Control method and control system of manipulator and electronic equipment

Abstract

The embodiment of the invention provides a control method and system of a manipulator and electronic equipment, and relates to the field of control. The method comprises the steps that a terminal layer collects real-time environment data and uploads the environment data to edge equipment; if the edge device determines that the environment is abnormal according to the environment data, the edge device reports alarm information of the environment abnormality to the cloud in real time; the cloud inputs the environmental data and the digital twin model of the manipulator into a pre-trained airing strategy model to obtain a control instruction, and issues the control instruction to the edge device; the edge device drives the manipulator according to the control instruction. The embodiment of the invention breaks through the limitation of a traditional fixed rule, can dynamically adapt to a complex and changeable environment, improves the airing efficiency and the clothes care quality, reduces the manual intervention cost, and achieves the intellectualization, automation and individuation of clothes airing.

Description

Technical Field

[0001] This application relates to the field of control technology, and more specifically, to a control method, control system, electronic device, computer-readable storage medium, and computer program product for a robotic arm. Background Technology

[0002] Robotic arms, with their telescopic arms having a variable working radius, enable spatial positioning of end effectors (such as grippers and welding torches), and are widely used in fields such as service robots.

[0003] In related technologies, the dynamic characteristics of robotic arms and the uncertainty of the network environment pose serious challenges to control accuracy and real-time performance. Summary of the Invention

[0004] This application provides a control method, control system, electronic device, computer-readable storage medium, and computer program product for a robotic arm, which can solve the above-mentioned problems of the prior art. The technical solution is as follows: According to one aspect of the embodiments of this application, a control method for a robotic arm is provided, the method comprising: The terminal layer collects real-time environmental data and uploads the environmental data to the edge device. The terminal layer includes a robotic arm and at least one environmental sensor. The robotic arm is used for drying clothes. If the edge device determines that the environment is abnormal based on the environmental data, it shall report the environmental abnormality alarm information to the cloud in real time, and the alarm information includes the environmental data. The cloud inputs the environmental data and the digital twin model of the robotic arm into a pre-trained drying strategy model, obtains the control commands output by the drying strategy model, and sends the control commands to the edge device. The control commands are used to indicate the running time and running parameters of the robotic arm. The edge device drives the robotic arm according to the control commands.

[0005] According to another aspect of the embodiments of this application, a control device for a robotic arm is provided, the device comprising: The terminal layer is used to collect real-time environmental data and upload the environmental data to the edge device. The terminal layer includes a robotic arm and at least one environmental sensor. The robotic arm is used to dry clothes. An edge device is used to report an environmental anomaly alarm to the cloud in real time if an environmental anomaly is determined based on the environmental data, wherein the alarm information includes the environmental data. In the cloud, the environmental data and the digital twin model of the robotic arm are input into a pre-trained drying strategy model to obtain control commands output by the drying strategy model. The control commands are then sent to the edge device to indicate the running time and running parameters of the robotic arm. The edge device is also used to drive the robotic arm according to the control instructions.

[0006] According to another aspect of the present application, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the above-described control method for a robotic arm.

[0007] According to another aspect of the embodiments of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the above-described control method for a robotic arm.

[0008] According to one aspect of the embodiments of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described control method for a robotic arm.

[0009] The beneficial effects of the technical solution provided in this application are as follows: the control method of the robotic arm forms an edge-cloud closed loop. This solution realizes intelligent clothes drying control based on a multi-level architecture, especially excelling in obtaining control commands based on models. The terminal layer collects environmental data such as temperature, humidity, light intensity, and wind speed in real time through environmental sensors, providing accurate basis for decision-making. Edge devices quickly identify and report environmental anomalies, ensuring timely response from the cloud. The cloud inputs environmental data and the robotic arm's digital twin model into the drying strategy model. This model, trained with a large amount of data, can deeply analyze the relationship between environmental factors and drying effects, and output scientific and reasonable control commands. The control commands precisely indicate the robotic arm's operating time and parameters. For example, the optimal drying time can be determined based on light intensity and wind speed, and the robotic arm's rotation frequency can be adjusted based on humidity to avoid excessive sun exposure or dampness and mold growth on clothes. This model-based control method breaks through the limitations of traditional fixed rules, dynamically adapts to complex and changing environments, improves drying efficiency and clothing care quality, while reducing manual intervention costs, and realizing intelligent, automated, and personalized clothes drying. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0011] Figure 1 A flowchart illustrating a control method for a robotic arm provided in an embodiment of this application; Figure 2A schematic diagram of the control system for a robotic arm provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0012] The embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the embodiments described below with reference to the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions of the embodiments of this application.

[0013] Those skilled in the art will understand that, unless otherwise stated, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. It should be further understood that the terms “comprising” and “including” as used in embodiments of this application mean that the corresponding feature can be implemented as the presented feature, information, data, step, operation, element, and / or component, but do not exclude implementation as other features, information, data, step, operation, element, component, and / or combinations thereof supported by the art. It should be understood that when we say that an element is “connected” or “coupled” to another element, the one element can be directly connected or coupled to the other element, or it can mean that the one element and the other element establish a connection relationship through an intermediate element. Furthermore, “connected” or “coupled” as used herein can include wireless connection or wireless coupling. The term “and / or” as used herein indicates at least one of the items defined by the term; for example, “A and / or B” can be implemented as “A,” or as “B,” or as “A and B.”

[0014] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0015] The robotic arm control methods in related technologies have at least one of the following problems: Command and feedback latency issues: There is a significant latency (tens to hundreds of milliseconds) in cloud command issuance, which greatly affects scaling operations that require rapid response or precise synchronization. It will severely reduce response speed and accuracy. There is also a delay in uploading critical status information to the cloud for monitoring or remote operation, causing the "real-time" screen seen by remote operators to deviate from the actual status and increasing the risk of misoperation.

[0016] The problem of over-reliance on network quality: It is sensitive to network bandwidth, jitter, and packet loss. When the network fluctuates, control commands may be lost or out of order, and status feedback may be interrupted, leading to a decline in control performance or even loss of control.

[0017] Limited functionality of edge devices: In existing solutions, edge devices are mainly responsible for real-time closed-loop control and safety logic. Their computing power is usually insufficient to run complex advanced control algorithms or real-time intelligent decision-making that require telescopic arm dynamics models. These advanced algorithms often need to be deployed in the cloud.

[0018] The problem of separating intelligence from real-time control: Intelligent decision-making takes place in the cloud, while real-time control is executed on the device. This separation prevents intelligent decisions from being adjusted in real time according to the precise and ever-changing dynamic state of the robotic arm. The optimal path / instruction generated in the cloud may no longer be optimal or even infeasible when executed by the edge device due to changes in the state.

[0019] Simplified model and accuracy loss problem: Existing common control schemes often simplify the telescopic arm to linear motion treatment or use overly simplified dynamic models, ignoring the significant changes in mass distribution, inertia, flexible deformation (especially when the manipulator extends) and strong nonlinear friction brought about by the telescopic motion. This results in poor trajectory tracking accuracy, large end-effector vibration, and difficulty in guaranteeing positioning accuracy during high-speed, high-acceleration or variable-load telescopic motion.

[0020] Lack of targeted optimization: The integration of compensation for the unique dynamic characteristics of telescopic booms (such as inertia compensation, friction compensation, and vibration suppression based on real-time length and load) and adaptive adjustment of control strategies (such as automatically adjusting control parameters according to the telescopic length) is low or ineffective in existing networking solutions.

[0021] State consistency is difficult to guarantee: During the collaborative process, the state information of each robot needs to be shared in real time, but network latency and uncertainty will cause inconsistencies in the "global state" obtained by each node, affecting the accuracy and robustness of collaborative decision-making.

[0022] Weak fault recovery and fault tolerance: When the network is interrupted or a robot / edge device fails, the existing system lacks a fast and effective fault detection, isolation and recovery mechanism, which can easily lead to the interruption of the entire collaborative task or trigger a chain of failures.

[0023] Predictive maintenance lag: Although cloud-based data analysis is possible, predictive maintenance models based on historical data lack the accuracy and real-time performance to predict progressive faults specific to telescopic booms that are related to dynamic loads and lengths (such as bearing wear and internal leakage in hydraulic cylinders at specific telescopic lengths).

[0024] The control method, device, electronic equipment, computer-readable storage medium, and computer program product for the robotic arm provided in this application are intended to solve the above-mentioned technical problems of the prior art.

[0025] The technical solutions of this application and their effects are described below through several exemplary embodiments. It should be noted that the following embodiments can be referenced, borrowed from, or combined with each other. Identical terms, similar features, and similar implementation steps in different embodiments will not be repeated.

[0026] This application provides a method for controlling a robotic arm, such as... Figure 1 As shown, the method includes: S101. The terminal layer collects real-time environmental data and uploads the environmental data to the edge device.

[0027] The terminal layer of this application embodiment includes a robotic arm and at least one environmental sensor. The robotic arm adopts a multi-stage sleeve arm structure driven by a lead screw, which can achieve stepless extension from 0.8 to 2.5 meters. Its single-rod load-bearing capacity is up to 30 kg. It is equipped with a dual horizontal reciprocating motion mechanism and also integrates a clothes drying rod that can be rotated and adjusted 180° as well as flexible grippers, which can adapt to different loads such as clothes and bedding.

[0028] Taking a drying scenario as an example, the environmental sensors in this application embodiment include raindrop sensors, photosensors, and temperature and humidity sensors, etc. The raindrop sensor collects rainfall data, can accurately sense whether rainfall has occurred, and can also reflect the intensity of rainfall through changes in the output signal; the photosensor collects ambient light intensity data, and can sense the brightness of the surrounding ambient light; the temperature and humidity sensor is used to collect ambient temperature and humidity data, and can accurately measure the temperature and relative humidity values ​​of the surrounding environment.

[0029] S102. If the edge device determines that the environment is abnormal based on the environmental data, it shall report the environmental abnormality alarm information to the cloud in real time, and the alarm information includes the environmental data.

[0030] Edge devices are hardware devices deployed at the network edge, close to data sources or user terminals, possessing certain data processing, storage, and communication capabilities, and capable of real-time decision-making and response locally. In this application's embodiments, the edge device can be a smart router, gateway, etc.

[0031] In this embodiment, after receiving environmental data collected by the terminal layer, the environmental data is analyzed to determine whether the environment is abnormal. Environmental anomalies may include sudden temperature changes, strong winds, rain, snow, etc. After determining that the environment is abnormal, the edge device in this embodiment reports the environmental anomaly alarm information to the cloud in real time, so that the cloud can promptly generate control commands for the robotic arm based on the alarm information.

[0032] S103. The cloud inputs the environmental data and the digital twin model of the robotic arm into the pre-trained drying strategy model, obtains the control instructions output by the drying strategy model, and sends the control instructions to the edge device. The control instructions are used to indicate the running time and running parameters of the robotic arm.

[0033] In this embodiment of the application, the cloud serves as the core for device access and management, responsible for global optimization and AI training. It trains the drying strategy model based on environmental data (humidity, probability of rainfall) and user habits to achieve dynamic task planning.

[0034] Furthermore, a digital twin model of the robotic arm is pre-built in the cloud, mapping it 1:1, integrating physical, kinematic, and dynamic properties to simulate energy consumption and extension trajectory under different loads and optimize operating parameters.

[0035] By inputting sensor data and digital twin models into the drying strategy model in the cloud, the drying strategy model can determine global operational decisions based on environmental data, namely whether to deploy or retract the robotic arm, and when to execute it. For example, if weather information indicates that a heavy rain is imminent, it can determine the optimal time to retract the robotic arm. On the other hand, the drying strategy model also determines the robotic arm's operating parameters, such as motor power and the robotic arm's operating speed, based on the digital twin model.

[0036] The drying strategy model in this application optimizes the start time and execution time based on real-time weather and the twin state of clothing weights and materials. In scenarios such as rain and strong winds, it allows for earlier collection or delayed drying, improving quality and energy efficiency. Furthermore, it can dynamically provide parameters such as rotation speed, swing amplitude, and spacing based on light, temperature, and humidity to suppress wrinkles, overexposure, and dampness, adapting to various clothing types and changing weather conditions. In case of anomalies, it triggers alarms and strategy switching (such as emergency collection or switching to indoor mode), and verifies the action sequence through twin simulation to reduce the risk of collisions and falls.

[0037] After receiving the control commands, the cloud sends the commands down to the edge devices, enabling the edge devices to drive the robotic arm according to the commands.

[0038] S104. The edge device drives the robotic arm according to the control command.

[0039] The robotic arm control method provided in this application forms an edge-cloud closed loop. This solution achieves intelligent clothing drying control based on a multi-level architecture, particularly excelling in obtaining control commands based on models. The terminal layer collects environmental data such as temperature, humidity, light intensity, and wind speed in real time through environmental sensors, providing accurate basis for decision-making. Edge devices quickly identify and report environmental anomalies, ensuring timely response from the cloud. The cloud inputs environmental data and the robotic arm's digital twin model into the drying strategy model. This model, trained with extensive data, can deeply analyze the relationship between environmental factors and drying effects, outputting scientifically sound control commands. The control commands precisely indicate the robotic arm's operating time and parameters. For example, the optimal drying time can be determined based on light intensity and wind speed, and the robotic arm's rotation frequency can be adjusted based on humidity to prevent excessive sun exposure or mold growth on clothing. This model-based control method breaks through the limitations of traditional fixed rules, dynamically adapting to complex and changing environments, improving drying efficiency and clothing care quality, while reducing manual intervention costs, achieving intelligent, automated, and personalized clothing drying.

[0040] Based on the above embodiments, as an optional embodiment, the terminal layer further includes a mechanical sensor for collecting the motion and force of the robotic arm, and the method further includes: The dynamic data of the robot arm under different motion states are collected using mechanical sensors. The dynamic data includes velocity, acceleration, and joint torque. The dynamic data is input into the dynamic model to obtain the updated dynamic model; Based on the updated dynamic model, the mass distribution and inertia changes of the robot are determined using the locally deployed lightweight model solver.

[0041] During the continuous operation of the robotic arm, multiple high-precision sensors deployed on it work synchronously. Accelerometers measure the acceleration data of each joint and end effector of the robotic arm in real time under different motion states, gyroscopes accurately capture its angular velocity information, and force sensors record the torques experienced by the joints. These sensors continuously collect data at a preset high frequency to ensure the acquisition of comprehensive and accurate dynamic information. The collected data covers various working conditions of the robotic arm under different arm lengths (L), loads (P), and speeds (v), providing a rich and representative data foundation for subsequent model updates.

[0042] When the preset update cycle is reached, the model update process is automatically initiated. The large amount of collected dynamic data is imported into a specially developed model update module. This module employs advanced parameter identification algorithms, such as the Kalman filter algorithm. This algorithm converts the dynamic model into a dynamic model: M(L,P)× +C(L,P,v) × +G(L,P)+F friction(L,v) = τ is used as the basis, where, For speed, Let τ be the acceleration, τ be the joint torque, M(L,P) be the time-varying inertia matrix, C(L,P,v) be the Coriolis force term, G(L,P) be the gravity term, and F be the acceleration. friction The friction force is nonlinear. Through iterative calculations on the input data, the parameters in the dynamic model are continuously adjusted, gradually reducing the error between the output of the dynamic model and the actually collected joint torque and other data. After multiple iterations of optimization, an updated dynamic model is finally obtained, which can more accurately describe the dynamic characteristics of the robot in its current state.

[0043] This application employs parameter identification methods, such as least squares or Kalman filtering, to re-estimate the parameters in the dynamic model, resulting in an updated dynamic model. Based on the updated dynamic model, the mass distribution and inertia variation of the robot are determined through mathematical derivation and calculation. For example, the mass distribution of the robot under different arm lengths and load conditions is analyzed using the element changes of the inertia matrix M(L,P).

[0044] Taking the least squares method as an example, the error function is constructed as follows:

[0045] The model is updated by finding the parameters that minimize J.

[0046] This application embodiment is based on the updated dynamic model, and adjusts the inertia matrix. M ( L , P A thorough analysis reveals that each element of the inertia matrix is ​​closely related to the mass and moment of inertia of different parts of the manipulator. Analysis of the diagonal elements clarifies the magnitude and variation of the manipulator's moment of inertia along different coordinate axes, thus inferring the mass distribution trend in different directions. The off-diagonal elements reflect the coupled inertia between different coordinate axes. Studying these elements further reveals the complex characteristics of the manipulator's mass distribution and the interrelationships between its parts, thereby comprehensively determining the manipulator's mass distribution and moment of inertia variation.

[0047] It should be noted that the inertia matrix M ( L , P The elements of ) are related to the mass and moment of inertia of each part of the robotic arm. Through the analysis of... M ( L , PAnalysis of the inertia matrix can determine the mass distribution of the robot under different positions and loads. For example, the diagonal elements of the inertia matrix represent the rotational inertia of the robot in different coordinate axis directions, and their changes reflect changes in mass distribution. The off-diagonal elements reflect the coupled inertia between different coordinate axis directions, which are also related to mass distribution.

[0048] The lightweight model in this application embodiment is a model obtained by lightweighting the drying strategy model in the cloud. This application embodiment can use quantization, pruning, and knowledge distillation to lightweight the drying strategy model. Taking quantization as an example, the model parameters are converted from high-precision floating-point numbers (such as FP32) to low-precision integers (such as INT4 / INT8), which significantly reduces the model size and computational load.

[0049] In actual operation, during the extension and retraction process of the robotic arm, positional deviation and nonlinear friction problems inevitably occur due to the influence of mechanical structure characteristics, manufacturing precision, and environmental factors. Positional deviation can prevent the end effector of the robotic arm from accurately reaching the target position, affecting the accuracy of operation; nonlinear friction can interfere with the motion state of the robotic arm, making its acceleration unstable and reducing the smoothness of motion and response speed. To address this problem, based on the above embodiments, as an optional embodiment, this application embodiment further includes: In response to disconnecting from the cloud, the edge device collects the deformation of the robotic arm and uses the deformation to determine the position compensation value and nonlinear friction compensation value of the robotic arm during the extension and retraction process based on the locally deployed lightweight model. The end effector position of the robot is updated based on the position compensation value, and the acceleration of the robot is updated based on the nonlinear friction compensation value. This application embodiment uses a mechanical sensor to collect the deformation of the robot arm, determine the position compensation value and nonlinear friction compensation value, and can capture these abnormal situations in real time and accurately during the extension and retraction process of the robot arm, providing a reliable basis for subsequent adjustments, thereby effectively improving the operating accuracy and motion performance of the robot arm.

[0050] In this embodiment of the application, during the design and manufacturing stage of the robotic arm, the installation positions of the mechanical sensors are precisely determined based on the structural characteristics and telescopic range of the robotic arm. These sensors are selected at key stress points in the telescopic mechanism, such as the connection points and joints of the telescopic rods, to ensure comprehensive and accurate capture of the deformation generated during the telescopic process. Simultaneously, high-precision, high-sensitivity mechanical sensors with good anti-interference capabilities are selected to meet the requirements for real-time and accurate data acquisition. After installation, the sensors are calibrated to eliminate initial errors and ensure the accuracy of the collected data.

[0051] In this embodiment, after obtaining the deformation data, a lightweight model is used for calculation and analysis. For calculating the position compensation value, the lightweight model, based on the linear or nonlinear relationship between the deformation and the extension / retraction position of the manipulator, converts the deformation data into position deviation values ​​using a specific algorithm (e.g., least squares curve fitting), thereby obtaining the position compensation value. For calculating the nonlinear friction compensation value, considering the complex relationship between friction and various factors such as deformation, velocity, and acceleration, the lightweight model can employ a friction model established based on experimental data. This model combines real-time collected deformation data and the manipulator's motion state information (such as velocity and acceleration) to derive the nonlinear friction compensation value through numerical calculation methods.

[0052] In this embodiment, the calculated position compensation value is fed back to the motion control system of the robot arm. The motion control system adjusts the position setting of the robot arm's end effector in real time based on this compensation value. By controlling the movement of the telescopic mechanism, the robot arm's end effector can correct deviations and accurately reach the target position. Simultaneously, the nonlinear friction compensation value is applied to acceleration control. In this embodiment, the acceleration command of the robot arm can be adjusted according to the compensation value to counteract the influence of nonlinear friction on the motion, enabling the robot arm to maintain smooth and accurate acceleration changes during extension and retraction, thus improving the smoothness and response speed of the motion.

[0053] In some embodiments, after the robotic arm completes one extension and retraction motion, the positional accuracy of the robotic arm's end effector is measured using a high-precision measuring device, while simultaneously monitoring the changes in acceleration during the movement. The measurement results are compared and analyzed with preset target values ​​to evaluate the effectiveness of position compensation and nonlinear friction compensation. If the compensation effect is found to be unsatisfactory, the lightweight model is adjusted and optimized, and data acquisition, calculation, and application updates are repeated until a satisfactory compensation effect is achieved. Through continuous iterative optimization, it is ensured that the robotic arm can maintain high-precision position control and stable motion performance throughout the extension and retraction process.

[0054] Based on the above embodiments, as an optional embodiment, the method further includes: The edge device, in response to disconnecting from the cloud, collects the current value and acceleration of the robot arm's motor; The actual torque of the motor is determined based on the current value, and the theoretical torque of the motor is determined through the updated dynamic model. Based on the actual torque, theoretical torque, and acceleration of the motor, the load mass is estimated using the disturbance observer algorithm. The maximum allowable acceleration for the robot's operation is determined based on the load mass and the robot's arm length.

[0055] In this embodiment, the motor's current sensor and accelerometer collect data in real time after the motor starts running. The motor current data reflects the actual output torque τ of the motor. measured The accelerometer records the robotic arm's acceleration 'a'. Simultaneously, based on the updated dynamic model, the theoretical torque 'τ' that the motor should output under the current control command is calculated. model .

[0056] By perturbating the observer, using the formula P=(τ) measured -τ model ) / a, which estimates the current load quality P in real time. The estimated load quality is the key basis for subsequent adaptive control and can quickly and accurately capture the dynamic changes of the load.

[0057] In this embodiment of the application, based on the real-time estimated current load P and the arm length L of the robotic arm, the system uses a function a max = f(P, L) calculates the maximum permissible acceleration. This calculation aims to prevent excessive acceleration during the robotic arm's acceleration process, thus ensuring smooth motion. At the trajectory planning level, the system introduces the concept of a virtual force field. When the robotic arm's end effector approaches a preset safety boundary, the virtual force field generates a repulsive force. This repulsive force corrects the robotic arm's trajectory, guiding it away from the boundary. Simultaneously, constraint parameters are updated in real-time to the motion planner. The motion planner dynamically adjusts the trajectory planning based on these parameters, ensuring that the robotic arm's entire trajectory remains within a stable workspace, avoiding collisions with the surrounding environment and guaranteeing operational safety and reliability.

[0058] Based on the above embodiments, as an optional embodiment, the method of estimating load mass using the disturbance observer algorithm further includes: Determine the difference between the estimated load quality and the load quality corresponding to the current load mode; If the difference exceeds a preset threshold, the load mode is switched to a load mode where the difference between the estimated load quality and the load mode is less than the preset threshold. The robotic arm is driven according to the operating parameters corresponding to the switched load mode.

[0059] To efficiently handle different load conditions, the system in this application embodiment presets five load modes: no load, light load, medium load, heavy load, and overload. Each mode corresponds to a dedicated set of dynamic parameters, which cover the motion characteristics of the robotic arm under different loads, such as inertia and friction.

[0060] When the load complexity exceeds a preset threshold (±10%), a mode switching mechanism is triggered. During the switching process, this embodiment of the application adopts a dual-buffer technology. One buffer stores the operating parameters of the current load mode, and the other buffer preloads the operating parameters of the target load mode. At the moment of switching, the two buffers transition smoothly to ensure the continuity of control commands and avoid robot arm shaking or loss of control caused by mode switching.

[0061] Based on the above embodiments, as an optional embodiment, the edge device drives the robotic arm according to the control command, and further includes: The edge device collects the working condition information of the robotic arm; it judges the difference between the working condition information and the predicted working condition information. If the difference is greater than a threshold, it uploads the working condition information and operating parameters to the cloud. The cloud platform updates the drying strategy model based on the uploaded working condition information and operating parameters; The updated drying strategy model is lightweighted to obtain an updated lightweight model, and the updated lightweight model is then sent to the edge device.

[0062] This solution aims to dynamically optimize the drying strategy of robotic arms through the collaborative work of edge devices and the cloud, thereby improving drying efficiency and quality.

[0063] The robot arm is equipped with a variety of high-precision sensors that can collect real-time information on its working conditions, including its motion trajectory, speed, load, and key parameters such as ambient light intensity, temperature, and humidity. This provides a comprehensive and accurate data foundation for subsequent analysis.

[0064] After acquiring the operating condition information, the edge device compares and analyzes it with pre-predicted operating condition information based on historical data and machine learning algorithms. If the difference exceeds a set threshold, it indicates a significant deviation between the actual operating conditions and expectations, potentially affecting the drying effect. In this case, the edge device quickly packages the current operating condition information and the robotic arm's operating parameters, and uploads them to the cloud server via a stable and reliable communication network. After receiving the uploaded data, the cloud server leverages its powerful computing capabilities and abundant data resources to update and optimize the drying strategy model using advanced machine learning algorithms. The new model can better adapt to changes in actual working conditions, improving the accuracy and practicality of the drying strategy.

[0065] Considering the limited computing resources of edge devices, the cloud performs lightweight processing on the updated drying strategy model. Through model compression and quantization techniques, the model's size and computational complexity are significantly reduced while maintaining performance. Finally, the processed, updated, lightweight model is distributed to the edge devices. After loading the new model, the edge devices can provide the robotic arm with a more scientific and reasonable drying strategy based on real-time operating information, achieving efficient and intelligent drying operations.

[0066] Based on the above embodiments, as an optional embodiment, the control command is sent to the edge device, including: The cloud sends the control commands to the network layer, and the network layer sends the control commands to the edge device through Time-Sensitive Networking (TSN).

[0067] The network layer acts as a bridge in the command transmission process, employing Time-Sensitive Networking (TSN) technology. TSN offers advantages such as high-precision time synchronization, low latency, and deterministic transmission, ensuring the stable, orderly, and rapid transmission of control commands within the network. Through TSN, the network layer can accurately send control commands from the cloud to edge devices according to strict time scheduling.

[0068] Upon receiving control commands, edge devices can respond quickly and execute corresponding operations, enabling real-time interaction and collaborative work with the cloud. This solution effectively solves problems such as latency and packet loss that may occur in the transmission of control commands in traditional networks, improving the overall control accuracy and response speed of the system.

[0069] This application discloses a control method for a robotic arm, which adopts a five-dimensional collaborative architecture of cloud, network, terminal, edge, and chip, integrating edge computing, IoT, and intelligent control technologies to construct a closed loop from global optimization to local real-time control. This meets the stringent requirements of predictive maintenance and remote operation and maintenance, and solves the pain points of traditional drying equipment such as low space utilization, cumbersome operation, and poor weather adaptability. It realizes the leap from passive execution to active intelligence of the telescopic robotic arm, providing an efficient, safe, and energy-saving drying solution for home scenarios.

[0070] Cloud layer As the core of device access and management, it is responsible for global optimization and AI training. Based on weather data (humidity, probability of rainfall) and user habits, it trains a drying strategy model to achieve dynamic task planning.

[0071] A digital twin platform is constructed, employing a virtual operating model that is mapped 1:1 to the physical robotic arm. This model integrates physical, kinematic, and dynamic properties to simulate energy consumption and extension / retraction trajectories, and optimizes control parameters.

[0072] User management, task scheduling, and data analysis functions are separated into independent services and communicated via RESTful APIs to ensure high availability and scalability of the system.

[0073] It receives sensor data uploaded by edge devices, generates globally optimal instructions (such as retracting the robotic arm before a rainstorm), and sends them to the edge device nodes for execution.

[0074] To address the issue of progressive fault prediction lag in telescopic manipulators under dynamic load and length variations, a predictive maintenance system based on "cloud-edge collaboration + digital twin" is constructed by combining technologies such as real-time analysis of edge devices, dynamic model updates, and multi-source data fusion. This system is then optimized through an edge device-cloud collaborative training mechanism.

[0075] Network layer It adopts 5G / WiFi dual-mode communication, supporting 5G remote control (latency <20ms) and local WiFi direct connection. Time-Sensitive Networking (TSN) ensures priority transmission of control commands with jitter <1ms.

[0076] Edge devices Real-time control and local decision-making: An industrial-grade gateway based on ARM architecture runs a lightweight ROS system to process sensor data and drive the robotic arm. The edge device also has a local fault tolerance mechanism that automatically switches to local mode when the network is interrupted and performs emergency operations according to preset rules (such as automatically retracting clothes in windy conditions).

[0077] Dynamic parameter updates: Dynamic parameters are updated at 1ms intervals via lightweight model solvers deployed on edge devices (such as FPGA-based parallel computing units), replacing traditional fixed-parameter models. A coupled dynamic model of arm length (L), load (P), and velocity (v) is established. M(L,P)· +C(L,P,v)· + G(L,P) + F friction (L,v) = τ Where M(L,P) is the time-varying inertia matrix, C(L,P,v) is the Coriolis force term, G(L,P) is the gravity term, and F... friction For nonlinear friction, it calculates mass distribution and inertia changes in real time, solving the accuracy loss problem caused by model simplification in telescopic manipulators under high-speed, high-acceleration or variable load scenarios.

[0078] Active compensation for flexible deformation: Fiber grating sensors are embedded in key nodes of the telescopic boom to monitor the deformation (δ) in real time under cantilever conditions; a deformation prediction model is established based on Euler-Bernoulli beam theory. δ= (P×L 3 ) / (3E×I(x)) Where E is the elastic modulus, I(x) is the moment of inertia of the cross section, and the sensor data is filtered by the edge device and then input into the compensation algorithm. The correction frequency is synchronized with the control cycle (≥1kHz).

[0079] Nonlinear friction dynamic compensation: The LuGre model is used to describe the friction nonlinearity: F friction =σ0z +σ1 +σ²v, where z is the bristle deformation. Model parameters (σ0, σ1, σ2) are identified online using the recursive least squares (RLS) method to adapt to different arm length-load conditions. High-frequency excitation signals are injected during the motion start-up / speed change phases to stimulate frictional characteristics and accelerate parameter convergence. The feedforward stage outputs a friction compensation torque τ based on the identified parameters. ff =-F friction The feedback loop employs adaptive sliding mode control (ASMC) to suppress residual errors, and the sliding surface is designed as s= +λe (e is the tracking error). Feedforward compensation dominates (over 80%), while feedback control only handles unmodeled dynamics, reducing energy consumption.

[0080] Variable load adaptive control: Based on motor current and accelerometer data, the load mass is estimated using a disturbance observer: P est =(τ measured -τ model ) / a Five load modes are preset (no load / light load / medium load / heavy load / overload), and the corresponding dynamic parameter set is called when switching. Mode switching is triggered when the load change exceeds a threshold (±10%), and a double buffer is used during the switching process to ensure control continuity. The maximum allowable acceleration is calculated based on the current load (P) and arm length (L). a max = f (P,L) is used to avoid excessive vibration; a virtual force field is introduced into the trajectory planning layer to generate a repulsive force when the end point approaches the safety boundary. The constraint parameters are updated to the motion planner in real time to ensure that the trajectory always stays within the stable working space.

[0081] Terminal layer It employs a lead screw-driven multi-stage telescopic arm, achieving stepless telescopic extension from 0.8 to 2.5 meters, with a single pole bearing a load of 30 kg, and a dual horizontal reciprocating motion mechanism. It integrates a rotatable clothesline (180° adjustable) and flexible grippers to accommodate different loads such as clothing and bedding. Additionally, it is equipped with rain, light, and temperature / humidity sensors to monitor weather changes in real time, with a response time of less than 10 seconds.

[0082] Chip layer It employs a heterogeneous computing chip with an integrated NPU (Neural Processing Unit) to run lightweight AI models (such as clothing recognition algorithms) locally, with inference latency <50ms. It also features a hardware encryption module to ensure end-to-end encryption of control commands and user data.

[0083] The flow of the robotic arm control method provided in this application embodiment includes: S1. Environmental Sensing and Data Upload Sensors at the terminal layer collect environmental data (such as rainfall signals) and upload them to edge devices via 5G / WiFi. Edge devices preprocess data (such as filtering and noise reduction), transmit critical information (such as emergency rain warnings) directly to the cloud, and store non-critical data locally.

[0084] S2 Cloud Policy Generation and Distribution The large model analyzes historical data and weather forecasts to generate optimization strategies, such as: unfolding clothes for drying before 14:00 and taking them in to protect them from rain at 16:00.

[0085] S3 Edge Device Decision Making and Robot Control Edge devices receive instructions from the cloud to drive the robotic arm to extend / remove / rotate; The local AI model identifies clothing type in real time and dynamically adjusts the clamping force. Users send commands via a mobile app, which are then distributed from the cloud to edge devices for execution, with an end-to-end latency of <100ms. The edge chip recognizes voice commands (such as "collect the clothes") and directly drives the robotic arm locally, without requiring cloud interaction. The edge device makes autonomous decisions based on sensor data (such as automatically starting the drying process when the light intensity is <100 lux).

[0086] This solution utilizes a five-dimensional collaborative architecture encompassing cloud, network, edge, and chip, with each level working closely together in a clearly defined sequence to complete a series of operations from environmental perception, data processing, decision generation to robotic arm control, providing an efficient, safe, and energy-saving clothes drying solution for home environments. The complete execution flow is as follows: first step: Various sensors at the terminal layer (rain sensor, photosensitive sensor, temperature and humidity sensor) collect environmental information in real time. For example, the rain sensor detects whether there is rain, the photosensitive sensor detects the light intensity, and the temperature and humidity sensor measures the temperature and humidity of the environment.

[0087] Users can send operation commands via a mobile app, or the edge chip can recognize voice commands (such as "collect the clothes").

[0088] The terminal layer uploads environmental information and operation instructions (APP instructions, voice instructions) to the edge device via a 5G / Wi-Fi communication network. Among them, critical information (such as emergency rain warnings) is pre-processed by the edge device and then directly transmitted to the cloud, while non-critical data is stored locally on the edge device.

[0089] Step Two: Data preprocessing: Edge devices perform preprocessing operations such as filtering and noise reduction on the data uploaded from the terminal layer to extract key information. For example, temperature and humidity data collected by sensors are smoothed to remove noise interference.

[0090] Local Response and Decision-Making: Edge devices respond and make decisions locally based on pre-processed data and preset rules stored locally. For example, they automatically start drying when the light intensity is <100 lux; when an emergency rain alarm is triggered, they perform emergency operations according to preset rules (such as automatically retracting clothes when there is wind).

[0091] Robotic arm control preparation: Receive cloud instructions (if any) and local decision results to prepare for driving the robotic arm.

[0092] This step requires determining the preparation information for controlling the robotic arm's movements, transmitting critical information (such as emergency rain warnings) directly to the cloud, and storing non-critical data locally. Simultaneously, processed data and local decision-making results are uploaded to the cloud.

[0093] Step 3: Data storage and analysis: The cloud stores and analyzes the data reported by edge devices to uncover potential patterns and trends in the data.

[0094] Global optimization and AI training: Based on environmental data and user habits, a large model is used for global optimization and AI training to train the drying strategy model and achieve dynamic task planning.

[0095] Digital Twin Platform Construction and Optimization: A digital twin platform is constructed, which adopts a virtual operating model that is mapped 1:1 to the physical robot arm, integrates physical, kinematic and dynamic properties, simulates energy consumption and extension trajectory, and optimizes control parameters.

[0096] Functional decomposition and service communication: User management, task scheduling, data analysis and other functions are decomposed into independent services and communicated through RESTful APIs to ensure high availability and scalability of the system.

[0097] Global optimal instruction generation: After analyzing and processing the data, generate global optimal instructions (such as retracting the robotic arm before a rainstorm) and send them to the edge devices. At the same time, feed back the analysis results and optimization strategies to the edge devices and the terminal layer to achieve continuous system optimization and intelligent decision-making.

[0098] Step 4: Edge Devices Edge devices receive instructions from the cloud, drive an industrial-grade gateway based on ARM architecture to run a lightweight ROS system, process environmental data, and drive a ball screw to drive a multi-stage sleeve arm to achieve telescopic / rotational movements. They also integrate a rotatable clothes drying rack and flexible grippers to complete the corresponding operations.

[0099] The lightweight model for edge devices can also identify clothing types in real time and dynamically adjust the clamping force.

[0100] When the network is interrupted, it automatically switches to local mode, and the edge device layer performs emergency operations according to preset rules. The edge device layer makes autonomous decisions based on environmental data (such as automatically starting drying when the light intensity is <100 lux).

[0101] Dynamic parameter updates and compensation: Lightweight model solvers deployed on edge devices (such as FPGA-based parallel computing units) update dynamic parameters at 1ms intervals, establishing a coupled dynamic model of arm length (L), load (P), and velocity (v), and calculating mass distribution and inertia changes in real time. Fiber grating sensors are embedded at key nodes of the telescopic arm to monitor deformation under cantilever conditions in real time. A deformation prediction model is established based on Euler-Bernoulli beam theory and active compensation is performed. The LuGre model is used to describe frictional nonlinearity, and model parameters are identified online using the recursive least squares (RLS) method for nonlinear frictional dynamic compensation. Based on motor current and accelerometer data, the load mass is estimated using a disturbance observer to achieve variable load adaptive control.

[0102] Input: Global optimal instructions from the cloud layer, real-time environmental data uploaded from the terminal layer, and some data processed by the terminal itself.

[0103] Output: Control the robotic arm to complete corresponding extension, rotation, clamping and other actions to operate the items being dried.

[0104] Compared with the prior art, the embodiments of this application have the following advantages: In terms of real-time performance and reliability, the local response latency of edge devices is <50ms, the cloud decision latency is <100ms, and the response speed in emergency scenarios (such as sudden rainfall) is improved by 90%. The dual-level reciprocating mechanism reduces wire rope wear and achieves an average mean time between failures (MTBF) of 20,000 hours (compared to approximately 8,000 hours for traditional products).

[0105] In terms of energy saving and intelligence, the AI ​​strategy dynamically adjusts the drying power based on weather forecasts, with an average daily power consumption of 0.8 kWh (70% more energy-efficient than a standalone dryer), a clothing recognition accuracy of >95%, avoiding pinch damage, and a 40% increase in user satisfaction.

[0106] In terms of scene adaptability: It can operate unmanned in all scenarios and supports multi-terminal collaboration. The weather response speed is less than 10 seconds. When going out, it can automatically deal with sudden rain and reduce the risk of clothes getting wet by 95%. It can also work with air conditioning and fresh air systems to automatically adjust the temperature and humidity when drying clothes, improving drying efficiency by 30%.

[0107] In terms of space utilization optimization: the telescopic arm covers a range of 0.8 to 2.5 meters, solving the problem of insufficient balcony space in small apartments and improving applicability by 60%.

[0108] This application provides a control system for a robotic arm, such as... Figure 2 As shown, the system may include: a terminal layer 101, an edge device 102, and a cloud 103, wherein, Terminal layer 101 is used to collect real-time environmental data and upload the environmental data to the edge device. The terminal layer includes a robotic arm and at least one environmental sensor. The robotic arm is used to dry clothes. Edge device 102 is used to report an alarm message about an environmental anomaly to the cloud in real time if an environmental anomaly is determined based on the environmental data. The alarm message includes the environmental data. Cloud 103 is used to input the environmental data and the digital twin model of the robotic arm into a pre-trained drying strategy model, obtain the control instructions output by the drying strategy model, and send the control instructions to the edge device. The control instructions are used to indicate the running time and running parameters of the robotic arm. Edge device 102 is also used to drive the robotic arm according to the control instructions.

[0109] The apparatus in this application embodiment can execute the method provided in this application embodiment, and the implementation principle is similar. The actions performed by each module in the apparatus of each embodiment of this application correspond to the steps in the method of each embodiment of this application. For detailed functional descriptions of each module of the apparatus, please refer to the descriptions in the corresponding methods shown above, which will not be repeated here.

[0110] This application provides an electronic device including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps of a robotic arm control method. Compared with related technologies, this method achieves an edge-cloud closed-loop control of the robotic arm. This solution realizes intelligent clothes drying control based on a multi-level architecture, especially excelling in obtaining control commands based on models. The terminal layer collects environmental data such as temperature, humidity, light intensity, and wind speed in real time through environmental sensors, providing accurate basis for decision-making. Edge devices quickly identify and report environmental anomalies, ensuring timely response from the cloud. The cloud inputs environmental data and a digital twin model of the robotic arm into a drying strategy model. This model, trained with a large amount of data, can deeply analyze the relationship between environmental factors and drying effects, outputting scientific and reasonable control commands. The control commands precisely indicate the robotic arm's operating time and parameters. For example, the optimal drying time can be determined based on light intensity and wind speed, and the robotic arm's rotation frequency can be adjusted based on humidity to avoid excessive sun exposure or dampness and mold growth on clothes. This model-based control method breaks through the limitations of traditional fixed rules, can dynamically adapt to complex and ever-changing environments, improve drying efficiency and clothing care quality, while reducing the cost of manual intervention, and realize intelligent, automated and personalized clothing drying.

[0111] In one alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of this application.

[0112] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0113] Bus 4002 may include a pathway for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, bus 4002 is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus.

[0114] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing computer programs and capable of being read by a computer, without limitation herein.

[0115] The memory 4003 stores computer programs that execute embodiments of this application, and its execution is controlled by the processor 4001. The processor 4001 executes the computer programs stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.

[0116] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the steps and corresponding content of the aforementioned method embodiments.

[0117] This application also provides a computer program product, including a computer program that, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments.

[0118] The terms "first," "second," "third," "fourth," "1," "2," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in a sequence other than that shown in the illustrations or text descriptions.

[0119] It should be understood that although arrows indicate various operation steps in the flowcharts of this application's embodiments, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of this application's embodiments, the implementation steps in each flowchart can be executed in other orders as required. Furthermore, some or all steps in each flowchart, based on the actual implementation scenario, may include multiple sub-steps or multiple stages. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage can also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured according to requirements, and this application's embodiments do not limit this.

[0120] The above description is only an optional implementation method for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application without departing from the technical concept of this application also fall within the protection scope of the embodiments of this application.

Claims

1. A control method for a robotic arm, characterized in that, include: The terminal layer collects real-time environmental data and uploads the environmental data to the edge device. The terminal layer includes a robotic arm and at least one environmental sensor. The robotic arm is used for drying clothes. If the edge device determines that the environment is abnormal based on the environmental data, it shall report the environmental abnormality alarm information to the cloud in real time, and the alarm information includes the environmental data. The cloud inputs the environmental data and the digital twin model of the robotic arm into a pre-trained drying strategy model, obtains the control commands output by the drying strategy model, and sends the control commands to the edge device. The control commands are used to indicate the running time and running parameters of the robotic arm. The edge device drives the robotic arm according to the control commands.

2. The method according to claim 1, characterized in that, The method further includes: In response to disconnecting from the cloud, the edge device collects dynamic data of the robotic arm under different motion states, including velocity, acceleration, and joint torque. The dynamic data is input into the dynamic model to obtain the updated dynamic model; Based on the locally deployed lightweight model, the updated dynamic model is used to determine the mass distribution and inertia changes of the robot arm.

3. The method according to claim 1, characterized in that, The method further includes: In response to disconnecting from the cloud, the edge device collects the deformation of the robotic arm and uses the deformation to determine the position compensation value and nonlinear friction compensation value of the robotic arm during the extension and retraction process based on the locally deployed lightweight model. The end effector position of the robot is updated based on the position compensation value, and the acceleration of the robot is updated based on the nonlinear friction compensation value.

4. The method according to claim 2, characterized in that, The method further includes: The edge device, in response to disconnecting from the cloud, collects the current value and acceleration of the robot arm's motor; The actual torque of the motor is determined based on the current value, and the theoretical torque of the motor is determined through the updated dynamic model. Based on the actual torque, theoretical torque, and acceleration of the motor, the load mass is estimated using the disturbance observer algorithm. The maximum allowable acceleration for the robot's operation is determined based on the load mass and the robot's arm length.

5. The method according to claim 4, characterized in that, The method of estimating load quality using the perturbation observer algorithm also includes: Determine the difference between the estimated load quality and the load quality corresponding to the current load mode; If the difference exceeds a preset threshold, the load mode is switched to a load mode where the difference between the estimated load quality and the load mode is less than the preset threshold. The robotic arm is driven according to the operating parameters corresponding to the switched load mode.

6. The method according to claim 2, characterized in that, The lightweight model is the model obtained by the cloud-based process of lightweighting the drying strategy model.

7. The method according to claim 2 or 6, characterized in that, The edge device drives the robotic arm according to the control command, and further includes: The edge device collects the working condition information of the robotic arm; it judges the difference between the working condition information and the predicted working condition information. If the difference is greater than a threshold, it uploads the working condition information and operating parameters to the cloud. The cloud platform updates the drying strategy model based on the uploaded working condition information and operating parameters; The updated drying strategy model is lightweighted to obtain an updated lightweight model, and the updated lightweight model is then sent to the edge device.

8. The method according to claim 1, characterized in that, Sending the control command to the edge device includes: The cloud sends the control commands to the network layer, and the network layer sends the control commands to the edge device through Time-Sensitive Networking (TSN).

9. A control system for a robotic arm, characterized in that, include: The terminal layer is used to collect real-time environmental data and upload the environmental data to the edge device. The terminal layer includes a robotic arm and at least one environmental sensor. The robotic arm is used to dry clothes. An edge device is used to report an environmental anomaly alarm to the cloud in real time if an environmental anomaly is determined based on the environmental data, wherein the alarm information includes the environmental data. In the cloud, the environmental data and the digital twin model of the robotic arm are input into a pre-trained drying strategy model to obtain control commands output by the drying strategy model. The control commands are then sent to the edge device to indicate the running time and running parameters of the robotic arm. The edge device is also used to drive the robotic arm according to the control instructions.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the control method of the robotic arm according to any one of claims 1-8.

Images