A multi-degree-of-freedom manipulator control method

Abstract

The present invention relates to the field of device control, and in particular to a control method for a multi-degree-of-freedom manipulator. The method includes: monitoring in real time the working environment where the multi-degree-of-freedom manipulator is located to obtain monitoring data; receiving an operation instruction of the multi-degree-of-freedom manipulator, and using Gaussian process regression to perform trajectory prediction based on the monitoring data and an operation target in the operation instruction to obtain operation path information of the multi-degree-of-freedom manipulator; adopting a fuzzy chance-constrained control model, and performing local solution on the operation paths corresponding to each motor according to the control information and constraint conditions of each motor to obtain a target control strategy for each motor; and executing the target control strategy to enable the robotic arm to perform optimized trajectory movement under the control cooperation of each motor. It can realize the intelligent control of the multi-degree-of-freedom manipulator, improve the environmental perception ability of the multi-degree-of-freedom manipulator, improve the accuracy of trajectory prediction, and improve the control precision of the multi-degree-of-freedom manipulator.

Description

Technical Field

[0001] The present invention relates to the technical field of device control, and in particular to a control method for a multi-degree-of-freedom manipulator. Background Art

[0002] With the continuous progress of technology, manipulators are developing towards the directions of intelligence, flexibility, miniaturization, and high precision.

[0003] A multi-degree-of-freedom manipulator is an automated device with multiple degrees of freedom developed on the basis of a traditional manipulator, and can complete diverse tasks in a more complex spatial environment. The degree of freedom refers to the number of independent movement directions of an object in space. For a multi-degree-of-freedom manipulator, the movement of each joint corresponds to a degree of freedom. For example, a rotary joint can rotate around a certain axis, and a translational joint can move linearly along a certain direction.

[0004] In related technologies, first, multi-degree-of-freedom manipulators often need to operate in complex and dynamically changing environments, such as the presence of numerous obstacles and unstable lighting conditions on industrial production lines. Traditional control methods may not be able to comprehensively and accurately perceive environmental information. Second, when performing operation tasks, it is necessary to accurately predict the movement trajectory of the manipulator to ensure that it can accurately reach the target position. However, due to the movement of the manipulator being affected by various factors, such as load changes and joint friction, traditional trajectory planning methods may not be able to accurately predict the actual movement trajectory of the manipulator. In addition, each joint of a multi-degree-of-freedom manipulator is usually driven by a motor, and the control accuracy of the motor directly affects the movement accuracy and stability of the manipulator. Traditional control methods may not be able to precisely control the movement of each motor, resulting in a large movement error of the manipulator.

[0005] Therefore, there is an urgent need to design a technical solution for realizing the intelligent control of a multi-degree-of-freedom manipulator and solving at least one of the above technical problems. Summary of the Invention

[0006] In view of the technical problems existing in the prior art, the present invention provides a control method for a multi-degree-of-freedom manipulator, which is used to realize the intelligent control of the multi-degree-of-freedom manipulator, improve the environmental perception ability of the multi-degree-of-freedom manipulator, improve the accuracy of trajectory prediction, and improve the control accuracy of the multi-degree-of-freedom manipulator.

[0007] In a first aspect, an embodiment of the present application provides a control method for a multi-degree-of-freedom manipulator, and this method is applied to the control scenario of the multi-degree-of-freedom manipulator; this method at least includes:

[0008] Real-time monitor the working environment where the multi-degree-of-freedom manipulator is located to obtain monitoring data;

[0009] Receive the operation instructions of the multi-degree-of-freedom manipulator, and use Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instructions to obtain the operation path information of the multi-degree-of-freedom manipulator; the operation path information includes: control information for each motor in the drive system and constraint conditions;

[0010] Adopt a fuzzy chance-constrained control model, and perform local solution on the operation paths corresponding to each motor according to the control information of each motor and the constraint conditions to obtain the target control strategy for each motor; in the fuzzy chance-constrained control model, construct fuzzy variables based on the control target and the constraint conditions, introduce a confidence level using the credibility theory to quantitatively calculate the feasibility of the operation paths of each motor, and evaluate the local control risks of each motor, and comprehensively obtain the target control strategy;

[0011] Execute the target control strategy to make the robotic arm perform optimized trajectory motion under the control cooperation of each motor.

[0012] In a second aspect, an embodiment of the present application provides a multi-degree-of-freedom manipulator control system, and the system is applied to a control scenario of a multi-degree-of-freedom manipulator; the system at least includes: a robotic arm, a drive system, a control module, and a data acquisition module; wherein,

[0013] The data acquisition module is used to monitor the working environment where the multi-degree-of-freedom manipulator is located in real time to obtain monitoring data:

[0014] The control module is used to receive the operation instructions of the multi-degree-of-freedom manipulator, and use Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instructions to obtain the operation path information of the multi-degree-of-freedom manipulator; the operation path information includes: control information for each motor in the drive system and constraint conditions; adopt a fuzzy chance-constrained control model, and perform local solution on the operation paths corresponding to each motor according to the control information of each motor and the constraint conditions to obtain the target control strategy for each motor; in the fuzzy chance-constrained control model, construct fuzzy variables based on the control target and the constraint conditions, introduce a confidence level using the credibility theory to quantitatively calculate the feasibility of the operation paths of each motor, and evaluate the local control risks of each motor, and comprehensively obtain the target control strategy;

[0015] The drive system is used to execute the target control strategy to make the robotic arm perform optimized trajectory motion under the control cooperation of each motor.

[0016] In a third aspect, an embodiment of the present application provides an electronic device, and the electronic device includes a memory for storing computer software programs;

[0017] A processor for reading and executing the computer software program to implement the multi-degree-of-freedom manipulator control method in the first aspect.

[0018] In a fourth aspect, a computer-readable storage medium is provided, which includes instructions that, when the instructions are run on a computer, cause the computer to execute the multi-degree-of-freedom manipulator control method in the first aspect.

[0019] The beneficial effects of the present invention are as follows: A multi-degree-of-freedom manipulator control method is provided. This technical solution is applied to the control scenario of a multi-degree-of-freedom manipulator. In this solution, the working environment of the multi-degree-of-freedom manipulator is first monitored in real time to obtain monitoring data. Then, the operation instructions of the multi-degree-of-freedom manipulator are received, and Gaussian process regression is used to perform trajectory prediction based on the monitoring data and the operation target in the operation instructions to obtain the operation path information of the multi-degree-of-freedom manipulator; the operation path information includes: control information for each motor in the drive system and constraint conditions. Next, a fuzzy chance-constrained control model is adopted to locally solve the operation paths corresponding to each motor according to the control information of each motor and the constraint conditions to obtain the target control strategy for each motor; in the fuzzy chance-constrained control model, fuzzy variables are constructed based on the control target and the constraint conditions, and the credibility theory is used to introduce a confidence level to quantitatively calculate the feasibility of the operation paths of each motor and evaluate the local control risks of each motor, and the target control strategy is comprehensively obtained. Finally, the target control strategy is executed to enable the robotic arm to perform an optimized trajectory movement under the control cooperation of each motor. In the embodiments of the present application, intelligent control of the multi-degree-of-freedom manipulator can be achieved, the environmental perception ability of the multi-degree-of-freedom manipulator can be improved, the accuracy of trajectory prediction can be improved, and the control accuracy of the multi-degree-of-freedom manipulator can be improved. Description of the Drawings

[0020] Figure 1 is a schematic flowchart of a multi-degree-of-freedom manipulator control method according to an embodiment of the present application;

[0021] Figure 2 is a schematic structural diagram of a multi-degree-of-freedom manipulator control system according to an embodiment of the present application;

[0022] Figure 3 is a schematic structural diagram of a medium device according to an embodiment of the present application. Detailed Embodiments

[0023] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0024] The embodiments of the present application provide a multi-degree-of-freedom manipulator control method. In this technical solution, first of all, multi-degree-of-freedom manipulators often need to operate in complex and dynamically changing environments, such as the presence of many obstacles and unstable lighting conditions on industrial production lines. In the related art, it may not be possible to comprehensively and accurately perceive environmental information. However, the embodiments of the present application obtain monitoring data by real-time monitoring of the working environment, and use various sensors (such as vision sensors, lidar, etc.) to perceive the environment from different dimensions, and can timely detect information such as obstacles and changes in the position of objects in the environment, so that the manipulator can better adapt to the complex environment and avoid collisions with obstacles. In addition, the working environment may change at any time, such as changes in the position and attitude of objects, the emergence of new obstacles, etc. The embodiments of the present application can continuously adjust the operation path of the manipulator according to the real-time monitoring data, so that the manipulator can quickly respond to environmental changes and ensure the smooth completion of the task.

[0025] Secondly, when performing an operation task, it is necessary to accurately predict the movement trajectory of the manipulator to ensure that it can accurately reach the target position. However, since the movement of the manipulator is affected by various factors, such as load changes, joint friction, etc., the trajectory planning methods in the related art may not be able to accurately predict the actual movement trajectory of the manipulator. The embodiments of the present application use Gaussian process regression to perform trajectory prediction in combination with monitoring data and operation targets, which can fully consider various uncertainty factors and improve the accuracy of trajectory prediction, thereby planning a more reasonable operation path. In addition, different operation tasks may require different movement paths, and traditional methods may be difficult to flexibly adjust the path according to the actual situation. The embodiments of the present application dynamically generate operation path information through the real-time acquired monitoring data and operation targets, so that the manipulator can flexibly plan the path according to the specific task requirements, improving work efficiency and the quality of task completion.

[0026] Thirdly, each joint of the multi-degree-of-freedom manipulator is usually driven by a motor, and the control accuracy of the motor directly affects the motion accuracy and stability of the manipulator. In the related art, it may not be possible to accurately control the motion of each motor, resulting in a large motion error of the manipulator. In the embodiment of the present application, a fuzzy chance-constrained control model is adopted to locally solve the operation paths of the respective motors, taking into account the fuzziness and uncertainty of the control objectives and constraints, and can more accurately determine the control parameters of each motor, such as rotational speed, torque, etc., thereby improving the control accuracy of the motor. In addition, during the actual operation process, the manipulator may face various risks, such as motor overload, joint wear, etc. In the related art, there is often a lack of effective evaluation and countermeasures for these risks. The embodiment of the present application can introduce the credibility theory, quantitatively calculate the feasibility of the operation paths of the respective motors, and evaluate the local control risks, can discover potential risk factors in advance, and take corresponding measures for prevention and control, improving the reliability and safety of the manipulator operation.

[0027] In summary, the embodiment of the present application can achieve intelligent control of the multi-degree-of-freedom manipulator, improve the environmental perception ability of the multi-degree-of-freedom manipulator, improve the accuracy of trajectory prediction, and improve the control accuracy of the multi-degree-of-freedom manipulator.

[0028] The multi-degree-of-freedom manipulator control solution provided by the embodiment of the present application can also be executed by an electronic device, and the electronic device can be a server, a server cluster, or a cloud server. The electronic device can also be a terminal device such as a mobile phone, a computer, a tablet computer, a wearable device, or a dedicated device (such as a dedicated terminal device with a multi-degree-of-freedom manipulator control system, etc.). These electronic devices can also be equipped with the chips introduced in the above embodiments. Or, these electronic devices can also install a service program for executing the multi-degree-of-freedom manipulator control solution.

[0029] Figure 1 It is a schematic diagram of a multi-degree-of-freedom manipulator control method provided by an embodiment of the present application. This method is applied to the control scenario of a multi-degree-of-freedom manipulator. As Figure 1 shown, this method includes:

[0030] 101, monitoring the working environment where the multi-degree-of-freedom manipulator is located in real time to obtain monitoring data;

[0031] 102, receiving the operation instruction of the multi-degree-of-freedom manipulator, and using Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instruction to obtain the operation path information of the multi-degree-of-freedom manipulator;

[0032] 103, adopting a fuzzy chance-constrained control model, and locally solving the operation paths corresponding to the respective motors according to the control information and constraint conditions of the respective motors to obtain the target control strategies of the respective motors;

[0033] 104. Execute the target control strategy to enable the robotic arm to perform an optimized trajectory movement under the control cooperation of each motor.

[0034] In the embodiment of the present application, the operation path information includes: control information for each motor in the drive system and constraint conditions. The motor control information in the operation path information covers parameter settings such as motor speed and torque, directly determining the operating state of the motor and precisely controlling the movement of each joint of the robotic arm. The constraint conditions are the key to ensuring the safe and stable operation of the system, including the upper limit of motor speed, maximum acceleration, position limit, etc., preventing the motor from overloading or the robotic arm from exceeding the working range. At the same time, based on the working environment and task requirements, the movement of the robotic arm is regulated to ensure that the robotic arm can efficiently complete the task on the premise of meeting the actual application requirements and avoid failures caused by improper control or out-of-bounds operation.

[0035] In the embodiment of the present application, in the fuzzy chance-constrained control model, fuzzy variables are constructed based on the control objectives and constraint conditions, and the credibility theory is used to introduce a confidence level to quantitatively calculate the feasibility of the operation path of each motor and evaluate the local control risk of each motor, and the target control strategy is obtained comprehensively.

[0036] In the fuzzy chance-constrained control model, by constructing the control objectives and constraint conditions as fuzzy variables and combining the credibility theory to introduce a confidence level, a scientific evaluation of the motor operation path and the formulation of the target control strategy are realized. The specific process is as follows: The control objectives (such as desired speed, position) and constraint conditions (maximum speed, maximum acceleration, etc.) of the motor are transformed into fuzzy variables. Taking the motor speed as an example, according to its numerical range and actual application scenarios, it is divided into fuzzy sets such as "low speed", "medium speed", "high speed", etc. By setting the membership function to describe the degree to which each value belongs to the corresponding fuzzy set, such as using a triangular membership function, the membership degree of the speed within a certain interval changes linearly, so as to map the precise value to the fuzzy concept and solve the uncertainty and fuzziness problems in the control process.

[0037] The credibility theory is used to introduce the confidence level, which represents the degree of completion of the motor operation path with respect to the constraint conditions. For important constraint conditions, such as the maximum motor speed limit, a higher confidence level is set to ensure the safety of the motor operation. For relatively less important conditions, a lower confidence level can be set. The setting of the confidence level provides a standard for quantitative evaluation, making the formulation of the control strategy more in line with the actual requirements and risk tolerance. The calculation of the feasibility of the operation path and the risk assessment are based on the confidence level and the membership function, and each predicted operation position in the manipulator movement path corresponding to the control information is analyzed. The feasibility of the operation path corresponding to each fuzzy set is calculated. For example, if the membership degree of the motor speed in the "medium speed" fuzzy set is high and the confidence level is high at a certain predicted position, then the feasibility of this position in terms of "medium speed" is high. At the same time, the local control risk is evaluated. If the motor speed is close to the maximum speed and the confidence level of the maximum speed constraint is high, it means that the local control risk at this position is high and the constraint condition may be violated.

[0038] Based on the comprehensive evaluation results of the feasibility of the operation path and the local control risk, the motor control parameters in the control information are adjusted. If it is found that the local control risk at a certain predicted position is high, such as the speed may exceed the upper limit, the motor speed is reduced; if the risk of the position deviating from the target is high, the torque is adjusted to make the manipulator approach the target. By continuously optimizing the motor control parameters, on the basis of meeting the constraint conditions, the motor operation is made to approach the expected control target, the control risk is reduced, and finally a scientific and reasonable target control strategy is formed to ensure that the manipulator can complete tasks safely and efficiently in a complex environment.

[0039] In step 101, through various sensors arranged in the working environment of the manipulator, various information in the environment is sensed in real time, including but not limited to the position of obstacles, the posture of objects, the illumination conditions of the environment, temperature and humidity, etc., providing basic data for subsequent trajectory prediction and control strategy formulation. The principle of step 101 is based on the physical characteristics of the sensors. For example, visual sensors use optical principles to capture image information, and force sensors measure the magnitude of force by sensing the action of force, etc.

[0040] Specifically, based on the spatial structure of the working environment and the corresponding business types of the working environment, determine the installation locations, quantities, and types of different types of sensors. For example, in an industrial production line, according to the layout of the production line and the operating range of the manipulator, install vision sensors to detect the position and posture of workpieces, and install force sensors to monitor the magnitude of the force when the manipulator contacts an object. Conduct an effectiveness evaluation of the information collection scope of the deployment plan, analyze whether the coverage range of the sensors meets the requirements, whether there are monitoring blind spots, etc., and optimize the deployment plan based on the evaluation results. Install different types of sensors in the working environment according to the optimized deployment plan. Collect real-time environmental information of the multi-degree-of-freedom manipulator from different dimensions through different types of sensors to obtain multi-dimensional monitoring data, and preprocess the collected data, such as filtering, noise reduction, feature extraction, etc.

[0041] Thus, it is possible to comprehensively and real-time obtain information about the manipulator's working environment, provide accurate data support for subsequent control decisions, enable the manipulator to better adapt to complex and changeable working environments, avoid collisions with obstacles, and improve the safety and reliability of work.

[0042] In step 102, Gaussian process regression is a probability-based machine learning method that assumes data follows a Gaussian distribution and measures the similarity between data points through a kernel function. In this step, use Gaussian process regression to predict the motion trajectory of the manipulator based on the monitoring data and operation objectives, taking into account the uncertainty factors in the data and being able to provide more accurate trajectory predictions. Organize and preprocess the monitoring data obtained in step 101, and at the same time clarify the operation objectives in the operation instructions and convert them into mathematical representations. Select an appropriate kernel function (such as a radial basis function, etc.) and determine the hyperparameters of the model (such as the variance of the kernel function, length scale parameter, etc.) to construct a Gaussian process regression model. Use the organized monitoring data and corresponding operation objectives as training data to train the Gaussian process regression model and adjust the parameters of the model to make it better fit the data. Use the current monitoring data as new input and use the trained model for trajectory prediction to obtain the operation path information of the manipulator, including the position, posture, etc. of the manipulator at each moment.

[0043] In this way, it is possible to fully consider the uncertainty of the working environment and the requirements of the operation objectives, provide more accurate and reliable trajectory predictions, enable the manipulator to move along the expected path, improve the accuracy and efficiency of the operation, and reduce motion errors.

[0044] In step 103, the fuzzy chance-constrained control model uses the method of fuzzy mathematics to transform the control objectives and constraints into fuzzy variables, and quantifies and calculates the feasibility of the operation path by introducing a confidence level. At the same time, the credibility theory is used to evaluate the local control risks of each motor, and these factors are comprehensively considered to solve for the optimal control strategy to ensure that the manipulator can approach the control objective as much as possible while meeting the constraint conditions.

[0045] Specifically, in step 103, according to the control objectives (such as the speed and position of the motor) and constraints (such as the maximum speed and maximum acceleration), the corresponding fuzzy sets are defined, and the membership functions of each fuzzy set are determined to transform the control objectives and constraints into fuzzy variables. According to the actual control requirements and the reliability requirements of the system, a confidence level is set for each constraint. For the operation path of each motor, combining the fuzzy variables and the confidence level, the feasibility of its operation path is calculated, and the local control risk is evaluated. Considering the feasibility of the operation path and the local control risk comprehensively, the local solution of the operation path corresponding to each motor is carried out through an optimization algorithm to obtain the target control strategies of each motor, such as control parameters such as the rotation speed and torque of the motor.

[0046] Thus, it can handle the uncertainties and fuzziness in the control process, more precisely control the movement of each motor, and improve the movement accuracy and stability of the manipulator. At the same time, by evaluating the control risks, potential problems can be discovered in advance, and corresponding measures can be taken for prevention and control to improve the reliability and safety of the system.

[0047] In step 104, the target control strategies of each motor obtained in step 103 are sent to the drive system of the manipulator, and the drive system controls the movement of each motor according to the control strategy, so that the robotic arm moves along the optimized trajectory to complete the predetermined operation task.

[0048] Specifically, the target control strategy is transformed into a drive signal through the control system and sent to each motor of the manipulator. The motor adjusts its own movement state, such as rotation speed and torque, according to the received signal to achieve the movement control of the robotic arm. At the same time, during the movement process, the movement state of the robotic arm is monitored in real time through sensors, and the monitoring data is fed back to the control system to adjust the control strategy in a timely manner to ensure the movement accuracy and stability of the robotic arm.

[0049] Thus, the robotic arm can move along the optimized trajectory, accurately complete the predetermined operation task, and improve the work efficiency and quality. At the same time, through real-time monitoring and feedback control, problems occurring during the movement process can be discovered and solved in a timely manner to ensure the stable operation of the robotic arm.

[0050] As an optional embodiment, in 101, the working environment of the multi-degree-of-freedom manipulator is monitored in real time to obtain monitoring data, including:

[0051] Based on the spatial structure of the working environment and the business type corresponding to the working environment, construct a deployment plan for different types of sensors; evaluate the effectiveness of the information collection range of the deployment plan, and optimize the deployment plan based on the evaluation results; based on the optimized deployment plan, deploy different types of sensors in the working environment; collect the real-time environment information of the multi-degree-of-freedom manipulator from different dimensions through different types of sensors to obtain multi-dimensional monitoring data.

[0052] Specifically, in 101, first, a detailed analysis of the space where the manipulator works is required, including the size, shape, layout of the space, and whether there are special areas or obstacles. For example, in a narrow workshop, the activity space of the manipulator is limited, and the arrangement of sensors needs to avoid obstacles while ensuring that the working range of the manipulator can be comprehensively monitored. Different business types have different operation requirements and environmental perception needs for the manipulator. For example, in the material handling business, it is necessary to focus on monitoring information such as the position, shape, and weight of the materials; in the welding business, in addition to monitoring the position and posture of the welded parts, it is also necessary to pay attention to information such as the temperature and arc light in the welding area. According to the spatial structure and business type, select appropriate sensor types, such as vision sensors for identifying the shape and position of objects, force sensors for measuring the force when the manipulator contacts the object, and temperature sensors for monitoring the temperature in the welding area. Then, according to the layout of the working space and the characteristics of the sensors, design a deployment plan for the sensors, determine the installation position, quantity, and orientation of the sensors, etc., to ensure that the required environmental information can be collected comprehensively and accurately.

[0053] Furthermore, evaluate the constructed deployment plan to analyze whether the information collection range of each sensor can cover the key areas in the working environment of the manipulator and whether there are monitoring blind spots. For example, for a vision sensor, evaluate whether its field of view can cover all possible positions of the objects that the manipulator needs to operate, and whether there are some areas that cannot be monitored due to occlusion or perspective problems.

[0054] Consider the accuracy and reliability of the information collected by sensors, including the precision, resolution, anti-interference ability, etc. of the sensors. For example, whether the measurement precision of the force sensor can meet the requirements of the manipulator for force control, and whether it can work stably and be free from interference in the electromagnetic environment of the workshop. According to the evaluation results, optimize the deployment plan. If monitoring blind spots are found, it may be necessary to add sensors or adjust the position and orientation of the sensors; if the precision or reliability of the sensors is insufficient, it may be necessary to select a more suitable sensor model or take anti-interference measures. Through continuous optimization, ensure that the deployment plan can effectively collect accurate and reliable environmental information.

[0055] According to the optimized deployment plan, accurately install different types of sensors in the working environment. Pay attention to the installation precision of the sensors during the installation process to ensure that they can work properly according to the design requirements. For example, the installation position of the vision sensor should ensure that its optical axis is perpendicular to the monitoring area to obtain accurate image information; the force sensor should be installed at the key parts where the manipulator contacts the object to ensure that the magnitude and direction of the force can be accurately measured. Connect the installed sensors to the control system of the manipulator to ensure that the sensors can accurately transmit the collected signals to the control system. Then, debug the sensors to check whether they can work properly and whether the collected data is accurate. For example, adjust the focal length and set the white balance of the vision sensor so that it can capture clear and accurate images; calibrate the force sensor to ensure the accuracy of its measurement results.

[0056] Finally, different types of sensors collect environmental information from different dimensions. The vision sensor collects information such as the shape, position, and posture of the object from the image dimension; the force sensor collects information about the force between the manipulator and the object from the mechanical dimension; the temperature sensor collects the temperature information of the working environment from the thermal dimension, etc. Through the collaborative work of multiple sensors, comprehensively and three-dimensionally perceive the working environment where the manipulator is located. Fuse and process the data collected by different types of sensors to obtain multi-dimensional monitoring data. Data fusion can adopt various methods, such as feature-level fusion, decision-level fusion, etc. Through data fusion, the advantages of each sensor can be fully utilized to improve the accuracy and reliability of the data. At the same time, preprocess the fused data, such as filtering, noise reduction, normalization, etc., for subsequent analysis and processing. The finally obtained multi-dimensional monitoring data will be used as the input of the manipulator control algorithm to provide a basis for trajectory prediction and control strategy formulation.

[0057] Further optionally, in the above steps, conduct an effectiveness evaluation of the information collection range of the deployment plan, including:

[0058] According to the field of view angle and / or detection distance of each sensor in the deployment plan, combined with the spatial structure of the working environment, determine the theoretical coverage range of each sensor through geometric calculation; compare the overlapping ratio between the theoretical coverage range of each sensor and the reference area to be monitored in the working environment; based on the accuracy requirements of the multi-dimensional monitoring data for different business types, analyze the spatial resolution and temporal resolution corresponding to each sensor in its respective position and the positional arrangement relationship; based on the overlapping ratio, spatial resolution, and temporal resolution, calculate the effectiveness score of the information collection range in the deployment plan.

[0059] Exemplarily, assume that in an automotive parts assembly workshop, there is a multi-degree-of-freedom manipulator responsible for grasping and assembling parts. The spatial structure of the workshop's working environment is a cuboid space with a length of 20 meters, a width of 15 meters, and a height of 5 meters, and the manipulator operates within this space. The main business types are the precise grasping and assembly of parts, with relatively high requirements for the accuracy of monitoring data, and it is necessary to accurately identify the position and posture of the parts. Three vision sensors (Camera1, Camera2, Camera3) are deployed to monitor the position and posture of the parts, each with a field of view angle of 60° and a detection distance of 10 meters. One force sensor (ForceSensor) is installed on the end effector of the manipulator to measure the grasping force. Two lidars (Lidar1, Lidar2), with a field of view angle of 360° and a detection distance of 15 meters, are used to detect obstacles in the space.

[0060] Among them, the end effector of the manipulator is installed at the end of the manipulator and is a device that directly contacts the working object and performs specific tasks. The gripper-type end effector is mainly used for grasping, clamping, and transporting objects. It can achieve the grasping of objects with different shapes and sizes through the opening and closing of the mechanical structure, and is widely used in fields such as logistics and assembly. The adsorption-type end effector includes vacuum suction cups and electromagnetic suction cups, etc. The vacuum suction cup generates negative pressure by pumping air to adsorb objects, and is suitable for grasping objects with flat and smooth surfaces, such as glass and plastic sheets; the electromagnetic suction cup uses electromagnetic force to adsorb ferromagnetic objects and is commonly used for workpiece clamping in machining.

[0061] For the vision sensor Camera1, it is installed in the upper left corner of the workshop at a height of 4 meters. According to its field of view angle and detection distance, through geometric calculation, its theoretical coverage range is the part of a cone with a vertex at the installation point, a field of view angle of 60°, and a radius of 10 meters within the workshop space. Similarly, calculate the theoretical coverage ranges of Camera2 and Camera3. The lidar Lidar1 is installed at the center of the workshop at a height of 3 meters, and its theoretical coverage range is a cylindrical space with a radius of 15 meters centered on the installation point. Lidar2 is similar.

[0062] The reference areas to be monitored in the working environment include the working area of the manipulator and the areas for storing and transporting parts. Calculate the overlapping ratio of the theoretical coverage of each vision sensor and lidar with the reference area. For example, the overlapping ratio of the theoretical coverage of Camera1 with the reference area is 70%, Camera2 is 60%, Camera3 is 50%, Lidar1 is 80%, and Lidar2 is 75%.

[0063] According to the accuracy requirements of multi-dimensional monitoring data for different business types, analyze the spatial resolution and temporal resolution of each sensor. The spatial resolution of the vision sensor can achieve an accuracy of 1 mm for an object at a distance of 5 meters under the current position and arrangement relationship, and the temporal resolution is 30 frames per second. The spatial resolution of the lidar can achieve an accuracy of 2 cm within a range of 10 meters, and the temporal resolution is 10 scans per second. The accuracy of the force sensor is 0.1 N, and the sampling frequency is 100 times per second.

[0064] Set the weights of the overlapping ratio, spatial resolution, and temporal resolution to 0.4, 0.3, and 0.3 respectively. For the vision sensor Camera1, its effectiveness score is: 0.4×70% + 0.3 (score corresponding to 1 mm accuracy) + 0.3 (score corresponding to 30 frames per second). Assuming the score corresponding to 1 mm accuracy is 80 points and the score corresponding to 30 frames per second is 70 points, then the effectiveness score of Camera1 is 0.4×0.7 + 0.3×0.8 + 0.3×0.7 = 0.28 + 0.24 + 0.21 = 0.73.

[0065] Similarly, calculate the effectiveness scores of other sensors, and then combine the scores of all sensors to obtain the total effectiveness score of the information collection range in the deployment plan.

[0066] Thus, through effectiveness evaluation, it can be found that the coverage of some sensors may be insufficient. For example, the overlapping ratio of Camera3 is relatively low, which indicates that it is necessary to adjust its position or increase the number of sensors to improve the monitoring coverage of the working area and ensure that the manipulator can obtain comprehensive environmental information during operation. Analyzing the resolution and incorporating it into the effectiveness score can ensure that the selected sensors can meet the business requirements for the accuracy of monitoring data at the current position and arrangement. If it is found that the resolution of a certain sensor does not meet the standard, such as the spatial resolution cannot accurately identify the fine features of components, a more suitable sensor can be replaced in a timely manner or its installation position can be adjusted, so as to ensure that the manipulator can accurately perform the assembly task. Calculating the effectiveness score by comprehensively considering factors such as the overlapping ratio and resolution can comprehensively evaluate the advantages and disadvantages of the deployment plan, discover potential problems in advance, avoid manipulator operation errors or failures caused by incomplete or inaccurate information collection, and improve the reliability and stability of the entire system. During the evaluation process, it can be judged whether there is a waste of sensor resources according to the effectiveness score. For example, if the coverage of some sensors overlaps excessively and the resolution is too high beyond the business requirements, the sensors can be appropriately adjusted or reduced to achieve reasonable utilization of resources and reduce costs.

[0067] Further optionally, in the above steps, the effectiveness evaluation of the information collection range of the deployment plan includes:

[0068] According to the field of view angle and / or detection distance and sensor characteristics of each sensor in the deployment plan, establish a mathematical model for each sensor; set the simulated operating conditions of the manipulator at different positions and actions in the working scenario corresponding to the working environment; integrate the mathematical model of each sensor into the environment model, and use the simulated operating conditions to simulate the perception of the environment by each sensor at different positions and postures in the environment model; based on the simulated monitoring data in the perception situation, obtain the effectiveness score of the information collection range in the deployment plan.

[0069] Exemplarily, assume that in an automated warehouse, a multi-degree-of-freedom manipulator is responsible for handling goods. The working environment of the warehouse is a space with a length of 50 meters, a width of 30 meters, and a height of 10 meters. Two types of sensors are selected, namely a laser ranging sensor and a vision sensor. For the laser ranging sensor, based on its field of view angle of 30°, detection distance of 20 meters, and measurement accuracy and other characteristics, a mathematical model is established to calculate the distance information that can be measured at different positions and angles. For example, the detection range boundaries in different directions are determined through trigonometric relationships. For the vision sensor, based on its field of view angle of 90°, detection distance of 15 meters, image resolution, and color perception characteristics, etc., a mathematical model is established. This model can calculate the imaging position and characteristics of an object in the image based on information such as the distance and angle between the object and the sensor, as well as the size and color of the object. Considering the different positions and movements of the manipulator in the warehouse, multiple simulation running conditions are set. For example, different scenarios such as the manipulator grasping goods in the corner of the warehouse, moving between shelves, and placing goods at the designated position. In each scenario, the manipulator has different postures and motion trajectories.

[0070] Furthermore, the mathematical models of each sensor are integrated into the environmental model of the warehouse. During the simulation run, according to the set simulation running conditions, the perception of the environment by each sensor at different positions and postures is calculated. For example, when the manipulator moves to a certain position, the laser ranging sensor calculates the distance information of the surrounding obstacles that it can detect through the mathematical model, forming simulation monitoring data. The vision sensor generates corresponding simulated image data as simulation monitoring data based on the lighting conditions, object color and shape, etc. in the current scenario. Analyze the simulation monitoring data to evaluate the effectiveness of the information collection range. For example, check whether the simulation monitoring data of the laser ranging sensor can cover the obstacle areas that the manipulator may encounter during movement, and whether the simulated image data of the vision sensor can clearly identify the position and characteristics of the goods. According to factors such as the coverage degree and data accuracy, a scoring standard is formulated. If the laser ranging sensor can accurately detect more than 90% of the potential obstacle areas, and the vision sensor can clearly identify the goods in more than 80% of the scenarios, a higher effectiveness score can be given. Assume that after comprehensively considering various factors, the effectiveness score of this deployment scheme is 85 points (out of 100 points).

[0071] In this way, by integrating the sensor mathematical model into the environmental model and performing simulation runs, the environmental perception ability of the sensor in various working scenarios can be comprehensively evaluated, including the information acquisition range and accuracy at different positions and postures. This helps to identify potential problems in the sensor deployment scheme in practical applications, such as certain areas that cannot be effectively monitored, or the performance of the sensor being affected under specific actions. According to the simulation results and effectiveness scores, the deployment scheme can be optimized accordingly. For example, if a monitoring blind spot frequently appears in a certain area during the simulation, additional sensors can be considered or the positions and angles of the existing sensors can be adjusted to improve the comprehensiveness and accuracy of information acquisition. By continuously adjusting and optimizing, the deployment scheme can be made more reasonable, improving the environmental perception ability of the multi-degree-of-freedom manipulator. Before actually installing and debugging the sensors, problems can be discovered and optimized in advance through simulation evaluation, avoiding the time and cost waste caused by frequently adjusting the sensor positions and parameters in practical applications. This helps to accelerate the project implementation progress and reduce the overall cost. The optimized sensor deployment scheme can provide more accurate and comprehensive environmental information for the multi-degree-of-freedom manipulator, enabling the manipulator to execute tasks more reliably, reducing the probability of operation errors and accidents caused by inaccurate environmental perception, and improving the reliability and stability of the entire automation system.

[0072] As an alternative embodiment, in 102, Gaussian process regression is used to perform trajectory prediction based on the monitoring data and the operation target in the operation instruction to obtain the operation path information of the multi-degree-of-freedom manipulator, including:

[0073] Taking the monitoring data obtained at the current moment as the input, performing trajectory prediction according to the Gaussian process regression algorithm to obtain the prediction mean and variance at the current moment; the prediction mean is used to represent the predicted operation target position of the multi-degree-of-freedom manipulator at the current moment; through Monte Carlo sampling, multiple sample points are randomly sampled from the predicted Gaussian distribution, and each sample point represents a candidate operation path; the Gaussian distribution is constructed from the prediction means and variances at multiple moments; based on the candidate operation paths, control information for each motor in the drive system is generated; the control information at least includes: motor speed, motor torque.

[0074] For example, assume that a multi-degree-of-freedom manipulator works in a two-dimensional plane, and its task is to move from the starting point A to the target point B while avoiding obstacles on the way. The sensor monitors the current position, speed, acceleration of the manipulator and the surrounding environment information (such as the position of obstacles, etc.) in real time, and takes these data as the monitoring data at the current moment. For example, at the first moment, the position, speed and acceleration of the manipulator are monitored, and at the same time, an obstacle is detected at point C. Take the monitoring data at the first moment as the input and calculate according to the Gaussian process regression algorithm. Assume that the predicted mean at the first moment is obtained through calculation, which represents the predicted operation target position of the manipulator at the current moment. At the same time, the variance is obtained.

[0075] Furthermore, during the sampling process, a Gaussian distribution is constructed based on the predicted means and variances at multiple moments. Multiple sample points are randomly sampled from this Gaussian distribution through Monte Carlo sampling. For example, N sample points S1, S2, ……, SN are sampled, and each sample point i = (xi, yi) represents a point on a candidate operation path. For each candidate operation path, according to the kinematic and dynamic models of the manipulator, control information for each motor in the drive system is generated. For example, assume that the manipulator is driven by two motors, which respectively control the movements in the x-axis and y-axis directions. According to the coordinate changes of the points on the candidate operation path and the load conditions of the manipulator, etc., the required rotational speeds and torques of each motor at different moments are calculated. For example, for the sample point S1, the rotational speed of motor 1 at time t1 is calculated as n{11}, and the torque is T{11}; the rotational speed of motor 2 is n{21}, and the torque is T{21}.

[0076] Gaussian process regression can comprehensively consider various information in the monitoring data, including the current state of the manipulator and environmental information, so as to make a relatively accurate prediction of the future trajectory of the manipulator. By continuously updating the monitoring data and making predictions, the operation path can be adjusted in real time, enabling the manipulator to adapt to environmental changes and accurately move towards the operation target. The variance obtained from the prediction and multiple candidate operation paths obtained through Monte Carlo sampling can well take into account the uncertainty in the prediction process. This enables the system to not only obtain the most likely operation path but also understand other possible path options, thus having better robustness when facing complex environments or unexpected situations. For example, when encountering sensor measurement errors or unexpected disturbances in the environment, the system can make more flexible decisions based on the candidate operation paths, avoiding operation failures caused by deviations in a single predicted path. The motor control information generated based on the candidate operation paths can precisely control the rotation speed and torque of the motor, enabling the manipulator to move along the predetermined trajectory. This helps to improve the motion accuracy and control performance of the manipulator, reduce jitter and errors during the motion process, and thus complete the operation task more accurately. For example, when grasping and placing an object, the position and posture of the manipulator can be precisely controlled, improving the success rate and efficiency of the operation. By evaluating and selecting multiple candidate operation paths, an operation path that can optimize certain performance indicators while meeting the operation target can be found. For example, a path that can minimize energy consumption, reach the target in the shortest time, or avoid more obstacles can be selected. This helps to improve the performance and efficiency of the entire system and reduce the operating cost.

[0077] Further optionally, in the above steps, trajectory prediction is performed according to the Gaussian process regression algorithm to obtain the prediction mean and variance at the current moment, including:

[0078] Obtain the real-time monitoring data at the current moment; calculate the covariance vector between the operation target position and the real-time monitoring position at the current moment based on the real-time monitoring data; calculate the mean of the predicted position points obtained through a preset kernel function and the corresponding variance at the current moment based on the covariance vector; use the mean of the predicted position points as the prediction mean at the current moment, and use the variance corresponding to the mean of the predicted position points as the variance at the current moment.

[0079] Gaussian process regression is a machine learning method based on probability statistics, which assumes that the data is generated by a Gaussian process. In the trajectory prediction of a multi-degree-of-freedom manipulator, it uses historical monitoring data and operation target information to establish a probability model for predicting future trajectories. By calculating the covariance vector between the operation target position at the current moment and the real-time monitoring position, the correlation between them is measured. The covariance reflects the change relationship between different positions, and based on this, the possible change trend of the operation target position can be inferred. A preset kernel function is used to calculate the covariance, which maps the data points in the input space to a high-dimensional feature space, making it easier to find the linear relationship between the data in the high-dimensional space. The choice of the kernel function determines the properties and prediction ability of the Gaussian process. For example, the kernel function has a radial basis function (RBF), etc. Here, the kernel function calculates the covariance between the operation target position and the monitoring data points by considering factors such as the distance and correlation between them, thus providing a basis for prediction.

[0080] In this way, based on the real-time monitoring data and historical information, the expected estimated value of the operation target position of the multi-degree-of-freedom manipulator at the current moment, that is, the predicted mean, can be accurately predicted. This predicted mean provides a clear target position for the motion control of the manipulator, which helps to achieve precise trajectory control. The variance is calculated to quantify the uncertainty of the prediction. The variance reflects the credibility of the prediction result. A higher variance indicates a greater uncertainty in the prediction result, and a lower variance indicates that the prediction result is relatively more reliable. This enables the control system to adjust the control strategy according to the variance information. For example, more conservative control measures are taken when the uncertainty is large to ensure the safe and stable operation of the manipulator. The Gaussian process regression algorithm can adapt to the complex working environment and diverse operation tasks of the multi-degree-of-freedom manipulator. It can handle non-linear and non-stationary data, and predict future trajectories by learning the patterns in the historical data, and has good adaptability to different operation targets and changes in the working environment.

[0081] In the embodiment of the present application, the preset kernel function is expressed as ; the current moment is represented as the th moment, represents the mean value of the predicted position points at the th moment, and the mean value of the predicted position points at the th moment is the expected estimated value of the operation target position predicted based on the real-time monitoring data at the th moment , represents the covariance vector calculated by the kernel function between the operation target position at the th moment and each monitoring data point in the real-time monitoring data, ​ Indicates the operational target position at the transpose of the covariance vector between each monitoring data point in the real-time monitoring data, Indicates the covariance matrix corresponding to the historical operational target position matched with the real-time monitoring data at the Indicates the inverse matrix of the sum of the covariance matrix corresponding to the historical operational target position value and the noise variance matrix at the th moment, where the noise variance matrix is a diagonal matrix obtained by multiplying the noise variance by the identity matrix.

[0082] In the embodiments of the present application, ; Indicates the predicted variance corresponding to the mean of the predicted position points at the is the signal variance matched by the sensor, is the length scale parameter matched by the sensor, is the operational target position predicted at the is the reference operational target position correlated with at the Indicates the correlation measure between and at the is the kernel function corresponding predicted uncertainty control term.

[0083] The formula involves the length scale parameter matched by the sensor, etc., and is used to calculate the variance more precisely, reflecting the uncertainty degree of the prediction result. By calculating the predicted variance, the control system can understand the reliability of the prediction so as to make corresponding decisions. The above formula system provides quantitative results for the trajectory prediction of the multi-degree-of-freedom manipulator by calculating the predicted mean and variance, enabling the control system to plan the operation path of the manipulator according to these results, realizing precise control and optimized motion of the manipulator.

[0084] As an alternative embodiment, in 103, a fuzzy chance-constrained control model is adopted, and according to the control information of each motor and the constraint conditions, local solutions are obtained for the operation paths corresponding to each motor, and the target control strategies for each motor are obtained, including:

[0085] Taking the speed, position, control target, maximum speed, and maximum acceleration of the motor as constraint conditions;

[0086] According to the numerical range and application scenario of the control information in each constraint condition, different fuzzy sets are defined respectively, and the membership function corresponding to each fuzzy set is determined, and the membership function is used to describe the digital distribution level of the specific numerical value of each constraint condition in the corresponding fuzzy set; a confidence level is set for each constraint condition, and the confidence level represents the degree of completion of the motor operation path to the constraint condition; based on the confidence level and the membership function, each predicted operation position in the manipulator motion path corresponding to the control information is analyzed to obtain the feasibility of the operation path and the local control risk corresponding to each fuzzy set; based on the evaluation result, the motor control parameters in the control information are adjusted so that the motor can approach the expected control target and reduce the control risk while satisfying the constraint conditions.

[0087] In the control of multi-DOF manipulators, the operation of the motor is restricted by many factors. The motor's speed, position, control target, maximum speed, and maximum acceleration are used as constraints. These constraints are key factors to ensure the safe and stable operation of the motor and the ability of the manipulator to accurately complete the operation task.

[0088] Speed constraint: The speed of the motor cannot exceed its designed maximum speed, otherwise it may cause damage to the motor or unstable movement of the robot.

[0089] Position constraints: The motors need to move the robot to the specified position accurately while avoiding exceeding its operable range.

[0090] Control target constraints: The operation of the motor needs to meet the control target set by the operation task, such as accurately grasping an object or completing a specific assembly action.

[0091] Maximum speed and maximum acceleration constraints: These two constraints limit the range of motor speed and acceleration to prevent the motor from generating excessive impact force during starting or stopping, which may damage the equipment.

[0092] Fuzzy sets and membership functions are the core concepts of the fuzzy chance constraint control model, which are used to deal with uncertainty and ambiguity in constraints. According to the numerical range and application scenario of the control information in each constraint, each constraint is divided into different fuzzy sets. For example, for the motor speed constraint, three fuzzy sets of "low speed", "medium speed" and "high speed" can be defined; for the position constraint, fuzzy sets such as "close to the target position" and "deviate from the target position" can be defined. The membership function is used to describe the numerical distribution level of the specific values of each constraint in the fuzzy set to which it belongs. Common membership functions include triangular membership function, trapezoidal membership function, Gaussian membership function, etc. Taking the "medium speed" fuzzy set of motor speed as an example, assuming that the medium speed range is 30-60 rpm, it can be represented by a triangular membership function. When the motor speed is 45 rpm, its membership in the “medium speed” fuzzy set is 1; when the speed is 30 rpm or 60 rpm, the membership is 0; between 30 and 60 rpm, the membership changes linearly with the speed.

[0093] The confidence level indicates the degree of completion of the motor operation path's contribution to the constraint. Setting a confidence level for each constraint reflects the importance of the constraint in the control process. For example, for the maximum speed constraint of the motor, a higher confidence level can be set, because exceeding the maximum speed may cause serious damage to the motor. For some relatively minor constraints, a lower confidence level can be set. The setting of the confidence level needs to be adjusted according to the specific application scenario and control requirements.

[0094] Based on the confidence level and membership function, each predicted operation position in the manipulator motion path corresponding to the control information is analyzed. For each fuzzy set, the feasibility of the operation path corresponding to the fuzzy set is calculated according to the membership function and confidence level. For example, for the "medium speed" fuzzy set of motor speed, the feasibility of the operation path corresponding to the fuzzy set is obtained by calculating the membership of the motor speed at each predicted operation position and combining the confidence level. If the membership of the motor speed in the "medium speed" fuzzy set is high at a certain predicted operation position, and the confidence level is also high, then the feasibility of the operation path in "medium speed" is high. Local control risk refers to the risk that the operation of the motor may violate the constraint conditions at a certain predicted operation position. By analyzing the membership function and confidence level, the local control risk corresponding to each fuzzy set can be evaluated. For example, if the motor speed is close to the maximum speed and the confidence level of the maximum speed constraint is high, the local control risk of the operation position is high.

[0095] According to the evaluation results of the operation path feasibility and local control risks, adjust the motor control parameters in the control information, such as the speed and torque of the motor. The goal of the adjustment is to make the motor approach the expected control target while meeting the constraint conditions and reduce the control risks. For example, if it is found through analysis that the local control risk of the motor speed at a certain predicted operation position is relatively high and may exceed the maximum speed constraint, the speed of the motor can be appropriately reduced; if it is found that the risk of the motor position deviating from the target position is relatively high, the torque of the motor can be adjusted to make the manipulator approach the target position faster. By continuously adjusting the motor control parameters, the operation path of the motor can be optimized, and the control performance and stability of the manipulator can be improved.

[0096] Thus, by fuzzifying the constraint conditions, the fuzzy chance-constrained control model comprehensively considers the confidence level and membership function to analyze and optimize the operation path of the motor, thereby obtaining the target control strategies for each motor to ensure that the manipulator can safely and accurately complete the operation task in a complex environment.

[0097] As an optional embodiment, in 104, execute the target control strategy to make the robotic arm perform optimized trajectory motion under the control cooperation of each motor.

[0098] The target control strategy is obtained based on the previous calculations and analyses. It stipulates how each motor should act at different times to achieve a specific motion of the robotic arm. Specifically, it adjusts the control parameters of the motor according to the constraint conditions such as the speed, position, and control target of the motor, and also considering the operation path feasibility and local control risks analyzed in the fuzzy chance-constrained control model. The motion of a multi-degree-of-freedom robotic arm is controlled by multiple motors working together. Each motor is responsible for driving one or more joints of the robotic arm. By precisely controlling parameters such as the speed and torque of each motor, the rotation and movement of each joint of the robotic arm can be achieved, and thus the robotic arm can move along the desired trajectory. For example, for a robotic arm with multiple joints, it may be necessary to simultaneously control the motors of the shoulder, elbow, and wrist to make them work at different speeds and torques at different times to complete complex actions.

[0099] When executing the target control strategy, the motor continuously adjusts its own motion state according to the control information, so that the robotic arm can move along the optimized trajectory. This optimized trajectory is obtained after considering various constraint conditions and control risks, aiming to enable the robotic arm to complete the task efficiently and accurately while avoiding exceeding the capabilities of the motor or generating excessive control risks. For example, in the task of carrying an object, the optimized trajectory may make the robotic arm reach the target position in the shortest time and with the least energy consumption, and remain stable during the movement to avoid colliding with surrounding objects.

[0100] Thus, by executing the target control strategy, the robot arm can move more accurately along the preset trajectory and reduce motion errors. This is very important for tasks that require high-precision operations, such as assembly and welding, and can improve product quality and production efficiency. The target control strategy that takes into account constraints and control risks can make the motor work within a safe range, avoiding motor overload or excessive movement that causes system instability. This helps to extend the service life of the robot arm and motor and reduce the probability of failure. The optimized trajectory motion allows the robot arm to complete tasks in a more reasonable way, saving time and energy. For example, in material handling tasks, the robot arm can choose the optimal path and movement speed, reduce empty strokes and unnecessary movements, and improve the efficiency of material handling. The target control strategy can be flexibly adjusted according to different task requirements and working environments, so that the robot arm can adapt to various complex operation tasks. For example, in the task of handling objects of different shapes and weights, the robot arm can complete the handling in the best way by adjusting the trajectory and control parameters.

[0101] The method provided in the embodiment of the present application can realize intelligent control of the multi-degree-of-freedom manipulator, enhance the environmental perception ability of the multi-degree-of-freedom manipulator, improve the accuracy of trajectory prediction, and improve the control precision of the multi-degree-of-freedom manipulator.

[0102] Figure 2 A schematic diagram of a multi-degree-of-freedom manipulator control system provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system is applied to the control scenario of a multi-degree-of-freedom manipulator. The system includes at least: a manipulator, a drive system, a control module, and a data acquisition module; wherein the data acquisition module is used to monitor the working environment of the multi-degree-of-freedom manipulator in real time to obtain monitoring data;

[0103] A control module, for receiving an operation instruction of the multi-degree-of-freedom manipulator, and using Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instruction, so as to obtain the operation path information of the multi-degree-of-freedom manipulator; the operation path information includes: control information and constraints of each motor in the drive system; using a fuzzy chance constraint control model, according to the control information and constraints of each motor, locally solving the operation path corresponding to each motor, and obtaining the target control strategy of each motor; in the fuzzy chance constraint control model, fuzzy variables are constructed based on the control target and the constraints, and the confidence level is introduced by using the credibility theory to quantitatively calculate the feasibility of the operation path of each motor, and the local control risk of each motor is evaluated to comprehensively obtain the target control strategy;

[0104] The drive system is used to execute the target control strategy so that the robot arm can move along an optimized trajectory under the control of each motor.

[0105] Further optionally, the data acquisition module monitors the working environment of the multi-degree-of-freedom manipulator in real time to obtain monitoring data, specifically used for:

[0106] Based on the spatial structure of the working environment and the business type corresponding to the working environment, construct a deployment plan for different types of sensors; evaluate the effectiveness of the information collection range of the deployment plan, and optimize the deployment plan based on the evaluation results; based on the optimized deployment plan, deploy different types of sensors in the working environment; collect the real-time environment information of the multi-degree-of-freedom manipulator from different dimensions through different types of sensors to obtain multi-dimensional monitoring data.

[0107] Further optionally, the data acquisition module evaluates the effectiveness of the information collection range of the deployment plan, specifically used for:

[0108] According to the field of view angle and / or detection distance of each sensor in the deployment plan, combined with the spatial structure of the working environment, determine the theoretical coverage range of each sensor through geometric calculation; compare the overlap ratio between the theoretical coverage range of each sensor and the reference area to be monitored in the working environment; based on the accuracy requirements of the multi-dimensional monitoring data for the business type, analyze the spatial resolution and time resolution corresponding to each sensor in its respective position and position arrangement relationship; based on the overlap ratio, spatial resolution and time resolution, calculate the effectiveness score of the information collection range in the deployment plan.

[0109] Further optionally, the data acquisition module evaluates the effectiveness of the information collection range of the deployment plan, specifically used for:

[0110] According to the field of view angle and / or detection distance, and sensor characteristics of each sensor in the deployment plan, establish a mathematical model for each sensor; set the simulation operation conditions of the manipulator at different positions and motions in the working scenario corresponding to the working environment; integrate the mathematical model of each sensor into the environment model, and use the simulation operation conditions to simulate the perception of the environment by each sensor at different positions and postures in the environment model; based on the simulated monitoring data in the perception situation, obtain the effectiveness score of the information collection range in the deployment plan.

[0111] Further optionally, the control module uses Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instruction to obtain the operation path information of the multi-degree-of-freedom manipulator, specifically used for:

[0112] Taking the monitoring data obtained at the current moment as input, trajectory prediction is performed according to the Gaussian process regression algorithm to obtain the predicted mean and variance at the current moment; the predicted mean is used to represent the predicted operation target position of the multi-degree-of-freedom manipulator at the current moment; through Monte Carlo sampling, multiple sample points are randomly sampled from the predicted Gaussian distribution, and each sample point represents a candidate operation path; the Gaussian distribution is constructed from the predicted means and variances at multiple moments; based on the candidate operation paths, control information for each motor in the drive system is generated; the control information at least includes: motor speed, motor torque.

[0113] Further optionally, the control module performs trajectory prediction according to the Gaussian process regression algorithm to obtain the predicted mean and variance at the current moment, specifically for:

[0114] Obtain the real-time monitoring data at the current moment; calculate the covariance vector between the operation target position and the real-time monitoring position at the current moment based on the real-time monitoring data; calculate the mean of the predicted position points obtained through a preset kernel function and the corresponding variance at the current moment; where the preset kernel function is expressed as ; the current moment is represented as the th moment, represents the mean of the predicted position points at the th moment, and the mean of the predicted position points at the th moment is the expected estimated value of the operation target position predicted based on the real-time monitoring data at the th moment , represents the covariance vector calculated through the kernel function between the operation target position at the th moment and each monitoring data point in the real-time monitoring data, represents the transpose of the covariance vector between the operation target position at the th moment and each monitoring data point in the real-time monitoring data, represents the covariance matrix corresponding to the historical operation target position matching the real-time monitoring data at the th moment, represents the inverse matrix of the sum of the covariance matrix corresponding to the historical operation target position value and the noise variance matrix at the th moment, and the noise variance matrix is a diagonal matrix obtained by multiplying the noise variance by the identity matrix; ; represents the predicted variance corresponding to the mean of the predicted position points at the th moment, is the signal variance of the sensor matching, is the length scale parameter for sensor matching, For the The predicted operation target position at the moment is For the At this moment The reference operation target position with correlation between Indicates At this moment and The correlation measure between is the kernel function Corresponding prediction uncertainty control items; taking the mean of the predicted position points as the predicted mean at the current moment, and taking the variance corresponding to the mean of the predicted position points as the variance at the current moment.

[0115] Further optionally, the control module adopts a fuzzy chance constraint control model to locally solve the operation path corresponding to each motor according to the control information and constraint conditions of each motor to obtain the target control strategy of each motor, which is specifically used for:

[0116] The speed, position, control target, maximum speed and maximum acceleration of the motor are taken as constraints; different fuzzy sets are defined according to the numerical range and application scenarios of the control information in each constraint, and the membership function corresponding to each fuzzy set is determined, and the membership function is used to describe the digital distribution level of the specific values of each constraint in the corresponding fuzzy set; a confidence level is set for each constraint, and the confidence level represents the degree of completion of the motor operation path to the constraint; based on the confidence level and the membership function, each predicted operation position in the manipulator motion path corresponding to the control information is analyzed to obtain the feasibility of the operation path and the local control risk corresponding to each fuzzy set; based on the evaluation results, the motor control parameters in the control information are adjusted so that the motor can approach the expected control target and reduce the control risk while satisfying the constraints.

[0117] The system provided in the embodiment of the present application can realize intelligent control of the multi-degree-of-freedom manipulator, improve the environmental perception ability of the multi-degree-of-freedom manipulator, improve the accuracy of trajectory prediction, and improve the control accuracy of the multi-degree-of-freedom manipulator.

[0118] The present application provides a computer-readable storage medium. Figure 3 As shown, this embodiment provides a computer-readable storage medium 600 on which a computer program 611 is stored. When the computer program 611 is executed by a processor, the aforementioned embodiment is implemented.

[0119] It should be noted that in the above embodiments, the descriptions of the various embodiments have their own emphases. For parts not described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0120] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one or more of the flows Figure 1 or a combination of multiple flows and / or blocks

[0122] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the specified functions in Figure 1 one or more of the flows Figure 1 or a combination of multiple flows and / or blocks

[0123] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in Figure 1 one or more of the flows Figure 1 or a combination of multiple flows and / or blocks

[0124] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn the basic inventive concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present invention.

[0125] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A multi-degree-of-freedom manipulator control method, characterized in that, The method is applied to the control scenario of a multi - degree - of - freedom manipulator; the method includes: Real - time monitoring of the working environment where the multi - degree - of - freedom manipulator is located to obtain monitoring data; Receiving the operation instruction of the multi - degree - of - freedom manipulator, and using Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instruction to obtain the operation path information of the multi - degree - of - freedom manipulator; the operation path information includes: control information for each motor in the drive system and constraint conditions, where the constraint conditions include the speed, position, control target, maximum speed, and maximum acceleration of the motor; Adopting a fuzzy chance - constrained control model, and locally solving the operation paths corresponding to each motor according to the control information of each motor and the constraint conditions to obtain the target control strategy for each motor; in the fuzzy chance - constrained control model, fuzzy variables are constructed based on the control target and the constraint conditions, and the credibility theory is used to introduce a confidence level to quantitatively calculate the feasibility of the operation paths of each motor and evaluate the local control risks of each motor, and the target control strategy is comprehensively obtained; Executing the target control strategy to enable the robotic arm to perform an optimized trajectory movement under the control cooperation of each motor; Among them, the real - time monitoring of the working environment where the multi - degree - of - freedom manipulator is located to obtain monitoring data includes: Based on the spatial structure of the working environment and the business type corresponding to the working environment, constructing a deployment plan for different types of sensors; Evaluating the effectiveness of the information collection range of the deployment plan, and optimizing the deployment plan based on the evaluation result; Based on the optimized deployment plan, deploying different types of sensors in the working environment; Collecting the real - time environment information where the multi - degree - of - freedom manipulator is located from different dimensions through different types of sensors to obtain multi - dimensional monitoring data; The using Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instruction to obtain the operation path information of the multi - degree - of - freedom manipulator includes: Taking the monitoring data obtained at the current moment as input, and performing trajectory prediction according to the Gaussian process regression algorithm to obtain the prediction mean and variance at the current moment; the prediction mean is used to represent the predicted operation target position of the multi - degree - of - freedom manipulator at the current moment; Through Monte Carlo sampling, randomly sampling multiple sample points from the predicted Gaussian distribution, and each sample point represents a candidate operation path; the Gaussian distribution is constructed by the prediction means and variances at multiple moments; Based on the candidate operation paths, generating control information for each motor in the drive system; the control information at least includes: motor speed, motor torque; The performing trajectory prediction according to the Gaussian process regression algorithm to obtain the prediction mean and variance at the current moment includes: Obtaining the real - time monitoring data at the current moment; Calculating the covariance vector between the operation target position and the real - time monitoring position at the current moment based on the real - time monitoring data; calculating the mean of the predicted position points obtained by the preset kernel function at the current moment and the corresponding variance based on the covariance vector; Among them, the preset kernel function is expressed as The current moment is represented as the p-th moment, μ p represents the mean value of the predicted position points at the p-th moment, and the mean value of the predicted position points at the p-th moment is the expected estimated value of the operation target position x predicted based on the real-time monitoring data at the p-th moment p , k p represents the covariance vector calculated by the kernel function k between the operation target position x at the p-th moment p and each monitoring data point in the real-time monitoring data; represents the transpose of the covariance vector between the operation target position x at the p-th moment p and each monitoring data point in the real-time monitoring data, y represents the covariance matrix corresponding to the historical operation target position matched with the real-time monitoring data at the p-th moment represents the inverse matrix of the sum of the covariance matrix corresponding to the historical operation target position value and the noise variance matrix at the p-th moment, and the noise variance matrix is a diagonal matrix obtained by multiplying the noise variance by the identity matrix; σ p 2 represents the predicted variance corresponding to the mean of the predicted position points at the p-th moment, σ f 2 is the signal variance matched by the sensor, l is the length scale parameter matched by the sensor, x p is the operation target position predicted at the p-th moment, x q at the p-th moment is related to x p is the reference operation target position with correlation, represents x at the p-th moment p and x q is the correlation measure between them, is the predicted uncertainty control term corresponding to the kernel function k; The predicted position point mean is used as the predicted mean at the current moment, and the variance corresponding to the predicted position point mean is used as the variance at the current moment.

2. The multi-degree-of-freedom manipulator control method according to claim 1, wherein, The effectiveness evaluation of the information collection scope of the deployment plan includes: Determine the theoretical coverage of each sensor through geometric calculation according to the field of view and / or detection distance of each sensor in the deployment scheme and in combination with the spatial structure of the working environment; Compare the overlap ratio between the theoretical coverage of each sensor and the reference area to be monitored in the working environment; Based on the accuracy requirements of the multi-dimensional monitoring data by the business type, the spatial resolution and temporal resolution corresponding to each sensor at its respective location and position arrangement relationship are analyzed; Based on the overlap ratio, spatial resolution and temporal resolution, an effectiveness score of the information collection range in the deployment scheme is calculated.

3. The multi-degree-of-freedom manipulator control method according to claim 1, wherein The effectiveness evaluation of the information collection scope of the deployment plan includes: Establishing a mathematical model of each sensor according to the field of view angle and / or detection distance and sensor characteristics of each sensor in the deployment scheme; Set the simulated operating conditions of the manipulator in different positions and actions in the working scene corresponding to the working environment; Integrate the mathematical model of each sensor into the environment model, and use the simulated operating conditions to simulate the perception of the environment by each sensor at different positions and postures in the environment model; Based on the simulated monitoring data in the perception situation, an effectiveness score of the information collection scope in the deployment plan is obtained.

4. The multi-degree-of-freedom manipulator control method according to claim 1, wherein, The fuzzy chance constraint control model is used to locally solve the operation path corresponding to each motor according to the control information and constraint conditions of each motor to obtain the target control strategy of each motor, including: According to the numerical range and application scenario of the control information in each constraint condition, different fuzzy sets are defined respectively, and the membership function corresponding to each fuzzy set is determined, wherein the membership function is used to describe the digital distribution level of the specific numerical value of each constraint condition in the fuzzy set to which it belongs; Setting a confidence level for each constraint, wherein the confidence level represents the degree of completion of the motor operation path to the constraint; Based on the confidence level and the membership function, each predicted operation position in the manipulator motion path corresponding to the control information is analyzed to obtain the feasibility of the operation path and the local control risk corresponding to each fuzzy set; The motor control parameters in the control information are adjusted based on the evaluation results, so that the motor can approach the expected control target while satisfying the constraint conditions and reduce the control risk.

5. A multi-degree-of-freedom manipulator control system, characterized in that, The system is applied to the control scenario of a multi-degree-of-freedom manipulator; the system at least includes: a manipulator, a drive system, a control module, and a data acquisition module; wherein, The data acquisition module is used to monitor the working environment of the multi-degree-of-freedom manipulator in real time to obtain monitoring data: The control module is used to receive the operation instructions of the multi-degree-of-freedom manipulator, and perform trajectory prediction based on the monitoring data and the operation target in the operation instructions by using Gaussian process regression to obtain the operation path information of the multi-degree-of-freedom manipulator; the operation path information includes: control information for each motor in the drive system and constraint conditions, and the constraint conditions include the speed, position, control target, maximum speed, and maximum acceleration of the motor; adopting a fuzzy chance-constrained control model, according to the control information of each motor and the constraint conditions, locally solve the operation paths corresponding to each motor to obtain the target control strategy for each motor; in the fuzzy chance-constrained control model, fuzzy variables are constructed based on the control target and the constraint conditions, and the credibility theory is used to introduce the confidence level to quantitatively calculate the feasibility of the operation paths of each motor and evaluate the local control risks of each motor, and the target control strategy is comprehensively obtained; The drive system is used to execute the target control strategy to enable the robotic arm to perform optimized trajectory motion under the control cooperation of each motor; Among them, the real-time monitoring of the working environment where the multi-degree-of-freedom manipulator is located to obtain monitoring data includes: Based on the spatial structure of the working environment and the business type corresponding to the working environment, construct a deployment plan for different types of sensors; Evaluate the effectiveness of the information collection range of the deployment plan, and optimize the deployment plan based on the evaluation results; Based on the optimized deployment plan, deploy different types of sensors in the working environment; Collect the real-time environment information of the multi-degree-of-freedom manipulator from different dimensions through different types of sensors to obtain multi-dimensional monitoring data; The use of Gaussian process regression to perform trajectory prediction based on the monitoring data and the operation target in the operation instructions to obtain the operation path information of the multi-degree-of-freedom manipulator includes: Take the monitoring data obtained at the current moment as the input, and perform trajectory prediction according to the Gaussian process regression algorithm to obtain the prediction mean and variance at the current moment; the prediction mean is used to represent the predicted operation target position of the multi-degree-of-freedom manipulator at the current moment; Through Monte Carlo sampling, randomly sample multiple sample points from the predicted Gaussian distribution, and each sample point represents a candidate operation path; the Gaussian distribution is constructed by the prediction means and variances at multiple moments; Based on the candidate operation paths, generate control information for each motor in the drive system; the control information at least includes: motor speed, motor torque; The performing trajectory prediction according to the Gaussian process regression algorithm to obtain the prediction mean and variance at the current moment includes: Obtain the real-time monitoring data at the current moment; Calculate the covariance vector between the operation target position and the real-time monitoring position at the current moment based on the real-time monitoring data; calculate the prediction position point mean and the corresponding variance obtained by the preset kernel function at the current moment based on the covariance vector; Among them, the preset kernel function is expressed as The current moment is expressed as the p-th moment, μ p represents the mean value of the predicted position points at the p-th moment, and the mean value of the predicted position points at the p-th moment is the expected estimated value of the operation target position x predicted based on the real-time monitoring data at the p-th moment p , k p represents the covariance vector calculated by the kernel function k between the operation target position x at the p-th moment p and each monitoring data point in the real-time monitoring data; represents the transpose of the covariance vector between the operation target position x at the p-th moment p and each monitoring data point in the real-time monitoring data, y represents the covariance matrix corresponding to the historical operation target position matched with the real-time monitoring data at the p-th moment, K y -1 represents the inverse matrix of the sum of the covariance matrix corresponding to the historical operation target position value and the noise variance matrix at the p-th moment, and the noise variance matrix is a diagonal matrix obtained by multiplying the noise variance by the identity matrix; σ p 2 represents the predicted variance corresponding to the mean of the predicted position points at the p-th moment, σ f 2 is the signal variance matched by the sensor, l is the length scale parameter matched by the sensor, x p is the predicted operating target position obtained at the p-th moment, x q at the p-th moment is related to x p The reference operating target position with correlation, represents x at the p-th moment p and x q The correlation measure between them, is the predicted uncertainty control term corresponding to the kernel function k; Take the prediction position point mean as the prediction mean at the current moment, and take the variance corresponding to the prediction position point mean as the variance at the current moment.

6. An electronic device, characterized in that, It includes a memory for storing a computer software program; a processor for reading and executing the computer software program to implement the multi-degree-of-freedom manipulator control method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, It includes instructions that, when run on a computer, cause the computer to execute the multi-degree-of-freedom manipulator control method according to any one of claims 1-4.

Images