Industrial robot control system and method based on deep learning

Abstract

The invention discloses an industrial robot control method based on deep learning, and relates to the technical field of industrial robot control. Comprising the steps of collecting operation data of a robot in a working process in real time through a sensor; dynamically adjusting control parameters according to real-time data changes; based on the reinforcement learning model, the complexity of the current task and the robot capability are analyzed in real time, and cooperative tasks among the multiple robots are intelligently distributed; the running state of the robot is continuously monitored, and early warning is given out when abnormity is found; according to the invention, real-time analysis and intelligent allocation are carried out on the task complexity and the robot capability through the reinforcement learning model, the task cooperation efficiency is optimized, and the resource utilization rate is improved; according to the method, an abnormal detection and early warning mechanism is constructed, an operation deviation state is rapidly identified, when an abnormity occurs, the abnormity is fed back to a control system through a graded early warning signal, task redistribution or parameter adjustment is carried out, and the self-adaptability, the cooperation capability and the operation safety of the industrial robot in a complex scene are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of industrial robot control technology, and in particular to an industrial robot control system and method based on deep learning. Background Art

[0002] In the field of industrial robot control, traditional control methods mainly rely on preset motion trajectories and fixed control parameters to achieve basic mechanical movements; these methods perform well in single, repetitive working environments, but when the task environment is complex and changeable or involves multi-robot collaboration, the limitations of traditional methods gradually become apparent; with the increasing diversification of industrial robot application scenarios and the continuous increase in task complexity, simple preset parameters and trajectories can no longer meet actual needs; especially in the face of real-time data changes, how to dynamically adjust control parameters to adapt to the real-time environment has become one of the important technical challenges of robot control.

[0003] However, traditional control methods find it difficult to effectively assess the complexity of tasks and the capabilities of the robots themselves, resulting in unreasonable allocation of multi-robot collaborative tasks and inefficient resource utilization. In addition, when faced with abnormal operating conditions (such as overload, abnormal speed, or abnormal energy consumption), traditional methods are unable to detect and issue early warning signals in a timely manner, which not only increases the risk of system operation, but may also cause irreversible damage to the equipment. Summary of the invention

[0004] In view of the problems existing in the existing industrial robot control technology, the present invention proposes an industrial robot control system and method based on deep learning.

[0005] Therefore, the problem to be solved by the present invention is how to solve the problem of intelligent task allocation and dynamic anomaly detection and early warning of multi-robot collaboration.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides an industrial robot control method based on deep learning, which includes collecting the robot's operating data during the work process in real time through sensors, including position, speed and load information; dynamically adjusting control parameters according to real-time data changes; based on a reinforcement learning model, analyzing the complexity of the current task and the robot's capabilities in real time, and intelligently allocating collaborative tasks among multiple robots; continuously monitoring the robot's operating status and issuing an early warning when an abnormality is found.

[0007] As a preferred solution of the deep learning-based industrial robot control method described in the present invention, multiple sensors are installed on the industrial robot, including position sensors, speed sensors, torque sensors and environmental sensors, to comprehensively monitor the operating status and working environment of the robot; real-time data of the robot when performing tasks is regularly collected, and the real-time data includes position data, speed data, torque data and environmental data.

[0008] The collected data are stored in a database and preprocessed, cleaned and standardized for subsequent analysis.

[0009] As a preferred solution of the industrial robot control method based on deep learning described in the present invention, wherein: feature extraction is performed on the real-time data, and the feature includes the current load , Current speed , Current ambient temperature And workpiece characteristic parameters , the workpiece characteristic parameters include material parameters and shape parameters.

[0010] Based on the extracted features, a dynamic parameter model is established to map the real-time data to the adjustment range of the control parameters and define the key control parameters, which include speed gain, acceleration limit and load response time, as follows: Based on the features, a mapping relationship is established to convert the extracted features into the adjustment range of the control parameters, and the control parameters are set, including the speed gain. , acceleration limit and load response time , the speed gain , acceleration limit and load response time The formulas are: ; ; ; in, represents the reference speed gain, represents the workpiece characteristic coefficient, represents the ambient temperature coefficient, Indicates the reference value of the workpiece characteristic, Indicates the ambient temperature reference value, represents the attenuation factor of the control speed, Indicates the base acceleration limit, represents the load influence coefficient, represents the ambient temperature influence coefficient, Indicates the upper limit reference value of load. represents the speed influence coefficient, Indicates the speed upper limit reference value, represents the benchmark response time, represents the load influence coefficient, represents the ambient temperature influence coefficient, Indicates the load reference value.

[0011] As a preferred solution of the industrial robot control method based on deep learning described in the present invention, the dynamic parameter model is applied to the control strategy, which is as follows: Use the control parameters generated by the dynamic parameter model to dynamically adjust the robot while it is running, and adjust the speed gain based on real-time monitoring data and acceleration limits , the adjustment formula is as follows: ; ; in, Indicates the preset safety threshold of the load, and Indicates the preset adjustment factor.

[0012] A performance feedback mechanism is introduced to evaluate the actual performance of the robot under different conditions, dynamically update the feature extraction and parameter models, and realize self-learning capabilities based on historical data.

[0013] As a preferred solution of the deep learning-based industrial robot control method described in the present invention, the task complexity index is extracted by analyzing the attributes of the current task.

[0014] Monitor the working status and performance of each robot in real time, including current load, energy consumption and remaining task capacity, to form a robot capability evaluation model.

[0015] Based on the evaluation results of task complexity and robot capabilities, a reinforcement learning decision model is constructed. The model is trained through historical data to optimize the task allocation strategy. The specific process is as follows: Collect and annotate the characteristics and historical allocation data of each task, including task complexity parameters and robot capability parameters. The task complexity parameters include task complexity , Task execution distance and task completion time requirements The robot capability parameters include the robot processing capability index , Current power level and maximum speed .

[0016] The task complexity parameter and the robot capability parameter are combined to design a state vector, which is recorded as ,in, .

[0017] The reward function is set according to the task completion efficiency and resource consumption. The reward formula is as follows: ; in, represents the task allocation strategy selected by the robot, represents the total energy of the robot, Indicates the actual time to complete the task. and represents the adjustment weight coefficient, The energy consumed to complete the task.

[0018] Record the state, action, reward, and next state of each round of task assignment to form an experience pool. Use the Q learning update formula to optimize the model to minimize the error between the strategy and the expected value. The loss function is as follows: ; Among them, r represents the immediate reward, represents the discount factor, represents the decay of future rewards, represents the model parameters, s represents the current state of the robot during the task execution, a represents the action selected by the robot in the current state, Expected function.

[0019] The model performance is evaluated by task completion rate, average resource consumption and task allocation time. The trained model is applied to real-time task allocation scenarios. Input, dynamically assign robots and tasks.

[0020] Reinforcement learning models are used to analyze in real time the match between task complexity and robot capabilities, and dynamically assign tasks to the most suitable robots to improve resource utilization and work efficiency, ensuring efficient collaboration among robots.

[0021] As a preferred solution of the deep learning-based industrial robot control method described in the present invention, the robot's operating status data is collected in real time, including position, speed, load and energy consumption.

[0022] Data analysis algorithms are used to analyze real-time monitoring data, identify states that deviate from normal operating modes, and form an anomaly detection mechanism, as follows: Building anomaly detection discriminant function , used to mark whether the current running status is abnormal. The function is as follows: ; in, represents the load of the robot at time t, Indicates the maximum allowable value of the robot load, represents the speed of the robot at time t, Indicates the maximum allowable value of the robot speed, represents the energy consumption of the robot at time t, Indicates the upper limit of the robot's energy consumption per unit time.

[0023] when , it indicates an abnormality has been detected.

[0024] when , the status is normal.

[0025] As a preferred solution of the industrial robot control method based on deep learning described in the present invention, an abnormal warning mechanism is established, which is as follows: When detected When , and record the abnormal type, abnormal parameters and detection time.

[0026] Set the warning level according to the abnormality type and the degree of parameter exceeding the limit.

[0027] ; Will and The information is fed back to the control system to determine whether tasks need to be reallocated or operating parameters need to be adjusted.

[0028] In the second aspect, an embodiment of the present invention provides an industrial robot control system based on deep learning, which includes an operation data acquisition module, which is used to comprehensively monitor the operation status and working environment of the robot by installing position, speed, torque, and environmental sensors on the industrial robot, regularly collect real-time data of the robot when performing tasks, and store the data in a database; a dynamic parameter control module, which is used to extract features from real-time data, and the extracted features include load, speed, ambient temperature, and workpiece characteristic parameters; a dynamic parameter model is established based on the features, the real-time data is mapped to the control parameter adjustment range, and key control parameters including speed gain, acceleration limit, and load response time are set; a reinforcement learning task allocation module, which is used to analyze the current task attributes and extract task complexity indicators, and construct a reinforcement learning decision model in combination with a robot capability evaluation model; a reward function is designed based on task completion efficiency and resource consumption and a Q learning optimization model is used to generate an optimal task allocation strategy; an abnormality monitoring and early warning module, which is used to collect the robot's operation status data in real time, mark abnormal states through an abnormality detection discriminant function, and trigger an early warning signal and set an early warning level based on the abnormality type and the degree of parameter exceeding the limit.

[0029] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the deep learning-based industrial robot control method as described in the first aspect of the present invention are implemented.

[0030] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the industrial robot control method based on deep learning as described in the first aspect of the present invention are implemented.

[0031] The beneficial effects of the present invention are as follows: the present invention collects robot operation data in real time through sensors to ensure that the subsequent deep learning model can be dynamically adjusted based on accurate data. In the control strategy, by dynamically adjusting the control parameters, the robot can maintain the best operating state under different task conditions. The task complexity and robot capabilities are analyzed and intelligently allocated in real time through the reinforcement learning model to optimize the task collaboration efficiency and improve resource utilization. At the same time, an abnormality detection and early warning mechanism is constructed, which can quickly identify the operation deviation state, and feedback to the control system through graded early warning signals when an abnormality occurs, and perform task reallocation or parameter adjustment to ensure safety; it significantly improves the adaptability, collaboration ability and operation safety of industrial robots in complex scenarios, and realizes efficient and intelligent operation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0033] Figure 1 Schematic diagram of the steps of the industrial robot control method based on deep learning.

[0034] Figure 2 Schematic diagram of the structure of an industrial robot control system based on deep learning. DETAILED DESCRIPTION

[0035] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0036] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0037] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments. Example

[0038] Reference Figure 1~Figure 2 , which is the first embodiment of the present invention, and provides an industrial robot control method based on deep learning, including: S1: The robot’s operating data, including position, speed and load information, is collected in real time through sensors during the working process to provide a basis for subsequent analysis.

[0039] Various sensors are installed on industrial robots, including position sensors, speed sensors, torque sensors and environmental sensors, to comprehensively monitor the robot's operating status and working environment; real-time data of the robot when performing tasks is regularly collected, and the real-time data includes position data, speed data, torque data and environmental data.

[0040] The collected data are stored in a database and preprocessed, cleaned and standardized for subsequent analysis.

[0041] S2: Dynamically adjust control parameters based on real-time data changes to optimize the robot's operating performance and ensure effective operation in different workpieces or environments.

[0042] Extract features from the real-time data, including the current load , Current speed , Current ambient temperature And workpiece characteristic parameters , the workpiece characteristic parameters include material parameters and shape parameters.

[0043] Based on the extracted features, a dynamic parameter model is established to map the real-time data to the adjustment range of the control parameters. Key control parameters including velocity gain, acceleration limit and load response time are defined to cope with different workpieces and environmental conditions, as follows: Based on the features, a mapping relationship is established to convert the extracted features into the adjustment range of the control parameters, and the control parameters are set, including the speed gain. , acceleration limit and load response time , the speed gain , acceleration limit and load response time The formulas are: ; ; ; in, represents the reference speed gain, represents the coefficient related to the workpiece characteristics, experimental adjustment, Indicates the coefficient related to the ambient temperature, experimental adjustment, Indicates the reference value of the workpiece characteristic (for example, the standard hardness of the workpiece), Indicates the ambient temperature reference value (for example, the set temperature at normal temperature), Indicates the attenuation factor of the control speed, which determines the sensitivity of the speed gain to the current speed. Indicates the base acceleration limit, represents the load influence coefficient, represents the ambient temperature influence coefficient, Indicates the upper limit reference value of load. represents the speed influence coefficient, Indicates the speed upper limit reference value, represents the benchmark response time, represents the load influence coefficient, represents the ambient temperature influence coefficient, Indicates the load reference value.

[0044] The dynamic parameter model is applied to the control strategy as follows: Use the control parameters generated by the dynamic parameter model to dynamically adjust the robot while it is running, and adjust the speed gain based on real-time monitoring data and acceleration limits , the adjustment formula is as follows: ; ; in, Indicates the preset safety threshold of the load, and Indicates the preset adjustment coefficient; By adjusting control parameters (such as velocity gain and acceleration limit) in real time, the robot can maintain the best operating state under different workpieces or environments, thereby optimizing execution efficiency, as follows: ; in, To control the output, is the error, , and are the proportional, integral and derivative gains respectively.

[0045] A performance feedback mechanism is introduced to evaluate the actual performance of the robot under different conditions, dynamically update the feature extraction and parameter models, so as to continuously improve the control strategy and realize self-learning capability based on historical data.

[0046] S3: Based on the reinforcement learning model, it analyzes the complexity of the current task and the robot's capabilities in real time, and intelligently allocates collaborative tasks among multiple robots to improve overall work efficiency.

[0047] Specifically, by analyzing the attributes of the current task, task complexity indicators such as operation time, accuracy requirements, and workpiece characteristics are extracted to provide basic data for subsequent task allocation.

[0048] Furthermore, the working status and performance of each robot, including the current load, energy consumption and remaining task capacity, are monitored in real time to form a robot capability assessment model to ensure that task allocation takes into account the actual operational capabilities of each robot.

[0049] Based on the evaluation results of task complexity and robot capabilities, a reinforcement learning decision model is constructed. The model is trained through historical data to enable it to make intelligent decisions in various task scenarios and optimize task allocation strategies. The specific process is as follows: Collect and annotate the characteristics and historical allocation data of each task, including task complexity parameters and robot capability parameters. The task complexity parameters include task complexity , Task execution distance and task completion time requirements The robot capability parameters include the robot processing capability index , Current power level and maximum speed .

[0050] The task complexity parameter and the robot capability parameter are combined to design a state vector, which is recorded as ,in, .

[0051] The reward function is set according to the task completion efficiency and resource consumption. The reward formula is as follows: ; in, Represents the task allocation strategy selected by the robot, that is, Assign to robot , It represents the energy consumed to complete the task. represents the total energy of the robot, Indicates the actual time to complete the task. and Represents the adjustment weight coefficient, which controls the importance of time and energy consumption respectively.

[0052] Furthermore, the state, action, reward, and next state of each round of task assignment are recorded to form an experience pool, and the Q-learning update formula is used to optimize the model to minimize the error between the strategy and the expected value. The loss function is as follows: ; Among them, r represents the immediate reward, represents the discount factor, represents the decay of future rewards, the current Q value estimate, represents the model parameters, s represents the current state of the robot during the task execution, a represents the action selected by the robot in the current state, Expected function.

[0053] The model performance is evaluated by task completion rate, average resource consumption and task allocation time. The trained model is applied to real-time task allocation scenarios. Based on the input, robots and tasks are dynamically allocated to achieve optimized decision-making.

[0054] Reinforcement learning models are used to analyze in real time the match between task complexity and robot capabilities, and dynamically assign tasks to the most suitable robots to improve resource utilization and work efficiency, ensuring efficient collaboration among robots.

[0055] S4: Continuously monitor the robot's operating status and issue warnings in a timely manner when abnormalities are detected to ensure safe operation.

[0056] First, the robot’s operating status data, including position, speed, load, and energy consumption, is collected in real time to ensure comprehensive monitoring of the robot’s operating process.

[0057] Data analysis algorithms are used to analyze real-time monitoring data to identify states that deviate from normal operating modes, such as overload, abnormal speed, or abnormal energy consumption, and form an abnormality detection mechanism, as follows: Building anomaly detection discriminant function , used to mark whether the current running status is abnormal. The function is as follows: ; in, represents the load of the robot at time t, Indicates the maximum allowable value of the robot load, represents the speed of the robot at time t, Indicates the maximum allowed value of the robot speed, represents the energy consumption of the robot at time t, Indicates the upper limit of the robot's energy consumption per unit time.

[0058] when , it indicates an abnormality has been detected.

[0059] when , the status is normal.

[0060] Furthermore, an abnormal warning mechanism is established, as follows: When detected When , and record the abnormal type, abnormal parameters and detection time. For example, different warning levels can be set for different types of abnormalities. , such as "load over limit" and "abnormal energy consumption" can be set as high priority.

[0061] According to the abnormality type and the degree of parameter exceeding the limit, set the warning level so as to take different levels of response measures: ; Will and The information is fed back to the control system, which helps the management system determine whether it needs to reallocate tasks or adjust operating parameters.

[0062] Furthermore, the present embodiment also provides an industrial robot control system based on deep learning, including: an operation data acquisition module, which is used to comprehensively monitor the operation status and working environment of the robot by installing multiple sensors such as position, speed, torque, and environmental sensors on the industrial robot; the module will regularly collect real-time data (position, speed, torque, and environmental data) of the robot when performing tasks, and store the data in a database to complete data cleaning and standardization processing; a dynamic parameter control module, which is used to extract features from real-time data, and the extracted features include load, speed, ambient temperature, and workpiece characteristic parameters; a dynamic parameter model is established based on the features, and the real-time data is mapped to the control parameter adjustment range, and key control parameters such as speed gain, acceleration limit, and load response time are set to ensure that the robot reaches the optimal operating state under different workpieces or environmental conditions; a reinforcement learning task allocation module, which is used to analyze the current task attributes and extract task complexity indicators (such as operation time, accuracy requirements, and workpiece characteristics), and build a reinforcement learning decision model in combination with a robot capability evaluation model (including load, energy consumption, task capability, etc.). The reward function is designed based on task completion efficiency and resource consumption, and the Q learning optimization model is used to generate the optimal task allocation strategy; the abnormal monitoring and early warning module is used to collect the robot's operating status data (including position, speed, load and energy consumption) in real time, and use data analysis algorithms to detect states that deviate from the normal operating mode, such as load overrun, speed abnormality and energy consumption abnormality. The abnormal state is marked by the abnormal detection discriminant function, and the early warning signal is triggered and the warning level is set according to the abnormal type and the degree of parameter overrun, and the early warning is fed back to the control system to adjust the task allocation or control parameters.

[0063] This embodiment also provides a computer device, which is suitable for the case of an industrial robot control method based on deep learning, and includes a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the industrial robot control method based on deep learning proposed in the above embodiment.

[0064] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0065] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the industrial robot control method based on deep learning as proposed in the above embodiment is implemented.

[0066] In summary, the present invention collects robot operation data in real time through sensors to ensure that the subsequent deep learning model can be dynamically adjusted based on accurate data. In the control strategy, by dynamically adjusting the control parameters, the robot can maintain the best operating state under different task conditions. The task complexity and robot capabilities are analyzed and intelligently allocated in real time through the reinforcement learning model to optimize the task collaboration efficiency and improve resource utilization. At the same time, an abnormality detection and early warning mechanism is constructed, which can quickly identify the operation deviation state, and feedback to the control system through graded early warning signals when an abnormality occurs, and perform task reallocation or parameter adjustment to ensure safety; it significantly improves the adaptability, collaboration ability and operation safety of industrial robots in complex scenarios, and realizes efficient and intelligent job management. Example

[0067] In the second embodiment, an industrial robot control method based on deep learning is provided. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0068] The experiment was conducted in a standard industrial environment, where multiple industrial robots performed similar assembly tasks. During the experiment, real-time data was collected from each robot at regular intervals and stored in a database. The data collection frequency was set to once per second to ensure that subtle changes in the robot's working state could be captured. After the data was stored, data preprocessing was performed, including cleaning and standardization, to ensure the accuracy of subsequent analysis.

[0069] Next, the control parameters are dynamically adjusted according to the changes in the collected real-time data to optimize the robot's operating performance. During the specific implementation process, features such as the current load, speed, ambient temperature, and the material and shape of the workpiece are extracted. Then, based on the extracted features, a dynamic parameter model is established to map the real-time data to the adjustment range of the control parameters. Based on these features, control parameters are set, including speed gain, acceleration limit, and load response time. The relevant coefficients and reference values ​​are defined in the formula to cope with different workpieces and environmental conditions.

[0070] In practical applications, the control parameters generated by the dynamic parameter model are applied to the robot control strategy in real time. When an abnormal situation is detected, such as load exceeding the limit or speed abnormality, the system will trigger a warning signal and record the abnormal parameters and detection time. By comparing the working status and performance of different robots, the task allocation is dynamically adjusted to ensure that each robot works in the best condition. In order to verify the effectiveness of this method, the experimental group was compared with the control group, and the differences in the robots' task completion rate, energy consumption and execution time were recorded.

[0071] Table 1 Industrial robot operation status and performance data Test subject name Location (m) Speed ​​(m / s) Load(kg) Energy consumption (kWh) Ambient temperature (°C) Task complexity (scoring) Robot 1 2.5 1.5 10 0.5 22 5 Robot 2 3.0 1.8 12 0.6 23 7 Robot 3 1.8 1.4 8 0.4 21 4 Robot 4 2.7 1.6 15 0.7 24 6 Robot 5 3.1 1.9 11 0.5 23 5 Robot 6 2.0 1.7 9 0.6 20 4 Robot 7 2.3 1.5 13 0.4 22 6 According to Table 1, we can observe the energy consumption performance of the robots under different load conditions. For example, robot 4 has the largest load (15 kg), but its energy consumption (0.7 kWh) is also relatively high; while robots with smaller loads (such as robot 3, 8 kg) have an energy consumption of only 0.4 kWh. This shows that there is a significant relationship between load control and energy consumption.

[0072] The monitoring results of speed and position data show that when the robot is at a faster speed (such as robot 5, 1.9 m / s), its task complexity score (5) does not exceed the scores of other robots; on the contrary, the slower robot (such as robot 3, 1.4 m / s) shows a large room for improvement in task complexity score (score 4), which shows that optimizing speed gain and acceleration limit is crucial to improving the robot's adaptability in complex task environments.

[0073] Through comprehensive analysis of various parameters, it is concluded that the dynamic adjustment control strategy adopted by the present invention has significant advantages; compared with the existing technology, this method optimizes the robot's operating efficiency and task allocation strategy through real-time feedback and data analysis; specifically, in the scenario of multiple robots working together, the implemented self-learning ability enables the robot to continuously adjust its own parameters according to historical data, thereby improving resource utilization and overall work efficiency.

[0074] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An industrial robot control method based on deep learning, characterized in that: include: S1: Collect the robot’s operating data in real time during the working process through sensors; S2: Dynamically adjust control parameters according to real-time data changes; S3: Based on the reinforcement learning model, it analyzes the complexity of the current task and the robot's capabilities in real time, and intelligently allocates collaborative tasks among multiple robots; S4: Continuously monitor the robot's operating status and issue an early warning when an abnormality is found; The specific implementation process of S2 includes: Extract features from the real-time data, including the current load , Current speed , Current ambient temperature And workpiece characteristic parameters , the workpiece characteristic parameters include material parameters and shape parameters; According to the extracted features, a dynamic parameter model is established, real-time data is mapped to the adjustment range of the control parameters, and key control parameters are defined. The key control parameters include speed gain, acceleration limit and load response time.

2. The industrial robot control method based on deep learning according to claim 1, characterized in that: The specific implementation process of step S1 includes: Install a variety of sensors on industrial robots, including position sensors, speed sensors, torque sensors, and environmental sensors; Regularly collect real-time data of the robot when performing tasks, the real-time data including position data, speed data, torque data and environmental data; The collected data are stored in a database and preprocessed, cleaned and standardized.

3. The industrial robot control method based on deep learning as claimed in claim 2, characterized in that: Establish a mapping relationship based on the features, convert the extracted features into the adjustment range of the control parameters, and set the control parameters, including speed gain , acceleration limit and load response time , the speed gain , acceleration limit and load response time The formulas are: ； ； ； in, represents the base speed gain, represents the workpiece characteristic coefficient, represents the ambient temperature coefficient, Indicates the reference value of the workpiece characteristic, Indicates the ambient temperature reference value, represents the attenuation factor of the control speed, represents the base acceleration limit, represents the load influence coefficient, represents the ambient temperature influence coefficient, Indicates the upper limit reference value of load. represents the speed influence coefficient, Indicates the speed upper limit reference value, represents the benchmark response time, represents the load influence coefficient, represents the ambient temperature influence coefficient, Indicates the load reference value.

4. The industrial robot control method based on deep learning as claimed in claim 3, characterized in that: The S2 further comprises the following steps: The dynamic parameter model is applied to the control strategy as follows: Use the control parameters generated by the dynamic parameter model to dynamically adjust the robot while it is running, and adjust the speed gain based on real-time monitoring data and acceleration limits , the adjustment formula is as follows: ； ； in, Indicates the preset safety threshold of the load, and Indicates the preset adjustment coefficient; A performance feedback mechanism is introduced to evaluate the actual performance of the robot under different conditions, dynamically update the feature extraction and parameter models, and realize self-learning capabilities based on historical data.

5. The industrial robot control method based on deep learning according to claim 4, characterized in that: The specific implementation process of S3 includes: By analyzing the attributes of the current task, task complexity indicators are extracted; Monitor the working status and performance of each robot in real time, including current load, energy consumption and remaining task capacity, to form a robot capability evaluation model; Based on the evaluation results of task complexity and robot capabilities, a reinforcement learning decision model is constructed. The model is trained through historical data to optimize the task allocation strategy. The specific process is as follows: Collect and annotate the characteristics and historical allocation data of each task, including task complexity parameters and robot capability parameters. The task complexity parameters include task complexity , Task execution distance and task completion time requirements The robot capability parameters include the robot processing capability index , Current power level and maximum speed ; The task complexity parameter and the robot capability parameter are combined to design a state vector, which is recorded as ,in, ; The reward function is set according to the task completion efficiency and resource consumption. The reward formula is as follows: ； in, represents the task allocation strategy selected by the robot, represents the total energy of the robot, Indicates the actual time to complete the task. and represents the adjustment weight coefficient, Energy consumed to complete the task; Record the state, action, reward, and next state of each round of task assignment to form an experience pool. Use the Q learning update formula to optimize the model to minimize the error between the strategy and the expected value. The loss function is as follows: ； Among them, r represents the immediate reward, represents the discount factor, represents the decay of future rewards, represents the model parameters, s represents the current state of the robot during the task execution, a represents the action selected by the robot in the current state, Expected function; The model performance is evaluated by task completion rate, average resource consumption and task allocation time. The trained model is applied to real-time task allocation scenarios. Input, dynamically allocate robots and tasks; Reinforcement learning models are used to analyze in real time the match between task complexity and robot capabilities, and dynamically assign tasks to the most suitable robots to improve resource utilization and work efficiency, ensuring efficient collaboration among robots.

6. The industrial robot control method based on deep learning according to claim 5, characterized in that: The specific implementation process of S4 includes: Collect the robot's operating status data in real time, including position, speed, load and energy consumption; Data analysis algorithms are used to analyze real-time monitoring data, identify states that deviate from normal operating modes, and form an anomaly detection mechanism, as follows: Building anomaly detection discriminant function , used to mark whether the current running status is abnormal. The function is as follows: ； in, represents the load of the robot at time t, Indicates the maximum allowable value of the robot load, represents the speed of the robot at time t, Indicates the maximum allowable value of the robot speed, represents the energy consumption of the robot at time t, Indicates the upper limit of the robot's energy consumption per unit time; when , it indicates that an abnormality is detected; when , the status is normal.

7. The industrial robot control method based on deep learning according to claim 6, characterized in that: The step S4 further comprises the following steps: Establish an abnormal early warning mechanism, as follows: When detected When , and record the abnormal type, abnormal parameters and detection time; Set the warning level according to the abnormality type and the degree of parameter exceeding the limit: ； Will and The information is fed back to the control system to determine whether tasks need to be reallocated or operating parameters need to be adjusted.

8. An industrial robot control system based on deep learning, based on the industrial robot control method based on deep learning according to any one of claims 1 to 7, characterized in that: Also includes, The operation data acquisition module is used to comprehensively monitor the operation status and working environment of the robot by installing sensors on the industrial robot, regularly collect real-time data of the robot when performing tasks, and store the data in the database; The dynamic parameter control module is used to extract the characteristics of real-time data, extract characteristic parameters, establish a dynamic parameter model based on the characteristics, map the real-time data to the control parameter adjustment range, and set key control parameters; The reinforcement learning task allocation module is used to analyze the current task attributes and extract task complexity indicators. It combines the robot capability evaluation model to build a reinforcement learning decision model, designs the reward function based on task completion efficiency and resource consumption, and finally generates the optimal task allocation strategy; The abnormality monitoring and early warning module is used to collect the robot's operating status data in real time, mark the abnormal status through the abnormality detection discriminant function, and trigger the early warning signal and set the early warning level according to the abnormality type and the degree of parameter exceeding the limit.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the deep learning-based industrial robot control method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the deep learning-based industrial robot control method described in any one of claims 1 to 7 are implemented.

Images