Multi-channel data acquisition terminal and distributed data acquisition synchronization system

Abstract

The present application relates to the technical field of data acquisition and distributed synchronization, and particularly relates to a multi-channel data acquisition terminal and a distributed data acquisition synchronization system, comprising an environment feature extraction module, an event detection and verification module, a consensus synchronization module, a local clock calibration module, a multi-channel data acquisition and time restoration module, and further comprising a sensor delay self-calibration module, an environment feature drift learning module, a terminal-to-terminal communication module and a new node guided access module; the modules are linked and the data flow direction is clear; cross-terminal data interaction and instruction transmission are realized through the terminal-to-terminal communication module, and a closed-loop distributed acquisition synchronization architecture is formed to complete the whole-process acquisition synchronization operation; in the present application, multi-modal environment physical field signal such as geomagnetic field and sound field is acquired to extract feature data, the consistency of environment feature events is verified through a cross-modal correlation model, an effective natural event is taken as a synchronization beacon, a time synchronization mechanism without satellite and wired infrastructure is constructed, and distributed terminal autonomous synchronization in a scene without external trusted sources is realized.

Description

Technical Field

[0001] This invention relates to the field of data acquisition and distributed synchronization technology, and in particular to a multi-channel data acquisition terminal and a distributed data acquisition and synchronization system. Background Technology

[0002] Multi-channel data acquisition terminals and distributed data acquisition synchronization systems are widely used in various scenarios such as outdoor environmental monitoring, underground engineering monitoring, and confined space detection. Their core function is to achieve synchronous data acquisition from multiple distributed terminals across multiple channels, ensuring that data acquired from different terminals and channels has a unified time reference. This guarantees the correlation and validity of the acquired data, providing reliable support for subsequent data processing, analysis, and applications. In practical applications, some acquisition scenarios often lack satellite signal coverage or cannot deploy wired infrastructure. These scenarios, lacking external signal sources, place higher demands on the autonomous synchronization capabilities of distributed terminals. Traditional synchronization methods are no longer suitable for the needs of such scenarios, necessitating a new synchronization mechanism that does not rely on external infrastructure to achieve accurate synchronous acquisition from distributed terminals. Currently, most existing multi-channel data acquisition terminals and distributed data acquisition synchronization systems rely on satellite time synchronization or wired infrastructure to achieve time synchronization between terminals. While these methods can generally meet synchronization requirements in scenarios with satellite signals and available wired infrastructure, they often lead to significant drops in synchronization accuracy and synchronization interruptions in scenarios without satellite signal coverage, wired infrastructure, or external signal sources. This results in a lack of a unified time reference for data collected from different terminals and channels, severely impacting the correlation and effectiveness of the collected data. This fails to meet the actual needs of multi-channel data synchronization in such special scenarios. Currently, there is no mature solution to achieve autonomous and accurate synchronization of distributed terminals without external signal sources. Therefore, there is an urgent need to develop a new type of multi-channel data acquisition terminal and distributed data acquisition synchronization system to solve this core technical challenge. Summary of the Invention

[0003] To overcome the above shortcomings, this invention provides a multi-channel data acquisition terminal and a distributed data acquisition synchronization system, aiming to improve the problem of synchronization anomalies in existing systems that rely on satellite or wired infrastructure and are in scenarios without external credit sources.

[0004] This invention provides the following technical solution: a multi-channel data acquisition terminal and a distributed data acquisition synchronization system, the system comprising: The feature extraction module is used to collect and extract multimodal physical field signals of the environment in which the terminal is located, in order to generate environmental feature data; The event verification module, connected to the feature extraction module, is used to receive environmental feature data, detect environmental feature events based on the environmental feature data, and perform consistency verification on the detected events. The consensus synchronization module, connected to the event verification module, is used to receive verified environmental feature event information, interact and reach consensus with other terminals in the network to achieve system-wide synchronization moment consensus, and generate and maintain a trust chain that records consensus events. The clock calibration module is connected to the consensus synchronization module and is used to calibrate the terminal's local clock according to the consensus of the synchronization time. The time restoration module, connected to the clock calibration module, is used to perform multi-channel data acquisition and generate accurate sensing time timestamps for the acquired data; The self-calibration module, connected to the time restoration module, is used to measure the end-to-end delay of each acquisition channel and provide delay compensation values. The drift learning module connects the feature extraction module and the event verification module, and is used to monitor statistical changes in environmental features and trigger model updates. The terminal communication module is used to provide data interaction support between terminals for the consensus synchronization module and the drift learning module.

[0005] Preferably, the feature extraction module includes: A multimodal sensor array unit is used to simultaneously acquire raw physical field signals, including at least the geomagnetic field, the acoustic field, and the atmospheric pressure field. The feature vector calculation unit is used to analyze the original physical field signal to form an environmental feature vector; The correlation model construction and storage unit is used to analyze the temporal correlation between feature vectors of different modal environments to construct a cross-modal correlation model, and to store environmental feature vector samples and cross-modal correlation models.

[0006] Preferably, the event verification module includes: An event triggering unit is used to trigger potential events based on changes in environmental feature vectors. The consistency verification unit is used to perform multimodal consistency verification on potential events by constructing a cross-modal association model with the storage unit after an event is triggered. The Synchronous Proposal Generation Unit is used to generate synchronous proposals for verified events.

[0007] Preferably, the consensus synchronization module includes: The proposal exchange unit is used to broadcast and receive synchronous proposals via the terminal communication module. The voting decision-making unit is used to verify the received synchronous proposals and form voting opinions based on the verification results; The consensus time calculation unit is used to calculate the final consensus synchronization time based on the local event detection times reported by all supporting terminals after collecting enough positive votes. The Trust Chain Maintenance Unit is used to update the Trust Chain after consensus is reached.

[0008] Preferably, the clock calibration module includes: The deviation calculation unit is used to calculate the clock deviation based on the consensus synchronization time and the locally recorded event detection time; A clock adjustment unit is used to adjust the local clock according to the clock deviation; The credibility update unit is used to update the historical credibility score of this terminal based on the consensus result.

[0009] Preferably, the time restoration module includes: The data acquisition and control unit is used to trigger multi-channel data acquisition; The delay compensation unit is used to store the delay compensation values ​​of each channel obtained from the self-calibration module; The timestamp generation unit is used to generate accurate sensing time timestamps based on the calibrated local time and delay compensation value.

[0010] Preferably, the self-calibration module includes: An embedded calibration source unit is used to generate physical excitation signals at known times; The delay measurement unit is used to measure the total time from excitation generation to signal output. The compensation value update unit is used to update the delay compensation value based on multiple measurement results.

[0011] Preferably, the drift learning module includes: Statistical monitoring unit, used to monitor the statistical distribution of environmental feature vectors; The drift detection unit is used to determine whether feature drift has occurred based on changes in statistical distribution. The model collaborative update unit is used to recalculate the cross-modal correlation model after drift occurs and to collaboratively update it with other terminals through the terminal communication module.

[0012] Preferably, the terminal communication module includes: The self-organizing network management unit is used to manage neighbor discovery and network topology maintenance among terminals. A reliable data transmission unit is used to transmit synchronous proposals, voting opinions, credibility chain update data, and model update information.

[0013] Preferably, the system further includes: a new node guidance access module, which connects the terminal communication module and the consensus synchronization module, to enable newly joined terminals to listen to the neighbor status, obtain the current trust chain copy and feature fingerprint database, and complete initial synchronization and model calibration by participating in the first consensus event.

[0014] The present invention has the following beneficial effects: 1. In this invention, by collecting multimodal environmental physical field signals such as geomagnetic field, acoustic field and atmospheric pressure field and extracting environmental feature data, multimodal consistency verification of detected environmental feature events is performed based on cross-modal correlation model, and the verified valid natural events are used as synchronization beacons, thereby constructing a new time synchronization mechanism that does not rely on satellites and wired infrastructure, realizing distributed terminal autonomous synchronization in scenarios without external credit sources.

[0015] 2. In this invention, a weighted voting decision mechanism that integrates multi-modal event verification scores and terminal historical credibility scores is designed. The consensus synchronization time is calculated by using the weighted median of the local event detection times of all supporting terminals. Finally, a credibility chain that records consensus events is formed and maintained. This mechanism significantly improves the robustness and anti-interference ability of the distributed consensus process and can effectively suppress the impact of abnormal or malicious terminals on synchronization accuracy.

[0016] 3. In this invention, by periodically measuring the end-to-end delay from the generation of physical excitation to the output of digital signal by embedding calibration sources in each terminal, and dynamically updating the delay compensation value, the timestamp generation unit can restore the precise physical sensing time of the data at the sensor end based on the calibrated local clock and the channel-specific delay compensation value, rather than the sampling time, thereby providing a high-precision time reference for the data of multimodal heterogeneous sensors that can achieve physical timing alignment.

[0017] 4. In this invention, by statistically monitoring the long-term distribution changes of environmental feature vectors and determining feature drift, the recalculation of the cross-modal association model and the collaborative update between terminals are triggered. At the same time, newly added terminals can obtain the current trust chain and feature fingerprint database by listening and complete the initialization by participating in the first consensus event. This enables the entire system to have the ability to adapt to the long-term evolution of the environment and to support the dynamic addition and removal of network nodes in a plug-and-play manner, thus ensuring the stability of long-term deployment. Attached Figure Description

[0018] Figure 1 This is a flowchart of the multi-channel data acquisition terminal and distributed data acquisition synchronization system proposed in this invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] In embodiments of the present invention, the present invention provides a multi-channel data acquisition terminal and a distributed data acquisition synchronization system, such as... Figure 1 As shown, the system includes: The feature extraction module is used to collect and extract multimodal physical field signals of the environment in which the terminal is located, in order to generate environmental feature data; Furthermore, the feature extraction module includes: A multimodal sensor array unit is used to simultaneously acquire raw physical field signals, including at least the geomagnetic field, the acoustic field, and the atmospheric pressure field. The feature vector calculation unit is used to analyze the original physical field signal to form an environmental feature vector; The correlation model construction and storage unit is used to analyze the temporal correlation between feature vectors of different modal environments to construct a cross-modal correlation model, and to store environmental feature vector samples and cross-modal correlation models.

[0021] Specifically, the environmental feature extraction module is the system's perception front end. Its core function is to transform raw environmental physical field signals into structured, computable environmental feature data, and to establish a benchmark model for subsequent event detection and multimodal verification. This module comprises three core technical units: a multimodal sensor array unit, a feature vector calculation unit, and a correlation model construction and storage unit. Its data processing flow is as follows: synchronous acquisition of raw signals, feature vector calculation, and cross-modal correlation modeling and storage.

[0022] The multimodal sensor array unit is responsible for highly sensitive and synchronous acquisition of various background physical field signals in the physical environment in which the terminal is located. To achieve data alignment at the "sensing moment" level, all sensor channels are triggered by a unified sampling clock from the same local, highly stable clock source. This embodiment acquires at least the following three physical field signals because they have good spatiotemporal consistency in local space and are difficult to be precisely forged: Geomagnetic field (B): A triaxial magnetoresistive sensor was used to collect the geomagnetic field vector B = [B x B y B z The micro-perturbation signal, sampling frequency Typically between 1 and 10 Hz, to capture its slow-changing characteristics.

[0023] Sound field (A): A wideband microphone is used to collect the sound pressure signal p(t) in the background infrasound to audible frequency range, with a sampling frequency of... The Nyquist sampling theorem must be satisfied.

[0024] Atmospheric pressure field (P): A high-resolution digital barometer was used to collect the micro-change signal of atmospheric pressure P(t), and the sampling frequency was... It is usually around 1Hz.

[0025] Synchronization mechanism example: Assume the terminal initiates a data acquisition at time t0, and the three sensor channels simultaneously record a time window T. w The data within the sequence were used to obtain discrete sequences: geomagnetic sequence B[n], sound pressure sequence p[n], and air pressure sequence P[n], where n is the sampling point index, and the starting sampling point of all sequences corresponds to the same physical time t0. This hardware-level synchronization forms the time basis for subsequent cross-modal correlation analysis.

[0026] The feature vector calculation unit extracts representative and discriminative features from the original synchronization signal, compressing high-dimensional time-series data into low-dimensional feature vectors for real-time processing and comparison. For each acquisition time window T... w The data within the range are used to calculate the characteristics of each modality.

[0027] Taking the geomagnetic field signal B[n] as an example, the calculated characteristics include: Total field intensity change: ,in F is the mean of the components of that axis within the window. ref The total field strength is the long-term average, used to capture the overall shift in magnetic field strength.

[0028] Vector direction change angle: Δθ is the angle difference between the magnetic field vector directions at the beginning and end of the calculation window.

[0029] Disturbance energy: Signal energy E within a specific frequency band band .

[0030] Similarly, features such as sound energy and zero-crossing rate in a specific frequency band are extracted from the sound pressure signal p[n]; features such as trend slope and minute-level fluctuation variance are extracted from the air pressure signal P[n]. Finally, the d features extracted from all modes within a time window are combined into a d-dimensional environmental feature vector F: ; The goal of the joint model construction and storage unit is to establish an intrinsic statistical correlation model between the characteristics of signals from different physical fields, i.e., a cross-modal correlation model. This model is the key basis for determining whether subsequent detected changes in multimode signals are caused by the same natural event. The model is established through a "learning period" during the initial stage of system deployment.

[0031] Data collection: During the initial learning period T init Within this timeframe, feature vectors {F1, F2, ..., F} are continuously collected for N time windows. N}

[0032] Correlation matrix calculation: Calculate any two features f i and f j Pearson correlation coefficient ρ over the entire learning period sequence ij : ; in, μ represents the i-th eigenvalue in the k-th eigenvector. i This is the mean of the feature during the learning period. From this, we obtain a d×d cross-modal correlation matrix R: ; Example: In matrix R, the geomagnetic disturbance energy characteristic f mag_e Infrasound energy characteristics f audio_e The correlation coefficient ρ between them mag_e,audio_e The value could be 0.7, which suggests that in this deployment environment, geomagnetic disturbances are strongly positively correlated with acoustic activity in a specific frequency band.

[0033] Storage: The mean vector μ, standard deviation vector σ, and correlation matrix R of all feature vectors during the learning period are stored together to form the terminal's "local environment feature fingerprint database". The mean μ and standard deviation σ are used for subsequent standardization processing of real-time features.

[0034] Through the above steps, the system not only transforms the original environmental signals into a compact feature representation, but more importantly, it establishes a statistical model R describing the "normal" correlation between multiple background environmental signals. This model serves as an objective benchmark for subsequent event consistency verification. When a real physical event occurs in the environment, it typically leaves coupled perturbation signals in multiple physical fields. The correlation between the features of these perturbation signals should conform to the pre-defined correlation pattern in model R. Conversely, localized, single-modal disturbances are unlikely to produce coordinated changes across multiple modes that meet the model's expectations, thus allowing for effective identification and filtering.

[0035] The event verification module, connected to the feature extraction module, is used to receive environmental feature data, detect environmental feature events based on the environmental feature data, and perform consistency verification on the detected events. Furthermore, the event verification module includes: An event triggering unit is used to trigger potential events based on changes in environmental feature vectors. The consistency verification unit is used to perform multimodal consistency verification on potential events by constructing a cross-modal association model with the storage unit after an event is triggered. The Synchronous Proposal Generation Unit is used to generate synchronous proposals for verified events.

[0036] Specifically, the event verification module is the system's decision-making center, responsible for identifying candidate events suitable for synchronization from a continuous stream of environmental features and performing rigorous multimodal cross-validation on their authenticity. This module receives a sequence of real-time environmental feature vectors from the feature extraction module. Its core process follows a trigger-verification-proposal logical chain: first, the event triggering unit performs preliminary screening; then, the consistency verification unit conducts in-depth multimodal correlation analysis; and finally, standardized synchronization proposals are generated for the verified reliable events. This process ensures that only real physical events that leave coupling imprints in multiple independent physical fields and conform to the inherent statistical laws of the environment can be adopted as benchmark beacons for time synchronization.

[0037] The event triggering unit acts as the system's "sentinel," responsible for continuously monitoring changes in environmental feature vectors to detect abnormal signals with extremely low latency. It does not determine the authenticity of an event, but rather quickly locates the time point of a "potential event" requiring further in-depth analysis.

[0038] This unit uses a sliding time window to process the real-time input environmental feature vector F. t Processing is then performed. First, the current feature vector F is calculated. t The Euclidean distance between the composite change D and a short-term historical baseline is used as the composite change D. t : ; in, Set a dynamic threshold θ trigger This threshold can be adaptively adjusted based on the statistical characteristics of recent changes.

[0039] When D t >θ trigger When triggered, the unit immediately marks a "potential event" at time point t. Upon triggering, the unit locks a fixed-length time window T centered at t. event and all the original feature vector sequences {F} within this window t ΔT ,...,F t+ΔT The data is then sent to the consistency verification unit for in-depth analysis.

[0040] In a calm environment, the characteristic values ​​of geomagnetism, sound waves, and air pressure all fluctuate within a small range. Suddenly, due to a distant lightning event, the terminal at time t...e The simultaneous and sharp increase in both the geomagnetic high-frequency disturbance characteristic value and the infrasound energy characteristic value led to a combined change in D. te The threshold is exceeded instantaneously, thus at t e A potential event is triggered at this point, and a data window containing a period of time before and after the spike is submitted to the next unit.

[0041] The consistency verification unit is the core of ensuring system robustness. It is responsible for "authenticating" the potential events submitted by the triggering unit. The core basis for its authentication is to examine whether the correlation shown by the event on the multimodal signal is consistent with the long-term statistical regularity described by the cross-modal correlation model R established in the feature extraction module.

[0042] Feature change sequence extraction: for event window T event For each feature dimension i within the data, calculate its value relative to the long-term baseline μ before the event. i The change sequence δf i [n], where n is the time index within the window.

[0043] δf i [n]=f i [n] μ i ; Actual correlation calculation: Within the event window, calculate δf for any two feature change sequences. i and δf j The actual Pearson correlation coefficient between them .

[0044] ; Model consistency comparison: The calculated actual correlation coefficient matrix R actual Compare the result with the expected association matrix R retrieved from the fingerprint database. For each pair of features (i,j), calculate the association consistency deviation e. ij : ; Where, ρ ij It is the corresponding value in the model matrix R.

[0045] Calculate the mean consistency deviation E for all feature pairs. avg Set a verification threshold θ verify If E avg ≤θ verify If the event is deemed a valid environmental feature event, it is considered a "valid environmental feature event"; otherwise, it is rejected as noise or local disturbance.

[0046] Example: Following the aforementioned lightning example, the consistency verification unit receives the event window data. It calculates and finds that the geomagnetic disturbance characteristic change sequence and the infrasound energy characteristic change sequence are highly positively correlated within the window. This is consistent with the historical strong positive correlation between these two features recorded in the correlation model R, with a deviation e. mag,audio The deviation is very small. Meanwhile, although the air pressure characteristics also change, their correlation coefficients with changes in geomagnetism and sound waves are consistent with model expectations. Therefore, the overall deviation E... avg The value is low, and the unit determines that this is a real and valid multimodal environmental feature event caused by the same natural phenomenon.

[0047] The synchronous proposal generation unit "creates an ID card" and "issues an invitation" for each valid event that has been verified, that is, it generates a structured synchronous proposal containing all the necessary consensus information in order to initiate a round of distributed time consensus.

[0048] Once the consistency verification unit confirms that an event is valid, this unit immediately generates a synchronization proposal data packet P, which must contain the following fields: Terminal Identifier ID: Uniquely identifies the terminal that initiated the proposal. Local event detection time t local : that is, the precise local clock time t marked by the event triggering unit. e Event feature vector F e : Feature vector at the core moment of the event, used for similarity comparison by other terminals.

[0049] Consistency verification score S verify For example, S verify =1 E avg This quantifies the credibility of the event's verification.

[0050] Preorder Trust Chain H prev The hash value of the latest node in the trust chain maintained by this terminal is used to associate new events with the historical consensus chain to form a trust transfer.

[0051] The generated proposal P will be broadcast through the terminal communication module. It is not just a timestamp statement, but a complete evidence package containing physical evidence, verification credentials, and historical trust context. Other terminals receiving the proposal can independently verify and vote based on this information, thereby initiating a decentralized consensus process to ultimately determine the system-recognized "consensus synchronization moment" corresponding to the event.

[0052] The consensus synchronization module, connected to the event verification module, is used to receive verified environmental feature event information, interact and reach consensus with other terminals in the network to achieve system-wide synchronization moment consensus, and generate and maintain a trust chain that records consensus events. Furthermore, the consensus synchronization module includes: The proposal exchange unit is used to broadcast and receive synchronous proposals via the terminal communication module. The voting decision-making unit is used to verify the received synchronous proposals and form voting opinions based on the verification results; The consensus time calculation unit is used to calculate the final consensus synchronization time based on the local event detection times reported by all supporting terminals after collecting enough positive votes. The Trust Chain Maintenance Unit is used to update the Trust Chain after consensus is reached.

[0053] Specifically, the consensus synchronization module is the core coordination engine of this system. It is responsible for achieving global consensus on the absolute time corresponding to a specific environmental characteristic event in a decentralized network environment. This module receives locally verified synchronization proposals from the event verification module and transforms the local observations of individual terminals into a network-recognized synchronization benchmark through an innovative distributed consensus protocol based on physical event credibility. Its core outputs are the "consensus synchronization moment" and a continuously growing "trust chain." The former provides a precise basis for clock calibration, while the latter establishes an accumulated trust history for the system. The module operates according to a rigorous process of "proposal exchange - independent voting - moment calculation - trust on-chain," ensuring the auditability, attack resistance, and eventual consistency of the synchronization process.

[0054] The proposal exchange unit acts as the "broadcasting station" and "inbox" of the consensus process, responsible for efficiently and reliably disseminating synchronous proposals within a self-organizing terminal network.

[0055] Once the synchronization proposal generation unit generates a valid proposal P, the proposal exchange unit immediately broadcasts it to all neighboring terminals via the reliable data transmission unit of the terminal communication module. The broadcast message includes the proposal P itself and a Time-to-Live (TTL) field to prevent unlimited flooding in the network. Simultaneously, the unit continuously monitors the network, receiving synchronization proposals broadcast from other terminals. For each received proposal, a preliminary verification of its format and signature is performed. If the verification passes, the proposal is placed in a pending proposal queue, and the voting decision unit is immediately notified for processing.

[0056] This unit enables the rapid dissemination of synchronization proposals throughout the network, ensuring that terminals physically sensing the same event receive the proposals in a timely manner, thus creating conditions for initiating consensus voting. Its design guarantees the effective propagation of critical time synchronization information even in dynamically changing network topologies.

[0057] The voting decision-making unit is an independently operating "jury" for each terminal. Based on local sensor data and the system trust model, it carefully evaluates each received synchronous proposal and makes a voting decision. Its decision is not a simple "yes / no", but an opinion that includes weight or confidence level. This is the key innovation that distinguishes this consensus mechanism from traditional majority decisions.

[0058] For a proposal P taken from the queue j This unit performs the following verification and evaluation steps: Retrieve local sensor data and examine the events claimed in the proposal. A reasonable time error window before and after Within the local area, has a similar environmental feature event also been detected? This is done by calculating the locally stored event feature vector and comparing it with the feature vector in the proposal. The similarity is calculated using a similarity metric, such as cosine similarity. If the similarity is below a threshold or there are no relevant local detections, this validation fails.

[0059] Extract the consistency verification score carried in the proposal. This score, generated by the event verification module of the proposal initiating terminal, reflects the reliability of the event's multimodal consistency.

[0060] Query the historical reliability score R maintained by this terminal j This score records the accuracy of terminal j's proposals or votes in past consensus processes and is a dynamically updated reputation value.

[0061] Calculate the current proposal P j The correlation strength L with the most recent events in the local credibility chain, for example, by comparing the similarity of feature vectors or the continuity of event types.

[0062] Based on the above verification and evaluation results, the voting weight W of this terminal for the proposal is calculated. This weight is not a fixed value, but is dynamically calculated using a weighted formula. An example is as follows: W=α·I(Sim>θ sim )+β· +γ·R j +δ·L; Where α, β, γ, and δ are preset weight coefficients, and I(·) is an indicator function, where 1 is true if local validation passes and 0 otherwise. A voting weight threshold θ is set. vote If W ≥ θ vote If the terminal votes "yes", it will assign the voting weight W and the corresponding event detection time recorded locally. The feedback should be sent to the proposal initiating terminal; otherwise, vote "against" or "abstain".

[0063] This unit integrates physical evidence, process quality, historical performance, and system context into a dynamic voting weight. This enables the consensus process to achieve not only quantitative consistency but also qualitative superiority, thereby greatly improving the reliability of the consensus results and the system's robustness against malicious nodes or false detections.

[0064] The consensus time calculation unit operates on the proposal initiation terminal and is responsible for calculating a unique "consensus synchronization time" that best represents the collective observation of the network after collecting enough votes, based on the information from all the affirmative votes. This unit uses a weighted median algorithm to calculate the final time, which is insensitive to outliers and is more robust than simple averaging.

[0065] The terminal that initiates the proposal sets a consensus timeout window. Within this window, the proposal exchange unit collects votes from other terminals.

[0066] When the number of affirmative votes collected reaches the preset quorum Q, the total support is calculated: ; Where V yes This is the set of affirmative votes. It must satisfy Support > θ. support Weighted median calculation: If both the number of voters and the support level meet the criteria, the final calculation begins. All affirmative votes are calculated according to their reported values. Sort the votes from smallest to largest. Simultaneously, calculate the weight W of all affirmative votes. k The sum of S. Then, the weights are accumulated from smallest to largest. When the accumulated weight first reaches or exceeds S / 2, the corresponding... Once determined as the weighted median, this point in time is declared as the consensus synchronization moment t for this event. consensus .

[0067] This unit aggregates scattered, noisy local observation times into a globally recognized, high-precision reference time. The weighted median method ensures that even if some terminals have large measurement errors, it will not have a decisive impact on the final result, thus achieving high-precision fault-tolerant synchronization.

[0068] The Trust Chain Maintenance Unit acts as the system's "notary office" and "historical archive." After each successful consensus, it records the event information in an immutable chain-like data structure. This "Trust Chain" is the system's innovative trust infrastructure, enabling trust to accumulate and be transferred over time.

[0069] Once the consensus time is determined, the computational unit t is determined. consensus This unit executes: Constructing a new block: Create a new chain node whose content includes at least: the consensus synchronization time t of this current block. consensus The event feature hash, the list of terminal IDs that participated in the vote and voted in favor, and the hash value H of the previous chain node. prev Local chain update: Append the new block to the end of the local trust chain.

[0070] Chain state synchronization: The chain update is broadcast to other terminals in the network via the terminal communication module. After verifying the new block content, other terminals synchronize it to their respective local chain copies.

[0071] The introduction of the trust chain enables several key functions: Trust transfer: When a new node joins, it can quickly establish trust in the network history and current time by verifying and synchronizing the existing trust chain, thus accelerating initialization.

[0072] Enhance subsequent consensus: In voting decisions, the correlation strength L between the proposal and the historical chain becomes an important basis, so that terminals with consistent historical behavior have a higher influence in the network.

[0073] Auditing and tamper prevention: All critical synchronization events are recorded sequentially and encrypted, providing a complete operation log. Any tampering with history will be detected by subsequent nodes.

[0074] System state anchoring: The chain itself becomes a distributed "anchor" for system time. Even if external events are lost in the short term, the system can maintain high-precision time estimation based on the last recorded time on the chain.

[0075] The clock calibration module is connected to the consensus synchronization module and is used to calibrate the terminal's local clock according to the consensus of the synchronization time. Furthermore, the clock calibration module includes: The deviation calculation unit is used to calculate the clock deviation based on the consensus synchronization time and the locally recorded event detection time; A clock adjustment unit is used to adjust the local clock according to the clock deviation; The credibility update unit is used to update the historical credibility score of this terminal based on the consensus result.

[0076] Specifically, the clock calibration module is the key execution component that transforms "logical consensus" into "physical synchronization" in the system. It receives the authoritative output from the consensus synchronization module, specifying the consensus synchronization time, and compares it with the terminal's initial local observation records to calculate the precise clock deviation. Subsequently, this module intelligently and smoothly corrects the terminal's local clock, while dynamically updating its credibility score based on the terminal's performance in this consensus process, forming a complete performance feedback loop. This module ensures the final implementation of time synchronization and optimizes the long-term consensus quality of the entire distributed network through continuous evaluation of terminal reliability.

[0077] The deviation calculation unit is the starting point of the calibration process. Its task is to align the network-recognized absolute time reference with the terminal's own relative time measurement and calculate the systematic deviation of the local clock.

[0078] This unit receives two key inputs: consensus synchronization time t consensus The local event detection time t is determined by the consensus time calculation unit of the consensus synchronization module and represents the network's shared recognition of the absolute time of occurrence of a certain environmental characteristic event. local The event triggering unit of this terminal initially records the local clock reading at the time of the event. The clock deviation Δt is calculated using direct interpolation. Δt=t consensus t local ; Here, Δt is a signed number. If Δt > 0, it indicates that the local clock is "slower" than the network consensus time; if Δt < 0, it indicates that the local clock is "faster".

[0079] The clock adjustment unit is responsible for performing the actual correction to the local clock. Directly and abruptly "flicking" the clock hand would cause the timestamp stream to backtrack or jump, which would be disastrous for many applications that rely on monotonically increasing timestamps. Therefore, this unit employs a gradual adjustment strategy, which smoothly absorbs the existing deviation Δt over a subsequent period by fine-tuning the frequency of the clock source, thereby maintaining the continuity and monotonicity of time.

[0080] Workflow and Innovative Adjustment Method: Assume the terminal's local clock is driven by an oscillator with a nominal frequency of f0. The goal of the adjustment unit is to calculate a temporary frequency adjustment Δf, such that within a predetermined adjustment period T... adj Within that time, the accumulated time difference exactly offsets the deviation Δt.

[0081] The basic relationship that needs to be satisfied for adjustment is: ; Therefore, the required relative frequency adjustment rate can be calculated: ; The actual applied frequency is f adj =f0+Δf.

[0082] For hardware-adjustable clock sources, such as voltage-controlled temperature-compensated crystal oscillators (VC-TCXO), this unit generates a voltage control signal proportional to Δf through a digital-to-analog converter, which is applied to the control terminal of the oscillator, thereby physically changing its output frequency.

[0083] For software clocks, this unit achieves equivalent rate adjustment by modifying the time increment rate of the operating system or underlying driver.

[0084] During the adjustment period T adj After completion, the frequency will be restored to the nominal value f0. At this point, the local clock is aligned with the consensus time, and the timestamp sequence generated by the entire adjustment process is smooth and continuous.

[0085] The credibility update unit implements a crucial feedback mechanism. Based on the terminal's behavior throughout the consensus synchronization process and the final calibration result, it dynamically updates the terminal's historical credibility score R. This score is a vital input to the voting decision-making unit in the consensus synchronization module, directly influencing the terminal's future influence in the consensus process. This design encourages terminals to provide accurate data and votes, fostering a virtuous cycle of system self-evolution.

[0086] This unit is triggered once consensus is reached and clock calibration is complete. It evaluates two core aspects: Voting accuracy: Compare whether the final voting opinion of the terminal in the voting decision unit on this consensus proposal is consistent with the consensus result of the entire network. If consistent, add points; otherwise, deduct points.

[0087] Calibration deviation: The magnitude of the final absolute clock deviation |Δt|. The smaller the deviation, the more accurate the original local observation of the terminal.

[0088] Based on the above assessment, a reputation update model is used to calculate the new credibility score R. new A typical model is a variant of the Exponentially Weighted Moving Average (EWMA): R new =λ·R old +(1 λ)·A; Where: R old This is the historical credibility score prior to this consensus. λ is the forgetting factor, which determines the rate of decay of historical credibility; it is typically set close to 1 to ensure smooth credibility changes. A is the performance score of this consensus, determined by both voting accuracy and calibration bias. For example, it can be defined as: ; Where I vote It is an indicator function: 1 when the vote matches the final result, and 0 or a negative value otherwise. τ is a normalization parameter used to map the bias |Δt| to the interval [0,1]. When the bias is greater than τ, this contribution is 0.

[0089] Through this unit, a terminal that consistently provides accurate observations and makes correct consensus judgments will have a higher performance score (A). After multiple consensus iterations, its credibility score (R) will gradually approach a high level. In future consensus processes, its voting weight will be greater, thus having a greater positive impact on the consensus outcome. Conversely, poorly performing or malicious terminals will have their credibility scores reduced, and their influence weakened. This mechanism endows the system with strong adaptability and long-term stability.

[0090] The time restoration module, connected to the clock calibration module, is used to perform multi-channel data acquisition and generate accurate sensing time timestamps for the acquired data; Furthermore, the time restoration module includes: The data acquisition and control unit is used to trigger multi-channel data acquisition; The delay compensation unit is used to store the delay compensation values ​​of each channel obtained from the self-calibration module; The timestamp generation unit is used to generate accurate sensing time timestamps based on the calibrated local time and delay compensation value.

[0091] Specifically, the time-reconstruction module is the final link in realizing the value of this system. It is responsible for executing user-specified multi-channel data acquisition tasks on a highly synchronized time base, overcoming the inherent signal transmission delays of the sensor system, and assigning a corresponding physical world "perception moment" timestamp to each acquired data point. The core innovation of this module lies in achieving precise backtracking from the "data sampling moment" to the "physical event perception moment." It receives a local clock signal from the clock calibration module, aligned with the network consensus time, and combines it with the channel-specific delay compensation values ​​provided by the sensor channel delay self-calibration module, ultimately outputting a data stream with high-precision, physically meaningful timestamps. Its operation follows a process of "synchronous triggering, delay compensation, and time generation," ensuring that data acquired from different modalities and terminals can be accurately compared and fused on the same physical time dimension.

[0092] The data acquisition and control unit is the "commander" of the data acquisition task, determining when to initiate a multi-channel synchronous acquisition. Its triggering mechanism is flexible and can respond to various synchronization commands, ensuring that the acquisition actions are coordinated with the system's synchronization state.

[0093] This unit monitors three main trigger sources: External trigger commands: responding to specific commands from outside the system, such as user manual commands or host computer control signals, to achieve on-demand data acquisition. Internal periodic timing: automatically generating periodic acquisition trigger pulses based on a calibrated local clock according to a preset fixed sampling interval. Trust chain event association triggering: an advanced triggering mode. When the consensus synchronization module generates a new trust chain node, this unit can be configured to initiate a round of data acquisition immediately or after a fixed delay. This allows the system to proactively capture relevant high-frequency data around consensus-reached events with clear physical significance.

[0094] Once the triggering condition is met, the unit generates a global acquisition enable signal and sends it synchronously to all data acquisition channels. Although there may be nanosecond-level differences in the hardware response of each channel, this synchronization signal ensures that the sampling start time of all channels is consistent with microsecond-level precision, laying a unified time starting point for subsequent processing.

[0095] Example: The system reached a consensus on a unique geomagnetic pulse event and recorded it in the trust chain. If the time-reconstruction module is configured in "event-related triggering" mode, it will automatically start a round of high-speed, multi-channel data acquisition 10 milliseconds after the consensus event occurs, aiming to capture the secondary physical effects that the geomagnetic pulse may induce. The sampling start commands of all channels are strictly synchronized.

[0096] The delay compensation unit is a key data storage and provider for reconstructing the "sensing moment." Different physical types of sensors, different signal conditioning circuits, and analog-to-digital converters (ADCs) all introduce unique, time-varying signal transmission delays. This unit is responsible for maintaining a dynamically updated "channel delay compensation value table," storing a unique delay estimate for each acquisition channel. .

[0097] Workflow and data association: This unit stores The value originates from the self-calibration results periodically performed by the sensor channel delay self-calibration module. The self-calibration module measures the total delay of each channel from the occurrence of the physical excitation to the output of the digital sample using an internal excitation source. After filtering and other processing, the updated best estimate is... Send to this unit for storage and replacement.

[0098] ← ; This value is typically on the order of microseconds to milliseconds, and is particularly significant for acoustic channels. This unit ensures that the timestamp generation unit can obtain the latest and most accurate delay compensation parameters for each channel at any time.

[0099] The existence of this unit enables the system to quantify and compensate for the inherent, non-ideal delays in the signal chain, shifting the time reference point from the "digital sampling moment" of the ADC to the "physical sensing moment" of the sensor's sensitive element.

[0100] The timestamp generation unit is the "processor" that performs the final "moment reconstruction" calculation. For each data sample obtained after each acquisition trigger, it uses an innovative formula to calculate the absolute time when the physical quantity represented by the sample was actually sensed by the sensor.

[0101] Workflow and the formula for restoring the moment of innovation: When the acquisition control unit triggers an acquisition, the local high-precision clock records the precise moment t when the trigger takes effect. trigger For the nth data sample output from the cth acquisition channel, its corresponding digital sampling time is... Based on sampling rate Estimation: ; However, This is not the moment of physical perception. The timestamp generation unit reads the delay compensation value for this channel from the delay compensation unit. And the following core formula is applied to calculate the sensing moment. : ; Formula analysis and data flow: enter : Calculated from the calibrated local clock and the known sampling interval, it represents the moment when the data was digitized.

[0102] enter : Obtained from the delay compensation unit, representing the estimated time consumed for the signal to travel from the physical interface to the digital interface.

[0103] Output By subtracting the transmission delay from the digitized time, the time reference point is backdated to the moment when the physical signal arrived at the sensor.

[0104] The output of this unit unifies the timestamps of data from different terminals and different types of sensors to the dimension of "physical world event occurrence," rather than the dimension of "data being recorded by the local system." This is crucial for applications such as multimodal data fusion and time-of-arrival analysis of distributed sensor arrays, achieving true physical-level distributed data synchronization.

[0105] The self-calibration module, connected to the time restoration module, is used to measure the end-to-end delay of each acquisition channel and provide delay compensation values. Furthermore, the self-calibration module includes: An embedded calibration source unit is used to generate physical excitation signals at known times; The delay measurement unit is used to measure the total time from excitation generation to signal output. The compensation value update unit is used to update the delay compensation value based on multiple measurement results.

[0106] Specifically, the self-calibration module is the cornerstone of the system's high-precision "sensing moment" reconstruction. Independent of the normal data acquisition process, it periodically performs end-to-end delay measurements on each sensor channel, aiming to quantify the total time consumed from the actual physical stimulus acting on the sensor to the acquisition of the corresponding digital sample by the system. This module dynamically measures and maintains a precise "channel delay compensation value table" through a built-in calibration source with known timing characteristics, combined with a high-precision time measurement circuit. Its core innovation lies in transforming the traditional offline sensor system delay calibration process, performed in a laboratory environment, into a routine, embedded, online, and automated operation. This allows for tracking and compensating for delay changes caused by factors such as device aging and temperature drift, ensuring the long-term accuracy of "sensing moment" reconstruction. The module's operation strictly follows a closed-loop process of "controllable stimulus → precise timing measurement → filter update."

[0107] The embedded calibration source unit is the excitation generator for the self-calibration process. It is designed and integrated into a miniaturized internal device that can generate physical excitations at known times for each type of sensor that needs calibration.

[0108] This unit receives a calibration start command from the system scheduler. Once triggered, it first obtains a precise trigger timestamp t from its internal high-precision reference clock. emit Immediately afterwards, it was in t emit A physical excitation signal with known characteristics is generated at all times. This excitation signal must match the sensing mechanism of the target sensor: For the acoustic channel: a short, specific-frequency single-tone pulse or linear frequency modulated chirp signal is generated using a miniature piezoelectric ceramic plate or loudspeaker. For the vibration / acceleration channel: a slight, transient mechanical vibration is generated using a miniature electromagnetic exciter or piezoelectric vibrator. For the magnetic field channel: a brief, controllable magnetic field pulse is generated using a miniature Helmholtz coil or current loop.

[0109] The absolute time t when the excitation signal is generated emit The time delay is precisely recorded and latched, serving as an absolute start time reference for subsequent delay measurements. This requires tight hardware coupling between the excitation circuit and the timing circuit to ensure the repeatability and measurability of the delay.

[0110] The delay measurement unit acts as the "timekeeper" in the self-calibration process. It accurately measures the known moment t when the excitation is emitted from the calibration source. emit The time t is when the analog-to-digital converter (ADC) of the corresponding sensor channel outputs the first stable digital sample. adc The time interval between them.

[0111] This unit simultaneously listens for two events: receiving the transmit time t latched by the embedded calibration source unit. emit The system monitors the ADC digital output stream of the target acquisition channel and uses digital signal processing algorithms to identify in real time the ADC sampling point corresponding to the first clearly identifiable signal feature point caused by the calibration excitation.

[0112] Suppose the feature point appears on the m-th ADC sample, and the ADC sampling period is T. s Then the corresponding digitized time t adc It can be calculated as: t adc =t trigger +m·T s ; Among them, t trigger This is the start time of this ADC sampling sequence.

[0113] End-to-end total delay d for a single measurement meas That is: d meas =t adc t emit This d meas This includes the sensor's own response delay, the analog transmission delay of the signal conditioning circuit, the group delay of the anti-aliasing filter, and the sample-hold and conversion delay of the ADC.

[0114] The compensation value update unit is the "data processing center" of the self-calibration process. Due to the single measurement d... meas Potentially affected by noise, interference, or minor fluctuations in the detection algorithm, this unit is responsible for statistically processing multiple consecutive calibration measurement results from the same channel to estimate the most reliable delay compensation value. And update the delay compensation unit of the time restoration module.

[0115] In a complete calibration cycle, the unit performs N consecutive excitation-measurement operations on each channel, obtaining a set of delayed observations. , ,..., }

[0116] To obtain robust estimates and track the slow time-varying delays, this unit employs a two-step update strategy that combines median filtering with exponentially weighted moving average (EWMA): First, calculate the median d of these N measurements. median The absolute deviation of the measurement from each value is calculated. Abnormal measurement points with an absolute deviation greater than a preset threshold are removed.

[0117] For the remaining K valid measurements, calculate their mean. Then, compare it with the latency estimate stored after the previous calibration. By merging, a new delay compensation value is obtained. : =α· +(1 α)· ; Here, α is a smoothing factor. A higher α value makes the estimation results have a stronger "memory" of historical values, resulting in smoother updates and effectively suppressing short-term measurement fluctuations; while (1 The weight of α) gives the current observations the ability to adjust for long-term trends.

[0118] Updated The value is immediately sent to the delay compensation unit of the time restoration module to replace the old value. Simultaneously, this unit records the statistical information from this calibration as a basis for assessing the health status of the sensor channel.

[0119] Through this periodic, online self-calibration mechanism, the system can automatically track and compensate for delay drift in each sensor channel over time and due to environmental changes. For example, the delay of an acoustic sensor may increase by several microseconds to tens of microseconds due to microphone diaphragm aging or preamplifier temperature rise. The self-calibration module can detect this change and update the calibration module accordingly. This ensures that the time-recovery module always uses the latest and most accurate delay parameters for sensing time calculation. This fundamentally guarantees the long-term, stable, and comparable timestamps of data from different terminals and different types of sensors in a distributed data acquisition system, and is a key technical guarantee for achieving a high-precision synchronization system.

[0120] The drift learning module connects the feature extraction module and the event verification module, and is used to monitor statistical changes in environmental features and trigger model updates. Furthermore, the drift learning module includes: Statistical monitoring unit, used to monitor the statistical distribution of environmental feature vectors; The drift detection unit is used to determine whether feature drift has occurred based on changes in statistical distribution. The model collaborative update unit is used to recalculate the cross-modal correlation model after drift occurs and to collaboratively update it with other terminals through the terminal communication module.

[0121] Specifically, the drift learning module is a key component for achieving long-term robustness and environmental adaptability in this system. The cross-modal correlation model established in the feature extraction module is essentially based on the statistical regularities of environmental characteristics within a specific time period during the initial deployment phase. However, the physical environment is not static. For example, background temperature and sound spectrum changes from summer to winter, disturbances to the geomagnetic field caused by nearby new buildings, or long-term geomagnetic reference drift, all of which cause slow changes in the overall statistical distribution of environmental feature vectors, i.e., "feature drift." If the initial model is not updated accordingly, the consistency verification based on a fixed model in the event verification module will generate more and more misjudgments, leading to a decline in system synchronization performance. This module enables the system to track and adapt to these long-term changes online through continuous monitoring, intelligent detection, and collaborative updates, ensuring that the cross-modal correlation model always reflects the true statistical characteristics of the current environment. Its workflow follows a closed loop of "continuous monitoring, drift detection, and collaborative updates."

[0122] The statistical monitoring unit acts as a "data logger" for drift learning. It does not participate in real-time event processing, but rather continuously and over a long period of time collects and compresses environmental feature vector data from the feature extraction module in the form of a background task, constructing and maintaining a historical statistical summary describing the feature distribution.

[0123] This unit processes the high-dimensional environmental feature vector F output by the feature extraction module at a relatively low frequency. t Sampling and archiving are performed. To efficiently store and process long-term data, it does not store all the original vectors, but instead computes and maintains a statistical summary within a sliding time window. For a window containing the L most recent feature vectors, the cell computes: Mean vector μ current The average value of each feature dimension within the window. Covariance matrix Σ current Covariance among the feature dimensions within the window, which contains information about the correlation between features. Feature edge distribution histogram: For key feature dimensions, a histogram of their values ​​is constructed to describe the distribution shape in a nonparametric manner.

[0124] Simultaneously, the unit permanently stores the "benchmark statistical summary" calculated by the system at the end of the initial learning period, including the benchmark mean μ. base , benchmark covariance Σ base And the baseline histogram.

[0125] This unit compresses long-term, continuous high-dimensional data streams into statistical summaries representing recent and initial environmental states, providing a directly comparable data foundation for subsequent drift detection.

[0126] The drift detection unit acts as a "sentinel" of environmental changes, periodically comparing the differences between current environmental statistics and historical baseline statistics, and scientifically determining whether significant "characteristic drift" has occurred.

[0127] This unit is periodically activated. It obtains the current statistical summary (μ) from the statistical monitoring unit. current ,Σ current ) and benchmark statistics summary ((μ base ,Σ base ).

[0128] The core of detection is quantifying the difference between two multivariate distributions. One effective method is to calculate the Jensen-Shannon divergence (JSD) between the current distribution and the baseline distribution. JSD is a symmetric, smoothed version of the Kullback-Leibler (KL) divergence, suitable for comparing two probability distributions P and Q.

[0129] First, by modeling the baseline distribution and the current distribution as multivariate normal distributions i, their empirical distributions can be calculated. To simplify the calculations and focus on key relationships, one approach is to calculate the distributional differences of multidimensional feature vectors across their principal component spaces.

[0130] A more operational and innovative method for determining this is to combine Mahalanobis distance with changes in correlation: Mean drift test: Calculate the current mean vector μ current Mahalanobis distance D relative to the baseline distribution M : ; D M It measures the degree of deviation of the current environmental state's "center point" from the historical normal range, taking into account the correlation between various feature dimensions.

[0131] Correlation structure change test: compare the current covariance matrix Σ current With reference covariance matrix Σ base The main eigenvalues ​​or matrix norm differences ΔΣ. Finally, a comprehensive drift score S is defined. drift : S drift =β1·D M +β2·ΔΣ; Where β1 and β2 are weighting coefficients. A drift threshold θ is set. drift If S drift >θ drift The drift detection unit determines that the system has experienced significant environmental feature drift and triggers the model collaborative update unit.

[0132] The model collaborative update unit is the "actor" that performs model evolution. Once it receives a trigger signal from the drift detection unit, it is responsible for organizing and executing the recalculation of the cross-modal associated model R and the secure synchronization across the network.

[0133] Local model recalculation: This unit instructs the feature extraction module to use all feature vector data from the most recent "representative" period recorded by the statistical monitoring unit to recalculate a new cross-modal correlation matrix R according to the correlation matrix calculation method. new Generating a model update proposal: To prevent malicious nodes from arbitrarily tampering with the model, updates must be agreed upon. This unit creates a model update proposal, which includes: the new model R. new The cryptographic hash value H(R) new ( ), used to calculate the data time period proof for the new model, and the digital signature of the initiating terminal.

[0134] Initiating a lightweight consensus: This model update proposal is broadcast to the network via the terminal communication module. Other terminals, upon receiving the proposal, can verify R using their stored contemporaneous feature data. new The validity of the model is determined through a simplified voting mechanism, allowing network nodes to reach a consensus on its validity.

[0135] Secure synchronization and handover: Once consensus is reached, the initiating unit will complete the R... new Data is distributed to all terminals. Each terminal, after verifying that the complete data matches the hash value, uses R... new Atomically replace the old model R in the associated model building and storage unit of the feature extraction module. old At the same time, the statistical monitoring unit sets the current statistical summary as the new "benchmark".

[0136] This unit enables the entire distributed system to smoothly upgrade its core model online in a coordinated and consistent manner. After the model is updated, the event verification module uses the new association model, which matches the current environment, to perform consistency verification, thus restoring and maintaining the system's event detection and synchronization accuracy. This gives the system the ability to be deployed for months or even years without manual intervention to reset or recalibrate, significantly improving the system's usability and reliability. The entire collaborative update process ensures the consistency of the models on all terminals in the network, which is a key mechanism for maintaining the overall performance stability of a large-scale distributed system.

[0137] The terminal communication module is used to provide data interaction support between terminals for the consensus synchronization module and the drift learning module.

[0138] Furthermore, the terminal communication module includes: The self-organizing network management unit is used to manage neighbor discovery and network topology maintenance among terminals. A reliable data transmission unit is used to transmit synchronous proposals, voting opinions, credibility chain update data, and model update information.

[0139] Specifically, the terminal communication module serves as the "neural network" and "information superhighway" for the entire distributed system. In field environments with no pre-defined infrastructure and dynamically changing topologies, it is responsible for establishing and maintaining physical and logical connections between terminals, providing reliable, orderly, and secure data transmission services for upper-layer core functional modules, especially the consensus synchronization module and drift learning module. The core design of this module lies in its self-organization and service-aware reliability. It ensures that critical synchronization proposals, voting, trust chain updates, and model update instructions can overcome the challenges of unstable wireless links and be delivered on time and accurately between the correct terminals, forming the physical foundation for the system's decentralized collaboration. Its operation is collaboratively completed by the self-organizing network management unit and the reliable data transmission unit.

[0140] The self-organizing network management unit is the "creator" and "maintainer" of the network. It enables each terminal to automatically discover neighbors within its communication range and jointly build and maintain a dynamic, multi-hop wireless network topology without any centralized access point or pre-configured network information.

[0141] Neighbor Discovery and Link Probe: Terminals periodically broadcast short "beacon frames" on a preset wireless channel. Each beacon frame contains the terminal's unique ID, current timestamp, and the latest block hash of its maintained trust chain. When terminal A receives a beacon frame from an unknown terminal B, it adds B to its "neighbor candidate list" and initiates a link quality assessment process. This includes measuring the received signal strength indicator, calculating the beacon frame reception success rate, and possibly actively sending probe packets to measure round-trip time.

[0142] Dynamic topology construction and maintenance: Based on link quality assessment results, each terminal maintains an "effective neighbor table." This table records not only the neighbor ID and link quality, but also other terminals reachable through that neighbor. The unit runs a simplified, on-demand routing protocol. For example, when it needs to communicate with a non-neighbor terminal, it queries its local neighbor table, selects the neighbor with the best link quality and fewest hops as the next hop, or initiates a simple route request broadcast.

[0143] Network self-healing: The unit continuously monitors the link status with its neighbors. If a neighbor's beacon frames are continuously lost or the link quality remains below a threshold, the link is considered failed, and the neighbor is removed from the valid neighbor table. Simultaneously, it recalculates reachability information and broadcasts link status updates as necessary, triggering local topology reconfiguration.

[0144] This unit gives the system plug-and-play capability and adaptability to environmental changes. It provides a basic "communication map" for reliable data transmission units, ensuring that data packets know which neighbor to send to in order to eventually reach the target terminal.

[0145] The reliable data transmission unit is the "guardian" of core business data. It operates on the topology provided by the self-organizing network management unit and provides different levels of transmission services, all of which guarantee eventual correctness, depending on the urgency and importance of different types of upper-layer data.

[0146] This unit provides dedicated data transmission interfaces for the upper-layer consensus synchronization module and drift learning module. Its core mechanisms include: Data fragmentation and serialization: For data exceeding the length of a single frame, the unit is responsible for fragmenting, numbering, and adding frame headers. Hybrid acknowledgment and retransmission mechanism: For synchronization proposals and voting, a "single-hop acknowledgment + limited retransmissions" strategy is adopted. After the sending terminal passes the data packet to its next-hop neighbor, it starts an acknowledgment timer. Upon receiving the timer, the neighbor must immediately reply with a link-layer acknowledgment frame. If the sending terminal does not receive an acknowledgment before the timeout, it retransmits. After exceeding the maximum number of retransmissions, it reports a transmission failure to the upper layer, which may trigger rerouting or waiting for the next opportunity.

[0147] For trust chain updates and model updates: an "end-to-end acknowledgment" strategy is adopted. Data packets are forwarded through multiple hops to the final destination terminal. After fully receiving and verifying all fragments, the destination terminal returns an application-layer acknowledgment message along the original path or a new path. The sending terminal considers the transmission successful only upon receiving this end-to-end acknowledgment. Otherwise, retransmission may be arranged through application-layer logic.

[0148] Priority queues and congestion control: Each unit maintains different priority queues for sending data. For example, synchronization proposals and votes have the highest priority to ensure low latency in the consensus process; chain updates are next; model updates and regular network maintenance messages have lower priority. When network congestion occurs, the unit schedules sending according to queue priorities and may report the congestion status to higher layers, which then decide whether to slow down the data generation rate.

[0149] Consensus synchronization scenario: The proposal exchange unit of the consensus synchronization module generates a synchronization proposal P. It submits P to the reliable data transmission unit of the terminal communication module, specifying it as "high priority, single-hop acknowledgment". The ad hoc network management unit selects a next-hop neighbor for it. The reliable data transmission unit packages and sends P, waiting for link-layer acknowledgment from the neighbor. After acknowledgment is received, the corresponding unit of that neighbor node continues to forward P to its neighbors until it is propagated throughout the network. The process of transmitting voting opinions is similar.

[0150] In the model collaborative update scenario, the model collaborative update unit of the drift learning module generates a model update proposal M. It submits M to the reliable data transmission unit, specifying it as "high reliability, end-to-end confirmation". The data packet is routed multi-hop to every participating terminal in the network. After each terminal fully receives M, its reliable data transmission unit sends an application-layer confirmation back to the source terminal. Only when the source terminal receives confirmations from a sufficient number of terminals does it determine that the model update distribution was successful, thus triggering subsequent consensus voting and switching processes.

[0151] The terminal communication module, through its self-organizing capabilities and hierarchical reliable transmission services, logically integrates physically dispersed independent terminals into a collaborative organic whole. It ensures that the core consensus and collaborative learning processes are not interfered with by the underlying unstable wireless environment, enabling the entire innovative distributed synchronization system to operate stably in real-world, complex scenarios without infrastructure. This provides the underlying communication guarantee for realizing all upper-layer technological innovations.

[0152] Furthermore, the system also includes: The new node access module connects the terminal communication module and the consensus synchronization module. It enables newly joined terminals to listen to the neighbor status, obtain the current trust chain copy and feature fingerprint database, and complete the initial synchronization and model calibration by participating in the first consensus event.

[0153] Specifically, the new node bootstrapping module is a key guarantee for the dynamic scalability and plug-and-play deployment of this distributed system. In a decentralized, self-organizing network, new data acquisition terminals must be able to autonomously and securely join the existing synchronized network cluster and quickly obtain a spatiotemporal reference and cognitive model consistent with the cluster. This module is responsible for coordinating and managing the entire bootstrapping process for new nodes, enabling them to complete the entire process from network discovery and state synchronization to final integration without manual configuration or reliance on any preset bootstrapping server. Its core innovation lies in utilizing the system's inherent trust chain and consensus mechanism as the trust anchor and synchronization tool for bootstrapping, achieving decentralized, secure, and reliable node bootstrapping. Its workflow follows a progressive step of "listening and discovery, state acquisition, and initial consensus fusion."

[0154] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-channel data acquisition terminal and a distributed data acquisition synchronization system, characterized in that, The system includes: The feature extraction module is used to collect and extract multimodal physical field signals of the environment in which the terminal is located, in order to generate environmental feature data; The event verification module, connected to the feature extraction module, is used to receive environmental feature data, detect environmental feature events based on the environmental feature data, and perform consistency verification on the detected events. The consensus synchronization module, connected to the event verification module, is used to receive verified environmental feature event information, interact and reach consensus with other terminals in the network to achieve system-wide synchronization moment consensus, and generate and maintain a trust chain that records consensus events. The clock calibration module is connected to the consensus synchronization module and is used to calibrate the terminal's local clock according to the consensus of the synchronization time. The time restoration module, connected to the clock calibration module, is used to perform multi-channel data acquisition and generate accurate sensing time timestamps for the acquired data; The self-calibration module, connected to the time restoration module, is used to measure the end-to-end delay of each acquisition channel and provide delay compensation values. The drift learning module connects the feature extraction module and the event verification module, and is used to monitor statistical changes in environmental features and trigger model updates. The terminal communication module is used to provide data interaction support between terminals for the consensus synchronization module and the drift learning module.

2. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The feature extraction module includes: A multimodal sensor array unit is used to simultaneously acquire raw physical field signals, including at least the geomagnetic field, the acoustic field, and the atmospheric pressure field. The feature vector calculation unit is used to analyze the original physical field signal to form an environmental feature vector; The correlation model construction and storage unit is used to analyze the temporal correlation between feature vectors of different modal environments to construct a cross-modal correlation model, and to store environmental feature vector samples and cross-modal correlation models.

3. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The event verification module includes: An event triggering unit is used to trigger potential events based on changes in environmental feature vectors. The consistency verification unit is used to perform multimodal consistency verification on potential events after an event is triggered by constructing a cross-modal association model with the storage unit using the association model. The Synchronous Proposal Generation Unit is used to generate synchronous proposals for verified events.

4. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The consensus synchronization module includes: The proposal exchange unit is used to broadcast and receive synchronous proposals via the terminal communication module. The voting decision-making unit is used to verify the received synchronous proposals and form voting opinions based on the verification results; The consensus time calculation unit is used to calculate the final consensus synchronization time based on the local event detection times reported by all supporting terminals after collecting enough positive votes. The Trust Chain Maintenance Unit is used to update the Trust Chain after consensus is reached.

5. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The clock calibration module includes: The deviation calculation unit is used to calculate the clock deviation based on the consensus synchronization time and the locally recorded event detection time; A clock adjustment unit is used to adjust the local clock according to the clock deviation; The credibility update unit is used to update the historical credibility score of this terminal based on the consensus result.

6. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The time restoration module includes: The data acquisition and control unit is used to trigger multi-channel data acquisition; The delay compensation unit is used to store the delay compensation values ​​of each channel obtained from the self-calibration module; The timestamp generation unit is used to generate accurate sensing time timestamps based on the calibrated local time and delay compensation values.

7. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The self-calibration module includes: An embedded calibration source unit is used to generate physical excitation signals at known times; The delay measurement unit is used to measure the total time from excitation generation to signal output. The compensation value update unit is used to update the delay compensation value based on multiple measurement results.

8. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The drift learning module includes: Statistical monitoring unit, used to monitor the statistical distribution of environmental feature vectors; The drift detection unit is used to determine whether feature drift has occurred based on changes in statistical distribution. The model collaborative update unit is used to recalculate the cross-modal correlation model after drift occurs and collaboratively update it with other terminals through the terminal communication module.

9. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The terminal communication module includes: The self-organizing network management unit is used to manage neighbor discovery and network topology maintenance among terminals. A reliable data transmission unit is used to transmit synchronous proposals, voting opinions, credibility chain update data, and model update information.

10. The multi-channel data acquisition terminal and distributed data acquisition synchronization system according to claim 1, characterized in that, The system also includes: The new node access module connects the terminal communication module and the consensus synchronization module. It enables newly joined terminals to listen to the neighbor status, obtain the current trust chain copy and feature fingerprint database, and complete the initial synchronization and model calibration by participating in the first consensus event.

Images