An automated production process and robot for metal sheet materials

Abstract

The present invention discloses an automated production process and robot for metal sheet materials, and relates to the field of robot control technology; the automated production process includes: obtaining sensor data of joint components; constructing a robot dynamics model, and constructing an observation model according to the sensor data and the robot dynamics model; obtaining an observation state data matrix through the observation model, and determining the robot control law according to the observation state data matrix; and generating robot control instructions according to the robot control law. The present application realizes state monitoring of the robot by acquiring sensor data of joint components, combining the sensor data and the robot dynamics model to construct an observation model; and by constructing an observation model, efficient and accurate control of the robot in the automated production process of metal sheet materials is realized.

Description

Technical Field

[0001] The invention relates to the technical field of robot control, and in particular to an automated production process for hardware sheet materials and a robot. Background Art

[0002] The demand for automated production in the modern manufacturing industry is constantly increasing, and the requirements for production efficiency and product quality are also increasing. In the traditional production process of sheet metal, due to the reliance on manual operation or simple automated equipment, it is not only inefficient, but also difficult to ensure the consistency and precision of the products. These traditional methods can no longer meet the needs of modern production, and efficient and high-precision automated production has become an important issue that the industry needs to solve urgently.

[0003] Multi-degree-of-freedom robot control technology is a key technology in the field of modern manufacturing and industrial automation. Among them, the degree of freedom refers to the number of directions in which the robot can move independently, including translation and rotation. Due to their multiple joints and rotation axes, these robots can perform various complex tasks in three-dimensional space, such as welding, assembly, and handling. With their flexibility and versatility, multi-degree-of-freedom robots not only improve work efficiency and precision, but also expand the ability of humans to work in various complex environments.

[0004] The structure of a multi-degree-of-freedom robot is complex, with multiple joints and rotation axes. The movement of each joint and axis will affect the position and posture of the entire robot. Therefore, efficient and accurate control methods are an important guarantee for achieving precise control of multi-degree-of-freedom robots. However, facing complex systems with multiple degrees of freedom, designing and implementing high-performance control methods is a high technical challenge. Existing control methods are difficult to ensure the high robustness and stability of the robot to cope with various uncertainties in actual operations. Summary of the invention

[0005] The specific technical solutions provided by this application are as follows:

[0006] The present invention provides an automated production process for metal sheet materials, comprising the steps of:

[0007] Acquire sensor data of joint components;

[0008] Construct the robot dynamics model and construct the observation model based on the sensor data and the robot dynamics model;

[0009] Obtain the observation state data matrix through the observation model, and determine the robot control law according to the observation state data matrix;

[0010] Generate robot control instructions according to robot control laws;

[0011] The step of constructing an observation model based on sensor data and a robot dynamics model comprises the following steps:

[0012] Establishing a number of state data matrices corresponding to sampling times according to the sensor data;

[0013] Determine a first correction matrix and a first torque error vector corresponding to each state data matrix by a genetic algorithm;

[0014] Inputting a plurality of sets of first correction matrices and first torque error vectors into a neural network model, and outputting a second correction matrix through the neural network model;

[0015] An observation model is constructed based on the second correction matrix and the robot dynamics model.

[0016] Preferably, the sensor data includes position data and speed data; the robot dynamics model is expressed as:

[0017] ,

[0018] in, is the joint torque vector; is the joint position vector; is the joint velocity vector; is the joint acceleration vector; is the inertia matrix; is the Coriolis force and centrifugal force matrix; is the gravity vector; is the friction force vector.

[0019] Preferably, the determining of the first correction matrix and the first torque error vector corresponding to each group of state data matrix by a genetic algorithm comprises the steps of:

[0020] The population is generated based on the initial state data matrix initialization; each individual in the population represents a correction scheme of the state data matrix;

[0021] Calculating the fitness function value of each individual in the population; the fitness function value is calculated by the Euclidean distance between the initial state data matrix and the individual and the moment error vector between the initial state data matrix and the individual;

[0022] Perform selection, crossover and mutation operations on individuals in the population according to the fitness function value of each individual;

[0023] Iterate until the number of iterations reaches the preset maximum value or the fitness function converges to obtain the optimal individual in the current population;

[0024] A first correction matrix is ​​obtained according to the optimal individual and the initial state data matrix, and the torque error vector of the optimal individual is used as the first torque error vector.

[0025] Preferably, the fitness function is expressed as:

[0026] ,

[0027] in, represents the initial state data matrix, G represents the current individual; represents the Euclidean distance between the initial state data matrix and the individual; represents the moment error vector between the initial state data matrix and the individual, is the preset weight parameter;

[0028] The torque error vector is expressed as:

[0029] ,

[0030] ,

[0031] ,

[0032] in, , and are the joint position vector, joint velocity vector and joint acceleration vector of the current individual respectively; the first term in the formula represents the predicted value of the torque vector of the current individual, and the second term represents the initial predicted value of the torque vector.

[0033] Preferably, after obtaining the observation state data matrix through the observation model and determining the robot control law according to the observation state data matrix, the method further includes: constructing an integral sliding surface based on a nonlinear disturbance observer, specifically:

[0034] The integral sliding surface is expressed as:

[0035]

[0036] in, is the parameter matrix of the integral sliding surface, is the current state value of the robot, is the initial value of the robot state, K is the robot state feedback matrix, and t is the running time of the robot.

[0037] Preferably, the observation model is expressed as:

[0038] ,

[0039] in, , and They respectively represent the joint position vector, joint velocity vector and joint acceleration vector corrected by the second correction matrix.

[0040] Preferably, the method of determining the robot control law according to the observed state data matrix further comprises the steps of:

[0041] Set priorities based on robot groups;

[0042] Predicting the observed state data by using a Kalman filter to obtain predicted state data;

[0043] The robot broadcasts the predicted state data to the domain robots according to the domain matrix; the domain matrix is ​​used to record the domain intersection of the domain robots;

[0044] Predicting the predicted paths of the domain robot and the main body robot based on the received predicted state data and its own predicted state data;

[0045] If it is identified that there is a path conflict between the main robot and the field robot, for each pair of robots in conflict, the task area is allocated according to the priority;

[0046] The control law of each robot is updated according to the task area allocation results.

[0047] Preferably, in the updating of the control law of each robot according to the task area allocation result, the updated control law is expressed as:

[0048] ,

[0049] in, represents the avoidance control item; and Represent the proportional gain and differential gain matrices respectively; and Respectively represent the joint position vector and joint velocity vector corrected by the second correction matrix; and They respectively represent the joint position vector and joint velocity vector corrected by the second correction matrix.

[0050] Preferably, the avoidance control item is expressed as:

[0051] ,

[0052] in, represents the gravitational coefficient of the target point, represents the obstacle rejection coefficient, Indicates the distance between the robot and the conflicting robot; Indicates the conflict distance.

[0053] The present invention also provides an automated production robot for metal sheet materials, comprising a joint assembly, a sensor group, a monitoring module, a communication module, a data processing module and a control module;

[0054] A sensor group, used to obtain sensor data of the joint component;

[0055] The monitoring module is configured to construct a robot dynamics model based on the sensor data and to construct an observation model based on the robot dynamics model;

[0056] The communication module is configured to transmit the torque data to the data processing module and obtain the robot control law transmitted by the data processing module;

[0057] The data processing module is configured to determine the robot control law according to the observed state data matrix;

[0058] The control module is configured to generate robot control instructions according to the robot control law.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] This application achieves state monitoring of the robot by acquiring sensor data of the joint components and building an observation model based on the sensor data and the robot dynamics model. By using a genetic algorithm to determine the first correction matrix and the first torque error vector, and further using a neural network model to output the second correction matrix, the accuracy and stability of the observation model can be optimized. The observation model is built based on the second correction matrix and the robot dynamics model to efficiently and accurately control the robot in the automated production process of metal sheet materials.

[0061] By analyzing the results of task area allocation, each robot can update its control law to obtain an efficient and coordinated control mode. First, it avoids damage or pauses caused by path conflicts, and increases the reliability and safety of task completion. Secondly, the robot can adjust its path planning and movement in real time according to real-time task requirements to adapt to the changing working environment and ensure the continuity of the task. In addition, through reasonable task allocation, the robot can efficiently complete its own tasks without interfering with the tasks of other robots, thereby improving resource utilization and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0064] Figure 1 A schematic diagram of a process flow of an automated production process for metal sheet materials provided by an embodiment of the present invention;

[0065] Figure 2 A schematic diagram of a process for constructing an observation model based on sensor data and a robot dynamics model provided in an embodiment of the present invention;

[0066] Figure 3 A schematic diagram of a flow chart of determining a first correction matrix and a first torque error vector corresponding to each group of state data matrices by a genetic algorithm provided in an embodiment of the present invention;

[0067] Figure 4 A schematic diagram of a flow chart for determining a robot control law based on an observed state data matrix provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0070] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0071] The demand for automated production in the modern manufacturing industry is constantly increasing, and the requirements for production efficiency and product quality are also increasing. In the traditional production process of sheet metal, due to the reliance on manual operation or simple automated equipment, it is not only inefficient, but also difficult to ensure the consistency and precision of the products. These traditional methods can no longer meet the needs of modern production, and efficient and high-precision automated production has become an important issue that the industry needs to solve urgently.

[0072] Multi-degree-of-freedom robot control technology is a key technology in the field of modern manufacturing and industrial automation. Among them, the degree of freedom refers to the number of directions in which the robot can move independently, including translation and rotation. Due to their multiple joints and rotation axes, these robots can perform various complex tasks in three-dimensional space, such as welding, assembly, and handling. With their flexibility and versatility, multi-degree-of-freedom robots not only improve work efficiency and precision, but also expand the ability of humans to work in various complex environments.

[0073] The structure of a multi-degree-of-freedom robot is complex, with multiple joints and rotation axes. The movement of each joint and axis will affect the position and posture of the entire robot. Therefore, efficient and accurate control methods are an important guarantee for achieving precise control of multi-degree-of-freedom robots. However, facing complex systems with multiple degrees of freedom, designing and implementing high-performance control methods is a high technical challenge. Existing control methods are difficult to ensure the high robustness and stability of the robot to cope with various uncertainties in actual operations.

[0074] To resolve this issue, see Figure 1 and Figure 2 The present invention provides an automated production process for metal sheet materials, comprising the steps of:

[0075] S1, obtaining sensor data of joint components;

[0076] S2, constructing a robot dynamics model, and constructing an observation model based on the sensor data and the robot dynamics model;

[0077] S3, obtaining an observation state data matrix through an observation model, and determining a robot control law according to the observation state data matrix;

[0078] S4. Generate robot control instructions according to the robot control law.

[0079] The step of constructing an observation model based on sensor data and a robot dynamics model comprises the following steps:

[0080] S21, establishing a number of state data matrices corresponding to sampling times according to the sensor data;

[0081] S22, determining a first correction matrix and a first torque error vector corresponding to each state data matrix by a genetic algorithm;

[0082] S23, inputting several groups of first correction matrices and first torque error vectors into a neural network model, and outputting a second correction matrix through the neural network model;

[0083] S24. Construct an observation model based on the second correction matrix and the robot dynamics model.

[0084] This application achieves state monitoring of the robot by acquiring sensor data of the joint components and building an observation model based on the sensor data and the robot dynamics model. By using a genetic algorithm to determine the first correction matrix and the first torque error vector, and further using a neural network model to output the second correction matrix, the accuracy and stability of the observation model can be optimized. The observation model is built based on the second correction matrix and the robot dynamics model to efficiently and accurately control the robot in the automated production process of metal sheet materials.

[0085] Among them, the construction and use of the robot dynamics model involves complex physical and mathematical relationships, but in practical applications, the accuracy of the model is often affected by various factors, and it is difficult to meet high-precision requirements by purely relying on theoretical models. Specifically, in practical applications, whether it is a position sensor or a speed sensor, there will be certain measurement errors, resulting in inaccuracy of the overall model; in the manufacturing and assembly process of the mechanical structure, there are small dimensional errors and accumulation of assembly deviations, which will also affect the accuracy of the overall dynamics model. Due to the nonlinear characteristics of the above errors, traditional linear calibration and correction methods are not enough to completely overcome and correct the errors in the dynamics model. Therefore, this application uses genetic algorithms and neural networks to process complex nonlinear relationships, and uses genetic algorithms to find better initial parameters to obtain several first correction matrices, provide preliminary corrections, reduce global nonlinear errors, and ensure basic accuracy; and further determine the second correction matrix based on several first correction matrices through neural networks to improve the accuracy and robustness of the observation model, and provide reliable guarantees for high-complexity and high-precision tasks.

[0086] The control instructions are specific operation commands generated based on the control law. Their function is to actually drive the robot's execution unit to complete specific physical actions, such as rotating 90 degrees, moving to a certain position, grabbing or placing a workpiece, etc.

[0087] Preferably, the sensor data includes position data and speed data; the robot dynamics model is expressed as:

[0088]

[0089] in, is the joint torque vector, which represents the driving torque of each joint; ; is the joint position vector, indicating the position of each joint, ; is the joint velocity vector, indicating the velocity of each joint, ; is the joint acceleration vector, which represents the acceleration of each joint. ; is a 6×6 inertia matrix, which is used to describe the inertia characteristics of the robot at each joint position; is a 6×6 Coriolis force and centrifugal force matrix, representing the Coriolis force and centrifugal force caused by the joint velocity; is a 6×1 gravity vector, which represents the gravity load on each joint; is the friction force vector, which represents the torque generated by friction at each joint.

[0090] This embodiment takes a six-degree-of-freedom (6-DOF) robot as an example and describes the relationship between the position, velocity, acceleration, and torque of the robot at different degrees of freedom.

[0091] For further information, see Figure 3 The method of determining the first correction matrix and the first torque error vector corresponding to each group of state data matrices by genetic algorithm comprises the steps of:

[0092] S231, generating a population according to the initial state data matrix initialization; each individual in the population represents a correction scheme of the state data matrix; each state data matrix is ​​composed of a set of joint position vectors, joint velocity vectors and joint acceleration vectors collected at a certain time point t, expressed as ;

[0093] S232, calculating the fitness function value of each individual in the population; the fitness function value is calculated by the Euclidean distance between the initial state data matrix and the individual and the moment error vector between the initial state data matrix and the individual;

[0094] S233, performing selection, crossover and mutation operations on individuals in the population according to the fitness function value of each individual;

[0095] S234, iterate until the number of iterations reaches a preset maximum value or the fitness function converges, and obtain the optimal individual in the current population;

[0096] S235, obtaining a first correction matrix according to the optimal individual and the initial state data matrix, and using the moment error vector of the optimal individual as the first moment error vector. The first correction matrix is ​​the product of the inverse matrix of the initial state data matrix and the optimal individual.

[0097] This embodiment determines the first correction matrix and the first moment error vector corresponding to each group of state data matrices through a genetic algorithm. The specific steps include initializing the population, calculating the fitness function value, performing selection crossover mutation, and obtaining the optimal individual until the iteration is terminated, and then solving the first correction matrix and the first moment error vector, thereby effectively reducing the nonlinear error.

[0098] Furthermore, the fitness function is expressed as:

[0099]

[0100] in, represents the initial state data matrix, G represents the current individual; represents the Euclidean distance between the initial state data matrix and the individual; represents the moment error vector between the initial state data matrix and the individual, is the preset weight parameter.

[0101] Furthermore, the Euclidean distance is expressed as:

[0102] ,

[0103] The torque error vector is expressed as:

[0104] ,

[0105] ,

[0106] ,

[0107] Wherein, n represents the degree of freedom, which is 6 in one embodiment; , and They are the joint position vector, joint velocity vector and joint acceleration vector of the current individual respectively; , and They represent the initial joint position vector, joint velocity vector and joint acceleration vector respectively; the first term in the formula represents the predicted value of the torque vector of the current individual, and the second term represents the predicted value of the initial torque vector.

[0108] Among them, the Euclidean distance is used to measure the similarity between the initial state data matrix and the modified matrix. The smaller the distance, the closer the modified matrix is ​​to the initial state data matrix, and the better the effect in maintaining the consistency of the data structure. The moment error vector is used to measure the error in the moment. The smaller the moment error, the smaller the deviation of the modified matrix in the moment, which can more accurately reflect the state of the actual physical system.

[0109] This embodiment combines the Euclidean distance between the initial state data matrix and the individual and the moment error vector between the initial state data matrix and the individual into a fitness function for measurement, thereby avoiding local optimal or misleading optimization that may be caused by relying on a single indicator, and guiding the genetic algorithm to search for a better solution in a clearer and more efficient way.

[0110] Based on the above content, in order to avoid local optimality or misleading optimization, the genetic algorithm introduces the Euclidean distance between the initial state data matrix and the individual as the fitness function. The first correction matrix is ​​difficult to accurately eliminate the error; and the genetic algorithm only realizes the correction of the state data matrix at a single time point, while the industrial environment is complex and changeable, and a single optimization is difficult to cover all situations. Therefore, this application learns a large number of samples through neural networks, which can better adapt to different working conditions, thereby improving the generalization ability of the system.

[0111] Preferably, the neural network model is a deep fully connected network model, and its input is a first correction matrix and a first moment error vector; the neural network model is trained offline by constructing a loss function to evaluate the moment error vector of the second correction matrix.

[0112] In one embodiment, the loss function is expressed as:

[0113]

[0114] in, represents the moment error vector obtained by the jth sample, represents the square of the second norm of the moment error vector; m is the total number of samples in the training set.

[0115] In this embodiment, the neural network model receives several sets of data including a first correction matrix and a first torque error vector, which describe the deviation between the current system state and the ideal state. In order to improve the stability and efficiency of the training process, the input data is normalized to ensure that the range of each eigenvalue is similar to avoid certain features dominating the training process.

[0116] Next, the neural network is modeled through the following architecture: the input layer inputs the first correction matrix and the first moment error vector to form a high-dimensional vector. Subsequently, the data flows into a hidden layer composed of multiple fully connected layers, which are nonlinearly transformed by using activation functions (such as ReLU) to capture the complex nonlinear relationship of the input data. Finally, the output layer generates a second correction matrix with the same dimension as the first correction matrix, which is used to adjust the system so that the moment error vector is zero in the ideal state.

[0117] In order to measure the performance of the output second correction matrix in eliminating the torque error, the model updates the parameters through the loss function and the Adam optimizer. The Adam optimizer can effectively handle the optimization problems of large-scale data and neural networks due to its adaptive learning rate adjustment mechanism.

[0118] During the training process, the input data passes through each layer of the neural network to generate a second corrected matrix of the output; then the loss value is calculated, and the loss function is minimized through back propagation and weight update.

[0119] After the model is trained, it needs to be evaluated and tuned. Use the validation set to evaluate the model performance, monitor overfitting or underfitting, and improve the model effect by adjusting hyperparameters such as learning rate, batch size, and network architecture. This can usually be done by optimizing hyperparameters through grid search or random search methods.

[0120] Finally, the trained model is transformed (such as quantization and pruning) to improve its operating efficiency and adapt to the actual application environment, and then integrated into the actual control system for real-time evaluation and adjustment to ensure the stability and accuracy of the model in practical applications.

[0121] Preferably, after obtaining the observation state data matrix through the observation model and determining the robot control law according to the observation state data matrix, it also includes: constructing an integral sliding surface based on a nonlinear disturbance observer, specifically:

[0122] The integral sliding surface is expressed as:

[0123]

[0124] in, is the parameter matrix of the integral sliding surface, is the current state value of the robot, is the initial value of the robot state, K is the robot state feedback matrix, and t is the running time of the robot.

[0125] Sliding surface refers to the process in control system design where a specific control strategy is used to make the system's state trajectory reach and remain on a specific hyperplane within a finite time. This hyperplane is the sliding surface. The sliding surface ensures that the system can maintain certain good dynamic qualities, such as asymptotic stability, after reaching the sliding surface. Sliding mode control is a special nonlinear control theory that selects a suitable sliding function and designs the corresponding control law so that the system's phase trajectory can converge to the equilibrium point at a certain speed before reaching the sliding surface.

[0126] In the embodiment of the present invention, the sliding surface theory is introduced to enhance the anti-interference ability of the robot during operation, suppress the uncertainty of the robot system, weaken chattering, and improve the control performance.

[0127] Furthermore, an observation model is constructed based on the second correction matrix and the robot dynamics model, and the observation model is expressed as:

[0128]

[0129] in, , and They respectively represent the joint position vector, joint velocity vector and joint acceleration vector corrected by the second correction matrix. Based on the observation model, the observation state data matrix includes the joint position vector, the joint velocity vector and the torque vector.

[0130] In one embodiment, the robot control law is determined according to the observed state data matrix, and the control law is expressed as:

[0131]

[0132] in, and Represent the proportional gain and differential gain matrices respectively.

[0133] This embodiment can optimize the response speed and stability of the system by adjusting the numerical values ​​of the proportional gain and the differential gain. In this solution, by constructing an observation model based on the sensor data and the robot dynamics model, combining the observation model with the robot control law, and using the PD control law to generate control instructions, precise control of the robot is achieved. The PD control law can quickly respond to changes in the observed state and generate corresponding control instructions based on the size and rate of change of the error, so that the robot can move accurately according to the predetermined trajectory or position.

[0134] In one embodiment, see Figure 4 , the robot control law is determined according to the observed state data matrix, comprising the steps of:

[0135] Priorities are set based on robot groups; groups can be based on their function, location, task type, etc. For example, transport robots, assembly robots, and inspection robots can be divided into different groups based on task type; and the same task type can also be prioritized based on the different business content they are responsible for. Each group is assigned a priority. For example, transport has a low priority, assembly has a medium priority, and inspection has the highest priority. Priorities are set based on the criticality and time sensitivity of the entire production process. For example, an inspection robot needs to complete the inspection work before the part can be further processed, so it has a higher priority.

[0136] The observed state data matrix is ​​predicted by a Kalman filter to obtain predicted state data; a Kalman filter is an efficient recursive filter, which is mainly used to estimate the state of a dynamic system and extract the optimal estimate of the system state from a series of measurement data containing noise. The Kalman filter can provide accurate predicted state data in the presence of noise based on the observed state data matrix.

[0137] The robot broadcasts the predicted state data to the domain robots according to the domain matrix; the domain matrix is ​​used to record the domain intersection of the domain robots;

[0138] Predicting the predicted paths of the domain robot and the main body robot based on the received predicted state data and its own predicted state data;

[0139] If it is identified that there is a path conflict between the main robot and the field robot, for each pair of conflicting robots, the task area is allocated according to the priority; if the conflicting robots have different priorities, the high-priority robot retains the path planning in the conflict area, and the low-priority robot adjusts its path plan to avoid the conflict area; if the conflicting robots have the same priority, the conflict area is divided into non-overlapping sub-blocks, and each block is allocated to a robot.

[0140] The control law of each robot is updated according to the task area allocation results.

[0141] In this embodiment, the updated control law is expressed as:

[0142]

[0143] in, represents an avoidance control item; the avoidance control item is constructed based on an artificial potential energy function;

[0144] The artificial potential energy function is expressed as:

[0145] ,

[0146] ,

[0147] ,

[0148] in, represents the gravitational coefficient of the target point, represents the obstacle rejection coefficient, Indicates the distance between the robot and the conflicting robot; Indicates the conflict distance.

[0149] The gradient of the artificial potential energy function is calculated to obtain the avoidance control term; the avoidance control term is expressed as:

[0150]

[0151] In this embodiment, by analyzing the results of task area allocation, each robot can update its control law to obtain an efficient and coordinated control mode. First, it avoids damage or pauses caused by path conflicts, and increases the reliability and safety of task completion. Secondly, the robot can adjust its path planning and movement in real time according to real-time task requirements to adapt to the changing working environment and ensure the continuity of the task. In addition, through reasonable task allocation, the robot can efficiently complete its own tasks without interfering with the tasks of other robots, thereby improving resource utilization and production efficiency.

[0152] The present invention also provides an automated production robot for metal sheet materials, comprising a joint assembly, a sensor group, a monitoring module, a communication module, a data processing module and a control module;

[0153] A sensor group, used to obtain sensor data of the joint component;

[0154] The monitoring module is configured to construct a robot dynamics model and to construct an observation model based on the sensor data and the robot dynamics model;

[0155] The communication module is configured to transmit the torque data to the data processing module and obtain the robot control law transmitted by the data processing module;

[0156] The data processing module is configured to determine the robot control law according to the observed state data matrix;

[0157] The control module is configured to generate robot control instructions according to the robot control law.

[0158] The joint assembly, specifically a six-degree-of-freedom joint, is used to achieve spatial positioning and posture adjustment; the joints are connected by connecting rods to ensure stable transmission and smooth movement; each joint is integrated with a high-precision motor and encoder to ensure precise control of movement.

[0159] The automated production robot of this application installs the sensor group on the six-degree-of-freedom joints of the robot, collects the sensor data of the joints in real time, and is used to reflect the motion state of each joint of the robot. The monitoring module uses the sensor data obtained by the sensor group to build a robot dynamic model to describe the position and motion of each joint and robotic arm in space. Then, an observation model is further established based on the dynamic model to estimate the torque of each joint required by the robot under different motion states. The communication module is responsible for transmitting the torque data in the monitoring module to the data processing module, and at the same time receives the robot control law generated by the calculation from the data processing module. The data processing module performs calculations based on multiple torque data to generate a robot control law suitable for the current operation requirements. The control module receives and parses the control law transmitted from the data processing module, and generates specific control instructions based on these control laws to guide the actions of each part of the robot, including joint movement, robotic arm extension, and the grasping or placement action of the end effector. The automated production robot of this application has real-time feedback and adjustment capabilities, and can accurately perform the production tasks of hardware sheet materials.

[0160] In one embodiment, the automated production robot of the present application is a six-degree-of-freedom robot, which includes a base, a robotic arm, an end effector, and a human-machine interface in addition to the above joint assembly, sensor group, monitoring module, communication module, data processing module, and control module. The sensor group is arranged at the position of the six-degree-of-freedom joint.

[0161] The base, made of high-strength material, provides a stable platform to ensure the balance and stability of the entire robot.

[0162] The robotic arm is composed of multiple segments connected by joints to provide flexible movement capabilities.

[0163] End effectors are used to configure different types of end effectors according to specific production tasks, such as clamps, suction cups, etc., for grabbing and moving metal sheets; end effectors can be quickly replaced to adapt to different types and specifications of metal sheets.

[0164] The human-machine interaction interface provides an interactive platform between the operator and the robot, including a touch screen, buttons and a graphical interface, which are used to display the robot's operating status, alarm information and operation command input.

[0165] In the embodiments provided in the present application, it should be understood that the disclosed robot can be implemented in other ways. For example, the robot embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of the modules may be electrical, mechanical or other forms.

[0166] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0167] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0168] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

[0169] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An automated production process for metal sheet materials, characterized in that: Includes steps: Acquire sensor data of joint components; Construct the robot dynamics model and construct the observation model based on the sensor data and the robot dynamics model; Obtain the observation state data matrix through the observation model, and determine the robot control law according to the observation state data matrix; Generate robot control instructions according to robot control laws; The step of constructing an observation model based on sensor data and a robot dynamics model comprises the following steps: Establishing a number of state data matrices corresponding to sampling times according to the sensor data; Determine a first correction matrix and a first torque error vector corresponding to each state data matrix by a genetic algorithm; Inputting a plurality of sets of first correction matrices and first torque error vectors into a neural network model, and outputting a second correction matrix through the neural network model; Building an observation model based on the second correction matrix and the robot dynamics model; The method of determining the robot control law according to the observed state data matrix also includes the steps of: Set priorities based on robot groups; Predicting the observed state data by using a Kalman filter to obtain predicted state data; The robot broadcasts the predicted state data to the domain robots according to the domain matrix; the domain matrix is ​​used to record the domain intersection of the domain robots; Predicting the predicted paths of the domain robot and the main body robot based on the received predicted state data and its own predicted state data; If it is identified that there is a path conflict between the main robot and the field robot, for each pair of robots in conflict, the task area is allocated according to the priority; Update the control law of each robot according to the task area allocation result; In the updating of the control law of each robot according to the task area allocation result, the updated control law is expressed as: Among them, τ avoid (t) represents the avoidance control item; K p and K d Represent the proportional gain and differential gain matrices respectively; θ and Respectively represent the joint position vector and joint velocity vector corrected by the second correction matrix; θ′ and Represent the desired joint position vector and the desired joint velocity vector respectively; The avoidance control term is expressed as: Among them, k att Represents the gravitational coefficient of the target point, k rep represents the obstacle rejection coefficient, d(θ) represents the distance between the robot and the conflicting robot; d0 represents the conflict distance.

2. The automated production process for metal sheet materials according to claim 1, characterized in that: The sensor data includes position data and speed data; the robot dynamics model is expressed as: Among them, τ is the joint torque vector; q is the joint position vector; is the joint velocity vector; is the joint acceleration vector; M(q) is the inertia matrix; is the Coriolis force and centrifugal force matrix; G(q) is the gravity vector; is the friction force vector.

3. The automated production process for metal sheet materials according to claim 2, characterized in that: The method of determining the first correction matrix and the first torque error vector corresponding to each group of state data matrices by using a genetic algorithm comprises the following steps: The population is generated based on the initial state data matrix initialization; each individual in the population represents a correction scheme of the state data matrix; Calculating the fitness function value of each individual in the population; the fitness function value is calculated by the Euclidean distance between the initial state data matrix and the individual and the moment error vector between the initial state data matrix and the individual; Perform selection, crossover and mutation operations on individuals in the population according to the fitness function value of each individual; Iterate until the number of iterations reaches the preset maximum value or the fitness function converges to obtain the optimal individual in the current population; A first correction matrix is ​​obtained according to the optimal individual and the initial state data matrix, and the torque error vector of the optimal individual is used as the first torque error vector.

4. The automated production process for metal sheet materials according to claim 3 is characterized in that: The fitness function is expressed as: f(G,G0)=ωd(G,G0)+(1-ω)ε(G,G0), Where G0 represents the initial state data matrix, G represents the current individual; d(G, G0) represents the Euclidean distance between the initial state data matrix and the individual; ε(G, G0) represents the moment error vector between the initial state data matrix and the individual, and ω is the preset weight parameter; The torque error vector is expressed as: ε(G,G0)=|τ(G0)-τ(G)|, Among them, q′, and are the joint position vector, joint velocity vector and joint acceleration vector of the current individual respectively; the first term in the formula represents the predicted value of the torque vector of the current individual, and the second term represents the initial predicted value of the torque vector.

5. The automated production process for metal sheet materials according to claim 1, characterized in that: After obtaining the observation state data matrix through the observation model and determining the robot control law according to the observation state data matrix, it also includes: constructing an integral sliding surface based on a nonlinear disturbance observer, specifically: The integral sliding surface is expressed as: Among them, C k is the parameter matrix of the integral sliding surface, x k is the current state value of the robot, is the initial value of the robot state, K is the robot state feedback matrix, and t is the running time of the robot.

6. The automated production process for metal sheet materials according to claim 1, characterized in that: The observation model is expressed as: Among them, θ, and They respectively represent the joint position vector, joint velocity vector and joint acceleration vector corrected by the second correction matrix.

7. An automated production robot for metal sheet materials, characterized in that: An automated production process for metal sheet materials as claimed in any one of claims 1 to 6, comprising a joint assembly, a sensor group, a monitoring module, a communication module, a data processing module and a control module; A sensor group, used to obtain sensor data of the joint component; The monitoring module is configured to construct a robot dynamics model based on the sensor data and to construct an observation model based on the robot dynamics model; The communication module is configured to transmit the torque data to the data processing module and obtain the robot control law transmitted by the data processing module; The data processing module is configured to determine the robot control law according to the observed state data matrix; The control module is configured to generate robot control instructions according to the robot control law.

Images