Micro-texture regulation and control method for anodic alumina-based robot touch sensor and touch sensor
By designing a sensing unit with symmetric stress decoupling and using a differential preprocessing algorithm, combined with static and dynamic models, the environmental interference and self-evaluation problems of robot tactile sensors were solved, achieving high-precision pressure sensing and improved stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-04-03
AI Technical Summary
Existing robotic tactile sensors cannot effectively combat environmental interference, cannot fully utilize signal characteristics, and lack self-evaluation capabilities, resulting in fixed perception performance and insufficient sensitivity and stability.
The design incorporates a symmetrical but stress-decoupled inductive sensing unit and a reference sensing unit. A differential preprocessing algorithm is employed, combined with a static mapping model and a dynamic correction model. By synchronously acquiring capacitance signals and extracting dynamic response features in real time, differential preprocessing and correction are performed to achieve accurate calculation of environmental noise suppression and pressure sensing.
It significantly improves the long-term stability and environmental adaptability of the sensor, enables high-precision sensing of rapidly changing pressure events, and has self-evaluation capabilities, supporting online status monitoring and predictive maintenance of the sensor.
Smart Images
Figure CN121783387A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of robot perception technology, specifically to a microtexture modulation method and a tactile sensor for anodized aluminum-based robot tactile sensors. Background Technology
[0002] Robotic tactile sensors are core components for intelligent grasping, precision assembly, and safe human-machine interaction. Capacitive tactile sensors are widely studied due to their advantages such as simple structure, high sensitivity, and low power consumption. The performance of the dielectric layer is a key factor determining the sensor's sensitivity, measurement range, and dynamic response.
[0003] Anodized aluminum oxide has attracted attention because it can be conveniently fabricated into dielectric layers with regular micro- and nanoporous structures through electrochemical processes. By controlling oxidation process parameters (such as voltage, electrolyte, and time), the "microtexture" parameters of the dielectric layer, such as pore size, pore depth, and porosity, can be precisely controlled, thereby allowing for customized design of the piezoresistive properties of sensors. For example, higher porosity usually means a lower equivalent Young's modulus, which may result in higher sensitivity.
[0004] However, existing technologies have the following limitations: Fixed performance and lack of adaptability: The microtexture parameters of the sensor are fixed after preparation, and its sensing performance (such as sensitivity and range) is also determined accordingly, making it impossible to adjust online according to different grasping tasks (such as pinching fragile items and grasping heavy tools).
[0005] Severe environmental interference: The dielectric constant and geometry of the anodic aluminum oxide dielectric layer are affected by ambient temperature and humidity, resulting in significant baseline drift and sensitivity changes in the sensor output, which seriously affects long-term stability and reliability.
[0006] Insufficient utilization of dynamic response: Existing methods mostly focus on the static mapping between pressure and steady-state capacitance, neglecting the rich dynamic response signals (such as rising edge, oscillation, and relaxation process) during pressure application. These dynamic characteristics contain information about the sensor's mechanical structure and the viscoelasticity of the dielectric layer, which has not been effectively utilized to improve sensing performance or diagnose sensor status.
[0007] Lack of self-calibration capability: Under long-term use or extreme environments, the micro-nano structure of sensors may undergo irreversible changes, leading to drift in their sensing characteristics. Existing systems lack effective online mechanisms to assess such changes in "health status".
[0008] Therefore, there is an urgent need for an intelligent tactile sensor solution that can suppress environmental interference, make full use of signal characteristics, and have self-evaluation capabilities. Summary of the Invention
[0009] In view of the above problems, embodiments of the present invention provide a microtexture control method and a tactile sensor for anodized aluminum-based robot tactile sensors, which solves the problems of existing tactile sensors that cannot effectively resist environmental interference, cannot fully utilize signal features, and do not have self-evaluation capabilities.
[0010] According to one aspect of the present invention, a method for microtexture modulation of an anodized aluminum-based robotic tactile sensor is provided. The tactile sensor includes a sensing unit, a reference sensing unit, and a signal analysis unit. The sensing unit and the reference sensing unit each include an identical flexible upper electrode layer, an anodized aluminum dielectric layer with micro / nanoporous texture, and a flexible lower electrode layer. The reference sensing unit is used to characterize changes in the micro / nanoporous texture caused by environmental factors. The anodized aluminum dielectric layer is prepared based on the target microtexture parameters. The target microtexture parameters are determined according to a target pressure sensing performance index. The method includes the following steps: The original capacitance change signal generated by the sensing unit due to pressure and the original reference capacitance signal of the reference sensing unit are collected simultaneously. Based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, and a net capacitance signal characterizing the pure pressure response is obtained. At least one dynamic response feature is extracted in real time from the net capacitance signal, and the dynamic response feature includes signal rise time, oscillation frequency or relaxation time constant. The net capacitance signal is input into a preset pressure-capacitance mapping model to obtain an initial pressure estimate. Based on the dynamic response characteristics, a real-time correction amount is generated for the initial pressure estimate. The initial pressure estimate is then corrected based on the correction amount, and the final pressure sensing result is output.
[0011] In one alternative approach, based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, resulting in a net capacitance signal characterizing the pure pressure response, including: The differential preprocessing is performed using the following formula: , in, This is the original capacitance change signal. The original reference capacitance signal is given, and k is the ratio coefficient of the base capacitance of the sensing unit and the reference sensing unit, obtained under a pressureless static environment. To calibrate the offset, This is the net capacitance signal.
[0012] In one alternative approach, the pressure-capacitance mapping model is a static calibration curve determined by the target micro / nanoporous texture parameters.
[0013] In one alternative approach, the step of generating a real-time correction amount for the initial pressure estimate based on the dynamic response characteristics, correcting the initial pressure estimate based on the correction amount, and outputting the final pressure sensing result further includes: The net capacitance signal is input into a preset pressure-capacitance mapping model to obtain an initial pressure estimate. The extracted dynamic response features are input into the dynamic correction model; the dynamic correction model generates a real-time correction amount for the initial pressure estimate based on the dynamic response features. The initial pressure estimate is corrected based on the correction amount, and the final pressure sensing result is output.
[0014] In one alternative approach, the dynamic correction model is a neural network algorithm trained based on test data from sensors with different microtexture parameters under various dynamic stimuli.
[0015] In one alternative approach, the dynamic correction model is trained in the following manner: A set of anodic aluminum oxide dielectric layer samples with different microtexture parameters were prepared; Each anodized aluminum dielectric layer sample was mounted on a standard dynamic force loading platform so that the standard dynamic force loading platform could apply a variety of controllable dynamic pressure excitations to each anodized aluminum dielectric layer sample. For each dynamic pressure excitation, the net sample capacitance signal after differential preprocessing is recorded, and the dynamic response sample feature vector is extracted from the net sample capacitance signal. Based on the pressure-capacitance mapping model corresponding to each anodic aluminum oxide dielectric layer sample, the initial sample pressure estimate is obtained; The actual pressure value of each anodic aluminum oxide dielectric layer sample is measured simultaneously, and the target correction amount is calculated based on the actual pressure value and the initial sample pressure estimate. Based on the dynamic response sample feature vectors of each anodic aluminum oxide dielectric layer sample and the corresponding target correction amount, a fully connected feedforward neural network is trained to obtain a dynamic correction model.
[0016] According to another aspect of the present invention, a tactile sensor is provided, the tactile sensor including a sensing unit, a reference sensing unit, and a signal analysis unit; The sensing unit and the reference sensing unit each include the same flexible upper electrode layer, an anodized aluminum dielectric layer with micro-nano porous texture, and a flexible lower electrode layer. The reference sensing unit is used to characterize the changes in micro / nano porous texture caused by environmental factors; the anodic aluminum oxide dielectric layer is prepared based on the target microtexture parameters; the target microtexture parameters are determined according to the target pressure sensing performance index; Signal analysis unit, used for: The original capacitance change signal generated by the sensing unit due to pressure and the original reference capacitance signal of the reference sensing unit are collected simultaneously. Based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, and a net capacitance signal characterizing the pure pressure response is obtained. At least one dynamic response feature is extracted in real time from the net capacitance signal, and the dynamic response feature includes signal rise time, oscillation frequency or relaxation time constant. The net capacitance signal is input into a preset pressure-capacitance mapping model to obtain an initial pressure estimate. Based on the dynamic response characteristics, a real-time correction amount is generated for the initial pressure estimate. The initial pressure estimate is then corrected based on the correction amount, and the final pressure sensing result is output.
[0017] In an alternative embodiment, the reference sensing unit is further provided with a rigid component to isolate the pressure above the reference sensing unit.
[0018] In one alternative approach, based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, resulting in a net capacitance signal characterizing the pure pressure response, including: The differential preprocessing is performed using the following formula: , in, This is the original capacitance change signal. The original reference capacitance signal is given, and k is the ratio coefficient of the base capacitance of the sensing unit and the reference sensing unit, obtained under a pressureless static environment. To calibrate the offset, This is the net capacitance signal.
[0019] In one alternative approach, the pressure-capacitance mapping model is a static calibration curve determined by the target micro / nanoporous texture parameters.
[0020] The tactile sensor of this invention includes a sensing unit, a reference sensing unit, and a signal analysis unit. The sensing unit and the reference sensing unit each include the same flexible upper electrode layer, an anodized aluminum dielectric layer with micro / nano porous texture, and a flexible lower electrode layer. The reference sensing unit is used to characterize the changes in the micro / nano porous texture caused by environmental factors. The anodized aluminum dielectric layer is prepared based on the target micro-texture parameters. The target micro-texture parameters are determined according to the target pressure sensing performance index. By synchronously acquiring the original capacitance change signal generated by the sensing unit under pressure and the original reference capacitance signal of the reference sensing unit; based on the original reference capacitance signal, differential preprocessing is performed on the original capacitance change signal to suppress common-mode noise and baseline drift caused by the environment, resulting in a net capacitance signal characterizing the pure pressure response; from the net capacitance signal, at least one dynamic response feature is extracted in real time, including signal rise time, oscillation frequency, or relaxation time constant; the net capacitance signal is input to a preset pressure-capacitance mapping model to obtain an initial pressure estimate; a real-time correction amount is generated for the initial pressure estimate based on the dynamic response feature; the initial pressure estimate is corrected based on the correction amount, and the final pressure sensing result is output, enabling accurate pressure calculation based on the micro-texture characteristics of the tactile sensor.
[0021] Compared with the prior art, the advantages of this invention are as follows: 1. The embodiments of the present invention, by designing a symmetrical but stress-decoupled sensing unit and a reference sensing unit, and by employing a differential preprocessing algorithm, effectively suppress common-mode drift caused by environmental factors such as temperature and humidity, and significantly improve the long-term stability and environmental adaptability of the sensor.
[0022] 2. Algorithm Fusion and Dynamic High Precision: This embodiment of the invention employs a dual-model fusion solution architecture combining a static mapping model and a dynamic correction model. This architecture not only utilizes steady-state capacitance values for preliminary pressure estimation but also fully leverages dynamic response characteristics to correct dynamic errors caused by factors such as sensor mechanical inertia and viscoelasticity in real time, thereby achieving high-precision perception of rapidly changing pressure events.
[0023] 3. The embodiments of the present invention use a dynamic correction model to inversely evaluate the performance of microtextures, and can use real-time sensing data to infer the effective state of the dielectric layer, providing possibilities for sensor factory quality inspection, online status monitoring and predictive maintenance.
[0024] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0025] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic diagram of the structure of the tactile sensor provided in an embodiment of the present invention is shown; Figure 2 A schematic diagram of the structure of a tactile sensor provided in another embodiment of the present invention is shown; Figure 3 A flowchart illustrating the microtexture modulation method for anodized aluminum-based robot tactile sensors provided in an embodiment of the present invention is shown. Detailed Implementation
[0026] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.
[0027] Figure 1 A schematic diagram of the structure of the anodized aluminum-based robot tactile sensor provided in an embodiment of the present invention is shown. Figure 1 and Figure 2 As shown, the device includes: the tactile sensor includes a sensing unit 110, a reference sensing unit 120, and a signal analysis unit 130.
[0028] The sensing unit 110 and the reference sensing unit 120 each include the same flexible upper electrode layer 1110, an anodized aluminum dielectric layer 1112 with micro- and nano-porous textures, and a flexible lower electrode layer. The reference sensing unit 120 is used to characterize changes in the micro- and nano-porous textures caused by environmental factors.
[0029] Specifically, the sensing unit 110 is the core pressure-sensing component of the anodized aluminum-based robotic tactile sensor, comprising, from top to bottom, a flexible upper electrode layer 1110, an anodized aluminum dielectric layer 1112, and a flexible lower electrode layer 1113. When external pressure is applied to the flexible upper electrode layer 1110, the anodized aluminum dielectric layer 1112 undergoes slight deformation, causing a change in the capacitance between the upper and lower electrode layers. The reference sensing unit 120 is identical to the sensing unit 110 in terms of material and stacked structure. Its key difference lies in that the entire reference sensing unit 120 is surrounded and supported by a rigid frame. This rigid frame ensures that externally applied pressure cannot be transmitted to the dielectric layer of the reference sensing unit 120, exposing it only to the same environment (temperature and humidity) as the sensing unit 110, without subjecting it to mechanical stress. Therefore, its capacitance change purely reflects the influence of environmental factors on the properties of the dielectric layer material itself.
[0030] The sensing unit 110 and the reference sensing unit 120 are integrated side-by-side within a mechanical package 140. The signal analysis unit 130 is connected to the electrode layers of the two units via wires, and is used to synchronously and in real-time acquire the raw capacitance change signal. and the original reference capacitance signal .
[0031] The anodic aluminum oxide dielectric layer 1112 is prepared based on the target microtexture parameters. These target microtexture parameters are determined according to the target pressure sensing performance indicators. It should be noted that the preparation of the anodic aluminum oxide dielectric layer 1112 is a conventional process, which can be obtained by anodizing aluminum foil in a specific electrolyte (such as sulfuric acid or oxalic acid solution) under constant voltage or constant current. By precisely controlling parameters such as voltage, temperature, and time, micro / nanoporous structures with specific pore sizes, depths, and porosities, i.e., "microtextures," can be obtained. This is well known to those skilled in the art and will not be elaborated further here.
[0032] In this embodiment of the invention, the signal analysis unit 130 may include a signal acquisition module 1301, a differential preprocessing module 1302, and a feature extraction and correction module 1303.
[0033] Specifically, for signal analysis unit 130: The signal acquisition module 1301 is used to synchronously acquire the original capacitance change signal generated by the sensing unit 110 under pressure, and the original reference capacitance signal of the reference sensing unit 120. Specifically, the signal acquisition module 1301 synchronously acquires the original capacitance change signal of the sensing unit 110 at the same sampling frequency. and the original reference capacitance signal of the reference sensing unit 120 The signal acquisition module 1301 can be a high-precision capacitance-to-digital converter chip. The original reference capacitance signal from the reference sensing unit 120 is included. It is used to reflect the impact of environmental factors on the properties of the dielectric layer material itself.
[0034] The differential preprocessing module 1302 is used to perform differential preprocessing on the original capacitance change signal based on the original reference capacitance signal to suppress common-mode noise and baseline drift caused by the environment, and obtain a net capacitance signal characterizing the pure pressure response.
[0035] In this process, the original capacitance change signal is differentially preprocessed by setting the original reference capacitance signal to eliminate environmental common-mode interference. Since the induction sensing unit 110 and the reference sensing unit 120 have the same material and structure and are in the same environment, the impact of environmental changes on them is similar.
[0036] The differential preprocessing is performed using the following formula: , in, This is the original capacitance change signal. The original reference capacitance signal is given, and k is the ratio coefficient of the base capacitance of the sensing unit and the reference sensing unit, obtained under a pressureless static environment. To calibrate the offset, This is the net capacitance signal. Where k is the proportional coefficient calibrated during the sensor initialization phase in a pressure-free, stable environment; it is equal to the value at this time... The ratio is used to offset the difference in base capacitance between the two cells due to minor manufacturing differences. It is a small calibration offset used to adjust the pressure-free condition. Reset to zero. Processing result. This is the pure, net capacitance response signal primarily caused by pressure.
[0037] The feature extraction and correction module 1303 is used to extract at least one dynamic response feature from the net capacitance signal in real time. The dynamic response feature includes the signal rise time, oscillation frequency, or relaxation time constant. In this embodiment of the invention, for... Analysis is performed to extract dynamic characteristics in real time. For example, at the instant pressure is applied, the rise time required for the signal to rise from 10% to 90% of its amplitude is extracted; during rapid pressure release, the decaying oscillation frequency of the signal is analyzed; or after the pressure holding phase ends, the exponential decay of the signal is fitted to obtain the relaxation time constant. These characteristics reflect the overall dynamic mechanical properties of the sensor, including the viscoelasticity of the dielectric layer.
[0038] The feature extraction and correction module 1303 is further used to input the net capacitance signal into a preset pressure-capacitance mapping model to obtain an initial pressure estimate. Specifically, it will... The steady-state or current value is input into a preset pressure-capacitance mapping model. This pressure-capacitance mapping model is a static calibration curve determined by the target micro / nanoporous texture parameters. It is obtained by calibrating the target micro-texture parameter sensor under quasi-static conditions and can be a lookup table or a polynomial fitting function, thereby outputting an initial pressure estimate.
[0039] The feature extraction and correction module 1303 is further configured to generate a real-time correction amount for the initial pressure estimate based on the dynamic response features, correct the initial pressure estimate based on the correction amount, and output the final pressure sensing result. Specifically, the net capacitance signal is input to a preset pressure-capacitance mapping model to obtain the initial pressure estimate; the extracted dynamic response features are input to a dynamic correction model; the dynamic correction model generates a real-time correction amount for the initial pressure estimate based on the dynamic response features; the initial pressure estimate is corrected based on the correction amount, and the final pressure sensing result is output.
[0040] Specifically, the extracted dynamic response features (such as rise time) are input into the dynamic correction model. This dynamic correction model is trained by subjecting sensor samples with different microtextures to various dynamic pressures (impact, pressure relief, etc.) and recording the relationship between their dynamic features and static measurement errors. Based on the current dynamic features, the dynamic correction model outputs a correction amount for the initial estimate. The initial estimate is added to the correction (or fused in other ways specified by the model) to obtain the final pressure sensing result. For example, when the dynamic characteristics show an extremely short rise time (impact), the dynamic correction model may output a positive correction. This is to compensate for the initial estimate being too low due to system inertia. The dynamic correction model is a neural network algorithm trained based on test data from sensors with different micro-texture parameters under various dynamic stimuli.
[0041] The dynamic correction model is trained in the following way: First, a set of anodized aluminum dielectric layer samples with different microtexture parameters are prepared. Specifically, a set (e.g., N≥20) of anodized aluminum dielectric layer samples can be prepared. These samples cover a wide range of microtexture parameters, including different pore sizes (e.g., 10nm-100nm), porosities (e.g., 20%-60%), and pore depths (i.e., dielectric layer thickness). These parameters are achieved by precisely controlling the voltage, electrolyte concentration, temperature, and time of anodizing. Each sample contains a sensing unit and a reference sensing unit with a rigid frame, and is encapsulated into an independent tactile sensor with a completely identical structure.
[0042] Subsequently, each anodized aluminum oxide dielectric layer sample was mounted on a standard dynamic force loading platform, allowing the platform to apply various controllable dynamic pressure excitations to each sample. Each training sensor was then mounted on a high-precision dynamic force loading platform. This platform can apply various controllable dynamic pressure excitations to simulate real-world robot operation scenarios, including step excitation, sinusoidal sweep excitation, and ramp excitation. Specifically: Step excitation: rapidly increasing pressure to different amplitudes and maintaining it, used to excite the sensor's transient response (rise time, overshoot). Sinusoidal sweep excitation: applying sinusoidal pressure at different frequencies (e.g., 0.1Hz-100Hz) and amplitudes, used to excite frequency-dependent oscillation characteristics. Pulse / impact excitation: applying short, high-amplitude pressure pulses to simulate collisions or rapid grasping. Ramp excitation: linearly increasing pressure at different rates to simulate slow squeezing.
[0043] Then, for each dynamic pressure excitation, the net sample capacitance signal after differential preprocessing is recorded, and the dynamic response sample feature vector is extracted from the net sample capacitance signal. The dynamic response sample feature vector may include: rise time Tr, peak overshoot Os, dominant oscillation frequency Freq, relaxation time constant Tau, and the energy distribution of the signal in a specific frequency band, wherein the energy distribution of the signal in a specific frequency band can be obtained by short-time Fourier transform.
[0044] Subsequently, based on the pressure-capacitance mapping model corresponding to each anodized aluminum dielectric layer sample, an initial sample pressure estimate is obtained. Simultaneously, the actual pressure value of each anodized aluminum dielectric layer sample is measured. Based on the actual pressure value and the initial sample pressure estimate, the target correction amount is calculated. A high-frequency response, high-precision standard force sensor built into the loading platform can be used to simultaneously measure the actual pressure applied to the tactile sensor.
[0045] Finally, based on the dynamic response sample feature vectors and corresponding target correction values of each anodized aluminum oxide dielectric layer sample, a fully connected feedforward neural network is trained to obtain a dynamic correction model. Data from all training sensors and all dynamic excitation experiments are then aggregated. Each data sample is in the form: Input = [Dynamic Feature Vector, Initial Pressure Estimate, Microtexture Parameter Encoding or Sensor ID]; Target Output = The fully connected feedforward neural network's input layer receives the input vector defined above. The hidden layers consist of 2-3 fully connected layers, each equipped with a non-linear activation function and a Dropout layer to prevent overfitting. The number of neurons in the hidden layers is adjusted according to data complexity, for example, 64, 128, or 64. The output layer is a linearly activated neuron that outputs the predicted correction. Iterative training is performed using mean squared error as the loss function.
[0046] In this embodiment of the invention, the tactile sensor further includes a feedback unit for continuously monitoring the net capacitance signal. The relationship with pressure sensing results is equivalent to monitoring the slope change of the pressure-capacitance mapping curve. Combined with the trend of dynamic response characteristics, an inversion model can be established. For example, a decrease in sensitivity and an increase in relaxation time may both indicate the collapse or blockage of the dielectric layer's pore structure. This inversion model can estimate the currently effective equivalent microtexture parameters and compare them with the target parameters at the factory. If the deviation exceeds a threshold, the signal analysis unit 130 can report a performance degradation alarm via the communication interface, indicating that calibration or replacement may be necessary.
[0047] The tactile sensor of this invention includes a sensing unit, a reference sensing unit, and a signal analysis unit. The sensing unit and the reference sensing unit each include the same flexible upper electrode layer, an anodized aluminum dielectric layer with micro / nano porous texture, and a flexible lower electrode layer. The reference sensing unit is used to characterize the changes in the micro / nano porous texture caused by environmental factors. The anodized aluminum dielectric layer is prepared based on the target micro-texture parameters. The target micro-texture parameters are determined according to the target pressure sensing performance index. By synchronously acquiring the original capacitance change signal generated by the sensing unit under pressure and the original reference capacitance signal of the reference sensing unit; based on the original reference capacitance signal, differential preprocessing is performed on the original capacitance change signal to suppress common-mode noise and baseline drift caused by the environment, resulting in a net capacitance signal characterizing the pure pressure response; from the net capacitance signal, at least one dynamic response feature is extracted in real time, including signal rise time, oscillation frequency, or relaxation time constant; the net capacitance signal is input to a preset pressure-capacitance mapping model to obtain an initial pressure estimate; a real-time correction amount is generated for the initial pressure estimate based on the dynamic response feature; the initial pressure estimate is corrected based on the correction amount, and the final pressure sensing result is output, enabling accurate pressure calculation based on the micro-texture characteristics of the tactile sensor.
[0048] Compared with the prior art, the advantages of this invention are as follows: 1. The embodiments of the present invention, by designing a symmetrical but stress-decoupled sensing unit and a reference sensing unit, and by employing a differential preprocessing algorithm, effectively suppress common-mode drift caused by environmental factors such as temperature and humidity, and significantly improve the long-term stability and environmental adaptability of the sensor.
[0049] 2. Algorithm Fusion and Dynamic High Precision: This embodiment of the invention employs a dual-model fusion solution architecture combining a static mapping model and a dynamic correction model. This architecture not only utilizes steady-state capacitance values for preliminary pressure estimation but also fully leverages dynamic response characteristics to correct dynamic errors caused by factors such as sensor mechanical inertia and viscoelasticity in real time, thereby achieving high-precision perception of rapidly changing pressure events.
[0050] 3. The embodiments of the present invention use a dynamic correction model to inversely evaluate the performance of microtextures, and can use real-time sensing data to infer the effective state of the dielectric layer, providing possibilities for sensor factory quality inspection, online status monitoring and predictive maintenance.
[0051] Figure 3 A flowchart illustrating a micro-texture modulation method for an anodized aluminum-based robotic tactile sensor provided in an embodiment of the present invention is shown. This method is executed by the tactile sensor, which can be an anodized aluminum-based robotic tactile sensor used in the robot described in the aforementioned embodiments. Specifically, it is executed by a signal analysis unit within the tactile sensor. The specific steps of this method are largely consistent with the functions performed by the signal analysis unit in the aforementioned embodiments. Figure 3 As shown, the method includes the following steps: Step 210: Synchronously acquire the original capacitance change signal generated by the sensing unit due to pressure, and the original reference capacitance signal of the reference sensing unit.
[0052] Specifically, the signal acquisition module 1301 synchronously acquires the original capacitance change signal of the sensing unit 110 at the same sampling frequency. and the original reference capacitance signal of the reference sensing unit 120 The signal acquisition module 1301 can be a high-precision capacitance-to-digital converter chip. The original reference capacitance signal from the reference sensing unit 120 is included. It is used to reflect the impact of environmental factors on the properties of the dielectric layer material itself.
[0053] Step 220: Based on the original reference capacitance signal, perform differential preprocessing on the original capacitance change signal to suppress common-mode noise and baseline drift caused by the environment, and obtain a net capacitance signal characterizing the pure pressure response.
[0054] In this process, the original capacitance change signal is differentially preprocessed by setting the original reference capacitance signal to eliminate environmental common-mode interference. Since the induction sensing unit 110 and the reference sensing unit 120 have the same material and structure and are in the same environment, the impact of environmental changes on them is similar.
[0055] The differential preprocessing is performed using the following formula: , in, This is the original capacitance change signal. The original reference capacitance signal is given, and k is the ratio coefficient of the base capacitance of the sensing unit and the reference sensing unit, obtained under a pressureless static environment. To calibrate the offset, This is the net capacitance signal. Where k is the proportional coefficient calibrated during the sensor initialization phase in a pressure-free, stable environment; it is equal to the value at this time... The ratio is used to offset the difference in base capacitance between the two cells due to minor manufacturing differences. It is a small calibration offset used to adjust the pressure-free condition. Reset to zero. Processing result. This is the pure, net capacitance response signal primarily caused by pressure.
[0056] Step 230: Extract at least one dynamic response feature from the net capacitance signal in real time. The dynamic response feature includes signal rise time, oscillation frequency, or relaxation time constant. In this embodiment of the invention, for Analysis is performed to extract dynamic characteristics in real time. For example, at the instant pressure is applied, the rise time required for the signal to rise from 10% to 90% of its amplitude is extracted; during rapid pressure release, the decaying oscillation frequency of the signal is analyzed; or after the pressure holding phase ends, the exponential decay of the signal is fitted to obtain the relaxation time constant. These characteristics reflect the overall dynamic mechanical properties of the sensor, including the viscoelasticity of the dielectric layer.
[0057] Step 240: Input the net capacitance signal into the preset pressure-capacitance mapping model to obtain the initial pressure estimate.
[0058] Among them, The steady-state or current value is input into a preset pressure-capacitance mapping model. This pressure-capacitance mapping model is a static calibration curve determined by the target micro / nanoporous texture parameters. It is obtained by calibrating the target micro-texture parameter sensor under quasi-static conditions and can be a lookup table or a polynomial fitting function, thereby outputting an initial pressure estimate.
[0059] Step 250: Generate a real-time correction amount for the initial pressure estimate based on the dynamic response characteristics, correct the initial pressure estimate based on the correction amount, and output the final pressure sensing result.
[0060] Specifically, the net capacitance signal is input to a preset pressure-capacitance mapping model to obtain an initial pressure estimate; the extracted dynamic response features are input to a dynamic correction model; the dynamic correction model generates a real-time correction amount for the initial pressure estimate based on the dynamic response features; the initial pressure estimate is corrected based on the correction amount, and the final pressure sensing result is output.
[0061] Specifically, the extracted dynamic response features (such as rise time) are input into the dynamic correction model. This dynamic correction model is trained by subjecting sensor samples with different microtextures to various dynamic pressures (impact, pressure relief, etc.) and recording the relationship between their dynamic features and static measurement errors. Based on the current dynamic features, the dynamic correction model outputs a correction amount for the initial estimate. The initial estimate is added to the correction (or fused in other ways specified by the model) to obtain the final pressure sensing result. For example, when the dynamic characteristics show an extremely short rise time (impact), the dynamic correction model may output a positive correction. This is to compensate for the initial estimate being too low due to system inertia. The dynamic correction model is a neural network algorithm trained based on test data from sensors with different micro-texture parameters under various dynamic stimuli.
[0062] The dynamic correction model is trained in the following way: First, a set of anodized aluminum dielectric layer samples with different microtexture parameters are prepared. Specifically, a set (e.g., N≥20) of anodized aluminum dielectric layer samples can be prepared. These samples cover a wide range of microtexture parameters, including different pore sizes (e.g., 10nm-100nm), porosities (e.g., 20%-60%), and pore depths (i.e., dielectric layer thickness). These parameters are achieved by precisely controlling the voltage, electrolyte concentration, temperature, and time of anodizing. Each sample contains a sensing unit and a reference sensing unit with a rigid frame, and is encapsulated into an independent tactile sensor with a completely identical structure.
[0063] Subsequently, each anodized aluminum oxide dielectric layer sample was mounted on a standard dynamic force loading platform, allowing the platform to apply various controllable dynamic pressure excitations to each sample. Each training sensor was then mounted on a high-precision dynamic force loading platform. This platform can apply various controllable dynamic pressure excitations to simulate real-world robot operation scenarios, including step excitation, sinusoidal sweep excitation, and ramp excitation. Specifically: Step excitation: rapidly increasing pressure to different amplitudes and maintaining it, used to excite the sensor's transient response (rise time, overshoot). Sinusoidal sweep excitation: applying sinusoidal pressure at different frequencies (e.g., 0.1Hz-100Hz) and amplitudes, used to excite frequency-dependent oscillation characteristics. Pulse / impact excitation: applying short, high-amplitude pressure pulses to simulate collisions or rapid grasping. Ramp excitation: linearly increasing pressure at different rates to simulate slow squeezing.
[0064] Then, for each dynamic pressure excitation, the net sample capacitance signal after differential preprocessing is recorded, and the dynamic response sample feature vector is extracted from the net sample capacitance signal. The dynamic response sample feature vector may include: rise time Tr, peak overshoot Os, dominant oscillation frequency Freq, relaxation time constant Tau, and the energy distribution of the signal in a specific frequency band, wherein the energy distribution of the signal in a specific frequency band can be obtained by short-time Fourier transform.
[0065] Subsequently, based on the pressure-capacitance mapping model corresponding to each anodized aluminum dielectric layer sample, an initial sample pressure estimate is obtained. Simultaneously, the actual pressure value of each anodized aluminum dielectric layer sample is measured. Based on the actual pressure value and the initial sample pressure estimate, the target correction amount is calculated. A high-frequency response, high-precision standard force sensor built into the loading platform can be used to simultaneously measure the actual pressure applied to the tactile sensor.
[0066] Finally, based on the dynamic response sample feature vectors and corresponding target correction values of each anodized aluminum oxide dielectric layer sample, a fully connected feedforward neural network is trained to obtain a dynamic correction model. Data from all training sensors and all dynamic excitation experiments are then aggregated. Each data sample is in the form: Input = [Dynamic Feature Vector, Initial Pressure Estimate, Microtexture Parameter Encoding or Sensor ID]; Target Output = The fully connected feedforward neural network's input layer receives the input vector defined above. The hidden layers consist of 2-3 fully connected layers, each equipped with a non-linear activation function and a Dropout layer to prevent overfitting. The number of neurons in the hidden layers is adjusted according to data complexity, for example, 64, 128, or 64. The output layer is a linearly activated neuron that outputs the predicted correction. Iterative training is performed using mean squared error as the loss function.
[0067] In this embodiment of the invention, the method further includes the following step: continuously monitoring the net capacitance signal. The relationship with pressure sensing results is equivalent to monitoring the slope change of the pressure-capacitance mapping curve. Combined with the trend of dynamic response characteristics, an inversion model can be established. For example, a decrease in sensitivity and an increase in relaxation time may both indicate the collapse or blockage of the dielectric layer's pore structure. This inversion model can estimate the currently effective equivalent microtexture parameters and compare them with the target parameters at the factory. If the deviation exceeds a threshold, the signal analysis unit 130 can report a performance degradation alarm via the communication interface, indicating that calibration or replacement may be necessary.
[0068] The tactile sensor of this invention includes a sensing unit, a reference sensing unit, and a signal analysis unit. The sensing unit and the reference sensing unit each include the same flexible upper electrode layer, an anodized aluminum dielectric layer with micro / nano porous texture, and a flexible lower electrode layer. The reference sensing unit is used to characterize the changes in the micro / nano porous texture caused by environmental factors. The anodized aluminum dielectric layer is prepared based on the target micro-texture parameters. The target micro-texture parameters are determined according to the target pressure sensing performance index. By synchronously acquiring the original capacitance change signal generated by the sensing unit under pressure and the original reference capacitance signal of the reference sensing unit; based on the original reference capacitance signal, differential preprocessing is performed on the original capacitance change signal to suppress common-mode noise and baseline drift caused by the environment, resulting in a net capacitance signal characterizing the pure pressure response; from the net capacitance signal, at least one dynamic response feature is extracted in real time, including signal rise time, oscillation frequency, or relaxation time constant; the net capacitance signal is input to a preset pressure-capacitance mapping model to obtain an initial pressure estimate; a real-time correction amount is generated for the initial pressure estimate based on the dynamic response feature; the initial pressure estimate is corrected based on the correction amount, and the final pressure sensing result is output, enabling accurate pressure calculation based on the micro-texture characteristics of the tactile sensor.
[0069] Compared with the prior art, the advantages of this invention are as follows: 1. The embodiments of the present invention, by designing a symmetrical but stress-decoupled sensing unit and a reference sensing unit, and by employing a differential preprocessing algorithm, effectively suppress common-mode drift caused by environmental factors such as temperature and humidity, and significantly improve the long-term stability and environmental adaptability of the sensor.
[0070] 2. Algorithm Fusion and Dynamic High Precision: This embodiment of the invention employs a dual-model fusion solution architecture combining a static mapping model and a dynamic correction model. This architecture not only utilizes steady-state capacitance values for preliminary pressure estimation but also fully leverages dynamic response characteristics to correct dynamic errors caused by factors such as sensor mechanical inertia and viscoelasticity in real time, thereby achieving high-precision perception of rapidly changing pressure events.
[0071] 3. The embodiments of the present invention use a dynamic correction model to inversely evaluate the performance of microtextures, and can use real-time sensing data to infer the effective state of the dielectric layer, providing possibilities for sensor factory quality inspection, online status monitoring and predictive maintenance.
[0072] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of the present invention are not directed to any particular programming language. It should be understood that the content of the invention described herein can be implemented using various programming languages, and the above description of specific languages is for the purpose of disclosing the best mode of implementation of the invention.
[0073] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.
[0074] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the invention. However, this disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim.
[0075] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.
[0076] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.
Claims
1. A method for microtexture modulation of an anodized aluminum-based robotic tactile sensor, characterized in that, The tactile sensor includes a sensing unit, a reference sensing unit, and a signal analysis unit; the sensing unit and the reference sensing unit each include an identical flexible upper electrode layer, an anodized aluminum dielectric layer with micro / nano porous texture, and a flexible lower electrode layer; the reference sensing unit is used to characterize changes in the micro / nano porous texture caused by environmental factors; the anodized aluminum dielectric layer is prepared based on the target micro-texture parameters; the target micro-texture parameters are determined according to the target pressure sensing performance index; the method includes the following steps: The original capacitance change signal generated by the sensing unit due to pressure and the original reference capacitance signal of the reference sensing unit are collected simultaneously. Based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, and a net capacitance signal characterizing the pure pressure response is obtained. At least one dynamic response feature is extracted in real time from the net capacitance signal, and the dynamic response feature includes signal rise time, oscillation frequency or relaxation time constant. The net capacitance signal is input into a preset pressure-capacitance mapping model to obtain an initial pressure estimate. Based on the dynamic response characteristics, a real-time correction amount is generated for the initial pressure estimate. The initial pressure estimate is then corrected based on the correction amount, and the final pressure sensing result is output.
2. The method according to claim 1, characterized in that, Based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, resulting in a net capacitance signal characterizing the pure pressure response, including: The differential preprocessing is performed using the following formula: ， in, This is the original capacitance change signal. The original reference capacitance signal is given, and k is the ratio coefficient of the base capacitance of the sensing unit and the reference sensing unit, obtained under a pressureless static environment. To calibrate the offset, This is the net capacitance signal.
3. The method according to claim 1, characterized in that, The pressure-capacitance mapping model is a static calibration curve determined by the target micro / nanoporous texture parameters.
4. The method according to claim 1, characterized in that, The step of generating a real-time correction amount for the initial pressure estimate based on the dynamic response characteristics, correcting the initial pressure estimate based on the correction amount, and outputting the final pressure sensing result further includes: The net capacitance signal is input into a preset pressure-capacitance mapping model to obtain an initial pressure estimate. The extracted dynamic response features are input into the dynamic correction model; the dynamic correction model generates a real-time correction amount for the initial pressure estimate based on the dynamic response features. The initial pressure estimate is corrected based on the correction amount, and the final pressure sensing result is output.
5. The method according to claim 4, characterized in that, The dynamic correction model is a neural network algorithm trained based on test data from sensors with different microtexture parameters under various dynamic stimuli.
6. The method according to claim 5, characterized in that, The dynamic correction model was trained in the following way: A set of anodic aluminum oxide dielectric layer samples with different microtexture parameters were prepared; Each anodized aluminum dielectric layer sample was mounted on a standard dynamic force loading platform so that the standard dynamic force loading platform could apply a variety of controllable dynamic pressure excitations to each anodized aluminum dielectric layer sample. For each dynamic pressure excitation, the net sample capacitance signal after differential preprocessing is recorded, and the dynamic response sample feature vector is extracted from the net sample capacitance signal. Based on the pressure-capacitance mapping model corresponding to each anodic aluminum oxide dielectric layer sample, the initial sample pressure estimate is obtained; The actual pressure value of each anodic aluminum oxide dielectric layer sample is measured simultaneously, and the target correction amount is calculated based on the actual pressure value and the initial sample pressure estimate. Based on the dynamic response sample feature vectors of each anodic aluminum oxide dielectric layer sample and the corresponding target correction amount, a fully connected feedforward neural network is trained to obtain a dynamic correction model.
7. A tactile sensor, characterized in that, The tactile sensor includes a sensing unit, a reference sensing unit, and a signal analysis unit. The sensing unit and the reference sensing unit each include the same flexible upper electrode layer, an anodized aluminum dielectric layer with micro-nano porous texture, and a flexible lower electrode layer. The reference sensing unit is used to characterize the changes in micro / nano porous texture caused by environmental factors; the anodic aluminum oxide dielectric layer is prepared based on the target microtexture parameters; the target microtexture parameters are determined according to the target pressure sensing performance index; Signal analysis unit, used for: The original capacitance change signal generated by the sensing unit due to pressure and the original reference capacitance signal of the reference sensing unit are collected simultaneously. Based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, and a net capacitance signal characterizing the pure pressure response is obtained. At least one dynamic response feature is extracted in real time from the net capacitance signal, and the dynamic response feature includes signal rise time, oscillation frequency or relaxation time constant. The net capacitance signal is input into a preset pressure-capacitance mapping model to obtain an initial pressure estimate. Based on the dynamic response characteristics, a real-time correction amount is generated for the initial pressure estimate. The initial pressure estimate is then corrected based on the correction amount, and the final pressure sensing result is output.
8. The tactile sensor according to claim 7, characterized in that, The reference sensing unit is also provided with a rigid component to isolate the pressure above the reference sensing unit.
9. The tactile sensor according to claim 7, characterized in that, Based on the original reference capacitance signal, the original capacitance change signal is subjected to differential preprocessing to suppress common-mode noise and baseline drift caused by the environment, resulting in a net capacitance signal characterizing the pure pressure response, including: The differential preprocessing is performed using the following formula: ， in, This is the original capacitance change signal. The original reference capacitance signal is given, and k is the ratio coefficient of the base capacitance of the sensing unit and the reference sensing unit, obtained under a pressureless static environment. To calibrate the offset, This is the net capacitance signal.
10. The tactile sensor according to claim 9, characterized in that, The pressure-capacitance mapping model is a static calibration curve determined by the target micro / nanoporous texture parameters.