A MEMS gyroscope control parameter self-tuning method and system based on intelligent optimization
The MEMS gyroscope control parameter self-tuning system, designed through intelligent optimization algorithms and hardware co-design, solves the problem of complex debugging of high-performance MEMS gyroscopes, realizes automated parameter tuning, reduces production costs and improves production efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF SCI & TECH
- Filing Date
- 2026-02-12
- Publication Date
- 2026-06-12
Smart Images
Figure CN122192365A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of MEMS gyroscope control, and specifically to a method and system for self-tuning MEMS gyroscope control parameters based on intelligent optimization. Background Technology
[0002] MEMS gyroscopes are inertial sensors that utilize the Coriolis effect to sense the angular motion of objects. They offer advantages such as small size, high reliability, ease of mass production, and low cost, and are widely used in military and civilian fields including aerospace, precision guidance, autonomous driving, and consumer electronics. After decades of development, the static performance of MEMS gyroscopes has continuously improved, reaching navigation-grade levels (BI < 0.01 ° / h). However, high-performance MEMS gyroscopes require a high degree of integration between the structure, circuitry, and control system, resulting in persistently high costs and hindering their widespread adoption as consumer-grade products. For example, debugging a high-performance MEMS four-mass gyroscope typically requires the following three steps: 1) Using an automated testing machine to screen for suitable die structures and obtain data such as resonant frequency, Q value, and actual capacitance values of each comb group; 2) Based on the capacitance test results, calculating the optimal circuit parameters (such as resistance, capacitance, bias voltage, etc.) to achieve a matching state, and determining the tuning voltage based on the structural parameter test results and expected bandwidth; 3) Fine-tuning the parameters of each control loop. This step is non-automatic and is usually done manually by professionals using formulas and experience several times until a better control performance is obtained. Therefore, achieving batch automatic parameter tuning of high-performance MEMS gyroscope control systems is a key approach to reducing production costs and improving production efficiency at this stage. Summary of the Invention
[0003] The purpose of this invention is to provide a method and system for self-tuning control parameters of MEMS gyroscopes based on intelligent optimization.
[0004] The technical solution to achieve the purpose of this invention is: a self-tuning system for MEMS gyroscope control parameters based on intelligent optimization, comprising the following steps:
[0005] Step 1: System initialization, setting the system steady-state conditions, initial control parameter combinations, step excitation signal form, excitation evaluation time, and iteration termination conditions;
[0006] Step 2, State initialization: By applying an initial combination of control parameters to the digital controller, the MEMS gyroscope system is driven to a preset initial stable state.
[0007] Step 3, Step Response Evaluation: The combination of control parameters to be evaluated generated by the online optimization engine is injected into the digital controller. A virtual step excitation is generated by the jump of the internal reference signal in the closed loop, and the response data of the control system is continuously recorded.
[0008] Step 4, fitness calculation: Based on the response data recorded in Step 3, feature extraction is performed to extract features that can quantitatively represent the performance of the control response process, and the fitness value of the current combination of control parameters is calculated using the fitness function.
[0009] Step 5: Update control parameters. Input the fitness value into the intelligent optimization algorithm model, iteratively generate a new combination of control parameters to be evaluated, and return to Step 3 for evaluation.
[0010] Step 6: Determine if the termination condition is met. If not, repeat steps 2 to 6 until the termination condition is met.
[0011] Furthermore, step 3 also includes: real-time monitoring of gyroscope resonance state information; when the control system shows signs of divergence, timely shutdown of the actuator to avoid structural damage; and giving the worst fitness evaluation for the current combination of control parameters.
[0012] Furthermore, in step 4, the extracted features are eight-dimensional features, namely: rise time, steady-state time, mean of steady-state error, variance of steady-state error, time of the first peak, amplitude of the first peak, time of the first trough, and amplitude of the first trough.
[0013] Furthermore, in step 4, the fitness function takes the following form:
[0014] (Formula 1);
[0015] Where B represents the preset maximum loss, and L represents the loss value;
[0016] The loss value is in the form of:
[0017] (Equation 2);
[0018] in, These are the extracted features. - These are: rise time, steady-state time, mean of steady-state error, variance of steady-state error, time of the first peak, amplitude of the first peak, time of the first trough, and amplitude of the first trough. These are the weighting coefficients for speed (F), accuracy (A), and stability (S), used to adjust the relative importance of these three performance dimensions; These are scaling factors for speed (F), accuracy (A), and stability (S), used to normalize feature values of different dimensions to make them comparable.
[0019] Furthermore, in step 6, the iteration termination condition includes:
[0020] 1) The preset running time has been reached;
[0021] 2) The control parameter is judged as a "bad parameter".
[0022] Furthermore, the definition criteria for the "bad parameter" are as follows:
[0023] 1) The rise time exceeds the threshold;
[0024] 2) The overshoot exceeds the threshold;
[0025] 3) The amplitude of the first trough exceeds the amplitude of the first peak, indicating that the system is diverging.
[0026] A self-tuning system for MEMS gyroscope control parameters based on intelligent optimization, used to execute the aforementioned self-tuning method for MEMS gyroscope control parameters based on intelligent optimization, includes:
[0027] The hardware execution layer includes a MEMS gyroscope, analog circuitry, and an FPGA, wherein the FPGA is configured to perform closed-loop digital control of the gyroscope.
[0028] The intelligent decision-making layer is deployed on an ARM processor and communicates with the FPGA via the AXI bus.
[0029] The intelligent decision-making layer includes:
[0030] An amplitude extractor is used to extract the amplitude of the excitation response signal.
[0031] Feature extractors are used to extract features from excitation response amplitude signals that can quantitatively represent the performance of the control response process;
[0032] The fitness evaluator has a built-in multi-objective weighted loss function, which is used to normalize and weight the extracted features and calculate the fitness value of the current combination of control parameters.
[0033] The online optimization engine, which integrates intelligent optimization algorithms, is configured to receive fitness values and automatically generate a new generation of control parameter combinations to be evaluated based on population iteration or gradient strategies.
[0034] The process controller is configured to: after the hardware execution layer drives the gyroscope to a steady state, schedule the entire process of excitation evaluation, feature extraction, fitness calculation and parameter optimization; and monitor the gyroscope resonance state in real time, interrupt the excitation and mark the current parameters as failed when divergence is detected.
[0035] Furthermore, the FPGA in the hardware execution layer is equipped with two independent control channels corresponding to the driving mode and the sensitive mode, respectively. Each channel contains a digital controller and an error synthesis unit.
[0036] Furthermore, the analog circuit includes an analog front-end, an analog back-end, an analog-to-digital converter, and a digital-to-analog converter.
[0037] Furthermore, the intelligent decision-making layer obtains control error signals and gyroscope resonance state information from the FPGA in real time via the AXI bus, and writes the generated new control parameter combination back to the FPGA's digital controller via the AXI bus.
[0038] Compared with existing technologies, the significant advantages of this invention are: 1) It designs a self-tuning system architecture for MEMS gyroscope control parameters based on intelligent optimization, which can be used for automatic calibration of MEMS gyroscope controllers. Compared with traditional trial-and-error, empirical, and formula methods, it has better parameter tuning efficiency and effect, and does not require human intervention in the loop, making it suitable for large-scale, batch production; 2) It adopts an intelligent optimization algorithm as a solution for automatic parameter update strategy, which has higher efficiency and accuracy when dealing with complex systems with multi-parameter coupling and non-convex relationships. Attached Figure Description
[0039] Figure 1 This is a system architecture diagram of a self-tuning method for MEMS gyroscope control parameters based on intelligent optimization, according to the present invention.
[0040] Figure 2 This is a flowchart of a self-tuning method for MEMS gyroscope control parameters based on intelligent optimization according to the present invention. Detailed Implementation
[0041] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0042] The system architecture diagram of the self-tuning method for MEMS gyroscope control parameters based on intelligent optimization of the present invention is as follows: Figure 1 As shown, the system adopts an architecture that coordinates the hardware execution layer and the intelligent decision-making layer to achieve a closed-loop parameter self-tuning process of "incentive-evaluation-optimization-re-incentive".
[0043] The hardware execution layer is used to implement the physical driving, signal detection, and closed-loop control of the MEMS gyroscope. Specifically, it includes: a MEMS gyroscope containing four key electrical interfaces: DDU (Drive-mode Drive Unit), DMU (Drive-mode Measurement Unit), SDU (Sense-mode Drive Unit), and SMU (Sense-mode Measurement Unit); analog circuitry including key modules such as AFE (Analog Front-End), ABE (Analog Back-End), ADC (Analog-to-Digital Converter), and DAC (Digital-to-Analog Converter); and an FPGA (Field Programmable Gate Array). The field-programmable gate array (FPGA) serves as the operating platform for the digital controller of MEMS gyroscopes. It integrates functional modules such as digital signal filtering, phase-sensitive demodulation, and digital controller. Internally, it has two independent control channels corresponding to the driving mode and the sensitive mode, respectively. Each channel contains a digital controller and an error synthesis unit. The digital controller receives Par. (control parameter combination) configuration and executes the control algorithm. The error synthesis unit compares Ref. (reference signal) with the gyroscope response signal to generate Err. (control error signal). This control error signal passes through the digital controller, DAC, and ABE to obtain the excitation voltage signal and is applied to the gyroscope drive interface (DDU or SDU). The response signal generated by the MEMS gyroscope is amplified and conditioned by AFE, then sampled and quantized into a digital signal by ADC, and returned to the input of the digital controller to form a control closed loop.
[0044] The intelligent decision-making layer, serving as the optimization core for parameter tuning, is used for performance evaluation, intelligent optimization, and process scheduling. Deployed on an ARM (Advanced RISC Machine) processor, this layer interacts with the hardware execution layer via the AXI (Advanced eXtensible Interface) bus and includes functional modules such as an amplitude extractor, feature extractor, fitness evaluator, intelligent optimization algorithm, and process controller. The amplitude extractor extracts the amplitude of the excitation response signal; depending on the form of the control error signal, amplitude extraction may require rectification, demodulation, and filtering. The feature extractor extracts features from the excitation response amplitude signal; the extracted features should quantitatively represent the performance of the control response process, such as overshoot, rise time, settling time, mean and variance of steady-state error, etc. The fitness evaluator maps and merges the extracted features to obtain the fitness value of the current control parameter combination. The intelligent optimization algorithm model iteratively generates new control parameter combinations to be evaluated based on the input fitness value. The process controller is used for module scheduling and process control of the parameter self-tuning task.
[0045] The hardware execution layer and the intelligent decision-making layer form a bidirectional data channel via the AXI bus: On one hand, the ARM processor acquires the control error signal and gyroscope resonance state information in the FPGA in real time via the AXI bus. After feature extraction and fitness evaluation, the intelligent optimization algorithm generates a new combination of control parameters. On the other hand, the new combination of control parameters is written back to the FPGA's digital controller via the AXI bus to update the control strategy. This forms a parameter self-tuning closed-loop architecture of "excitation-evaluation-optimization-re-excitation".
[0046] The flowchart of a self-tuning method for MEMS gyroscope control parameters based on intelligent optimization is shown below. Figure 2 As shown, it includes the following steps:
[0047] Step 1: System initialization, setting the system steady-state conditions, initial control parameter combinations, step excitation signal form, excitation evaluation time, and iteration termination conditions;
[0048] Step 2, State Initialization: First, the pre-tuned initial control parameter combination is loaded into the digital controller via the AXI bus. Then, the error synthesis unit inputs the difference between the reference signal (control target) and the gyroscope response signal (amplified and conditioned by AFE, sampled and quantized by ADC) as the control error signal into the digital controller. This control error signal passes through the digital controller, DAC, and ABE to obtain the excitation voltage signal, which is applied to the gyroscope drive interface (DDU or SDU), thereby driving the gyroscope to the preset initial state. Simultaneously, the control error signal is monitored in real time. When its mean, variance, and key state variables simultaneously meet the preset steady-state conditions, the gyroscope system is determined to have reached a stable operating state.
[0049] Step 3, Step Response Evaluation: After confirming that the gyroscope system has reached steady state, the process controller sends the newly generated combination of control parameters to be evaluated via the AXI bus to the digital controller. Subsequently, by changing the reference signal (control target) of the digital controller, a step control error signal is generated at its input. This error signal passes through the digital controller, DAC, and ABE to obtain a step excitation voltage signal, which is applied to the gyroscope drive interface (DDU or SDU). The response signal generated by the MEMS gyroscope is amplified and conditioned by the AFE, then sampled and quantized by the ADC into a digital signal, which returns to the input of the digital controller to form a control closed loop. At the same time, the excitation response error signal is obtained. The excitation response error signal of the control system is continuously recorded, and the gyroscope resonance state information is continuously monitored. When the control system shows signs of divergence, the actuator is shut down in time to avoid structural damage, and the worst fitness evaluation of the current control parameter combination is given.
[0050] Step 4, Fitness Calculation: The excitation response error signal recorded in Step 3 is transmitted to the amplitude extractor of the ARM processor via the AXI bus (if the excitation error signal is in amplitude form, the amplitude extractor is skipped). The extracted excitation response error amplitude signal is input to the feature extractor, which calculates eight-dimensional features that quantitatively represent the performance of the control response process: rise time, steady-state time, mean of steady-state error, variance of steady-state error, time of the first peak, amplitude of the first peak, time of the first trough, and amplitude of the first trough. Then, the extracted features are mapped and merged using a fitness function to obtain the fitness value of the current control parameter combination.
[0051] Step 5: Update control parameters. Input the fitness value into the intelligent optimization algorithm model, iteratively generate a new combination of control parameters to be evaluated, and return to Step 3 for evaluation.
[0052] Step 6: Determine if the termination condition is met. If not, repeat steps 2 to 6 until the termination condition is met.
[0053] Furthermore, the termination conditions include:
[0054] 1) The preset running time has been reached;
[0055] 2) The control parameter is judged as a "bad parameter".
[0056] Furthermore, the definition criteria for the "bad parameter" are as follows:
[0057] 1) The rise time exceeds the threshold;
[0058] 2) The overshoot exceeds the threshold;
[0059] 3) The amplitude of the first trough exceeds the amplitude of the first peak, indicating that the system is diverging.
[0060] Furthermore, the fitness function is exemplified by Equation 1.
[0061] (Equation 1)
[0062] Where B represents the preset maximum loss, and L represents the loss value. An example of the form of the loss value is shown in Equation 2.
[0063] (Equation 2)
[0064] in, These are the extracted features. - These are: rise time, steady-state time, mean of steady-state error, variance of steady-state error, time of the first peak, amplitude of the first peak, time of the first trough, and amplitude of the first trough. These are the weighting coefficients for speed (F), accuracy (A), and stability (S), used to adjust the relative importance of these three performance dimensions; These are scaling factors for speed (F), accuracy (A), and stability (S), used to normalize feature values of different dimensions to make them comparable.
[0065] In summary, this embodiment adopts a hardware and software co-engineering architecture. Through efficient information interaction between the ARM platform and the FPGA, online updates of control parameters and real-time acquisition of errors can be achieved, forming a closed-loop self-tuning. This solution has low requirements for core chips, high system integration, and good reliability, making it highly valuable for engineering applications and very conducive to deployment and promotion in mass-produced products.
[0066] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0067] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for self-tuning control parameters of a MEMS gyroscope based on intelligent optimization, characterized in that, Includes the following steps: Step 1: System initialization, setting the system steady-state conditions, initial control parameter combinations, step excitation signal form, excitation evaluation time, and iteration termination conditions; Step 2, State initialization: By applying the initial control parameter combination to the digital controller, the MEMS gyroscope system is driven to a preset initial stable state. Step 3, Step Response Evaluation: The combination of control parameters to be evaluated generated by the online optimization engine is injected into the digital controller. A virtual step excitation is generated by the jump of the internal reference signal in the closed loop, and the response data of the control system is continuously recorded. Step 4, fitness calculation: Based on the response data recorded in Step 3, feature extraction is performed to extract features that can quantitatively represent the performance of the control response process, and the fitness value of the current combination of control parameters is calculated using the fitness function. Step 5: Update control parameters. Input the fitness value into the intelligent optimization algorithm model, iteratively generate a new combination of control parameters to be evaluated, and return to Step 3 for evaluation. Step 6: Determine if the termination condition is met. If not, repeat steps 2 to 6 until the termination condition is met.
2. The self-tuning method for MEMS gyroscope control parameters based on intelligent optimization according to claim 1, characterized in that, Step 3 also includes: real-time monitoring of gyroscope resonance state information, timely shutdown of actuators to avoid structural damage when the control system shows signs of divergence, and giving the worst fitness evaluation for the current combination of control parameters.
3. The self-tuning method for MEMS gyroscope control parameters based on intelligent optimization according to claim 1, characterized in that, In step 4, the extracted features are eight-dimensional features, namely: rise time, steady state time, mean of steady state error, variance of steady state error, time of the first peak, amplitude of the first peak, time of the first trough, and amplitude of the first trough.
4. The method according to claim 1 or 3, characterized in that, In step 4, the fitness function takes the following form: (Formula 1); Where B represents the preset maximum loss, and L represents the loss value; The loss value is in the form of: (Equation 2); in, These are the extracted features. - These are: rise time, steady-state time, mean of steady-state error, variance of steady-state error, time of the first peak, amplitude of the first peak, time of the first trough, and amplitude of the first trough. These are the weighting coefficients for speed (F), accuracy (A), and stability (S), used to adjust the relative importance of these three performance dimensions; These are scaling factors for speed (F), accuracy (A), and stability (S), used to normalize feature values of different dimensions to make them comparable.
5. The self-tuning method for MEMS gyroscope control parameters based on intelligent optimization according to claim 1, characterized in that, In step 6, the iteration termination condition includes: 1) The preset running time has been reached; 2) The control parameter is judged as a "bad parameter".
6. The self-tuning method for MEMS gyroscope control parameters based on intelligent optimization according to claim 5, characterized in that, The definition criteria for the "bad parameter" are as follows: 1) The rise time exceeds the threshold; 2) The overshoot exceeds the threshold; 3) The amplitude of the first trough exceeds the amplitude of the first peak, indicating that the system is diverging.
7. A self-tuning system for MEMS gyroscope control parameters based on intelligent optimization, characterized in that, The method for performing the self-tuning method for MEMS gyroscope control parameters based on intelligent optimization as described in any one of claims 1-6 includes: The hardware execution layer includes a MEMS gyroscope, analog circuitry, and an FPGA, wherein the FPGA is configured to perform closed-loop digital control of the gyroscope. The intelligent decision-making layer is deployed on an ARM processor and communicates with the FPGA via the AXI bus. The intelligent decision-making layer includes: An amplitude extractor is used to extract the amplitude of the excitation response signal. Feature extractors are used to extract features from excitation response amplitude signals that can quantitatively represent the performance of the control response process; The fitness evaluator has a built-in multi-objective weighted loss function, which is used to normalize and weight the extracted features and calculate the fitness value of the current combination of control parameters. The online optimization engine, which integrates intelligent optimization algorithms, is configured to receive fitness values and automatically generate a new generation of control parameter combinations to be evaluated based on population iteration or gradient strategies. The process controller is configured to: after the hardware execution layer drives the gyroscope to a steady state, schedule the entire process of excitation evaluation, feature extraction, fitness calculation and parameter optimization; and monitor the gyroscope resonance state in real time, interrupt the excitation and mark the current parameters as failed when divergence is detected.
8. The MEMS gyroscope control parameter self-tuning system based on intelligent optimization according to claim 7, characterized in that, The FPGA in the hardware execution layer has two independent control channels, corresponding to the driving mode and the sensitive mode respectively. Each channel contains a digital controller and an error synthesis unit.
9. The MEMS gyroscope control parameter self-tuning system based on intelligent optimization according to claim 7, characterized in that, The analog circuit includes an analog front-end, an analog back-end, an analog-to-digital converter, and a digital-to-analog converter.
10. The MEMS gyroscope control parameter self-tuning system based on intelligent optimization according to claim 7, characterized in that, The intelligent decision-making layer obtains control error signals and gyroscope resonance state information from the FPGA in real time via the AXI bus, and writes the generated new control parameter combination back to the FPGA's digital controller via the AXI bus.