A collaborative operation and inspection control system based on fusion of AR virtual superimposition perception feedback and artificial inspection

Abstract

The application relates to the technical field of operation and inspection control or regulation systems, and discloses a collaborative operation and inspection control system based on AR virtual superimposed perception feedback and artificial inspection fusion, which comprises a feedback data sampling unit, a state evolution mapping unit, a logical fusion arbitration unit and an instruction execution driving unit. The logical fusion arbitration unit determines a feedback divergence characteristic value according to statistical distribution characteristics of a physical perception sequence, determines a confidence state and a fusion weight coefficient of a perception channel based on the feedback divergence characteristic value, selects a weighted fusion mapping relationship for the physical perception sequence and a predicted state feedback stream, and generates a collaborative control instruction. The application uses advanced feedback information generated by a dynamic mathematical model to offset the sampling lag of the perception channel, suppresses the step overshoot of an execution mechanism of a controlled object through a logical damping feature, solves the phase delay of closed-loop feedback in an unstructured dynamic disturbance scene, and improves the operation and inspection operation control precision and operation continuity.

Description

Technical Field

[0001] This invention relates to a collaborative operation and maintenance control system based on the fusion of AR virtual overlay perception feedback and manual inspection, belonging to the field of operation and maintenance control or adjustment technology. Background Technology

[0002] The accuracy of current closed-loop feedback relies heavily on the real-time perception of the controlled object's state by the sensing unit. In industrial settings, current sensors, vibration sensors, and temperature sensors are typically used to construct direct feedback loops. By collecting physical state parameters, the trajectory of the controlled object is corrected. Such systems are based on the fact that the feedback signal can characterize the transient features of the controlled object in real time and in all dimensions, which is the fundamental means to maintain the stable operation of the closed-loop loop. The effectiveness of the closed-loop control law depends on the spatiotemporal consistency between the sampled data and the physical entity's state. The control unit calculates the adjustment increment based on the received feedback parameters and drives the actuator to generate deviation compensation. In terms of the sensing mechanism, the sensing element captures the physical field changes generated by the controlled object and converts them into electrical signals, which are then sent to the controller via an analog-to-digital conversion process. This process involves physical response delays and sampling window periods.

[0003] However, controlled objects are typically located in unstructured environments containing random interference sources such as electromagnetic pulses and mechanical resonances. The sensing process of physical sensors inherently suffers from physical response delays and sampling windows, causing the system state represented by the feedback signal to deviate from the current state of the controlled object when it reaches the controller. This sensing lag means the controller can only adjust based on historical deviations, making it difficult to capture the transient evolution trend of the controlled object. This manifests as response lag or unexpected overshoot in the closed-loop control process. Industry attempts have attempted to use low-pass filtering or increasing the sampling frequency, but low-pass filtering introduces additional signal phase lag, and simply increasing the sampling frequency increases the deployment cost and computational burden of the system, without eliminating physical sampling lag and real-time compensation. The core contradiction between compensation and demand is exemplified by the Chinese invention patent with authorization announcement number CN113452962B, which discloses a data center enhanced inspection system and method with spatial collaborative perception. It utilizes a high-definition binocular camera and augmented reality equipment to match the physical image of the data center with the spatial pose of a virtual three-dimensional model, thereby enhancing the visualization of the inspection process. While it has made progress in risk avoidance and interactive perception through multi-source information visual fusion, it still relies on the passive capture and mapping of physical sensor data. It lacks dynamic prediction and hedging mechanisms at the control law level for issues such as random disturbances in the feedback channel caused by unstructured environmental noise and oscillations in the actuator's response to sudden stress adjustment. This results in unsatisfactory closed-loop stability of the system under extreme dynamic conditions.

[0004] Therefore, the technical problem to be solved by this invention is how to eliminate physical perception lag and improve the stability of the closed-loop system under dynamic disturbances through the coordination of virtual and real feedback at the logical level without relying on hardware stacking. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A collaborative operation and maintenance control system based on the fusion of AR virtual overlay perception feedback and manual inspection, the system comprising:

[0006] The feedback data sampling unit is used to acquire the physical sensing sequence and environmental disturbance data that characterize the real-time state of the controlled object;

[0007] The state evolution mapping unit is used to construct a dynamic mathematical model of the controlled object based on the physical perception sequence, and to use the dynamic mathematical model to generate a predicted state feedback flow that is ahead of the sampling time of the physical perception sequence.

[0008] The logical fusion arbitration unit, connected to the feedback data sampling unit and the state evolution mapping unit respectively, is used to perform the following steps: Step S101, extract the statistical distribution characteristics of the physical sensing sequence and calculate the feedback divergence feature value, which characterizes the degree of dispersion of the statistical distribution characteristics; Step S102, extract the evolution characteristics of the feedback divergence feature value with the sampling period, and determine the confidence state of the sensing channel in combination with a preset confidence threshold, and determine the fusion weight coefficient based on the confidence state; Step S103, perform a weighted fusion operation on the physical sensing sequence and the estimated state feedback stream based on the fusion weight coefficient to generate a cooperative control command;

[0009] The instruction execution drive unit is used to receive cooperative control instructions and convert them into control pulse signals to adjust the physical operating state of the controlled object.

[0010] Preferably, the state evolution mapping unit is equipped with a transfer function self-correction loop, which is used to perform the following steps: acquiring the drive feedback data output by the instruction execution drive unit, and using the drive feedback data as the basis for model inversion to correct the transfer function coefficients of the dynamic mathematical model in real time; adjusting the frequency response characteristic of the predicted state feedback flow according to the corrected transfer function coefficients, so that the dynamic trend of the predicted state feedback flow evolves synchronously with the mechanical wear characteristics of the controlled object.

[0011] Preferably, when executing step S102, the logic fusion arbitration unit determines the fusion weight coefficient in the following manner. When the feedback divergence feature value exceeds the preset confidence threshold, the increment of the feedback divergence feature value between the current sampling period and the previous sampling period is calculated. The fusion weighting coefficient is determined using the following formula. : , where α is the slope of the autocorrelation spectrum; the logic fusion arbitration unit adjusts the weight of the predicted state feedback flow in the collaborative control command generation logic according to the fusion weight coefficient ω.

[0012] Preferably, the logic fusion arbitration unit is also used to perform the following steps: Step S401, monitor the rate of change of the physical sensing sequence, and when the rate of change of the physical sensing sequence exceeds a preset jump threshold and the feedback divergence characteristic value is in the non-steady-state range, activate the elastic gating of the feedback path; Step S402, introduce logic damping characteristics into the cooperative control command through the elastic gating to limit the output slope of the cooperative control command.

[0013] Preferably, the state evolution mapping unit uses the estimated state feedback flow to pre-set the prediction compensation amount before the instruction execution driving unit takes action, and compensates for the phase delay of the sensing channel according to the feedback lag time of the physical sensing sequence.

[0014] Preferably, the feedback data sampling unit includes a sensor health monitoring module for monitoring the power spectral density of the sampled signal. When the correlation of the power spectral density is less than 0.75, the state evolution mapping unit is triggered to switch to the dynamic mathematical model inertial maintenance mode.

[0015] Preferably, when executing step S103, the logic fusion arbitration unit uses a nonlinear saturation operator to restrict the state control parameters of the controlled object within the logical safety boundary defined by the estimated state feedback flow.

[0016] Preferably, the instruction execution drive unit includes an execution deviation closed loop, used to perform residual compensation on the control pulse signal based on the deviation between the measured displacement of the instruction execution drive unit and the calculated expected value of the cooperative control instruction.

[0017] Preferably, the system further includes an environmental stress processing unit for converting environmental disturbance data into disturbance gain coefficients and performing amplitude weighting on the logical safety boundary of the estimated state feedback flow.

[0018] Preferably, the system also includes an operation and maintenance status tracking unit, which maps the timing waveform trajectory of the coordinated control command to the health assessment value of the controlled object, and outputs a status maintenance warning signal when the health assessment value is continuously lower than 0.85.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. In collaborative operation and maintenance with artificial logic fusion, by constructing a dynamic mathematical model synchronized with the dynamic characteristics of the controlled object, the traditional passive physical feedback is transformed into a real-time superposition of virtual enhanced feedback and physical sampling signals. Under this mechanism, the system no longer relies in isolation on physical sensor data with physical response time delay or sampling window period. Instead, it uses the theoretical expected state prediction compensation amount generated by the calculation module in the dynamic mathematical model to make up for the information loss of the feedback loop within the sampling interval. This method of pre-compensating the physical perception residual through virtual prediction potential energy enables the regulator to respond to the transient trend of the controlled object and eliminates the regulation dead zone of the closed-loop system when dealing with unstructured disturbances.

[0021] 2. This system uses a logical fusion arbitration unit to perceive the temporal topological characteristics of the physical sensing sequence. It uses the autocorrelation spectrum slope of the physical state parameters within the sliding sampling window to characterize the confidence level of the feedback channel. When the controlled object faces extreme random interference that reduces the autocorrelation of the physical signal, the logical fusion arbitration unit uses evolution damping generated by the autocorrelation spectrum slope to forcibly lock the dominance of the virtual enhanced predicted state feedback flow. This allows the deterministic logic of the dynamic mathematical model to anchor the random jitter of the physical sensing. Without introducing the phase lag of the low-pass filter, it filters high-frequency nonlinear noise in the physical sensing link, avoiding the blind response of the controlled object's actuator to transient spike signals, and reducing energy loss and mechanical wear of the controlled object's actuator during the closed-loop regulation process.

[0022] 3. The system has online self-calibration capability based on the electrical echo signal of the actuator. By extracting the parasitic response characteristics of the instruction execution drive unit during the instruction pulse interval, it realizes real-time updates of the dynamic mathematical model parameters. In this collaborative mechanism, the back electromotive force fluctuation or current response lag of the actuator is transformed into a data source for verifying the accuracy of the dynamic mathematical model, enabling the state evolution mapping unit to automatically correct its transfer function coefficients according to the evolution of the mechanical characteristics of the physical entity. This method of using the natural echo information of the execution instruction for reverse inversion eliminates the mechanism mismatch between the dynamic mathematical model and the controlled object during service life, ensuring the consistency of the virtual-real mapping throughout the system's entire life cycle, and enabling the operation and maintenance task to maintain stable control accuracy even under long-term physical environment drift. Attached Figure Description

[0023] Figure 1 This is a block diagram illustrating the closed-loop feedback control principle and data flow of the system of the present invention;

[0024] Figure 2 This is a diagram showing the three-layer functional domain division and vertical deployment architecture of the system of this invention. Detailed Implementation

[0025] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, illustrates a collaborative operation and maintenance control system based on the fusion of AR virtual overlay perception feedback and manual inspection provided by the present invention. It should be noted that the following embodiments are intended to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0026] This invention provides a collaborative operation and maintenance control system based on the fusion of AR virtual overlay sensing feedback and manual inspection. It comprises a feedback data sampling unit, a state evolution mapping unit, a logic fusion arbitration unit, and an instruction execution drive unit. The feedback data sampling unit acquires physical sensing sequences characterizing the real-time state (trajectory) of the controlled object, as well as environmental disturbance data. The state evolution mapping unit constructs a dynamic mathematical model of the controlled object using the physical sensing sequences and generates an estimated state feedback stream that precedes the sampling timing of the physical sensing sequences. The logic fusion arbitration unit, as the decision layer, performs dynamic weighted fusion of the physical sensing sequences and the estimated state feedback stream based on the confidence state of the feedback channel, generating collaborative control instructions to correct the trajectory of the controlled object. The instruction execution drive unit converts the collaborative control instructions into control pulse signals, thereby driving the controlled object to complete the closed-loop adjustment process. In general operation and maintenance control or adjustment systems, closed-loop feedback... The accuracy of the feedback relies heavily on the real-time perception of the controlled object's state by the feedback data sampling unit. However, the controlled object is often in an unstructured environment containing random interference sources. The physical sensor's perception process has inherent physical response delays and sampling windows, causing the system state represented by the feedback signal to deviate from the current state of the controlled object when it is transmitted to the regulator. This phase delay contradiction means that the controller can only adjust based on historical deviations, making it difficult to capture the transient evolution trend of the controlled object. To address the collaborative interaction needs in manual inspection scenarios, this invention utilizes AR technology to overlay the estimated state feedback stream generated by the state evolution mapping unit as a virtual shadow trajectory onto the real-time monitoring screen of the inspection terminal. This virtual-real overlay perception enhances the inspector's intuitive understanding of the transient trend of the controlled object. To address this challenge, the feedback data sampling unit in this invention establishes a synchronous sampling procedure to sample the physical state parameters of the controlled object. Real-time capture is performed; the feedback data sampling unit includes a sensor health monitoring module, which is used to monitor the power spectral density of the sampled signal. When the correlation of the power spectral density is less than 0.75, the subsequent state evolution mapping unit is triggered to switch to the inertial maintenance mode of the dynamic mathematical model.

[0027] Because traditional control systems suffer from a physical response time delay between the hysteresis of sampled signals and the real-time compensation requirements of the system when dealing with unstructured dynamic disturbances, the system employs a state evolution mapping unit to construct a dynamic mathematical model of the controlled object and uses this model to generate an estimated state feedback flow. The unit features a self-correcting transfer function loop that uses the drive feedback data output by the instruction execution drive unit as the basis for model inversion, correcting the transfer function coefficients of the dynamic mathematical model in real time. Specifically, the state evolution mapping unit adjusts the frequency response characteristics of the estimated state feedback flow based on the corrected transfer function coefficients, ensuring that the dynamic trend of the estimated state feedback flow evolves synchronously with the mechanical wear characteristics of the controlled object. In the motion control scenario of the inspection robot, the system uses the current damping coefficient and rotational inertia model to obtain the theoretical displacement after 20ms through integration, and constructs the estimated state feedback flow accordingly. This allows the system to utilize the predicted potential energy to address the physical perception residual. Pre-compensation compensates for the information loss of physical sensors within the sampling interval; the state evolution mapping unit acquires the back electromotive force E generated during the pulse interval of the command execution drive unit to correct the transfer function coefficients of the dynamic mathematical model, extracts the attenuation rate of the back electromotive force E, and calculates the transient damping ratio ζ of the controlled object according to the formula ζ=η⋅(dE / dt), where ζ is the transient damping ratio, η is the preset conversion coefficient, and dE / dt is the rate of change of the back electromotive force with time. The deviation between the transient damping ratio ζ and the preset damping coefficient of the dynamic mathematical model is calculated, and the deviation is used as the correction increment factor μ to adjust the coefficients of the polynomial in the denominator of the transfer function, compensates for model mismatch caused by mechanical wear, and makes the predicted state feedback flow more accurate. With physical perception sequence The time-domain response characteristics are consistent.

[0028] In the process of multi-source feedback signal fusion, the multi-source feedback signals include the physical sensing sequence acquired by the feedback data sampling unit and the estimated state feedback stream generated by the state evolution mapping unit. Simply relying on numerical deviation for weight switching can easily introduce high-frequency oscillations in the closed loop. The logic fusion arbitration unit extracts the statistical distribution characteristics of the physical sensing sequence and calculates the feedback divergence characteristic value, which characterizes the dispersion of the statistical distribution characteristics. This invention not only allocates weights based on the magnitude of the deviation, but also judges whether the disturbance of the physical environment has exceeded the system stability region by real-time monitoring of the autocorrelation decay characteristics of the feedback signal. The logic fusion arbitration unit calculates the physical state parameters within the sliding sampling window. The short-time autocorrelation function R(m) of the sequence is obtained, where m is the delay step size. The logarithmic decay curve of R(m) within a preset step size range is fitted by the least squares method to obtain the autocorrelation spectrum slope α. This autocorrelation spectrum slope α is used to characterize the energy consistency of the physical feedback channel and the steady-state deviation trend of the controlled object. When the loss of autocorrelation of the physical signal is detected, that is, when the absolute value of the spectral slope exceeds a preset threshold, the system forcibly locks the estimated state feedback flow through logic damping. The dominant position is maintained by virtual inertia to keep the operating frequency of the actuator; the logical fusion arbitration unit calculates the instantaneous variance of the physical sensing sequence within the sliding sampling window. And compare it with the historical baseline variance of the controlled object when it is in steady state. The feedback divergence eigenvalue is calculated using the following formula after comparison. : ,in, To provide feedback divergence eigenvalues, Let be the instantaneous variance of the physical perception sequence. The historical baseline variance is used; the logical fusion arbitration unit is based on the feedback divergence characteristic value. The numerical range is selected for weighted fusion mapping relationship, in the feedback divergence eigenvalue. When the value is below a preset stability threshold, the system employs a linear weighted mapping relationship to maintain the real-time performance of the physical feedback, based on the feedback divergence characteristic value. When the system exceeds the stationarity threshold and the autocorrelation spectrum slope α becomes distorted, the system automatically switches to an exponentially weighted mapping relationship based on logistic damping characteristics, injecting physical state parameter weighting coefficients by limiting the physical state parameters of the physical sampled signal. This is used to suppress the step response at the execution end.

[0029] To achieve a smooth evolution of weights, the logical fusion arbitration unit determines the fusion weight coefficient ω using the following formula: Where ω is the fusion weight coefficient, ΔH is the increment of the feedback divergence characteristic value between the current sampling period and the previous sampling period, and α is the slope of the autocorrelation spectrum; the logical fusion arbitration unit adjusts the predicted state feedback flow according to the fusion weight coefficient ω. The weights in the collaborative control instruction generation logic are calculated using the following formula to determine the collaborative adjustment vector. : ,in, For the coordinated adjustment vector, These are the weighting coefficients for the physical state parameters. To estimate the weight coefficients of the state feedback flow, and and The change in a single iteration is constrained by a weight-adjusted damping factor determined based on ω, to ensure... Second-order continuity on the time-series evolution curve; system sampling frequency When the Hz frequency is 200Hz, α is set to 0.5. The slope of the autocorrelation spectrum, which reflects the confidence level of the physical feedback channel, is defined as the variable α. α is compared with the judgment threshold in real time. The quantitative relationship executes weight allocation between the virtual prediction path and the physical feedback path. When the controlled object faces random disturbances that degrade the physical signal quality, the system activates the elastic gating of the feedback path. The logic fusion arbitration unit monitors the rate of change of the physical sensing sequence. When the rate of change exceeds the preset jump threshold and the feedback divergence characteristic value is in the non-steady-state range, the system introduces a logic damping feature to the cooperative control command through elastic gating to limit the output slope of the cooperative control command. In this process, the logic fusion arbitration unit uses a nonlinear saturation operator to confine the state control parameters of the controlled object within the logical safety boundary defined by the estimated state feedback flow, eliminating the step disturbance caused by the sudden change in the feedback path. The command execution drive unit includes an execution deviation closed loop, which is used to perform residual compensation based on the deviation between the measured displacement and the calculated expected value of the cooperative control command.

[0030] The logical safety boundary is determined, and the logical fusion arbitration unit calculates the covariance of the N sets of historical state vectors of the controlled object to generate a covariance matrix P. The standard deviation σ, which represents the distribution law of each controlled parameter, is extracted from the main diagonal position of matrix P. The amplitude of the cooperative control command output is limited to the physical sensing sequence using a nonlinear saturation operator. Centered on the moving average and within a radius of three standard deviations σ, the logical safety boundary not only suppresses step disturbances caused by unstructured environmental noise, but also serves as a safety reference threshold for the fusion of human operation and automated commands during manual inspections. This ensures that even under manual intervention or sudden disturbances, the output slope of the coordinated control commands maintains second-order continuity and mechanical stress safety; it also suppresses step disturbances caused by unstructured environmental noise and guarantees the coordinated adjustment vector. The continuity in the time-series evolution curve; in the numerical demonstration of collaborative control command generation, assuming the sampling frequency is set to 10Hz, the physical sensing sequence at the current moment. The instantaneous sample value is 10.2, and the predicted state feedback flow is... The autocorrelation spectrum slope α is set to 0.5, with a value of 10.0. If ΔH is 0.05, the calculated ω is 0.90, and the system allocates accordingly. 0.1 The value is 0.9, and the collaborative adjustment vector is calculated. The value is 10.02; if the environmental disturbance data suddenly increases in the next sampling period, leading to... When the instantaneous jump to 15.0, ΔH increases to 0.8, and ω decreases to 0.20, the system shifts the control of the controlled object's actuators towards the predicted state feedback flow, thus increasing the output coordinated adjustment vector. The system smoothly migrates to the new target value, effectively suppressing step overshoot in the actuator of the controlled object. It should be noted that all functional modules in the above system are deployed in industrial control computers or embedded microprocessors. The nonlinear saturation operator executed by the logic fusion arbitration unit is used to confine the state control parameters of the controlled object within the logical safety boundary defined by the estimated state feedback flow. The logical safety boundary is obtained by real-time calculation of the covariance matrix of the historical running trajectory, ensuring that the output slope of the cooperative control command is always within the mechanical stress tolerance range of the actuator of the controlled object. Through the deep coupling of modules such as state acquisition, virtual mapping, logic arbitration, and execution drive, the system constructs a closed-loop control system with self-evolution capabilities.

[0031] Example 1: In a substation inspection scenario involving high-voltage circuit breaker operation and high-frequency transient electromagnetic interference, the controlled object is an inspection robot performing a preset path planning task. When it enters the monitoring area of ​​the switchgear where the circuit breaker is located, the pulse group interference generated in the environment electromagnetically couples with the sensing link, causing the physical perception sequence captured by the feedback data sampling unit to deviate from the actual motion trajectory of the controlled object in a nonlinear jump. Specifically, this manifests as changes in physical state parameters. The numerical value generates an abnormal increment exceeding the trajectory prediction threshold within a single sampling period; to address the feedback loop instability problem caused by distortion of the physical perception sequence, the system initiates a virtual-real coordinated adjustment mechanism. The state evolution mapping unit retrieves the drive feedback data output by the instruction execution drive unit at the previous moment and inputs it into the dynamic mathematical model of the controlled object to generate the predicted state feedback flow. The logical fusion arbitration unit performs feature analysis on the physical sensing sequence acquired by the feedback data sampling unit, and calculates the short-time autocorrelation function R(m) of the sequence within the sliding sampling window. The calculation formula is as follows: Where R(m) is the short-time autocorrelation function, m is the delay step size, n is the sampling sequence index, and N is the total number of sampling points. As physical state parameters, the logic fusion arbitration unit fits the logarithmic decay trend of the short-time autocorrelation function R(m) using the least squares method to determine the autocorrelation spectrum slope α. When the autocorrelation spectrum slope α deviates from the preset stable range, it indicates that the energy distribution of the physical sampling signal has undergone random distortion. The logic fusion arbitration unit determines the fusion weight coefficient ω based on the autocorrelation spectrum slope α, and the calculation formula is as follows: , where ω is the fusion weight coefficient, ΔH is the increment of the feedback divergence eigenvalue between the current sampling period and the previous sampling period, and α is the slope of the autocorrelation spectrum.

[0032] The logical fusion arbitration unit adjusts the predicted state feedback flow based on the fusion weight coefficient ω. Predicted state feedback flow weight coefficient And reduce the physical state parameters accordingly. Physical state parameter weighting coefficients This allows the generation logic of collaborative control commands to switch from passive perception to active prediction and guidance; the command execution drive unit receives the collaborative adjustment vector output by the logic fusion arbitration unit. The system outputs the back EMF fluctuation characteristics at the execution end in real time after the adjustment action is performed. The state evolution mapping unit compares the back EMF fluctuation characteristics with the expected response of the dynamic mathematical model. When the deviation between the two continues to accumulate, the damping coefficient of the dynamic mathematical model is corrected. This calibration process makes the predicted state feedback flow more accurate. By continuously approximating the evolution of the physical properties of the controlled object, under this working condition, the inspection robot is in a 100ms interference window period when the sensor sampling fails, but still maintains the preset heading according to the estimated state feedback flow, limiting the displacement overshoot to within 3mm.

[0033] Example 2: When verifying the trajectory control stability of the collaborative operation and maintenance system under strong electromagnetic interference, an experimental platform including a physical simulation guide rail and a servo control terminal was used. The data of the experimental platform comes from the physical sensing sequence (sensor measurement sequence), which captures the physical state parameters of the controlled object through an electromagnetically isolated displacement gauge with a sampling accuracy of 0.01 mm and a maximum sampling rate of not less than 200 Hz. To balance the real-time performance of system regulation with the logic processing load, the system sets the sampling period to 5ms. The logic for determining this parameter is as follows: when the motion bandwidth frequency of the controlled object is in the range of 10Hz to 20Hz during operation and maintenance, in order to satisfy the Nyquist sampling law and reserve more than three times the prediction phase margin, the sampling frequency is selected as 10 times the signal bandwidth. Under this test condition, the system actively injects broadband noise with a signal-to-noise ratio of 15dB and second harmonic interference with a frequency of 50Hz into the feedback channel to simulate the transient electromagnetic stress generated when the high-voltage equipment inside the substation operates.

[0034] The experimental process evaluated the synergistic effect by setting up a sample group of the present invention and two control groups. The sample group of the present invention ran the logic of the complete implementation of the specific method. Control group one removed the state evolution mapping unit, and control group two disabled the elastic arbitration mechanism in the logic fusion arbitration unit and used a fixed weighting coefficient of 0.5 to process the physical signal and virtual prediction value. During the observation period with injected transient interference, due to the nonlinear distortion of the physical feedback channel caused by noise stress, the logic fusion arbitration unit in the sample group of the present invention detected that the autocorrelation spectrum slope α jumped from 0.12 under normal conditions to 0.92, triggering the fusion weighting coefficient ω to decay from 0.88 to 0.08. The system automatically adjusted the predicted state feedback flow based on ω. Predicted state feedback flow weight coefficient The compensation was reduced to 0.96, thus relying on the inertial characteristics of the dynamic mathematical model to maintain the trajectory of the controlled object. In the gradient verification of the interference intensity, as the signal-to-noise ratio of the injected noise decreased from 30dB to 10dB, the distortion of physical perception increased. Monitoring data showed that when the signal-to-noise ratio was at a moderate interference intensity of 20dB, the measured trajectory deviation of control group 1 was 3.42mm, the measured trajectory deviation of control group 2 was 1.85mm, while the deviation of the present invention sample group was locked at 0.24mm. When the interference intensity was further increased to 10dB, the control group 1 experienced mechanical oscillation of the servo motor due to the failure of physical signal feedback, and control group 2 showed an overshoot drift of 5.12mm. The present invention sample group limited the command output slope through logic damping characteristics, and its measured trajectory deviation was 0.41mm. This phenomenon confirms the predicted state feedback flow generated by the state evolution mapping unit. It provides a logical reference for the logical fusion arbitration unit, and the trajectory stability generated by the deep coupling between the two is better than the simple sum of the functions of each unit.

[0035] Empirical studies on the boundaries of key parameters show that when the threshold for the autocorrelation spectrum slope α is set within the range of 0.85 to 0.95, the system's sensitivity to unstructured disturbances reaches its peak, manifested in the dynamic division of feedback path weights within 8ms after the disturbance occurs. If the threshold for the autocorrelation spectrum slope α is lowered to 0.60, the system experiences false triggering, leading to accumulated tracking errors in the controlled object under stable conditions due to over-reliance on the dynamic mathematical model. If the threshold for the autocorrelation spectrum slope α is raised to 1.10 (out-of-range operation), the system's perception of energy distortion decreases, resulting in a lag in the weight switching process. The transient overshoot of the controlled object at the moment of disturbance increases from 0.41mm to 2.68mm. This indicates that the threshold range of 0.85 to 0.95 is a working window that balances perception sensitivity and control continuity, suppressing step responses in the controlled object's actuators and ensuring coordinated adjustment vectors. Under dynamic stress interference, this experiment, through the analysis of deviation data under different noise gradients, confirms that the advanced feedback information generated by the dynamic mathematical model can offset the physical lag of the sensing channel. The logical fusion arbitration unit executes decisions based on the evolution of autocorrelation characteristics, enabling the system to shift from lag-based passive sensing to predictive active adjustment.

[0036] Example 3: This example combines Figures 1 to 2 This section describes a collaborative operation and maintenance control system based on the fusion of AR virtual overlay perception feedback and manual inspection, such as... Figure 1As shown, a closed-loop feedback loop includes a feedback data sampling unit, a state evolution mapping unit, a logic fusion arbitration unit, and an instruction execution drive unit. The feedback data sampling unit is directly connected to the controlled object and is responsible for receiving the physical signals generated by the controlled object. By acquiring the physical sensing sequence and extracting statistical distribution characteristics, the generated physical sensing sequence is sent to the state evolution mapping unit and the logic fusion arbitration unit respectively. The state evolution mapping unit constructs a dynamic mathematical model based on the received physical sensing sequence and generates an advanced estimated state feedback stream, which is then transmitted to the logic fusion arbitration unit. The logic fusion arbitration unit determines the feedback divergence characteristic value based on the physical sensing sequence and calculates the fusion weight coefficient. After performing a weighted fusion operation on the physical sensing sequence and the estimated state feedback stream, it generates a cooperative control instruction. The instruction execution drive unit receives the cooperative control instruction and converts it into a control pulse signal, which is then applied to the controlled object through physical adjustment actions. After the controlled object generates real-time state and environmental disturbance data, it feeds back to the feedback data sampling unit in the form of physical signals, thus forming a complete cooperative operation and maintenance control loop.

[0037] like Figure 2 As shown, the overall logical structure is divided into three functional domains: top-level, middle-level, and bottom-level. The top-level domain is defined as the logical decision-making domain, deployed in the cloud or main control environment, and its core includes a logical fusion arbitration unit. The middle-level domain is defined as the virtual mapping domain, which relies on digital twin or edge computing technology and has a state evolution mapping unit deployed at its core. The bottom-level domain is defined as the physical execution domain, located at the field terminal, and includes a feedback data sampling unit, controlled objects, and an instruction execution drive unit. In terms of data flow between layers, the feedback data sampling unit of the bottom layer provides the data required for model building to the middle-level state evolution mapping unit, and sends physical perception sequences to the top-level logical fusion arbitration unit. The middle-level state evolution mapping unit transmits the generated predicted state feedback stream to the top-level logical fusion arbitration unit. After completing arbitration and fusion, the top-level logical fusion arbitration unit sends the collaborative control instructions across levels to the bottom-level instruction execution drive unit, realizing vertical collaboration between the physical execution domain, the virtual mapping domain, and the logical decision-making domain.

[0038] Example 4: In a scenario where a wheeled inspection robot carrying a variable mass load detection device operates on a non-uniform friction surface, the dynamic characteristics of the controlled object undergo transient drift as the load is grasped, and the physical displacement gauge generates a nonlinear feedback signal containing high-frequency random spikes on the rough surface. The system is set to an initial technical state, wherein the system sampling frequency is... The controlled object's dynamic frequency components during the acceleration phase are between 5Hz and 15Hz, with a frequency of 200Hz. To balance the calculation accuracy of the short-time autocorrelation function R(m) with the response time delay of the controlled object's actuator, the system configures a window length calibration procedure to determine the sliding window length L. The logic fusion arbitration unit monitors the acceleration increment of the controlled object's command execution, and if the acceleration increment exceeds a preset dynamic threshold... In the initial stage, L is set to 10 sampling points to shorten the feature extraction delay. During steady-state operation, L is expanded to 30 sampling points. The averaging effect of autocorrelation operations is used to suppress random fluctuations in the physical sensing sequence, thereby filtering out unstructured disturbances caused by ground unevenness at the physical level. To eliminate the mismatch risk between the dynamic mathematical model in the state evolution mapping unit and the evolution of the controlled object's physical properties, the system executes a correction step size determination procedure based on residual variance estimation. The state evolution mapping unit receives the drive feedback data output by the instruction execution drive unit and retrieves the physical state parameters at the current moment. With the predicted state feedback flow The instantaneous deviation e between the samples is used to extract the instantaneous deviation sequence over 20 consecutive sampling periods using the transfer function self-correction loop, and the discrete standard deviation σ of the instantaneous deviation sequence is calculated. The correction increment factor is then determined according to the following formula. : Where μ is the correction increment factor. The preset base correction factor is 0.001, and σ is the discrete standard deviation of the instantaneous deviation sequence. The system has a preset maximum deviation tolerance limit; the state evolution mapping unit adjusts the coefficients of the denominator polynomial of the dynamic mathematical model according to the correction increment factor μ, so that the predicted state feedback flow... The actual evolution trend of the physical entity was refitted within three sampling periods after a sudden change in the load mass of the controlled object, thus solving the problem of asynchronous virtual and real feedback caused by mechanical parameter drift.

[0039] Under these conditions, the inspection robot moves from a smooth surface onto a gravel surface, and the feedback data sampling unit... Performing a 128-point Discrete Fourier Transform, an abnormal increase in the power spectral density distribution of the signal was detected near 100Hz. The logic fusion arbitration unit quickly calculated the autocorrelation spectrum slope α using an adaptively adjusted sliding window and identified its deviation from the stationary region. The system forced the weighting coefficients of the physical state parameters of the physical sampled signal to be adjusted by the fusion weighting coefficient ω. The value is reduced to 0.12, and a high-confidence predicted state feedback flow is output from the dynamic mathematical model after correction using the self-correcting loop of the transfer function. The logic fusion arbitration unit takes over the control of the closed-loop control circuit, preventing sampling jumps caused by random ground impacts from being transmitted to the command execution drive unit. Actual test results show that the overshoot of the controlled object's trajectory decreased from 12.5mm to 1.8mm, confirming that by deeply coupling the parameter adaptive calibration procedure with virtual and real feedback, the dynamic determinism of the control system can be improved without adding external physical vibration damping devices. In substation control scenarios where the controlled object faces intermittent packet loss in the feedback data sampling unit's communication link, a feedback flow maintenance procedure is executed for the sampling gap period, and the feedback data sampling unit monitors the physical state parameters. If the update frequency is lower than the preset 200Hz sampling reference value, the logic fusion arbitration unit determines that the physical feedback channel has entered a data missing state and forcibly sets the fusion weight coefficient ω to 0; the state evolution mapping unit uses the real-time corrected dynamic mathematical model to perform integral calculations to compensate for the displacement increment during the sampling gap period, so that the coordinated adjustment vector The calculations depend entirely on the predicted state feedback flow. The provided predictive potential energy maintains the inertial trajectory of the controlled object during the interruption of the sensing link, avoiding uncontrolled response of the actuator due to loss of feedback data.

[0040] Example 5: In the static parameter calibration scenario before the collaborative operation and maintenance system performs its first task, the system runs the consistency verification procedure of the feedback path, and the controlled object performs a stepping motion within the laboratory track; the feedback data sampling unit acquires the physical state parameters of the controlled object. The logic fusion arbitration unit extracts the discretized statistical distribution characteristics of the sampled signal; the state evolution mapping unit aligns the preset control impulse response curve with the statistical distribution characteristics of the measured physical sensing sequence in the time domain to determine the fixed phase difference of the physical sensing; the logic fusion arbitration unit calculates the background energy distribution characteristics of the physical sensing sequence in an interference-free environment, and determines the confidence threshold for distortion judgment based on the fluctuation amplitude of the autocorrelation spectrum slope α in a steady state; this procedure establishes the synchronization relationship between physical feedback and dynamic mathematical model in the initial state, providing a reference zero point for the weight allocation logic in the dynamic adjustment process.

[0041] When the system encounters a situation where the actuator of the controlled object experiences mechanical backlash drift due to service time evolution, the system operates according to the field recalibration procedure for the command response characteristics. The command execution drive unit sends a small-range excitation pulse when the controlled object is in a standby state. The state evolution mapping unit synchronously captures the back electromotive force fluctuation characteristics fed back from the motor to the feedback data sampling unit and calculates the real-time rotational inertia of the controlled object based on these characteristics. The logic fusion arbitration unit retrieves the transfer function coefficients of the dynamic mathematical model and performs incremental correction based on the deviation between the measured rotational inertia and the model's preset value, converting the physical losses at the actuator end into damping compensation parameters within the model. The corrected dynamic mathematical model is then adjusted through a coordinated vector... Maintain the operational accuracy of the controlled object, so that the output slope of the cooperative control command is controlled by the physical response speed of the current hardware entity.

[0042] Example 6: In the feedback path benchmark calibration scenario during the initial deployment of the collaborative operation and maintenance control system, the system runs a quantitative evaluation procedure for the feedback channel noise floor. The feedback data sampling unit continuously acquires the physical state parameters of 1000 sampling points when the controlled object is in a static reference state. The logic fusion arbitration unit calculates the physical state parameters of 1000 sampling points. The statistical variance was determined and used as the background noise baseline. The logic fusion arbitration unit extracts the autocorrelation spectrum slope under the reference state through a sliding sampling window with a length of 30 sampling points. The system determines the judgment threshold of the sensing channel according to the following formula. : The coefficient 1.2 is a safety redundancy operator determined based on the floating-point error calculated by the embedded processor. This procedure, by capturing the characteristics of the electromagnetic background noise in the field, transforms the statistical criteria in the feedback logic into quantitative indicators coupled with the physical environment, and establishes the synchronization relationship between physical feedback and dynamic mathematical model in the initial state.

[0043] When the system encounters a situation where the actuators of the controlled object experience mechanical backlash drift due to service time evolution, the system operates by recalibrating the response characteristics of the command on-site, and the logical fusion arbitration unit calculates the physical state parameters in real time. With the predicted state feedback flow Normalized residual norm between In the normalized residual norm When the mismatch exceeds the preset limit of 0.15 for five consecutive sampling periods, the state evolution mapping unit is triggered to enter the model parameter reset state. In this state, the instruction execution drive unit sends excitation pulses to the controlled object, and the state evolution mapping unit simultaneously captures the back electromotive force fluctuation characteristics fed back by the motor, and calculates the real-time rotational inertia of the controlled object based on these characteristics, thereby performing a full update of the transfer function coefficients of the dynamic mathematical model. This procedure utilizes the statistical distribution characteristics of the measured physical sensing sequence to perform phase alignment on the drifting dynamic mathematical model, ensuring the predicted state feedback flow... The temporal characteristics are corrected to the range of the hardware entity's response by coordinating the adjustment vector. Maintain the adjustment accuracy of the controlled object.

[0044] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A collaborative operation and maintenance control system based on the fusion of AR virtual overlay perception feedback and manual inspection, characterized in that, The system includes: The feedback data sampling unit is used to acquire the physical sensing sequence and environmental disturbance data that characterize the real-time state of the controlled object; The state evolution mapping unit is used to construct a dynamic mathematical model of the controlled object based on the physical perception sequence, and to use the dynamic mathematical model to generate an estimated state feedback flow that is ahead of the sampling time sequence of the physical perception sequence. The logical fusion arbitration unit, connected to the feedback data sampling unit and the state evolution mapping unit respectively, is used to perform the following steps: Step S101, extract the statistical distribution characteristics of the physical sensing sequence and calculate the feedback divergence feature value, which characterizes the degree of dispersion of the statistical distribution characteristics; Step S102, extract the evolution characteristics of the feedback divergence feature value with the sampling period, and determine the confidence state of the sensing channel in combination with a preset confidence threshold, and determine the fusion weight coefficient based on the confidence state; Step S103, perform a weighted fusion operation on the physical sensing sequence and the estimated state feedback stream based on the fusion weight coefficient to generate a cooperative control command; The instruction execution drive unit is used to receive cooperative control instructions and convert them into control pulse signals to adjust the physical operating state of the controlled object.

2. The collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, The state evolution mapping unit is equipped with a transfer function self-correction loop, which is used to perform the following steps: obtain the driving feedback data output by the instruction execution driving unit, and use the driving feedback data as the basis for model inversion to correct the transfer function coefficients of the dynamic mathematical model in real time; adjust the frequency response characteristic of the predicted state feedback flow based on the corrected transfer function coefficients.

3. The collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, When executing step S102, the logic fusion arbitration unit determines the fusion weight coefficient in the following manner. When the feedback divergence feature value exceeds the preset confidence threshold, calculate the increment of the feedback divergence feature value between the current sampling period and the previous sampling period. ; The fusion weight coefficient is determined by the following formula. : , where α is the slope of the autocorrelation spectrum; the logic fusion arbitration unit adjusts the weight of the predicted state feedback flow in the collaborative control command generation logic according to the fusion weight coefficient ω.

4. The collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, The logic fusion arbitration unit is also used to perform the following steps: Step S401, monitor the rate of change of the physical sensing sequence, and when the rate of change of the physical sensing sequence exceeds the preset jump threshold and the feedback divergence characteristic value is in the non-steady-state range, activate the elastic gating of the feedback path. Step S402: Introduce logic damping features into the cooperative control command through elastic gating to limit the output slope of the cooperative control command.

5. A collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, The state evolution mapping unit uses the estimated state feedback flow to pre-set the prediction compensation amount before the instruction execution driving unit takes action, and compensates for the phase delay of the sensing channel based on the feedback lag time of the physical sensing sequence.

6. The collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, The feedback data sampling unit includes a sensor health monitoring module, which monitors the power spectral density of the sampled signal. When the correlation of the power spectral density is less than 0.75, the state evolution mapping unit is triggered to switch to the dynamic mathematical model inertial maintenance mode.

7. A collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, When executing step S103, the logic fusion arbitration unit uses a nonlinear saturation operator to restrict the state control parameters of the controlled object within the logical safety boundary defined by the estimated state feedback flow.

8. A collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, The instruction execution drive unit includes an execution deviation closed loop, which is used to perform residual compensation on the control pulse signal based on the deviation between the measured displacement of the instruction execution drive unit and the calculated expected value of the cooperative control instruction.

9. A collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1 or 7, characterized in that, The system also includes an environmental stress processing unit, which converts environmental disturbance data into disturbance gain coefficients and performs amplitude weighting on the logical safety boundary of the estimated state feedback flow.

10. A collaborative operation and maintenance control system based on AR virtual overlay perception feedback and manual inspection fusion as described in claim 1, characterized in that, The system also includes an operation and maintenance status tracking unit, which maps the timing waveform trajectory of the coordinated control commands to the health assessment value of the controlled object.

Images