Remote operation control method and system for crane

Abstract

The invention relates to the technical field of crane remote operation control, and particularly discloses a crane remote operation control method and system.The method comprises the steps that theoretical feedback force and actual feedback force of crane remote operation are collected, and a cumulative distortion sequence is extracted through a joint distortion criterion; secondly, extracting coupling analysis characteristics of dynamic parameters of the hanging object and an accumulative distortion sequence, and customizing a dynamic calibration baseline of actual feedback force for different risks; adjusting the output force of the operating rod based on the calibration base line, analyzing the matching between the behavior characteristics of the operator and distortion accumulation, and judging whether cognitive overload occurs or not; if overload occurs, an overload scene data set is constructed to determine early warning feature time, and a personalized optimal feedback force mapping relation is generated through a relation adaptation model. The system comprises a distortion extraction module, a baseline calibration module, an overload identification module and a feedback adjustment module. The method is beneficial to correcting feedback force distortion and relieving cognitive overload of an operator.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote operation control of cranes, and particularly relates to a remote operation control method and system of cranes. BACKGROUND

[0002] Although the current remote operation technology of cranes has broken through the spatial limitation of traditional on-site operation, in the core interaction link of remote operation, there is a problem of coordinated out-of-control of feedback force distortion and cognitive overload of operators, which directly restricts the control accuracy and safety of remote operation.

[0003] Feedback force distortion makes the force information obtained by the operator disengaged from the actual state of the hoisted object, and cognitive overload further causes the operator to be difficult to judge the operation timing and force, thereby reducing the control accuracy of remote operation, and easily causing problems such as hoisted object shaking and positioning deviation in scenarios such as precise hoisting and narrow space operation, and even existing safety hazards.

[0004] Therefore, the present application provides a remote operation control method and system of cranes. SUMMARY

[0005] The present application aims to provide a remote operation control method and system of cranes to solve the above background problems.

[0006] The purpose of the present application can be achieved by the following technical solutions: A remote operation control method of cranes: comprising the following steps: Collecting a feedback deviation gradient of remote operation of the crane, establishing a joint distortion criterion based on the feedback deviation gradient and extracting a cumulative distortion sequence; Through coupling analysis of the hoisted object dynamic parameters and the cumulative distortion sequence, a historical risk group is constructed based on the coupling analysis features, and a real-time risk label type is extracted, and a dynamic calibration baseline of actual feedback force is formulated for different types; Based on the dynamic calibration baseline, the output force of the operation lever is adjusted in real time, the matching of the length of the continuous same direction correction chain of the operator after adjustment and the distortion accumulation is evaluated, if they are not matched, the cognitive load of the operator is analyzed, the cognitive load degree is obtained, and it is determined whether the cognitive load of the operator is overloaded; If the cognitive load is in an overloaded state, an overload scene data set of distortion accumulation and calibration lag is extracted, a warning time is extracted from the scene data set to determine a warning feature time, and a relationship adaptation model is constructed to formulate an optimal feedback force mapping relationship for different operators.

[0007] As a further scheme of the present application: the way of establishing the joint distortion criterion is: Obtaining a window average gradient in a sliding window; The instantaneous pulse distortion criterion is established based on the feedback deviation gradient and the window average gradient, and the instantaneous pulse distortion existing in the sliding window is identified and removed; Based on the removed instantaneous pulse distortion, the cumulative distortion criterion is constructed to identify the cumulative distortion in the sliding window. If the feedback deviation gradient meets the cumulative distortion criterion, a cumulative distortion sequence containing the feedback deviation gradient meeting the cumulative distortion criterion is constructed.

[0008] As a further scheme of the present application, the way to formulate the dynamic calibration baseline of the actual feedback force is: The sub-features of the historical coupling analysis features are marked as the label type of risk, and the historical label type of risk and the coupling analysis features are taken as a historical risk group; The number proportion of different label types of risk in all historical risk groups is extracted as the prior distribution of each label type of risk; The joint probability density function of the coupling analysis features containing the sub-features is extracted; The current coupling analysis features are obtained to establish a new real-time risk group, and the posterior probability is obtained by combining the prior distribution, the joint probability density function and the Bayes formula; The risk label type of the real-time risk group is extracted based on the posterior probability, and different dynamic calibration baselines of the actual feedback force are formulated based on different label types of risk.

[0009] As a further scheme of the present application, the way to obtain the sub-features of the coupling analysis features is: The peak value of the cumulative distortion sequence and the time stamp corresponding to the extreme value of the swing phase sequence are obtained, the peak value of the feedback deviation gradient is matched with the extreme value of the swing phase sequence closest to the time stamp based on the nearest neighbor pairing method, and a synchronous matching pair is obtained; The difference between the time stamps corresponding to the peak value and the extreme value in the synchronous matching pair is calculated to obtain the swing phase difference; The swing phase difference of the synchronous matching pair is subjected to phase synchronization analysis to obtain a phase synchronization coefficient; The moment of inertia of the hoisted object and the shape coefficient of the hoisted object are extracted from the dynamic parameters of the hoisted object, and a distortion amplification coefficient is obtained by constructing a distortion amplification equation; The length of the continuous same direction gradient chain in the cumulative distortion sequence is extracted to obtain the maximum gradient chain length; The maximum gradient chain length, the distortion amplification coefficient and the phase synchronization coefficient are taken as the sub-features of the coupling analysis features of the feedback force and the state of the hoisted object of the remote operation of the crane.

[0010] As a further scheme of the present application, the way to perform the cognitive load analysis is: The probability that the standard deviation of the feedback correlation coefficient falls in different sub-intervals is obtained, and the fluctuation entropy of the feedback correlation coefficient is obtained by inputting an entropy value equation; extracting the feedback bias gradient in the corrected cumulative distortion sequence, calculating the distortion complexity entropy; obtaining the cognitive load degree by constructing a cognitive load equation based on the distortion complexity entropy and the fluctuation entropy of the feedback correlation coefficient; comparing and analyzing the cognitive load degree to obtain a determination result of whether the cognitive load of the operator is in an overload state.

[0011] As a further scheme of the present application, the feedback correlation coefficient is obtained in the following manner: extracting the length of the continuous same-direction correction chain of the operator, performing chain length correction analysis on the length of the continuous same-direction correction chain in combination with the maximum gradient chain length, obtaining a chain length correction ratio, and evaluating the matching of the continuous correction behavior of the operator and the distortion accumulation; if the matching is not matched, performing correlation analysis based on the behavior characteristics of the operator in combination with the dynamic adjustment characteristics of the actual feedback force to obtain the feedback correlation coefficient.

[0012] As a further scheme of the present application, the chain length correction analysis is performed in the following manner: obtaining the maximum number of continuous same-direction corrections performed by the operator to obtain the length of the continuous same-direction correction chain; obtaining the maximum gradient chain length of the cumulative distortion sequence, calculating the deviation ratio of the length of the continuous same-direction correction chain and the maximum gradient chain length, and obtaining the chain length correction ratio; evaluating the matching of the direction and the continuous number of characteristics of the continuous same-direction correction behavior of the operator and the distortion accumulation based on the chain length correction ratio.

[0013] As a further scheme of the present application, the preferred feedback force mapping relationship is formulated in the following manner: collecting overload scene data during the cognitive overload period to construct an overload scene data set; based on the overload scene data set, establishing a time chain correlation of the distortion accumulation and the calibration, and extracting a lag feature vector forming each overload scene; based on the lag feature vector of each overload scene, determining the early warning characteristic time of the cognitive overload of different operators; and based on the early warning characteristic time window, extracting the operator behavior adaptation type; based on the determined operator behavior adaptation type, constructing a relationship adaptation model, inputting the behavior characteristics of the operator in the overload scene data set into the relationship adaptation model, and generating the preferred feedback force mapping relationship.

[0014] As a further scheme of the present application, the operator behavior type is divided in the following manner: intercepting a time window from T seconds before the early warning characteristic time point to the early warning time point as an early warning analysis window; wherein T is the early warning characteristic time value; Extract the continuous same direction correction chain length and the force change rate original sequence in the early warning analysis window from the overload preposition data; calculate the distribution density of the continuous same direction correction chain length and the peak value coefficient of the force change rate in the early warning analysis window; Combined with the time sequence change trend of the feedback correlation coefficient in the early warning analysis window, the operator behavior type is divided: the operator behavior adaptation type is determined based on the distribution density and the peak value coefficient of the force change rate.

[0015] A crane remote operation control system comprises the following modules: A distortion extraction module is used for collecting a feedback deviation gradient of the crane remote operation, establishing a joint distortion criterion based on the feedback deviation gradient and extracting a cumulative distortion sequence; A baseline calibration module is used for extracting a coupling analysis feature of the object dynamic parameter and the cumulative distortion sequence, constructing a historical risk group based on the coupling analysis feature, extracting a real-time risk label type, formulating a dynamic calibration baseline for different types, and adjusting the output force of the operating lever based on the dynamic calibration baseline; An overload identification module is used for evaluating the matching of the continuous same direction correction chain length and the distortion accumulation of the operator after adjustment, extracting an adjusted feedback correlation coefficient if the matching is not matched, performing a cognitive load analysis on the operator based on the feedback correlation coefficient, obtaining a cognitive load degree, and determining whether the cognitive load of the operator is overloaded; A feedback adjustment module is used for extracting an overload scene data set of the distortion accumulation and the calibration lag if the cognitive load is in an overload state, extracting a pre-warning time to determine a pre-warning feature time, constructing a relationship adaptation model, and formulating an optimal feedback force mapping relationship for different operators.

[0016] The present application has the following beneficial effects: The actual weight of the object is collected by a crane hook pressure sensor to correlate the theoretical feedback force, the actual feedback force of the operating lever is collected by a remote control device sensor, the feedback deviation gradient is calculated after time alignment and sliding window calculation, accidental interference is removed by an instantaneous pulse criterion, and a continuous distortion sequence is extracted by a cumulative distortion criterion. The cumulative distortion sequence extracted is beneficial to realize the real long-term change law of the feedback force, and provide data support for subsequent calibration.

[0017] The object swing phase, inertia moment and other dynamic parameters are synchronously extracted, the phase synchronization coefficient is calculated by nearest neighbor pairing, the distortion amplification coefficient is constructed in combination with the object characteristics, and the maximum gradient chain length is extracted as the coupling feature; the risk label is determined based on the historical risk group and the Bayesian formula, and the differentiated baseline is customized for high, medium and low risk scenes, so that the baseline is adapted to the object dynamics and distortion risk depth, which is beneficial to the calibration strategy to be independent of the actual working condition, and improve the flexibility and accuracy of the remote operation feedback force adjustment.

[0018] After adjusting the operating lever output force based on the calibration baseline, the operator's continuous same direction correction chain length and force change rate are collected, the matching of the behavior and distortion is evaluated through the chain length correction ratio, and the feedback correlation coefficient is calculated when they are not matched, and the cognitive load degree is calculated by combining the fluctuation entropy and distortion complexity entropy. It is beneficial to quantify the adaptability of the operator's behavior and feedback force, can identify signs of cognitive overload in time, and reduce operation errors caused by heavy cognitive burden of the operator.

[0019] A dataset containing overload preposition, scene and calibration data is constructed, the early warning time is determined by calculating the lag feature vector through time chain, the operator type is divided according to the behavior index, and the personalized feedback mapping is generated by using the gradient boosting regression model. It is beneficial to early warning of potential overload risk, and can adapt to different operator habits and reduce the probability of subsequent cognitive overload. BRIEF DESCRIPTION OF DRAWINGS

[0020] The application will be further described below in conjunction with the drawings.

[0021] Figure 1 is a flowchart of a crane remote operation control method of the application; Figure 2 is a determination flowchart of whether the cognitive load of the operator is overloaded in the application; Figure 3 is a module diagram of a crane remote operation control system of the application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the application will be clearly and completely described below in conjunction with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0023] Embodiment 1

[0024] As shown in Figure 1 , the application is a crane remote operation control method, comprising the following steps: S1, collecting the feedback deviation gradient of the crane remote operation, establishing a joint distortion criterion based on the feedback deviation gradient and extracting the cumulative distortion sequence; The way of collecting the feedback deviation gradient of the crane remote operation is: Preferably, the actual weight of the hoisted object is collected by a pressure sensor at the crane hook, and the actual weight of the hoisted object is input into the crane control system; The mapping relationship between the actual weight of the hoisted object and the preset theoretical feedback force is obtained, and the theoretical feedback force corresponding to the actual weight of the hoisted object is extracted; The pressure sensor of the remote control device of the crane collects the actual feedback force of the operating lever on the remote control device; The theoretical feedback force and the actual feedback force are data preprocessed, and the theoretical feedback force and the actual feedback force are time-aligned; Based on the time-aligned theoretical feedback force and the actual feedback force, the deviation of the theoretical feedback force and the actual feedback force in each sliding window is calculated by setting a sliding window, and a feedback deviation value is obtained; All feedback deviation values in the sliding window are obtained, the change gradient of adjacent feedback deviation values is calculated, and a feedback deviation gradient is obtained; Among them, the joint distortion criterion is established based on the feedback deviation gradient and the cumulative distortion sequence is extracted; The mean of the absolute value of the feedback deviation gradient in the sliding window is calculated as the window average gradient ; Among them, the way to establish the joint distortion criterion and extract the cumulative distortion sequence is: It should be noted that the joint distortion criterion includes: instantaneous pulse distortion criterion and cumulative distortion criterion; S101, based on the feedback deviation gradient and the window average gradient, an instantaneous pulse distortion criterion is established, and instantaneous pulse distortion existing in the sliding window is identified and removed; Preferably, the instantaneous pulse criterion is established by the equation: ; Preferably, ; Among them, is the k feedback deviation gradient in the sliding window, is the window average gradient, and k is the number of feedback deviation gradients; If the feedback deviation gradient of the feedback deviation value meets the instantaneous pulse criterion, the feedback deviation value is determined as an instantaneous pulse distortion; The instantaneous pulse distortion in the sliding window is removed; S102, based on the removed instantaneous pulse distortion, a cumulative distortion criterion is constructed to identify the cumulative distortion in the sliding window; Preferably, the cumulative distortion criterion is established by the equation: ; Among them, is the feedback deviation gradient of the removed instantaneous pulse distortion, b is the number of feedback deviation gradients, and m is the number of feedback deviation gradients; Preferably, ; M is the total number of instantaneous pulse distortions in the sliding window; If the feedback deviation gradient satisfies the cumulative distortion criterion, the corresponding feedback deviation gradient is obtained, and a cumulative distortion sequence containing the feedback deviation gradient satisfying the cumulative distortion criterion is constructed. It can be understood that the function of obtaining the feedback deviation gradient is to: Function one: by establishing an instantaneous pulse distortion criterion with the window average gradient to eliminate incidental interference, and then constructing a cumulative distortion criterion based on the feedback deviation gradient after eliminating interference, the cumulative distortion sequence is finally extracted, which provides effective data support for subsequent feedback force distortion correction. Function two: feedback deviation gradient supports coupling analysis feature extraction and cognitive load quantification analysis: on the one hand, the cumulative distortion sequence generated by it can provide sub-features such as peak value and maximum gradient chain length, which can help risk label determination and dynamic calibration baseline development; on the other hand, the feedback deviation gradient in the corrected cumulative distortion sequence can be used to calculate distortion complexity entropy, which can further assist in cognitive load calculation and overload determination.

[0025] S2, by extracting the coupling analysis features of the hoisted object dynamic parameters and the cumulative distortion sequence, constructing a historical risk group based on the coupling analysis features, and extracting real-time risk label types, a dynamic calibration baseline of actual feedback force is developed for different types; The starting time and the terminal time of the cumulative distortion sequence are obtained, and the hoisted object dynamic parameters of the crane from the starting time to the terminal time are extracted synchronously; The hoisted object dynamic parameters include: swing phase, swing acceleration, and moment of inertia of the hoisted object. Based on the hoisted object dynamic parameters of the crane from the starting time to the terminal time, the swing phase of the hoisted object is extracted, and a swing phase sequence is established. The peak value of the feedback deviation gradient in the cumulative distortion sequence and the extreme value of the swing phase sequence are calculated. The time stamps corresponding to the peak value and the extreme value are obtained, and the extreme value of the swing phase sequence closest to the peak value of the feedback deviation gradient in the time stamp is matched based on the nearest neighbor pairing method, to obtain a synchronous matching pair. The difference between the time stamps corresponding to the peak value and the extreme value in the synchronous matching pair is calculated to obtain the swing phase difference. The swing phase difference of the synchronous matching pair is analyzed in phase synchronization to obtain a phase synchronization coefficient. The phase synchronization analysis is performed in the following manner: It should be noted that the peak value and the extreme value can only be uniquely matched when matched by the nearest neighbor pairing method; if the number of peak values and extreme values is not equal, the nearest neighbor pairing is performed according to the parameter with the smallest number (peak value or extreme value). The swing phase difference of the synchronous matching pair is screened and processed, and the number of strong synchronization synchronous matching pairs is extracted. Preferably, when the swing phase difference is lower than or equal to 0.5s, the synchronous matching pair is considered as a strong synchronous synchronous matching pair; The number of strong synchronous synchronous matching pairs is ratio processed with the number of all synchronous matching pairs to obtain a phase synchronization coefficient S; The lifting object inertia moment I and the lifting object shape coefficient Xz of the lifting object dynamic parameter are extracted, and a distortion amplification equation is constructed: A distortion amplification coefficient A is obtained; wherein, is a calibration coefficient, which is obtained by fitting historical data; It should be noted that the lifting object shape coefficient is based on the geometric shape of the lifting object (such as block, column, irregular, etc.), the lifting force mode (such as single-point lifting, multi-point lifting), and is obtained by querying a preset lifting object shape coefficient database (containing coefficient values corresponding to different shapes and lifting modes) or using an industry general shape coefficient calculation model, and finally used to construct a distortion amplification equation together with the lifting object inertia moment to calculate the distortion amplification coefficient; The calibration coefficient is an empirical fitting coefficient, which is obtained based on distortion experiments of 100 different types of lifting objects (including block, column, irregular shape); is a reference inertia moment, for example, the inertia moment of a 100kg standard block-shaped lifting object; The length of the continuous same direction gradient chain in the cumulative distortion sequence is extracted to obtain the maximum gradient chain length L; It should be noted that the maximum gradient chain length refers to the maximum continuous number of times that the feedback deviation gradient direction continuously remains consistent (all positive or all negative) in the cumulative distortion sequence; The maximum gradient chain length, the distortion amplification coefficient, and the phase synchronization coefficient are used as sub-features of the coupling analysis features of the feedback force of the remote operation of the crane and the state of the lifting object; The label type d of the historical coupling analysis feature that is marked as a risk is obtained, and the historical risk label type and the coupling analysis feature are used as a historical risk group; wherein, the risk label type d=0, 1, 2, respectively corresponding to high risk, medium risk, and low risk; The number proportion of different risk label types in all historical risk groups is extracted as the prior distribution of each risk label type; For each risk label type d, the joint probability density function of the maximum gradient chain length L, the distortion amplification coefficient A, and the phase synchronization coefficient S of the historical risk group is fitted by kernel density estimation ; The skilled in the art can understand that, for each risk label type d, the risk scenarios marked as d in the historical data are first screened, the maximum gradient chain length L, the distortion amplification coefficient A, and the phase synchronization coefficient S under the scenario are extracted to form a three-dimensional sample set, and then a kernel function such as Gaussian kernel is selected, and the sample set is fitted by a kernel density estimation algorithm (such as non-parametric kernel regression) to generate a joint probability density function of the three; Obtain the current coupling analysis feature to establish a new real-time risk group , through the Bayes formula: Obtain the posterior probability ; Through the formula: Obtain the risk label type D of the real-time risk group; Different actual feedback forces are formulated based on different risk label types to establish a dynamic calibration baseline; Preferably, the way to establish a dynamic calibration baseline of different actual feedback forces is: When the risk label type D=0, that is, corresponding to high risk, the equation of the dynamic calibration baseline is: Obtain the calibrated actual feedback force ; Wherein, is the theoretical feedback force corresponding to the weight of the hoisted object, is the cumulative difference value between the actual feedback force and the theoretical feedback force; It can be understood that in the high-risk scenario, the feedback deviation is caused by long-term same-direction accumulation L and strong coupling with the hoisted object dynamics, which has formed a vicious cycle of distortion amplification-oscillation aggravation-more serious distortion. The equation of the dynamic calibration baseline is aimed at the error accumulation divergence risk, and the corrects the stock deviation, A matches the distortion amplification characteristics of the inertia and shape of the hoisted object, weights the accumulation trend of continuous same-direction gradient chain, and blocks the loss through strong robust compensation; When the risk label type D=1, that is, corresponding to medium; Through formula one: Obtain the predicted next sliding window bias prediction increment ; Wherein, is the historical average maximum gradient chain length under the same risk label; It can be understood that medium risk is a precursor of high risk, and the deviation has been accumulated but not out of control, and the core is to block the deterioration trend. The formula first predicts the deviation increment: the higher S is, the more closely the distortion-oscillation coupling is, and the more certain the future deviation growth is; Compare the gradient chain length of the current and historical medium-risk to determine if the accumulation is overspeed. Then, perform correction through equation one, take 70% of the stock and increment to intervene, which is conducive to both early containment of deviation deterioration and leaving 30% buffer to avoid system shock, balancing safety and stability; Through equation two: Obtain the calibrated actual feedback force ; When the risk label type D=2, that is, corresponding to low-risk; The equation for dynamically calibrating the baseline is: Obtain the calibrated actual feedback force ; It can be understood that in the low-risk scenario, the feedback distortion is weakly coupled with the dynamic coupling of the suspended object, the deviation accumulation is slow, and the calibration needs to be lightly intervened to preserve the operation experience. In the equation for dynamically calibrating the baseline: Weaken the correction weight of the coupling-related deviation (the lower the synchronization, the weaker the coupling, and the less correction); Inversely reduce the calibration amplitude of heavy load (the greater the inertia, the more need to avoid sudden changes in feedback force to interfere with operation), within the controllable range of risk, priority is given to maintaining the smoothness of the operator's feel.

[0026] Embodiment 2

[0027] As shown in Figure 1 The present application is a kind of crane remote operation control method, further comprising the following steps: S3, based on the dynamic calibration baseline, real-time adjustment of the output force of the operating rod, evaluate the matching of the operator's continuous same direction correction chain length and distortion accumulation after adjustment, if not match then extract the feedback correlation coefficient, based on the feedback correlation coefficient, the cognitive load of the operator is analyzed, the cognitive load degree is obtained, and it is judged whether the cognitive load of the operator is overloaded; Based on the dynamic calibration baseline as the initial adjustment reference of the output force of the operating rod, the output force of the operating rod is real-time adjusted by the servo motor as the force feedback execution mechanism; Obtain the actual feedback force of the operating rod in the monitoring period after real-time adjustment, and the behavior characteristics of the operator; Among them, the behavior characteristics of the operator include: the change rate of the operating force of the operator for two times, the continuous same direction correction chain length of the operator; The fitness of the actual feedback force and the behavior characteristics of the operator is analyzed, and the behavior fitness is obtained; Among them, the fitness analysis method is: S301, extract the continuous same direction correction chain length of the operator, combine the maximum gradient chain length to analyze the chain length correction of the continuous same direction correction chain length, obtain the chain length correction ratio and evaluate the matching of the continuous correction behavior of the operator and the distortion accumulation; Obtain the maximum number of consecutive unidirectional corrections (such as continuously increasing or decreasing the operating force) performed by the operator, and obtain the length of the consecutive unidirectional correction chain; Obtain the maximum gradient chain length of the cumulative distortion sequence, calculate the deviation ratio between the continuous in-direction correction chain length and the maximum gradient chain length, and obtain the chain length correction ratio. The matching between the operator's continuous unidirectional correction behavior and the direction and number of consecutive distortion accumulation characteristics is evaluated based on the chain length correction ratio. Understandably, if the chain length correction ratio is in the range of 0.8-1.2, it indicates that the number of consecutive corrections made in the same direction by the operator is close to the characteristics of the consecutive number of distortion accumulations. At the same time, comparing the direction of the operator's consecutive corrections with the direction of the feedback deviation gradient chain of distortion accumulation, if both are in the same direction (both positive or both negative), it is determined that the operator's consecutive correction behavior in the same direction matches the direction and consecutive number characteristics of distortion accumulation. If the ratio deviates from this range (e.g., much greater than 1 indicates excessively frequent corrections, much less than 1 indicates lagging corrections), or the directions are inconsistent, it is determined to be a mismatch, and further analysis of cognitive fit is needed through feedback correlation coefficient analysis. S302. If there is no match, a correlation analysis is performed based on the operator's behavioral characteristics and the dynamic adjustment characteristics of the actual feedback force to obtain the feedback correlation coefficient. Construct a force change sequence corresponding to the rate of change of force within the monitoring period; Obtain the rate of change of two consecutive actual feedback force adjustments during dynamic baseline adjustment, and construct a feedback correction sequence; Correlation analysis was performed on the feedback change sequence and the feedback correction sequence to obtain the feedback correlation coefficient; Preferably, correlation analysis is performed by calculating the Pearson correlation coefficient between the feedback change sequence and the feedback correction sequence to obtain the feedback correlation coefficient; S303. Based on the feedback correlation coefficient, perform human-machine cognitive collaboration analysis to determine whether the operator's cognitive load is overloaded. Preferably, the standard deviation of the feedback correlation coefficients within N sliding windows is obtained, and i sub-intervals are divided according to the size of the standard deviation, where i ranges from [1, M] and M is the maximum number of sub-intervals. Calculate the probability that the standard deviation of the feedback correlation coefficient falls within different sub-intervals, and obtain... ; Through the entropy equation: Obtain the fluctuation entropy of the feedback correlation coefficient ; Extract the feedback bias gradient within the corrected cumulative distortion sequence and calculate the distortion complexity entropy. ; The preferred method for calculating the distortion complex entropy is as follows: Based on the cumulative distortion sequence, combined with the phase synchronization coefficient S, the modified gradient = feedback deviation gradient is obtained by the formula: ; Wherein, the phase synchronization coefficient reflects the coupling of distortion and the swing of the suspended object, and the distortion amplification coefficient reflects the amplification effect of the suspended object characteristics on the distortion, which describes the influence of distortion from the 'dynamic coupling' and'static characteristics' dimensions respectively; The introduction of S in the modified gradient is to highlight the influence of'real-time coupling state' on the distortion analysis, and A supports the cognitive load calculation together, and the logic of the two is: S focuses on real-time dynamics, and A focuses on the inherent characteristics of the suspended object; It can be understood that the role of the modified gradient is: Role one, by combining the feedback deviation gradient in the cumulative distortion sequence with the phase synchronization coefficient for modification, and then calculating the standard deviation probability based on the modified gradient and substituting it into the entropy equation, the distortion complexity entropy obtained can support the cognitive load calculation together with the fluctuation entropy of the feedback correlation coefficient, providing a key quantitative basis for determining the cognitive overload of the operator; Role two, the modified gradient can better reflect the actual influence of the feedback force distortion, by incorporating the phase synchronization coefficient, it reduces the deviation caused by ignoring the distortion-swing coupling relationship when only using the original feedback deviation gradient analysis, providing distortion data that fits the actual working conditions for subsequent cognitive load analysis, and ensuring the rationality of the cognitive overload determination result; Get the standard deviation of the modified gradient in the N sliding windows, divide i sub-intervals according to the size of the standard deviation of the modified gradient, and the value range of i is [1, M], M is the maximum value of the number of sub-interval division; Calculate the probability of the modified gradient falling into different sub-intervals to obtain ; Get the distortion complexity entropy through the entropy equation: ; Get the cognitive load degree by constructing the cognitive load equation ; As Figure 2 shown, the cognitive load degree is compared with the preset cognitive load threshold value, if the cognitive load degree is lower than the cognitive load threshold value, it is determined that the cognitive load of the operator is in an overload state; If the cognitive load degree is higher than or equal to the preset cognitive load threshold value, the cognitive load degree is continuously monitored; Wherein, the purpose of obtaining the cognitive load degree is; Purpose one, determine whether the cognitive load of the operator is overloaded, by comparing the calculated cognitive load degree with the preset threshold value, if the cognitive load degree is lower than the threshold value, it is determined that the operator is in a cognitive overload state, which provides a judgment basis for subsequent targeted intervention; ​​Objective II. When the cognitive load is determined to be overloaded, the accumulated distortion sequence, dynamic calibration baseline, feedback correlation coefficient and other data in the overload period can be collected to construct an overload scenario dataset, providing data support for subsequent analysis of overload causes; Objective III. Obtaining the cognitive load degree is a prerequisite for generating personalized feedback force mapping relationship. Only after determining the cognitive overload state, can the warning feature time of the overload scenario dataset be extracted, the operator behavior adaptation type be divided, and the relationship adaptation model be constructed to customize the optimal feedback force mapping relationship for different operators.

[0028] S4. If the cognitive load is in an overloaded state, the overload scenario dataset of the accumulated distortion and calibration lag is extracted, the warning time of the scenario dataset is extracted to determine the warning feature time, and the relationship adaptation model is constructed to develop the optimal feedback force mapping relationship for different operators; If the cognitive load is in an overloaded state, the start and end times in the overloaded state are taken as the cognitive overload period of the cognitive overload period; S401. Collect the overload scenario data of the cognitive overload period and construct the overload scenario dataset; Preferably, the accumulated distortion sequence, the start and mutation time stamp of the accumulated distortion of the cognitive overload period are collected, and the accumulated distortion sequence and the start and mutation time stamp are taken as the overload scenario data; The dynamic calibration baseline and the phase synchronization coefficient are taken as the calibration scenario data; The feedback correlation coefficient sequence before the cognitive overload period, the continuous same direction correction chain length, and the force change rate are taken as the pre-overload data; The overload scenario dataset containing the pre-overload data, the overload scenario data, and the calibration scenario data is constructed; S402. Based on the overload scenario dataset, the time chain correlation after the distortion accumulation and calibration is established, and the lag feature vector forming each overload scenario is extracted; Preferably, from the overload scenario data, the start time stamp (the first time the feedback deviation gradient meets the accumulated distortion criterion) and the mutation time stamp (the time when the absolute value of the feedback deviation gradient increases by ≥50%) of the accumulated distortion sequence are located; From the calibration scenario data, the first adjustment time stamp of the dynamic calibration baseline (the first calibration action for the accumulated distortion) is extracted; From the pre-overload data, the determination time stamp of the cognitive overload (the time when Rz first reaches the threshold) is determined; Based on the start time stamp, the mutation time stamp, the first adjustment time stamp, and the determination time stamp, the time chain of distortion start-distortion mutation-calibration adjustment-cognitive overload is constructed; The time difference of each node (such as the lag length of distortion start to calibration adjustment , calibration adjustment to the evolution length of cognitive overload ), forming the lag feature vector of each overload scenario , ]; S403, based on the lag feature vector of each overload scenario, determine the early warning feature time of cognitive overload of different operators; Preferably, the lag feature vectors of all overload scenarios of the operator are counted, and the typical lag mode with the highest frequency is screened out; For example, 70% of the overload of a certain operator corresponds And ; Take the total evolution length of the typical mode ( + ) as the basis, reserve a safety redundancy (such as taking 60% of the total length as the early warning advance), and get the early warning feature time of the operator (such as the total evolution length 5s, the early warning time = 5s×60%=3s); Dynamic updating mechanism: every 3 new overload scenarios, recalculate the typical lag mode and adjust the early warning feature time to ensure the adaptation to the change of operator behavior habits (such as shortening the lag length after being skilled, and reducing the early warning time at the same time); S404, extract the operator behavior adaptation type based on the early warning feature time window; Preferably, the time window from T seconds before the early warning feature time point to the early warning time point (T is the early warning feature time value) is intercepted as the early warning analysis window; Extract the continuous same direction correction chain length and the original sequence of force change rate in the early warning analysis window from the pre-overload data; Calculate the distribution density of the continuous same direction correction chain length and the peak value coefficient (the ratio of the peak value to the average value) of the force change rate in the early warning analysis window; Combined with the time sequence change trend of the feedback correlation coefficient in the early warning analysis window, the operator behavior type is divided: Determine the operator behavior adaptation type based on the distribution density and the peak value coefficient of the force change rate; It can be understood that based on the parameter calculation results in the early warning feature time window, the adaptation type is determined by three threshold values: If the distribution density of the continuous same direction correction chain length is ≥3 times / second (high frequency correction) and the peak value coefficient of the force change rate is >1.5 (strong fluctuation), it is determined that the operator is sensitive, that is, the operator is too sensitive to the feedback force fluctuation, and is easy to correct frequently due to small deviation, and needs to be mapped by the model to reduce the gradient smoothing (such as reducing the feedback force change slope in the low weight section), to reduce the cognitive overload caused by excessive correction; If the distribution density is ≤1 time / second (low frequency correction) and the peak coefficient is ≤1.2 (weak fluctuation), it is determined that the operator is stable, that is, the operator corrects the rhythm slowly, and the key segment needs to be strengthened through model customization (such as amplifying the feedback force difference of the high weight segment) to reduce the correction lag caused by insufficient feedback; If the distribution density is 1-3 times / second (medium frequency correction) and the peak coefficient is 1.2-1.5 (medium fluctuation), it is determined that the operator is adaptive, that is, the operator's behavior characteristics match the existing feedback force mapping, and the existing mapping can meet the cognitive load control requirements; S405, based on the determined operator behavior adaptation type, a relationship adaptation model is constructed, the behavior characteristics of the operator in the overload scene data set are input into the relationship adaptation model, and an optimized feedback force mapping relationship is generated; Preferably, if the operator is sensitive and stable, the actual weight of the hoisted object in the overload scene data set is used as the input feature, and the ideal feedback force of the operator when there is no overload is used as the output label (the ideal feedback force is obtained by back-propagating the operation period of the "continuous same direction correction chain length stable and force change rate ≤10%"). The mean value of the continuous same direction correction chain length and the standard deviation of the force change rate of the operator are included as model adjustment parameters, a gradient boosting regression model with regularization is used to train the relationship adaptation model, and a personalized mapping function is output , W is the actual weight of the hoisted object. The mapping is tested in a non-overload scene, and if the operator's feedback correlation coefficient improves by ≥15% and the matching degree of the continuous same direction correction chain length and the maximum gradient chain length improves by ≥20%, the optimized feedback force mapping relationship is determined. Based on the optimized feedback force mapping relationship, the operator's feedback force mapping relationship is updated. It should be noted that when the operator's feedback force mapping relationship is updated based on the optimized feedback force mapping relationship, the behavior adaptation type (sensitive, stable, or adaptive) corresponding to the operator is determined, the personalized feedback force mapping function generated for this type (such as the gradient smoothing mapping for sensitive type and the key segment strengthening mapping for stable type) is imported into the force feedback control module of the crane remote control system, and the original general feedback force mapping logic is replaced. In subsequent operations, the system will input the actual weight of the hoisted object, calculate and output the actual feedback force of the operating lever in real time through the updated mapping function, and retrain the relationship adaptation model combined with new overload data every time 3 new overload scenes are added to dynamically optimize the mapping function, ensuring that the updated feedback force mapping relationship continuously fits the operator's behavior habits, and achieving the effect of relieving cognitive overload and correcting feedback force distortion.

[0029] Embodiment 3

[0030] As Figure 3As shown, the crane remote operation control system also comprises the following modules: A distortion extraction module is configured to collect a feedback deviation gradient of the crane remote operation, establish a joint distortion criterion based on the feedback deviation gradient, and extract a cumulative distortion sequence; A baseline calibration module is configured to extract a coupling analysis feature of the dynamic parameters of the hoisted object and the cumulative distortion sequence, construct a historical risk group based on the coupling analysis feature, extract a real-time risk label type, formulate a dynamic calibration baseline for different types, and adjust the actual feedback force; An overload identification module is configured to adjust the output force of the operating lever in real time based on the dynamic calibration baseline, evaluate the matching of the length of the continuous same direction correction of the operator after adjustment and the distortion accumulation, extract a feedback correlation coefficient if the matching is not met, perform a cognitive load analysis on the operator based on the feedback correlation coefficient, obtain a cognitive load degree, and determine whether the cognitive load of the operator is overloaded; A feedback adjustment module is configured to extract overload scenario data sets of the distortion accumulation and the calibration lag if the cognitive load is in an overloaded state, extract a warning time to determine a warning feature time, construct a relationship adaptation model, and formulate an optimal feedback force mapping relationship for different operators.

[0031] The above describes one embodiment of the present application in detail, but the content described is only a preferred embodiment of the present application and cannot be considered as limiting the scope of the present application. Any equivalent changes and improvements made in the scope of the present application should still belong to the scope of the present application.

Claims

1. A method for remote operation and control of a crane, characterized in that: Includes the following steps: The feedback deviation gradient of remote operation of crane is collected, and a joint distortion criterion is established based on the feedback deviation gradient and the cumulative distortion sequence is extracted. By extracting the coupling analysis features of the dynamic parameters of the suspended object and the cumulative distortion sequence, historical risk groups are constructed based on the coupling analysis features and real-time risk label types are extracted. Dynamic calibration baselines for actual feedback force are formulated for different types. The output force of the control lever is adjusted in real time based on the dynamic calibration baseline. The matching between the operator's continuous unidirectional correction chain length and the cumulative distortion is evaluated. If there is a mismatch, the operator's cognitive load is analyzed to obtain the cognitive load level and determine whether the operator's cognitive load is overloaded. If the cognitive load is overloaded, the overload scenario dataset with distortion accumulation and calibration lag is extracted. The warning time is extracted from the scenario dataset to determine the warning feature time. A relationship adaptation model is constructed to formulate the preferred feedback force mapping relationship for different operators.

2. The remote operation control method for a crane according to claim 1, characterized in that: The method for establishing the aforementioned joint distortion criterion is as follows: Obtain the average gradient of the sliding window; Instantaneous pulse distortion criteria are established based on feedback deviation gradient and window average gradient to identify and eliminate instantaneous pulse distortion within the sliding window; Based on the removed instantaneous pulse distortion, a cumulative distortion criterion is constructed to identify cumulative distortion within a sliding window; If the feedback bias gradient satisfies the cumulative distortion criterion, construct a cumulative distortion sequence containing the feedback bias gradient that satisfies the cumulative distortion criterion.

3. The remote operation control method for a crane according to claim 1, characterized in that: The method for establishing the dynamic calibration baseline of the actual feedback force is as follows: The sub-features of the historical coupling analysis features are labeled as risk label types, and the historical risk label types and coupling analysis features are used as historical risk groups; Extract the proportion of different risk label types in all historical risk groups as the prior distribution for each risk label type; Extract the joint probability density function of the sub-features contained in the coupling analysis features; Obtain the current coupling analysis features to establish a new real-time risk group, and combine the prior distribution and joint probability density function to obtain the posterior probability through Bayes' formula; Risk label types for real-time risk groups are extracted based on posterior probability, and dynamic calibration baselines with different actual feedback forces are formulated based on different risk label types.

4. A remote operation control method for a crane according to claim 3, characterized in that: The method for obtaining the sub-features of the coupling analysis features is as follows: Obtain the timestamps corresponding to the peak value of the cumulative distortion sequence and the extreme value of the oscillating phase sequence. Based on the nearest neighbor pairing method, match the peak value of the feedback deviation gradient with the extreme value of the oscillating phase sequence whose timestamp is closest to the peak value to obtain the synchronization matching pair. The difference between the timestamps corresponding to the peak and extreme values ​​in the synchronous matching pair is calculated to obtain the swing phase difference; Phase synchronization analysis is performed on the swing phase difference of the synchronized matching pair to obtain the phase synchronization coefficient; Extract the moment of inertia and shape coefficient of the suspended object as dynamic parameters, and obtain the distortion amplification coefficient by constructing a distortion amplification equation; Extract the lengths of consecutive gradient chains in the same direction from the cumulative distortion sequence to obtain the maximum gradient chain length; The maximum gradient chain length, distortion amplification factor, and phase synchronization factor are used as sub-features in the coupled analysis of feedback force and load status of remote crane operation.

5. A remote operation control method for a crane according to claim 1, characterized in that: The cognitive load analysis is performed as follows: Obtain the probability that the standard deviation of the feedback correlation coefficient falls in different sub-intervals, and input the entropy equation to obtain the fluctuation entropy of the feedback correlation coefficient; Extract the feedback bias gradient within the corrected cumulative distortion sequence and calculate the distortion complexity entropy; The degree of perceived load is obtained by constructing a perception load equation based on the fluctuation entropy of distortion complex entropy and feedback correlation coefficient. Based on the comparative analysis of cognitive load, the results of the determination of whether the operator's cognitive load is in an overload state are obtained.

6. A remote operation control method for a crane according to claim 5, characterized in that: The feedback correlation coefficient is obtained as follows: Extract the continuous same-direction correction chain length of the operator, and perform chain length correction analysis on the continuous same-direction correction chain length in combination with the maximum gradient chain length to obtain the chain length correction ratio and evaluate the matching between the operator's continuous correction behavior and distortion accumulation. If there is a mismatch, a correlation analysis is performed based on the operator's behavioral characteristics and the dynamic adjustment characteristics of the actual feedback force to obtain the feedback correlation coefficient.

7. A remote operation control method for a crane according to claim 6, characterized in that: The chain length correction analysis is performed as follows: Obtain the maximum number of consecutive same-direction corrections performed by the operator, and thus obtain the length of the consecutive same-direction correction chain. Obtain the maximum gradient chain length of the cumulative distortion sequence, calculate the deviation ratio between the continuous in-direction correction chain length and the maximum gradient chain length, and obtain the chain length correction ratio. The chain length correction ratio is used to assess the matching between the operator's continuous unidirectional correction behavior and the characteristics of the direction and number of consecutive distortion accumulation.

8. A remote operation control method for a crane according to claim 1, characterized in that: The preferred method for determining the feedback force mapping relationship is as follows: Collect overload scenario data during cognitive overload periods and construct an overload scenario dataset; Based on the overload scenario dataset, a correlation is established between distortion accumulation and the time chain after calibration, and the hysteresis feature vector is extracted to form each overload scenario. Based on the hysteresis feature vector of each overload scenario, the warning feature time of cognitive overload for different operators is determined; Extract operator behavior adaptation types based on early warning feature time windows; Based on the determined operator behavior adaptation type, a relationship adaptation model is constructed. The operator behavior characteristics in the overload scenario dataset are then input into the relationship adaptation model to generate the preferred feedback force mapping relationship.

9. A remote operation control method for a crane according to claim 8, characterized in that: The operator behavior types are categorized as follows: The time window from T seconds before the warning feature time point to the warning time point is extracted as the warning analysis window; Where T is the warning characteristic time value; Extract the original sequences of continuous unidirectional correction chain length and force change rate within the early warning analysis window from the overload pre-data; calculate the distribution density of continuous unidirectional correction chain length and the peak coefficient of force change rate within the early warning analysis window; Based on the temporal variation trend of the correlation coefficient within the early warning analysis window, operator behavior types are classified: the operator behavior adaptation type is determined based on the peak coefficient of distribution density and intensity change rate.

10. A crane remote operation control system, used to implement the crane remote operation control method according to any one of claims 1-9, characterized in that: Includes the following modules: Distortion extraction module: used to collect feedback deviation gradients of remote crane operation, establish joint distortion criteria based on feedback deviation gradients and extract cumulative distortion sequences; Baseline calibration module: By extracting the coupling analysis features of the dynamic parameters of the suspended object and the cumulative distortion sequence, historical risk groups are constructed based on the coupling analysis features, and real-time risk label types are extracted to formulate dynamic calibration baselines for actual feedback forces for different types; Overload identification module: Based on the dynamic calibration baseline, the output force of the operating lever is adjusted in real time. The matching between the operator's continuous unidirectional correction chain length and the cumulative distortion is evaluated after the adjustment. If there is a mismatch, the feedback correlation coefficient of the adjustment is extracted. Based on the feedback correlation coefficient, the cognitive load of the operator is analyzed to obtain the cognitive load degree and determine whether the operator's cognitive load is overloaded. Feedback Adjustment Module: If the cognitive load is in an overload state, the overload scenario dataset with distortion accumulation and calibration lag is extracted, the warning time is extracted from the scenario dataset to determine the warning feature time, and a relationship adaptation model is constructed to formulate the preferred feedback force mapping relationship for different operators.

Images