Multi-network integrated internet of things card signal strength real-time comparison method

Abstract

The present application relates to the technical field of Internet of Things communication, in particular to a multi-network integrated Internet of Things card signal strength real-time comparison method, comprising: after the multi-network integrated Internet of Things card is started or network switching, initiating an attachment request to all resident networks in parallel. In a signal collection period, collecting real-time signal data streams of all attached networks simultaneously, and maintaining a dynamic signal data table to store recent structured signal records. Based on the current real-time data and recent historical records, a dynamic signal quality score is generated for each network through a multi-dimensional real-time scoring model. According to the score, all available networks are sorted to generate a real-time network quality ranking list. This method realizes the synchronous collection and instant evaluation of multi-operator and multi-standard network signals, enabling the Internet of Things terminal to quickly and accurately identify the current optimal network.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) communication technology, and in particular to a method for real-time comparison of signal strength of multi-network integrated IoT cards. Background Technology

[0002] In IoT applications, multi-network unified IoT SIM cards are designed to access networks from different operators and with different standards to enhance connection reliability. Existing network selection and signal evaluation mechanisms typically rely on sequential scanning or network attachment strategies based on preset priorities. Terminal devices first attempt to attach to a preferred network, only switching to alternative networks if the connection fails or the signal is extremely poor. This serial processing method results in lengthy network discovery and initial registration times, especially during network switching scenarios after power-on or cross-regional movement, where devices cannot quickly obtain the connection status of all available networks.

[0003] Existing signal monitoring methods mostly focus on the signal strength of the currently serving network, lacking the ability to continuously and synchronously monitor other potentially available networks in the vicinity. Even if a device is camped on a particular network, the signal quality of other operators' networks may be better, but the device is unaware of this. Because signal acquisition is asynchronous and isolated, it is impossible to obtain a snapshot of the signal of all networks at the same point in time. Therefore, network quality comparisons are subject to time lags and cannot truly and fairly reflect the actual performance of each network at the same time and location. This limits the accuracy and timeliness of the device's real-time, optimal network switching decisions. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a real-time signal strength comparison method for multi-network integrated IoT cards.

[0005] To achieve the above objectives, the present invention employs the following technical solution: a real-time signal strength comparison method for multi-network integrated IoT cards, comprising: After the multi-network integrated IoT card is powered on or the network is switched, it simultaneously initiates network attach requests to all the host networks, including multiple operator networks and multiple network standards. Within a preset signal acquisition period, real-time signal data streams for all the registered networks that have responded and attached are simultaneously acquired from the baseband chip and radio frequency circuit of the multi-network integrated IoT card. Create and maintain a dynamic signal data table. After each signal acquisition cycle, convert the acquired real-time signal data stream into a structured network signal record, update the network signal record to the dynamic signal data table, and delete historical records that exceed the preset valid duration. Based on the real-time signal data stream of the current acquisition period and the recent network signal records stored in the dynamic signal data table, a multi-dimensional real-time scoring model is invoked to generate a dynamic signal quality score for the current period for each attached residential network. All attached residing networks are sorted according to their dynamic signal quality scores to generate a real-time network quality ranking list.

[0006] As a further aspect of the present invention, the simultaneous initiation of network attach requests to all residing networks in parallel includes: Read the operator network configuration information and network frequency band information pre-set in the multi-network integrated IoT card; Based on the operator network configuration information and network frequency band information, a unique network identifier and initial access parameters are generated for each available network; The network attach requests carrying different network identifiers are sent synchronously via the radio frequency channel, and an independent response timer is started for each request; Receive the attachment responses returned by each stationary network, record their response delays, and mark the stationary networks that successfully respond as attached networks and include them in the subsequent signal acquisition range.

[0007] As a further aspect of the present invention, the simultaneous acquisition of real-time signal data streams for all the registered networks that have responded and attached from the baseband chip and radio frequency circuit of the multi-network integrated IoT card includes: The real-time signal data stream includes received signal strength indication values, signal-to-noise ratio values, and signal bit error rate data; Configure the multi-channel signal demodulation unit in the baseband chip to simultaneously lock onto the communication channels of all attached networks; At the beginning of each signal acquisition cycle, the signal measurement units corresponding to each communication channel are synchronously triggered; Read the received signal strength indication value and the signal-to-noise ratio value under the synchronization timestamp from each signal measurement unit; While the baseband chip is decoding data, the signal bit error rate data is extracted from the decoder feedback loop, and the signal bit error rate data is bound to the corresponding network identifier code.

[0008] As a further aspect of the present invention, the creation and maintenance of the dynamic signal data table includes: A circular buffer is allocated in the memory as the physical storage space for the dynamic signal data table; Define the data structure of the network signal record, the data structure includes a network identifier field, a timestamp field, a received signal strength indication value field, a signal-to-noise ratio value field, and a signal error rate data field; During each update, the new network signal record is written to the current write pointer position of the circular buffer, and the write pointer is moved forward. If the write pointer reaches the end of the buffer, it wraps back to the beginning. A separate background cleanup thread periodically scans the dynamic signal data table and deletes historical records that exceed the preset valid duration based on the difference between the timestamp field in the record and the current time.

[0009] As a further aspect of the present invention, the invocation of the multi-dimensional real-time scoring model includes: Extract historical network signal records of the target network over the most recent consecutive periods from the dynamic signal data table; Weighting coefficients are configured for the three dimensions: the received signal strength indication value, the signal-to-noise ratio value, and the signal bit error rate data. Using the real-time component calculation function in the multi-dimensional real-time scoring model, the real-time signal scoring component is calculated based on the real-time signal data stream collected in the current period; Using the stability component calculation function in the multi-dimensional real-time scoring model, the signal stability score component is calculated based on the numerical change trend of each dimension in the extracted historical network signal records. The real-time signal score component and the signal stability score component are combined according to a preset ratio to obtain the dynamic signal quality score. The construction steps of the multi-dimensional real-time scoring model include: Define model input parameters, which include the received signal strength indication value, signal-to-noise ratio value, and signal bit error rate data in the real-time signal data stream, as well as historical network signal records extracted from the dynamic signal data table; The computational framework of the multi-dimensional real-time scoring model is determined, and the computational framework includes the real-time component calculation function and the stable component calculation function; Configure threshold parameters and weight parameters for the real-time component calculation function. The threshold parameters include ideal threshold ranges and acceptable threshold ranges set for the received signal strength indication value, the signal-to-noise ratio value, and the signal bit error rate data, respectively. The weight parameters include weight coefficients used in calculating the real-time signal scoring component for the three dimensions of the received signal strength indication value, the signal-to-noise ratio value, and the signal bit error rate data. Configure stability evaluation parameters for the stable component calculation function, the stability evaluation parameters including the type of statistical index used to calculate the volatility of historical signals and the mapping function rules that map the volatility to stability coefficients; The synthesis ratio of the real-time signal scoring component and the signal stability scoring component is set to complete the construction of the multi-dimensional real-time scoring model.

[0010] As a further aspect of the present invention, the workflow of the real-time component calculation function includes: Set the ideal threshold range and the acceptable threshold range for the received signal strength indication value; If the current received signal strength indicator value falls within the ideal threshold range, then it is assigned the highest score in the dimension. If the score falls within the acceptable threshold range but is not ideal, then the dimension score is calculated using linear interpolation. If the score is below the acceptable threshold, then the lowest score in that dimension will be assigned. Similar calculation logic based on threshold range and interpolation is used for the signal-to-noise ratio (SNR) and bit error rate (BER) data to obtain scores for their respective dimensions. The real-time signal score components are obtained by multiplying the scores of the three dimensions by their respective weighting coefficients and then summing them.

[0011] As a further aspect of the present invention, the workflow of the stable component calculation function includes: For each dimension of the received signal strength indication value, signal-to-noise ratio value, and signal bit error rate data, calculate the variance or standard deviation of its numerical sequence over the most recent consecutive periods. The variance or standard deviation of each dimension is input into a preset monotonically decreasing mapping function, which maps the degree of fluctuation to a stability coefficient between zero and one. The stability coefficients obtained from the three dimensions are weighted and averaged, with the weight configuration consistent with the weight coefficients in the real-time component calculation, to obtain a comprehensive stability coefficient. The signal stability score component is obtained by multiplying a fixed baseline score by the comprehensive stability coefficient.

[0012] As a further aspect of the present invention, the generation of the real-time network quality ranking list includes: Sort the dynamic signal quality scores calculated from all attached networks in descending order; The sorted list is traversed to check the dynamic signal quality score difference between adjacent networks; If the score difference between adjacent networks is less than the preset small difference threshold, the ranking order of the two networks will be slightly adjusted according to their historical average handover success rate, with the network with the higher historical average handover success rate ranked first. The final real-time network quality ranking list is generated, which includes network identifiers, dynamic signal quality scores, and recommended network handover priority tags.

[0013] As a further aspect of the present invention, the method also includes network handover decision triggering based on a dynamic signal data table: Continuously monitor the difference between the network ranked first in the real-time network quality ranking list and the current service network; If the current service network is not the top-ranked network, then query the signal quality trend of the current service network and the top-ranked network over a period of time from the dynamic signal data table. When the signal quality trend of the top-ranked network is consistently better than that of the current serving network, and the duration of this advantage exceeds a preset stability judgment time threshold, the network switching decision process is triggered.

[0014] As a further aspect of the present invention, the method also includes adaptive adjustment of the parameters of the multi-dimensional real-time scoring model: Record the actual communication quality performance after each network switch, including data transmission rate and number of connection interruptions; The actual communication quality performance is compared with the prediction made based on the dynamic signal quality score before the handover, and the prediction deviation is calculated. When the prediction deviation continues to exceed the allowable range, the parameter adjustment program is initiated to perform small-scale iterative optimization on the weight coefficients of each dimension in the multi-dimensional real-time scoring model and the mapping function parameters in the stable component calculation function, so that the model's prediction output is closer to the actual communication quality performance.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: After power-on or network switching, attach requests are initiated in parallel to all carriers and network standards, replacing the traditional sequential attach process. This method allows the device to establish initial connections with multiple networks almost simultaneously, changing the registration process of all potential networks from serial to parallel. This shortens the total time from power-on to obtaining a complete list of available networks, creating the prerequisite for subsequent real-time signal comparison and avoiding the long wait caused by sequential attempts.

[0016] By coordinating the baseband chip and RF circuitry, real-time signal data streams from all successfully attached networks are acquired concurrently within the same acquisition cycle. This technical solution obtains a synchronous signal parameter set covering multiple operators and standards, with all data sharing the same timestamp and physical location background. It eliminates the time skew caused by traditional asynchronous acquisition, enabling signal quality assessments of different networks to be based on a completely consistent environmental benchmark, resulting in highly real-time and comparable signal quality scores.

[0017] Dynamic scoring is performed based on synchronously collected multi-dimensional real-time data and recent historical records, and a real-time network ranking is generated according to the scores. This process can reflect the current status and short-term stability trends of each network in real time and accurately. Based on this ranking list, terminal devices or network management systems can quickly switch to the optimal network or perform load balancing, thereby improving the overall online rate, communication quality, and connection reliability of IoT devices. Attached Figure Description

[0018] Figure 1 This is a flowchart of the real-time signal strength comparison method for multi-network integrated IoT cards described in this invention; Figure 2 A flowchart for initiating network attach requests in parallel; Figure 3 A flowchart for creating and maintaining dynamic signal data tables. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1After powering on or switching networks, the multi-network integrated IoT SIM card simultaneously initiates network attach requests to all its host networks in parallel. These host networks include multiple operator networks and various network standards. Within a preset signal acquisition cycle, it simultaneously acquires real-time signal data streams for all responding host networks from the baseband chip and radio frequency circuit of the multi-network integrated IoT SIM card. It creates and maintains a dynamic signal data table. At the end of each signal acquisition cycle, it converts the acquired real-time signal data streams into structured network signal records and updates the network signal records to the dynamic signal data table, while deleting historical records that exceed the preset valid duration. Based on the real-time signal data stream of the current acquisition cycle and the recent network signal records stored in the dynamic signal data table, it calls a multi-dimensional real-time scoring model to generate a dynamic signal quality score for each attached host network for the current cycle. Finally, it sorts all attached host networks according to their dynamic signal quality scores to generate a real-time network quality ranking list.

[0022] See Figure 2 In one embodiment of the present invention, the multi-network integrated IoT card is pre-loaded with operator network configuration information and standard network frequency band information. After this information is read, a unique network identifier and initial access parameters for each available network are generated based on the operator network configuration information and standard network frequency band information. Network attachment requests carrying different network identifiers are synchronously sent through the radio frequency channel, and an independent response timer is started for each request. The attachment responses returned by each camped network are received, their response delays are recorded, and the camped networks that respond successfully are marked as attached networks and included in the subsequent signal acquisition range. The real-time signal data stream includes received signal strength indication value, signal-to-noise ratio value, and signal error rate data. The multi-channel signal demodulation unit in the baseband chip is configured to simultaneously lock the communication channels of all attached networks, and synchronously trigger the signal measurement unit corresponding to each communication channel at the beginning of each signal acquisition cycle. The received signal strength indication value and signal-to-noise ratio value under the synchronization timestamp are read from each signal measurement unit. While the baseband chip is decoding the data, the signal error rate data is extracted from the decoder feedback loop, and the signal error rate data is bound to the corresponding network identifier.

[0023] In practical implementation, after the multi-network integrated IoT card is powered on, a parallel network attach process is executed. The multi-network integrated IoT card reads pre-set operator network configuration information and network standard frequency band information from non-volatile memory. The operator network configuration information includes the Access Point Names (APNs) and authentication parameters of the three major domestic mobile network operators. The network standard frequency band information covers the GSM 900MHz band for 2G networks, the LTE Band3 and Band8 bands for 4G networks, and the NRn1 and n28 bands for 5G networks. Based on the operator network configuration information and network standard frequency band information... The system generates a unique network identifier for each combination of operator and standard. The unique network identifier is composed of operator code, standard code and frequency band code. At the same time, it generates corresponding initial access parameters according to the standard protocol, such as paging channel configuration and random access preamble format. The system sends network attach request messages carrying the corresponding unique network identifier to all identified available networks through multiple subcarriers of the radio frequency channel. For each network attach request sent, the baseband processing unit starts an independent response timer to record the time interval from the time the request is sent to the time the response is received. In some embodiments, the multi-network integrated IoT SIM card receives attach response messages from each hosted network. The attach response message contains an indication of whether the network accepts or rejects the attach. The baseband processing unit records the response timer reading when each attach response message arrives as the response delay. For a hosted network that responds successfully, the multi-network integrated IoT SIM card records its unique network identifier in the attached network list and marks the hosted network as an attached network. Hosted networks marked as attached networks are included in the scope of subsequent signal acquisition. For a specific example scenario, assuming the multi-network integrated IoT SIM card is pre-configured with operators A, B, and C, after parallel attaching, if operator A's 4G network, operator B's 2G and 4G networks, and operator C's 5G network return a successful response, the attached network list will contain these four entries, with their corresponding unique network identifiers being "A_LTE_B3", "B_GSM_900", "B_LTE_B8", and "C_NR_n28", respectively. The multi-network integrated IoT SIM card will no longer initiate subsequent signal acquisition actions for hosted networks that do not respond or whose responses fail.

[0024] In practical implementation, the multi-network integrated IoT card enters the signal acquisition stage. The multi-channel signal demodulation unit in the baseband chip is configured so that its multiple parallel processing channels lock the communication channels of all attached networks. The communication channels are allocated by the attached networks in the attachment response. At the beginning of a signal acquisition cycle, the main controller of the baseband chip sends a global synchronization trigger signal to all signal measurement units, triggering the signal measurement units to perform real-time signal measurement on their respective locked communication channels. The signal measurement units read the received signal strength indicator value and signal-to-noise ratio value under the synchronization timestamp. The received signal strength indicator value is in dBm and the signal-to-noise ratio value is in dB. While the baseband chip decodes the received data frame, it extracts the signal bit error rate data from the forward error correction feedback loop of the decoder. The signal bit error rate data is expressed as a percentage of the bit error rate. The received signal strength indication (RSI) and signal-to-noise ratio (SNR) values ​​read from each signal measurement unit, along with the bit error rate (BER) data extracted from the decoder feedback loop, together constitute the real-time signal data stream. Each data item in the real-time signal data stream is bound to a unique network identifier corresponding to the communication channel that generated it at the time of generation. This binding operation ensures the clear attribution of data in subsequent processing. For example, for a network with the unique network identifier "B_LTE_B8", a sample point of its real-time signal data stream may contain: an RSI of -85dBm, an SNR of 15dB, and a BER of 0.01%. These three values ​​form a data packet, and its metadata clearly marks the source network "B_LTE_B8". In a data comparison example, a data packet from network "A_LTE_B3" at the same time may contain an RSI of -78dBm, an SNR of 20dB, and a BER of 0.005%. Real-time signal data streams from different networks are output in parallel after timestamp alignment for use in the subsequent creation and updating of dynamic signal data tables. To optimize resource utilization and select networks with rapid responses, after recording response latency, the multi-network integrated IoT SIM card performs an initial screening of successfully attached networks based on a screening threshold. The screening latency threshold is the product of the standard deviation of the response latency and a configurable scaling factor, plus the arithmetic mean of the response latency of all successfully attached networks in this round. When the response latency of an attached network is less than or equal to the screening latency threshold, the network is retained for subsequent in-depth signal acquisition and scoring. When the response latency of an attached network is greater than the screening latency threshold, the multi-network integrated IoT SIM card can release some of the baseband processing resources reserved for that network, maintaining only minimal monitoring of it. This step helps to prioritize the accuracy of signal quality assessment for faster-responding networks when the processing capacity of the multi-network integrated IoT SIM card is limited.

[0025] See Figure 3In one embodiment of the present invention, a circular buffer is allocated in the memory as the physical storage space for the dynamic signal data table. A data structure for network signal records is defined, which includes a network identifier field, a timestamp field, a received signal strength indicator field, a signal-to-noise ratio field, and a signal error rate field. During each update, a new network signal record is written to the current write pointer position of the circular buffer, and the write pointer is moved forward. If the write pointer reaches the end of the buffer, it wraps back to the beginning. An independent background cleanup thread periodically scans the dynamic signal data table and deletes historical records that exceed a preset valid duration based on the difference between the timestamp field in the record and the current time. When calling the multi-dimensional real-time scoring model, historical network signal records of the target network in the most recent consecutive periods are extracted from the dynamic signal data table; weight coefficients are configured for the three dimensions of received signal strength indication value, signal-to-noise ratio value, and signal bit error rate data; the real-time component calculation function in the multi-dimensional real-time scoring model is used to calculate the real-time signal score component based on the real-time signal data stream collected in the current period; the stability component calculation function in the multi-dimensional real-time scoring model is used to calculate the signal stability score component based on the numerical change trend of each dimension in the extracted historical network signal records; the real-time signal score component and the signal stability score component are synthesized according to a preset ratio to obtain the dynamic signal quality score. The construction steps of the multi-dimensional real-time scoring model include: defining model input parameters, which include received signal strength indication (RSI), signal-to-noise ratio (SNR), and bit error rate (BER) data from the real-time signal data stream, as well as historical network signal records extracted from the dynamic signal data table; determining the computational framework of the multi-dimensional real-time scoring model, which includes real-time component calculation functions and stable component calculation functions; configuring threshold and weight parameters for the real-time component calculation functions, where the threshold parameters include ideal and acceptable threshold ranges for RSI, SNR, and BER data respectively, and the weight parameters include weight coefficients used for calculating the real-time signal scoring components in the three dimensions; configuring stability evaluation parameters for the stable component calculation functions, where the stability evaluation parameters include the type of statistical index used to calculate historical signal volatility and the mapping function rule for mapping volatility to stability coefficients; and setting the synthesis ratio of real-time signal scoring components and signal stability scoring components to complete the construction of the multi-dimensional real-time scoring model.

[0026] In practice, the creation of the dynamic signal data table is completed during the system initialization phase. A fixed-size contiguous storage area is allocated in the random access memory of the multi-network integrated IoT card as a circular buffer. The capacity of the circular buffer is preset to store N network signal records. The data structure of each network signal record is clearly defined, and the data structure includes the following fields: network identifier field, used to store a string that uniquely identifies the source network; timestamp field, used to record the precise time when the signal data acquisition was completed; received signal strength indication value field, used to store the measurement value in dBm; signal-to-noise ratio value field, used to store the measurement value in dB; and signal error rate data field, used to store the bit error rate expressed as a percentage. During initialization, the write pointer of the circular buffer points to the starting address, and a background cleanup thread independent of the signal acquisition process is created and started. The background cleanup thread is woken up at fixed time intervals to perform cleanup tasks.

[0027] In some embodiments, after each signal acquisition cycle, the multi-network integrated IoT card converts the acquired real-time signal data stream into structured network signal records. The conversion process includes attaching a high-precision timestamp to each set of real-time signal data from the attached network and encapsulating the data according to the aforementioned data structure. The encapsulated network signal records are written to the storage location pointed to by the current write pointer in the circular buffer. After the write operation is completed, the write pointer moves forward by an address offset equal to the size of one network signal record. When the write pointer moves to the end address of the circular buffer, it wraps back to the beginning address of the circular buffer to continue writing. New network signal records will overwrite the oldest historical records. The background cleanup thread periodically scans all valid network signal records in the circular buffer, reads the timestamp field of each record and compares it with the current system time, calculates the time difference, and if the difference between the timestamp of a network signal record and the current time is greater than a preset valid duration, such as 300 seconds, the network signal record is marked as invalid, and the storage space it occupies can be overwritten by subsequent new records.

[0028] In practice, a multi-dimensional real-time scoring model is invoked to generate a dynamic signal quality score for each attached resident network. This invocation occurs after each signal acquisition cycle ends and the dynamic signal data table is updated. For the target network, the scoring model first extracts the historical network signal records of the network in the most recent M consecutive cycles from the dynamic signal data table, where M is a preset integer. The scoring model configures weight coefficients for the three dimensions of received signal strength indication (RSI), signal-to-noise ratio (SNR), and bit error rate (BER). The weight coefficients are additive and equal to 1. For example, the weight coefficient for RSI is 0.4, the weight coefficient for SNR is 0.35, and the weight coefficient for BER is 0.25. The multi-dimensional real-time scoring model includes real-time component calculation functions and stable component calculation functions. The real-time component calculation function calculates based on the real-time signal data stream acquired in the current period. The function inputs are the received signal strength indicator value, signal-to-noise ratio value, and signal error rate data for the current period. The stable component calculation function calculates based on the historical network signal records of the most recent M consecutive periods extracted from the dynamic signal data table. The function analyzes the changing trends of the three dimensions in these historical records. The real-time component calculation function outputs the real-time signal score component, and the stable component calculation function outputs the signal stability score component. The real-time signal score component and the signal stability score component are combined according to a preset ratio, such as a 7:3 ratio. The combined result is the dynamic signal quality score of the network in the current period. The dynamic signal quality score is a normalized value.

[0029] In practical implementation, the construction steps of the multi-dimensional real-time scoring model are completed offline during the system configuration phase. Model input parameters are defined, including the received signal strength indicator (RSI), signal-to-noise ratio (SNR), and bit error rate (BER) data from the real-time signal data stream, as well as a historical network signal record sequence containing M records extracted from the dynamic signal data table. The computational framework of the multi-dimensional real-time scoring model is determined, consisting of parallel real-time component calculation functions and stable component calculation functions. The output is finally obtained through weighted synthesis. Threshold parameters and weight parameters are configured for the real-time component calculation functions. Threshold parameters include ideal and acceptable threshold ranges for the RSI, SNR, and BER data. The weight parameters are the weighting coefficients used for the three dimensions—RSI, SNR, and BER—when calculating the real-time signal scoring components.

[0030] In some embodiments, stability evaluation parameters are configured for the stable component calculation function. These parameters include the type of statistical indicator used to calculate historical signal volatility, such as specifying the standard deviation as the statistical indicator. The stability evaluation parameters also include a mapping function rule that maps volatility to stability coefficients. This mapping function rule can be a piecewise linear function or a monotonically decreasing exponential function. The synthesis ratio of the real-time signal score component and the signal stability score component is set, for example, the synthesis ratio is set to 'a' and (1-a'), where 'a' is 0.7. After completing the above parameter configuration, the multi-dimensional real-time scoring model is constructed and can be loaded into the running memory of the multi-network integrated IoT card for periodic calls. When synthesizing the real-time signal score component and the signal stability score component, a synthesis formula with a compensation term can be used. This can provide an extra reward when the signal is highly stable. One synthesis weight coefficient is multiplied by the real-time signal score component, and another synthesis weight coefficient is multiplied by the signal stability score component. These two items, along with the stability compensation term, are then added to obtain the final dynamic signal quality score. The sum of the two synthesis weight coefficients is 1, used to balance real-time performance and stability considerations. The stability compensation term... The calculation relies on the comprehensive stability coefficient output by the stable component calculation function (the closer the coefficient is to 1, the more stable the historical signal is). The specific calculation method is as follows: when the comprehensive stability coefficient is greater than a high stability threshold, the stability compensation term takes a small positive compensation value; otherwise, the stability compensation term takes zero. This compensation mechanism aims to give a slight bonus to the score of networks that maintain excellent stability in the long term, so that they are more advantageous in the ranking.

[0031] In one embodiment of the present invention, the workflow of the real-time component calculation function includes: setting an ideal threshold range and a qualified threshold range for the received signal strength indication value; if the current received signal strength indication value falls within the ideal threshold range, it is assigned the highest dimension score; if it falls within the qualified threshold range but is not within the ideal range, the dimension score is calculated according to the linear interpolation method; if it is below the qualified threshold range, it is assigned the lowest dimension score; similar calculation logic based on threshold range and interpolation is applied to the signal-to-noise ratio value and the signal bit error rate data to obtain the scores of their respective dimensions; the scores of the three dimensions are multiplied by the corresponding weight coefficients and then summed to obtain the real-time signal scoring component. The workflow of the stable component calculation function includes: calculating the variance or standard deviation of the numerical sequence of the received signal strength indication value, signal-to-noise ratio value, and signal bit error rate data over the most recent consecutive periods for each dimension; inputting the variance or standard deviation of each dimension into a preset monotonically decreasing mapping function, which maps the degree of fluctuation to a stability coefficient between zero and one; weighting the stability coefficients obtained from the three dimensions, with the weight configuration consistent with the weight coefficients in the real-time component calculation, to obtain a comprehensive stability coefficient; and multiplying a fixed benchmark score by the comprehensive stability coefficient to obtain the signal stability score component.

[0032] In practice, the workflow of the real-time component calculation function begins by setting an ideal threshold range and a qualified threshold range for the received signal strength indication value. The upper and lower limits of the ideal threshold range and the qualified threshold range are pre-stored in a configuration file. For example, the ideal threshold range for the received signal strength indication value is set to -70dBm to -50dBm, and the qualified threshold range is set to -100dBm to -70dBm. The current received signal strength indication value is provided by the signal acquisition stage. If the current received signal strength indication value falls within the ideal threshold range, it is assigned the highest dimension score, which is a fixed full score. If the current received signal strength indication value falls within the qualified threshold range but is not in the ideal range, the dimension score is calculated according to the linear interpolation method. The linear interpolation method maps the current received signal strength indication value to a score between the lowest and highest scores proportionally based on its relative position within the qualified threshold range. If the current received signal strength indication value is below the qualified threshold range, it is assigned the lowest dimension score, which is a fixed base score.

[0033] In some embodiments, similar threshold-range and interpolation-based calculation logic is used for both signal-to-noise ratio (SNR) values ​​and bit error rate (BER) data. The ideal threshold range for SNR values ​​is set above 20 dB, and the acceptable threshold range is set to 10 dB to 20 dB. For BER data, the ideal threshold range is set to 0% to 0.001%, and the acceptable threshold range is set to 0.001% to 0.1%. For SNR values, if the current value falls within the ideal threshold range, it is assigned the highest score in its dimension; if it falls within the acceptable threshold range, the score is calculated using linear interpolation; and if it is below the acceptable threshold, it is assigned the lowest score in its dimension. For BER data, since a smaller value indicates better quality, the calculation logic needs to be reversed. If the current signal bit error rate falls within the ideal threshold range (low bit error rate), it is assigned the highest score in that dimension. If it falls within the acceptable threshold range (higher bit error rate), the score is calculated using linear interpolation, but the interpolation direction is opposite to that of the received signal strength indicator and signal-to-noise ratio (SNR) values. That is, the higher the bit error rate, the lower the score. If the current signal bit error rate is higher than the acceptable threshold range, it is assigned the lowest score in that dimension. After obtaining the scores for the received signal strength indicator, SNR, and bit error rate, the scores for the three dimensions are multiplied by their corresponding weighting coefficients and then summed to obtain the real-time signal scoring components. The weighting coefficients are set during model construction, for example, 0.4, 0.35, and 0.25.

[0034] In practical implementation, after the stable component calculation function starts its workflow, it calculates the variance or standard deviation of the numerical sequence over the most recent consecutive periods for each dimension of the received signal strength indication value, signal-to-noise ratio value, and bit error rate data. The most recent consecutive periods are determined by a preset parameter M. The historical network signal records of the target network over the most recent M periods are extracted from the dynamic signal data table. The received signal strength indication value, signal-to-noise ratio value, and bit error rate data in each record are parsed to form three numerical sequences of length M. The variance or standard deviation calculation formula is applied to each numerical sequence to obtain three scalar values ​​representing the volatility of the received signal strength indication value, the volatility of the signal-to-noise ratio value, and the volatility of the bit error rate data, respectively. The variance or standard deviation of each dimension is input into a pre-defined monotonically decreasing mapping function. The input to the mapping function is a volatility value, and the output is a stability coefficient between zero and one. The stability coefficient decreases as the input volatility value increases. The specific form of the mapping function can be a piecewise linear function or a negative exponential function, and its parameters are defined during the model building stage. The mapping function maps the degree of volatility to a stability coefficient between zero and one. For example, if the standard deviation of the received signal strength indicator numerical sequence is 3dB, the stability coefficient calculated by the mapping function is 0.8; if the standard deviation of the signal-to-noise ratio numerical sequence is 2dB, the stability coefficient is 0.9; and if the standard deviation of the signal bit error rate data sequence is 0.02%, the stability coefficient is 0.7. In some embodiments, the stability coefficients obtained from the three dimensions are weighted and averaged. The weight configuration is consistent with the weight coefficients in the real-time component calculation, that is, the weights corresponding to the received signal strength indication value, signal-to-noise ratio value, and signal bit error rate data are 0.4, 0.35, and 0.25, respectively. The weighted average is used to calculate a comprehensive stability coefficient, which reflects the overall stability of the target network signal quality in multiple dimensions. A fixed benchmark score is multiplied by the comprehensive stability coefficient to obtain the signal stability score component. The fixed benchmark score is a predefined constant used to convert the stability coefficient into a score that matches the dimensions of the real-time signal score component.The monotonically decreasing mapping function used in the stable component calculation function can take a smooth nonlinear form to more finely distinguish different levels of volatility. The scaling factor is multiplied by the volatility value, and the result is taken as the exponent of the natural constant e, yielding the stability coefficient. Here, the stability coefficient ranges from 0 to 1; the volatility value is the standard deviation of a historical sequence in a certain dimension (received signal strength indicator, signal-to-noise ratio, or bit error rate); the scaling factor controls the rate at which the stability coefficient decreases as the volatility value increases. When the volatility value is 0, the stability coefficient is 1, representing complete stability. As the volatility value increases, the stability coefficient decays towards 0. This form can produce a clear coefficient distinction for low to medium levels of volatility, while mapping extremely high volatility uniformly to a coefficient value close to zero.

[0035] In one embodiment of the present invention, when generating a real-time network quality ranking list, the dynamic signal quality scores calculated for all attached networks are sorted in descending order. The sorted list is then traversed to check the difference in dynamic signal quality scores between adjacent networks. If the score difference between adjacent networks is less than a preset small difference threshold, the ranking order of the two networks is fine-tuned based on their historical average handover success rates, placing the network with the higher historical average handover success rate first. A final real-time network quality ranking list is generated, which includes network identifiers, dynamic signal quality scores, and recommended network handover priority markers. The network handover decision triggering based on the dynamic signal data table includes: continuously monitoring the difference between the top-ranked network in the real-time network quality ranking list and the current serving network; if the current serving network is not the top-ranked network, querying the signal quality trends of the current serving network and the top-ranked network in the dynamic signal data table over a period of time; when the signal quality trend of the top-ranked network is consistently better than that of the current serving network, and the duration of the advantage exceeds a preset stability judgment duration threshold, the network handover decision process is triggered.

[0036] In practice, the process of generating a real-time network quality ranking list is automatically executed after each round of signal acquisition and scoring. The dynamic signal quality scores calculated for all attached networks are sorted in descending order. The sorting algorithm is based on simple numerical comparison, placing the network with the highest dynamic signal quality score at the top of the list and the network with the lowest dynamic signal quality score at the bottom, forming a preliminary sorted list. The sorted list is then traversed, and the difference in dynamic signal quality scores between each pair of adjacent networks is checked. The difference in scores between adjacent networks is calculated by subtracting the dynamic signal quality score of the next network from the dynamic signal quality score of the previous network.

[0037] In some embodiments, the preset minor difference threshold is a fixed value, such as 0.5 points. If the score difference between adjacent networks is less than the preset minor difference threshold, their ranking order is fine-tuned based on the historical average handover success rate of the two networks. The historical average handover success rate is stored in the persistent memory of the multi-network integrated IoT card, recording the ratio of the historical successful handovers from the current network to the total number of attempts. The fine-tuning rule is to rank the network with the higher historical average handover success rate first, even if its dynamic signal quality score is slightly lower. If the score difference is greater than or equal to the preset minor difference threshold, the original sorting order based on the dynamic signal quality score is maintained. After the traversal and fine-tuning process is completed, the final real-time network quality ranking list is generated. The real-time network quality ranking list is a data structure containing network identifiers, dynamic signal quality scores, and recommended network handover priority markers. The network handover priority marker can be a simple integer sequence number, where 1 represents the highest recommended handover priority, and the larger the sequence number, the lower the priority. In a specific example scenario, after sorting and possible fine-tuning, the generated real-time network quality ranking list may be stored in memory in tabular form, as shown in Table 1.

[0038] Table 1: Real-time Network Quality Ranking List In practice, the network handover decision triggering based on the dynamic signal data table is an independently running monitoring process. This process continuously monitors the difference between the top-ranked network in the real-time network quality ranking list and the current service network. The current service network is the network on which the multi-network converged IoT card is currently transmitting business data. If the current service network is the top-ranked network, no handover decision is triggered. If the current service network is not the top-ranked network, the signal quality trend of the current service network and the top-ranked network over a period of time is queried from the dynamic signal data table.

[0039] In some embodiments, the past period is a configurable duration window, such as the most recent 60 seconds. Querying the signal quality trend requires extracting all historical network signal records of the current service network and the top-ranked network within this duration window from the dynamic signal data table. The average value or change curve of each signal dimension of the two networks within this window is calculated. The dominance duration refers to the cumulative time during which the dynamic signal quality score of the top-ranked network continuously exceeds the dynamic signal quality score of the current service network. The stability judgment duration threshold is a preset value, such as 15 seconds. When the signal quality trend of the top-ranked network is consistently better than that of the current service network, and the dominance duration exceeds the preset stability judgment duration threshold, the network switching decision process is triggered. The network switching decision process includes a switching feasibility assessment and a switching command execution sequence. To handle network ranking boundary situations more smoothly, a switching propensity index can be introduced for calculation before triggering a network switching decision. The switching propensity index is jointly determined by the dynamic signal quality score of the top-ranked network, the dynamic signal quality score of the current serving network, and the duration of the difference between the two scores. Specifically, the calculation of the switching propensity index is to first adjust the difference between the dynamic signal quality scores of the top-ranked network and the current serving network by a score normalization factor, and then combine it with the hyperbolic tangent function value obtained by normalizing the duration of the advantage by a time constant to obtain the final result.

[0040] In one embodiment of the present invention, the adaptive adjustment of the parameters of the multi-dimensional real-time scoring model includes: recording the actual communication quality performance after each network handover, the actual communication quality performance including data transmission rate and number of connection interruptions; comparing the actual communication quality performance with the prediction made before the handover based on the dynamic signal quality score, and calculating the prediction deviation; when the prediction deviation continues to exceed the allowable range, initiating a parameter adjustment program to perform small-scale iterative optimization of the weight coefficients of each dimension in the multi-dimensional real-time scoring model and the mapping function parameters in the stable component calculation function, so as to make the model's prediction output closer to the actual communication quality performance. In specific implementation, the adaptive adjustment process of the parameters of the multi-dimensional real-time scoring model is initiated after the network handover event is completed, recording the actual communication quality performance after each network handover, the actual communication quality performance including two specific indicators: data transmission rate and number of connection interruptions. The data transmission rate is obtained by dividing the number of bytes successfully received by the application layer within a fixed time window after a successful handover by the length of the time window, in kilobits per second. The number of connection interruptions is accumulated by monitoring the abnormal disconnection indication in the link layer protocol or the application layer heartbeat timeout event. The record of the actual communication quality performance is associated with and stored with the source network identifier and the target network identifier of the handover.

[0041] In some embodiments, the actual communication quality performance is compared with the prediction made before the handover based on the dynamic signal quality score. The predicted communication quality performance is calculated by a preset mapping function based on the dynamic signal quality score of the target network before the handover. The mapping function maps the dynamic signal quality score to the expected data transmission rate range and the connection interruption probability. The prediction deviation is calculated, which includes the data transmission rate deviation and the connection interruption number deviation. The data transmission rate deviation is the relative difference between the predicted data transmission rate and the actual data transmission rate. The connection interruption number deviation is a Boolean comparison or probability difference between the predicted connection interruption probability and the actual number of connection interruptions. The combined prediction deviation is obtained by weighting and combining these two sub-deviations to obtain a scalar value.

[0042] In practice, the system monitors whether the overall prediction deviation continuously exceeds the allowable range. The allowable range is a preset deviation threshold interval, such as [-0.1, 0.1]. Continuously exceeding the allowable range means that in multiple consecutive network handover events, the overall prediction deviation falls outside the allowable range. When the prediction deviation continuously exceeds the allowable range, the system starts the parameter adjustment program. The parameter adjustment program first locks the set of parameters that need to be optimized. The parameter set includes the weight coefficients of each dimension in the multi-dimensional real-time scoring model, namely the weight coefficients of the received signal strength indicator, the signal-to-noise ratio, and the bit error rate, as well as the mapping function parameters in the stable component calculation function, such as the scaling factor or segmentation point of the mapping function. The parameter adjustment procedure employs a small-amplitude iterative optimization method. Iterative optimization is based on the deviation sequence between actual and predicted communication quality performance recorded in historical handover events. For weight coefficients, the adjustment direction aims to reduce future prediction deviations. The adjustment process follows a constraint: all weight coefficients must sum to 1 after adjustment, and each coefficient must remain within reasonable upper and lower limits. For the mapping function parameters in the stable component calculation function, the adjustment is based on the correlation between the signal stability score component and actual connection stability in historical data. In an example scenario, if historical data indicates that a network with a high signal stability score component frequently experiences connection interruptions, it may be necessary to adjust the mapping function parameters to make them more sensitive to volatility, thereby reducing overly optimistic stability scores. The iterative optimization process can use an update rule based on gradient descent, but the step size is set very small to ensure stability. Update rule: The adjusted parameter vector is obtained by multiplying the iteration step size, the sign function, and the parameter preset adjustment increment vector, and then adding the result to the parameter vector before adjustment.

[0043] The sign function is applied to the arithmetic mean of the comprehensive prediction deviations of the most recent switching events, and is used to determine the direction of parameter adjustment; the iteration step size is a positive number much less than 1, and is used to control the magnitude of parameter adjustment; the parameter preset adjustment increment vector defines the maximum basic unit that each parameter can be changed in a single iteration.

[0044] The parameter tuning procedure can incorporate a cooling mechanism. This mechanism gradually reduces the iteration step size as the number of tuning iterations increases, preventing over-tuning and oscillations in the later stages of optimization. The cooling mechanism uses a decay factor applied to the iteration step size; for example, after each successful parameter tuning, the iteration step size is multiplied by a decay factor slightly less than 1. This decay factor ensures that the optimization process eventually converges to a stable set of parameters. In comparing actual communication quality performance data, a log of historical parameter versions and their corresponding average prediction deviations can be maintained. The log records the parameter values ​​before each tuning, the adjusted parameter values, and the average prediction deviation observed over a subsequent period. Data comparison is used to verify the effectiveness of parameter tuning. If the average prediction deviation significantly decreases after tuning and regresses to an acceptable range, the tuning is confirmed as effective optimization. If the deviation does not improve or worsens, a rollback operation may be triggered, restoring the parameters to the previous version and attempting another tuning strategy, such as modifying the direction or size of the tuning increment vector.

[0045] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for real-time comparison of signal strength of multi-network integrated IoT cards, characterized in that, The method includes: After the multi-network integrated IoT card is powered on or the network is switched, it simultaneously initiates network attach requests to all the host networks, including multiple operator networks and multiple network standards. Within a preset signal acquisition period, real-time signal data streams for all the registered networks that have responded and attached are simultaneously acquired from the baseband chip and radio frequency circuit of the multi-network integrated IoT card. Create and maintain a dynamic signal data table. After each signal acquisition cycle, convert the acquired real-time signal data stream into a structured network signal record, update the network signal record to the dynamic signal data table, and delete historical records that exceed the preset valid duration. Based on the real-time signal data stream of the current acquisition period and the recent network signal records stored in the dynamic signal data table, a multi-dimensional real-time scoring model is invoked to generate a dynamic signal quality score for the current period for each attached residential network. All attached residing networks are sorted according to their dynamic signal quality scores to generate a real-time network quality ranking list.

2. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 1, characterized in that, The simultaneous initiation of network attach requests to all residing networks in parallel includes: Read the operator network configuration information and network frequency band information pre-set in the multi-network integrated IoT card; Based on the operator network configuration information and network frequency band information, a unique network identifier and initial access parameters are generated for each available network; The network attach requests carrying different network identifiers are sent synchronously via the radio frequency channel, and an independent response timer is started for each request; Receive the attachment responses returned by each stationary network, record their response delays, and mark the stationary networks that successfully respond as attached networks and include them in the subsequent signal acquisition range.

3. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 1, characterized in that, The simultaneous acquisition of real-time signal data streams for all the registered networks that have responded and attached from the baseband chip and radio frequency circuit of the multi-network integrated IoT card includes: The real-time signal data stream includes received signal strength indication values, signal-to-noise ratio values, and signal bit error rate data; Configure the multi-channel signal demodulation unit in the baseband chip to simultaneously lock onto the communication channels of all attached networks; At the beginning of each signal acquisition cycle, the signal measurement units corresponding to each communication channel are synchronously triggered; Read the received signal strength indication value and the signal-to-noise ratio value under the synchronization timestamp from each signal measurement unit; While the baseband chip is decoding data, the signal bit error rate data is extracted from the decoder feedback loop, and the signal bit error rate data is bound to the corresponding network identifier code.

4. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 1, characterized in that, The creation and maintenance of the dynamic signal data table includes: A circular buffer is allocated in the memory as the physical storage space for the dynamic signal data table; Define the data structure of the network signal record, the data structure includes a network identifier field, a timestamp field, a received signal strength indication value field, a signal-to-noise ratio value field, and a signal error rate data field; During each update, the new network signal record is written to the current write pointer position of the circular buffer, and the write pointer is moved forward. If the write pointer reaches the end of the buffer, it wraps back to the beginning. A separate background cleanup thread periodically scans the dynamic signal data table and deletes historical records that exceed the preset valid duration based on the difference between the timestamp field in the record and the current time.

5. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 3, characterized in that, The invocation of the multi-dimensional real-time scoring model includes: Extract historical network signal records of the target network over the most recent consecutive periods from the dynamic signal data table; Weighting coefficients are configured for the three dimensions: the received signal strength indication value, the signal-to-noise ratio value, and the signal bit error rate data. Using the real-time component calculation function in the multi-dimensional real-time scoring model, the real-time signal scoring component is calculated based on the real-time signal data stream collected in the current period; Using the stability component calculation function in the multi-dimensional real-time scoring model, the signal stability score component is calculated based on the numerical change trend of each dimension in the extracted historical network signal records. The real-time signal score component and the signal stability score component are combined according to a preset ratio to obtain the dynamic signal quality score. The construction steps of the multi-dimensional real-time scoring model include: Define model input parameters, which include the received signal strength indication value, signal-to-noise ratio value, and signal bit error rate data in the real-time signal data stream, as well as historical network signal records extracted from the dynamic signal data table; The computational framework of the multi-dimensional real-time scoring model is determined, and the computational framework includes the real-time component calculation function and the stable component calculation function; Configure threshold parameters and weight parameters for the real-time component calculation function. The threshold parameters include ideal threshold ranges and acceptable threshold ranges set for the received signal strength indication value, the signal-to-noise ratio value, and the signal bit error rate data, respectively. The weight parameters include weight coefficients used in calculating the real-time signal scoring component for the three dimensions of the received signal strength indication value, the signal-to-noise ratio value, and the signal bit error rate data. Configure stability evaluation parameters for the stable component calculation function, the stability evaluation parameters including the type of statistical index used to calculate the volatility of historical signals and the mapping function rules that map the volatility to stability coefficients; The synthesis ratio of the real-time signal scoring component and the signal stability scoring component is set to complete the construction of the multi-dimensional real-time scoring model.

6. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 5, characterized in that, The workflow of the real-time component calculation function includes: Set the ideal threshold range and the acceptable threshold range for the received signal strength indication value; If the current received signal strength indicator value falls within the ideal threshold range, then it is assigned the highest score in the dimension. If the score falls within the acceptable threshold range but is not ideal, then the dimension score is calculated using linear interpolation. If the score is below the acceptable threshold, then the lowest score in that dimension will be assigned. Similar calculation logic based on threshold range and interpolation is used for the signal-to-noise ratio (SNR) and bit error rate (BER) data to obtain scores for their respective dimensions. The real-time signal score components are obtained by multiplying the scores of the three dimensions by their respective weighting coefficients and then summing them.

7. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 5, characterized in that, The workflow of the stable component calculation function includes: For each dimension of the received signal strength indication value, signal-to-noise ratio value, and signal bit error rate data, calculate the variance or standard deviation of its numerical sequence over the most recent consecutive periods. The variance or standard deviation of each dimension is input into a preset monotonically decreasing mapping function, which maps the degree of fluctuation to a stability coefficient between zero and one. The stability coefficients obtained from the three dimensions are weighted and averaged, with the weight configuration consistent with the weight coefficients in the real-time component calculation, to obtain a comprehensive stability coefficient. The signal stability score component is obtained by multiplying a fixed baseline score by the comprehensive stability coefficient.

8. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 1, characterized in that, The generated real-time network quality ranking list includes: Sort the dynamic signal quality scores calculated from all attached networks in descending order; The sorted list is traversed to check the dynamic signal quality score difference between adjacent networks; If the score difference between adjacent networks is less than the preset small difference threshold, the ranking order of the two networks will be slightly adjusted according to their historical average handover success rate, with the network with the higher historical average handover success rate ranked first. The final real-time network quality ranking list is generated, which includes network identifiers, dynamic signal quality scores, and recommended network handover priority tags.

9. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 8, characterized in that, The method also includes network handover decision triggering based on dynamic signal data tables: Continuously monitor the difference between the network ranked first in the real-time network quality ranking list and the current service network; If the current service network is not the top-ranked network, then query the signal quality trend of the current service network and the top-ranked network over a period of time from the dynamic signal data table. When the signal quality trend of the top-ranked network is consistently better than that of the current serving network, and the duration of this advantage exceeds a preset stability judgment time threshold, the network switching decision process is triggered.

10. The method for real-time comparison of signal strength of multi-network integrated IoT cards as described in claim 7, characterized in that, The method also includes adaptive adjustment of the parameters of the multi-dimensional real-time scoring model: Record the actual communication quality performance after each network switch, including data transmission rate and number of connection interruptions; The actual communication quality performance is compared with the prediction made based on the dynamic signal quality score before the handover, and the prediction deviation is calculated. When the prediction deviation continues to exceed the allowable range, the parameter adjustment program is initiated to perform small-scale iterative optimization on the weight coefficients of each dimension in the multi-dimensional real-time scoring model and the mapping function parameters in the stable component calculation function, so that the model's prediction output is closer to the actual communication quality performance.

Images