Intelligent networking system of six-dimensional force sensor array and data transmission and storage method

Abstract

The invention discloses an intelligent networking system of a six-dimensional force sensor array and a data transmission and storage method, and the system comprises the six-dimensional force sensor array which comprises a plurality of six-dimensional force sensor nodes, is used for collecting corresponding six-dimensional force time sequence data, and is internally provided with an edge model; the wireless intelligent networking and synchronization module is used for establishing a logic cluster, electing a main node, and performing intra-cluster synchronization and data aggregation to obtain feature data; the self-adaptive data transmission module is used for dynamically switching data transmission modes of the wireless intelligent networking and the synchronization module; and the cloud intelligent analysis platform performs deep analysis on each piece of transmission data through a biomechanical large model, and is used for feeding back the abnormity of the sole and the kinematics parameters of the sole and generating a corresponding report chart. Full-field, high-synchronization and high-resolution measurement of the plantar six-dimensional force field under complex gaits is achieved, key mechanical events can be intelligently focused, a deep analysis report with clinical guiding significance is generated, and the method has extremely high effectiveness and advancement.

Description

Technical Field

[0001] This invention relates to the field of advanced sensing and Internet of Things (IoT) technology, specifically to an intelligent networking system for a six-dimensional force sensor array and a data transmission and storage method. Background Technology

[0002] As scientific research deepens, researchers' exploration of the mechanics of human movement is gradually expanding from small-scale to large-scale movements. For small-scale movements such as standing balance and large-scale movements such as walking a fixed distance, the commonly used two-plate weighing force sensors are insufficient for measurement applications, leading to the development of sensor arrays. Simultaneously, in rehabilitation training in the medical field, force sensor arrays are expected to assist rehabilitation physicians in monitoring the recovery of lower limb motor disorders caused by various pathological reasons, and in monitoring patients' rehabilitation progress during periodic training and testing. Therefore, to achieve full-field, high-resolution six-dimensional force measurement of the entire sole, the use of large-scale six-dimensional force sensor arrays has become an inevitable trend. However, integrating dozens or even hundreds of high-data-dimensional sensor nodes into a stable and reliable measurement system presents challenges in system networking, data transmission, and storage management that are difficult to overcome with current technologies.

[0003] Currently, there are two main system architecture solutions for intelligent sensor networking applications in medical rehabilitation training: Firstly, a centralized direct-connection acquisition architecture is adopted, which is the most traditional solution. Each six-dimensional force sensor node is directly connected to the central data acquisition box or industrial control computer through an independent analog or digital cable (e.g., each sensor requires 6 analog signal lines or 1 CAN bus). All data analog-to-digital conversion, filtering, and calculation are completed in the central processing unit. However, this approach has at least the following defects or shortcomings: (1) The wiring is extremely complex and the system is bloated: As the number of sensor nodes increases, the number of cables increases year by year. This not only makes system deployment difficult and reduces reliability (the cables are easily damaged), but also seriously limits its application in space-constrained scenarios. (2) Synchronization performance bottleneck: Synchronization relies on a central clock. Long-distance and inconsistent transmission paths will cause small timing deviations (microseconds to milliseconds) in the data acquisition of each node, making it difficult to achieve true high-precision synchronization and affecting the accuracy of dynamic gait analysis. (3) Concentrated data transmission pressure: The raw high-speed six-dimensional force data of all nodes simultaneously floods the central processor, which puts great pressure on the processing capacity and bus bandwidth of the central node, and is very likely to become a system performance bottleneck, limiting the expansion of the array size. (4) Lack of intelligence and fault tolerance: Any node or cable failure may affect the entire system, and the self-identification of nodes and self-diagnosis of the system cannot be achieved.

[0004] Secondly, a simple distributed bus architecture is adopted. This scheme attempts to simplify wiring by connecting all six-dimensional force sensor nodes to a shared communication bus (such as a CAN bus) and transmitting data by polling or contention based on address. However, this approach also has the following problems and shortcomings: (1) Real-time performance and bandwidth contradiction: The shared bus bandwidth is limited. When the number of nodes increases and the amount of data is large, bus conflicts intensify or the polling cycle becomes longer, resulting in a significant increase and uncertainty in network transmission delay. It is impossible to guarantee the real-time performance of each node's data and it is difficult to meet the requirements of high-speed dynamic measurement. (2) Lack of intelligent data scheduling: The "one-size-fits-all" transmission strategy is adopted. Regardless of whether the data is important (e.g., data in a static state and data at the moment of impact), the same communication resources are occupied, resulting in a waste of network bandwidth and low overall system efficiency. (3) Data silos and fusion difficulties: Although the sensors are physically connected, the data of each node is still logically independent. The central node receives a series of discrete data packets with timestamps. There is a lack of an effective mechanism to correlate, align and fuse data from different nodes in real time at the network level to directly generate full-field mechanical distribution information. The workload of post-processing data is huge. (4) Single storage structure: The collected data is usually stored linearly in chronological order or node address. When it is necessary to backtrack and analyze data of a specific area (spatial dimension) or a specific event (such as the moment when all nodes exceed the torque limit), the retrieval efficiency is extremely low and the data value is not effectively mined.

[0005] As can be seen from the above, existing technologies have the following core defects in their networking and data processing methods when constructing large-scale six-dimensional force sensor arrays: (1) The network topology is rigid and cannot achieve a balance between complexity, reliability and scalability; (2) The data transmission strategy is simple and cannot meet the requirements of high synchronization, low latency and high concurrency, and the bandwidth utilization efficiency is low; (3) The data management and storage methods are primitive and fail to reflect the inherent spatial attributes of the sensor array, which makes data fusion and advanced analysis difficult.

[0006] Therefore, there is an urgent need in this field for an intelligent networking system for large-scale six-dimensional force sensor arrays and its supporting data transmission and storage methods to solve the systemic problem of moving from "connectivity" to "intelligent efficiency". Summary of the Invention

[0007] To address the aforementioned problems, the purpose of this invention is to propose an intelligent networking system for a six-dimensional force sensor array and a data transmission and storage method.

[0008] This was achieved through the following technical solutions: A smart networking system for a six-dimensional force sensor array includes a six-dimensional force sensor array, a wireless smart networking and synchronization module, an adaptive data transmission module, and a cloud-based intelligent analysis platform; The six-dimensional force sensor array comprises multiple six-dimensional force sensor nodes, each used to acquire corresponding six-dimensional force time-series data, and includes a built-in edge model for vector event detection and local feature extraction; the six-dimensional force time-series data of any six-dimensional force sensor node includes [F x(t) , F y(t) , F z(t) M x(t) M y(t) M z(t) F represents force, M represents torque, x and y correspond to two coordinate axes parallel to the six-dimensional force sensor array, and z is the coordinate axis perpendicular to the six-dimensional force sensor array. The wireless intelligent networking and synchronization module uses a six-dimensional force sensor of foot stepping as a trigger event. Each sensor node that is triggered simultaneously and is adjacent to the other will be dynamically assembled into a logical cluster. With the help of the edge model, the master node is elected within the logical cluster according to the multi-attribute decision mechanism. Then, the cluster synchronization and data aggregation are performed to obtain feature data. The adaptive data transmission module is used to dynamically switch the data transmission mode of the wireless intelligent networking and synchronization module based on the vector event detection results of the edge model. The data transmission modes include at least sleep mode, normal mode, and event mode. The cloud-based intelligent analysis platform receives and stores each transmission data from the adaptive data transmission module, and performs in-depth analysis of each transmission data through a large biomechanical model to provide feedback on foot abnormalities, foot kinematic parameters, and generate corresponding reports and charts.

[0009] Optionally, the multi-attribute decision-making mechanism includes electing the master node in the following order of priority: a) maximum load; b) maximum signal strength; c) maximum remaining power; d) minimum node ID.

[0010] Optionally, each logical cluster performs intra-cluster synchronization independently, and each master node initiates wireless clock synchronization based on PTP; when data is aggregated within any logical cluster, the timing data of each non-master node is sent to the master node for spatial data fusion to obtain the corresponding total resultant force, total resultant torque, pressure center trajectory and as feature data.

[0011] Optionally, the hibernation mode is set to run every F for a set time period. z(t) When all values ​​are below the preset threshold I; the normal mode is when F occurs. z(t) Greater than the preset threshold I and F z(t)The rate of change exceeds the preset threshold II; the event mode is that the result of vector event detection satisfies a specific event, which includes at least severe impact and foot slippage.

[0012] Optionally, the formula for determining a severe impact is: ,in, Let be the resultant force vector at time t. The membrane is the resultant force vector. The membrane as the resultant force vector in the time window The average rate of change within, The threshold value corresponds to a severe impact.

[0013] Optionally, foot sliding is based on horizontal shear force ( F x ，F y The covariance and the abrupt change value of the resultant force direction angle β are used to determine the resultant force.

[0014] Optionally, the adaptive data transmission module is also equipped with misoperation protection, which filters out accidental interference based on time window, average rate of change and duration.

[0015] Secondly, a data transmission and storage method is proposed, which is implemented using the aforementioned intelligent networking system of a six-dimensional force sensor array. This method includes the following steps: S1. Dynamic Triggering and Network Synchronization: Based on the mechanical events generated by foot stomping, the corresponding multiple sensor nodes in the six-dimensional force sensor array are dynamically triggered and logical clusters are formed. Within the logical cluster, the election of the master node and the synchronization within the cluster are completed. S2, Intelligent Acquisition and Adaptive Transmission: Synchronously acquire data from multiple sensor nodes within the logical cluster, perform vector event detection based on the edge model, and adaptively switch between sleep mode, normal mode and event mode according to the detection results, and prepare to transmit data according to the corresponding mode. S3. Intra-cluster fusion and unified reporting: In normal mode or event mode, the master node receives and fuses intra-cluster data, calculates the overall mechanical parameters, and uploads the fusion results and the original data in the corresponding mode to the cloud intelligent analysis platform. S4. Semantic Storage and Intelligent Analysis: The cloud-based intelligent analysis platform calls upon a large biomechanical model to perform in-depth analysis of the uploaded data and generate corresponding feedback reports.

[0016] Optionally, the edge model is a temporal neural network model based on the Transformer architecture, which performs vector event detection through a self-attention mechanism and a feedforward neural network.

[0017] The beneficial effects of this invention compared to the prior art are: The technical solution of this invention adopts a wireless networking and event-triggered dynamic logic cluster formation mechanism, eliminating complex wiring, supporting rapid deployment in various complex scenarios, effectively avoiding single points of failure, and ensuring the continuity of data acquisition and system reliability. Through a synchronization strategy oriented towards dynamic logic clusters, it solves the problem of multi-node timing deviation, laying the foundation for accurate mechanical analysis. Through edge models and multi-mode collaboration, the system can cope with various different situations and select the most suitable transmission method, achieving intelligent scheduling and optimizing system energy efficiency. Based on a large biomechanical model, it identifies complex abnormal patterns, inverts deep joint mechanical parameters, and generates accurate personalized reports, realizing the transformation from massive data to clinical decision support. Attached Figure Description

[0018] Figure 1 A schematic diagram of the framework of an intelligent networking system for a six-dimensional force sensor array; Figure 2 A flowchart of a data transmission and storage method; Figure 3 This is a schematic diagram of a six-dimensional force sensor array. Detailed Implementation

[0019] The following will be based on embodiments of the present invention. Figures 1 to 3 The technical solutions in the embodiments of the present invention will be described in detail below.

[0020] like Figure 1 The diagram shown is a schematic of the framework of an intelligent networking system for a six-dimensional force sensor array; as shown... Figure 2 The diagram shown is a flowchart of a data transmission and storage method; combined with Figure 1 and Figure 2 As shown, this solution constructs an intelligent networking system architecture for a six-dimensional force sensor array, a high-concurrency data transmission protocol under this architecture, and a space-time correlation storage method oriented towards the mechanical data characteristics of the array. With the help of a six-dimensional force sensor array, a wireless intelligent networking and synchronization module, an adaptive data transmission module, and a cloud-based intelligent analysis platform, it successfully achieves full-field, high-synchronization, and high-resolution measurement of the six-dimensional force field of the foot under complex gait. This enables the system to intelligently focus on key mechanical events and automatically generate in-depth analysis reports with clinical guidance significance, verifying the effectiveness and advancement of the technical solution of this invention.

[0021] In this embodiment, the six-dimensional force sensor array includes multiple six-dimensional force sensor nodes (hereinafter referred to as nodes). Each six-dimensional force sensor node is used to collect six-dimensional force time-series data (e.g., at a sampling rate of 1000Hz) and has a built-in edge model for vector event detection and local feature extraction. The six-dimensional force time-series data of any six-dimensional force sensor node includes [F x(t) , Fy(t) , F z(t) M x(t) M y(t) M z(t) F represents force, M represents torque, x and y correspond to two coordinate axes parallel to the six-dimensional force sensor array, and z is a coordinate axis perpendicular to the six-dimensional force sensor array; in addition, when transmitting six-dimensional force time-series data, the six-dimensional force sensor node will also transmit its own ID and corresponding timestamp together.

[0022] The wireless intelligent networking and synchronization module, designed for the working characteristics of force plate arrays (multiple plates arranged horizontally or at an angle, with one foot potentially covering multiple adjacent plates in a single measurement), abandons the static networking strategy based on fixed physical locations and adopts a dynamic logical networking and synchronization mechanism based on event triggering. Using the six-dimensional force sensor of foot stepping as the trigger event, each sensor node that is simultaneously triggered and adjacent is dynamically assembled into a logical cluster. With the help of an edge model, a master node is elected within the logical cluster according to a multi-attribute decision-making mechanism, and then intra-cluster synchronization and data aggregation are performed to obtain feature data.

[0023] In this embodiment, after system initialization, all six-dimensional force sensor nodes are in a low-power listening (sleep) state. When a foot is stepped on, one or more adjacent six-dimensional force plate sensor nodes that are covered and subjected to effective load will be automatically triggered. These simultaneously triggered nodes dynamically and temporarily form a logical cluster for this measurement through a wireless communication protocol. The members of this logical cluster are determined solely by the foot coverage situation and are strongly related to the physical installation location of the nodes, but the organization method is dynamically variable. This logical cluster automatically elects a master node, which is the six-dimensional force sensor node with the largest load in this measurement, responsible for coordinating intra-cluster synchronization and data aggregation for this measurement.

[0024] When any dynamic logical cluster is formed, the corresponding master node within the cluster immediately initiates a high-precision wireless clock synchronization process (based on PTP) to ensure that the data acquisition clocks of all nodes under the corresponding footprint coverage are strictly aligned at the microsecond level. This synchronization mechanism is performed independently for each measurement within each dynamic logical cluster, effectively avoiding data fusion errors caused by asynchronous measurements and laying the foundation for accurate calculation of parameters such as overall resultant force and pressure center. Within the logical cluster, each non-master node sends the synchronously acquired six-dimensional force time-series data to the master node. The master node does not simply forward the data but performs real-time spatial data fusion, including, but not limited to, vector synthesis of the force-moment vectors of each node, to obtain the total resultant force and resultant moment time-series data of the entire sole contact surface as feature data. Then, based on the moment and spatial coordinates of each node, the time-series data of the pressure center trajectory of the entire sole can be calculated. The master node uploads the fused high-level data results to the cloud gateway via the adaptive data transmission module, thereby performing corresponding analysis on the cloud intelligent analysis platform, reducing data congestion on the wireless channel and the total system power consumption.

[0025] To address the possibility of two six-dimensional force sensor nodes having the same maximum load value, the multi-attribute decision-making mechanism can elect a master node according to the following priority: a) The node with the largest load is selected based on the current trigger value. z a) The node with the largest component reading; b) The node with the largest signal strength. If the load is the same, compare the corresponding node with the wireless signal strength and select the strongest node; c) The node with the largest remaining power. If the aforementioned signal strengths are still the same, select the node with the highest remaining power among the six-dimensional force sensor nodes; d) The node with the smallest node ID number. If the aforementioned contents are all the same, select the node with the smallest physical ID number.

[0026] In this embodiment, the edge model can perform the following real-time analysis on this six-dimensional force time-series data: 1) Vector event detection: It can not only identify complex mechanical events such as "foot slip" and "severe impact" based on the threshold of a single component, but also based on the magnitude and direction of the resultant force / resultant torque.

[0027] 2) Local feature extraction: Time series data can be calculated online at the node, such as statistical features (mean, peak value, variance) or frequency domain features within a time window, providing a basis for decision-making for subsequent multi-mode transmission strategies and reducing data transmission pressure from the source.

[0028] For severe impacts, the determination formula is as follows: ,in, Let be the resultant force vector at time t. The membrane is the resultant force vector. The membrane as the resultant force vector in the time window The average rate of change within, The threshold value corresponds to a severe impact.

[0029] Regarding foot slippage, this event was analyzed by comprehensively examining the horizontal shear force. The determination is based on the covariance characteristics and the abrupt change in the direction of the resultant force. Its mathematical definition includes two cooperative criteria: Criterion 1: Shear force covariance ,in, To a sliding time window Within 200ms (can be set according to different acquisition requirements), shear force component and covariance, The covariance threshold can be obtained through the calibration process.

[0030] Calibration procedure: Subjects perform known, slight sliding motions on the target test surface, and the shear force covariance is recorded and analyzed. The peak distribution is then used to determine an optimal threshold that can effectively distinguish between steady states and sliding events.

[0031] Criterion 2: Direction angle of resultant force mutation ,in, The threshold for determining a sudden change in direction can be set to 30°, and "slide" indicates sliding.

[0032] Therefore, the complete formula for determining foot slip events is: .

[0033] The adaptive data transmission module is used to dynamically switch the data transmission mode of the wireless intelligent networking and synchronization module based on the vector event detection results of the edge model. The data transmission modes include at least sleep mode, normal mode, and event mode.

[0034] Sleep mode is set to run every F for a set time period. z(t) When all values ​​are below the preset threshold I, i.e., when there is no load, each node only sends a preset heartbeat signal, resulting in extremely low power consumption. The normal mode is characterized by F... z(t) Greater than the preset threshold I and F z(t) When the rate of change exceeds the preset threshold II, it corresponds to the situation where, during steady-state movements such as walking, each node prioritizes sending compressed feature parameters (such as resultant force and central pressure), and then transmits the original data (six-dimensional force time series data) to the cloud when idle. The event mode is that the results of vector event detection satisfy specific events, which at least include severe impact and foot slippage.

[0035] The adaptive data transmission module is also equipped with misoperation protection, which filters out interference caused by false triggering based on time window, average rate of change and signal duration of corresponding data.

[0036] Each piece of data transmitted via the adaptive data transmission module contains a corresponding timestamp and node ID number, and is forcibly associated with the corresponding global physical space coordinates (x, y, z) and biomechanical semantic labels (such as pressure center and gait cycle, which can be obtained from the feature parameters fused within the cluster).

[0037] The cloud-based intelligent analysis platform receives and stores each transmission data from the adaptive data transmission module, and performs in-depth analysis of each transmission data through a large biomechanical model to provide feedback on foot abnormalities, foot kinematic parameters, and generate corresponding reports and charts.

[0038] The large-scale biomechanical model utilizes massive amounts of pre-collected gait analysis, pathological gait, and sports biomechanical data. Based on the Transformer architecture, a pre-trained basic model is applicable to the following intelligent analysis scenarios: 1) Anomaly detection and early warning: The model can analyze the dynamic changes of six-dimensional forces across the entire field in real time, identifying abnormal patterns that are difficult to detect with the naked eye. For example, it can detect abnormal plantar pressure distribution in diabetic patients early, providing an early warning of the risk of foot ulcers; or detect abnormal joint torques at the moment of landing in athletes in real time, providing an early warning of the risk of ankle sprains or anterior cruciate ligament injuries. 2) Biomechanical parameter inversion: It can directly inversely derive plantar force and other gait-related kinematic parameters from the six-dimensional force sensor array data. 3) Personalized feedback and guidance: It generates natural language reports and charts to intuitively explain gait problems and assist physicians in providing personalized rehabilitation training or exercise posture adjustment suggestions for patients or athletes.

[0039] Secondly, a data transmission and storage method is proposed, which is implemented using the aforementioned intelligent networking system of a six-dimensional force sensor array. This method includes the following steps: S1. Dynamic Triggering and Network Synchronization: Based on the mechanical events generated by foot stomping, the corresponding multiple sensor nodes in the six-dimensional force sensor array are dynamically triggered and logical clusters are formed. Within the logical cluster, the election of the master node and the synchronization within the cluster are completed. S2. Intelligent Acquisition and Adaptive Transmission: Simultaneously acquire data from multiple sensor nodes within the logical cluster, perform vector event detection based on the edge model, and adaptively switch between sleep mode, normal mode, and event mode according to the detection results, preparing to transmit data according to the corresponding mode; the edge model is a temporal neural network model based on the Transformer architecture, which performs vector event detection through self-attention mechanism and feedforward neural network; S3. Intra-cluster fusion and unified reporting: In normal mode or event mode, the master node receives and fuses intra-cluster data, calculates the overall mechanical parameters (characteristic parameters), and uploads the fusion results and the original data in the corresponding mode to the cloud intelligent analysis platform. S4. Semantic Storage and Intelligent Analysis: The cloud-based intelligent analysis platform invokes a large-scale biomechanical model to perform in-depth analysis of the uploaded data and generate corresponding feedback reports. The function and effects of this method can be found in the description of the aforementioned system, and will not be repeated here.

[0040] like Figure 3 The diagram shows a schematic of a six-dimensional force sensor array, consisting of 30 force plates (six-dimensional force sensors) used for gait and steering mechanics analysis. Figure 3 As shown, here is another example of this application: Thirty six-dimensional force sensors are used as the core sensing elements of the force measuring plate. Each node (marked as 1) can simultaneously measure the force (F) in three directions. x , F y , F z ) and torque (M) x M y M z Each force plate is equipped with a microprocessor (ARM Cortex-M4 series microprocessor) responsible for data acquisition, preliminary processing, and communication control. Data transmission uses a module supporting the IEEE 802.11ac (Wi-Fi 5) protocol and integrates a hardware clock supporting the IEEE 1588-2008 (PTP) protocol for high-precision time synchronization.

[0041] Each force measuring board has a built-in rechargeable lithium polymer battery (10000mAh) and is equipped with a low-power management circuit. Each force measuring board is a square structure with a side length of 40mm, and each board has a unique physical number (such as NP01 to NP30) and MAC address. The 30 force measuring board units are laid out in a 3 (row) × 10 (column) matrix to form a large force measuring area. This area is divided into three functional areas: forward walking area (2), turning area (3), and reverse walking area (4). During laying, it is necessary to ensure that the surface of each unit is on the same horizontal plane and that the gaps are minimized. Each force measuring board is also equipped with an edge gateway. The main processor adopts an embedded AI computing module (NVIDIA Jetson Nano), which can perform calculations using an edge model. It has dual-band Wi-Fi functionality and serves as a wireless access point to communicate with all force measuring board nodes. At the same time, it connects to the cloud intelligent analysis platform through a gigabit Ethernet or 5G module.

[0042] The embedded AI computing module incorporates a lightweight one-dimensional convolutional neural network (1D-CNN) model for real-time vector event detection. The model architecture is as follows: Input layer: Six-dimensional force time-series data, window length 200ms. Convolutional layers: 3 layers, with 32, 64, and 64 filters respectively, filter size = 5, stride = 1, activation function = ReLU. Pooling layers: Each convolutional layer is followed by a max-pooling layer, window size = 2. Fully connected layers: 2 layers, with 128 and 3 neurons respectively. Output layer: Softmax function, outputting probabilities for 3 categories: normal, severe impact, foot slip. Event definition and judgment: Severe impact, rate of change of the magnitude of the resultant force vector within a 10ms time window; Foot slip, based on F... x F y The covariance of the components exceeds a set threshold within a continuous time window, and the change in the direction angle of the resultant force exceeds 30°.

[0043] The cloud-based intelligent analysis platform incorporates a dedicated large-scale biomechanical model, the core of which is a time-series network based on the Transformer architecture. The model architecture is as follows: Input embedding: A six-dimensional force sequence, mapped to a high-dimensional vector through a linear layer. Encoder: Six Transformer coding layers, each containing an eight-head self-attention mechanism and a feedforward neural network. Output head: Connected to a fully connected layer, outputting biomechanical parameters (such as ankle torque, knee flexion-extension angle, etc.). Training method: The basic model is pre-trained using massive amounts of existing gait analysis data, and then supervised fine-tuning is performed using a portion of the array data collected by this system.

[0044] In this embodiment, 30 force plates are arranged horizontally in the test area as shown in the diagram above. Upon system power-up, each force plate node starts up and automatically connects to the edge gateway via Wi-Fi. The gateway records the IDs of all online nodes, initial connections, and battery level information. When the subject steps into the force plate array area, their right foot simultaneously lands on four force plates numbered NP05, NP06, NP15, and NP16. The force of these four plates is... z If a component instantaneously exceeds the preset threshold (5N) and duration threshold (2 seconds), it indicates that they were triggered simultaneously. The triggered nodes elect NP05 as the master node for this measurement through a multi-attribute decision-making mechanism. The master node NP05 immediately initiates a precise PTP clock synchronization process with NP06, NP15, and NP16 under the coordination of the gateway, ensuring that the sampling clock deviation of all nodes within this dynamic logical cluster is less than 10 microseconds.

[0045] Then, all nodes within the logical cluster acquire data at a sampling rate of 500Hz, inputting it into the edge model every 200ms as a time window. The edge model calculates and outputs the probability values ​​of three events. If the probability of "severe impact" or "foot slip (steering)" exceeds a confidence threshold of 0.8, an event marker is immediately triggered, and a mode switching process is initiated. Definition and calculation of severe impact: The "severe impact" identified by the model corresponds to a rapid physical load change, mathematically defined as the differential value of the resultant force modulus exceeding 20 N / ms within an extremely short time (e.g., 10ms). This feature is automatically learned and extracted by the CNN model from the original six-dimensional force time-series data, comprising six channels.

[0046] The subject suddenly turned, and the master node NP05 detected F. y If the vector magnitude of the (shear force) changes drastically by more than 20N within 10ms, an instruction is immediately broadcast to all nodes in the cluster, switching to event mode. Within the following 500ms, all nodes mark the original 500Hz full data stream as high priority and continuously send it to the master node.

[0047] The master node NP05 receives data from cluster members and performs real-time spatial fusion: it vectorizes the resultant forces of the four nodes to obtain the total three-dimensional force F_total(t) of the right foot = [F x _total(t), F y _total(t), F z [_total(t)] Calculate the global pressure center coordinates and trajectory of the foot based on the torque and spatial coordinates of each node. The master node NP05 uploads the fused high-level results (total resultant force, pressure center) and the original data fragments in event mode to the edge gateway, which then forwards them to the cloud platform.

[0048] Below are some of the function names used in actual testing when a turning event occurred. The function descriptions are also marked after each function to facilitate understanding of the function (this is only for the purpose of assisting in understanding this solution and the function names below are not protected).

[0049] { "_comment": { "test_id": "A unique identifier for this test, used to uniquely identify a complete experiment or data acquisition session." "subject_id": "Subject ID, identifying which subject the data originated from." "time_segment": "The start and end timestamps of a data segment, accurate to milliseconds. It is an array, where the first element is the start time and the second element is the end time." "node_list": "A list of sensor nodes that participated in generating this data record, i.e., members of the dynamic logical cluster." "spatial_region": "The foot region corresponding to the data, which is automatically inferred or marked by the system based on the physical location of the node." "biomechanical_phase": "The gait or motion event phase corresponding to the data is automatically labeled by the system based on mechanical characteristic rules or large model analysis." "data_type": "The data type stored in this record, such as raw event data, compressed feature data, etc." "time_series": "The core six-dimensional force time series data ontology, storing the raw arrays synchronously collected by each node within a time segment." }, "test_id": "T20251001001", "subject_id": "S01", "time_segment": ["10:05:23.100", "10:05:23.600"], "node_list": ["NP05", "NP06", "NP15", "NP16"], "spatial_region": "right_forefoot", "biomechanical_phase": "cutting_maneuver", "data_type": "raw_event_data", "time_series": { "NP05": { "Fx": [0.12, 0.15, 0.18], "Fy": [-1.05, -1.08, -1.10], "Fz": [352.1, 355.6, 350.8], "Mx": [0.1, 0.2, 0.3], "My": [0.4, 0.5, 0.6], "Mz": [0.7, 0.8, 0.9] }, "NP06": { "Fx": [0.13, 0.16, 0.19], "Fy": [-1.06, -1.09, -1.11], "Fz": [353.1, 356.6, 351.8], "Mx": [0.2, 0.3, 0.4], "My": [0.5, 0.6, 0.7], "Mz": [0.8, 0.9, 1.0] } } } Finally, the cloud-based intelligent analysis platform invoked a pre-trained biomechanical model to perform in-depth analysis of the raw six-dimensional force time-series data of this turning event. The model output an analysis report stating: "At time 10:05:23.350, an abnormally high eversion torque (M) was detected on the lateral side of the right forefoot." z The peak value was approximately 25% higher than normal turning, indicating a risk of ankle inversion injury. Strengthening of the peroneal muscles is recommended. This report, along with the original, time- and space-labeled data, was presented to researchers and clinicians.

[0050] In summary, this invention employs wireless networking and an event-triggered dynamic logical cluster formation mechanism, eliminating complex wiring, supporting rapid deployment in various complex scenarios, effectively avoiding single points of failure, and ensuring the continuity of data acquisition and system reliability. Through a synchronization strategy oriented towards dynamic logical clusters, it solves the problem of multi-node timing deviations, laying the foundation for accurate mechanical analysis. Through edge models and multi-mode collaboration, the system can cope with various situations and select the most suitable transmission method, achieving intelligent scheduling and optimizing system energy efficiency. Based on a large biomechanical model, it identifies complex abnormal patterns, inverts deep joint mechanical parameters, and generates accurate personalized reports, realizing the transformation from massive data to clinical decision support, demonstrating significant progress.

[0051] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solutions based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. An intelligent networking system for a six-dimensional force sensor array, characterized in that, It includes a six-dimensional force sensor array, a wireless intelligent networking and synchronization module, an adaptive data transmission module, and a cloud-based intelligent analysis platform; The six-dimensional force sensor array comprises multiple six-dimensional force sensor nodes, each used to acquire corresponding six-dimensional force time-series data, and includes a built-in edge model for vector event detection and local feature extraction; the six-dimensional force time-series data of any six-dimensional force sensor node includes [F x(t) , F y(t) , F z(t) M x(t) M y(t) M z(t) F represents force, M represents torque, x and y correspond to two coordinate axes parallel to the six-dimensional force sensor array, and z is the coordinate axis perpendicular to the six-dimensional force sensor array. The wireless intelligent networking and synchronization module uses a six-dimensional force sensor of foot stepping as a trigger event. Each sensor node that is triggered simultaneously and is adjacent to the other will be dynamically assembled into a logical cluster. With the help of the edge model, the master node is elected within the logical cluster according to the multi-attribute decision mechanism. Then, the cluster synchronization and data aggregation are performed to obtain feature data. The adaptive data transmission module is used to dynamically switch the data transmission mode of the wireless intelligent networking and synchronization module based on the vector event detection results of the edge model. The data transmission modes include at least sleep mode, normal mode, and event mode. The cloud-based intelligent analysis platform receives and stores each transmission data from the adaptive data transmission module, and performs in-depth analysis of each transmission data through a large biomechanical model to provide feedback on foot abnormalities, foot kinematic parameters, and generate corresponding reports and charts.

2. The intelligent networking system of a six-dimensional force sensor array according to claim 1, characterized in that, The multi-attribute decision-making mechanism includes electing the master node in the following order of priority: a) maximum load; b) maximum signal strength; c) maximum remaining power; d) minimum node ID.

3. The intelligent networking system of a six-dimensional force sensor array according to claim 1, characterized in that, Each logical cluster performs intra-cluster synchronization independently, and each master node initiates wireless clock synchronization based on PTP. When data is aggregated within any logical cluster, the timing data of each non-master node is sent to the master node for spatial data fusion to obtain the corresponding total resultant force, total resultant torque, and pressure center trajectory as feature data.

4. The intelligent networking system of a six-dimensional force sensor array according to claim 1, characterized in that, Sleep mode is set to run every F for a set time period. z(t) When all values ​​are below the preset threshold I; the normal mode is when F occurs. z(t) Greater than the preset threshold I and F z(t) The rate of change exceeds the preset threshold II; the event mode is that the result of vector event detection satisfies a specific event, which includes at least severe impact and foot slippage.

5. The intelligent networking system of a six-dimensional force sensor array according to claim 4, characterized in that, The formula for determining severe impact is: ,in, Let be the resultant force vector at time t. The membrane is the resultant force vector. The membrane as the resultant force vector in the time window The average rate of change within, The threshold value corresponds to a severe impact.

6. The intelligent networking system of a six-dimensional force sensor array according to claim 4, characterized in that, Foot sliding is based on horizontal shear force ( F x ，F y The covariance and the abrupt change value of the resultant force direction angle β are used to determine the resultant force.

7. The intelligent networking system of a six-dimensional force sensor array according to claim 1, characterized in that, The adaptive data transmission module is also equipped with misoperation protection, which filters out interference triggered by false triggers based on time window, average rate of change and duration.

8. A data transmission and storage method, operating using an intelligent networking system of a six-dimensional force sensor array as described in any one of claims 1 to 7, characterized in that, Includes the following steps: S1. Dynamic Triggering and Network Synchronization: Based on the mechanical events generated by foot stomping, the corresponding multiple sensor nodes in the six-dimensional force sensor array are dynamically triggered and logical clusters are formed. Within the logical cluster, the election of the master node and the synchronization within the cluster are completed. S2, Intelligent Acquisition and Adaptive Transmission: Synchronously acquire data from multiple sensor nodes within the logical cluster, perform vector event detection based on the edge model, and adaptively switch between sleep mode, normal mode and event mode according to the detection results, and prepare to transmit data according to the corresponding mode. S3. Intra-cluster fusion and unified reporting: In normal mode or event mode, the master node receives and fuses intra-cluster data, calculates the overall mechanical parameters, and uploads the fusion results and the original data in the corresponding mode to the cloud intelligent analysis platform. S4. Semantic Storage and Intelligent Analysis: The cloud-based intelligent analysis platform calls upon a large biomechanical model to perform in-depth analysis of the uploaded data and generate corresponding feedback reports.

9. A data transmission and storage method according to claim 8, characterized in that, The edge model is a temporal neural network model based on the Transformer architecture, which performs vector event detection through a self-attention mechanism and a feedforward neural network.

Images