Fatigue state recognition-based self-adaptive resistance adjustment method for human power generation vehicle

Abstract

The invention discloses a human-powered generator car self-adaptive resistance adjusting method based on fatigue state recognition, which comprises the following steps: monitoring pedal pressure and rotating speed signals of a user in real time, recognizing a fatigue state by using a multi-dimensional physiological feature extraction and gradient recursive analysis algorithm, and automatically adjusting resistance based on a self-adaptive PID control strategy. Therefore, the exercise time is prolonged and the generating capacity is increased. The method comprises the following steps: step 1, data acquisition and power calculation; step 2, fatigue feature extraction and index modeling; 3, judging a dynamic fatigue state; step 4, performing optimal power target calculation and dynamic adjustment; step 5, performing adaptive PID control and resistance execution; and step 6, performing data feedback and adaptive continuous optimization.

Description

Technical Field

[0001] This invention belongs to the field of fitness equipment control technology, specifically relating to an adaptive resistance adjustment method for a human-powered generator based on fatigue state recognition. Background Technology

[0002] Exercise bikes, also known as stationary bikes, are a common and traditional form of fitness equipment. They typically use resistance components to provide resistance and achieve the desired power output. A common type of exercise bike in gyms today is the constant resistance type, where the resistance is the force exerted while riding; the faster the riding speed, the greater the power output. Higher riding speeds result in greater power consumption, and lower riding speeds result in less power output.

[0003] Currently, exercise bikes use brake pads, magnetic attraction, or magnetic force generated by coils and rollers to create resistance. These methods cannot sense the user's current riding status. When the user feels fatigued after riding for a while, the resistance cannot be automatically reduced, forcing the user to stop training. However, during exercise, fat burning only begins when a certain intensity is reached, body temperature rises, and muscles feel particularly fatigued. This invention intelligently and automatically reduces resistance during the fat-burning phase when the user is too fatigued to continue, allowing the user to persevere and achieve better training results.

[0004] Therefore, the existing human-powered generators that use manual resistance adjustment have the following problems: (1) Users often adjust the resistance only after fatigue has occurred; (2) There is a lack of objective fatigue assessment indicators; (3) Frequent manual adjustment will interrupt the rhythm of exercise; (4) It is impossible to dynamically optimize the resistance according to the user's real-time status, resulting in short exercise time and low power generation. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the present invention aims to provide an adaptive resistance adjustment method for a human-powered generator based on fatigue state recognition. By monitoring the user's pedal pressure and speed signals in real time, the method uses multi-dimensional physiological feature extraction and gradient recursive analysis algorithms to identify fatigue state, and automatically adjusts the generator resistance based on an adaptive PID control strategy, so that the user is always kept in the optimal exercise range, thereby extending exercise time and increasing power generation.

[0006] To achieve the above objectives, the present invention can be implemented using the following specific technical solutions: The adaptive resistance adjustment method for a human-powered generator based on fatigue state recognition includes the following steps: Step 1, Data Acquisition and Power Calculation: Acquire pedal pressure using a pedal pressure sensor. The rear wheel speed is collected by a Hall effect speed sensor. The data is filtered to remove noise and outliers, and the instantaneous output power is calculated. ; Step 2, Fatigue Feature Extraction and Exponential Modeling: Extract the power attenuation coefficient based on the data from Step 1. Pressure variation coefficient Rotational speed attenuation coefficient As a fatigue characteristic, the second-order power gradient is calculated. And establish a fatigue index model to calculate the fatigue index. ; Step 3, Dynamic Fatigue State Determination: Calculate the dynamic threshold that varies with exercise time and accumulated work. ,when Greater than the dynamic threshold and When the value is less than the negative threshold, the state is determined to be fatigued, and a fatigue flag is output. ; Step 4: Calculation and Dynamic Adjustment of Optimal Power Target: Based on the Power Generation Efficiency Function and fatigue risk function Calculate the optimal power Adjust the target power based on fatigue indicators. According to fatigue state Reduce target power; Step 5, Adaptive PID Control and Resistance Execution: Calculate the error between the target power and the actual power. ,according to Calculate the adaptive proportional gain, use a PID controller to calculate the control quantity and map it to the resistance level, and generate a PWM signal to drive the electromagnetic resistance coil. Step 6, Data Feedback and Adaptive Continuous Optimization: Feed back the power, fatigue index, resistance level and cumulative power generation to the display terminal, and repeat steps 1-5; when the cumulative data reaches the threshold, re-optimize the fatigue index model weight coefficients, efficiency function and risk function.

[0007] Furthermore, in step 1, the pedal pressure is collected by a pedal pressure sensor. The rear wheel speed is collected by a Hall effect speed sensor. Kalman filtering was applied to the pedal pressure and rear wheel speed signals for noise reduction. The criteria remove outliers and normalize to... Range; based on the collected pressure and rotation speed, calculate the instantaneous output power during the user's movement process.

[0008] Furthermore, in step 2, the fatigue characteristics include the power attenuation coefficient. Pressure variation coefficient and speed decay coefficient The fatigue index model is as follows: , Weighting coefficient Optimization using least squares: ,in, Let be the predicted fatigue index value for the i-th sample. Let i be the experimental calibration value for the i-th sample. This represents the number of historical samples.

[0009] Furthermore, in step 3, firstly, the dynamic fatigue judgment threshold is calculated; then, a judgment is made: when both conditions are met... and When the condition is determined to be fatigued, output... Otherwise, output .

[0010] Furthermore, in step 4, a power generation efficiency function is fitted based on historical data. and fatigue risk function Calculate the optimal power; and adjust the target power according to the fatigue state: when hour, ,in, Value ;when hour, .

[0011] Furthermore, in step 5, the power error is used... and adaptive proportional gain Calculate PID control input Map the control input to the resistance level: The resistance change rate is limited to ≤ 2 gears / second, when the rotational speed... Forced ;in, For output resistance settings, This is the minimum resistance setting. This is the gear with the greatest resistance. This is the theoretical minimum value of the control quantity. This represents the theoretical maximum value of the control quantity.

[0012] Furthermore, in step 5, the PWM duty cycle is calculated based on the resistance level. A PWM signal with a frequency of 1kHz is generated to drive the electromagnetic reluctance coil.

[0013] Furthermore, in step 6, steps 1-6 are executed cyclically with a control period of 0.1-0.5 seconds; complete data for each movement is recorded, and when the accumulated data reaches a threshold, the weighting coefficients in step 2 are re-optimized. And the efficiency curve in step 4 Risk function .

[0014] Compared with the prior art, the present invention has the following advantages: This invention achieves early fatigue identification through multi-dimensional feature and gradient analysis, and adopts an adaptive PID strategy to dynamically adjust resistance according to the degree of fatigue, thereby extending exercise time and directly increasing power generation. This invention can automatically control and improve the exercise experience, and supports personalized model adaptive continuous optimization. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] like Figure 1 As shown, the present invention provides an adaptive resistance adjustment method for a human-powered generator based on fatigue state recognition, comprising the following steps: (a) Step 1: Data acquisition and power calculation.

[0018] The collected data includes pedal pressure collected via a pedal pressure sensor. (unit: sampling frequency And the rear wheel speed is collected via a Hall effect speed sensor. (unit: sampling frequency Kalman filtering was applied to the pedal pressure and rear wheel speed signals for noise reduction. The criteria remove outliers and normalize to... Interval.

[0019] Based on the collected pressure and rotation speed data, the user's instantaneous output power is calculated. : , in, The pressure (in N) applied by the user to the pedal at time t. Effective radius of the rear wheel (unit: (Determined through measurement). Average power is also calculated. As a benchmark reference: , in, The moment when the motion begins (unit: s). The current time (in seconds). The integral time variable (unit: s).

[0020] This step completes data acquisition, preprocessing, and power calculation, providing high-quality basic data for fatigue feature extraction.

[0021] (II) Step 2: Fatigue feature extraction and index modeling.

[0022] Fatigue characteristics include: power attenuation coefficient Pressure variation coefficient (Window length N=50), speed decay coefficient .Will Normalization .in, The standard deviation of the pedal pressure within the sliding window (unit: N). This represents the average pressure applied to the pedal within the sliding window (unit: N). The maximum rotational speed (in rpm) recorded at the beginning of the motion.

[0023] Calculate the first gradient of power and second gradient ,in, A persistently negative value is a key indicator for early warning of fatigue. The time period is short, and the control period is in seconds.

[0024] Establish a fatigue index model: , Weighting coefficient Optimization using least squares: ,in, Let be the predicted fatigue index value for the i-th sample. Let i be the experimental calibration value for the i-th sample. This represents the number of historical samples.

[0025] This step completes multi-dimensional feature extraction, gradient analysis, and fatigue index modeling, providing quantitative indicators for fatigue assessment.

[0026] (III) Step 3: Determination of dynamic fatigue state.

[0027] Calculate the dynamic fatigue judgment threshold: , in, This is the initial threshold (default 0.7). This is the time decay coefficient (unit: 1 / s, calibrated experimentally). The cumulative work effect coefficient (unit: 1 / J, calibrated experimentally) and the other two were obtained through experimental calibration. The calibration method was as follows: using an initial threshold... With a baseline of 0.7, and under a balanced performance configuration, parameters are adjusted to match the fatigue state identification results with the actual user state. At the same time, the system records the user's subjective fatigue score (level 1-5) for different riding durations, and adjusts accordingly. Make the identified fatigue level consistent with the score; adjust At the same time, the user's pedal force attenuation performance is recorded according to different cumulative power records, and adjustments are made accordingly. Match the recognition result with the performance, and lock the parameters once the target is met.

[0028] When both conditions are met and ( Recommended value When the condition reaches 0, it is determined to be in a state of fatigue, and the output is... Otherwise, output .

[0029] This step uses dynamic thresholds and dual-condition constraints to determine fatigue state and trigger the target power adjustment mechanism.

[0030] (iv) Step 4: Calculation and dynamic adjustment of the optimal power target.

[0031] Fitting a power generation efficiency function based on historical data and fatigue risk function Calculate the optimal power: , in, is the power generation efficiency function (dimensionless, range [0,1], fitted based on historical data). The fatigue risk function is dimensionless, ranging from [0,1], and is fitted based on historical data. Risk preference coefficient ( When historical data is unavailable, empirical formulas are used. ,in, The maximum power (in W) recorded in the user's historical exercise history.

[0032] Adjust target power based on fatigue status :when hour, ,in Recommended value ;when hour, .

[0033] This step completes the calculation of optimal power and the adjustment of the target under fatigue conditions, providing target values ​​for PID control.

[0034] (v) Step 5: Adaptive PID control and resistance execution.

[0035] Calculate power error Calculate adaptive proportional gain ,in, The initial proportional gain coefficient is dimensionless. Then, the PID control input is calculated. : , in, This is the integral gain coefficient (unit: 1 / s). This represents the differential gain coefficient (unit: s). Integral gain. and differential gain Calibrate using the attenuation curve method: with a fixed initial proportional gain. With the integral and derivative channels disabled, tune the system under the balanced performance configuration described in the document (pedal pressure 239±5N, speed 69±2rpm) to ensure the power response meets the control requirements, and then determine the final calibration value. .

[0036] Map the control input to the resistance level. : The rate of change of resistance is limited to ≤ 2 gears / second, when the rotational speed... Forced (3 gears recommended). Among them, The output resistance level is an integer. This is the minimum resistance level (integer, value 1). This is the maximum resistance level (an integer, with a value of 10, determined based on the actual hardware). This is the theoretical minimum value of the control quantity (dimensionless, determined based on the PID output range). This is the theoretical maximum value of the control quantity (dimensionless, determined based on the PID output range).

[0037] Calculate the PWM duty cycle based on the resistance level. A PWM signal with a frequency of 1kHz is generated to drive the electromagnetic reluctance coil.

[0038] This step completes the adaptive PID control calculation, resistance mapping and limiting, and PWM signal generation, realizing the actual execution of the resistance.

[0039] (vi) Step 6: Data feedback and adaptive continuous optimization.

[0040] Calculate and output current power in real time Fatigue index Resistance levels Cumulative power generation These parameters are transmitted to the display terminal via the UART / I2C interface.

[0041] Return to step 1 to continuously collect data, and repeat steps 1-6 in a loop with a control cycle of 0.1-0.5 seconds. Record complete data for each exercise (pressure, speed, power, fatigue index, etc.). When the accumulated data reaches a threshold (e.g., 10 exercises), re-optimize the weighting coefficients in step 2. And the efficiency curve in step 4 Risk function .

[0042] This step establishes a data feedback mechanism and a cyclical execution framework, continuously learning and optimizing the personalized model to achieve long-term stable adaptive resistance adjustment.

[0043] Experimental Example: To verify the technical effect of this invention, an experimental system was built and the following parameters were set: The permanent magnet generator was configured as 24V / 300W, 12 poles; a pedal pressure sensor (sampling frequency ≥100Hz) and a Hall effect speed sensor (sampling frequency ≥50Hz) were used for data acquisition; the control cycle was set to 0.2 seconds. The system resistance range was 1-10 levels, and the resistance change rate was limited to ≤2 levels / second. The weighting coefficients of the fatigue index model were optimized using the least squares method. The initial threshold for fatigue assessment Time decay coefficient The initial proportional gain of the PID controller was calibrated experimentally. Set to 1.2, integral gain Differential gain The experiment covered four battery voltage configurations (12V, 24V, 36V, 48V) and four load powers (10W, 30W, 50W, 70W), and recruited seven personnel (five men and two women, aged 22-30) to complete 623 experiments.

[0044] Test results show that, under balanced performance configuration (pedal pressure 239±5N, speed 69±2rpm), the system output power reaches 316±19W, the efficiency is 0.891±0.016, and the fatigue accumulation rate is reduced by 49% compared to the maximum power configuration. Model prediction accuracy verification shows: root mean square error RMSE=3.52W, mean absolute percentage error MAPE=3.9±1.1%, consistency correlation coefficient CCC=0.9959, and coefficient of determination R²=0.9923. The multi-component fatigue model has an R²=0.971, which is 55.9% higher than the single exponential model (R²=0.847). Voltage optimization results show that the 48V configuration improves efficiency by 68% compared to the 12V baseline, and the overall system efficiency increases from the traditional 60%-70% to 80%-90%. The power output coefficient of variation decreases from 25%-35% to <15%. Long-term operation tests show that the system's response time to automatically adjust the load after detecting fatigue is less than 3 seconds, extending the user's continuous exercise time by 35-48% and increasing cumulative power generation by 40%-60%. No safety incidents occurred in 623 tests, verifying the effectiveness and safety of the invention.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A human-powered rickshaw adaptive resistance adjustment method based on fatigue state recognition, characterized in that, Comprising the following steps: Step 1, data collection and power calculation: pedal pressure is collected by pedal pressure sensor , rear wheel speed is collected by Hall speed sensor , data is filtered and denoised, and abnormal values are removed, and instantaneous output power is calculated ; Step 2, Fatigue feature extraction and index modeling: Based on the data of step 1, the power attenuation coefficient is extracted , the coefficient of variation of pressure , the speed attenuation coefficient As a fatigue feature, the second-order gradient of power is calculated , and a fatigue index model is established to calculate the fatigue index ; Step 3, dynamic fatigue state determination: calculate dynamic threshold value varying with motion time and accumulated work When greater than the dynamic threshold value and less than the negative threshold value, determine fatigue state and output fatigue flag ; Step 4, Optimal power target calculation and dynamic adjustment: based on power generation efficiency function and fatigue risk function Calculate optimal power Adjust target power according to fatigue flag Lower target power according to fatigue state Step 5, adaptive PID control and resistance execution: calculate the error of target power and actual power , according to Calculate the adaptive proportional gain, calculate the control amount through the PID controller and map to the resistance gear, generate the PWM signal to drive the electromagnetic resistance coil; Step 6, data feedback and adaptive continuous optimization: feedback the power, fatigue index, resistance level and cumulative power generation to the display terminal, and execute steps 1-5 cyclically; when the cumulative data reaches a threshold, re-optimize the fatigue index model weight coefficient and the efficiency function and risk function.

2. The human-powered rickshaw adaptive resistance adjustment method based on fatigue state recognition according to claim 1, characterized in that, In step 1, pedal pressure is collected by pedal pressure sensor Rear wheel speed is collected by Hall speed sensor Kallman filter is applied to pedal pressure and rear wheel speed signal to remove noise, outliers are removed by criteria, and normalized to interval; based on collected pressure and speed, instantaneous output power during user motion is calculated.

3. The human-powered rickshaw adaptive resistance adjustment method based on fatigue state recognition according to claim 1, characterized in that, The fatigue characteristic in step 2 includes a power decay coefficient , a pressure variation coefficient , and a rotation speed decay coefficient , and the fatigue index model is: ， weighting factor Optimized by least square method: wherein, is the fatigue index prediction value of the i-th sample, is the experimental calibration value of the i-th sample, is the number of historical samples.

4. The human-powered vehicle adaptive resistance adjustment method based on fatigue state recognition according to claim 3, characterized in that, In the step 3, first, a dynamic fatigue determination threshold is calculated; then, a determination is made: when the following conditions are simultaneously satisfied and , the fatigue state is determined, and is output; otherwise, is output.

5. The human-powered vehicle adaptive resistance adjustment method based on fatigue state recognition according to claim 4, characterized in that, In step 4, the power generation efficiency function is fitted based on historical data and the fatigue risk function , the optimal power is calculated; and the target power is adjusted according to the fatigue state: when , , the target power is adjusted according to the fatigue state: when , , the target power is adjusted according to the fatigue state: when , .

6. The human-powered vehicle adaptive resistance adjustment method based on fatigue state recognition according to claim 1, characterized in that, In step 5, the PID control amount is calculated by power error and adaptive proportional gain The PID control amount is calculated by power error The control amount is mapped to the resistance level: ; the resistance change rate is limited to ≤ 2 levels / s, and when the rotation speed is forced to ; wherein, is the output resistance level, is the minimum resistance level, is the maximum resistance level, is the control amount theoretical minimum value, is the control amount theoretical maximum value.

7. The human-powered vehicle adaptive resistance adjustment method based on fatigue state recognition according to claim 6, characterized in that, In step 5, the PWM duty cycle is calculated according to the resistance level A PWM signal with a frequency of 1 kHz is generated to drive the electromagnetic resistance coil.

8. The human-powered vehicle adaptive resistance adjustment method based on fatigue state recognition according to claim 1, characterized in that, The step 6, the step 1-6 is executed circularly, the control period is 0.1-0.5 seconds;The complete data of each movement is recorded, when the accumulated data amount reaches the threshold value, the weight coefficient in the step 2 is re-optimized And the efficiency curve in the step 4 , the risk function .

Images