-making capabilities: real-time response to device status (battery / temperature), dynamic adaptation to channel characteristics (RSSI / bit error rate), and intelligent matching of task attributes (real-time performance / data volume). Compared with traditional fixed rules, this significantly improves the communication success rate.

[0021] In a specific feasible implementation, if a prediction task exists in the next time period, the active pre-authorization mechanism is activated, and a virtual authorization resource pool is allocated based on the authorization probability distribution and historical channel utilization, including:

[0022] Determine the predicted events for the next moment and assess the predicted resource requirements;

[0023] The channel historical utilization rate is calculated based on the historical work logs to obtain the idle frequency bands;

[0024] Based on the available frequency bands and the predicted resource requirements, a virtual authorized resource pool corresponding to each predicted event is allocated;

[0025] The process of determining the predicted event at the next moment and judging the predicted resource requirements includes:

[0026] Based on the spatiotemporal convolutional neural network, the spatiotemporal patterns of the task are extracted, and the task is classified into hard real-time, soft real-time and non-real-time tasks by combining the state space model.

[0027] Based on the authorized probability distribution and the resource consumption per mode unit, the resource requirements of each task are quantitatively calculated. The resource consumption per mode unit includes 2.4G bandwidth and Bluetooth power consumption parameters.

[0028] In one specific implementation, the step of calculating the historical channel utilization rate based on the historical work log to obtain the idle frequency band includes:

[0029] The number of CSMA / CA collisions in the 2.4G channel and the frequency hopping success rate of the Bluetooth channel were analyzed, and the channel occupancy rate was calculated by combining dynamic sliding window.

[0030] Generate a heatmap of channel utilization in the spatiotemporal dimensions, associate it with time slice index and spatial location fingerprint, and identify high-interference areas and busy periods;

[0031] The system filters idle frequency bands based on a dual threshold, which includes a basic channel quality threshold and a dynamic correction threshold. The dynamic correction threshold is associated with device status and task type.

[0032] In one specific implementation, the allocation of a virtual licensed resource pool corresponding to each predicted event based on the idle frequency band and the predicted resource demand includes:

[0033] A three-tier resource pool architecture is established, with resource allocation based on real-time control, data transmission, and low-power tasks. The real-time control resource pool is given priority in allocating high signal-to-noise ratio 2.4G channels.

[0034] When the predicted resource demand exceeds the available idle frequency bands, the low-priority task resources are preempted based on the near-end strategy optimization algorithm, and the resource gap is compensated through a channel dynamic switching mechanism.

[0035] In one specific implementation, the step of performing the prediction task based on the virtual authorized resource pool includes:

[0036] The communication mode is dynamically calibrated through a dual-verification mechanism for mode selection. When the deviation between the pre-assigned mode and the authorized probability distribution exceeds a threshold, a reselection is triggered.

[0037] The Hungarian algorithm is used to map logical resources in the virtual authorized resource pool to physical layer parameters, and the transmission parameters are dynamically adjusted in combination with the device status.

[0038] When channel quality changes abruptly or task priority is upgraded, reinforcement learning strategies are invoked to reallocate resources and the authorization probability distribution is updated synchronously.

[0039] In one specific implementation, the method further includes:

[0040] Real-time detection of temporary emergencies, and activation of resource reallocation mechanism in the event of such an emergency.

[0041] In one specific implementation scheme, the real-time detection of temporary emergencies and the activation of the resource reallocation mechanism in the presence of such emergencies include:

[0042] Detecting equipment malfunctions using an isolated forest model;

[0043] The rate of change of channel parameters is calculated based on a sliding window, and the presence of sudden environmental changes is determined based on the rate of change of channel parameters.

[0044] If the determination is yes, then the abnormal communication pattern is identified by parsing the protocol layer, and resource reallocation is completed according to the abnormal communication pattern.

[0045] Secondly, this application provides an electronic device that adopts the following technical solution:

[0046] An electronic device is characterized by comprising a memory and a processor, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement a multi-mode converged communication optimization method based on 2.4G and Bluetooth as described in the first aspect.

[0047] Thirdly, this application provides a computer-readable storage medium, which adopts the following technical solution:

[0048] A computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement a multi-mode converged communication optimization method based on 2.4G and Bluetooth as described in the first aspect.

[0049] In summary, this application includes at least one of the following beneficial technical effects:

[0050] 1. This application constructs a data-driven intelligent decision-making system, transforming the device's historical communication experience into a dynamic authorization probability distribution, enabling the communication mode selection to adapt to the device status, channel environment, and task requirements in real time, thus solving the problem of insufficient adaptability of traditional fixed rule switching to complex environments;

[0051] 2. This application introduces an innovative spatiotemporal joint optimization mechanism, which realizes the spatiotemporal consistency verification of probability distribution through cluster analysis and interpolation model. It can quickly adapt to the communication patterns of different regions and time periods without retraining, breaking through the adaptation bottleneck of traditional models for cross-regional deployment.

[0052] 3. This application designs a prediction-driven proactive pre-authorization mechanism, which predicts task requirements and pre-allocates virtual authorization resource pools based on spatiotemporal convolutional networks, transforming traditional passive scheduling into proactive planning, significantly improving resource scheduling efficiency and real-time response capabilities, optimizing energy consumption and resource utilization while ensuring communication quality, and meeting the differentiated needs of different scenarios. Attached Figure Description

[0053] Figure 1 A flowchart of a multi-mode converged communication optimization method based on 2.4G and Bluetooth, provided in an embodiment of this application, is shown.

[0054] Figure 2 This is a schematic diagram of the structure of an electronic device disclosed in another embodiment of this application.

[0055] Explanation of reference numerals in the attached figures: 500, electronic device; 501, processor; 502, communication bus; 503, user interface; 504, network interface; 505, memory. Detailed Implementation

[0056] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0057] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0058] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0059] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0060] Reference Figure 1 An exemplary embodiment of this application provides a multi-mode fusion communication optimization method based on 2.4G and Bluetooth, including:

[0061] S100: Obtain the historical working logs of all target devices, and obtain the authorization probability distribution corresponding to each target device based on the historical working logs;

[0062] The authorization probability distribution refers to the probability model of choosing between 2.4G and Bluetooth communication modes in different scenarios, generated through in-depth analysis of the target device's historical communication data. The authorization probability distribution is not a fixed value, but a dynamic probability mapping relationship that associates multiple dimensions such as device operating status, channel quality, and task type. For example, when the device's battery is below 20%, the selection probability of Bluetooth mode will be given a higher weight based on historical power consumption data. When 2.4G band interference sources such as microwave ovens are detected, the system will automatically reduce the authorization probability of 2.4G mode based on the communication success rate when interference occurs in historical logs.

[0063] It is important to note that the authorization probability distribution analyzed in step S100 essentially transforms the environmental adaptation experience from the device's past communication behavior into a mathematical distribution containing mode selection probabilities. This distribution then guides intelligent decision-making for the current communication mode, achieving a balance between multiple objectives such as improved success rate, reduced latency, and optimized power consumption. Specifically, the authorization probability distribution can learn state transition patterns through a Hidden Markov Model, fit the probability density of performance indicators using a Gaussian Mixture Model, and continuously iterate using reinforcement learning algorithms. Ultimately, it forms a dynamic probability mapping system that can respond in real time to device status (such as battery level and temperature), channel characteristics (such as RSSI and bit error rate), and task attributes (such as real-time performance and data volume). This enables communication mode selection to possess "data-driven intelligent predictive capabilities." Specifically, S100 may include the following sub-steps:

[0064] S110, construct a multi-dimensional data feature system to perform in-depth analysis and preprocessing of historical work logs;

[0065] The multi-dimensional data feature system needs to cover three major dimensions: device status, communication environment, and task characteristics. Specifically, this includes hardware status parameters such as device power consumption and CPU load; channel quality indicators such as RSSI signal strength and SNR signal-to-noise ratio; and task attributes such as data transmission volume and real-time requirements. Specifically, S110 can be further subdivided into:

[0066] S111, Design an adaptive sliding window algorithm for data cleaning;

[0067] In this application, a dynamic time window (the window size is automatically adjusted according to the communication frequency, ranging from 5 to 30 minutes) is mainly used to sample the original logs. The 3σ criterion and the K-nearest neighbor algorithm are combined to identify and remove outliers. For example, when an RSSI value mutation is detected to exceed 3 times the historical standard deviation, it is automatically marked as an anomaly and corrected by interpolation using the nearest neighbor value of K=5.

[0068] S112, Establish the spatiotemporal correlation feature matrix;

[0069] The cleaned data is expanded according to time series to construct a three-dimensional feature matrix that includes time slice indexes (such as weekdays / weekends, morning, noon and evening time periods), spatial location fingerprints (located jointly by Bluetooth beacons and 2.4G signal strength), and device status vectors. For example, it can identify that a device is in the morning in the conference room area, and the success rate of Bluetooth communication is 27% higher than that of 2.4G.

[0070] S113 enables multimodal data fusion encoding;

[0071] Among them, multimodal data fusion coding refers to Z-score standardization of numerical parameters (such as power and latency) and one-hot coding of categorical parameters (such as task type and interference source type), and finally mapping multidimensional features into feature vectors of fixed length. In addition, for time-series features, the use of Long Short-Term Memory (LSTM) network also requires the extraction of sequence dependencies, such as capturing the time lag effect between device power decline and communication mode switching.

[0072] It should be noted that in step S110, the construction of the feature system must follow the principle of dynamic expansion. When a new type of interference source (such as newly added Wi-Fi channel overlap) or device status parameters (such as abnormal temperature affecting communication quality) are detected, the system will automatically trigger feature space reconstruction, expanding the feature dimension through incremental learning without retraining the full model. This mechanism can effectively ensure that the system can maintain the integrity and adaptability of feature expression when facing complex and ever-changing communication environments, providing a solid data foundation for subsequent probability distribution modeling.

[0073] S120, Construct a state-space model of communication behavior and realize dynamic probabilistic modeling of communication patterns based on Markov decision process (MDP);

[0074] The state-space model needs to accurately represent the dynamic relationship between device state, channel quality, and communication mode selection, specifically including:

[0075] S121, Define the multimodal state transition function;

[0076] The main approach maps device status (battery level, temperature, etc.), channel status (interference level, available bandwidth), and task type (real-time control, data transmission) to state vector S in MDP. Communication mode selection (2.4G / Bluetooth) is defined as action A. The state transition probability P(S'|S,A) is obtained through historical data statistics. For example, when the device is in a "low battery + high interference" state, the probability of transitioning to the "communication successful" state after selecting Bluetooth mode is 0.78, which is significantly higher than 0.42 for 2.4G mode.

[0077] S122, Design the instant reward function;

[0078] The reward function R(S,A,S') is constructed primarily by considering three dimensions: communication success rate, transmission delay, and power consumption. Its specific form is as follows:

[0079] ;

[0080] ;

[0081] in, This serves as the basic reward item for scene adaptation, reflecting the weighted balance of multiple objectives under normal communication conditions. As a multi-dimensional anomaly penalty item, negative constraints are imposed on issues such as communication failure, latency exceeding limits, and power consumption exceeding limits to avoid overall performance degradation caused by optimization in a single dimension;

[0082] This is a dynamic weight adjustment factor used to amplify / reduce the weight of the basic reward item, adapting to the priority of different scenarios. The value range of is [0.336, 2.184]; in this embodiment, real-time tasks (such as control commands) Take 1.2, for non-real-time tasks (such as data backup). Use 0.8; Low battery (<20%) Take 1.3, medium battery level (20%-80%). Use 1.0, high battery level (>80%). Use 0.7; High interference (e.g., microwave oven + Wi-Fi overlap). Take 1.4, medium interference Set to 1.0 for low interference. Take 0.6;

[0083] This is the basic weighting coefficient for communication success rate, reflecting the basic priority of "success rate first". In this embodiment, it is fixed at a value of 40. This is the basic weighting coefficient for transmission delay, reflecting the basic priority of "low latency first". In this embodiment, it is fixed at a value of 35. This is the basic weighting coefficient for power consumption, reflecting the basic priority of "low power consumption first". In this embodiment, it is fixed at a value of 25.

[0084] r represents the success rate of a single communication in the current communication mode (2.4G / Bluetooth), with a value range of [0,1]. It is statistically determined by the "ACK feedback from the receiving end" (r=1 for success, r=0 for failure). It is updated based on the physical layer feedback after each communication and is the core indicator for judging the effectiveness of communication (e.g., r is usually higher for Bluetooth when the battery is low than for 2.4G). t represents the actual transmission delay of the current communication mode (the total time from when the sender sends data to when the receiver returns an ACK), with a value range of [10,1000]. It can be calculated in real time by using the difference between the timestamps of the sender and receiver. For example, t≈50ms for 2.4G when there is no interference, and t≈300ms when there is interference.

[0085] The transmission delay threshold (the maximum acceptable delay for a task) is dynamically configured by the upper-layer application based on the task type. In this embodiment, Tt = 100ms for real-time tasks and Tt = 500ms for non-real-time tasks. p is the power consumption per unit of data in the current communication mode (the power consumed to transmit 1KB of data), with the unit being mWh / KB and a value range of [3, 20] (for Bluetooth, p ≈ 5, and for 2.4G, p ≈ 12). p is collected in real time by the device's power management module. When the battery is low, the device will automatically report the critical value of p (e.g., when the battery is <20%, p exceeding 5mWh / KB will trigger a low battery alarm).

[0086] As a power consumption threshold based on device power (the maximum power consumption per unit of data that the device can accept), in this embodiment, low power (<20%) Pb=5mWh / KB, medium power Pb=10mWh / KB, and high power Pb=15mWh / KB;

[0087] The communication failure penalty coefficient (triggered when r=0) is fixed at -50 in this embodiment (negative penalty, the absolute value is much greater than the maximum possible value of the basic reward). Communication failure will cause the task to be interrupted (such as loss of control commands), so the penalty is the strongest, guiding the system to prioritize the "high success rate mode" (such as prioritizing Bluetooth in case of high interference).

[0088] The power consumption over-limit penalty coefficient (when (Triggered by time), in this embodiment the value is fixed at -15. Exceeding the power limit only affects the device's battery life (not an immediate interruption), so the penalty is minimal. The penalty effect is amplified by α only in low power scenarios (e.g., when the power is low, η=1.3, which enhances the actual impact of δ).

[0089] I(condition)∈{0,1} is an indicator function (1 when the condition is met, 0 when the condition is not met), which means that the penalty is triggered only when an anomaly occurs, and there is no additional negative impact during normal communication, so as to ensure the "normal scenario guidance" and "abnormal scenario constraint" of the reward function.

[0090] The following example illustrates the immediate reward function in S122 above:

[0091] Assuming a device is performing a real-time control task (τ=1.2), has 15% battery (η=1.3), and is in a high-interference environment (ι=1.4), when Bluetooth mode is selected:

[0092] α = 1.2 × 1.3 × 1.4 = 2.184

[0093] r=0.9,

[0094] t=80ms< =100ms

[0095] p=4mWh / KB< =5mWh / KB;

[0096] Basic bonus item: 40×0.9−35×(80 / 100)−25×(4 / 5)=36−28−20=−12;

[0097] Abnormal penalty terms: I(r=0)=0, I(t> )=0、I(p> Since ) = 0, the penalty term = 0;

[0098] Total reward: 2.184 × (−12) + 0 ≈ −26.21;

[0099] If you choose the 2.4G mode (poor performance under high interference):

[0100] r=0.5,

[0101] t=120ms> =100ms

[0102] p=6mWh / KB> =5mWh / KB;

[0103] Basic bonus: 40×0.5−35×(120 / 100)−25×(6 / 5)=20−42−30=−52;

[0104] Abnormal penalty term: β×1+γ×1+δ×1=−50−20−15=−85;

[0105] Total reward: 2.184×(−52)−85≈−113.57−85=−198.57.

[0106] It is evident that the reward for Bluetooth mode (-26.21) is much higher than that for 2.4G (-198.57), and the function can effectively guide the system to select the optimal communication mode, which meets the design goals.

[0107] S123 implements deep reinforcement learning optimization. A dual-deep Q-network (DDQN) architecture is used to train the policy network π(a|s). An experience replay mechanism addresses data correlation issues, and priority experience replay is introduced to improve sample utilization efficiency. Specifically, to handle the continuous state space, the Deep Deterministic Policy Gradient (DDPG) algorithm is used to optimize the probability distribution in the continuous action space, outputting the selection probability vectors [P_2.4G, P_BLE] for 2.4G and Bluetooth modes.

[0108] It is worth noting that in step S120, the state space model adopts a hierarchical structure. The bottom layer maintains a basic state space, containing general parameters such as RSSI and power level; the upper layer constructs a domain-specific state space, dynamically loading dedicated parameters for different application scenarios (such as industrial monitoring and smart homes). When new environmental features are detected (such as 5G NR-U interference), the system adapts quickly through a meta-learning mechanism, requiring only a small number of samples to update the state transition function, ensuring the model's generalization ability in complex heterogeneous network environments. This layered architecture enables the system to maintain stable modeling of basic communication modes while flexibly responding to the special needs of vertical domains, providing accurate probabilistic decision support for multi-mode converged communication.

[0109] S130, Construct a multi-objective probabilistic optimization model;

[0110] The multi-objective probabilistic optimization model is used to achieve dynamic weight allocation and real-time decision-making for communication mode selection. This model needs to balance multiple objectives such as communication success rate, latency, and power consumption, specifically including:

[0111] S131, Design an adaptive weight adjustment mechanism;

[0112] The main approach involves dynamically adjusting the weight coefficients of each objective based on real-time task requirements, constructing a weighted utility function to calculate the utility values ​​of the two modes, and selecting the mode with the higher U(S,A) (a higher utility value indicates a better fit for the current task, device, and channel scenario):

[0113] ;

[0114] in, The utility value of communication mode A (2.4G / Bluetooth) in the current state S; This is a positive indicator of communication success rate in the [0,1] interval. ; This is a positive indicator of transmission delay in the [0,1] interval. ; This indicates that exceeding the delay limit results in no positive contribution. This is a positive indicator of unit power consumption in the [0,1] interval. , This indicates that the power consumption exceeded the limit and made no positive contribution. All weights are dynamic and can be automatically assigned based on task type using a fuzzy logic controller. For example, when real-time video streaming is detected, the system automatically assigns weights accordingly. Increased to 0.6; when the device battery level is below 20%, It automatically adjusts to 0.7, etc.

[0115] The following example illustrates the weighted utility function in S131 above:

[0116] The scenario input is as follows:

[0117] Task type: Real-time video streaming (real-time, triggered by rule 1);

[0118] Device status: SOC=15% (low battery, but the task is real-time, so latency will be prioritized).

[0119] Channel quality: Medium interference (ι=1.0);

[0120] Mode performance parameters:

[0121] Bluetooth mode: r=0.9, t=180ms ( =200ms), p=4.5mWh / KB ( =5mWh / KB);

[0122] 2.4G mode: r=0.6, t=220ms ( =200ms), p=6mWh / KB ( =5mWh / KB);

[0123] The processor's calculation process is as follows:

[0124] Fuzzy controller assigns weights: matching rule 1, output ;

[0125] Indicator normalization:

[0126] Bluetooth: ;

[0127] 2.4G: ;

[0128] Utility value calculation:

[0129] Bluetooth: U = 0.3 × 0.9 + 0.6 × 0.1 + 0.1 × 0.1 = 0.27 + 0.06 + 0.01 = 0.34;

[0130] 2.4G: U=0.3×0.6+0.6×0+0.1×0=0.18;

[0131] In summary, the utility value of Bluetooth mode (0.34) is much higher than that of 2.4G (0.18). At this time, the system automatically selects Bluetooth mode, which not only meets the low latency requirements of real-time video streaming, but also takes into account power consumption control in low-battery scenarios.

[0132] S132, implements Bayesian update of probability distribution;

[0133] Specifically, based on the current state observations, the authorization probability distribution is updated using a Bayesian inference framework:

[0134] ;

[0135] Where M represents the communication mode (2.4G / Bluetooth), and O represents the observation vector (RSSI, SNR, etc.). By calculating the posterior probability, the system corrects the mode selection bias in real time. For example, when the observed SNR is <15dB, the posterior probability of the 2.4G mode decreases by 40%.

[0136] S133, Design a conflict resolution algorithm;

[0137] In cases where multi-objective optimization results conflict (e.g., the 2.4G mode has a high success rate but high power consumption), the Pareto optimality principle is used to generate a non-dominated solution set, and the comprehensive utility value of each solution is calculated using the entropy weight method. Specifically, for edge computing scenarios, a computational offloading factor is introduced to extend the utility function.

[0138] ;

[0139] Where δ is the calculation offload weight, so that the system takes into account the computing resource status of edge devices when making communication decisions.

[0140] It is worth noting that in step S130, the model employs a hybrid architecture combining online learning and offline training. In the offline phase, an initial policy library is generated using Monte Carlo Tree Search (MCTS), while in the online phase, Thompson sampling is used to balance exploration and exploitation. When a change in environmental distribution is detected (such as differences in communication patterns between weekdays and weekends), the system triggers concept drift detection, achieving a smooth transfer of the probabilistic model through incremental support vector machines. This mechanism ensures that the system, during long-term operation, can both effectively utilize historical experience and quickly adapt to dynamically changing communication environments, ultimately achieving optimal probabilistic decision-making for multi-modal fusion communication.

[0141] To address the probability distribution deviation caused by cross-regional deployment and time evolution, S100 of this embodiment further includes the following steps:

[0142] S140, Construct a spatiotemporal consistency verification and optimization mechanism for probability distribution;

[0143] The spatiotemporal consistency verification and optimization mechanism is used to ensure the stability and adaptability of the authorization probability under different spatiotemporal dimensions; specifically, S140 includes:

[0144] S141, Design a spatiotemporal clustering analysis framework;

[0145] Specifically, the DBSCAN algorithm is used to spatially cluster the geographical locations of devices, and K-means is used to divide the time dimension into time periods to form a spatiotemporal grid. For example, the office area is divided into three spatial clusters: meeting room, office area, and corridor. Each cluster is further subdivided according to weekday / weekend and peak / off-peak time periods. Devices in each spatiotemporal grid are modeled separately, and a special pattern is found that the Bluetooth success rate in the meeting room area is 15% higher on weekend afternoons than on weekdays.

[0146] S142 performs consistency verification of the probability distribution;

[0147] The system uses the Walshstein distance to measure the difference in probability distribution between different spatiotemporal grids. When the difference exceeds a threshold (such as 0.25), local retraining is triggered. For example, if a newly deployed Wi-Fi device on a certain floor causes a shift in the probability distribution of the 2.4G mode, the system will automatically start fine-tuning of the model in that area.

[0148] S143, Construct a spatiotemporal interpolation model for the probability distribution;

[0149] Specifically, based on the Kriging interpolation algorithm, the probability values ​​of unknown regions are estimated using the probability distribution of adjacent spatiotemporal grids, enabling rapid modeling of uncovered areas; in particular, a spatiotemporal weight decay factor is introduced:

[0150] Wherein, d iτ represents spatial distance. i Let α be the time distance, and β be the decay coefficients. The optimal values ​​are determined through cross-validation.

[0151] S150, establish a continuous learning and adaptive update mechanism for probability distribution;

[0152] This mechanism is used to ensure the effectiveness of the model in long-term operation, and specifically includes:

[0153] S151, Design concept drift detection algorithm;

[0154] The ADWIN algorithm is used to monitor changes in data distribution in real time. When concept drift is detected, the sliding window size is dynamically adjusted (minimum 500 samples, maximum 5000 samples). For example, when the variance of Bluetooth success rate increases by 20% for 100 consecutive samples, the window is shrunk to focus on recent data.

[0155] S152, based on the online gradient boosting framework, uses weighted gradient updates for new samples;

[0156] Among them, the weight of historical samples decays exponentially over time (half-life = 7 days); in particular, when new data contains rare events (such as extreme interference scenarios), its weight is automatically increased to 3 times that of ordinary samples.

[0157] S160, generates the final authorization probability distribution based on all mechanisms;

[0158] S160, by integrating multi-dimensional features, dynamic weight adjustment, and adaptive model updates, forms a complete probabilistic decision-making system, specifically including:

[0159] S161, Construct the probability fusion function;

[0160] Specifically, the multi-objective optimization probability obtained from S130, the spatiotemporal interpolation probability from S140, and the incremental learning probability from S150 are weighted and fused together:

[0161] ;

[0162] Wherein, ω1, ω2, and ω3 are the fusion weights, and the optimal values ​​are determined through cross-validation (typical configuration: 0.5, 0.3, 0.2). This function ensures that the probability distribution contains both the theoretical optimal solution of multi-objective optimization and retains spatiotemporal characteristics and model evolution information.

[0163] S162, performs normalization of the probability distribution;

[0164] Specifically, a Softmax transformation is performed on the fused probabilities to ensure... Simultaneously, a temperature parameter T is introduced to adjust the smoothness of the probability distribution. When T=1, it is the standard Softmax distribution; when T>1, the distribution is smoother; and when T<1, the distribution is sharper. The optimal temperature parameter T=0.85 is determined through grid search, achieving the best balance between exploration and utilization of the probability distribution.

[0165] S163, Establish confidence interval estimates for the probability distribution;

[0166] In this process, a nonparametric bootstrapping method is used to calculate the confidence interval for the probability estimate in each state. For example: States with confidence intervals exceeding a threshold (e.g., ±0.1) will be marked as high uncertainty regions, triggering an active learning mechanism to collect more samples.

[0167] Finally, through the complete process of steps S110 to S160, the system successfully generates the authorization probability distribution P(m|s) corresponding to each target device. This distribution is stored in the form of a conditional probability table (CPT) on the edge node, while the parameter set of the probability generation model is maintained in the cloud. When communication mode decision is required, the corresponding probability value can be quickly queried by simply inputting the current state vector s, thus realizing intelligent mode selection based on historical experience and real-time environment.

[0168] S200: If there is a predicted task in the next time period, the active pre-authorization mechanism will be activated to allocate a virtual authorization resource pool.

[0169] The proactive pre-authorization mechanism refers to the system allocating communication resources in advance for tasks in future time periods based on historical work logs and prediction models, forming a virtual authorization queue to reduce real-time scheduling latency. The virtual authorization resource pool is a dynamically reserved set of communication resources, including 2.4G channels, Bluetooth connection slots, etc., and its capacity is dynamically adjusted according to the priority, data volume, and real-time requirements of the predicted tasks. Specifically, S200 may include the following sub-steps:

[0170] S210, determine the predicted event for the next moment and assess the predicted resource requirements;

[0171] This step relies on the multi-dimensional feature system and authorization probability distribution model built by S100 to achieve deep coupling between task prediction and resource demand assessment:

[0172] S211, Spatiotemporal correlation prediction modeling;

[0173] In this embodiment, based on the time slice index (e.g., weekday / weekend, morning / noon / evening time slots) and spatial location fingerprint (Bluetooth beacon + 2.4G signal positioning) preprocessed by S110, a spatiotemporal convolutional neural network (STCNN) is used to extract the spatiotemporal patterns of task occurrence. For example, by analyzing the historical logs of a smart factory, it was found that real-time control tasks of production line equipment surged at 10:00 AM every day, at which time the demand for latency-sensitive resources in 2.4G mode increased by 30%.

[0174] S212, Dynamic classification of task types;

[0175] In this embodiment, the state space model constructed by S120 is used to divide the prediction tasks into three categories: hard real-time tasks (such as robotic arm control): require 2.4G mode transmission latency <10ms, and have the highest priority in resource requirements; soft real-time tasks (such as video surveillance): allow Bluetooth mode latency <50ms, and resource requirements need to take into account both bandwidth and power consumption; and non-real-time tasks (such as device log upload): can be flexibly scheduled, and prioritize the use of Bluetooth Low Energy mode.

[0176] S213, Quantitative calculation of resource requirements;

[0177] Among them, the resource demand calculation formula is established by combining the mode selection probability output by the S130 multi-objective optimization model:

[0178] ;

[0179] Where Q(m) is the unit resource consumption of mode m (500kbps bandwidth per unit in 2.4G mode, 10mW power consumption per unit in Bluetooth mode), and P(m|s_t) is the mode selection probability in the current state generated by S100.

[0180] S220: Calculate the historical channel utilization rate based on the historical work log to obtain the idle frequency band;

[0181] This step deeply reuses the channel state parameters (RSSI, SNR, bit error rate) and communication event records (number of collisions, number of retransmissions) collected by the S100 to construct a dynamic channel evaluation system.

[0182] S221, Multimodal channel utilization analysis;

[0183] Specifically, for the 2.4G channel, based on the number of CSMA / CA collisions and the transmission failure rate, a sliding window (the window size is the same as S111, dynamically adjusted from 5 to 30 minutes) is used to calculate the channel occupancy rate. When the occupancy rate is >80%, it is marked as a busy frequency band. For the Bluetooth channel, interference frequency points are identified by the historical success rate of frequency hopping sequences and ACK response time (e.g., when the success rate of the channel near 2450MHz commonly used in microwave ovens is <60%, it is marked as an interference segment).

[0184] S222, Spatiotemporal Dimension Utilization Mapping;

[0185] Among them, a channel utilization heatmap is generated by combining the spatiotemporal correlation feature matrix established in S112:

[0186] Time dimension: During the morning peak (9:00-11:00), the utilization rate of the 2.4G channel increases by an average of 25%, and the available time slots for Bluetooth mode increase by 18% during this period;

[0187] Spatial dimension: Due to the high density of Wi-Fi hotspots in the conference room area, the utilization rate of 2.4G channels 1-3 has been consistently >90%. The system automatically shifts the pre-authorized resources in this area to Bluetooth mode.

[0188] S223, intelligent filtering of idle frequency bands;

[0189] The system employs a dual-threshold screening mechanism, consisting of a basic channel quality threshold and a dynamic correction threshold. The basic channel quality threshold requires a 2.4G channel utilization rate of <30% and an SNR >20dB, while the Bluetooth channel bit error rate is <5%. The dynamic correction threshold adjusts the screening criteria based on the device status in the S100 (e.g., prioritizing idle Bluetooth frequency bands when battery level is <20%) and the task type (avoiding frequency bands with a historical conflict rate >15% for real-time tasks).

[0190] S230, allocates a virtual licensed resource pool corresponding to each predicted event based on the idle frequency band and predicted resource demand;

[0191] In step S230, the authorization probability distribution from S100 is combined with the real-time calculation results from S210 / S220 to achieve accurate pre-allocation of resources.

[0192] S231, Tiered resource pre-allocation;

[0193] Specifically, a three-tier resource pool architecture is established to achieve deep collaboration with the state space model of S120: For the real-time control resource pool (accounting for 30%), the high signal-to-noise ratio 2.4G channel selected by S220 is prioritized for allocation, and fixed time slot pre-allocation is adopted (such as reserving a 1ms transmission window every 10ms) to ensure that the authorization probability of hard real-time tasks is ≥95%; For the data transmission resource pool (accounting for 50%), the 2.4G / Bluetooth ratio is dynamically adjusted according to the multi-objective optimization results of S130. For example, when the prediction task is a video stream, the 2.4G resource ratio is increased to 70%, and 10% elastic resources are reserved to cope with sudden traffic; For the low-power resource pool (accounting for 20%), only Bluetooth LE (Low Power Mode) time slots are allocated. Combined with the probability distribution of device battery <20% in S100 (Bluetooth mode probability +40%), an asynchronous wake-up mechanism is adopted to reduce standby power consumption.

[0194] S232, Conflict Resolution and Flexible Adjustment;

[0195] When the predicted resource demand exceeds the available idle frequency band, the reinforcement learning strategy verified in S100 is activated for dynamic adjustment. In this embodiment, a resource preemption model based on the PPO algorithm can be used to prioritize hard real-time tasks and allow temporary borrowing of non-real-time task resources (borrowing ratio ≤ 20%, with double compensation after task completion). Alternatively, a channel dynamic switching mechanism can be adopted: if there is sudden interference in the 2.4G pre-allocated channel (SNR drops sharply > 10dB), the probability distribution update of S100 is automatically triggered, the channel resource is released to the Bluetooth pool, and the equivalent resource amount is compensated.

[0196] S300 completes the actual prediction tasks based on the virtual authorized resource pool;

[0197] Specifically, S300 can be further divided into the following sub-steps: based on the pre-allocation of the virtual authorized resource pool in S200, it realizes the closed-loop execution of the prediction task from "resource pre-allocation" to "actual communication", while forming a dynamic linkage with the authorization probability distribution in S100:

[0198] S310, activate the virtual authorized resource pool and establish a real-time mapping relationship;

[0199] This step transforms pre-allocated virtual resources (such as 2.4G channel slots and Bluetooth connection handles) into actual communication parameters. The core of this process is dynamic calibration of resource configuration based on the license probability generated by the S100.

[0200] S311, selects a dual verification mechanism based on the mode;

[0201] Among them, the pre-authorization verification prioritizes the mode pre-allocated by S200 (e.g., 2.4G by default for hard real-time tasks), but the authorization probability of S100 under its current state needs to be calculated in real time. If the deviation between the two exceeds a threshold (e.g., 20%), mode reselection is triggered. For example, a 2.4G mode may be pre-assigned, but channel interference is detected in real time, causing... If the resource level drops to 40%, it will automatically switch to Bluetooth mode and request resource pool compensation.

[0202] S312, dynamic parameter adaptation;

[0203] Specifically, resource usage parameters are adjusted based on the device status (such as battery level and temperature) extracted in S110. When the device battery level is less than 15%, even if 2.4G resources are pre-allocated, Bluetooth Low Energy (BLE) mode is forcibly enabled, and the transmission rate is dynamically reduced according to the probability distribution of this state in S100 (Bluetooth mode + 30%).

[0204] S313, resource mapping algorithm;

[0205] Specifically, a resource matching model based on the Hungarian algorithm is used to map logical resources (such as "high-priority 2.4G channels") in the virtual authorized resource pool to physical layer parameters:

[0206] ;

[0207] Among them, c i,j The matching cost between task i and resource j (calculated based on the multi-objective optimization function of S130) ensures that real-time tasks are preferentially matched with low-latency resources, and low-power tasks are preferentially matched with Bluetooth LE time slots.

[0208] S320, implements dynamic resource adjustment and conflict resolution;

[0209] Specifically, to address potential prediction biases during actual task execution (such as sudden interference or changes in task priority), a three-layer response mechanism is constructed, deeply reusing the S100 state-space model and reinforcement learning strategies.

[0210] Parameters such as RSSI and SNR are collected in real time (sampling frequency 100Hz). When a sudden change in channel quality is detected (such as a sudden drop in SNR of 15dB), the state transition function of S120 is triggered to recalculate the grant probability. If the difference between the new probability and the pre-allocated mode is >30%, resource reconfiguration is immediately initiated. In addition, the spatiotemporal consistency check of S140 is introduced. When the actual mode selection of a certain area for five consecutive task executions deviates from the predicted probability of S100 by >25%, the area is automatically marked as a "spatiotemporal anomaly area", and 20% of elastic resources are temporarily added for fault tolerance.

[0211] If an emergency event is triggered during the execution of the actual task (such as an industrial sensor detecting a device malfunction), the resource priority should be increased by following these steps:

[0212] Call the reward function of S120 to recalculate the immediate reward value of the emergency task (temporarily increase the success rate weight α to 0.8) and force resource preemption; dynamically allocate 30% of resources from the bronze layer (elastic resources) of S200. If it is still insufficient, perform cross-layer scheduling according to the optimal mode of similar tasks in the historical data of S100 (such as the emergency control task using 2.4G with a 92% probability) to temporarily reduce the resource quota of non-real-time tasks.

[0213] It should be noted that when the resource utilization rate of a certain mode (such as 2.4G) exceeds 90%, load migration based on S130 multi-objective optimization is initiated: the "mode migration cost" (including latency increment and power consumption increment) of the task to be migrated is calculated, and tasks with a cost < threshold (such as latency increase < 10ms) are migrated to Bluetooth mode first; at the same time, the licensing probability distribution of S100 is updated in real time during the migration process, and the mode selection after load balancing is used as a new sample input to the incremental learning module of S151 to form an "execution-feedback-optimization" closed loop.

[0214] S400 can detect the presence of temporary emergencies in real time and initiate a resource reallocation mechanism based on the detection results.

[0215] Temporary emergencies refer to urgent communication needs not captured by the S200 prediction model, including but not limited to equipment failure alarms, security event triggers, and sudden data uploads. The resource reallocation mechanism ensures high-priority processing of emergencies by dynamically adjusting the virtual authorized resource pool (S200) and actual communication parameters (S300). Specifically, S400 may include the following sub-steps:

[0216] S410 constructs a multi-dimensional emergency event detection model to achieve real-time perception of emergency communication needs;

[0217] This model integrates three dimensions: device status anomaly detection, channel environment mutation identification, and protocol layer anomaly capture, specifically including:

[0218] S411, Equipment status anomaly detection;

[0219] Specifically, based on the device status parameters (battery power, temperature, CPU load, etc.) extracted from S110, an isolated forest model is constructed to detect abnormal states.

[0220] ;

[0221] Where h(x) is the path length of sample x in the isolated tree, and c(n) is the average path length of the tree. An emergency alarm is triggered when the device temperature suddenly rises above the historical 3σ range (e.g., industrial sensor temperature > 75℃), or the battery power decreases at a rate > 0.5% / minute.

[0222] S412, Channel environment change identification;

[0223] Specifically, a sliding window (window size 50ms) is used to calculate the rate of change of channel parameters such as RSSI and SNR in S110 in real time:

[0224] ;

[0225] When an SNR change rate > 10dB / ms (such as sudden microwave oven interference) or RSSI fluctuation amplitude > 15dB (such as large metal object moving and blocking), it is determined to be a channel environment change event.

[0226] S413, Protocol layer exception capture;

[0227] Among them, the MAC layer and application layer protocol data units (PDU) are parsed to identify abnormal communication patterns: three consecutive ACK timeouts (corresponding to a physical layer packet loss rate > 30%), an application layer heartbeat packet loss rate > 50% (such as Modbus protocol response timeout in industrial equipment), and control commands being sent repeatedly more than twice (such as retransmission of robotic arm control signals).

[0228] S420 establishes a tiered response mechanism based on the urgency of emergencies and resource requirements.

[0229] Based on the state-space model of S120 and the multi-objective optimization function of S130, sudden events are divided into three priority levels:

[0230] Level 1 Emergency (Highest Priority): Emergency situations involving personal safety or system crashes (such as fire alarms, equipment malfunctions); Resource Requirements: Immediately allocate the highest priority channel in the 2.4G Gold Layer resources (S200), with a pre-allocated bandwidth ≥1Mbps and a latency tolerance of <5ms; Response Strategy: Automatically interrupt all current non-critical tasks (such as log uploads), preempt their resources (up to 60%), and trigger an instant probability distribution update of S100 (2.4G mode probability in this area +50%).

[0231] Level 2 Emergency (Medium Priority): Faults affecting local functions but not endangering the overall system (e.g., abnormal data from a single sensor); Resource Requirements: Allocate 2.4G Silver Layer or Bluetooth high-priority resources, pre-allocated bandwidth ≥256kbps, latency tolerance <20ms; Response Strategy: Reduce resource quotas for non-real-time tasks (e.g., video surveillance) (by 30%), and migrate some low-priority traffic to idle channels through the S300's load balancing mechanism;

[0232] Level 3 Emergency Event (Low Priority): Temporary needs that can tolerate short delays (such as sudden file transfers); Resource Requirements: Utilize Bluetooth Bronze Layer Elastic Resources (S200), pre-allocate bandwidth ≥64kbps, and delay tolerance <100ms; Response Strategy: Wait for the current time slot to end before allocating resources, or schedule the next idle period predicted by the S140 spatiotemporal interpolation model.

[0233] S430 implements a dynamic resource reallocation and recovery mechanism to ensure system stability and fairness;

[0234] This mechanism combines reinforcement learning and game theory to minimize the impact on normal tasks while meeting the needs of unexpected events.

[0235] S431, resource reallocation based on deep deterministic policy gradient;

[0236] In this embodiment, S431 is implemented by constructing a Markov decision process. The state space S is defined as follows: it includes the priority of emergencies, current resource utilization, and the completion rate of each task; the action space A is for adjusting the resource allocation ratio of each mode (2.4G / Bluetooth); and the reward function R is for balancing the response speed to emergencies with the loss from normal task interruptions. ;

[0237] Among them, T response Loss is the event response time. i The interruption loss for task i is calculated based on the utility function of S130, and α and β are balance coefficients (typical values ​​α=0.7, β=0.3).

[0238] S432, Establish a resource acquisition compensation mechanism;

[0239] This embodiment provides three compensation strategies for tasks whose resources are preempted:

[0240] Priority scheduling: After the emergency is handled, double the resources are immediately allocated to the affected tasks (for 3 scheduling cycles). Delay compensation: The delay weight β in S130 is dynamically adjusted according to the task delay time (β increases by 0.05 for every 100ms delay). Power consumption compensation: For devices that consume extra power due to frequent mode switching, the allocation probability of Bluetooth Low Energy mode is increased (+15%) in subsequent scheduling.

[0241] S433, Design a system recovery decision tree;

[0242] In this embodiment, after an emergency is resolved, the resource recovery strategy can be determined based on the following conditions: if the resource pool utilization rate is less than 70% and there are no new events within 5 consecutive minutes, then a gradual recovery is performed, such as restoring 5% of resources every 200ms; if there are currently events of level 2 or above in the queue, then the current resource allocation is maintained until the queued events are resolved; otherwise, the preempted resources are restored immediately.

[0243] Based on the same inventive concept described above, this application also discloses an electronic device, which includes a memory and a processor. The memory stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement a multi-mode fusion communication optimization method based on 2.4G and Bluetooth as provided in the above method embodiments.

[0244] Based on the same inventive concept described above, this application also discloses a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set can be loaded and executed by a processor to implement the multi-mode fusion communication optimization method based on 2.4G and Bluetooth provided in the above method embodiments.

[0245] Reference Figure 2 This application also discloses an electronic device 500, which may include: at least one processor 501, at least one network interface 504, user interface 503, memory 505, and at least one communication bus 502.

[0246] The communication bus 502 is used to enable communication between these components.

[0247] The user interface 503 may include a display screen and a camera. Optionally, the user interface 503 may also include a standard wired interface and a wireless interface.

[0248] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0249] The processor 501 may include one or more processing cores. The processor 501 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 505, and by calling data stored in memory 505. Optionally, the processor 501 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 501 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 501 and may be implemented as a separate chip.

[0250] The memory 505 may include random access memory (RAM) or read-only memory. Optionally, the memory 505 may include non-transitory computer-readable storage medium. The memory 505 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 505 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 505 may also be at least one storage device located remotely from the aforementioned processor 501.

[0251] The memory 505, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program based on a multi-mode converged communication optimization method using 2.4G and Bluetooth.

[0252] exist Figure 2 In the illustrated electronic device 500, the user interface 503 is mainly used to provide an input interface for the user and acquire user input data; while the processor 501 can be used to call an application program stored in the memory 505 that is an optimization method for multi-mode fusion communication based on 2.4G and Bluetooth. When executed by one or more processors 501, the electronic device 500 performs one or more methods as described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0253] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0254] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0255] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units. That is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0256] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0257] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0258] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will readily conceive of those skilled in the art upon consideration of the specification and the disclosure of practical truths.

[0259] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A 2.4G and Bluetooth-based multi-mode fusion communication optimization method, characterized in that, The application comprises the following steps: acquiring historical working logs of all target devices, and obtaining an authorization probability distribution corresponding to each target device according to the historical working logs, the authorization probability distribution being a selection probability model of 2.4G and Bluetooth modes in different scenes, dynamically correlating device state, channel quality and task type; if there is a predicted task in the next period, starting an active pre-authorization mechanism, and allocating a virtual authorization resource pool based on the authorization probability distribution and historical channel utilization rate; performing the predicted task according to the virtual authorization resource pool, wherein the step of obtaining the authorization probability distribution corresponding to each target device according to the historical working logs comprises the following steps: constructing a multi-dimensional data feature system, analyzing and preprocessing the historical logs, wherein the feature system covers device state, communication environment and task characteristics; 2. The multi-modal converged communication optimization method of claim 1, wherein, constructing a communication behavior state space model, defining a state transition function and a reward function based on a Markov decision process, and generating a mode selection probability vector through reinforcement learning; constructing a multi-objective probability optimization model, dynamically adjusting the mode selection weight by combining Bayesian update and conflict resolution algorithm; constructing a spatio-temporal consistency verification mechanism, ensuring the spatio-temporal adaptability of the probability distribution through cluster analysis and interpolation model; obtaining the authorization probability distribution based on the communication behavior state space model, the multi-objective probability optimization model and the spatio-temporal consistency verification mechanism.

3. The multi-modal converged communication optimization method of claim 2, wherein, The step of starting the active pre-authorization mechanism based on the authorization probability distribution and historical channel utilization rate to allocate a virtual authorization resource pool if there is a predicted task in the next period comprises the following steps: determining a predicted event at the next moment, and judging the predicted resource demand; calculating the channel historical utilization rate according to the historical working logs to obtain an idle frequency band; allocating a virtual authorization resource pool corresponding to each predicted event according to the idle frequency band and the predicted resource demand; the step of determining a predicted event at the next moment and judging the predicted resource demand comprises the following steps: extracting the spatio-temporal law of the task based on a spatio-temporal convolutional neural network, and classifying the task into hard real-time, soft real-time and non-real-time tasks based on a state space model; quantitatively calculating the resource demand of each task according to the authorization probability distribution and the mode unit resource consumption, wherein the mode unit resource consumption includes 2.4G bandwidth and Bluetooth power consumption parameters. The step of calculating the channel historical utilization rate according to the historical working logs to obtain an idle frequency band comprises the following steps: analyzing the CSMA / CA conflict times of the 2.4G channel and the frequency hopping success rate of the Bluetooth channel, and calculating the channel occupancy rate in combination with a dynamic sliding window; generating a channel utilization rate heat map in spatio-temporal dimensions, correlating time slice index and spatial location fingerprint, and identifying high interference areas and busy periods; screening an idle frequency band based on double threshold values, wherein the double threshold values include a basic channel quality threshold value and a dynamic correction threshold value, and the dynamic correction threshold value is associated with device state and task type.

4. The multi-modal converged communication optimization method of claim 3, wherein, The virtual authorization resource pool corresponding to each prediction event is allocated according to the idle frequency band and the predicted resource requirement, including: a three-level resource pool architecture is established, and resource proportions are divided according to real-time control, data transmission and low-power consumption tasks, wherein the real-time control resource pool preferentially allocates a high signal-to-noise ratio 2.4G channel; when the predicted resource requirement exceeds the available idle frequency band, a near-end strategy optimization algorithm is used to preempt low-priority task resources, and a channel dynamic switching mechanism is used to compensate for resource gaps.

5. The multi-modal converged communication optimization method of claim 4, wherein, The predicted task is executed according to the virtual authorization resource pool, including: a mode selection double-checking mechanism is used to dynamically calibrate the communication mode, and reselection is triggered when the deviation of the pre-allocated mode from the authorization probability distribution exceeds a threshold value; a Hungarian algorithm is used to map logical resources in the virtual authorization resource pool to physical layer parameters, and transmission parameters are dynamically adjusted in combination with device states; when the channel quality mutates or the task priority upgrades, a reinforcement learning strategy is called to re-allocate resources, and the authorization probability distribution is synchronously updated.

6. The multi-modal converged communication optimization method of claim 5, wherein, The method further includes: detecting a temporary burst event in real time, and starting a resource re-allocation mechanism in the presence of the temporary burst event.

7. The multi-modal converged communication optimization method of claim 6, wherein, The real-time detection of the temporary burst event and the starting of the resource re-allocation mechanism in the presence of the temporary burst event include: detecting device state abnormalities through an isolated forest model; calculating a channel parameter change rate based on a sliding window, and determining whether there is an environmental mutation according to the channel parameter change rate; if the determination is yes, an abnormal communication mode is identified through a protocol layer, and resource re-allocation is completed according to the abnormal communication mode.

8. An electronic device, comprising: A memory and a processor are included, and the memory stores at least one instruction, at least one program, a code set or an instruction set, the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to realize the multi-mode fusion communication optimization method based on 2.4G and Bluetooth as claimed in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to realize the multi-mode fusion communication optimization method based on 2.4G and Bluetooth as claimed in any one of claims 1 to 7.

Images