Eight-axis linkage control system and method for flying probe test machine

Abstract

The present application relates to the technical field of flying probe testers, and specifically discloses an eight-axis linkage control system and method for flying probe testers, which receives a test instruction sent by a host computer and performs instruction parsing to obtain motion parameters of a PCB board to be tested; based on the motion parameters, determines the motion trajectories of multiple motion axes; performs corresponding actions based on the motion trajectories of multiple motion axes, and monitors the motion data of multiple motion axes in real time; transmits the motion data of multiple motion axes to a central processing unit; in the central processing unit, performs collaborative analysis on the motion data of multiple motion axes to obtain control instructions; transmits the control instructions to a PID controller to control the motor speed corresponding to the first motion axis. In this way, complex nonlinear relationships and dynamic changes can be captured, and the inherent connections between the motion axes can be revealed to accurately understand the dynamic behavior of the system, thereby supporting more sophisticated and intelligent control decisions.

Description

Technical Field

[0001] The present application relates to the technical field of flying probe testers, and more specifically, to an eight-axis linkage control system and method for a flying probe tester. Background Art

[0002] With the rapid development of the electronics manufacturing industry, flying probe testers, as high-precision and highly flexible automated testing equipment, play a vital role in the manufacturing process of printed circuit boards (PCBs). By adopting multi-axis linkage control technology, they can detect and test each component on the PCB board.

[0003] However, traditional PID controllers face significant limitations in multi-axis linkage systems. Its linear control logic has difficulty coping with complex nonlinear and dynamic changes, resulting in response lag or over-adjustment, affecting operational accuracy and stability, especially in high-density, miniaturized PCB board inspection. This deficiency is more obvious. In addition, traditional methods rely on simple statistical analysis and empirical rules to process motion data, and are unable to deeply explore the implicit associations and dynamic characteristics between the various motion axes, which limits the intelligence level of the system and makes it difficult to optimize the control strategy. Due to the lack of understanding of complex interaction patterns, traditional systems find it difficult to achieve efficient and accurate multi-axis collaborative control, further affecting overall performance and reliability.

[0004] Therefore, an optimized eight-axis linkage control scheme for a flying probe tester is desired. Summary of the invention

[0005] The present application provides an eight-axis linkage control system and method for a flying probe tester, which can capture complex nonlinear relationships and dynamic changes, reveal the intrinsic connections between the various motion axes, so as to accurately understand the dynamic behavior of the system, thereby supporting more sophisticated and intelligent control decisions.

[0006] In a first aspect, an eight-axis linkage control method for a flying probe tester is provided, comprising:

[0007] Receive test instructions sent by the host computer;

[0008] Parsing the test instruction to obtain motion parameters of the PCB board being tested, wherein the motion parameters include moving distance, direction and speed;

[0009] Based on the motion parameters, determining motion trajectories of multiple motion axes;

[0010] Execute corresponding actions based on the motion trajectories of the multiple motion axes, and monitor the motion data of the multiple motion axes in real time, wherein the motion data includes position data and speed data;

[0011] Transmitting the motion data of the plurality of motion axes to a central processor;

[0012] In the central processor, the motion data of the plurality of motion axes are collaboratively analyzed to obtain a control instruction, wherein the control instruction is a recommended value for adjusting the motor speed of the first motion axis;

[0013] The control instruction is transmitted to a PID controller to control the motor speed corresponding to the first motion axis.

[0014] In a second aspect, an eight-axis linkage control system for a flying probe tester is provided, comprising:

[0015] A test instruction receiving module is used to receive the test instruction sent by the host computer;

[0016] A test instruction parsing module is used to parse the test instruction to obtain the motion parameters of the PCB board being tested, wherein the motion parameters include moving distance, direction and speed;

[0017] A motion trajectory determination module, used to determine the motion trajectories of multiple motion axes based on the motion parameters;

[0018] A motion data monitoring module, used to perform corresponding actions based on the motion trajectories of the multiple motion axes, and monitor the motion data of the multiple motion axes in real time, wherein the motion data includes position data and speed data;

[0019] A motion data transmission module, used for transmitting the motion data of the multiple motion axes to a central processing unit;

[0020] A motion data collaborative analysis module, used for collaboratively analyzing the motion data of the plurality of motion axes in the central processor to obtain a control instruction, wherein the control instruction is a recommended value for adjusting the motor speed of the first motion axis;

[0021] A control instruction transmission module is used to transmit the control instruction to a PID controller to control the motor speed corresponding to the first motion axis.

[0022] The present application provides an eight-axis linkage control system and method for a flying probe tester, which receives a test instruction sent by a host computer and performs instruction parsing to obtain motion parameters of a PCB board to be tested; based on the motion parameters, determines the motion trajectories of multiple motion axes; performs corresponding actions based on the motion trajectories of multiple motion axes, and monitors the motion data of multiple motion axes in real time; transmits the motion data of multiple motion axes to a central processing unit; in the central processing unit, performs collaborative analysis on the motion data of multiple motion axes to obtain control instructions; transmits the control instructions to a PID controller to control the motor speed corresponding to the first motion axis. In this way, complex nonlinear relationships and dynamic changes can be captured, and the internal connections between the motion axes can be revealed to accurately understand the dynamic behavior of the system, thereby supporting more sophisticated and intelligent control decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0024] Figure 1 This is a schematic flow chart of an eight-axis linkage control method for a flying probe tester according to an embodiment of the present application.

[0025] Figure 2 This is a data flow diagram of an eight-axis linkage control method for a flying probe tester according to an embodiment of the present application.

[0026] Figure 3 This is a schematic flow chart of step S6 in the eight-axis linkage control method for a flying probe tester according to an embodiment of the present application.

[0027] Figure 4 It is a schematic flow chart of step S61 in the eight-axis linkage control method for a flying probe tester according to an embodiment of the present application.

[0028] Figure 5 It is a schematic flow chart of step S63 in the eight-axis linkage control method for a flying probe tester according to an embodiment of the present application.

[0029] Figure 6 It is a schematic flow chart of step S64 in the eight-axis linkage control method for a flying probe tester according to an embodiment of the present application.

[0030] Figure 7 It is a schematic block diagram of an eight-axis linkage control system for a flying probe tester according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0032] Based on this, this application proposes an eight-axis linkage control method for a flying probe tester, which effectively solves many problems existing in the control of traditional flying probe testers by integrating advanced AI technology and data analysis algorithms, and provides a more efficient, accurate and flexible solution. Figure 1 and Figure 2 As shown, the eight-axis linkage control method for a flying probe tester includes: S1, receiving a test instruction sent by a host computer; S2, parsing the test instruction to obtain motion parameters of a PCB board to be tested, wherein the motion parameters include moving distance, direction and speed; S3, determining motion trajectories of a plurality of motion axes based on the motion parameters; S4, performing corresponding actions based on the motion trajectories of the plurality of motion axes, and monitoring the motion data of the plurality of motion axes in real time, wherein the motion data includes position data and speed data; S5, transmitting the motion data of the plurality of motion axes to a central processing unit; S6, in the central processing unit, collaboratively analyzing the motion data of the plurality of motion axes to obtain a control instruction, wherein the control instruction is a recommended value for adjusting the motor speed of the first motion axis; S7, transmitting the control instruction to a PID controller to control the motor speed corresponding to the first motion axis.

[0033] Exemplarily, in step S1, a test instruction sent by a host computer is received. It should be understood that the host computer usually refers to a computer or controller that monitors and manages the entire system, which is responsible for planning tasks, setting parameters, and issuing execution commands. For a flying probe tester, the role of the host computer is to generate corresponding test plans based on the specific specifications and requirements of the PCB board to be tested, and convert these plans into specific test instructions and send them to the slave computer (i.e., the hardware device that actually performs the test action).

[0034] In one embodiment of the present application, an effective communication connection is established between the host computer and the flying probe tester, and this connection can be based on different physical layers and technical standards such as serial communication, Ethernet, and field bus. When the host computer is ready to execute the test instructions, it will package the data in a data format agreed upon by both parties in advance, and then send it out through the selected communication interface. The flying probe tester needs to have a corresponding receiving module to monitor the information flow from the host computer. Once a valid instruction packet is detected, the content of the instruction is immediately read and parsed.

[0035] Exemplarily, in step S2, the test instruction is parsed to obtain the motion parameters of the PCB board to be tested, and the motion parameters include moving distance, direction and speed. It should be understood that each PCB board to be tested has its unique design layout and component configuration, which determines that the flying probe needs to move in a specific way to contact each test point. By parsing the test instruction sent by the host computer, the system can understand the specific test requirements, such as the position where the probe should reach, how to approach the target position and the speed of movement. This information is crucial for planning the motion path of the robot arm, and directly affects the test efficiency and accuracy.

[0036] After receiving the test instruction sent by the host computer, the test instruction parsing module in the flying probe tester will start working. The task of this module is to conduct an in-depth analysis of the content of the instruction data packet and extract all the necessary information required for PCB board testing. First, it needs to identify the format and structure of the instruction to confirm that it is a valid test instruction and contains all the necessary information. Next, it will separate the data describing the geometric dimensions of the PCB board, component distribution, test point coordinates, etc. from the instruction according to the predefined data mapping rules. For each test point, the instruction parsing module needs to clearly indicate the distance the probe should move, the direction of travel (such as along the X-axis or Y-axis), and the speed setting to be adopted at different stages. Such an analysis process can ensure that the subsequent motion trajectory determination module can formulate the optimal motion plan based on accurate data, so that the robot arm can efficiently complete various test operations.

[0037] Exemplarily, in step S3, based on the motion parameters, the motion trajectories of multiple motion axes are determined. It should be understood that in a flying probe tester, each motion axis is responsible for movement in different directions or dimensions, and they work together to achieve accurate contact of the probe with each test point on the PCB board. In order to achieve this goal, the system must plan the specific motion path, i.e., the motion trajectory, of each motion axis based on the resolved motion parameters (such as moving distance, direction, and speed). This step is crucial to ensuring the accuracy and efficiency of the test because it directly determines how the robot arm moves to most effectively complete a series of test operations.

[0038] In one embodiment, first, the motion parameters obtained from the instruction parsing module are used as input, which define where the probe needs to go and how it should move. Then, through a series of calculations, including but not limited to path planning algorithms, dynamic analysis, etc., the system designs an optimal motion route for each motion axis. This route must not only take into account the target position, but also other factors, such as avoiding collisions with other parts, maintaining stable speed changes to reduce vibration, optimizing time to improve efficiency, etc. The final output is a detailed trajectory plan that describes where each motion axis should be at each point in time, in which direction it should move, and how fast it should move.

[0039] In a specific embodiment, suppose there is a PCB board with several dispersed components that need to be tested for electrical performance. According to the test instructions sent by the host computer, the instruction parsing module learns that the probe needs to access these components in sequence and gives the exact coordinate position of each component. Next, the motion trajectory determination module starts working. It first analyzes the relative position relationship between all components to be tested and builds a map containing all possible paths. Then, using the shortest path algorithm such as the A* search algorithm or the Dijkstra algorithm, combined with the dynamic obstacle avoidance strategy, calculates an optimal path that is both safe and can quickly reach all test points without repeating the same location. For each test point, the module will further refine the path and specify the speed curve when the probe approaches to ensure that it can slow down at the right time and accurately stop at the right position. In addition, considering the physical limitations in actual operation, such as maximum acceleration and angular velocity, the module will also adjust the trajectory parameters to make the whole process smooth and smooth.

[0040] Exemplarily, in step S4, corresponding actions are performed based on the motion trajectories of the multiple motion axes, and the motion data of the multiple motion axes are monitored in real time, wherein the motion data includes position data and speed data. It should be understood that in actual operation, each motion axis moves the probe to the specified position for testing according to a pre-planned path, and real-time monitoring provides the necessary feedback mechanism to adjust or correct any behavior that deviates from the expected, thereby ensuring the safety and accuracy of the entire process. After the motion trajectory of each motion axis is determined, the next step is to guide the action of the robot arm according to these trajectories. Specifically, the control system converts the motion trajectory into a series of specific commands and sends them to the drivers (such as motors) corresponding to each motion axis. These commands contain information about when to start, stop, accelerate or decelerate, so that each motion axis can move accurately along the predetermined path. For example, when approaching a test point, the system may instruct an axis to slow down and fine-tune its position to ensure that the probe can stably contact the target without causing damage; when away from the test point, a higher speed is allowed to save time. In this way, all motion axes are coordinated and work together to complete precise positioning in a complex three-dimensional space.

[0041] At the same time, in order to monitor the execution status and respond to possible changes in a timely manner, it is also necessary to collect and analyze the motion data from each motion axis in real time, mainly including position data and speed data. This is usually achieved with the help of sensors installed on each motion axis, such as encoders used to measure the rotation angle and then calculate the linear displacement, and accelerometers used to sense instantaneous acceleration. These sensors continuously generate data streams reflecting the current status, which are then transmitted back to the central processor through a high-speed communication interface.

[0042] Exemplarily, in step S5, the motion data of the multiple motion axes are transmitted to the central processing unit. It should be understood that each motion axis will generate a large amount of real-time data when performing its specific task, including information such as position and speed. These data are crucial to understanding the operating status of the current system. They not only reflect whether each motion axis moves accurately according to the predetermined trajectory, but also reveal possible deviations or abnormalities. By aggregating these scattered data to the central processing unit, global analysis and processing can be performed to ensure the safety, accuracy and efficiency of the entire test process.

[0043] Exemplarily, in step S6, in the central processing unit, the motion data of the multiple motion axes are collaboratively analyzed to obtain a control instruction, and the control instruction is a recommended value for adjusting the motor speed of the first motion axis. It should be understood that when the real-time motion data of all motion axes are transmitted to the central processing unit, these data contain key information such as position and speed, which reflect the current working state of each motion axis and the interaction between each other. Through the collaborative analysis of these data, the central processing unit can capture the subtle but important relationship between the various motion axes, such as how the speed change of one axis affects the position accuracy of other axes, or whether the degree of synchronization between different axes meets expectations. This deep understanding helps to reveal the hidden dynamic characteristics inside the system and provide a basis for the subsequent formulation of more reasonable control strategies.

[0044] Exemplarily, in step S7, the control instruction is transmitted to the PID controller to control the motor speed corresponding to the first motion axis. It should be understood that when the central processing unit completes the collaborative analysis of the motion data of multiple motion axes and generates specific control instructions based on these analysis results, such as the recommended value for adjusting the motor speed of the first motion axis, the next task is to effectively convey this instruction to the actuator, that is, the motor and its driver installed on the motion axis. The PID controller plays the role of a bridge here: it receives the set value (i.e., the target speed) from the central processing unit and monitors the actual operating status of the motor (such as the current speed). Based on the difference between the two, the PID controller calculates the appropriate correction signal and sends it to the motor driver, thereby adjusting the working parameters of the motor so that the actual speed is as close to the set target value as possible.

[0045] Correspondingly, in the central processing unit, the motion data of the multiple motion axes are analyzed in a coordinated manner to obtain a control instruction, wherein the control instruction is a recommended value for adjusting the motor speed of the first motion axis. The technical concept of the present application is that in the central processing unit, the motion data of the multiple motion axes are integrated using data analysis and coding algorithms based on artificial intelligence, and then the time series implicit association analysis is performed on the time queue of the motion data of each motion axis after integration, and the time series dynamic compensation aggregation representation between the motion time series implicit association features of each motion axis is calculated, and then the time series implicit association features corresponding to the first motion axis are extracted, so as to intelligently obtain the control instruction for controlling the first motion axis according to the response relationship between the first motion time series implicit association features and the time series dynamic compensation aggregation representation. The present application can handle the complexity of nonlinear and dynamic changes, and effectively avoids the problem of slow response or over-regulation that may be caused by traditional PID controllers based on linear logic. This is particularly important for detecting high-density and miniaturized PCB boards, which ensures the accuracy and stability of the operation. At the same time, it can capture complex nonlinear relationships and dynamic changes, reveal the inherent connection between each motion axis, and accurately understand the dynamic behavior of the system, thereby supporting more sophisticated and intelligent control decisions.

[0046] In one embodiment, Figure 3 As shown, in step S6, in the central processing unit, the motion data of the multiple motion axes are collaboratively analyzed to obtain a control instruction, and the control instruction is a recommended value for adjusting the motor speed of the first motion axis, including: S61, data integration of the motion data of the multiple motion axes to obtain a motion data time queue of the multiple motion axes; S62, timing implicit association of the motion data time queue of the multiple motion axes to obtain motion timing implicit association feature vectors of the multiple motion axes; S63, sequence hub timing dynamic compensation aggregation of the motion timing implicit association feature vectors of the multiple motion axes to obtain a motion timing aggregation representation vector of the motion axis; S64, extracting a first motion timing implicit association feature vector from the motion timing implicit association feature vectors of the multiple motion axes, and calculating the response characteristics of the first motion timing implicit association feature vector relative to the motion timing aggregation representation vector of the motion axis to obtain a first motion timing response feature matrix; S65, obtaining the control instruction based on the first motion timing response feature matrix.

[0047] In one embodiment, Figure 4As shown, in step S61, the motion data of the multiple motion axes are integrated to obtain a time queue of motion data of the multiple motion axes, including: S611, the motion data of the multiple motion axes are regularized according to the time dimension and the motion parameter sample dimension to obtain a time queue of position data of the multiple motion axes and a time queue of speed data of the multiple motion axes; S612, speed data of the corresponding position is added after each position data in the time queue of the position data of the multiple motion axes to obtain a time queue of motion data of the multiple motion axes.

[0048] Exemplarily, in step S611, the motion data of the multiple motion axes are data regularized according to the time dimension and the motion parameter sample dimension to obtain the time queue of the position data of the multiple motion axes and the time queue of the speed data of the multiple motion axes. It should be understood that, considering that the motion data of the multiple motion axes of the flying probe tester are complex and dynamically change over time during operation, and each motion data presents different degrees of temporal dynamics, in order to clearly see the data situation at each time point, it is helpful to directly compare the relative motion situation between the axes in subsequent analysis. In the technical solution of the present application, the motion data of the multiple motion axes are data regularized according to the time dimension and the motion parameter sample dimension to organize different types of motion parameters (such as position and speed) into time series forms, respectively, to obtain the time queue of the position data of the multiple motion axes and the time queue of the speed data of the multiple motion axes.

[0049] Exemplarily, in step S612, each position data in the time queue of the position data of the multiple motion axes is added with the speed data of the corresponding position to obtain the motion data time queue of the multiple motion axes. It should be understood that, considering that in the actual operation process of the flying probe tester, the position data and speed data of the motion axis are closely related. The change of position is mutually influenced and restricted by the speed situation at the corresponding moment. For example, the motion axis is in a specific position at a certain moment, and its trend of reaching the position and continuing to move subsequently depends largely on the speed at that time. Therefore, in order to more comprehensively describe the state of the motion axis and provide richer dynamic information, in the technical solution of the present application, each position data in the time queue of the position data of the multiple motion axes is added with the speed data of the corresponding position to obtain the motion data time queue of the multiple motion axes, so that the actual motion state of the motion axis at each time point can be intuitively and completely reflected, so that the deep-level associations and laws between the motion axes can be more comprehensively and richly excavated.

[0050] Exemplarily, in step S62, the motion data time queues of the multiple motion axes are implicitly associated in time series to obtain motion time series implicit association feature vectors of the multiple motion axes. It should be understood that, considering that the motion data time queues of the multiple motion axes of the flying probe tester have obvious time series dynamic characteristics, their position and speed data change over time and there is correlation between the data of the previous and next moments, and there are different time series feature correlations between different time scales, such as in a short time window, the correlation between the motion data of the motion axis reflects the immediate operation adjustment and response; while in a longer time period, it can reveal more macroscopic trends and patterns, such as long-term stability or periodic behavior. Therefore, in order to mine the deep-level time series implicit associations contained in the motion data, so as to more comprehensively understand these complex time dynamic association characteristics, in the technical solution of the present application, the motion data time queues of the multiple motion axes are implicitly associated in time series to obtain motion time series implicit association feature vectors of the multiple motion axes, including: inputting the motion data time queues of the multiple motion axes into a sequence encoder based on the RNN-LSTM hybrid model to obtain motion time series implicit association feature vectors of the multiple motion axes. It is understandable that LSTM (Long Short-Term Memory Network) is particularly good at dealing with problems with long-term dependencies, while RNN (Recurrent Neural Network) can capture short-term patterns in sequences very well. Therefore, the RNN-LSTM hybrid model integrates the advantages of RNN in capturing short-term patterns and LSTM in dealing with long-term dependencies. Specifically, for the time queue of motion data of multiple motion axes of the flying probe tester, this hybrid model can focus on both short-term and long-term motion characteristics. In the short term, RNN can capture the rapid change pattern of position and speed data of the motion axis at a small time scale, such as in a tiny action link. For example, when the flying probe performs a small-scale precise positioning, RNN can identify the law of rapid alternating changes in position and speed in this short process. In the long term, due to the fact that the motion state of the motion axis of the flying probe tester may change over a long time span due to factors such as equipment wear and task process changes during the entire detection process. LSTM can handle this long-term dependency, such as identifying the long-term trend that the speed of the motion axis may gradually decrease as the equipment is used for a longer time, as well as the long-term synergistic change relationship between the motion axes.

[0051] Exemplarily, in step S63, the motion timing implicit association feature vectors of the multiple motion axes are aggregated by sequence hub timing dynamic compensation to obtain the motion timing aggregate representation vector of the motion axis. It should be understood that, considering that the motion timing implicit association feature vectors of the multiple motion axes contain complex time-dependent information, which reflects the correlation of various factors such as the position and speed of each motion axis at different times, and under different operating conditions, the core importance of each motion axis may change. Based on this, in order to better grasp the main features of the motion axis under different motion states, thereby enhancing the ability to understand the distribution of complex features, the present application aggregates the motion timing implicit association feature vectors of the multiple motion axes by sequence hub timing dynamic compensation aggregation to obtain the motion timing aggregate representation vector of the motion axis. In particular, this method can not only capture the complex temporal associations between the motion axes, but also dynamically compensate for the differences between the feature vectors of different motion axes according to the core information, ensuring that the aggregation process is more accurate. Specifically, this adaptive compensation mechanism can automatically correct the misjudgment that may be caused by the change of a single factor during the aggregation process, so that the final aggregate representation vector more accurately reflects the real state of the motion axis in the actual dynamic environment.

[0052] In one embodiment, Figure 5 As shown, in step S63, the motion timing implicit association feature vectors of the multiple motion axes are subjected to sequence hub timing dynamic compensation aggregation to obtain the motion timing aggregated representation vector of the motion axis, including: S631, inputting the motion timing implicit association feature vectors of the multiple motion axes into the sequence hub extraction network to obtain the motion axis motion timing sequence hub feature vector; S632, based on the motion axis motion timing sequence hub feature vector, performing attention saliency modulation on the motion timing implicit association feature vectors of each motion axis in the motion timing implicit association feature vectors of the multiple motion axes to obtain the salient motion timing node-hub complementary information embedding coding vector of the multiple motion axes; S633, fusing the motion axis motion timing sequence hub feature vector and the salient motion timing node-hub complementary information embedding coding vector of the multiple motion axes to obtain the motion timing aggregated representation vector of the motion axis.

[0053] Exemplarily, in step S631, the motion timing implicit association feature vectors of the plurality of motion axes are input into the sequence hub extraction network to obtain the motion timing sequence hub feature vector of the motion axis. Specifically, the process can be expressed by the formula:

[0054] ;

[0055] in, is the motion timing implicit correlation feature vector of the multiple motion axes, , , and are the first, second, and third implicit correlation feature vectors of the motion timing of the multiple motion axes. and The motion sequence of the motion axes implicitly correlates the eigenvectors, For calculation The sequence pivot feature vector of yes The corresponding motion time series feature weight matrix, is matrix multiplication, yes The corresponding motion timing feature bias vector, is the transformed rating vector, It is the first among the motion timing transformation scores of multiple motion axes. The motion timing transformation score of each motion axis, is the normalization function, The motion timing weight values ​​of multiple motion axes The motion timing weight value of each motion axis, is the number of vectors in the motion time series implicit association feature vectors of the multiple motion axes, It is the pivotal feature vector of the motion timing sequence of the motion axis.

[0056] Specifically, when these motion time series implicit correlation feature vectors are fed into the sequence hub extraction network, the core task of the network is to identify and extract those "hub" features that can summarize the behavior of the entire sequence. These hub features not only reflect the instantaneous state at a single time point, but more importantly, they reveal the intrinsic connection between different time points and the trend of overall dynamic evolution. For example, in the operation of a flying probe tester, although the actions of each motion axis seem independent, they actually influence each other and work together to complete a specific task. Therefore, the feature vector obtained by the sequence hub extraction network not only contains the specific performance of each motion axis at different moments, but also reflects the synergistic relationship between them and its impact on the overall performance of the system. The extraction of such hub feature vectors is of great significance for understanding the operating mechanism of complex systems. First, it helps us simplify the original high-dimensional time series data, making information that was originally difficult to process directly easy to manage and interpret. Second, because the hub feature vectors emphasize global structures and patterns rather than local details, they can resist noise interference to a certain extent and deal with outliers or missing parts in the data more robustly. This means that even if the data quality of some individual time points is not high, it will not significantly affect the quality of the hub features finally extracted.

[0057] In one embodiment, in step S632, based on the motion axis motion timing sequence hub feature vector, the motion timing implicit association feature vector of each motion axis in the motion timing implicit association feature vectors of the multiple motion axes is subjected to attention saliency modulation to obtain saliency motion timing node-hub complementary information embedding coding vectors of the multiple motion axes, including: calculating complementary information of the motion timing implicit association feature vector of each motion axis in the motion timing implicit association feature vectors of the multiple motion axes relative to the motion axis motion timing sequence hub feature vector to obtain motion timing node-hub complementary information embedding coding vectors of the multiple motion axes. Specifically, the process can be expressed as follows:

[0058] ;

[0059] in, is the point convolutional coding, is the motion timing weight matrix of the motion axis, is the activation function, yes The corresponding motion axis motion sequence implicitly associates the transformation eigenvector, is the pivot transformation feature vector of the motion axis motion timing sequence, It is subtracted by position point. Indicates taking the absolute value, is the first node-hub difference vector in the motion timing of multiple motion axes. The motion timing node-hub difference vector of the motion axis, It is the first part of the embedding encoding vector of the motion timing node-hub complementary information of multiple motion axes. The motion timing node-hub complementary information of each motion axis is embedded in the encoding vector.

[0060] The motion timing node-hub complementary information embedding coding vectors of each motion axis in the motion timing node-hub complementary information embedding coding vector of the multiple motion axes are respectively input into the complementary information saliency identification module based on the attention mechanism to obtain the motion timing complementary information attention weights of the multiple motion axes. Specifically, the process can be expressed by the formula:

[0061] ;

[0062] ;

[0063] in, yes The corresponding motion timing node-hub complementary information weight matrix, yes The corresponding motion timing node-hub complementary information bias vector, is the complementary information transformation score vector, It is the first of the motion timing attention values ​​of multiple motion axes. The motion timing attention value of each motion axis, represents the exponential function value with the natural constant e as the base, It is the attention weight of the complementary information of motion timing of multiple motion axes. Attention weights for complementary information of motion timing of the motion axes.

[0064] Based on the attention weights of the motion timing complementary information of the multiple motion axes, the motion timing node-hub complementary information embedded coding vectors of the multiple motion axes are modulated with attention to obtain the salient motion timing node-hub complementary information embedded coding vectors of the multiple motion axes. Specifically, the process can be expressed by the formula:

[0065] ;

[0066] in, , and are the first, second and third nodes of the motion timing node-hub complementary information embedding coding vector of the multiple motion axes respectively. The motion timing node-hub complementary information of the motion axis is embedded in the encoding vector, , and are the first, second and third attention weights of the complementary information of the motion timing of the multiple motion axes. The attention weights of the complementary information of the motion timing of the motion axes, It is an embedding encoding vector of the salient motion timing node-hub complementary information of multiple motion axes.

[0067] Specifically, first, the complementary information of the motion timing implicit correlation feature vector of each motion axis relative to the hub feature vector of the motion timing sequence of the motion axis is calculated to capture the unique contribution of each motion axis beyond the global pattern. This complementary information reveals the special state or change trend of a single motion axis at a specific point in time, which may be overlooked when considering the hub feature alone. The motion timing node-hub complementary information embedding encoding vector generated in this way not only contains the time dependency and local details of the original feature distribution, but also supplements important information that is not fully expressed by the hub feature. This step ensures that all valuable and unique dynamic features are retained, thereby enriching the feature representation of the entire system.

[0068] Next, these complementary information are embedded in the encoding vector and input into the complementary information saliency identification module based on the attention mechanism, with the aim of evaluating the importance of the complementary information carried by each motion axis. The attention mechanism allows the model to automatically adjust the degree of attention to different parts according to task requirements, so that important features receive higher weights, while less relevant features are given lower weights. This process is essentially a soft selection process that helps us identify which motion axis behaviors are most critical to the current task. By assigning appropriate attention weights, we can better focus on the most relevant information, reduce the impact of noise and other irrelevant factors, and further improve the quality of feature representation.

[0069] Finally, based on the calculated attention weights of the motion timing complementary information, the motion timing node-hub complementary information embedding coding vectors of the multiple motion axes are modulated for attention. This modulation process is actually a weighted summation operation on the complementary information embedding coding vectors, which enhances the influence of those features that are considered more important while suppressing less relevant features. The result is a significant motion timing node-hub complementary information embedding coding vector, which not only better reflects the actual importance of each motion axis in the entire sequence, but also retains the spatial distribution characteristics of the original features. Such encoding helps to reveal relationships and patterns that are not easy to detect when considering node features alone, ensuring that all unique and meaningful information is not overlooked.

[0070] Exemplarily, in step S633, the motion timing sequence hub feature vector of the motion axis and the significant motion timing node-hub complementary information embedding coding vector of the multiple motion axes are fused to obtain the motion timing aggregate representation vector of the motion axis. Specifically, the process can be expressed by the formula:

[0071] ;

[0072] in, For cascade processing, is the motion temporal aggregation representation vector of the motion axis.

[0073] Specifically, the motion timing sequence hub feature vector of the motion axis and the significant motion timing node-hub complementary information of the multiple motion axes are embedded in the coding vector for fusion to construct a feature representation that can reflect both global trends and local details. In a specific embodiment, feature aggregation can be achieved by feature cascading, so that the motion timing aggregation representation vector of the motion axis finally output can truly present the complementary information of the feature core information and each feature node in the feature sequence, so as to integrate the global and local information of the original feature distribution, improve the quality and relevance of the feature expression, and thus enhance the model's ability to understand complex sequence data and the expression accuracy.

[0074] In one embodiment, Figure 6 As shown, in step S64, a first motion timing implicit association feature vector is extracted from the motion timing implicit association feature vectors of the multiple motion axes, and a response feature of the first motion timing implicit association feature vector relative to the motion timing aggregate characterization vector of the motion axis is calculated to obtain a first motion timing response feature matrix, including: S641, extracting the motion timing implicit association feature vector corresponding to the first motion axis from the motion timing implicit association feature vectors of the multiple motion axes to obtain the first motion timing implicit association feature vector; S642, calculating a transfer matrix between the first motion timing implicit association feature vector and the motion timing aggregate characterization vector of the motion axis to obtain the first motion timing response feature matrix.

[0075] Exemplarily, in step S641, the motion timing implicit association feature vector corresponding to the first motion axis is extracted from the motion timing implicit association feature vectors of the multiple motion axes to obtain the first motion timing implicit association feature vector. It should be understood that, considering that each motion axis plays a different role in the multi-axis linkage system, its behavior pattern and influencing factors are also different. Therefore, in order to be able to control each motion axis more centrally and meticulously, in the technical solution of the present application, the motion timing implicit association feature vector corresponding to the first motion axis is extracted from the motion timing implicit association feature vectors of the multiple motion axes to obtain the first motion timing implicit association feature vector. In this way, extracting the feature vector of the first motion axis alone helps to more deeply understand the unique dynamic characteristics of the axis and its impact on the overall system.

[0076] Exemplarily, in step S642, the transfer matrix between the first motion timing implicitly associated feature vector and the motion timing aggregate representation vector of the motion axis is calculated to obtain the first motion timing response feature matrix. It should be understood that the first motion timing implicitly associated feature vector reflects the first motion axis's own unique motion association characteristics at different times, while the first motion timing implicitly associated feature vector comprehensively reflects the overall motion timing characteristics of all motion axes. Therefore, in order to quantify the dynamic response relationship between the first motion axis and all other motion axes, so as to understand how the changes in other axes affect the behavior of the first motion axis at different time points, in the technical solution of the present application, the transfer matrix between the first motion timing implicitly associated feature vector and the motion timing aggregate representation vector of the motion axis is calculated to obtain the first motion timing response feature matrix, so that it can be clarified how the characteristics of the first motion axis interact with the overall motion characteristics. For example, what kind of dynamic response relationship exists between the speed change characteristics of the first motion axis at a specific time point (reflected in the implicit correlation feature vector of the first motion sequence) and the comprehensive position and speed coordinated changes of the entire motion axis at the same time (reflected by the motion sequence aggregation representation vector of the motion axis), which helps to understand the impact of changes in other axes on the behavior of the first motion axis at different time points.

[0077] Exemplarily, in step S65, the control instruction is obtained based on the first motion timing response characteristic matrix. In one embodiment, the control instruction is obtained based on the first motion timing response characteristic matrix, including: the first motion timing response characteristic matrix is ​​passed through a decoder-based control instruction generator to obtain the control instruction. That is, the first motion timing response characteristic matrix obtained by transfer calculation using the first motion timing implicit association characteristic vector and the motion timing aggregate representation vector of the motion axis is decoded, so as to use its unique structure and algorithm, and the mapping mode constructed by learning a large amount of sample data to intelligently obtain the control instruction for controlling the first motion axis.

[0078] In a specific embodiment, the first motion timing response feature matrix is ​​expanded into a first motion timing response feature vector, and then a fully connected layer is used as the decoder-based control instruction generator. The fully connected layer is composed of multiple neurons, each of which is connected to all elements in the input layer to form a densely connected network structure. After receiving the data from the first motion timing response feature vector, the fully connected layer performs a linear transformation on the input according to the pre-trained weights and biases, introduces nonlinear characteristics through the activation function, and finally outputs a set of numerical values ​​representing the control instructions. In the present application, this output is the recommended value for adjusting the motor speed of the first motion axis.

[0079] Preferably, in one example, the step of passing the first motion timing response characteristic matrix through a decoder-based control instruction generator to obtain the control instruction comprises:

[0080] Expanding the first motion timing response characteristic matrix into a first motion timing response characteristic vector;

[0081] Determine the number of zero eigenvalues ​​in the first motion time series response eigenvector and zero eigenvalue , and the zero eigenfactor value and the number of zero eigenvalues Multiply and divide respectively to obtain the weight value of the first motion timing response field and the second motion time series response field weight value ;

[0082] Calculate the square root of the sum of squares of all eigenvalues ​​of the first motion timing response eigenvector to obtain the motion timing response full mode weight ,in, represents each eigenvalue of the first motion time series response eigenvector, represents the weight of the whole mode of motion timing response;

[0083] With the zero characteristic factor value Each eigenvalue of the first motion time series response eigenvector is calculated as an index The power function , and multiplied by the weight value of the first motion timing response field To obtain the intermediate value of the first motion timing response ;

[0084] Calculate each eigenvalue of the first motion time series response eigenvector The second motion timing response field weight value and the motion timing response full mode weights The product of the second motion timing response is obtained by ;

[0085] The first motion timing response median The second motion timing response is the middle value The weighted difference The first motion timing response feature vector that constitutes the optimization, and represents different weight hyperparameters;

[0086] The optimized first motion timing response feature vector is passed through a decoder-based control instruction generator to obtain the control instruction.

[0087] Here, the first motion timing implicit association feature vector and the motion timing aggregation representation vector of the motion axis respectively represent the motion pattern timing implicit association characteristics of the first motion axis and the motion pattern timing collaborative aggregation characteristics of all motion axes. When calculating the transfer matrix of the first motion timing implicit association feature vector compared with the motion timing aggregation representation vector of the motion axis, the imbalance of motion pattern timing of each motion axis will cause the transfer response richness of the first motion timing response feature vector, resulting in repetition of unstructured probability mapping, affecting the accuracy of the decoding result.

[0088] Therefore, based on the vector field fitting law of high-dimensional manifolds, the isolated zeros and their differentiable factors in the first motion timing response feature vector obtained by expanding the first motion timing response feature are used as the differentiable dimension of the high-dimensional manifold, so as to fix the normal direction pointing to the center of the feature distribution near the local coordinates represented by the eigenvalues ​​of the first motion timing response feature vector, thereby avoiding invalid repetitions in the probability mapping process caused by the unstructured characteristics of the distribution of the first motion timing response feature vector with rich interactive responses by effectively aligning the feature distribution pattern of the first motion timing response feature vector to the probability density differentiable field, thereby improving the accuracy of the control instructions obtained by the decoder-based control instruction generator.

[0089] In summary, the eight-axis linkage control method for a flying probe tester according to the embodiment of the present application is explained, which uses artificial intelligence-based data analysis and coding algorithms to integrate the motion data of multiple motion axes, and then performs a time series implicit association analysis on the motion data time queues of each motion axis after integration, and calculates the time series dynamic compensation aggregation representation between the motion time series implicit association features of each motion axis, and then extracts the time series implicit association features corresponding to the first motion axis, so as to intelligently obtain the control instructions for controlling the first motion axis according to the response relationship between the first motion time series implicit association features and the time series dynamic compensation aggregation representation. The present application can handle the complexity of nonlinear and dynamic changes, and effectively avoids the problems of slow response or over-regulation that may be caused by traditional PID controllers based on linear logic. This is particularly important for detecting high-density and miniaturized PCB boards, which ensures the accuracy and stability of the operation. At the same time, it can capture complex nonlinear relationships and dynamic changes, reveal the inherent connections between the motion axes, and accurately understand the dynamic behavior of the system, thereby supporting more sophisticated and intelligent control decisions.

[0090] Figure 7 Schematic flow chart of an eight-axis linkage control system for a flying probe tester according to an embodiment of the present application. Figure 7As shown, the eight-axis linkage control system 100 for a flying probe tester includes: a test instruction receiving module 110, which is used to receive a test instruction sent by a host computer; a test instruction parsing module 120, which is used to parse the test instruction to obtain the motion parameters of the PCB board being tested, wherein the motion parameters include moving distance, direction and speed; a motion trajectory determination module 130, which is used to determine the motion trajectories of multiple motion axes based on the motion parameters; a motion data monitoring module 140, which is used to perform corresponding actions based on the motion trajectories of the multiple motion axes and monitor the motion data of the multiple motion axes in real time, wherein the motion data includes position data and speed data; a motion data transmission module 150, which is used to transmit the motion data of the multiple motion axes to a central processing unit; a motion data collaborative analysis module 160, which is used to perform collaborative analysis on the motion data of the multiple motion axes in the central processing unit to obtain a control instruction, wherein the control instruction is a recommended value for adjusting the motor speed of the first motion axis; a control instruction transmission module 170, which is used to transmit the control instruction to a PID controller to control the motor speed corresponding to the first motion axis.

[0091] In one embodiment, the motion data collaborative analysis module includes: a motion data integration unit, which is used to integrate the motion data of the multiple motion axes to obtain the motion data time queues of the multiple motion axes; a motion data timing implicit association analysis unit, which is used to perform timing implicit association on the motion data time queues of the multiple motion axes to obtain motion timing implicit association feature vectors of the multiple motion axes; a sequence hub timing dynamic compensation aggregation unit, which is used to perform sequence hub timing dynamic compensation aggregation on the motion timing implicit association feature vectors of the multiple motion axes to obtain the motion timing aggregation representation vector of the motion axes; a response feature calculation unit, which is used to extract a first motion timing implicit association feature vector from the motion timing implicit association feature vectors of the multiple motion axes, and calculate the response feature of the first motion timing implicit association feature vector relative to the motion timing aggregation representation vector of the motion axis to obtain a first motion timing response feature matrix; and a control instruction determination unit, which is used to obtain the control instruction based on the first motion timing response feature matrix.

[0092] Here, those skilled in the art can understand that the specific operations of the various modules and units in the eight-axis linkage control system for the flying probe tester have been described in the above reference. Figures 1 to 6 The eight-axis linkage control method for a flying probe tester has been described in detail, and therefore, its repeated description will be omitted.

[0093] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0094] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0095] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0096] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An eight-axis linkage control method for a flying probe tester, characterized in that: include: Receive test instructions sent by the host computer; Parsing the test instruction to obtain motion parameters of the PCB board being tested, wherein the motion parameters include moving distance, direction and speed; Based on the motion parameters, determining motion trajectories of multiple motion axes; Execute corresponding actions based on the motion trajectories of the multiple motion axes, and monitor the motion data of the multiple motion axes in real time, wherein the motion data includes position data and speed data; Transmitting the motion data of the plurality of motion axes to a central processor; In the central processing unit, the motion data of the plurality of motion axes are collaboratively analyzed to obtain a control instruction, wherein the control instruction is a recommended value for adjusting the motor speed of the first motion axis; Transmitting the control instruction to a PID controller to control the motor speed corresponding to the first motion axis; Wherein, in the central processor, the motion data of the plurality of motion axes are collaboratively analyzed to obtain a control instruction, and the control instruction is a recommended value for adjusting the motor speed of the first motion axis, including: Performing data integration on the motion data of the multiple motion axes to obtain a time queue of motion data of the multiple motion axes; Performing implicit temporal association on the motion data time queues of the multiple motion axes to obtain motion temporal implicit association feature vectors of the multiple motion axes; Performing sequence hub timing dynamic compensation aggregation on the motion timing implicit correlation feature vectors of the multiple motion axes to obtain the motion timing aggregate representation vector of the motion axis; Extracting a first motion timing implicit association feature vector from the motion timing implicit association feature vectors of the plurality of motion axes, and calculating a response feature of the first motion timing implicit association feature vector relative to the motion timing aggregate representation vector of the motion axis to obtain a first motion timing response feature matrix; The control instruction is obtained based on the first motion timing response characteristic matrix.

2. The eight-axis linkage control method for a flying probe tester according to claim 1, characterized in that: Integrating the motion data of the plurality of motion axes to obtain a motion data time queue of the plurality of motion axes includes: Regularizing the motion data of the multiple motion axes according to the time dimension and the motion parameter sample dimension to obtain a time queue of the position data of the multiple motion axes and a time queue of the speed data of the multiple motion axes; The speed data of the corresponding position is added after each position data in the time queue of the position data of the multiple motion axes to obtain the time queue of the motion data of the multiple motion axes.

3. The eight-axis linkage control method for a flying probe tester according to claim 2, characterized in that: Performing implicit temporal association on the motion data time queues of the multiple motion axes to obtain motion timing implicit association feature vectors of the multiple motion axes, including: inputting the motion data time queues of the multiple motion axes into a sequence encoder based on an RNN-LSTM hybrid model to obtain motion timing implicit association feature vectors of the multiple motion axes.

4. The eight-axis linkage control method for a flying probe tester according to claim 3, characterized in that: Performing sequence hub temporal dynamic compensation aggregation on the motion timing implicit correlation feature vectors of the multiple motion axes to obtain the motion timing aggregate representation vector of the motion axes includes: Inputting the motion timing implicit correlation feature vectors of the plurality of motion axes into a sequence hub extraction network to obtain a motion timing sequence hub feature vector of the motion axis; Based on the motion axis motion timing sequence hub feature vector, performing attention saliency modulation on the motion timing implicit association feature vectors of each motion axis in the motion timing implicit association feature vectors of the multiple motion axes to obtain saliency motion timing node-hub complementary information embedding coding vectors of the multiple motion axes; The motion timing sequence hub feature vector of the motion axis and the significant motion timing node-hub complementary information embedding coding vector of the multiple motion axes are fused to obtain the motion timing aggregate representation vector of the motion axis.

5. The eight-axis linkage control method for a flying probe tester according to claim 4, characterized in that: Based on the motion axis motion timing sequence hub feature vector, the motion timing implicit association feature vector of each motion axis in the motion timing implicit association feature vectors of the multiple motion axes is subjected to attention saliency modulation to obtain saliency motion timing node-hub complementary information embedding coding vectors of the multiple motion axes, including: Calculating complementary information of the motion timing implicit association feature vector of each motion axis in the motion timing implicit association feature vectors of the multiple motion axes relative to the motion timing sequence hub feature vector of the motion axis to obtain the motion timing node-hub complementary information embedding coding vectors of the multiple motion axes; The motion timing node-hub complementary information embedding coding vectors of each motion axis in the motion timing node-hub complementary information embedding coding vector of the multiple motion axes are respectively input into the complementary information saliency identification module based on the attention mechanism to obtain the motion timing complementary information attention weights of the multiple motion axes; Based on the attention weights of the motion timing complementary information of the multiple motion axes, the motion timing node-hub complementary information embedded coding vectors of the multiple motion axes are attention modulated to obtain the salient motion timing node-hub complementary information embedded coding vectors of the multiple motion axes.

6. The eight-axis linkage control method for a flying probe tester according to claim 5, characterized in that: Extracting a first motion timing implicit association feature vector from the motion timing implicit association feature vectors of the plurality of motion axes, and calculating a response feature of the first motion timing implicit association feature vector relative to the motion timing aggregate representation vector of the motion axis to obtain a first motion timing response feature matrix, including: Extracting the motion timing implicit association feature vector corresponding to the first motion axis from the motion timing implicit association feature vectors of the plurality of motion axes to obtain the first motion timing implicit association feature vector; A transfer matrix between the first motion timing implicit association feature vector and the motion timing aggregate representation vector of the motion axis is calculated to obtain the first motion timing response feature matrix.

7. The eight-axis linkage control method for a flying probe tester according to claim 6, characterized in that: Based on the first motion timing response characteristic matrix, the control instruction is obtained, including: the first motion timing response characteristic matrix is ​​passed through a decoder-based control instruction generator to obtain the control instruction.

8. An eight-axis linkage control system for a flying probe tester, comprising: A test instruction receiving module is used to receive the test instruction sent by the host computer; A test instruction parsing module is used to parse the test instruction to obtain the motion parameters of the PCB board being tested, wherein the motion parameters include moving distance, direction and speed; A motion trajectory determination module, used to determine the motion trajectories of multiple motion axes based on the motion parameters; A motion data monitoring module, used to perform corresponding actions based on the motion trajectories of the multiple motion axes, and monitor the motion data of the multiple motion axes in real time, wherein the motion data includes position data and speed data; A motion data transmission module, used for transmitting the motion data of the multiple motion axes to a central processing unit; A motion data collaborative analysis module, used for collaboratively analyzing the motion data of the plurality of motion axes in the central processing unit to obtain a control instruction, wherein the control instruction is a recommended value for adjusting the motor speed of the first motion axis; A control instruction transmission module, used for transmitting the control instruction to a PID controller to control the motor speed corresponding to the first motion axis; The motion data collaborative analysis module includes: A motion data integration unit, used for integrating the motion data of the plurality of motion axes to obtain a motion data time queue of the plurality of motion axes; A motion data time series implicit association analysis unit, used for performing time series implicit association on the motion data time queues of the multiple motion axes to obtain motion time series implicit association feature vectors of the multiple motion axes; A sequence hub temporal dynamic compensation aggregation unit is used to perform sequence hub temporal dynamic compensation aggregation on the motion temporal implicit correlation feature vectors of the plurality of motion axes to obtain a motion temporal aggregation representation vector of the motion axis; a response feature calculation unit, configured to extract a first motion timing implicit association feature vector from the motion timing implicit association feature vectors of the plurality of motion axes, and calculate a response feature of the first motion timing implicit association feature vector relative to the motion timing aggregate representation vector of the motion axis to obtain a first motion timing response feature matrix; A control instruction determination unit is used to obtain the control instruction based on the first motion timing response characteristic matrix.

Images