A human-computer interaction control method, device and equipment applied to a robot

Abstract

The application provides a human-robot interaction control method, device and equipment applied to a robot. The method provided by the application comprises the following steps: collecting an interaction force time domain signal and a pose signal of the end of the robot, and obtaining an interaction force frequency domain signal corresponding to the interaction force time domain signal; determining a first evaluation index; extracting a first time domain feature of the interaction force time domain signal, and determining a second evaluation index; extracting a second time domain feature of the pose signal, and determining a third evaluation index; inputting the interaction force time domain signal into a trajectory prediction model, and outputting an expected interaction trajectory; determining a fourth evaluation index; fusing the evaluation indexes to obtain a composite evaluation index; when the composite evaluation index exceeds a stability threshold, updating a mobility controller parameter; determining a control amount of the robot at a next moment according to the updated mobility controller parameter, and controlling the robot to perform corresponding movement. The method provided by the application is used to ensure that the robot can reliably and stably interact.

Description

Technical Field

[0001] This application relates to the field of robot control technology, and in particular to a human-computer interaction control method, device and equipment for robots. Background Technology

[0002] In the field of human-computer interaction, with the continuous advancement of robotics technology and the improvement of intelligence, achieving smooth and stable interaction between robots and humans has become a research hotspot. The efficiency and stability of human-computer interaction are directly related to the performance and application effect of the system. Therefore, exploring an efficient and stable human-computer interaction control method is particularly important.

[0003] Currently, existing human-machine interaction control methods mainly rely on a strategy combining force feedback and position control. This method collects the interaction force and pose signals from the robot's end effector to control the robot in real time, enabling interaction with humans. Force feedback technology can sense the force applied by the human to the robot, thereby adjusting the robot's trajectory and force, while position control technology ensures that the robot can follow a predetermined trajectory during interaction. However, this existing technology still has some obvious shortcomings in practical applications. When facing complex and dynamic interaction environments, the position control strategy may not be able to adjust the robot's motion state in a timely and accurate manner, leading to delays or instability in the interaction process. Existing methods often lack effective differentiation and evaluation of force and pose signals under different interaction states, which limits the robot's adaptability and compliance in different interaction scenarios. Therefore, there is an urgent need for a new human-machine interaction control method that can improve the robot's control accuracy and stability, and ensure the robot's reliability in different interaction scenarios. Summary of the Invention

[0004] In view of this, this application provides a human-computer interaction control method, apparatus and device for robots, so as to ensure that robots can interact reliably and stably.

[0005] Specifically, this application is implemented through the following technical solution:

[0006] The first aspect of this application provides a human-computer interaction control method for robots, the method comprising:

[0007] Collect the time-domain signal and pose signal of the interaction force at the end of the robot, and obtain the frequency-domain signal of the interaction force corresponding to the time-domain signal of the interaction force;

[0008] A first evaluation index is determined for evaluating the stability of the interactive force frequency domain signal; the first evaluation index is used to evaluate the energy characteristic difference between the unstable and stable parts of the interactive force frequency domain signal.

[0009] Extract a first time-domain feature from the interaction force time-domain signal, and determine a second evaluation index based on the first time-domain feature to evaluate the fluctuation degree of the interaction force time-domain signal; the second evaluation index is used to evaluate the fluctuation of the interaction force time-domain signal in the time dimension.

[0010] Extract the second time-domain features of the pose signal, and determine a third evaluation index based on the second time-domain features to evaluate the stability of the pose signal changes;

[0011] The interaction force time-domain signal is input into a pre-constructed trajectory prediction model, and the trajectory prediction model outputs the desired interaction trajectory based on the interaction force time-domain signal.

[0012] A fourth evaluation index is determined based on the expected interaction trajectory and the actual interaction trajectory to evaluate the degree of consistency between the actual interaction trajectory and the expected interaction trajectory; the actual interaction trajectory is determined based on the pose signal at each time point.

[0013] The first evaluation index, the second evaluation index, the third evaluation index, and the fourth evaluation index are integrated to obtain a composite evaluation index.

[0014] When the composite evaluation index exceeds the stability threshold, the admittance controller parameters of the robot are updated;

[0015] The control quantity for the robot at the next moment is determined based on the updated admittance controller parameters, and the robot is controlled to perform corresponding movements according to the control quantity to achieve smooth and stable human-machine interaction.

[0016] A second aspect of this application provides a human-computer interaction control device for robots, the device comprising an acquisition module, a determination module, a fusion module, and a processing module; wherein...

[0017] The acquisition module is used to collect the interaction force time-domain signal and pose signal of the robot end effector, and to acquire the interaction force frequency-domain signal corresponding to the interaction force time-domain signal.

[0018] The determining module is used to determine a first evaluation index for evaluating the stability of the interactive force frequency domain signal; the first evaluation index is used to evaluate the energy characteristic difference between the unstable part and the stable part in the interactive force frequency domain signal.

[0019] The determining module is further configured to extract a first time-domain feature of the interaction force time-domain signal, and determine a second evaluation index for evaluating the fluctuation degree of the interaction force time-domain signal based on the first time-domain feature; the second evaluation index is used to evaluate the fluctuation of the interaction force time-domain signal in the time dimension.

[0020] The determining module is further configured to extract a second time-domain feature of the pose signal and determine a third evaluation index for evaluating the stability of the pose signal change based on the second time-domain feature.

[0021] The determining module is further configured to input the interaction force time-domain signal into a pre-constructed trajectory prediction model, and the trajectory prediction model outputs the desired interaction trajectory based on the interaction force time-domain signal.

[0022] The determining module is further configured to determine a fourth evaluation index for evaluating the consistency between the actual interaction trajectory and the expected interaction trajectory based on the expected interaction trajectory and the actual interaction trajectory; the actual interaction trajectory is determined based on the pose signal at each time point.

[0023] The fusion module is used to fuse the first evaluation index, the second evaluation index, the third evaluation index, and the fourth evaluation index to obtain a composite evaluation index.

[0024] The processing module is used to update the admittance controller parameters of the robot when the composite evaluation index exceeds the stable threshold.

[0025] The processing module is also used to determine the control quantity of the robot at the next moment based on the updated admittance controller parameters, and control the robot to perform corresponding movements according to the control quantity, so as to achieve smooth and stable human-machine interaction.

[0026] A third aspect of this application provides a human-computer interaction control device for robots, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of any of the methods provided in the first aspect of this application.

[0027] The human-machine interaction control method, apparatus, and equipment for robots provided in this application first acquire the interaction force time-domain signal and pose signal of the robot's end effector, and obtain the interaction force frequency-domain signal corresponding to the interaction force time-domain signal; then, determine a first evaluation index for evaluating the stability of the interaction force frequency-domain signal; the first evaluation index is used to evaluate the energy characteristic difference between the unstable and stable parts of the interaction force frequency-domain signal; then, extract the first time-domain feature of the interaction force time-domain signal, and determine a second evaluation index for evaluating the fluctuation degree of the interaction force time-domain signal based on the first time-domain feature; the second evaluation index is used to evaluate the fluctuation of the interaction force time-domain signal in the time dimension; then, extract the second time-domain feature of the pose signal, and determine a third evaluation index for evaluating the stability of pose signal changes based on the second time-domain feature. The evaluation criteria are as follows: First, the first, second, third, and fourth evaluation criteria are fused to obtain a composite evaluation criterion. Then, when the composite evaluation criterion exceeds a stable threshold, the robot's admittance controller parameters are updated. Finally, the control quantity for the robot at the next moment is determined based on the updated admittance controller parameters, and the robot is controlled to perform corresponding movements according to the control quantity to achieve compliant and stable human-machine interaction.In this way, firstly, by collecting the time-domain and pose signals of the interaction force at the robot's end effector, key information in the human-machine interaction process can be comprehensively obtained, and the corresponding frequency-domain signal of the interaction force can be acquired. The frequency-domain signal can reveal the frequency components in the force signal, which helps to identify stable and unstable interaction force components. Secondly, by determining the first evaluation index for assessing the stability of the frequency-domain signal of the interaction force, the energy characteristic difference between the unstable and stable components in the frequency-domain signal of the interaction force can be quantified, and a second evaluation index for assessing the fluctuation degree of the time-domain signal of the interaction force can be determined accordingly. This reflects the fluctuation of the interaction force in the time dimension and provides an important reference for evaluating the smoothness of the interaction process. Thirdly, by inputting the time-domain signal of the interaction force into a pre-built trajectory prediction model, the model can output based on the interaction force signal. The expected interaction trajectory enables accurate prediction of the robot's motion trajectory. Simultaneously, by comparing the expected and actual interaction trajectories, a fourth evaluation index is determined to assess the consistency between the actual and expected trajectories. This quantifies the deviation between the actual and expected trajectories, providing a direct basis for evaluating control effectiveness. Fourthly, the first, second, third, and fourth evaluation indices are integrated to obtain a composite evaluation index, which comprehensively reflects the stability and smoothness of the human-machine interaction process. Thus, through precise signal acquisition and processing, comprehensive evaluation index construction, accurate trajectory prediction and consistency evaluation, intelligent controller parameter updates, and smooth and stable human-machine interaction, the stability and smoothness of human-machine interaction are significantly improved, providing strong support for the widespread application of robotics technology. Attached Figure Description

[0028] Figure 1 A flowchart of an embodiment of the human-computer interaction control method for robots provided in this application;

[0029] Figure 2 This is a hardware structure diagram of a human-computer interaction control device for robots, which is part of the human-computer interaction control equipment for robots applied in this application.

[0030] Figure 3 This is a schematic diagram of the structure of a human-computer interaction control device for robots provided in this application. Detailed Implementation

[0031] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0032] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0033] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0034] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0035] Figure 1 This is a flowchart of an embodiment of the human-machine interaction control method for robots provided in this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0036] S101. Collect the time-domain signal and pose signal of the interaction force at the end of the robot, and obtain the frequency-domain signal of the interaction force corresponding to the time-domain signal of the interaction force.

[0037] Specifically, robots can be robots that interact closely with humans. These robots can possess a high degree of flexibility and adaptability to work collaboratively with human users in different interaction scenarios. In specific implementations, for example, in one embodiment, robots can include collaborative robots, service robots, or medical robots, etc.

[0038] Furthermore, the robot end effector refers to the foremost part of the robot that performs interactive actions. When a robot collaborates with a human, the end effector directly interacts with the human. In specific implementations, for example, in one embodiment, the robot end effector can be a gripper, a claw, or a tool end effector with sensors: a collaborative robot uses a claw to hand a screwdriver to a worker; the claw is the robot end effector.

[0039] Furthermore, the time-domain signal represents a signal that changes over time. For example, in one embodiment, the time-domain signal can be obtained by determining the curve of the force at the robot's end effector fluctuating over time based on the values ​​of the interaction force at consecutive time points.

[0040] Furthermore, the pose signal contains the robot's position and orientation, which can represent the robot's end effector's trajectory in three-dimensional space.

[0041] Furthermore, the time-domain signal can be converted into a frequency-domain signal through Fourier transform in order to observe whether there are certain periodic oscillations or abnormal frequency components in the signal and to identify unstable phenomena.

[0042] In practice, this step involves using sensors to read the contact force data of the robot end effector when it collaborates with a human, obtaining the interaction force time domain signal, and then using the robot's position encoder or external vision tracking system to obtain the pose signal of the robot end effector. The interaction force time domain signal is then converted into the interaction force frequency domain signal through Fourier transform.

[0043] S102. Determine a first evaluation index for evaluating the stability of the interactive force frequency domain signal; the first evaluation index is used to evaluate the energy characteristic difference between the unstable part and the stable part in the interactive force frequency domain signal.

[0044] Specifically, the primary evaluation index is a numerical indicator used to quantitatively assess the difference between the stable and unstable components of a frequency domain signal. In practice, high-frequency energy proportion, dominant frequency shift, high-frequency peak value, and high-frequency energy standard deviation can all be used as the primary evaluation index.

[0045] Furthermore, the energy characteristic difference between the unstable and stable components can be characterized by a spectrum diagram. Components less than 5 Hz are identified as stable low-frequency components, while components greater than 20 Hz are identified as unstable high-frequency components. For example, in one embodiment, 95% of the energy in a force signal spectrum is concentrated between 0 and 5 Hz, which can be identified as a stable component, while 30% of the energy in another force signal spectrum is concentrated between 20 and 50 Hz, which can be identified as an unstable component. The energy characteristic difference between these two components is evaluated using a first evaluation index.

[0046] The following is a specific example to illustrate the process of determining the first evaluation index:

[0047] (1) Obtain the energy spectral density of the frequency domain signal of the interaction force.

[0048] Specifically, energy spectral density represents the energy intensity distribution of the interactive force frequency domain signal at different frequencies, and the commonly used unit is N² / Hz.

[0049] In practice, a fast Fourier transform can be performed on the time-domain signal of the interaction force, and the power spectral density at each frequency point can be calculated based on the following formula:

[0050] PSD(f) = |F(f)| 2

[0051] Where PSD(f) is the power spectral density and F(f) is the time-domain signal of the interaction force.

[0052] (2) Divide the energy spectral density into multiple sub-regions and obtain the frequency domain characteristics of each sub-region.

[0053] In specific implementation, for example, in one embodiment, the entire frequency domain (e.g., 0–50Hz) can be divided into multiple frequency bands of equal width or specific intervals. Sub-region 1: 0–5Hz (stable operating range) has an average energy of 0.82 N² / Hz, a peak value of 1.0, and a narrow bandwidth; Sub-region 2: 5–15Hz (transition band) has an average energy of 0.15 N² / Hz, a peak value of 0.4, and a moderate bandwidth; Sub-region 3: 15–30Hz (potentially unstable region) has an average energy of 0.35 N² / Hz, a peak value of 1.2, and a wide bandwidth; Sub-region 4: 30–50Hz (unstable region).

[0054] (3) For each sub-region, determine the category of the sub-region based on its frequency domain characteristics; the category of the sub-region includes stable and unstable.

[0055] In practice, frequency band characteristics can be analyzed to determine whether the frequency band represents a stable interaction or oscillation state. For example, in one embodiment, if the peak value of the high-frequency band is greater than a preset threshold, it is determined to be an unstable sub-region; if the average value of the low-frequency band is greater than a preset threshold, it is determined to be a stable sub-region.

[0056] (4) Determine the first evaluation index based on the frequency domain characteristics of the unstable sub-region and the frequency domain characteristics of the stable sub-region.

[0057] In practice, the frequency domain characteristics of each sub-region include the main frequency, bandwidth, and spectral peak value; the comprehensive numerical index calculated based on the frequency domain characteristics of the stable and unstable regions quantifies the stability state of the interactive system.

[0058] The first evaluation index can be determined based on the frequency domain characteristics of both the unstable and stable sub-regions using the following method:

[0059] Step 1: For each sub-region, determine the fusion characteristics of that sub-region based on its dominant frequency, bandwidth, and spectral peak.

[0060] Specifically, each sub-region can be traversed, the main frequency, bandwidth and spectral peak of each sub-region can be calculated, and they can be fused into a fused feature through linear fusion or nonlinear fusion.

[0061] In practice, the fusion features can be obtained using the following formula:

[0062] ;

[0063] In this context, Fi represents the fusion features, where α, β, and γ are preset weights, and the sum of α, β, and γ is 1. BW is the normalized maximum value of the dominant frequency of the sub-region, and BW is the normalized maximum value of the bandwidth of the sub-region. This represents the normalized maximum value of the spectral peak value in the sub-region.

[0064] Step 2: Calculate the first weighted average of the fusion characteristics of all unstable sub-regions and the second weighted average of the fusion characteristics of all stable sub-regions.

[0065] In practice, the first weighted average can be calculated based on the following formula:

[0066] ;

[0067] Where μ1 is the first weighted average, U is the number of unstable sub-regions, ωi is the preset weight of the i-th unstable sub-region, Fi is the fusion feature of the i-th unstable sub-region, i∈[1,U] and i is a positive integer.

[0068] ;

[0069] Where μ2 is the second weighted average, M is the number of stable sub-regions, ωj is the preset weight of the j-th stable sub-region, Fj is the fusion feature of the j-th stable sub-region, j∈[1,M] and j is a positive integer.

[0070] Step 3: Determine the first evaluation index based on the first weighted average and the second weighted average.

[0071] In practice, the difference between the first weighted average and the second weighted average can be determined as the first evaluation index, or the ratio of the first weighted average and the second weighted average can be determined as the first evaluation index. For example, in one embodiment, the first evaluation index can be calculated based on the following formula:

[0072] P1 = ;

[0073] Wherein, P1 is the first evaluation index, μ1 is the first weighted average, and μ2 is the second weighted average.

[0074] Understandably, the interaction force signal acquired by the robot's end effector is first subjected to a Fourier transform to obtain its energy spectral density in the frequency domain, thereby revealing the energy distribution of the signal in various frequency ranges. Then, the entire frequency domain is divided into multiple sub-regions, and frequency domain features, including the main frequency, bandwidth, and spectral peak, are extracted for each sub-region to comprehensively reflect the dynamic characteristics of the signal in that frequency band. Then, by setting a threshold for the frequency domain features or adopting classification rules, each sub-region is divided into a stable region or an unstable region, which significantly enhances the ability to identify unstable behaviors such as system oscillations and abnormal fluctuations. Finally, by combining the frequency domain features of the stable and unstable regions, a first evaluation index reflecting the difference between the two is constructed to achieve a quantitative evaluation of the stability state of the interaction force signal. This allows the system to maintain compliance while effectively suppressing high-frequency oscillations or interference, improving the safety, naturalness, and overall robustness of the human-computer interaction process and the control system.

[0075] S103. Extract the first time-domain feature of the interaction force time-domain signal, and determine a second evaluation index for evaluating the fluctuation degree of the interaction force time-domain signal based on the first time-domain feature; the second evaluation index is used to evaluate the fluctuation of the interaction force time-domain signal in the time dimension.

[0076] Specifically, the mean square value of the interaction force time-domain signal at each time point can be selected as the first time-domain feature. It should be noted that the first time-domain feature can be determined by the mean square value of variance, peak-to-peak value, or sliding window standard deviation, etc.

[0077] Furthermore, the degree of fluctuation of the interaction force signal on the time axis reflects whether the interaction force changes drastically in a short period of time, which can determine whether the system is in a stable state.

[0078] Furthermore, the second evaluation index is an index calculated based on the first time-domain characteristics, used to measure the degree of fluctuation of the interactive force signal on the time axis. For example, in one embodiment, the second evaluation index can be calculated by the ratio of the first time-domain characteristics to the maximum force value.

[0079] It should be noted that the larger the value of the second evaluation index, the more violent the force signal fluctuations and the more unstable the trend; the smaller the value of the second evaluation index, the smoother the interaction process and the better the system status.

[0080] Furthermore, the first time-domain feature includes the mean square error of the interaction force signal within the current preset time window; a specific embodiment is given below to describe in detail the process of determining the second evaluation index:

[0081] (1) Obtain the mean square error values ​​of the interaction force signal within multiple preset time windows before the current preset time window.

[0082] Specifically, the exact length of the preset time window is determined according to actual needs, and this embodiment does not limit it. For example, in one embodiment, the time window can be defined as 0.5 seconds.

[0083] Furthermore, the mean square error value within the current preset time window represents the statistical value of the energy magnitude of the interaction force signal within the current time window, and is used to describe its volatility.

[0084] In specific implementation, for example, in one embodiment, when the signals within the current preset time window are 2, 3, and 4, the corresponding mean square error is . That is .

[0085] (2) Find the largest mean square error value among the plurality of mean square error values.

[0086] In practice, all mean squared errors can be compared to obtain the maximum mean squared error value. For example, in one embodiment, the maximum mean squared error value can be determined based on the following formula:

[0087] RMSmax=max(RMS1, RMS2,..., RMS M );

[0088] Where RMSmax is the maximum mean squared error, RMS M This represents the mean squared error value corresponding to the Mth preset time window.

[0089] (3) The normalization process of the plurality of mean squared differences is performed using the maximum mean squared value to obtain the normalization result, and the normalization result is determined as the second evaluation index.

[0090] In practice, the mean square value of the latest time window is obtained, and the ratio of this value to the maximum mean square difference is determined as the second evaluation index. For example, in one embodiment, the mean square value of the current window is 3.2, and the maximum mean square value is 4.2. After normalization, the second evaluation index is determined to be 3.2 / 4.2 = 0.76.

[0091] Furthermore, after calculating the second evaluation index, when the value of the second index is greater than 1, the current preset time window fluctuates violently; when the value of the second index approaches 1, it is determined that the current preset time window is close to the historical maximum value, and therefore the fluctuation is strong; when the value of the current second evaluation index approaches 0, it is determined that the interaction force of the current preset time window is relatively stable, and the fluctuation is small and relatively stable.

[0092] Understandably, the mean square error of the interaction force signal within the current preset time window is first extracted as the first time-domain feature to describe the signal energy and fluctuation level within that window. Then, several historical time windows are traced back, the mean square error of each window is calculated, and the historical maximum value is extracted as a reference benchmark. Finally, the mean square error of the current window is normalized to this historical maximum value to obtain the second evaluation index. In this way, the second evaluation index can reflect the proportion of the current fluctuation relative to the historical fluctuation peak, effectively identifying sudden and violent fluctuation signals in the interaction process. This provides a fast and accurate time-domain stability judgment basis for the adaptive parameter adjustment of the admittance controller, improving the robot's stable control performance and the naturalness and safety of human-machine collaboration in complex interactive environments.

[0093] S104. Extract the second time-domain feature of the pose signal, and determine a third evaluation index for evaluating the stability of the pose signal change based on the second time-domain feature.

[0094] Specifically, the second time-domain feature is a feature extracted from the pose signal to measure its stability over time.

[0095] In practice, the specific data types included in the second time-domain feature are determined according to actual needs, and this embodiment does not limit this. For example, in one embodiment, the second time-domain feature may be the speed standard deviation / acceleration mean, local trajectory offset, jitter amplitude, etc., determined by speed, acceleration, jitter amplitude, trajectory deviation, etc.

[0096] Furthermore, the second time-domain feature is standardized or normalized to form a measurable third evaluation index, which is used to quantify whether the pose signal change is stable.

[0097] The following is a specific example to illustrate the calculation process of the third evaluation index:

[0098] (1) Extract the velocity, acceleration and jitter amplitude of the pose signal within a preset time window.

[0099] In practice, taking a preset time window length of 0.1 seconds as an example, sampling is performed every 0.01 seconds to obtain 100 sets of pose data for the end position.

[0100] Furthermore, the speed within a preset time window can be calculated based on the following formula:

[0101] ;

[0102] Among them, v i It is the speed within the i-th preset time window, x i It is the i-th end position, x i-1It is the (i-1)th end position, and Δt is the sampling interval.

[0103] Furthermore, the acceleration within a preset time window can be calculated based on the following formula:

[0104] ;

[0105] Among them, a i It is the acceleration within the i-th preset time window, v i It is the speed within the i-th preset time window, v i-1 It is the velocity within the (i-1)th preset time window, and Δt is the sampling interval.

[0106] Furthermore, the jitter amplitude within the preset time window can be calculated based on the following formula:

[0107] ;

[0108] Where jitter is the jitter amplitude within the preset time window, max(x(t)) is the maximum position within the preset time window, and min(x(t)) is the minimum position within the preset time window.

[0109] Furthermore, all pose signals can be high-pass filtered to calculate the maximum amplitude of the residual signal, which is then determined as the jitter amplitude.

[0110] (2) The speed, acceleration and jitter amplitude are weighted and fused to obtain the third evaluation index.

[0111] In practice, the three features can be standardized. For example, in one embodiment, Z-score or maximum-minimum normalization can be used to process speed, acceleration, and jitter amplitude.

[0112] Furthermore, fixed weights can be preset for speed, acceleration, and jitter amplitude. By multiplying the corresponding weights by these three features, the corresponding third evaluation index can be obtained.

[0113] It should be noted that by extracting the velocity, acceleration, and jitter amplitude of the pose signal within a preset time window and weighting and fusing them into a third evaluation index, the system can comprehensively and in real time perceive the motion stability of the robot's end effector. The third evaluation index integrates dynamic features at different levels: velocity reflects the overall trajectory trend, acceleration captures abrupt changes, and jitter amplitude detects minor disturbances, thereby establishing a dynamic stability evaluation mechanism that is both sensitive and robust. This mechanism can not only be used to identify whether the robot is undergoing uncontrolled motion, but also serve as a basis for adaptive adjustment of admittance controller parameters, effectively improving the safety, compliance, and naturalness of human-robot collaboration.

[0114] S105. The interaction force time domain signal is input into the pre-constructed trajectory prediction model, and the trajectory prediction model outputs the desired interaction trajectory based on the interaction force time domain signal.

[0115] Specifically, the pre-built trajectory prediction model can be a model trained using historical data or machine learning methods, used to infer the trajectory of a person's intention based on force signals.

[0116] Furthermore, the expected interaction trajectory is the trajectory output by the trajectory prediction model, used to guide the robot's actions, representing the system's inference of the user's intention.

[0117] In practice, multiple interactive task sample datasets can be collected. The datasets can include force signals and corresponding human operation trajectories. A trajectory prediction model of force and trajectory is established using machine learning methods. Then, the current interactive force time-domain signal is input into the trained model, and the model outputs a future expected interactive trajectory, which is the predicted robot motion plan.

[0118] Understandably, by inputting the time-domain signal of the interactive force into a pre-trained trajectory prediction model, and outputting the expected interactive trajectory reflecting human intentions in real time, a bridge is established from force perception to behavioral reasoning. This enables the robot to accurately identify and respond to human operational intentions. Compared with traditional methods that only provide feedback based on the current force value, this method can integrate temporal context information, possessing predictability and foresight. It can effectively improve the robot's adaptability to natural interactions. The expected interactive trajectory can not only serve as the target signal for trajectory tracking, but also be used to determine whether the current behavior is reasonable, thereby dynamically adjusting the control gain and ensuring the safety, compliance, and intelligence of human-robot collaboration.

[0119] S106. Determine a fourth evaluation index for evaluating the consistency between the actual interaction trajectory and the expected interaction trajectory based on the expected interaction trajectory and the actual interaction trajectory; the actual interaction trajectory is determined based on the pose signal at each time point.

[0120] Specifically, the actual interaction trajectory is the real trajectory of the robot's end effector during actual interaction, which can be obtained by splicing pose signals.

[0121] Furthermore, the fourth evaluation metric is used to quantify the similarity between the actual trajectory and the expected trajectory. In practice, one or more of the following can be used as the fourth evaluation metric: average Euclidean distance, trajectory similarity, and path offset rate.

[0122] The following is a specific implementation example to illustrate in detail the process of obtaining the fourth evaluation index:

[0123] A fourth evaluation index is determined based on the expected interaction trajectory and the actual interaction trajectory to assess the degree of consistency between the actual interaction trajectory and the expected interaction trajectory, including:

[0124] (1) Time-align the expected interaction trajectory and the actual interaction trajectory to obtain the processed expected interaction trajectory and the processed actual interaction trajectory.

[0125] In practice, the actual interaction trajectory and the expected interaction trajectory can be resampled to have the same time resolution and length.

[0126] Furthermore, methods such as linear interpolation and time standardization can be used to ensure that the expected interaction trajectory and the actual interaction trajectory can correspond one-to-one.

[0127] (2) For each time point, calculate the Euclidean distance between the processing expected trajectory point corresponding to the processing expected interaction trajectory and the processing actual trajectory point corresponding to the processing actual interaction trajectory, obtain the position error at that time point, and determine the average position error based on the position errors at all time points.

[0128] In practice, according to the order of the time points, the positions corresponding to the expected interaction trajectory and the actual interaction trajectory are determined at each time point, thus obtaining the expected trajectory point and the actual trajectory point.

[0129] Furthermore, the Euclidean distance between the desired trajectory point and the actual trajectory point can be calculated based on the following formula:

[0130] ;

[0131] Where di is the Euclidean distance at the i-th time point. The actual trajectory point corresponding to the i-th time point; Let be the expected trajectory point corresponding to the i-th time point.

[0132] Furthermore, the Euclidean distances calculated at all time points are accumulated, and the average value is calculated to obtain the average position error.

[0133] (3) The optimal matching distance between the processing expected interaction trajectory and the processing time interaction trajectory in time sequence is calculated using the DTW algorithm.

[0134] Specifically, the DTW algorithm is used to calculate the optimal matching path and distance between two trajectory sequences on the time axis.

[0135] In practice, the DTW algorithm is used to process the desired interaction trajectory and the time interaction trajectory to obtain the optimal matching distance.

[0136] (4) For each time point, calculate the angular deviation between the expected trajectory point and the actual trajectory point in the tangent direction to obtain the angular deviation at that time point, and determine the average angular deviation based on the angular deviation at all time points.

[0137] Specifically, the local tangent vector for each trajectory point can be calculated based on the following formula:

[0138] ;

[0139] in, It is the local tangent vector of the i-th trajectory point. It is the position of the (i+1)th trajectory point. It is the position of the i-th trajectory point.

[0140] Furthermore, the angle deviation can be calculated based on the following formula:

[0141] ;

[0142] in, Let be the angle deviation of the i-th trajectory point. Let i be the local tangent vector for the i-th actual trajectory point. Let be the local tangent vector for the i-th point of the desired trajectory.

[0143] (5) Normalize the average position error, the optimal matching distance and the average angle deviation respectively to obtain the normalized average position error, the normalized optimal matching distance and the normalized average angle deviation.

[0144] In practice, the average angle deviation can be normalized based on the following formula:

[0145] ;

[0146] in, The normalized average angle deviation is θ, where θ is the average angle deviation being processed. It is the minimum value among the average angular deviations. This represents the maximum value in the average angular deviation.

[0147] Furthermore, the same calculation formula can be used to normalize the average position error and the optimal matching distance to obtain the corresponding normalized average position error and normalized optimal matching distance.

[0148] (6) Determine the fourth evaluation index based on the normalized average position error, the normalized optimal matching distance, and the normalized average angle deviation.

[0149] Specifically, the normalized average position error, normalized optimal matching distance, and normalized average angle deviation can be fused together to obtain the fourth evaluation index; alternatively, the normalized average position error, normalized optimal matching distance, and normalized average angle deviation can be multiplied by preset weights respectively, and the three weighted data can be fused together to obtain the fourth evaluation index.

[0150] Understandably, the expected trajectory and the actual trajectory are first time-aligned to ensure comparability. Then, at each time point, the Euclidean distance of the spatial position and the angular deviation of the trajectory tangent direction are calculated, and the DTW algorithm is introduced to calculate the optimal matching distance of the global temporal structure, evaluating the consistency characteristics of the trajectory from multiple perspectives. Based on this, the three types of error indices are normalized to eliminate dimensional differences and improve the fairness and robustness of the fusion. Finally, the normalized average position error, optimal matching distance, and average angular deviation are weighted and fused to obtain a fourth evaluation index with comprehensive judgment capabilities. This not only comprehensively reflects the degree of deviation between the actual trajectory and the human intention trajectory in terms of spatial offset, temporal rhythm, and directional control, but also serves as an important reference input for the admittance controller or interactive intent discrimination module, significantly improving the naturalness, sensitivity, and adaptability of the human-computer interaction system.

[0151] S107. The first evaluation index, the second evaluation index, the third evaluation index and the fourth evaluation index are integrated to obtain a composite evaluation index.

[0152] Specifically, the first, second, third, and fourth evaluation indicators can be normalized to have the same dimensions.

[0153] Furthermore, based on the weights pre-set for the first, second, third, and fourth evaluation indicators, the four indicators can be weighted and merged to obtain a composite evaluation indicator.

[0154] S108. When the composite evaluation index exceeds the stable threshold, update the admittance controller parameters of the robot.

[0155] Specifically, the composite evaluation index is compared with the stability threshold. If the composite evaluation index does not exceed the stability threshold, the robot system is considered stable and no adjustment is needed. If the composite evaluation index exceeds the stability index, the robot's parameters need to be adjusted, and the robot's admittance control parameters are updated.

[0156] The following is a specific example to illustrate the process of determining the stability threshold:

[0157] (1) For each of the preset multiple interaction states, collect multiple sets of data of the robot end in the interaction state; each set of data includes the interaction force time domain signal and pose signal of the robot end; the multiple interaction states include stable interaction states and unstable interaction states.

[0158] Specifically, the interaction state is a state that describes the overall stable behavioral pattern of the human-computer interaction process.

[0159] In practice, the interaction state can include a stable interaction state and an unstable interaction state. For example, in one embodiment, when the human and robot cooperate well and there are no sudden vibrations, the interaction state can be determined to be a stable interaction state. For example, in another embodiment, when vibrations or severe vibrations occur during the interaction between the human and robot, or when there is a deviation from the original interaction intention, the interaction state can be determined to be an unstable interaction state.

[0160] Furthermore, under a defined interaction state, the robot collects the interaction force signal and pose signal based on the sensors carried on it, and determines the interaction force time domain signal through the interaction force signal.

[0161] (2) For each set of data in each interactive state, determine the composite evaluation index of that set of data in that interactive state.

[0162] In practice, for each interaction state, four evaluation indicators are determined based on the interaction force time domain signal and pose signal, using the method of determining the first evaluation indicator, the second evaluation indicator, the third evaluation indicator and the fourth evaluation indicator. After normalization and weighting, they are fused to obtain the evaluation indicators that meet the requirements.

[0163] (3) According to the type of interaction state, the determined composite evaluation indexes are divided into the first dataset corresponding to the stable interaction state and the second dataset corresponding to the unstable interaction state.

[0164] Specifically, the first dataset is a set of composite evaluation indicators obtained under all stable interaction states; the second dataset is a set of composite evaluation indicators obtained under all unstable interaction states. The first and second datasets are determined based on the interaction state type from which the composite evaluation indicators are obtained.

[0165] (4) Statistically analyze the distribution of the first dataset and the second dataset to obtain the first distribution characteristics of the first dataset and the second distribution characteristics of the second dataset.

[0166] In practice, the distribution of the first and second datasets can be determined by judging the mean and variance of the first and second datasets, or by using kernel density estimation and probabilistic modeling. For example, in one embodiment, the stable interaction composite index is concentrated in [0.1, 0.3], and the unstable interaction is concentrated in [0.4, 0.9].

[0167] (5) Determine a stability threshold for distinguishing between stable and unstable interaction states based on the first distribution feature and the second distribution feature; wherein the stability threshold is used to determine whether the current interaction is in a stable interaction state based on a composite evaluation index calculated in real time.

[0168] Specifically, the stability threshold is a critical value used to classify composite indicators into stable and unstable states. For example, if the composite evaluation indicators calculated in real time do not exceed the stability threshold, the current state can be defined as a stable interaction state; otherwise, it is an unstable interaction state.

[0169] Furthermore, the first and second distribution features can be processed using the mean plus or minus the standard deviation, the minimum cross error point, the optimal classification point of the ROC curve, or the equal error method (to minimize the probability of misclassification of the two classes) to obtain a stable threshold.

[0170] It should be noted that the calculation methods for the stability threshold of unstable interaction states include, but are not limited to, the mean, standard deviation, confidence interval, KS test, or kernel density estimation. For example, in one embodiment, based on the distribution differences of the datasets, a stability threshold that distinguishes between stable and unstable interaction states can be determined. The stability threshold can be the intersection of the upper limit of the confidence interval of the first dataset, the lower limit of the unstable dataset, or the index value corresponding to the maximum inter-class discrimination.

[0171] Understandably, by collecting multiple sets of interaction force and pose data of the robot's end effector under stable and unstable interaction states, calculating composite evaluation indicators for each set of samples, and combining statistical analysis to obtain the distribution characteristics of the two states, a stability threshold is finally determined, thus realizing the construction of a real-time discrimination mechanism for the robot's current interaction state. This process has good interpretability and adaptability. When facing complex and changing human-robot interaction environments, it can autonomously identify whether it is in a stable state, thereby driving the admittance controller to make timely and effective response adjustments, significantly enhancing the robustness, safety, and intelligence of the human-robot collaborative system.

[0172] The following is a specific example to illustrate in detail the process of updating the admittance controller parameters of a robot:

[0173] (1) Construct an objective function based on the composite evaluation index corresponding to the robot; wherein the objective function is a weighted composite of multiple indexes, and the objective function is used to evaluate the control stability and control compliance of the admittance controller.

[0174] Specifically, composite evaluation indicators may include force frequency domain characteristics, force time domain fluctuations, pose stability, and intention. Figure 1 Four evaluation indicators for consistency.

[0175] In practical implementation, force frequency domain characteristics, force time domain fluctuations, pose stability, and intention can be considered. Figure 1 The four evaluation indicators of consistency are weighted and then fused to form the objective function.

[0176] (2) Define the energy tank model of the admittance controller, and determine the parameter adjustment boundary in combination with the performance limitations of the robot; wherein, the energy tank model is used to provide feedback on the correlation between the energy input to the robot and the energy consumed by the robot.

[0177] Specifically, the energy tank model is used to constrain the system's ability to absorb / release energy, preventing instability caused by excessive parameter adjustments.

[0178] In practice, the energy tank model corresponding to the robot can be determined based on traditional energy tank model construction methods. For example, in one embodiment, it is determined that adjusting parameters such as increasing damping and increasing inertia will consume energy.

[0179] Furthermore, reasonable ranges are set for each admittance parameter. In specific implementation, the virtual mass is determined to be 0.5 to 3.0, the damping to be 10 to 100, and the stiffness to be 0 to 500.

[0180] Furthermore, by combining the energy tank model and performance limitations, the adjustable boundaries of each parameter are determined.

[0181] (3) Using the parameter adjustment boundary as a constraint, solve the objective function to obtain the updated admittance controller parameters.

[0182] Specifically, under the above objective function and constraints, a parameter optimization problem is established, and the optimization problem is solved to obtain the new admittance controller parameters.

[0183] In practice, a multivariate equation containing all parameters can be constructed and solved using quadratic programming to obtain a new set of admittance control parameters.

[0184] Understandably, the process begins by constructing a multi-index weighted objective function based on real-time calculated composite evaluation metrics, reflecting the performance of control under the current interactive state. Then, by combining the energy tank model of the admittance controller with the performance limitations of the robot itself, reasonable parameter adjustment boundaries are set to ensure that the parameter adjustment process meets energy constraints and physical safety requirements. Finally, the objective function is solved under boundary constraints to obtain a set of optimal admittance parameters, which are used to update the controller in real time. This not only enables intelligent adjustment of the control response according to different interactive states but also effectively avoids problems such as excessive rigidity or unstable oscillations, significantly improving the human-computer interaction system's adaptability, control robustness, and naturalness of interaction in complex task scenarios.

[0185] S109. Determine the control quantity of the robot at the next moment based on the updated admittance controller parameters, and control the robot to perform corresponding movements according to the control quantity to achieve smooth and stable human-machine interaction.

[0186] In practice, the updated admittance control parameters may include virtual mass, virtual damping, and virtual stiffness, and the control quantity for the next time step is determined based on these parameters. For example, in one embodiment, the control quantity for the next time step can be calculated based on the following formula:

[0187] ;

[0188] Where M is virtual mass, D is virtual damping, and K is virtual stiffness. and As an intermediate quantity, and Here are the parameters of the admittance controller, where Δt is the time interval from the current moment to the next moment, and x(t+1) is the position at the next moment. The control quantity can be determined based on the position at the next moment.

[0189] Furthermore, combining the above example, the controller sends the calculated x(t+1) to the robot actuator, which controls the joint position according to the instruction to realize the robot's motion response and achieve compliant and stable cooperation.

[0190] The human-machine interaction control method for robots provided in this embodiment has the following aspects: First, by collecting the interaction force time-domain signal and pose signal from the robot's end effector, key information in the human-machine interaction process can be comprehensively acquired, and the interaction force frequency-domain signal corresponding to the interaction force time-domain signal can be obtained. The frequency-domain signal can reveal the frequency components in the force signal, which helps to identify stable and unstable interaction force components. Second, by determining a first evaluation index for evaluating the stability of the interaction force frequency-domain signal, the energy characteristic difference between the unstable and stable components in the interaction force frequency-domain signal can be quantified, and a second evaluation index for evaluating the fluctuation degree of the interaction force time-domain signal can be determined accordingly. This reflects the fluctuation of the interaction force in the time dimension and provides an important reference for assessing the smoothness of the interaction process. Third, by inputting the interaction force time-domain signal into a pre-built trajectory prediction model, the model can... Based on the interaction force signal, the desired interaction trajectory is output, enabling accurate prediction of the robot's motion trajectory. Simultaneously, by comparing the desired and actual interaction trajectories, a fourth evaluation index is determined to assess the consistency between the actual and desired trajectories. This quantifies the deviation between the actual and desired trajectories, providing a direct basis for evaluating control effectiveness. Fourthly, the first, second, third, and fourth evaluation indices are integrated to obtain a composite evaluation index, which comprehensively reflects the stability and smoothness of the human-machine interaction process. Thus, through precise signal acquisition and processing, comprehensive evaluation index construction, accurate trajectory prediction and consistency evaluation, intelligent controller parameter updates, and smooth and stable human-machine interaction, the stability and smoothness of human-machine interaction are significantly improved, providing strong support for the widespread application of robotics technology.

[0191] Corresponding to the aforementioned embodiment of a human-computer interaction control method for robots, this application also provides an embodiment of a human-computer interaction control device for robots.

[0192] This application discloses an embodiment of a human-machine interaction control device for robots, which can be applied to human-machine interaction control devices for robots. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of the human-machine interaction control device used in the robot loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 2 The diagram shown is a hardware structure diagram of a human-machine interaction control device for robots, as described in this application. Except for... Figure 2In addition to the processor, memory, network interface, and non-volatile memory shown, the human-machine interaction control device for robots in the embodiments may also include other hardware depending on the actual function of the human-machine interaction control device for robots, which will not be described in detail here.

[0193] Figure 3 This is a schematic diagram of the structure of a human-machine interaction control device for robots provided in this application, according to Embodiment 1. Please refer to... Figure 3 The apparatus provided in this embodiment includes an acquisition module 310, a determination module 320, a fusion module 330, and a processing module 340; wherein,

[0194] The acquisition module 310 is used to collect the interaction force time-domain signal and pose signal of the robot end effector, and to acquire the interaction force frequency-domain signal corresponding to the interaction force time-domain signal.

[0195] The determining module 320 is used to determine a first evaluation index for evaluating the stability of the interactive force frequency domain signal; the first evaluation index is used to evaluate the energy characteristic difference between the unstable part and the stable part in the interactive force frequency domain signal.

[0196] The determining module 320 is further configured to extract a first time-domain feature of the interaction force time-domain signal, and determine a second evaluation index for evaluating the fluctuation degree of the interaction force time-domain signal based on the first time-domain feature; the second evaluation index is used to evaluate the fluctuation of the interaction force time-domain signal in the time dimension.

[0197] The determining module 320 is further configured to extract a second time-domain feature of the pose signal and determine a third evaluation index for evaluating the stability of the pose signal change based on the second time-domain feature.

[0198] The determining module 320 is further configured to input the interaction force time domain signal into a pre-constructed trajectory prediction model, and the trajectory prediction model outputs the desired interaction trajectory based on the interaction force time domain signal;

[0199] The determining module 320 is further configured to determine a fourth evaluation index for evaluating the consistency between the actual interaction trajectory and the expected interaction trajectory based on the expected interaction trajectory and the actual interaction trajectory; the actual interaction trajectory is determined based on the pose signal at each time point.

[0200] The fusion module 330 is used to fuse the first evaluation index, the second evaluation index, the third evaluation index and the fourth evaluation index to obtain a composite evaluation index;

[0201] The processing module 340 is used to update the admittance controller parameters of the robot when the composite evaluation index exceeds the stability threshold.

[0202] The processing module 340 is further configured to determine the control quantity of the robot at the next moment based on the updated admittance controller parameters, and control the robot to perform corresponding movements according to the control quantity, so as to achieve smooth and stable human-machine interaction.

[0203] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0204] Please continue to refer to Figure 2 This application also provides a human-computer interaction control device for robots, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of any of the methods provided in the first aspect of this application.

[0205] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods provided in this application.

[0206] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0207] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0208] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A human-computer interaction control method for robots, characterized in that, The human-computer interaction control method for robots is applied to robots in human-computer interaction systems; the method includes: Collect the time-domain signal and pose signal of the interaction force at the end of the robot, and obtain the frequency-domain signal of the interaction force corresponding to the time-domain signal of the interaction force; A first evaluation index is determined for evaluating the stability of the interactive force frequency domain signal; the first evaluation index is used to evaluate the energy characteristic difference between the unstable and stable parts of the interactive force frequency domain signal. Extract a first time-domain feature from the interaction force time-domain signal, and determine a second evaluation index based on the first time-domain feature to evaluate the fluctuation degree of the interaction force time-domain signal; the second evaluation index is used to evaluate the fluctuation of the interaction force time-domain signal in the time dimension. Extract the second time-domain features of the pose signal, and determine a third evaluation index based on the second time-domain features to evaluate the stability of the pose signal changes; The interaction force time-domain signal is input into a pre-constructed trajectory prediction model, and the trajectory prediction model outputs the desired interaction trajectory based on the interaction force time-domain signal. A fourth evaluation index is determined based on the expected interaction trajectory and the actual interaction trajectory to evaluate the degree of consistency between the actual interaction trajectory and the expected interaction trajectory; the actual interaction trajectory is determined based on the pose signal at each time point. The first evaluation index, the second evaluation index, the third evaluation index, and the fourth evaluation index are integrated to obtain a composite evaluation index. When the composite evaluation index exceeds the stability threshold, the admittance controller parameters of the robot are updated; The control quantity for the robot at the next moment is determined based on the updated admittance controller parameters, and the robot is controlled to perform corresponding movements according to the control quantity to achieve smooth and stable human-machine interaction.

2. The method according to claim 1, characterized in that, The method for determining the stability threshold includes: For each of the multiple preset interaction states, the robot end effector collects multiple sets of data in that interaction state; each set of data includes the interaction force time-domain signal and pose signal of the robot end effector; the multiple interaction states include stable interaction states and unstable interaction states; For each set of data in each interaction state, determine the composite evaluation index for that set of data in that interaction state; Based on the type of interaction state, the determined composite evaluation indicators are divided into the first dataset corresponding to stable interaction states and the second dataset corresponding to unstable interaction states. The distribution of the first dataset and the second dataset is statistically analyzed to obtain the first distribution characteristic of the first dataset and the second distribution characteristic of the second dataset. A stability threshold is determined based on the first distribution feature and the second distribution feature to distinguish between stable and unstable interaction states; wherein, the stability threshold is used to determine whether the current interaction is in a stable interaction state based on a composite evaluation index calculated in real time.

3. The method according to claim 1, characterized in that, The determination of the first evaluation index for evaluating the stability of the interactive force frequency domain signal includes: Obtain the energy spectral density of the frequency domain signal of the interactive force; The energy spectral density is divided into multiple sub-regions, and the frequency domain characteristics of each sub-region are obtained; For each sub-region, the category of the sub-region is determined based on its frequency domain characteristics; the category of the sub-region includes stable and unstable. The first evaluation index is determined based on the frequency domain characteristics of the unstable sub-region and the frequency domain characteristics of the stable sub-region.

4. The method according to claim 1, characterized in that, The step of extracting the second time-domain features of the pose signal and determining a third evaluation index for evaluating the stability of pose signal changes based on the second time-domain features includes: Extract the velocity, acceleration, and jitter amplitude of the pose signal within a preset time window; The speed, acceleration, and jitter amplitude are weighted and fused to obtain the third evaluation index.

5. The method according to claim 1, characterized in that, The first time-domain feature includes the mean square error of the interaction force signal within the current preset time window; the step of determining a second evaluation index for evaluating the fluctuation degree of the interaction force time-domain signal based on the first time-domain feature includes: Obtain multiple mean squared errors of the interaction force signal within multiple preset time windows prior to the current preset time window; Find the largest mean squared error value from the plurality of mean squared error values; The normalization process of the plurality of mean squared errors is performed using the maximum mean squared value to obtain a normalized result, and the normalized result is determined as the second evaluation index.

6. The method according to claim 1, characterized in that, A fourth evaluation index is determined based on the expected interaction trajectory and the actual interaction trajectory to assess the degree of consistency between the actual interaction trajectory and the expected interaction trajectory, including: The expected interaction trajectory and the actual interaction trajectory are time-aligned to obtain the processed expected interaction trajectory and the processed actual interaction trajectory. For each time point, calculate the Euclidean distance between the processing expected trajectory point corresponding to the processing expected interaction trajectory and the processing actual trajectory point corresponding to the processing actual interaction trajectory to obtain the position error at that time point, and determine the average position error based on the position errors at all time points. The DTW algorithm is used to calculate the optimal temporal matching distance between the expected interaction trajectory and the time interaction trajectory. For each time point, the angular deviation between the expected trajectory point and the actual trajectory point in the tangential direction is calculated to obtain the angular deviation at that time point, and the average angular deviation is determined based on the angular deviations at all time points. The average position error, the optimal matching distance, and the average angle deviation are normalized respectively to obtain the normalized average position error, the normalized optimal matching distance, and the normalized average angle deviation. The fourth evaluation index is determined based on the normalized average position error, the normalized optimal matching distance, and the normalized average angle deviation.

7. The method according to claim 1, characterized in that, The updating of the robot's admittance controller parameters includes: An objective function is constructed based on the composite evaluation index corresponding to the robot; wherein, the objective function is a weighted composite of multiple indexes, and the objective function is used to evaluate the control stability and control compliance of the admittance controller; Define an energy tank model for the admittance controller, and determine the parameter adjustment boundaries based on the robot's performance limitations; wherein, the energy tank model is used to provide feedback on the correlation between the energy input to the robot and the energy consumed by the robot; Using the parameter adjustment boundary as a constraint, the objective function is solved to obtain the updated admittance controller parameters.

8. The method according to claim 3, characterized in that, The frequency domain characteristics of each sub-region include the dominant frequency, bandwidth, and spectral peak value; the determination of the first evaluation index based on the frequency domain characteristics of unstable and stable sub-regions includes: For each sub-region, the fusion characteristics of that sub-region are determined based on its dominant frequency, bandwidth, and spectral peak. Calculate the first weighted average of the fusion characteristics of all unstable sub-regions and the second weighted average of the fusion characteristics of all stable sub-regions; The first evaluation index is determined based on the first weighted average and the second weighted average.

9. A human-computer interaction control device for robots, characterized in that, The device includes an acquisition module, a determination module, a fusion module, and a processing module; wherein... The acquisition module is used to collect the interaction force time-domain signal and pose signal of the robot end effector, and to acquire the interaction force frequency-domain signal corresponding to the interaction force time-domain signal. The determining module is used to determine a first evaluation index for evaluating the stability of the interactive force frequency domain signal; the first evaluation index is used to evaluate the energy characteristic difference between the unstable part and the stable part in the interactive force frequency domain signal. The determining module is further configured to extract a first time-domain feature of the interaction force time-domain signal, and determine a second evaluation index for evaluating the fluctuation degree of the interaction force time-domain signal based on the first time-domain feature; the second evaluation index is used to evaluate the fluctuation of the interaction force time-domain signal in the time dimension. The determining module is further configured to extract a second time-domain feature of the pose signal and determine a third evaluation index for evaluating the stability of the pose signal change based on the second time-domain feature. The determining module is further configured to input the interaction force time-domain signal into a pre-constructed trajectory prediction model, and the trajectory prediction model outputs the desired interaction trajectory based on the interaction force time-domain signal. The determining module is further configured to determine a fourth evaluation index for evaluating the consistency between the actual interaction trajectory and the expected interaction trajectory based on the expected interaction trajectory and the actual interaction trajectory; the actual interaction trajectory is determined based on the pose signal at each time point. The fusion module is used to fuse the first evaluation index, the second evaluation index, the third evaluation index, and the fourth evaluation index to obtain a composite evaluation index. The processing module is used to update the admittance controller parameters of the robot when the composite evaluation index exceeds the stable threshold. The processing module is also used to determine the control quantity of the robot at the next moment based on the updated admittance controller parameters, and control the robot to perform corresponding movements according to the control quantity, so as to achieve smooth and stable human-machine interaction.

10. A human-computer interaction control device for robots, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the program, implements the steps of the method according to any one of claims 1-8.

Images