Intelligent Exoskeleton System Integrating Environmental Sensing and Physiological Indicator Monitoring

Abstract

This invention relates to an intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring, belonging to the field of intelligent exoskeleton protection. The solution first synchronizes multi-source sensor data in the time domain, and then inverts the theoretically expected physiological values ​​corresponding to the current work intensity based on joint movement speed and interaction torque. By calculating the physiological condition residual between the measured physiological data and the theoretical expected values, the system can dynamically isolate physiological indicator fluctuations caused by the movement itself, thereby extracting the true risk characteristics caused only by environmental risks or physical abnormalities. Based on this, it combines environmental parameters to perform correlation risk discrimination, accurately identifying specific risk types such as environmental risks or abnormal physiological indicators. Finally, the system automatically matches differentiated target impedance parameters and tactile cue signals according to the identification results, achieving active safety control and precise risk warning of the exoskeleton without interfering with high-intensity normal work.

Description

Technical Field

[0001] This invention relates to the field of intelligent exoskeleton protection, specifically to an intelligent work exoskeleton system that integrates environmental perception and physiological indicator monitoring. Background Technology

[0002] In high-load, high-risk work scenarios such as emergency rescue, disaster relief, and deep-mine operations, powered exoskeletons have become key equipment for enhancing workers' physical capabilities and pushing the limits of human physiology. However, these work environments are often characterized by complex unstructured features and potential hazards, posing a dual threat to workers from extreme physical exertion and harsh environmental conditions. Therefore, building an intelligent protection system capable of real-time perception of external environmental risks and precise monitoring of internal physiological indicators is a core requirement for achieving safe human-machine collaboration and protecting personnel's health and safety, and is of great significance for improving the safety and sustainability of special operations.

[0003] Although some existing high-end exoskeleton equipment has attempted to integrate environmental sensors or health monitoring modules, their motion control unit and health monitoring unit are usually separate in terms of physical architecture and logical algorithms, generally employing simple logic triggered by a single-channel threshold. This monitoring method based on static thresholds cannot effectively solve the problem of ambiguity and misjudgment of physiological signal sources; that is, the system struggles to decouple normal operational physiological load from abnormal environmental or pathological physiological loads from complex physiological data. Specifically, under high-intensity work conditions such as heavy-duty handling, the worker's physiological indicators, such as heart rate, will naturally increase due to metabolic demands. If the system sets a low alarm threshold, it is very easy to misjudge normal exercise load as a danger signal, leading to frequent false shutdowns of the exoskeleton and severely disrupting the normal work process. Conversely, if the threshold is raised to adapt to high-intensity work, when a worker experiences a sudden illness causing an abnormal increase in heart rate, the system often fails to detect it because it has not reached the high threshold, missing the best opportunity for early warning and protection.

[0004] Existing technologies lack a mechanism to model the correlation between energy metabolism, environmental parameters, and physiological data. They cannot invert theoretical physiological values ​​based on current exercise intensity to calculate physiological residuals, resulting in exoskeletons being unable to implement differentiated and precise protection strategies based on the true source of risk. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems existing in the prior art. According to one aspect of this application, an intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring is provided, comprising: a data fusion and synchronization module, used to perform temporal hard synchronization and sliding window resampling on the original sensor data stream to obtain a synchronization frame dataset, wherein the original sensor data stream includes exoskeleton motion parameters, environmental parameters, and human physiological parameters; an expected physiological value calculation module, used to perform dynamic metabolic consumption inversion based on a human basal metabolic rate model on the joint motion velocity and joint interaction torque in the synchronization frame dataset to obtain theoretical expected physiological values; a physiological condition residual calculation module, used to calculate the physiological condition residual index between the measured physiological data and the theoretical expected physiological values ​​in the synchronization frame dataset; a risk discrimination module, used to input the physiological condition residual index and the environmental parameters in the synchronization frame dataset into a correlation risk discrimination model to obtain a risk type enumeration value characterizing the current risk attribute; and a control command generation module, used to retrieve a pre-set differentiated active protection strategy library based on the risk type enumeration value, and calculate a target impedance parameter and tactile cue signal adapted to the current risk scenario to obtain a control command structure, wherein the control command structure is used to be sent to the joint actuators to perform physical actions.

[0006] Compared with existing technologies, this application addresses the problem of misjudgment and missed judgment caused by the inability of existing technologies to decouple normal working load from abnormal pathological environmental load. This application proposes an intelligent protection scheme based on physiological condition residual analysis. The scheme first synchronizes multi-source sensor data in the time domain. Using a human basal metabolic rate model, it inverts the theoretical expected physiological value corresponding to the current working intensity based on joint movement speed and interaction torque. By calculating the physiological condition residual between measured physiological data and theoretical expected values, the system can dynamically isolate fluctuations in physiological indicators caused by the movement itself, thereby extracting the true risk characteristics caused only by environmental risks or physical abnormalities. Based on this, it combines environmental parameters to perform correlation risk discrimination, accurately identifying specific risk types such as environmental risks or abnormal physiological indicators. Finally, the system automatically matches differentiated target impedance parameters and tactile prompts based on the identification results, achieving active safety control and accurate risk warning of the exoskeleton without interfering with high-intensity normal work. Attached Figure Description

[0007] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. Several embodiments of this application are illustrated in the drawings by way of example and not limitation, wherein: in the drawings, the same or corresponding reference numerals denote the same or corresponding parts.

[0008] Figure 1 This is a block diagram of an intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application; Figure 2 This is a schematic diagram of data flow in an intelligent work exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application; Figure 3 This is a block diagram of the expected physiological value calculation module in an intelligent work exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application; Figure 4 This is a schematic diagram of the data flow of the control command generation module in an intelligent work exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application; Figure 5 This is a schematic diagram of the overall wearing state of the exoskeleton system.

[0009] The components include: 1. Back support structure; 2. Joint actuator; 3. Assist linkage; 4. Leg connecting rod; and 5. Leg fixation assembly. Detailed Implementation

[0010] To further illustrate the technical means and effects adopted by this application in order to achieve the intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of this application is provided in conjunction with the accompanying drawings and preferred embodiments.

[0011] To address the shortcomings in the aforementioned technical fields, this application provides an intelligent exoskeleton system 100 that integrates environmental perception and physiological indicator monitoring, such as... Figure 1 and Figure 2 As shown, Figure 1 This is a block diagram of an intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application. Figure 2 This is a schematic diagram of data flow in an intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application.

[0012] Specifically, the data fusion synchronization module 110 is used to perform temporal hard synchronization and sliding window resampling on the original sensor data stream to obtain a synchronization frame dataset. The original sensor data stream includes exoskeleton motion parameters, environmental parameters, and human physiological parameters. It is understandable that when multi-source heterogeneous sensors work collaboratively in complex operating environments, they face the severe challenge of inconsistent spatiotemporal data benchmarks. The motor encoders in the exoskeleton system typically output motion states at high frequencies in the kilohertz range to ensure real-time control, while environmental gas sensors, limited by chemical reaction rates, often update data at low frequencies in the hertz range. Meanwhile, human physiological signals such as photoplethysmography (PPG) are non-stationary quasi-periodic signals and are highly susceptible to motion artifacts. If these raw data streams, with their vastly different sampling frequencies, varying noise characteristics, and transmission delays, are directly input into the subsequent risk assessment model, it will lead to temporal misalignment of the data, causing a physiological response at a certain moment to fail to correctly correspond to the physical action or environmental trigger, thus resulting in misjudgment of causality. Therefore, the data fusion synchronization module eliminates the time phase difference caused by sensor heterogeneity through time-domain hard synchronization and sliding window resampling mechanism, and constructs a unified spatiotemporal coordinate system, providing a high-fidelity data foundation for subsequent decoupling of physiological load sources and accurate identification of risk types.

[0013] In one exemplary embodiment, the data fusion synchronization module 110 includes: an original data alignment unit, used to map motion parameters and environmental parameters with different sampling frequencies in the original sensor data stream to a unified time to obtain a time alignment matrix; a physiological feature extraction and cleaning unit, used to extract physiological waveform features and clean signals from the time alignment matrix to obtain a feature enhancement matrix; and a data verification and encapsulation unit, used to perform outlier truncation verification on the feature enhancement matrix based on a physical limit threshold, and encapsulate the verified data into a synchronization frame dataset containing physiological features, torque features, motion features, and environmental features in a time sequence.

[0014] Its primary task is to process the raw sensor data stream acquired in real time. This raw sensor data stream is a collection of information transmitted from the exoskeleton system's hardware layer to the upper algorithm layer. Specifically, it includes joint positions obtained through absolute encoders, joint motion velocities obtained through differentiating circuits or observers, joint interaction torques sensed by series elastic actuators or torque sensors, acceleration information acquired by a nine-axis inertial measurement unit (IMU), raw PPG voltage values ​​acquired by a blood oxygen probe attached to the body surface, and data on the concentration of gases such as carbon monoxide and methane, as well as ambient temperature, acquired through an electrochemical sensor array. This data is transmitted concurrently via a CAN bus or UART interface. Due to the slight drift of the internal clocks of each sensor and the bus arbitration mechanism, the time when the data arrives at the processing unit often has random jitter.

[0015] The raw data alignment unit first establishes a globally unified discrete time axis. This time axis is based on a high-precision timer of the system's main control chip and adopts a hierarchical frequency architecture: the bottom-level motion control loop maintains a high-frequency operation of 1000Hz (1ms cycle) to ensure real-time response of human-machine interaction and stability of impedance control, avoiding human-machine conflict or system oscillation caused by excessive delay; while for the upper-level physiological perception and decision-making loop, the reference resampling frequency is set to 50Hz, that is, the time step between two adjacent decision timestamps is 20 milliseconds. This frequency selection satisfies the sampling theorem requirement that the energy of physiological signals such as heart rate is concentrated below 5Hz, and effectively filters high-frequency noise and reduces the computational load of complex metabolic models. To handle continuously incoming data, a first-in-first-out (FIFO) circular buffer of length S is opened inside the unit. The size of S is determined according to the maximum allowable system delay, for example, S is set to 100 frames. The raw sensor data stream is pushed into the buffer sequentially according to the order in which it arrives at the interrupt service routine, and each data point is accompanied by a local hardware timestamp when the data was generated. Subsequently, the unit traverses the heterogeneous data in the buffer and performs resampling mapping based on the global time axis. For high-frequency and physically continuously changing mechanical quantities such as joint position, interaction torque, and IMU acceleration, linear interpolation algorithms are used for downsampling or phase alignment. Linear interpolation calculates the slope of the original data on the time axis and proportionally extrapolates the value at the target time, thus ensuring the continuity and smoothness of the motion trajectory. For environmental parameters such as gas concentration and ambient temperature, which change slowly and have low sampling frequencies, a zero-order hold strategy is adopted. This is because gas diffusion and heat conduction have significant physical inertia, and the probability of abrupt numerical changes in a short period of time is extremely low. If the global time is between two sampling times of the environmental sensor, the system directly assigns the measurement value of the previous time to the current time until a new measured value appears. This processing method effectively avoids spurious fluctuations introduced by over-interpolation of low-frequency signals. For the original PPG voltage value, in order to preserve subtle features such as diaplasmic notches in its waveform for subsequent feature extraction, the alignment unit uses a cubic spline interpolation algorithm. This algorithm constructs a smooth curve between sampling points using piecewise cubic polynomials, ensuring not only the continuity of function values ​​at interpolation points but also the continuity of their first and second derivatives, thus maximizing the reproduction of the nonlinear morphology of the pulse wave. After completing all the above mappings, the data in each dimension are merged at the current time to form a row of a matrix, which is continuously generated over time, eventually forming a time-aligned matrix.

[0016] Subsequently, the physiological feature extraction and cleaning unit receives the time alignment matrix and focuses on the original PPG voltage sequence. Because the operator is in motion, the PPG signal often contains aliasing caused by baseline drift (low-frequency noise) due to muscle contraction and skin deformation, as well as environmental electromagnetic interference (high-frequency noise). To address this issue, a Butterworth bandpass filter is deployed within the unit. This filter is set to a range of 0.5Hz to 5Hz, precisely covering the human body's extreme heart rate range, i.e., 30 to 300 beats per minute. The filter's transfer function is designed to maintain a flat amplitude-frequency response in the passband to avoid attenuating the effective signal, while providing a steep attenuation rate in the stopband to filter out noise. The filtered waveform is cleaner, and the baseline returns to zero. Next, the system uses an adaptive thresholding method to extract features from the filtered PPG waveform. The system dynamically calculates the moving average of the waveform amplitude as a floating threshold; when the waveform signal crosses this threshold and passes through a local maximum detection, it is determined to be the R-wave peak moment. For example, if the system detects two consecutive R-wave peaks occurring at 10.5 seconds and 11.25 seconds on the time axis, then the time difference between the two peaks, i.e., the RR interval, is... It takes 0.75 seconds. Based on this, the instantaneous heart rate is calculated using the formula... The calculation yields a result of 80 beats per minute in this example. Simultaneously, the unit also calculates blood oxygen saturation by analyzing the ratio of the DC to AC components of the waveform. The calculated instantaneous heart rate and blood oxygen values ​​replace the original voltage data columns in the original matrix and are then recombine with other unprocessed motion and environmental data (such as directly transmitted joint torques and gas concentrations) to generate a feature enhancement matrix.

[0017] Finally, the data verification and encapsulation unit performs a rigorous logical review of the feature enhancement matrix. This step relies on a pre-set physical limit threshold library, which is constructed based on human physiological standards, sensor range manuals, and historical operational safety data statistics. For example, the set effective heart rate range threshold is [30, 220] bpm, the joint torque threshold is [-100, 100] Nm, and the ambient temperature threshold is [-20, 60] ℃. The unit scans the values ​​in the matrix one by one. If it finds that the instantaneous heart rate calculated at a certain moment is 250 bpm, which significantly exceeds the human physiological limit, the system judges that the value is an abnormal value caused by poor sensor contact or algorithm divergence. It then truncates and marks it as invalid (NaN) or fills it with the valid value of the previous frame due to inertia, preventing abnormal data from contaminating subsequent risk models. The verified and cleaned data is standardized and encapsulated according to the time sequence. Each frame of data is instantiated into a structure object containing specific attributes. This structure is the synchronization frame dataset, which is logically strictly aligned, and each data frame contains At any given moment: interpolated synchronized joint interaction torque (characterizing the torque characteristics of external load), joint movement velocity (characterizing the motion characteristics of movement intensity), measured heart rate extracted from the PPG waveform (characterizing the physiological characteristics of the body's response), and current gas concentration and ambient temperature (characterizing the environmental characteristics of the environment). For example, a typical synchronized frame might show: {Time: 10:05:20.020, Joint movement velocity: 2.0 rad / s, Interaction torque: 35 Nm, Measured heart rate: 88 bpm, CO concentration: 15 ppm, Temperature: 26℃}.

[0018] Specifically, the expected physiological value calculation module 120 is used to perform dynamic metabolic consumption inversion based on the human basal metabolic rate model on the joint movement velocity and joint interaction torque in the synchronous frame dataset to obtain the theoretical expected physiological values. Correspondingly, in high-intensity application scenarios such as heavy-duty operations and disaster relief, the physiological indicators of workers (such as heart rate and blood pressure) are mainly affected by two factors: one is the metabolic physiological load generated by muscle work, and the other is the abnormal physiological load caused by environmental stress or physical illness. Due to the complex metabolic lag effect and individual differences in the human body, simply relying on the absolute value of measured physiological data cannot accurately determine whether the current physiological state is on the verge of danger. For example, the normal heart rate rise during high-intensity exercise may be higher than the heart rate of poisoning warning at rest. Without a quantitative assessment of the normal physiological changes caused by exercise, exoskeleton protection strategies will fall into a dilemma of misjudgment and underjudgment. Therefore, the expected physiological value calculation module aims to construct a dynamic physiological twin model, using mechanical motion data to invert the physiological indicators that the human body should theoretically possess under the current working conditions, and to achieve accurate judgment of abnormal physiological states by establishing a dynamic physiological baseline.

[0019] Figure 3 This is a block diagram of the expected physiological value calculation module in an intelligent work exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application. Figure 3 As shown, in an exemplary embodiment, the expected physiological value calculation module 120 includes: a data instantaneous calculation unit 121, used to perform instantaneous calculation of exoskeleton interaction power and motion intensity on joint interaction torque and joint motion velocity in the synchronous frame dataset to obtain instantaneous interaction power and resultant acceleration magnitude; an interaction power integration unit 122, used to perform metabolic hysteresis effect integration on instantaneous interaction power based on a sliding window to obtain short-term cumulative work; and a theoretical expected physiological value generation unit 123, used to determine the theoretical expected physiological value based on the short-term cumulative work and resultant acceleration magnitude.

[0020] First, the instantaneous data processing unit 121 receives the synchronous frame dataset and analyzes the kinematic and dynamic parameters frame by frame. In human biomechanics, metabolic consumption occurs not only during muscle contraction (concentric contraction) but also during muscle braking against external forces (eccentric contraction). Therefore, calculating only traditional positive mechanical work is insufficient to characterize real human physical exertion. This unit traverses all drive joints of the exoskeleton. (Taking the hip and knee joints as examples), extract the current moment. Joint interaction torque With joint movement speed Unit application formula The instantaneous exoskeleton interaction power is calculated. In the formula, absolute value calculation is introduced to ensure that the energy exchange generated, whether the human body drives the exoskeleton movement (positive work) or the human body uses muscle strength to restrict the exoskeleton movement (negative work), is included in the total metabolic cost. For example, at the moment mentioned in the previous embodiment, the interaction torque of a single hip joint is 35 Nm, and the movement speed is 2.0 rad / s, so the instantaneous interaction power generated by this single joint is 70W. At this time, the system simultaneously collects and accumulates the instantaneous power of the four driving joints (hips and knees of both legs), and the total instantaneous interaction power of the whole body at this moment is calculated to be approximately 140 W. Simultaneously, to quantify the intensity of non-work-related body tremors or overall movement, the unit extracts the IMU three-axis acceleration components from the synchronization frame and calculates the resultant acceleration magnitude using the Euclidean norm formula. This model can capture movement characteristics such as running in place and bouncing, which do not produce significant joint displacement but consume physical energy. In this way, each frame of raw data is transformed into a representation of the intensity of muscle work on the exoskeleton at the current moment and a representation of the overall intensity of the movement.

[0021] Secondly, the interactive power integration unit 122 receives the instantaneous interactive power data stream and addresses the lag in human physiological responses. Physiological studies show that there is a time delay of tens of seconds between the onset of muscle work and the cardiopulmonary system's response (increased heart rate, increased ventilation). Instantaneous high power output does not immediately correspond to the highest heart rate; the current physiological state reflects the accumulated load over a period of time. Therefore, this unit establishes a time sliding window of length W in memory. The window length W is set according to the time constant of the human body's anaerobic and aerobic metabolic transition, with a value of 30 seconds or 60 seconds. (The system sampling period is also specified.) If the interval is 0.02 seconds, or 50Hz, then a 30-second window corresponds to N=1500 historical data frames. The unit uses a first-in-first-out (FIFO) queue to maintain these 1500 latest instantaneous interactive power values. And perform discrete integration. The calculation formula is expressed as follows: This involves performing a Riemann summation on the power within the window to obtain the short-time cumulative work. This physical quantity is the short-time cumulative work. The unit is joule (J), which eliminates the interference of instantaneous burst force on the assessment and smoothly represents the total mechanical energy output by the worker to the exoskeleton in the most recent physiological reaction cycle. For example, even if the current instantaneous power fluctuates to 140 W, if the worker's smoothed average interactive power has remained at 100 W over the past 30 seconds, the calculated short-term cumulative work is 3000 J.

[0022] Finally, the theoretical expected physiological value generation unit 123, based on the calculated short-term cumulative work and the magnitude of the resultant acceleration, uses pre-calibrated individualized parameters to invert the theoretical heart rate. In the first exemplary embodiment, the theoretical expected physiological value generation unit 123 includes: a multi-factor linear regression mapping subunit, used to perform multi-factor linear regression mapping on the short-term cumulative work and the magnitude of the resultant acceleration to obtain the heart rate increment; and an incremental heart rate superposition subunit, used to superimpose the heart rate increment with the worker's resting heart rate baseline to obtain the theoretical expected physiological value.

[0023] The multifactor linear regression mapping subunit relies on a pre-built individualized metabolic model, which takes the form of: .in, The work conversion factor, with dimensions bpm / J, represents the increase in heart rate caused by a unit of work done by a specific worker, reflecting their work efficiency. The intensity coefficient is measured in bpm / (m / s). 2 ), representing the additional heart rate burden caused by body vibration; For the predicted heart rate increment. Model parameters. , and resting heart rate benchmark This data is obtained during the calibration phase before the exoskeleton is delivered to a specific worker: the wearer is required to undergo standard tests under different working conditions, such as static, low load (e.g., walking without load), and high load (e.g., weighted squats), and the stability of the exoskeleton is recorded. , The weighting coefficients were obtained by fitting a regression equation to the actual heart rate using the least squares method. This individualized calibration effectively eliminated model bias caused by differences in age, weight, and physical fitness. Taking an adult male worker of moderate physical fitness as an example, after calibration... =0.01 bpm / J, =0.5 bpm / (m / s 2 Resting heart rate =70 bpm. The system substitutes the current operating data into the above model. Using the aforementioned data, the short-term cumulative work is calculated. =3000 J, and the current resultant acceleration magnitude The average is 2 m / s 2 (When representing slight vibrations) calculate the heart rate increment. =0.01×3000+0.5×2=31 bpm.

[0024] Subsequently, the incremental heart rate superposition subunit superimposes this increment with the worker's static baseline, i.e. =70+31=101 bpm. The output of 101 bpm is the theoretical expected physiological value. It represents the heart rate that should be at the worker's level if the worker is healthy and the environment is non-toxic, under the current working conditions of performing 3000 joules of work and accompanied by a certain amount of vibration.

[0025] In particular, in complex and ever-changing field work environments and high-intensity industrial production sites, using traditional linear models to estimate the theoretical heart rate of a human wearing an exoskeleton has significant mechanistic flaws. Linear models simply weight and sum the work done and exercise intensity, neglecting the nonlinear coupling effect of human muscles under the dual pressures of load and exercise. For example, when a worker carries a heavy load and makes rapid posture adjustments, they not only need to overcome gravity but also perform high-intensity isometric muscle contractions to counteract the inertia of the exoskeleton and environmental disturbances, leading to an exponential rather than nonlinear increase in metabolic consumption. Linear models often severely underestimate the physiological costs of such human-machine interaction, misjudging normal work-related heart rate spikes as pathological abnormalities. Furthermore, the human body is not a constant energy conversion machine. With the accumulation of work time, influenced by heat stress and dehydration, even if the work output remains constant, the baseline heart rate will naturally rise nonlinearly over time—a phenomenon known as cardiovascular drift. The fixed-coefficient model in the first embodiment cannot distinguish between the normal heart rate rise caused by physiological efficiency decline and the abnormal heart rate surge caused by hypoxia, which is very likely to cause frequent false alarms in the later stages of operation, leading to incorrect intervention behavior of the exoskeleton. To address these limitations, this application introduces a nonlinear prediction mechanism based on load-motion synergistic coupling in the preferred embodiment. By constructing a dynamic model that includes synergistic coupling factors and metabolic drift coefficients, it achieves accurate inversion of theoretically expected physiological values ​​under complex working conditions, thereby ensuring personnel safety while maintaining the continuity of operation to the greatest extent.

[0026] In a second exemplary embodiment, the theoretical expected physiological value generation unit 123 includes: The cooperative coupling factor calculation subunit is used to determine the cooperative coupling factor characterizing the complexity of human-machine confrontation conditions based on the short-term cumulative work and the magnitude of the resultant acceleration. This unit aims to quantify the complexity of human-machine confrontation conditions and capture the metabolic mutation characteristics when both load and exercise intensity are high. To mathematically describe this nonlinear process, a sigmoid activation function model is introduced as the core of the calculation. Using the aforementioned values, the unit receives the short-term cumulative work from the upstream module. 3000 J and resultant acceleration modulus 2 m / s 2 Using the formula: Calculate the cooperative coupling factor. In this formula, The dimensionless coupling factor is the output, and its value is strictly limited to the interval (0,1). Defined as the coupling gain coefficient, it is used to adjust the model's sensitivity to changes in operating conditions, for example, set to 0.002; Defined as the activation threshold, this threshold indicates that only when the product of load and exercise intensity exceeds this threshold is the system considered to be in a high-energy-consumption zone. For example, a threshold of 5000 is set based on calibration data. Substituting specific values, the product of current load and exercise intensity = 3000 × 2 = 6000. Since 6000 > 5000, the activation condition is met. The exponent for the exponent term is -0.002 × (6000 - 5000) = -2. Calculation... .therefore, =1 / (1+0.135)≈0.88. This coefficient of 0.88 indicates that there is not only a heavy load (3000J), but also accompanying vibration (2m / s²). 2 The synergistic effect of the two results in the human body being in a state of intense resistance at about 88% saturation, and the metabolic rate will be far greater than the result of simple linear addition.

[0027] The ratio calculation subunit is used to calculate the ratio between the short-term cumulative work and a preset critical fatigue work to obtain the metabolic drift coefficient, which characterizes cardiovascular drift. Secondly, the ratio calculation subunit is responsible for calculating the metabolic drift coefficient to address the problem of fatigue accumulation during long-term work. This unit introduces the critical fatigue work... This key physiological parameter represents the inflection point in physical fitness at which a specific worker enters a period of significant fatigue. =300,000 J, or 300 kJ. The unit continuously calculates the total cumulative work done from the start of the task to the present (note the difference from the short-term 3000 J). If the task has been ongoing for a longer period, the total cumulative work will be... =150,000 J (approaching half of the fatigue limit). Using the formula Calculate the drift coefficient. Where, The drift intensity constant, reflecting individual endurance level, is set to 0.2. Substituting the values, the ratio is calculated as: 150,000 / 300,000 = 0.5. (Exponential term) Drift coefficient This coefficient means that, due to the considerable physical exertion already expended, the current baseline heart rate should be 13% (1.13 times) higher than in a completely resting state under the same physical work, which is a normal manifestation of the decline in physiological function.

[0028] A dynamic nonlinear synthesis subunit is used to dynamically and nonlinearly synthesize short-term cumulative work and resultant acceleration magnitude based on a cooperative coupling factor and a metabolic drift coefficient to obtain the heart rate increment. Subsequently, the dynamic nonlinear synthesis subunit performs a three-level nested synthesis operation based on the aforementioned two key factors. This unit constructs a deeply fused prediction model, expressed by the following formula: In this formula, (Work conversion factor) and The (exercise intensity coefficient) remains the same as before, at 0.01 bpm / J and 0.5 bpm / (m / s). 2 ); The coupling penalty weight is used to define the additional metabolic proportion under severe adversarial conditions, and is set as follows: =0.5, meaning an additional 50% increase in metabolic rate. The calculation steps are as follows: 1. Basic linear component: 0.01×3000+0.5×2=31 bpm. This is the increment obtained from the traditional linear model. 2. Coupled nonlinear correction: Considering a synergistic factor of 0.88 and a weight of 0.5, the amplification factor is 1+0.5×0.88=1.44. The corrected increment is 31×1.44=44.64 bpm. This reflects the additional energy expenditure caused by the combination of heavy and fast metabolism. 3. Fatigue drift correction: Considering a fatigue coefficient of 1.13. Final heart rate increment. =44.64×1.13≈50.4 bpm.

[0029] The incremental heart rate superposition subunit is used to superimpose the heart rate increment with the worker's resting heart rate baseline to obtain the theoretically expected physiological value. Finally, the incremental heart rate superposition subunit combines this nonlinearly synthesized heart rate increment with the worker's resting heart rate baseline. (Set to 70 bpm) The values ​​are superimposed to obtain the theoretically expected physiological values. =70 + 50.4 = 120.4 bpm, rounded to 120 bpm. That is, if the first embodiment (linear model) is used, the calculation result is only 70 + 30 = 101 bpm. However, in this preferred embodiment, by introducing a coupling and drift mechanism, the predicted value reaches 120 bpm. This means that if the measured heart rate is 125 bpm, the linear model will calculate a huge residual of 24 bpm, which may be falsely reported as environmental poisoning; while the residual calculated by this model is only 5 bpm, and the system will determine that this is a normal physiological response under high-intensity work, thereby avoiding unnecessary downtime intervention. Through this deep nonlinear fusion calculation, the system can output a theoretical benchmark with extremely high physiological fidelity, realizing the precise decoupling of physiological load caused by work and physiological abnormalities caused by environment / pathology. Specifically, the aforementioned coupling gain coefficient, activation threshold, critical fatigue work, and drift intensity constant are all individualized configuration parameters determined by collecting measured metabolic data of operators under gradient loads and long-term continuous operations during the system calibration phase before the exoskeleton is delivered, and then using nonlinear least squares method to perform regression fitting and optimization on the aforementioned dynamic coupling model and fatigue drift model.

[0030] Specifically, the physiological condition residual calculation module 130 is used to calculate the physiological condition residual index between the measured physiological data and the theoretically expected physiological values ​​in the synchronous frame dataset. It is understandable that in high-intensity work scenarios involving exoskeleton human-machine coupling, measured physiological indicators are often a comprehensive response of the body to external exercise load, environmental pressure, and internal pathological states; a single absolute value lacks specificity. For example, when a worker's heart rate reaches 140 beats per minute, this could be due to normal metabolic demands from carrying 50 kg of ore while climbing a slope, or it could be due to pathological compensation occurring under other conditions while at rest. If the system judges solely based on the current heart rate value, it is highly likely to confuse these two drastically different working conditions, leading to misjudgment and shutdown when strong assistance is needed, or missed risk reports. Therefore, the physiological condition residual calculation module aims to establish a differentiated evaluation mechanism by mathematically separating the normal motion component predicted by the dynamic model from the measured total physiological load, extracting the residual component that cannot be explained by physical motion.

[0031] In one exemplary embodiment, the physiological condition residual calculation module 130 includes: a physiological difference calculation unit, used to calculate the original physiological difference between the measured heart rate data and the theoretical expected physiological value in the synchronous frame dataset; a physiological difference normalization unit, used to perform dimensionless normalization of the original physiological difference based on the individual limit based on the maximum safe heart rate threshold of the worker to obtain the normalized instantaneous residual; and a time-domain smoothing processing unit, used to perform time-domain smoothing processing on the normalized instantaneous residual based on a moving average filter to obtain the physiological condition residual index.

[0032] First, the physiological difference calculation unit, serving as the entry point for residual analysis, is responsible for receiving and processing two core data streams. One stream is the current moment data from the data fusion and synchronization module, which has been filtered, cleaned, and aligned in the time domain. Actual heart rate data The other path is the theoretical expected physiological value derived from the expected physiological value calculation module, based on the basal metabolic rate model and short-term cumulative work output. This unit performs algebraic difference operations at each time step, and the calculation formula is as follows: In this step, the sign of the difference is preserved. A positive difference, i.e. A value >0 indicates that the current actual heart rate is significantly higher than the theoretically expected value. This is due to cardiovascular overload caused by hypoxia, heat stress, or acute internal diseases (such as arrhythmias), and is a primary risk indicator of concern. A negative difference, i.e. A value less than 0 indicates that the measured heart rate is lower than the expected value based on exercise intensity. Aside from excluding reduced heart rate variability due to extreme fatigue, this often suggests potential issues such as poor sensor contact, loose wear, or drift in physiological model calibration parameters over time, falling under the category of equipment reliability anomalies. Taking the specific values ​​from the first embodiment as an example, at a certain moment, a worker is performing a handling operation. Based on the kinetic model, the theoretical expected physiological value under this condition is... It should be 101 bpm. This is the actual heart rate data measured at this time due to sudden physical discomfort or potential pathological reasons among the worker. The value rose to 135 bpm. Despite the possible minor environmental disturbances at the scene, the physiological difference calculation unit performed a subtraction operation: 135 - 101, to obtain the original physiological difference value. The value is +34 bpm. This +34 bpm value intuitively represents the absolute magnitude of the deviation between the actual situation and the theoretical model. That is, there are 34 heartbeats per minute that cannot be explained by the current handling work. This excess load is the potential source of risk.

[0033] Subsequently, the data stream enters the physiological difference normalization unit. Due to significant individual differences in human physiological functions, the same 34 bpm deviation represents drastically different levels of danger for a healthy 20-year-old and a seasoned 55-year-old worker. To eliminate the impact of individual physiological baseline differences on risk assessment and construct a universal dimensionless evaluation index, this unit introduces the maximum safe heart rate threshold for workers. This is a key parameter. It is a physiological constant characterizing an individual's cardiac pumping limit, and it can be obtained in two ways: one is through actual measurement during the exoskeleton delivery and adaptation phase using standard incremental load exercise tests (such as the Bruce protocol treadmill test); the other is based on general physiological empirical formulas when actual measurement conditions are lacking. =220-Age is used for estimation. In this example, if the worker is a 30-year-old male, the maximum safe heart rate threshold is calculated using an empirical formula. =220-30=190 bpm. The physiological difference normalization unit first extracts the original values ​​of physiological differences. The absolute value, that is This is because the focus in the risk quantification phase is on the degree of deviation rather than its direction. Subsequently, the following calculations are performed. Perform normalization. Substitute the numerical values ​​into the calculation: =|34| / 190≈0.1789. This result... It is a dimensionless floating-point number between 0 and 1 (may be greater than 1 in extreme critical situations), and it has a clear physical meaning: it represents that at the current moment, the amount of physiological deviation caused by non-motor factors (i.e., potential abnormal factors) accounts for 17.89% of the human body's maximum physiological tolerance. Compared with the absolute bpm value, this ratio can more objectively and uniformly reflect the degree of threat of risk to personal safety, and achieve the comparability of risk indicators for people of different ages and physical conditions.

[0034] Finally, the normalized instantaneous residual is fed into the time-domain smoothing unit. Because the human body's instantaneous heart rate is influenced by autonomic nervous system regulation, it naturally exhibits a certain degree of beat-by-beat variability (HRV). Furthermore, even after pre-processing cleaning, the sensor may still have occasional minor noise during strenuous exercise, leading to variations in the calculated HRV. It exhibits high-frequency jitter on a microscopic timescale. To avoid frequent oscillations in the control commands of the protection system due to instantaneous numerical jumps, such as sudden changes in the magnitude of the assist, the signal needs to be smoothed. This unit internally establishes a timescale with a duration of... For example, a 5-second moving average filtering window is set. At a sampling frequency of 50Hz, this window contains N=250 of the most recent normalized instantaneous residual samples. The time-domain smoothing unit uses a recursive moving average algorithm to improve computational efficiency. Let the output at time t be... The input is The calculation logic is as follows: The arithmetic mean within the window is calculated by continuously moving in the latest data and removing the oldest data. For example, if the normalized instantaneous residual fluctuates between 0.17 and 0.19 within the current 5-second window, after smoothing, the resulting physiological condition residual index... It will stabilize at around 0.18. When the index is close to 0, it indicates that the measured heart rate is highly consistent with the theoretical heart rate, and the body is in a fully controlled and normal working state; when the index (such as 0.18 in this example) deviates significantly from 0, it clearly indicates the presence of abnormal non-exercise load.

[0035] Specifically, the risk discrimination module 140 is used to input the physiological condition residual index and environmental parameters from the synchronous frame dataset into the correlation risk discrimination model to obtain risk type enumeration values ​​representing the current risk attribute. It should be understood that simply obtaining the physiological condition residual index, while revealing that the human body is in an abnormal physiological load state, only answers the question of whether there is an abnormality, not the crucial question of where the abnormality originates. In complex work scenarios, abnormal fluctuations in physiological indicators may originate from internal pathological emergencies or from acute external environmental attacks. These two drastically different causes correspond to completely opposite protection logics: the former requires the exoskeleton to immediately provide high-damping support to prevent falls due to personnel disability, while the latter may require the exoskeleton to enter a zero-impedance mode to assist personnel in quickly evacuating the danger zone. Without accurate identification of the source of the abnormality, the control system will be unable to issue appropriate tactical commands. Therefore, the risk discrimination module is introduced by fusing and analyzing the physiological residual representing the internal state with environmental parameters representing the external conditions, using a logic inference engine to decouple the causal relationship of risk at the data level, thereby outputting risk types with clear physical orientation.

[0036] In one exemplary embodiment, the risk discrimination module 140 includes: a risk state vector construction unit, used to calculate the exceedance rate and feature fusion of gas concentration and ambient temperature in the synchronous frame dataset to obtain a normalized environmental severity index, and to construct a risk state vector together with the physiological condition residual index; a physiological environment logic decision unit, used to input the risk state vector into a logic decision unit containing a hysteresis comparator to map the physiological component and the environmental component into binary logic state tuples representing whether they are abnormal; and a type determination unit, used to perform truth table matching reasoning based on the logic state tuples, and to determine the type as environmental injury when both physiological logic and environmental logic are abnormal, as sudden illness when only physiological logic is abnormal, and as environmental warning when only environmental logic is abnormal, so as to generate a risk type enumeration value.

[0037] First, the risk state vector construction unit is activated. This unit first reads the occupational safety standard thresholds stored in the system's non-volatile memory. This threshold library is pre-collected based on national occupational health standards and industry-specific safety regulations. Specifically, for gas concentrations, either the time-weighted average permissible concentration (TWA) or the short-term exposure limit (STEL) is set as the limit threshold. Regarding ambient temperature, a heat stress threshold was set according to the high-temperature operation classification standard. For example, the short-term exposure limit for carbon monoxide (CO) is set at 30 mg / m³. 3 (Approximately 24 ppm), with an ambient temperature warning threshold of 35°C. The unit accesses the real-time gas concentration in the synchronization frame dataset. With ambient temperature To eliminate the influence of different physical dimensions and comprehensively evaluate the severity of the environment, the element uses the weighted root mean square method for feature fusion. The calculation formula is as follows: In this formula, and These are dimensionless weighting coefficients that sum to a value of 1. The weights are set based on the priority of the hazard factors in the work scenario. For example, in a gas mine, the risk of death from gas toxicity is far higher than that from high temperature; therefore, a weight of 1 is set. =0.8, =0.2; however, in outdoor high-temperature handling scenarios, it may be set to... =0.3, =0.7. For example, if the CO concentration in the current synchronization frame is 15 ppm and the ambient temperature is 26℃, with each weighted at 0.5, the system first performs dimensionless processing: the exceedance rate of the gas is 15 / 24 = 0.625, and the exceedance rate of the temperature is 26 / 35 ≈ 0.74. Substituting these values ​​into the formula, the normalized environmental severity index is calculated: The calculated result is the Normalized Environmental Severity Index, which indicates that the current environment is at a moderately high stress level, but has not yet reached the critical threshold of complete danger. Subsequently, the unit will use the physiological condition residual index output from the previous module. By combining this with the environmental severity index, a current two-dimensional risk state vector is constructed. This vector accurately locates the current human-loop state point in a two-dimensional Cartesian coordinate system.

[0038] Next, the physiological environment logic decision unit receives the risk state vector and is responsible for mapping the continuous analog quantity into a discrete binary logic state. To prevent frequent system state jumps due to small fluctuations in the signal near the threshold, i.e., the ping-pong effect, this unit integrates a hysteresis comparator logic, which has a high threshold and a low threshold. For the physiological component (0.18) in the vector, its high threshold is set to 0.15, i.e., a residual exceeding 15% is considered a serious anomaly, and the low threshold is set to 0.10. The decision logic is as follows: if the current value is greater than 0.15, output logic 1 (abnormal); if the current value is less than 0.10, output logic 0 (normal); if it is between 0.10 and 0.15, the logic state of the previous moment remains unchanged. In this embodiment, the input physiological residual is 0.18. The previous state was normal (0), and since 0.18 is within the dead zone and has not exceeded the high threshold, the logic state remains 0; however, if the previous state was already abnormal (1), then 0.18 has not yet fallen below the low threshold, and the logic state remains 1. To illustrate the typical risk assessment process, we assume that the physiological residual has just exceeded 0.15 due to continuous increases, therefore the physiological logical component is judged as True (1). Similarly, for the environmental component (0.63), we set the high threshold of the environment to 0.8, and the low threshold to 0.6, as the environment has reached a severe level due to various factors. Since 0.68 < 0.8, the environmental logical component is judged as False (0). Finally, this unit outputs a binary logical state tuple: (physiological = 1, environmental = 0). This tuple clearly expresses that there is a serious abnormality in the body, but the external environment is relatively safe.

[0039] Finally, the type determination unit performs truth table matching reasoning based on the logical tuple. The truth table is a logical lookup matrix pre-installed in the microcontroller code, which exhaustively lists all possible logical combinations and their corresponding risk semantics. The specific implementation logic of the unit is as follows: 1. Input tuple (1,1): Physiological abnormality and environmental abnormality. The system determines it as [environmental injury type]. This means that toxic gas or high temperature in the environment directly caused the deterioration of the human body's physiological indicators, and the causal relationship is clear. The strategy to be executed at this time is to evacuate immediately. 2. Input tuple (1,0): Physiological abnormality but normal environment. This is the situation in this embodiment. The system determines it as [sudden illness type]. Logical deduction shows that since there are no significant harmful factors in the external environment (environmental index is not exceeded), and the physiological residual after removing the exercise load is still serious, the only explanation is that the worker's own health system has experienced a primary failure, such as a heart attack, heatstroke, or fainting. This is an extremely critical judgment, which directly guides the exoskeleton to increase damping to support the user who may lose consciousness. 3. Input tuple (0,1): Physiological normality but environmental abnormality. The system determines this to be an [Environmental Warning Type]. This means the environment has deteriorated, but the human body has not yet shown significant physiological reactions or the reactions are delayed. An early warning should be issued to prevent harm. 4. Input tuple (0,0): Physiologically normal and environmentally normal. The system determines this to be a [Safe Type]. In this embodiment, the type determination unit matches tuple (1,0), therefore locking the output result as a sudden illness.

[0040] Specifically, the control command generation module 150 is used to retrieve a pre-set differentiated active protection strategy library based on the risk type enumeration value, and calculate the target impedance parameters and tactile prompt signals adapted to the current risk scenario to obtain a control command structure. The control command structure is used to be sent to the joint actuator to perform physical actions. That is, in the control loop of most intelligent exoskeletons, accurate risk identification only completes the cognitive process, while the final line of defense for ensuring the personal safety of workers lies in the correct execution of physical actions. Since risk sources with different attributes correspond to diametrically opposed physical assistance needs—for example, hazardous factors in the environment require workers to be able to run and evacuate quickly without restraint, which requires the exoskeleton to minimize movement resistance; while sudden cardiogenic syncope of workers requires the exoskeleton to immediately provide high-strength rigid support or damping buffer to prevent secondary injuries caused by falls due to disability—therefore, a single shutdown protection or fixed assistance mode cannot adapt to dynamically changing extreme working conditions. To this end, the control command generation module retrieves differentiated active protection schemes from the pre-set strategy library based on the specific risk type determined in the previous step, and transforms the abstract risk semantics into virtual impedance parameters, torque control quantities, and tactile feedback signals that are adapted to the current scenario, thereby ensuring that the joint actuator can make the optimal physical response that meets the current survival requirements.

[0041] Figure 4 This is a schematic diagram of the data flow in the control command generation module of an intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to an embodiment of this application. Figure 4 As shown, in an exemplary embodiment, the control command generation module 150 includes: a pattern matching unit, used to retrieve a strategy lookup table based on the risk type enumeration value, matching zero-impedance mode parameters when the risk type is environmental damage, and matching high-damping support mode parameters when the risk type is sudden illness, to generate a target impedance configuration including virtual stiffness coefficient and virtual damping coefficient; an assist torque calculation unit, used to perform adaptive torque calculation based on the current joint posture data in the synchronization frame dataset based on the target impedance configuration to obtain the assist torque; a tactile vibration waveform synthesis unit, used to synthesize the corresponding tactile vibration waveform based on the risk type enumeration value; and a structure encapsulation unit, used to encapsulate the tactile vibration waveform, the target impedance configuration, and the assist torque together to obtain a control command structure.

[0042] First, the pattern matching unit initiates the strategy retrieval mechanism. This unit internally stores a pre-defined strategy lookup table. This lookup table is based on a parameter database constructed from numerous biomechanical experiments and safety specifications, covering the mechanical impedance characteristics that the exoskeleton should exhibit under different hazardous conditions. In this table, each risk type (such as safety, environmental warning, environmental injury, and sudden illness) is indexed to a specific set of control parameter tuples, including virtual stiffness coefficients. Virtual damping coefficient and gravity compensation gain Following the judgment results of the aforementioned embodiment, the current risk type enumeration value is sudden illness, indicating that the worker is about to lose control of their body due to physiological reasons. After receiving this enumeration value, the pattern matching unit traverses the lookup table for matching. If the input is an environmental hazard type, the system will match a zero-impedance pattern. and Set all to 0, and Setting it to 0 aims to eliminate the back electromotive force of the motor and the friction of the gearbox, allowing the exoskeleton to be in a completely transparent, responsive state, minimizing the mechanical resistance of the human-machine system, and facilitating full-speed escape for personnel. If the input is a safety or environmental warning type, then a transparent assist mode is matched, with a medium stiffness setting. and low damping The medium stiffness is set based on the exoskeleton linkage dynamics model as a gravity compensation stiffness to counteract the mechanical arm's own gravitational torque, while the low damping is calibrated through frequency response experiments as the minimum critical damping value that can suppress closed-loop oscillations of the control system without producing a viscous drag sensation during normal joint movement. In the sudden illness scenario of this embodiment, the unit is matched to a gravity compensation + high damping support mode. In this mode, to prevent the person from falling, the unit retrieves high damping parameters. Setting it to 50 Nms / rad increases the viscous resistance of joint rotation, acting as a hydraulic damper; simultaneously, to support a person who has lost muscle strength, the unit activates the maximum gravity compensation gain. If set to 1.5, it provides a counterforce 1.5 times the limb's weight. Finally, the pattern matching unit generates an intermediate object containing the target control law parameters, i.e., the target impedance configuration. Here, virtual stiffness will be used. Set to 0 or a specific lock value, depending on the specific support strategy, such as whether the joint needs to be locked at the current angle.

[0043] Subsequently, the data stream enters the assist torque calculation unit. In one exemplary embodiment, the assist torque calculation unit includes: an adaptive torque calculation subunit, used to perform adaptive torque calculation based on the current joint posture data in the synchronization frame dataset based on the target impedance configuration to obtain a calculated torque value; and a peak limiting subunit, used to perform motor peak limiting processing on the calculated torque value to obtain the assist torque.

[0044] The adaptive torque calculation subunit is responsible for converting the abstract impedance parameters mentioned above, combined with the current motion state, into specific motor current commands. The unit reads the current joint posture data, including joint angles, from the synchronization frame dataset in real time. and speed of movement For example, at the current moment, due to the worker's fainting, the knee joint is involuntarily flexing (kneeling tendency) at a speed of 2.0 rad / s, that is... The current velocity is -2.0 rad / s, and the current angle is 30 degrees. The element substitutes these state variables into the impedance control equation: In response to sudden illness patterns, to prevent the body from collapsing, the system will adjust the balance point. The joint angle is dynamically set to the previous moment, attempting to maintain the current state. The first term in the formula is the gravity compensation term: The gravitational torque at the current posture is calculated based on the mass distribution of the exoskeleton links and an anthropometry model, which is 20 Nm at this time, multiplied by the gain. =1.5, resulting in an active lifting force of 1.5 × 20 = 30 Nm. This torque is greater than gravity, generating a net upward lift. For the sudden illness mode, since the virtual stiffness coefficient is... The pattern matching unit is set to 0 to avoid secondary damage caused by mechanical stiffness locking. Therefore, the second term of the formula (stiffness recovery term) does not contribute torque, and the equilibrium point is... The settings are temporarily ineffective in this mode. The system primarily relies on the counteracting torque generated by the third damping term, which is opposite in direction to the motion to resist buckling. The upward damping torque is -50 × (-2.0) = 100 Nm. The negative sign indicates that the torque direction is opposite to the velocity direction, generating a drag torque of up to 100 Nm to counteract the rapid bending of the knee joint. The calculated torque value is obtained by superimposing the various torque components. With a value of 130 Nm, this instruction is designed to powerfully slow down the collapse velocity of a human body, acting like an invisible wall. Following this, the peak limiting subunit calculates... A safety check is performed. Since the calculated theoretical damping torque may be very large, such as the aforementioned 130 Nm, it may exceed the physical limits of the motor or reducer. The unit reads the preset rated peak value of the motor. This threshold is determined based on the motor's torque-current characteristic curve and the rated destructive torque of the harmonic reducer, for example, set to 80 Nm. If the calculated value of 130 Nm exceeds 80 Nm, the unit will force a hard cutoff, outputting a final assist torque of 80 Nm, ensuring maximum protection without damaging the actuator.

[0045] Meanwhile, the tactile vibration waveform synthesis unit operates in parallel. This unit synthesizes tactile feedback signals to notify workers based on risk type enumeration values. Different signal waveform designs follow human tactile psychology characteristics to ensure accurate information transmission. For environmental hazards, to instantly trigger a fight-or-flight response in noisy industrial environments and mask background noise, the unit synthesizes continuous high-frequency pulses, such as 200Hz square waves with a 50% duty cycle, simulating a strong electric shock to prompt escape. For environmental warnings, given that the environment is not yet lethal but needs to alert workers, the unit synthesizes intermittent mid-frequency pulses, such as 50Hz, striking three times per second, simulating the sensation of a finger lightly tapping the shoulder, prompting workers to pay attention to changes in the surrounding environment or check instrument readings. For safety hazards, to avoid sensory desensitization caused by the "boy who cried wolf" effect, the system outputs a zero-amplitude silence signal, effectively shutting down the vibration motor, ensuring workers are not disturbed during normal operations. However, in the scenario of a sudden illness in this embodiment, considering that the worker may be in a state of confusion or extreme physical vulnerability due to pathological reasons, and at this time they should not be frightened but rather soothed or awakened, the unit synthesizes intermittent low-frequency pulses, such as a 1Hz sine wave, similar to the rhythm of deep breathing, which are applied to the vibration motor in contact with the skin to provide a gentle but firm tactile cue, thus obtaining a tactile vibration waveform. Finally, the structure encapsulation unit summarizes the results of the above parallel processing. It packages the generated target impedance configuration (for the closed-loop adjustment of the underlying controller), the limited boost torque (as feedforward torque), and the synthesized tactile vibration waveform (for the tactile driver) according to a specific communication protocol format. For example, it encapsulates it into a control instruction structure containing a Header, Payload, and CRC (checksum). The Payload specifically includes: {Mode:MEDICAL_SUPPORT,Torque_FF:80.0,Damping:50.0,Vibration:PATTERN_ISO_1HZ}. The command instructs the exoskeleton to immediately enter emergency medical support mode, using motors to output a powerful 80 Nm feedforward torque to support the human body, while simultaneously applying high damping of 50 Nm / s / radian to buffer the impact of falling. This is combined with a 1 Hz breathing-style vibration alert, thus providing proactive reassurance and fall protection for the incapacitated worker. This structure is then pushed to the real-time bus. Upon receiving the command, the exoskeleton's hip and knee joint motors instantly increase their output torque and motion damping, effectively supporting the worker about to fall, while the vibration unit emits a low-frequency warning.

[0046] like Figure 5 The diagram shown illustrates the overall wearing state of the intelligent exoskeleton system described in this application. The exoskeleton system is worn by the worker, forming a human-machine integrated protective work unit.

[0047] Specifically, the back support structure 1 is attached to the back area of ​​the operator and fixed to the torso by a strap mechanism, providing a stable support benchmark for the upper body. It also houses core processing units such as the data fusion synchronization module 110, the expected physiological value calculation module 120, the physiological condition residual calculation module 130, the risk assessment module 140, and the control command generation module 150. The joint actuator 2 is located on both sides of the hip and integrates a servo motor, harmonic reducer, absolute encoder, and torque sensor. It is the core physical execution unit for achieving impedance control and assist output. The assist linkage 3 connects the back support structure 1 and the joint actuator 2, transmitting the back support force to the hip joint. The leg connecting rod 4 extends downwards to the thigh. Its upper end is connected to the output shaft of the joint actuator 2 via a hinge mechanism, and its lower end is constrained and fixed to the thigh by the leg fixing component 5, thus establishing a kinematic coupling relationship between the human lower limbs and the exoskeleton mechanical structure.

[0048] When the wearer is in wearable mode, when the worker performs actions such as bending over, lifting, or squatting, the leg connecting rod 4 moves relative to the thigh, driving the angle encoder and torque sensor inside the joint actuator 2 to collect the joint movement speed and joint interaction torque in real time. These kinematic and dynamic data are processed by the data fusion synchronization module 110 to form a synchronization frame dataset, providing the basic input for subsequent calculation of expected physiological values, physiological condition residual analysis, and risk assessment. When the system identifies abnormal risks such as sudden illness or environmental injury through the risk assessment module 140, the control command structure generated by the control command generation module 150 is sent to each joint actuator 2 in real time. By precisely adjusting the motor output torque and virtual impedance parameters, high-damping support protection or zero-impedance transparent follow-up is achieved for the worker, ensuring personal safety and work continuity in complex and changing working environments.

[0049] In summary, the intelligent exoskeleton protection method integrating environmental perception and vital sign monitoring based on the embodiments of this application is clarified, addressing the problem of misjudgment and missed judgment caused by the inability of existing technologies to decouple normal work load from abnormal pathological environmental load. This solution first synchronizes multi-source sensor data in the time domain, and uses a human basal metabolic rate model to invert the theoretical expected physiological value corresponding to the current work intensity based on joint movement speed and interaction torque. By calculating the physiological condition residual between measured physiological data and theoretical expected values, the system can dynamically isolate physiological index fluctuations caused by the movement itself, thereby extracting the true risk characteristics caused only by environmental risks or physical abnormalities. Based on this, it combines environmental parameters to perform correlation risk discrimination, accurately identifying specific risk types such as environmental risks or abnormal physiological indicators. Finally, the system automatically matches differentiated target impedance parameters and tactile prompts based on the identification results, achieving active safety control and accurate risk warning of the exoskeleton without interfering with high-intensity normal work.

[0050] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring, characterized in that, include: The data fusion and synchronization module is used to perform temporal hard synchronization and sliding window resampling on the raw sensor data stream to obtain a synchronization frame dataset. The raw sensor data stream includes exoskeleton motion parameters, environmental parameters, and human physiological parameters. The expected physiological value calculation module is used to perform dynamic metabolic consumption inversion based on the human basal metabolic rate model on the joint motion velocity and joint interaction torque in the synchronous frame dataset to obtain the theoretical expected physiological value. The physiological condition residual calculation module is used to calculate the physiological condition residual index between the measured physiological data and the theoretical expected physiological values ​​in the synchronous frame dataset; The risk discrimination module is used to input the physiological condition residual index and environmental parameters in the synchronous frame dataset into the correlation risk discrimination model to obtain the risk type enumeration value that represents the current risk attribute; The control command generation module is used to retrieve a pre-set differentiated active protection strategy library based on the risk type enumeration value, and calculate the target impedance parameters and tactile prompt signals adapted to the current risk scenario to obtain a control command structure. The control command structure is used to be sent to the joint actuator to perform physical actions.

2. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 1, characterized in that, The data fusion and synchronization module includes: The raw data alignment unit is used to map motion parameters and environmental parameters with different sampling frequencies in the raw sensor data stream to a unified time to obtain a time alignment matrix. The physiological feature extraction and cleaning unit is used to extract physiological waveform features and clean the signal from the time alignment matrix to obtain the feature enhancement matrix. The data verification and encapsulation unit is used to perform outlier truncation verification on the feature enhancement matrix based on the physical limit threshold, and encapsulate the verified data into a synchronous frame dataset containing physiological features, torque features, motion features and environmental features in a time series.

3. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 1, characterized in that, The expected physiological value calculation module includes: The instantaneous data calculation unit is used to perform instantaneous calculation of exoskeleton interaction power and motion intensity on the joint interaction torque and joint motion velocity in the synchronous frame dataset to obtain the instantaneous interaction power and resultant acceleration magnitude. The interactive power integration unit is used to integrate the instantaneous interactive power based on the metabolic hysteresis effect through a sliding window to obtain the short-term cumulative work done. The theoretical expected physiological value generation unit is used to determine the theoretical expected physiological value based on the short-term cumulative work done and the resultant acceleration magnitude.

4. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 3, characterized in that, The theoretically expected physiological value generation unit includes: The multi-factor linear regression mapping subunit is used to perform multi-factor linear regression mapping on the short-term cumulative work and the resultant acceleration magnitude to obtain the heart rate increment. The incremental heart rate superposition subunit is used to superimpose the heart rate increment with the worker's resting heart rate baseline to obtain the theoretical expected physiological value.

5. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 3, characterized in that, The theoretically expected physiological value generation unit includes: The cooperative coupling factor calculation subunit is used to determine the cooperative coupling factor characterizing the complexity of human-machine confrontation conditions based on the short-term cumulative work and the resultant acceleration magnitude. The ratio calculation subunit is used to perform a ratio calculation between the short-term cumulative work and the preset critical fatigue work to obtain the metabolic drift coefficient, which characterizes the cardiovascular drift characteristics. A dynamic nonlinear synthesis subunit is used to perform dynamic nonlinear synthesis of short-term cumulative work and resultant acceleration magnitude based on the cooperative coupling factor and metabolic drift coefficient to obtain the heart rate increment. The incremental heart rate superposition subunit is used to superimpose the heart rate increment with the worker's resting heart rate baseline to obtain the theoretical expected physiological value.

6. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 1, characterized in that, The physiological condition residual calculation module includes: The physiological difference calculation unit is used to calculate the raw physiological difference between the measured heart rate data and the theoretical expected physiological value in the synchronous frame dataset. The physiological difference normalization unit is used to perform dimensionless normalization of the original physiological difference values ​​based on individual limits, based on the maximum safe heart rate threshold of the workers, to obtain the normalized instantaneous residual. The time-domain smoothing unit is used to perform time-domain smoothing on the normalized instantaneous residuals based on the moving average filter to obtain the physiological condition residual index.

7. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 1, characterized in that, The risk assessment module includes: The risk state vector construction unit is used to calculate the exceedance rate and feature fusion of gas concentration and ambient temperature in the synchronous frame dataset to obtain the normalized environmental severity index, and to construct the risk state vector together with the physiological condition residual index. The physiological and environmental logic decision unit is used to input the risk state vector into the logic decision unit containing the hysteresis comparator to map the physiological component and the environmental component into binary logic state tuples representing whether they are abnormal. The type determination unit is used to perform truth table matching reasoning based on logical state tuples. When both physiological logic and environmental logic are abnormal, it is determined to be an environmental injury type. When only physiological logic is abnormal, it is determined to be a sudden illness type. When only environmental logic is abnormal, it is determined to be an environmental warning type, so as to generate risk type enumeration values.

8. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 1, characterized in that, The control command generation module includes: The pattern matching unit is used to retrieve the strategy lookup table based on the risk type enumeration value. When the risk type is environmental damage, it matches the zero impedance mode parameter, and when the risk type is sudden illness, it matches the high damping support mode parameter to generate a target impedance configuration that includes virtual stiffness coefficient and virtual damping coefficient. The assist torque calculation unit is used to perform adaptive torque calculation based on the current joint posture data in the synchronization frame dataset to obtain the assist torque based on the target impedance configuration. The tactile vibration waveform synthesis unit is used to synthesize the corresponding tactile vibration waveform based on the risk type enumeration value. The structure encapsulation unit is used to encapsulate the tactile vibration waveform, target impedance configuration, and assist torque to obtain the control command structure.

9. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 8, characterized in that, The assist torque calculation unit includes: The adaptive torque calculation subunit is used to perform adaptive torque calculation based on the current joint posture data in the synchronization frame dataset based on the target impedance configuration to obtain the calculated torque value. The peak limiting subunit is used to perform motor peak limiting processing on the calculated torque value to obtain the assist torque.

10. The intelligent exoskeleton system integrating environmental perception and physiological indicator monitoring according to claim 1, characterized in that, Also includes: The back support structure is used to fix the back of the worker and carry the data fusion and synchronization module, the expected physiological value calculation module, the physiological condition residual calculation module, the risk judgment module, and the control command generation module. A hip fixation structure is installed around the worker's hips; The joint actuator is located on both sides of the hip fixation structure. It integrates a motor, reducer and torque sensor to receive control commands and output assist torque. The assist linkage connects the back support structure to the joint actuator; The leg connecting rod is hinged at the upper end to the output shaft of the joint actuator, and at the lower end is constrained and fixed to the worker's thigh by the leg fixing component, which is used to establish the kinematic coupling relationship between the human lower limb and the mechanical structure of the exoskeleton.

Images