Flexible force feedback sensing system for substation inspection robot

Abstract

The invention relates to the technical field of inspection robots, and particularly discloses a flexible force feedback sensing system for a substation inspection robot, which comprises a flexible sensor array, a signal processing module, a dynamic compensation mechanism module, a contact force prediction module and a transmission control module. According to the scheme, adaptive Kalman filtering is adopted, force signal dynamic characteristics are modeled through a state equation, sliding window overlapping processing is used, updating delay of the characteristics is effectively reduced, and it is ensured that the system can extract reliable contact force characteristics in a complex inspection environment; dynamic compensation is carried out on inertia force and temperature, random noise in signals is effectively removed, an improved Kalman filtering algorithm is adopted to carry out real-time processing on the signals, and the precision and stability of the contact force signals are further improved; the contact force prediction value is predicted based on the convolutional neural network and the long-short-term memory network, and the active prediction capability of the contact force information is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of inspection robots, and in particular to a flexible force feedback sensing system for a substation inspection robot. Background Art

[0002] During intelligent inspections of substation equipment, robots require precise force feedback for safe interaction. However, conventional filters require manual parameter adjustment and have a fixed frequency band. Conventional filters have control delays and are prone to losing transient characteristics. During robot motion, the inertial force caused by acceleration can be easily misinterpreted as contact force. The sensitivity of piezoresistive materials varies with temperature, resulting in significant errors at high temperatures and multiple interference coupling. Contact force information is typically acquired through passive response control during robot inspections. Fixed-parameter models struggle to cope with interference such as temperature drift and mechanical vibration, resulting in poor environmental adaptability and low computational efficiency. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a flexible force feedback sensing system for a substation inspection robot. In view of the problem that general filters require manual parameter adjustment and have a fixed frequency band, and conventional filters have control delays and are prone to losing transient features, this solution adopts adaptive Kalman filtering, models the dynamic characteristics of force signals through state equations, and uses sliding window overlapping processing to effectively reduce feature update delays, ensuring that the system can extract reliable contact force features in complex inspection environments; when the robot is moving, the inertial force caused by acceleration is easily misjudged as contact force, the sensitivity of piezoresistive materials changes with temperature, the error is significant at high temperatures, and there is a problem of multi-interference coupling, this solution adjusts the inertial force and temperature Dynamic compensation is performed to effectively remove random noise in the signal, and an improved Kalman filter algorithm is used to process the signal in real time to further improve the accuracy and stability of the contact force signal. In view of the problems that passive response control is usually used to obtain contact force information during robot inspection, the fixed parameter model is difficult to cope with interference such as temperature drift and mechanical vibration, has poor environmental adaptability, and has low computational efficiency, this solution predicts the contact force value based on CNN (convolutional neural network) and LSTM (long short-term memory network), effectively improving the active prediction ability of contact force information, and extracting spatiotemporal features from the distribution data and dynamic features of the optimal contact force value estimate, thereby improving the collaborative anti-interference ability of the system's multiple physical quantities and effectively improving the prediction accuracy.

[0004] The present invention provides a flexible force feedback sensing system for a substation inspection robot, comprising a flexible sensor array, a signal processing module, a dynamic compensation mechanism module, a contact force prediction module, and a transmission control module;

[0005] The flexible sensor array is composed of a plurality of pressure-sensitive units, each of which comprises a composite structure of a piezoresistive film material and a conductive filler, which are connected by a flexible circuit board and arranged in a grid to form an array;

[0006] The signal processing module acquires raw data from the flexible sensor array, and uses an adaptive Kalman filter algorithm to model the dynamic characteristics of the raw data, extracts the contact force value, mean and variance as dynamic features, and sends the dynamic features to the dynamic compensation mechanism module and the contact force prediction module;

[0007] The dynamic compensation mechanism module installs an inertial measurement unit and a temperature sensor inside the inspection robot to obtain acceleration data and temperature data, performs real-time correction on the contact force value in the dynamic characteristics, and outputs an optimal contact force value estimate to be sent to the contact force prediction module;

[0008] The contact force prediction module is based on a convolutional neural network and combined with a long short-term memory network to extract spatial and temporal series features from the distribution data and dynamic features of the optimal contact force value estimation, and predict the contact force change trend in the short term.

[0009] The transmission control module is installed with a wireless communication module to transmit the contact force value to the central control system inside the robot through the wireless communication protocol, and adjusts the robot's posture and strength in real time according to the contact force value through the CAN bus protocol.

[0010] Furthermore, the signal processing module includes a signal acquisition unit, a noise suppression unit and a feature extraction unit;

[0011] The signal acquisition unit acquires raw data from the flexible sensor array and converts the resistance change of each pressure sensitive unit into a voltage signal through a Wheatstone bridge with a sampling frequency of 1 kHz;

[0012] The noise suppression unit is based on the adaptive Kalman filter algorithm. It models the dynamic characteristics of the voltage signal through the state equation and the observation equation, adjusts the filter parameters in real time, and performs dual suppression on high-frequency noise and low-frequency drift to obtain a filtered voltage signal. The formula used is as follows: ; ;

[0013] Where, Indicates the moment, 、 is the state vector, is the state transition matrix, is the process noise, is the observation vector, is the observation matrix, is the observation noise;

[0014] The feature extraction unit converts the filtered voltage signal into a contact force value, extracts local features through a sliding window algorithm, sets the window size and sliding step size, calculates the mean and variance of the contact force in each window, and outputs the contact force value, mean and variance as dynamic features.

[0015] Furthermore, the dynamic compensation mechanism module includes an acceleration compensation unit, a temperature compensation unit and a force feedback compensation unit;

[0016] The acceleration compensation unit installs an inertial measurement unit inside the end effector of the inspection robot to monitor acceleration data, processes the acceleration data using a low-pass filter, and establishes a mapping relationship between the change in acceleration data and the deviation of dynamic characteristics based on a linear regression algorithm to correct inertial interference during the robot's motion. The formula used is as follows: ;

[0017] Where, Indicates the inertia force compensation amount, is the change in acceleration data, and is the regression coefficient;

[0018] The temperature compensation unit embeds a temperature sensor in the flexible sensor array to detect ambient temperature data. By looking up the table using bilinear interpolation, it determines the influence coefficient of temperature on the sensitivity of the flexible sensor array, performs real-time correction on the contact force value, and outputs the contact force value after temperature compensation. The formula used is as follows: ;

[0019] Where, is the contact force value after temperature compensation, is the original contact force value in the dynamic feature, represents the temperature sensitivity coefficient, is the change in temperature data;

[0020] The force feedback compensation unit is based on an improved Kalman filter algorithm to suppress environmental noise and mechanical vibration noise. It outputs the optimal contact force value estimate based on the historical estimated value and the temperature-compensated contact force value. The formula used is as follows: ; ;

[0021] Where, represents the Kalman gain, represents the state covariance matrix at the previous moment, represents the observation noise covariance matrix, Indicates the The optimal contact force value estimation at time , represents the observation matrix, represents the transpose of the observation matrix.

[0022] Furthermore, the contact force prediction module includes a CNN branch unit, an LSTM branch unit, an objective function construction unit and a fusion output unit;

[0023] The CNN branch unit extracts spatial features from the distribution data and dynamic features of the optimal contact force value estimation, constructs a convolution layer to output a local feature map, performs dimensionality reduction on the local feature map through a pooling layer, repeats convolution and pooling operations, and finally outputs a global feature vector;

[0024] The LSTM branch unit performs temporal dependency modeling on the distribution data and dynamic features of the optimal contact force value estimation, constructs a bidirectional LSTM layer to learn the dynamic evolution law of the contact force value, weights the key time steps through the attention mechanism, and outputs the temporal features;

[0025] The objective function construction unit constructs an overall loss function, and minimizes the overall loss function by adjusting the parameters of the CNN branch unit and the LSTM branch unit while avoiding overfitting problems. The formula used is as follows: ;

[0026] Where, represents the overall loss function, The distribution data representing the optimal contact force value estimate and the total number of samples included in the dynamic features, represents the sample index, and Represent the predicted force value and the true optimal contact force value of the sample, represents the regularization coefficient, represents the model parameters;

[0027] The fusion output unit concatenates the global feature vector and the time series feature, performs dimensionality reduction, and outputs a contact force prediction value.

[0028] Furthermore, in the data transmission module, the wireless communication module is integrated at one end of the flexible circuit board, uses a wireless communication protocol to transmit signals through a flexible antenna, operates in a low power consumption mode, and enters a sleep state during the data transmission interval.

[0029] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0030] (1) In view of the fact that general filters require manual parameter adjustment and have a fixed frequency band, and conventional filters have control delays and are prone to losing transient features, this solution adopts adaptive Kalman filtering, models the dynamic characteristics of force signals through state equations, and uses sliding window overlapping processing to effectively reduce the update delay of features, ensuring that the system can extract reliable contact force features in complex inspection environments.

[0031] (2) When the robot is moving, the inertial force caused by acceleration is easily misjudged as contact force. The sensitivity of the piezoresistive material changes with temperature, and the error is significant at high temperature. There is also the problem of multi-interference coupling. This scheme dynamically compensates for the inertial force and temperature, effectively removes the random noise in the signal, and uses an improved Kalman filter algorithm to process the signal in real time, further improving the accuracy and stability of the contact force signal.

[0032] (3) In view of the problems that passive response control is usually adopted for obtaining contact force information during robot inspection, the fixed parameter model is difficult to cope with interference such as temperature drift and mechanical vibration, has poor environmental adaptability, and has low computational efficiency, this scheme predicts the contact force value based on convolutional neural network and long short-term memory network, effectively improving the active prediction capability of contact force information, and extracting spatiotemporal features from the distribution data and dynamic features of the optimal contact force value estimate, thereby improving the collaborative anti-interference capability of the system's multiple physical quantities and effectively improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of a flexible force feedback sensing system for a substation inspection robot proposed by the present invention;

[0034] Figure 2 Schematic diagram of the contact force prediction module.

[0035] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0037] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0038] Embodiment one, refer to Figure 1 The present application provides a flexible force feedback sensing system for a substation inspection robot, comprising a flexible sensor array, a signal processing module, a dynamic compensation mechanism module, a contact force prediction module and a transmission control module.

[0039] The flexible sensor array is composed of pressure-sensitive units, each pressure-sensitive unit containing a piezoresistive film material and a conductive filler composite structure, connected by a flexible circuit board and arranged in a grid array.

[0040] The signal processing module obtains raw data from the flexible sensor array, and uses an adaptive Kalman filter algorithm to model the dynamic characteristics of the raw data, extracts the contact force value, mean and variance as dynamic characteristics, and sends the dynamic characteristics to the dynamic compensation mechanism module and the contact force prediction module.

[0041] The dynamic compensation mechanism module installs an inertial measurement unit and a temperature sensor inside the inspection robot, obtains acceleration data and temperature data, and corrects the contact force value in the dynamic characteristics in real time, and outputs the optimal contact force value estimate to the contact force prediction module.

[0042] The contact force prediction module is based on a convolutional neural network, combined with a long short-term memory network, extracts spatial and temporal sequence features from the distribution data of the optimal contact force value estimate and the dynamic characteristics, and predicts the contact force change trend in the near future.

[0043] The transmission control module installs a wireless communication module, transmits the contact force value to the central control system inside the robot through the wireless communication protocol, and adjusts the posture and force of the robot in real time according to the contact force value through the CAN bus protocol.

[0044] Embodiment two, refer to Figure 1 Based on the above embodiment, in the flexible sensor array, the size of the pressure-sensitive unit is 1mmx1mm, each pressure-sensitive unit contains a piezoresistive film material and a conductive filler composite structure, connected by a flexible circuit board and arranged in a grid array, and the flexible circuit board uses polyimide material as the substrate.

[0045] The flexible sensor array is fixed to the surface of the end effector of the inspection robot by silicone adhesive to alleviate the stress concentration problem caused by surface unevenness; a porous silicone buffer layer with a thickness of 0.5 mm is set in the edge area to absorb external impact and protect the flexible sensor array from mechanical damage.

[0046] Example 3, see Figure 1 , this embodiment is based on the above embodiment, and the signal processing module includes a signal acquisition unit, a noise suppression unit and a feature extraction unit;

[0047] The signal acquisition unit acquires raw data from the flexible sensor array and converts the resistance change of each pressure sensitive unit into a voltage signal through a Wheatstone bridge with a sampling frequency of 1 kHz;

[0048] The noise suppression unit is based on the adaptive Kalman filter algorithm, constructs state equations and observation equations to model the dynamic characteristics of the voltage signal, adjusts the filter parameters in real time, and performs dual suppression on high-frequency noise and low-frequency drift to obtain a filtered voltage signal. The formula used is as follows: ; ;

[0049] Where, Indicates the moment, 、 is the state vector, is the state transition matrix, is the process noise, is the observation vector, is the observation matrix, is the observation noise;

[0050] The feature extraction unit converts the filtered voltage signal into a contact force value, extracts local features through a sliding window algorithm, sets the window length to 10ms, and the sliding step to 1ms. The mean and variance of the contact force are calculated within each window, and the contact force value, mean, and variance are output as dynamic features.

[0051] By performing the above operations, the general filter requires manual adjustment of parameters and has a fixed frequency band, while conventional filtering has control delays and is prone to losing transient features. In this solution, adaptive Kalman filtering is used to model the dynamic characteristics of the force signal through state equations and use sliding window overlapping processing to effectively reduce the update delay of the feature, ensuring that the system can extract reliable contact force features in complex inspection environments.

[0052] Example 4, see Figure 1, this embodiment is based on the above embodiment, and the dynamic compensation mechanism module includes an acceleration compensation unit, a temperature compensation unit and a force feedback compensation unit;

[0053] The acceleration compensation unit installs an inertial measurement unit inside the end effector of the inspection robot to monitor acceleration data, processes the acceleration data using a low-pass filter, and establishes a mapping relationship between the change in acceleration data and the deviation of dynamic characteristics based on a linear regression algorithm to correct inertial interference during the robot's motion. The formula used is as follows: ;

[0054] Where, Indicates the inertia force compensation amount, is the change in acceleration data, and is the regression coefficient;

[0055] The temperature compensation unit embeds a temperature sensor in the buffer layer of the flexible sensor array to detect ambient temperature data. By using a bilinear interpolation table lookup, it determines the influence coefficient of temperature on the sensitivity of the flexible sensor array, performs real-time correction on the contact force value, and outputs the contact force value after temperature compensation. The formula used is as follows: ;

[0056] Where, is the contact force value after temperature compensation, is the original contact force value in the dynamic feature, represents the temperature sensitivity coefficient, is the change in temperature data;

[0057] The force feedback compensation unit is based on an improved Kalman filter algorithm to suppress environmental noise and mechanical vibration noise. It outputs the optimal contact force value estimate based on the historical estimated value and the temperature-compensated contact force value. The formula used is as follows: ; ;

[0058] Where, represents the Kalman gain, represents the state covariance matrix at the previous moment, represents the observation noise covariance matrix, Indicates the The optimal contact force value estimation at time , represents the observation matrix, represents the transpose of the observation matrix.

[0059] By performing the above operations, the inertial force caused by acceleration during robot movement is easily misjudged as contact force. The sensitivity of piezoresistive materials changes with temperature, the error is significant at high temperatures, and there is a problem of multi-interference coupling. This solution dynamically compensates for the inertial force and temperature, effectively removes random noise in the signal, and uses an improved Kalman filter algorithm to process the signal in real time, further improving the accuracy and stability of the contact force signal.

[0060] Example 5, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the contact force prediction module includes a CNN branch unit, an LSTM branch unit, an objective function construction unit and a fusion output unit;

[0061] The CNN branch unit extracts spatial features from the distribution data and dynamic features of the optimal contact force value estimation, constructs a convolution layer to output a local feature map: the convolution kernel width is set to 5, the number of channels is 32, and the step size is 1. The local feature map is reduced in dimension through the pooling layer, and the convolution and pooling operations are repeated to finally output a 128-dimensional global feature vector;

[0062] The LSTM branch unit performs temporal dependency modeling on the distribution data and dynamic features of the optimal contact force value estimation, constructs a bidirectional LSTM layer to learn the dynamic evolution law of the contact force value, sets 128 hidden units, and weights the key time steps through the attention mechanism to output 256-dimensional temporal features;

[0063] The objective function construction unit constructs an overall loss function, and minimizes the overall loss function by adjusting the parameters of the CNN branch unit and the LSTM branch unit while avoiding overfitting problems. The formula used is as follows: ;

[0064] Where, represents the overall loss function, The distribution data representing the optimal contact force value estimate and the total number of samples included in the dynamic features, represents the sample index, and Represent the predicted force value and the true optimal contact force value of the sample, represents the regularization coefficient, represents the model parameters;

[0065] The fusion output unit concatenates the 128-dimensional global feature vector and the 256-dimensional temporal feature, performs dimensionality reduction, and outputs the contact force prediction value within the next 50 ms.

[0066] By performing the above operations, passive response control is usually adopted for obtaining contact force information during robot inspection. Fixed parameter models are difficult to cope with interference such as temperature drift and mechanical vibration, have poor environmental adaptability, and have low computational efficiency. This solution predicts contact force values ​​based on convolutional neural networks and long short-term memory networks, effectively improving the active prediction capability of contact force information. By extracting spatiotemporal features from the distribution data and dynamic features of the optimal contact force value estimate, the collaborative anti-interference capability of multiple physical quantities of the system is improved, and the prediction accuracy is effectively improved.

[0067] Example 6, see Figure 1 This embodiment is based on the above embodiment. In the transmission control module, the wireless communication module is integrated at one end of the flexible circuit board. The wireless communication protocols include Zigbee and Bluetooth wireless communication protocols. Signal transmission is achieved through a flexible antenna, and it operates in a low-power mode. It enters a dormant state during the data transmission interval. Through the CAN bus protocol, the contact force information is transmitted to the central control system inside the robot in real time, and the posture and strength of the end effector are adjusted according to the size and direction of the contact force.

[0068] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0069] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

[0070] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A flexible force feedback sensing system for a substation inspection robot, characterized by: It includes a flexible sensor array, a signal processing module, a dynamic compensation mechanism module, a contact force prediction module and a transmission control module; The flexible sensor array is composed of a plurality of pressure-sensitive units, each of which comprises a composite structure of a piezoresistive film material and a conductive filler, which are connected by a flexible circuit board and arranged in a grid to form an array; The signal processing module acquires raw data from the flexible sensor array, and uses an adaptive Kalman filter algorithm to model the dynamic characteristics of the raw data, extracts the contact force value, mean and variance as dynamic features, and sends the dynamic features to the dynamic compensation mechanism module and the contact force prediction module; The dynamic compensation mechanism module installs an inertial measurement unit and a temperature sensor inside the inspection robot to obtain acceleration data and temperature data, performs real-time correction on the contact force value in the dynamic characteristics, and outputs an optimal contact force value estimate to be sent to the contact force prediction module; The contact force prediction module is based on a convolutional neural network and combined with a long short-term memory network to extract spatial and temporal series features from the distribution data and dynamic features of the optimal contact force value estimation, and predict the contact force change trend in the short term. The transmission control module is installed with a wireless communication module to transmit the contact force value to the central control system inside the robot through the wireless communication protocol, and adjusts the robot's posture and strength in real time according to the contact force value through the CAN bus protocol.

2. The flexible force feedback sensing system for a substation inspection robot according to claim 1, characterized in that: The signal processing module includes a signal acquisition unit, a noise suppression unit and a feature extraction unit; The signal acquisition unit acquires raw data from the flexible sensor array and converts the resistance change of each pressure sensitive unit into a voltage signal through a Wheatstone bridge with a sampling frequency of 1 kHz; The noise suppression unit is based on an adaptive Kalman filter algorithm, which models the dynamic characteristics of the voltage signal through state equations and observation equations, adjusts the filter parameters in real time, and performs dual suppression on high-frequency noise and low-frequency drift to obtain a filtered voltage signal; The feature extraction unit converts the filtered voltage signal into a contact force value, extracts local features through a sliding window algorithm, sets the window size and sliding step size, calculates the mean and variance of the contact force in each window, and outputs the contact force value, mean and variance as dynamic features.

3. The flexible force feedback sensing system for a substation inspection robot according to claim 1, characterized in that: The dynamic compensation mechanism module includes an acceleration compensation unit, a temperature compensation unit and a force feedback compensation unit; The acceleration compensation unit installs an inertial measurement unit inside the end effector of the inspection robot to monitor acceleration data, processes the acceleration data using a low-pass filter, and establishes a mapping relationship between the change in acceleration data and the deviation of dynamic characteristics based on a linear regression algorithm to correct inertial interference during the robot's movement. The temperature compensation unit embeds a temperature sensor in the flexible sensor array to detect ambient temperature data, determines the influence coefficient of temperature on the sensitivity of the flexible sensor array through bilinear interpolation table lookup, makes real-time corrections to the contact force value, and outputs the contact force value after temperature compensation; The force feedback compensation unit suppresses environmental noise and mechanical vibration noise based on an improved Kalman filter algorithm, and outputs an optimal contact force value estimate according to historical estimated values ​​and temperature-compensated contact force values.

4. The flexible force feedback sensing system for a substation inspection robot according to claim 1, characterized in that: The contact force prediction module includes a CNN branch unit, an LSTM branch unit, an objective function construction unit and a fusion output unit; The CNN branch unit extracts spatial features from the distribution data and dynamic features of the optimal contact force value estimation, constructs a convolution layer to output a local feature map, performs dimensionality reduction on the local feature map through a pooling layer, repeats convolution and pooling operations, and finally outputs a global feature vector; The LSTM branch unit performs temporal dependency modeling on the distribution data and dynamic features of the optimal contact force value estimation, constructs a bidirectional LSTM layer to learn the dynamic evolution law of the contact force value, weights the key time steps through the attention mechanism, and outputs the temporal features; The objective function construction unit constructs an overall loss function and minimizes the overall loss function by adjusting parameters of the CNN branch unit and the LSTM branch unit; The fusion output unit concatenates the global feature vector and the time series feature, performs dimensionality reduction, and outputs a contact force prediction value.

5. The flexible force feedback sensing system for a substation inspection robot according to claim 1, characterized in that: In the transmission control module, the wireless communication module is integrated at one end of the flexible circuit board, uses the wireless communication protocol to transmit signals through the flexible antenna, operates in a low power consumption mode, and enters a dormant state during the data transmission interval.

Images