An interaction control method and system of an intelligent robot based on an internet of things

Abstract

The application relates to the technical field of machine control, in particular to an interactive control method and system of an intelligent robot based on the Internet of Things, which comprises the following steps: in response to a change amount of any environmental parameter being greater than a first threshold value, calculating an adjustment coefficient of each control parameter according to an initialized sensitivity matrix and the change amount of the environmental parameter; taking the control parameter with an adjustment coefficient greater than a second threshold value as an adjustment parameter; taking each adjustment parameter as a particle, determining a target value of each adjustment parameter by using a particle swarm optimization algorithm, and realizing the control of the robot; and calculating the convergence degree of each adjustment parameter in the particle swarm optimization algorithm, and updating the sensitivity matrix based on the convergence degree. Through the technical scheme, the precise control of the intelligent robot can be realized by adjusting multiple control parameters.

Description

Technical Field

[0001] This application relates to the field of machine control technology, and in particular to an interactive control method and system for an intelligent robot based on the Internet of Things. Background Technology

[0002] In the field of industrial internet, intelligent robots are increasingly being used in industrial production processes such as welding, dispensing, grinding, and spraying. In these processes, intelligent robots need to continuously sense changes in the external environment so that they can adjust their control parameters in a timely manner when the external environment changes, thus enabling them to interact with the outside world. Therefore, how intelligent robots can quickly and accurately adjust their control parameters to adapt to the new external environment is a problem that urgently needs to be solved.

[0003] Currently, patent application CN117389416A discloses an interactive control method, device, and robot for an intelligent robot. The method includes the following steps: an environment detection step, in which the robot senses other robots in its vicinity and obtains the distance L between itself and the other robots; an interaction detection step, in which the service robot senses the user's interactive behavior and determines the interaction type corresponding to the interactive behavior. The interaction type includes a daily interaction subclass representing interaction by non-administrator users, an administrator control subclass representing interaction by administrator users, and a non-interaction subclass representing interaction by no personnel. When the interaction type is the non-interaction subclass, the environment detection and this step are repeated; otherwise, the subsequent steps are performed; a mode switching step, in which the interaction type is the daily interaction subclass, the robot enters the daily interaction mode; and in the administrator control mode, the robot enters the administrator control mode; and an interaction response step, in which the robot responds to the interactive behavior. In the daily interaction mode, daily interaction is performed according to daily interaction rules; and in the administrator control mode, administrator management interaction is performed according to administrator management rules.

[0004] The above method adjusts the robot's mode by sensing the environment around it, achieves mode switching, and performs intelligent interaction according to the preset interaction strategy in the corresponding mode. However, in industrial production environments, intelligent robots have multiple control parameters, and simple mode switching cannot achieve precise control of intelligent robots. Summary of the Invention

[0005] To address the technical problem of achieving precise control of intelligent robots, this application provides an interactive control method and system for intelligent robots based on the Internet of Things, which can achieve precise control of intelligent robots by adjusting multiple control parameters.

[0006] In a first aspect, this application provides an interactive control method for an intelligent robot based on the Internet of Things (IoT). The control method includes: responding to a change in any environmental parameter exceeding a first threshold, calculating adjustment coefficients for each control parameter based on an initialized sensitivity matrix and the change in the environmental parameter, wherein the sensitivity matrix includes the global sensitivity of any control parameter to any environmental parameter; using control parameters whose adjustment coefficients exceed a second threshold as adjustment parameters; treating each adjustment parameter as a particle and using a particle swarm optimization algorithm to determine the target value of each adjustment parameter to achieve robot control; and calculating the convergence degree of each adjustment parameter in the particle swarm optimization algorithm and updating the sensitivity matrix based on the convergence degree.

[0007] By pre-screening the adjustment parameters most sensitive to changes in the current environment using a sensitivity matrix, and then performing particle swarm optimization only on these key parameters, the search space dimension and computational complexity of the optimization algorithm are greatly reduced, improving the response speed of the intelligent robot, meeting the needs of industrial real-time control, and the control accuracy can be continuously improved through the sensitivity matrix update mechanism.

[0008] Preferably, the steps for obtaining the initial sensitivity matrix include: collecting historical data showing that the control effect meets preset requirements, wherein the historical data includes various environmental parameters and various control parameters; clustering multiple historical data using the K-means algorithm, and determining the number of clusters using the elbow method; and calculating the control parameters for any cluster pair. Environmental parameters Local sensitivity, calculate the centering control parameters for all cluster pairs. Environmental parameters The mean of local sensitivity is used to obtain the control parameters in the sensitivity matrix. Environmental parameters The global sensitivity; the local sensitivity is the cluster centering control parameter. and environmental parameters The ratio of the changes in the quantity.

[0009] By clustering historical data that meet control requirements using the K-means algorithm, the inherent correlation between environmental parameters and control parameters is extracted from real operational experience. This allows the initialized sensitivity matrix to objectively reflect the degree of influence of any control parameter on any environmental parameter, providing a data foundation for subsequent selection of adjustment parameters.

[0010] Preferably, the cluster pairs Control parameters Environmental parameters Local sensitivity Satisfying the relation: ;in, and Clusters and clusters Control parameters in cluster centers The value of , and Clusters and clusters Environmental parameters in cluster centers The value of .

[0011] Preferably, control parameters adjustment coefficient for:

[0012] ;in, This represents the total number of environmental parameters. Environmental parameters The change It is the sum of the changes in all environmental parameters. Control parameters after normalization Environmental parameters Global sensitivity.

[0013] By using the normalized global sensitivity as a weight, the changes in all environmental parameters are summed in a weighted manner, thereby accurately measuring the contribution of adjusting each control parameter to maintaining the control effect under the current environmental changes.

[0014] Preferably, determining the target values ​​of each adjustment parameter using the particle swarm optimization algorithm includes: initializing multiple particles, standardizing the real-time environmental parameters, particles, and control parameters other than the adjustment parameters, inputting them into the evaluation model, and outputting the control effect; iteratively updating the multiple particles until the change in the global optimal solution is less than a preset change or the number of iterations is greater than a preset number, and then ending the optimization process; and determining the target values ​​of each adjustment parameter based on the global optimal solution.

[0015] The control effect is obtained by inputting real-time environmental parameters, particles (adjustment parameters), and control parameters other than the adjustment parameters into the evaluation model. The control effect is used as the fitness function of the particle swarm optimization algorithm, and the target value of each adjustment parameter is determined with the goal of achieving the optimal control effect.

[0016] Preferably, before determining the target values ​​of each adjustment parameter using the particle swarm optimization algorithm, the control method further includes: using environmental parameters and control parameters from historical operation processes as training samples and labeling the historical operation processes with quality tags; inputting the training samples into the evaluation model to obtain the output results; iteratively updating the evaluation model based on the cross-entropy loss function of the output results and quality tags until the number of updates equals the maximum number of iterations, thus obtaining the trained evaluation model.

[0017] Preferably, the evaluation model is a BP neural network or a logistic regression model.

[0018] Preferably, the convergence of any adjustment parameter is negatively correlated with the variance of the adjustment parameter in each particle when determining the target value of each adjustment parameter.

[0019] The degree of convergence can quantify the influence of the adjustment parameter on the control effect after environmental changes. If an adjustment parameter is densely distributed near the global optimal solution among multiple particles (i.e., the variance is small), it means that any value of the adjustment parameter that deviates from this optimal solution will lead to a sharp decline in the control effect, and the value of the adjustment parameter has a significant impact on the control effect after environmental changes.

[0020] Preferably, updating the sensitivity matrix based on the convergence degree includes: for the global sensitivity of an adjustment parameter to any environmental parameter, in response to the change in the environmental parameter being greater than a first threshold, weighted summing of the convergence degree of the adjustment parameter and the global sensitivity to update the global sensitivity of the adjustment parameter to the environmental parameter.

[0021] By weighting and summing the convergence of the adjustment parameters obtained during the optimization process with the original global sensitivity, the sensitivity matrix is ​​updated online. This allows the sensitivity matrix to dynamically and continuously absorb new experiences, making the intelligent robot's response to environmental changes increasingly accurate.

[0022] In a second aspect, this application also provides an interactive control system for an IoT-based intelligent robot, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an interactive control method for an IoT-based intelligent robot according to the first aspect of this application is implemented.

[0023] The technical solution of this application has the following beneficial technical effects:

[0024] First, by analyzing historical data that meets the preset control effect, a sensitivity matrix is ​​initialized using the K-means clustering algorithm to quantify the influence of each control parameter on each environmental parameter. When a significant change in any environmental parameter is detected, the adjustment coefficient of each control parameter is calculated using the sensitivity matrix and the change in each environmental parameter. Control parameters with adjustment coefficients greater than a second threshold are then selected as adjustment parameters. Subsequently, using only these key adjustment parameters as particles, a particle swarm optimization algorithm is applied, guided by a pre-trained evaluation model, to efficiently find the target values ​​of the adjustment parameters, thereby achieving precise control of the robot.

[0025] Furthermore, the convergence degree of each adjustment parameter during the optimization process is analyzed, and the sensitivity matrix is ​​dynamically updated using this convergence degree, so that the sensitivity matrix has the ability to learn online, and the sensitivity matrix becomes more accurate and efficient in responding to environmental changes with the accumulation of experience. Attached Figure Description

[0026] Figure 1 This is a flowchart of an interactive control method for an IoT-based intelligent robot according to an embodiment of this application.

[0027] Figure 2 This is a schematic diagram visualizing the clustering results according to an embodiment of this application.

[0028] Figure 3 This is a schematic diagram of the loss change curve during the training process of the evaluation model according to an embodiment of this application.

[0029] Figure 4 This is a structural block diagram of an interactive control system for an IoT-based intelligent robot according to an embodiment of this application. Detailed Implementation

[0030] According to the first aspect of this application, this application provides an interactive control method for intelligent robots based on the Internet of Things (IoT), used to realize the interactive control of intelligent robots in industrial production processes. In industrial production processes, intelligent robots are used to complete operations such as welding, dispensing, and spraying on production lines. Taking an intelligent robot used for welding as an example, environmental parameters include at least workshop temperature, material hardness, workshop humidity, and production line transmission speed; control parameters include at least welding voltage, welding current, wire feeding speed, and shielding gas flow rate. Taking an intelligent robot used for spraying as an example, environmental parameters include at least workshop temperature, workshop humidity, workpiece surface roughness, and paint properties; control parameters include at least spraying pressure, spraying flow rate, paint preheating temperature, and nozzle opening. Thus, the environmental parameters and control parameters of the intelligent robot can be determined according to the application scenario of the intelligent robot.

[0031] Figure 1 This is a flowchart illustrating an interactive control method for an IoT-based intelligent robot according to an embodiment of this application. Figure 1 As shown, the interactive control method for the IoT-based intelligent robot includes steps S101 to S104, which are described in detail below.

[0032] S101, in response to any change in an environmental parameter exceeding a first threshold, calculate the adjustment coefficient of each control parameter based on the initialized sensitivity matrix and the change in the environmental parameter. The sensitivity matrix includes the global sensitivity of any control parameter to any environmental parameter.

[0033] In one embodiment, when the change in any environmental parameter exceeds a first threshold, indicating a change in the environment, the control parameters of the intelligent robot need to be adjusted to ensure optimal control performance. The change in the environmental parameter is the absolute value of the difference between the environmental parameter at the current moment and the previous moment, and the first threshold is 1.

[0034] Understandably, the control effect refers to the work quality of the intelligent robot. When the intelligent robot is used for welding, the control effect is the welding quality; when the intelligent robot is used for spraying, the control effect is the spraying quality.

[0035] When the environment changes, adjusting all control parameters would not only be computationally intensive but would also reduce the real-time performance of the intelligent robot. Therefore, it is necessary to calculate the adjustment coefficients of each control parameter. These adjustment coefficients can reflect the effectiveness of the control parameters in maintaining the control effect and ensuring that the control effect is not reduced after the environment changes.

[0036] Before calculating the adjustment coefficients of each control parameter, it is necessary to obtain an initial sensitivity matrix, which includes the global sensitivity of each control parameter to each environmental parameter. The steps for obtaining the initial sensitivity matrix include: collecting historical data showing that the control effect meets preset requirements; the historical data includes each environmental parameter and each control parameter; clustering multiple historical data points using the K-means algorithm, and determining the number of clusters using the elbow method, with each cluster corresponding to a cluster center; calculating the control parameters for any cluster pair. Environmental parameters Local sensitivity, calculate the centering control parameters for all cluster pairs. Environmental parameters The mean of local sensitivity is used to obtain the control parameters in the sensitivity matrix. Environmental parameters The global sensitivity; the local sensitivity is the cluster centering control parameter. and environmental parameters The ratio of the changes in the quantity.

[0037] Please see Figure 2 This is a schematic diagram visualizing the clustering results according to an embodiment of this application; from Figure 2 It can be seen that the number of clusters determined by the elbow method is 5. By using the PCA algorithm to obtain two principal component directions (principal component 1 and principal component 2), and mapping all historical data and the cluster centers of each cluster to the two-dimensional coordinates formed by principal component 1 and principal component 2, a schematic diagram of the clustering results can be obtained.

[0038] Specifically, the cluster pairs Control parameters Environmental parameters Local sensitivity Satisfying the relation: ;in, and Clusters and clusters Control parameters in cluster centers The value of , and Clusters and clusters Environmental parameters in cluster centers The value of .

[0039] Understandably, the cluster centers of each cluster are the environmental parameters and control parameters required to achieve the preset control effect. Within a cluster pair, the environmental parameters... When changes occur, in order to ensure that the control effect meets the preset requirements, the control parameters... The control parameters will also change accordingly. Changes and environmental parameters The ratio of changes can reflect the control parameters Environmental parameters Local sensitivity.

[0040] Control parameters Environmental parameters global sensitivity for:

[0041] ;

[0042] in, For cluster pairs Control parameters Environmental parameters Local sensitivity, The set consisting of all cluster pairs. Let be the number of cluster pairs. Understandably, the number of clusters is denoted as . ,but A cluster pair.

[0043] Thus, the sensitivity matrix is ​​initialized based on historical data, thereby quantifying the global sensitivity of any control parameter to any environmental parameter.

[0044] In one embodiment, before calculating the adjustment coefficients of each control parameter, the global sensitivity needs to be normalized. This normalization can be either row normalization or global normalization. The row normalization process involves: adjusting the control parameters... Environmental parameters Global sensitivity divided by control parameters The sum of global sensitivities to various environmental parameters; the steps for global normalization are: to control parameters Environmental parameters The global sensitivity is divided by the sum of the global sensitivities of each control parameter to each environmental parameter.

[0045] Control parameters adjustment coefficient for:

[0046] ;

[0047] in, This represents the total number of environmental parameters. Environmental parameters The change It is the sum of the changes in all environmental parameters. Control parameters after normalization Environmental parameters Global sensitivity.

[0048] Thus, the adjustment coefficient The larger the value, the more the control parameter needs to be adjusted after environmental changes. It can effectively guarantee the control effect.

[0049] S102, the control parameter with an adjustment coefficient greater than the second threshold is used as the adjustment parameter.

[0050] In one embodiment, the second threshold is 0.5. A control parameter with an adjustment coefficient greater than the second threshold is selected as the adjustment parameter. Subsequently, only the adjustment parameter needs to be adjusted, realizing the transformation from blind global optimization to precise local optimization. The computing resources are concentrated on the most critical adjustment parameter, which shortens the response time and meets the needs of industrial real-time control.

[0051] S103: Treat each adjustment parameter as a particle and use the particle swarm optimization algorithm to determine the target value of each adjustment parameter to achieve robot control.

[0052] In one embodiment, the adjustment parameters are treated as particles, and the dimension of the particles is smaller than the total number of control parameters. Compared with the prior art, which treats all control parameters of the intelligent robot as particles, the particle dimension can be effectively reduced, thus reducing the computational load of the particle swarm optimization algorithm.

[0053] The process of determining the target values ​​of each adjustment parameter using the particle swarm optimization algorithm includes: initializing multiple particles, standardizing the real-time environmental parameters, particles, and control parameters other than the adjustment parameters, inputting them into the evaluation model, and outputting the control effect; iteratively updating the multiple particles until the change in the global optimal solution is less than a preset change or the number of iterations is greater than a preset number, at which point the optimization process ends, and the target values ​​of each adjustment parameter are determined based on the global optimal solution.

[0054] The number of particles is set to 30, the preset number of times is 100, and the preset change is 1.

[0055] Understandably, the control effect is the fitness function in the particle swarm optimization algorithm. The input to the evaluation model is the standardized real-time environmental parameters, particles, and control parameters other than the adjustment parameters, i.e., the values ​​of each environmental parameter and each control parameter. The output is the control effect, i.e., the operational quality of the intelligent robot. The accuracy of the evaluation model's output of the control effect directly affects the accuracy of the target value. Therefore, before using the particle swarm optimization algorithm to determine the target values ​​of each adjustment parameter, the control method further includes: using the environmental parameters and control parameters from historical operations as training samples and labeling them with quality tags; inputting the training samples into the evaluation model to obtain the output result; iteratively updating the evaluation model based on the cross-entropy loss function of the output result and the quality tags until the number of updates equals the maximum number of iterations, thus obtaining the trained evaluation model. The maximum number of iterations is 100. Please refer to [link to relevant documentation]. Figure 3 This is a schematic diagram of the loss change curve during the training process of the evaluation model according to the embodiments of this application. The value of the cross-entropy loss continuously decreases during the training process.

[0056] The quality label is manually labeled. When the intelligent robot's work quality meets the standard, the quality label is 1, and when the intelligent robot's work quality does not meet the standard, the quality label is 0. The closer the output result of the evaluation model is to 1, the better the work quality of the intelligent robot. The evaluation model is a BP neural network or a logistic regression model, which is not limited in this application.

[0057] In this way, the target values ​​of each adjustment parameter are determined by using the particle swarm optimization algorithm, and the values ​​of the adjustment parameters are adjusted to the corresponding target values, so that the adjusted control parameters can adapt to changes in the environment and realize interactive control of the robot.

[0058] S104 calculates the convergence degree of each adjustment parameter in the particle swarm optimization algorithm and updates the sensitivity matrix based on the convergence degree.

[0059] In one embodiment, during the optimization process of the particle swarm optimization algorithm, multiple particles search for the optimal solution in the solution space (a multidimensional space composed of all adjustment parameters). If an adjustment parameter among the multiple particles is densely distributed near the global optimal solution, it indicates that after repeated attempts, all particles have arrived at a unified optimal solution. Any value deviating from this optimal solution will lead to a sharp decline in control performance, and the value of this adjustment parameter has a significant impact on the control performance after environmental changes. Conversely, if the distribution of this adjustment parameter among the multiple particles is relatively dispersed, it means that within a certain range, many values ​​of this adjustment parameter can achieve similar control results. That is, although the value of this adjustment parameter changes, the improvement in control performance is not significant, meaning that after environmental changes, the adjustment parameter has a small impact on the control performance.

[0060] Specifically, the convergence degree of any adjustable parameter is negatively correlated with the variance of the adjustable parameter in each particle when determining the target value of each adjustable parameter, and the convergence degree can quantify the influence of the adjustable parameter on the control effect after environmental changes. (Using adjustable parameters...) As an example, adjust the parameters degree of convergence for:

[0061] ;

[0062] in, The total number of particles, When determining the target values ​​of each adjustment parameter, the particles Adjusting parameters The value of ; To determine the target values ​​for each adjustment parameter, all particles are at the adjustment parameters. The mean of the above; To adjust parameters The search range is equal to the number of particles with respect to the adjustment parameters when the target values ​​of each adjustment parameter are determined. The difference between the maximum and minimum values ​​is used to achieve normalization.

[0063] Understandably, The value ranges from 0 to 1, and when the value approaches 1, it indicates an adjustment parameter. It has a large convergence, i.e., adjusting the parameters Densely distributed near the global optimum, adjusting parameters The effectiveness of control measures after environmental changes has a significant impact.

[0064] Furthermore, the sensitivity matrix is ​​updated based on the convergence degree of each adjustment parameter in the particle swarm optimization algorithm, so that when the environment changes again, the global sensitivity of any control parameter to any environmental parameter can be accurately quantified, thereby improving the accuracy of the next interactive control.

[0065] Specifically, updating the sensitivity matrix based on the convergence degree includes: for the global sensitivity of an adjustment parameter to any environmental parameter, in response to the change in the environmental parameter being greater than a first threshold, weighted summing of the convergence degree of the adjustment parameter and the global sensitivity to update the global sensitivity of the adjustment parameter to the environmental parameter.

[0066] Wherein, if the change in the environmental parameter is not greater than the first threshold, it means that the environmental parameter has not changed; here, only the global sensitivity corresponding to the adjustment parameter and the changed environmental parameter is updated.

[0067] Responding to environmental parameters If the change is greater than the first threshold, the adjustment parameter in the sensitivity matrix... The global sensitivity of environmental parameters is updated, and the update process is as follows:

[0068] ;

[0069] in, Adjust parameters before updating Environmental parameters global sensitivity, To adjust parameters The degree of convergence, For learning rate, Adjusting parameters after the update Environmental parameters The global sensitivity. The learning rate controls the update speed of the global sensitivity; in this embodiment, the learning rate... The value is 0.1.

[0070] This endows the sensitivity matrix with the ability to learn and evolve, enabling it to be continuously updated through online learning, thus making the intelligent robot's response to environmental changes increasingly accurate and efficient.

[0071] According to a second aspect of this application, this application also provides an interactive control system for an intelligent robot based on the Internet of Things. Figure 4 This is a structural block diagram of an interactive control system for an IoT-based intelligent robot according to an embodiment of this application. Figure 4 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an interactive control method for an IoT-based intelligent robot according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configurations and functions are known in the art and will not be described further here.

[0072] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. An interactive control method for an intelligent robot based on the Internet of Things, characterized in that, The control method includes: in response to any change in an environmental parameter exceeding a first threshold, calculating the adjustment coefficient of each control parameter based on an initialized sensitivity matrix and the change in the environmental parameter, wherein the sensitivity matrix includes the global sensitivity of any control parameter to any environmental parameter; the step of obtaining the initialized sensitivity matrix includes: Historical data showing that the control effect meets preset requirements is collected. This historical data includes various environmental parameters and control parameters. The K-means algorithm is used to cluster multiple historical data sets, and the elbow method is used to determine the number of clusters. Control parameters are calculated for any pair of clusters. Environmental parameters Local sensitivity, calculate the centering control parameters for all cluster pairs. Environmental parameters The mean of local sensitivity is used to obtain the control parameters in the sensitivity matrix. Environmental parameters Global sensitivity; local sensitivity is the cluster centering control parameter. and environmental parameters The ratio of the changes in the quantity; Cluster pairs Control parameters Environmental parameters Local sensitivity satisfy: ； and Clusters and clusters Control parameters in cluster centers The value of , and Clusters and clusters Environmental parameters in cluster centers The possible values ​​of ; The control parameter with an adjustment coefficient greater than the second threshold is used as the adjustment parameter; control parameter adjustment coefficient for: ; This represents the total number of environmental parameters. Environmental parameters The change It is the sum of the changes in all environmental parameters. Control parameters after normalization Environmental parameters Global sensitivity; Each adjustment parameter is treated as a particle, and the target value of each adjustment parameter is determined by the particle swarm optimization algorithm to achieve robot control; the convergence degree of each adjustment parameter in the particle swarm optimization algorithm is calculated, and the sensitivity matrix is ​​updated based on the convergence degree.

2. The interactive control method for an intelligent robot based on the Internet of Things according to claim 1, characterized in that, The target values ​​for each adjustment parameter are determined using the particle swarm optimization algorithm, including: Multiple particles are initialized, and the control parameters other than the real-time environmental parameters, particles, and adjustment parameters are standardized and then input into the evaluation model to output the control effect. The multiple particles are iteratively updated until the change in the global optimal solution is less than the preset change or the number of iterations is greater than the preset number, at which point the optimization process ends, and the target value of each adjustment parameter is determined based on the global optimal solution.

3. The interactive control method for an intelligent robot based on the Internet of Things according to claim 2, characterized in that, Before using the particle swarm optimization algorithm to determine the target values ​​of each adjustment parameter, the control method further includes: using environmental parameters and control parameters from historical operation processes as training samples, and labeling the historical operation processes with quality tags; The training samples are input into the evaluation model to obtain the output results. The evaluation model is iteratively updated based on the output results and the cross-entropy loss function of the quality labels until the number of updates equals the maximum number of iterations, at which point the trained evaluation model is obtained.

4. The interactive control method for an intelligent robot based on the Internet of Things according to claim 3, characterized in that, The evaluation model is a BP neural network or a logistic regression model.

5. The interactive control method for an intelligent robot based on the Internet of Things according to claim 1, characterized in that, The convergence of any adjustable parameter is negatively correlated with the variance of the adjustable parameter in each particle when determining the target value of each adjustable parameter.

6. The interactive control method for an intelligent robot based on the Internet of Things according to claim 1, characterized in that, Updating the sensitivity matrix based on the convergence degree includes: For the global sensitivity of an adjustment parameter to any environmental parameter, in response to the change in the environmental parameter being greater than a first threshold, the convergence degree of the adjustment parameter and the global sensitivity are weighted and summed to update the global sensitivity of the adjustment parameter to the environmental parameter.

7. An interactive control system for an intelligent robot based on the Internet of Things, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement an interactive control method for an Internet of Things-based intelligent robot according to any one of claims 1 to 6.

Images