Multi-modal Motion Control System and Method for CNC Machine Tool Loading and Unloading Robots

Abstract

The present invention discloses a multimodal motion control system and method for a CNC machine tool loading and unloading robot, relating to the field of robot motion control technology. The method comprises the following steps: during the path planning process, first, dynamically generating the motion path of each robot arm through real-time sensor feedback based on the task requirements and current workspace state of each robot arm; after generating the complete motion path, dividing the robot arm's path into a number of sub-path segments according to a fixed time window, each sub-path segment representing a small continuous motion area of the robot arm. The present invention combines real-time sensor feedback with a machine learning model to perform dynamic path prediction and conflict detection. This method not only enables timely identification and assessment of potential path conflicts, but also intelligently adjusts the time window and task execution order based on the severity of the conflict, ensuring that each robot arm performs its task within a safe time and space, reducing the occurrence of deadlock and improving overall efficiency and stability.

Description

Technical Field

[0001] The present invention relates to the technical field of robot motion control, and in particular to a multimodal motion control system and method for a CNC machine tool loading and unloading robot. Background Art

[0002] The multimodal motion of CNC machine loading and unloading robots refers to the robot's ability to coordinate multiple motion modes during the automatic loading and unloading process of CNC machine tools. It can switch or integrate different perception and motion modes according to task requirements to achieve efficient and precise operation. Specifically, multimodal motion not only includes traditional mechanical motion modes such as linear, rotational, and interpolation, but also integrates data from multiple sensors such as vision, force perception, and displacement. It uses intelligent algorithms to achieve dynamic perception and adjustment of workpiece position, posture, clamping force, etc. For example, when grabbing a workpiece, the robot can first locate the approximate position through vision, and then use force sensors to accurately adjust the clamping force and angle to ensure stable clamping of the workpiece. This technology significantly improves the flexibility and intelligence level of loading and unloading, and is suitable for a variety of complex processing environments.

[0003] In motion control, the path planning control system is mainly responsible for calculating and generating the optimal or feasible motion path from the starting point to the target point based on the robot's task requirements and environmental constraints. The system combines the robot's kinematic and dynamic models with sensor feedback information to ensure that the robot can accurately perform loading and unloading operations in complex working environments. The role of the path planning control system is to optimize the robot's motion trajectory to avoid collisions with obstacles, while meeting goals such as time efficiency and energy consumption control, and improving work accuracy and speed. Through intelligent path planning, the system can also adjust the motion path in real time to cope with different workpiece sizes, positions and posture changes, ensuring that the robot can complete various tasks flexibly and efficiently during execution. Especially in multimodal motion, it can effectively coordinate different motion modes and sensor inputs to ensure safe and efficient operations.

[0004] The existing technology has the following deficiencies:

[0005] When there are a large number of workpieces and their distribution is chaotic, existing path planning and control systems may fail to effectively predict and avoid path conflicts between multiple robotic arms. Especially when multiple workpieces are located on the same plane, multiple robotic arms may approach the same position simultaneously, causing a deadlock and preventing subsequent operations. Such problems not only cause work to stall, but can also lead to the following serious consequences:

[0006] Robotic arm collision: The path planning system fails to adjust the path in time, which may cause multiple robotic arms to collide with each other, thereby damaging the equipment and increasing repair costs;

[0007] Workpiece damage: Collision may cause the workpiece position to shift or impact during the clamping process, which may lead to damage or dislocation of the workpiece, seriously affecting the processing accuracy.

[0008] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0009] The purpose of the present invention is to provide a multimodal motion control system and method for CNC machine tool loading and unloading robots. This system utilizes dynamic path prediction and conflict detection, combining real-time sensor feedback with machine learning models. This system not only promptly identifies and assesses potential path conflicts, but also intelligently adjusts the time window and task execution sequence based on conflict severity, ensuring that each robot arm executes its task within a safe time and space. The application of time delay technology effectively prevents multiple robots from entering the same area simultaneously, reducing the occurrence of deadlock and improving overall efficiency and stability, thereby ensuring the continuity of the production line and the machining accuracy of the workpiece, thereby addressing the aforementioned background art issues.

[0010] In order to achieve the above object, the present invention provides the following technical solution: a multimodal motion control method for a CNC machine tool loading and unloading robot, comprising the following steps:

[0011] During the path planning process, the motion path of each robot arm is first dynamically generated through real-time sensor feedback based on the task requirements of each robot arm and the current workspace state;

[0012] After generating the complete motion path, the robot's path is divided into several sub-path segments according to a fixed time window. Each sub-path segment represents a small continuous motion area of the robot.

[0013] When the robot generates a new sub-path segment, it uses the pre-trained machine learning model to predict the collision of the sub-path segment.

[0014] When a path conflict is detected in a new sub-path segment, the severity of the conflict is evaluated based on the prediction results of the machine learning model. Based on the severity evaluation results of the path conflict, the original fixed time window is divided into several sub-time periods;

[0015] In each sub-time period, time delay technology is used to control the execution order and timing of tasks, ensuring that the motion path corresponding to each sub-time period only allows one robotic arm to perform the task.

[0016] Preferably, the motion path of each robotic arm is dynamically generated through real-time sensor feedback according to the task requirements of each robotic arm and the current workspace state. The specific steps are as follows:

[0017] The surrounding environment data is continuously collected through the sensors integrated on each robotic arm;

[0018] Based on the real-time collected environmental data, a dynamic environmental model is constructed. The environmental model is used to perceive the current environmental state and generate a feasible preliminary path to ensure that the robotic arm can perform tasks in a dynamic environment.

[0019] After the preliminary path is generated, the path is optimized according to the current workspace state to ensure that the path of the robot arm is always in the optimal state during execution.

[0020] Preferably, when the robotic arm generates a new sub-path segment, the path data of all sub-path segments within the fixed time window is collected, and after pre-processing the collected path data, features reflecting the conflict between the sub-path segment and other paths are extracted therefrom, wherein the extracted features include the degree of overlap between the current sub-path segment and other paths and the degree of time intersection between the current sub-path segment and other paths. Within the fixed time window, after in-depth analysis of the extracted features through feature engineering technology, path overlap index and time intersection index are generated respectively, and the path overlap index and time intersection index are used to quantify the conflict risk between the current sub-path segment and other paths.

[0021] Preferably, a feature vector consisting of a path overlap index and a time intersection index is input into a pre-trained machine learning model, a conflict coefficient is output by the machine learning model, and a conflict prediction is performed on the sub-path segment based on the conflict coefficient to determine whether the new sub-path segment has a risk of path conflict with other path segments.

[0022] Preferably, the conflict coefficient generated when the pre-trained machine learning model is used to predict the conflict of the sub-path segment is compared with a pre-set conflict coefficient reference threshold to determine whether the new sub-path segment has a risk of path conflict with other path segments. The specific determination steps are as follows:

[0023] If the conflict coefficient generated when the conflict prediction is performed on the sub-path segment is greater than the conflict coefficient reference threshold, it is determined that the new sub-path segment has a risk of path conflict with other path segments; if the conflict coefficient generated when the conflict prediction is performed on the sub-path segment is less than or equal to the conflict coefficient reference threshold, it is determined that the new sub-path segment has no risk of path conflict with other path segments.

[0024] Preferably, when a new sub-path segment is detected to have a path conflict, the original fixed time window is divided into several sub-time periods. The specific steps are as follows:

[0025] When a path conflict is detected in a new sub-path segment, the severity of the conflict is first evaluated based on the conflict coefficient predicted by the machine learning model. The conflict severity evaluation is completed using the following formula:

[0026]

[0027] ,in: Represents the conflict severity score, which is used to quantify the risk of path conflict. is the conflict coefficient predicted by the model, which reflects the conflict risk between the path segment and other paths. is the reference threshold of the conflict coefficient, which indicates the set conflict risk tolerance;

[0028] Based on the conflict severity assessment results, i.e., the conflict severity score , the original fixed time window is divided into multiple sub-time periods, each of which corresponds to a task execution phase. The specific division process is as follows:

[0029]

[0030] ,in: Indicates the length of the new sub-time period. The length of each sub-time period is dynamically adjusted according to the severity of the conflict. is the original fixed time window length.

[0031] Preferably, in each sub-time period, a time delay technique is used to control the execution order and timing of tasks, ensuring that the motion path corresponding to each sub-time period allows only one robot arm to perform the task. The specific steps are as follows:

[0032] In each sub-time period, the task execution order and priority of each robot arm are first assigned based on the conflict severity assessment results. The task execution time of each robot arm is dynamically adjusted according to the length of different sub-time periods and the urgency of the tasks, ensuring that only one robot arm executes the task in each sub-time period.

[0033] Time delay technology is used to precisely control the order of task execution to ensure that the task execution time of each robot arm does not overlap;

[0034] During the execution process, the risk of path conflict is monitored in real time. If the execution path of a certain robot arm has a potential conflict with the tasks of other robot arms, the time delay parameters are adjusted to ensure that only one robot arm is active in the conflict area.

[0035] Preferably, in a fixed time window, the specific steps of generating a path overlap index after performing an in-depth analysis of the overlap between the current sub-path segment and other paths using feature engineering technology are as follows:

[0036] First, the overlapping area is preliminarily calculated by the intersection of the path shapes. Assuming that the trajectory of each path segment is represented as a set of coordinate points, the degree of overlap between the current sub-path segment and other paths is calculated using the following formula:

[0037]

[0038] ,in: Indicates the overlapping area between the current sub-path segment and other paths, Indicates the path and The spatial intersection of and is the discretized representation of the path segment, Used to adjust the weights of paths within a given area (e.g., their length, complexity, or speed), which affect the risk of path overlap;

[0039] Based on the calculation results of the overlapping area, the path overlap index is generated by the spatial occupancy and overlapping area of the path to quantify the conflict risk between the current sub-path segment and other paths. The generation expression of the path overlap index is:

[0040]

[0041] ,in: represents the path overlap index, Is the path The area occupied in the workspace, Indicates all paths The total area occupied.

[0042] Preferably, in a fixed time window, the specific steps of generating a time intersection index after performing an in-depth analysis of the time intersection degree between the current sub-path segment and other paths through feature engineering technology are as follows:

[0043] First, the time intersection degree between the current sub-path segment and other paths within a fixed time window is analyzed by real-time path data. To this end, the time intersection interval is defined as the intersection of the time intervals of the two path segments. The time interval of the first path segment is set to The time interval of the second path segment is ,in 、 are the start and end times of the current subpath segment, 、 are the start and end times of the other path segments. If these two time intervals intersect, the length of the time intersection interval is expressed as:

[0044]

[0045] , where: The time intersection interval between two path segments measures the degree of temporal overlap between the two path segments. A longer time intersection interval indicates a higher temporal overlap between the two path segments and a greater risk of conflict.

[0046] After identifying the time intersection interval, the time intersection index is calculated by the time intersection interval length to quantify the conflict between the current sub-path segment and other paths. Assuming that within a fixed time window, the time intersection interval length of path i is , then the time intersection lengths of the current sub-path segment and all other paths can be recorded as , where n represents the total number of the current sub-path segment and other path segments, that is, how many other paths the current sub-path segment has time intersections with within a fixed time window. The expression for calculating the time intersection index using the time intersection interval is:

[0047]

[0048] ,in: is the time crossover indicator, is a weighting factor reflecting the priority of path i or the severity of the conflict (for example, paths with higher priority may be weighted more heavily), T is the total length of the fixed time window, is the length of the time crossing interval of path i at time t.

[0049] The multimodal motion control system for CNC machine tool loading and unloading robots includes a path generation module, a path division module, a conflict prediction module, a time window division module, and a time delay module;

[0050] The path generation module dynamically generates the motion path of each robot arm through real-time sensor feedback based on the task requirements of each robot arm and the current workspace state;

[0051] The path division module divides the path of the robot into several sub-path segments according to a fixed time window. Each sub-path segment represents a small continuous motion area of the robot.

[0052] The collision prediction module uses a pre-trained machine learning model to predict collisions when the robot generates a new sub-path segment.

[0053] The time window partitioning module, when a path conflict is detected in a new sub-path segment, evaluates the severity of the conflict based on the prediction results of the machine learning model and divides the original fixed time window into several sub-time segments based on the path conflict severity assessment results;

[0054] The time delay module uses time delay technology to control the execution order and timing of tasks in each sub-time period, ensuring that the motion path corresponding to each sub-time period only allows one robot arm to perform the task.

[0055] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0056] This invention combines real-time sensor feedback with machine learning models to implement dynamic path prediction and conflict detection. This not only enables timely identification and assessment of potential path conflicts, but also intelligently adjusts time windows and task execution order based on conflict severity, ensuring that each robot executes its task within a safe time and space. The application of time delay technology effectively prevents multiple robots from entering the same area simultaneously, reducing deadlock and improving overall efficiency and stability, thereby ensuring production line continuity and workpiece processing accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0058] Figure 1 This is a flow chart of the multimodal motion control method for the loading and unloading robot on a CNC machine tool of the present invention.

[0059] Figure 2 The figure is a schematic diagram of the modules of the multimodal motion control system of the CNC machine tool loading and unloading robot of the present invention. DETAILED DESCRIPTION

[0060] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0061] The present invention provides Figure 1 The multimodal motion control method of the CNC machine tool loading and unloading robot shown includes the following steps:

[0062] During the path planning process, the motion path of each robot arm is first dynamically generated through real-time sensor feedback based on the task requirements of each robot arm and the current workspace state;

[0063] In the first step of path planning, each robotic arm uses integrated sensors (such as vision sensors, depth cameras, and force sensors) to perceive its surroundings in real time. These sensors continuously collect environmental data, such as the position of the workpiece, dynamic changes in obstacles, and the robotic arm's current position and status. After data collection, the sensors transmit this information to a central control system, which uses this real-time data to build a model of the current environment. Ensuring the real-time and accuracy of this data is crucial to subsequent path planning. Through precise real-time data collection, the system can dynamically understand changes in the workspace, ensuring that path planning can adapt promptly to environmental changes.

[0064] Based on the environmental data collected in real time, the system will build a dynamic environmental model. This model usually includes the current position of the robot arm, obstacles in the workspace, the position of the workpiece, and other elements that may affect path planning. The path planning system will then use this information to generate a preliminary motion path. Taking into account the requirements of the task (such as grasping, carrying or placing), the robot arm must avoid colliding with obstacles and try to choose the optimal path. The preliminary path planning will combine kinematic constraints and task requirements to determine a reasonable motion trajectory for the robot arm. Environmental modeling ensures that the system can accurately perceive the current state and generate a feasible preliminary path to ensure that the robot arm can perform tasks in a dynamic environment.

[0065] After the initial path is generated, the system optimizes the path based on the current workspace state. At this point, the system considers multiple factors, such as the shortest path, obstacle avoidance efficiency, and the smoothness of the robot's motion. To address sudden environmental changes (such as the movement of other robots or changes in the position of the workpiece), path planning is adjusted in real time. This adjustment is achieved through sensor feedback mechanisms and dynamic optimization algorithms (such as the A algorithm and the D algorithm), ensuring that the robot's path remains optimal during execution. Real-time path optimization and adjustment can cope with uncertain environmental changes, ensuring that the robot can perform tasks flexibly and efficiently, and avoiding possible path conflicts or collisions.

[0066] After generating the complete motion path, the robot's path is divided into several sub-path segments according to a fixed time window. Each sub-path segment represents a small continuous motion area of the robot.

[0067] A subpath segment is typically a straight or curved section within a path. This division is based on two primary considerations: the physical characteristics of the path (such as displacement, velocity, and acceleration limits); and the complexity of the task (such as grasping, carrying, and placing). Within each subpath segment, the robot's motion is relatively stable, facilitating subsequent collision detection and path optimization. By dividing the path into subpath segments, the system can analyze and optimize the path segment by segment, improving the accuracy and flexibility of path planning.

[0068] Considering the physical characteristics of the path (such as displacement, speed, and acceleration limits) and the complexity of the task (such as grasping, carrying, placing, etc.) is to ensure that the path division can adapt to the robot's motion capabilities and task requirements. The physical characteristics determine the motion limitations of the robot arm, ensuring that the divided sub-path segments do not exceed the robot's motion capabilities, thereby avoiding collisions, jitters, or errors caused by too fast or too large movements. The complexity of the task determines how the path is divided. Complex tasks such as grasping and placing usually require more sophisticated path planning to ensure that the robot arm can accurately reach the target location and complete the operation. Therefore, considering these two factors helps to ensure that the path division is consistent with the robot's motion capabilities and meets the high-precision requirements of task execution, thereby improving overall work efficiency and accuracy.

[0069] When the robot arm generates a new sub-path segment, the path planning system will use the pre-trained machine learning model to predict conflicts for the sub-path segment;

[0070] When the robot arm generates a new sub-path segment, it collects path data for all sub-path segments within a fixed time window. After preprocessing the collected path data, it extracts features that reflect whether the sub-path segment conflicts with other paths. The extracted features include the degree of overlap between the current sub-path segment and other paths and the degree of temporal intersection between the current sub-path segment and other paths. Within the fixed time window, after in-depth analysis of the extracted features using feature engineering techniques, a path overlap index and a temporal intersection index are generated, respectively. The path overlap index and temporal intersection index are used to quantify the conflict risk between the current sub-path segment and other paths.

[0071] A higher degree of overlap between the current sub-path segment and other paths generally indicates a greater risk of path conflict between the current sub-path segment and other paths. This is because when multiple paths overlap spatially, the robots may occupy the same area when executing these paths, leading to collisions or interference. Path overlap increases the likelihood that the robots will enter the same space at the same time, significantly increasing the risk of path conflict. Conversely, a lower degree of path overlap means that the robots' motion trajectories and spatial occupancy are more dispersed, reducing the likelihood of conflict. Therefore, the degree of path overlap directly reflects the potential for spatial competition between robots; greater overlap indicates a higher risk of conflict.

[0072] In a fixed time window, the specific steps for generating a path overlap index by deeply analyzing the overlap between the current sub-path segment and other paths using feature engineering techniques are as follows:

[0073] First, the overlapping area is preliminarily calculated by the intersection of the path shapes. Assuming that the trajectory of each path segment is represented as a set of coordinate points, the degree of overlap between the current sub-path segment and other paths is calculated using the following formula:

[0074]

[0075] ,in: Indicates the overlapping area between the current sub-path segment and other paths, Indicates the path and The spatial intersection of and is the discretized representation of the path segment, Used to adjust the weights of paths within a given area (e.g., their length, complexity, or speed), which affect the risk of path overlap;

[0076] By calculating the path intersection and evaluating the space occupancy, the preliminary path overlap degree was obtained, reflecting the areas where conflicts may occur between paths.

[0077] Based on the calculation results of the overlapping area, the path overlap index is generated by the spatial occupancy and overlapping area of the path to quantify the conflict risk between the current sub-path segment and other paths. The generation expression of the path overlap index is:

[0078]

[0079] ,in: represents the path overlap index, Is the path The area occupied in the workspace, Indicates all paths Total area occupied;

[0080] This formula quantifies the risk of path overlap by calculating the ratio of the overlapping area to the total area occupied by all paths. Specifically, when paths occupy a larger area and have more overlapping areas, the overlap index is higher, indicating a greater risk of conflict between paths. Conversely, when paths have less overlap, the conflict risk is lower. By optimizing the path overlap index, the system can clearly reflect the degree of spatial conflict between paths.

[0081] The path overlap index, generated through feature engineering techniques that deeply analyze the degree of overlap between the current sub-path segment and other paths within a fixed time window, shows that a higher value indicates a greater risk of path conflict between the current sub-path segment and other paths. This is because the path overlap index measures the risk of path conflict by calculating the spatial overlap area between paths. A large overlap area between the current sub-path segment and other paths indicates that multiple paths are operating simultaneously in the same area, increasing the probability of collision or interference. Therefore, a high path overlap index value indicates a greater risk of conflict between these paths. A lower overlap index value indicates a lower degree of overlap between paths and a lower probability of conflict. Therefore, the path overlap index directly reflects the severity of potential conflicts between paths.

[0082] A high degree of temporal overlap between the current subpath segment and other paths indicates a greater risk of path conflict. This is because when multiple paths intersect within the same time period, the robot may enter the same work area simultaneously, resulting in spatial overlap and interference, which increases the risk of collision. If the paths do not overlap in time, or overlap only slightly, the robot's operations in the same area are staggered, reducing the possibility of path conflict. In short, a high degree of temporal overlap means that the execution timing of the paths is close or overlapping, increasing the possibility of path conflict; conversely, a low degree of temporal overlap means that the chance of conflict is low and the paths can be executed safely.

[0083] In a fixed time window, the specific steps for generating a time intersection index by deeply analyzing the degree of temporal intersection between the current sub-path segment and other paths through feature engineering technology are as follows:

[0084] First, the time intersection degree between the current sub-path segment and other paths within a fixed time window is analyzed by real-time path data. To this end, the time intersection interval is defined as the intersection of the time intervals of the two path segments. The time interval of the first path segment is set to The time interval of the second path segment is ,in 、 are the start and end times of the current subpath segment, 、 are the start and end times of the other path segments. If these two time intervals intersect, the length of the time intersection interval is expressed as:

[0085]

[0086] , where: The time intersection interval between two path segments measures the degree of temporal overlap between the two path segments. A longer time intersection interval indicates a higher temporal overlap between the two path segments and a greater risk of conflict.

[0087] The purpose of this step is to quantify the degree of intersection between two path segments in the time dimension and provide basic data for subsequent conflict assessment.

[0088] After identifying the time intersection interval, the time intersection index is calculated by the length of the time intersection interval to quantify the conflict between the current sub-path segment and other paths. Assuming that within a fixed time window, the length of the time intersection interval of path i is , then the time intersection lengths of the current sub-path segment and all other paths can be recorded as , where n represents the total number of the current sub-path segment and other path segments, that is, how many other paths the current sub-path segment has time intersections with within a fixed time window. The expression for calculating the time intersection index using the time intersection interval is:

[0089]

[0090] ,in: is the time crossover indicator, is a weighting factor reflecting the priority of path i or the severity of the conflict (for example, paths with higher priority may be weighted more heavily), T is the total length of the fixed time window, is the length of the time crossing interval of path i at time t;

[0091] Through integral calculation, the time intersection index provides a quantitative conflict risk assessment value, which quantifies the overall conflict risk through integral calculation and provides a basis for path optimization and conflict avoidance.

[0092] The time intersection index shows that within a fixed time window, the greater the performance value of the time intersection index, generated through feature engineering techniques that deeply analyze the degree of temporal intersection between the current sub-path segment and other paths, the greater the risk of path conflict between the current sub-path segment and other path segments. This is because the time intersection index quantifies the degree of temporal overlap between the current sub-path segment and other path segments within a fixed time window, reflecting their temporal intersection area. If the intersection time is long, it means that multiple path segments occupy the same space or working area within the same time period, increasing the risk of collision or interference. Therefore, a higher time intersection index means more severe temporal overlap and a greater possibility of path conflict; conversely, a lower time intersection index indicates less temporal overlap between paths and a lower risk of conflict.

[0093] When the robotic arm generates a new sub-path segment, the feature vector consisting of the path overlap index and the time intersection index is input into the pre-trained machine learning model. The machine learning model outputs the conflict coefficient, and the conflict prediction of the sub-path segment is performed based on the conflict coefficient to determine whether the new sub-path segment has the risk of path conflict with other path segments.

[0094] A pre-trained machine learning model is one that is trained using historical or simulated data in a path planning system, optimizing its learning process and enabling it to accurately predict path conflicts in real-world scenarios. In the context of this problem, the primary task of this machine learning model is to predict the risk of conflict between different path segments based on input feature vectors (such as path overlap and time intersection indices). This is achieved by outputting a conflict coefficient. The training process is based on known path data, which includes the temporal and spatial characteristics of different path segments and their conflict situations. The model uses this data to continuously adjust its internal weighting parameters, thereby learning to identify the relationships between path characteristics and conflict risk.

[0095] Machine learning models are typically trained using a series of supervised or unsupervised learning algorithms (such as regression analysis, decision trees, neural networks, and support vector machines). The training data includes various scenarios of path overlap and time intersection, as well as observed path conflict outcomes (whether collisions or interference occurred). Using this data, the model gradually adjusts its parameters to minimize the error between the predicted results and the actual situation, thereby acquiring the ability to predict new path segments. The training process for this model relies not only on static input features (such as path geometry and timing), but also considers dynamically changing factors such as the robot arm's speed and acceleration. As training progresses, the model gradually understands the key factors that lead to path conflicts under different operating conditions. Consequently, a well-trained model can more efficiently and accurately predict new path conflicts in actual operations.

[0096] In the path planning system, the pre-trained machine learning model primarily analyzes and processes the input feature vectors. When the robot generates a new sub-path segment, the system calculates the path overlap index and the temporal intersection index based on the spatial overlap and temporal intersection between the current path and other paths. These eigenvalues are combined into a feature vector, which is then input into the trained machine learning model. The model's output, the conflict coefficient, allows the system to determine whether the sub-path segment is at risk of conflict with other path segments.

[0097] In practical applications, the conflict coefficient, as the output of the machine learning model, represents the probability or risk level of path conflict. The higher the value of the conflict coefficient, the greater the risk of conflict between the current path segment and other path segments. Conversely, if the conflict coefficient is low, it can be considered that the execution risk of this path segment is low. Through continuous learning and optimization, the model can quickly predict path conflicts based on similar situations in historical data when faced with new path data and provide corresponding conflict assessment values. Through the application of this machine learning model, the path planning system can not only handle the complex path conflict problems in multi-robot collaborative tasks, but also cope with uncertain and dynamically changing working environments. As the system continues to accumulate new path data, the predictive ability of the machine learning model will continue to improve, further enhancing the path conflict warning capability.

[0098] The machine learning model is not specifically limited here, and can achieve the path overlap index and time crossover indicators Conduct comprehensive analysis to generate conflict coefficients In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; the conflict coefficient The generated expression is: , where 、 Path overlap index and time crossover indicators The weight coefficient of 、 are all greater than 0. The weight coefficient refers to the parameter used to adjust the influence of different features in the model. Specifically, and They are the path overlap index ( ) and time-crossing indicators ( ) are weight coefficients. They are used to quantify the conflict coefficients generated by these feature pairs ( ) by adjusting these weights. By adjusting these weights, the model learns which features (such as path overlap and time intersection) are more important in predicting path conflicts. The size of the weights directly affects the final conflict coefficient calculation, thereby affecting the accuracy of path conflict prediction. During model training, the weights are optimized using a learning algorithm to ensure that the model can generate accurate conflict predictions based on the input features.

[0099] It can be seen from the conflict coefficient that, within a fixed time window, the greater the performance value of the path overlap index generated after in-depth analysis of the degree of overlap between the current sub-path segment and other paths through feature engineering technology, the greater the performance value of the time intersection index generated after in-depth analysis of the degree of time intersection between the current sub-path segment and other paths through feature engineering technology. In other words, the greater the performance value of the conflict coefficient generated when conflict prediction is performed on the sub-path segment using a pre-trained machine learning model, the greater the risk of path conflict between the current sub-path segment and other paths, and vice versa.

[0100] The conflict coefficient generated when using the pre-trained machine learning model to predict the sub-path segment is compared with the pre-set conflict coefficient reference threshold to determine whether the new sub-path segment has the risk of path conflict with other path segments. The specific judgment steps are as follows:

[0101] If the conflict coefficient generated when the conflict prediction is performed on the sub-path segment is greater than the conflict coefficient reference threshold, it is determined that the new sub-path segment has a risk of path conflict with other path segments; if the conflict coefficient generated when the conflict prediction is performed on the sub-path segment is less than or equal to the conflict coefficient reference threshold, it is determined that the new sub-path segment has no risk of path conflict with other path segments.

[0102] When a path conflict is detected in a new sub-path segment, the severity of the conflict is evaluated based on the prediction results of the machine learning model. Based on the severity evaluation results of the path conflict, the original fixed time window is divided into several sub-time periods;

[0103] When a new sub-path segment is detected to have a path conflict, the original fixed time window is divided into several sub-time periods. The specific steps are as follows:

[0104] When a path conflict is detected in a new sub-path segment, the severity of the conflict is first evaluated based on the conflict coefficient predicted by the machine learning model. The conflict severity evaluation is completed using the following formula:

[0105]

[0106] ,in: Represents the conflict severity score, which is used to quantify the risk of path conflict. is the conflict coefficient predicted by the model, which reflects the conflict risk between the path segment and other paths. is the reference threshold of the conflict coefficient, which indicates the set conflict risk tolerance;

[0107] This step can normalize the conflict coefficient with the conflict coefficient reference threshold to obtain a ratio that represents the severity of the conflict. When the value of is large, it indicates that the path conflict is more serious, and the system will adopt stricter path adjustment and time window division for this sub-path segment.

[0108] Based on the conflict severity assessment results, i.e., the conflict severity score , the original fixed time window is divided into multiple sub-time periods, each of which corresponds to a task execution phase. The specific division process is as follows:

[0109]

[0110] ,in: Indicates the length of the new sub-time period. The length of each sub-time period is dynamically adjusted according to the severity of the conflict. is the original fixed time window length;

[0111] By dynamically adjusting the time window based on conflict severity scores, the system can precisely control the robot's execution time based on task complexity and the risk of path conflicts. More severe conflicts result in the time window being subdivided into multiple sub-time periods. This helps prevent multiple robots from entering the same area to perform tasks during the same time period, effectively reducing the occurrence of path conflicts.

[0112] In each sub-time period, time delay technology is used to control the execution order and timing of tasks, ensuring that the motion path corresponding to each sub-time period only allows one robot arm to perform the task;

[0113] The specific steps are as follows:

[0114] Step 1: Task scheduling and time window allocation;

[0115] Within each sub-timeframe, the order and priority of task execution for each robot arm are first assigned based on the conflict severity assessment results. Based on the length of each sub-timeframe and the urgency of the task, the task execution time of each robot arm is dynamically adjusted to ensure that only one robot arm executes a task within each sub-timeframe.

[0116] Step 2: Time delay introduction and execution synchronization;

[0117] To prevent multiple robotic arms from executing tasks simultaneously in the same area, time delay technology is used to precisely control the order in which tasks are executed. The time delay is dynamically calculated based on each robotic arm's current state and path requirements, ensuring that each robotic arm's task execution time does not overlap.

[0118] Step 3: Path conflict monitoring and adjustment;

[0119] During execution, the risk of path conflicts is monitored in real time. If a robot's execution path potentially conflicts with another robot's task, the time delay parameters are adjusted to ensure that only one robot is active within the conflicting area, thus avoiding overlapping tasks.

[0120] The above solution effectively improves the path planning accuracy and conflict avoidance capabilities of multi-arm collaboration, significantly reducing the risk of robot collisions and workpiece damage. Dynamic path prediction and conflict detection, which combines real-time sensor feedback with machine learning models, not only promptly identifies and evaluates potential path conflicts, but also intelligently adjusts the time window and task execution sequence based on the severity of the conflict, ensuring that each robot performs its task within a safe time and space. The application of time delay technology effectively prevents multiple robots from entering the same area simultaneously, reduces deadlock, and improves the efficiency and stability of the overall system, thereby ensuring the continuity of the production line and the processing accuracy of the workpiece.

[0121] The method for capturing the flight status of a flying object provided in an embodiment of the present invention is implemented by the above-mentioned system for capturing the flight status of a flying object. The specific methods and processes of the system for capturing the flight status of a flying object are detailed in the embodiment of the method for capturing the flight status of a flying object, and will not be repeated here.

[0122] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0123] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0124] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A multimodal motion control method for a CNC machine tool loading and unloading robot, characterized in that: The following steps are involved: During the path planning process, the motion path of each robot arm is dynamically generated through real-time sensor feedback based on the task requirements of each robot arm and the current workspace status; After generating the complete motion path, the robot's path is divided into several sub-path segments according to a fixed time window. Each sub-path segment represents a small continuous motion area of the robot. When the robot generates a new sub-path segment, it uses the pre-trained machine learning model to predict the collision of the sub-path segment. When a path conflict is detected in a new sub-path segment, the severity of the conflict is evaluated based on the prediction results of the machine learning model. Based on the severity evaluation results of the path conflict, the original fixed time window is divided into several sub-time periods; In each sub-time period, time delay technology is used to control the execution order and timing of tasks, ensuring that the motion path corresponding to each sub-time period only allows one robot arm to perform the task; When the robot arm generates a new sub-path segment, it collects path data for all sub-path segments within a fixed time window. After preprocessing the collected path data, it extracts features that reflect whether the sub-path segment conflicts with other paths. The extracted features include the degree of overlap between the current sub-path segment and other paths and the degree of temporal intersection between the current sub-path segment and other paths. Within the fixed time window, after in-depth analysis of the extracted features using feature engineering techniques, a path overlap index and a temporal intersection index are generated, respectively. The path overlap index and temporal intersection index are used to quantify the conflict risk between the current sub-path segment and other paths. When a new sub-path segment is detected to have a path conflict, the original fixed time window is divided into several sub-time periods. The specific steps are as follows: When a path conflict is detected in a new sub-path segment, the severity of the conflict is evaluated based on the conflict coefficient predicted by the machine learning model. The conflict severity evaluation is completed using the following formula: ,in: Represents the conflict severity score, which is used to quantify the risk of path conflict. is the conflict coefficient predicted by the model, reflecting the conflict risk between the path segment and other paths. is the reference threshold of the conflict coefficient; Based on the conflict severity assessment results, i.e., the conflict severity score , the original fixed time window is divided into multiple sub-time periods, each of which corresponds to a task execution phase. The specific division process is as follows: ,in: Indicates the length of the new sub-time period. The length of each sub-time period is dynamically adjusted according to the severity of the conflict. is the original fixed time window length.

2. The multimodal motion control method for a CNC machine tool loading and unloading robot according to claim 1, characterized in that: Based on the task requirements of each robot arm and the current workspace state, the robot arm's motion path is dynamically generated through real-time sensor feedback. The specific steps are as follows: The surrounding environment data is continuously collected through the sensors integrated on each robotic arm; Based on the real-time collected environmental data, a dynamic environmental model is constructed. The environmental model is used to perceive the current environmental state and generate a feasible preliminary path to ensure that the robotic arm can perform tasks in a dynamic environment. After the preliminary path is generated, the path is optimized according to the current workspace state to ensure that the path of the robot arm is always in the optimal state during execution.

3. The multimodal motion control method for a CNC machine tool loading and unloading robot according to claim 1, characterized in that: The feature vector consisting of the path overlap index and the time intersection index is input into a pre-trained machine learning model. The conflict coefficient is output by the machine learning model, and a conflict prediction is performed on the sub-path segment based on the conflict coefficient to determine whether the new sub-path segment has the risk of path conflict with other path segments.

4. The multimodal motion control method for a CNC machine tool loading and unloading robot according to claim 3, characterized in that: The conflict coefficient generated when using the pre-trained machine learning model to predict the sub-path segment is compared with the pre-set conflict coefficient reference threshold to determine whether the new sub-path segment has the risk of path conflict with other path segments. The specific judgment steps are as follows: If the conflict coefficient generated when the conflict prediction is performed on the sub-path segment is greater than the conflict coefficient reference threshold, it is determined that the new sub-path segment has a risk of path conflict with other path segments; if the conflict coefficient generated when the conflict prediction is performed on the sub-path segment is less than or equal to the conflict coefficient reference threshold, it is determined that the new sub-path segment has no risk of path conflict with other path segments.

5. The multimodal motion control method for a CNC machine tool loading and unloading robot according to claim 1, characterized in that: In each sub-time period, time delay technology is used to control the execution order and timing of tasks, ensuring that the motion path corresponding to each sub-time period allows only one robot arm to perform the task. The specific steps are as follows: In each sub-time period, the task execution order and priority of each robot arm are first assigned based on the conflict severity assessment results. The task execution time of each robot arm is dynamically adjusted according to the length of different sub-time periods and the urgency of the tasks, ensuring that only one robot arm executes the task in each sub-time period. Time delay technology is used to precisely control the order of task execution to ensure that the task execution time of each robot arm does not overlap; During the execution process, the risk of path conflict is monitored in real time. If the execution path of a certain robot arm has a potential conflict with the tasks of other robot arms, the time delay parameters are adjusted to ensure that only one robot arm is active in the conflict area.

6. The multimodal motion control method for a CNC machine tool loading and unloading robot according to claim 1, characterized in that: In a fixed time window, the specific steps for generating a path overlap index by deeply analyzing the overlap between the current sub-path segment and other paths using feature engineering techniques are as follows: First, a preliminary calculation of the overlapping area is performed through the intersection of the path shapes; According to the calculation results of the overlapping area, the path overlap index is generated by the spatial occupancy and overlapping area of the path to quantify the conflict risk between the current sub-path segment and other paths.

7. The multimodal motion control method for a CNC machine tool loading and unloading robot according to claim 1, characterized in that: In a fixed time window, the specific steps for generating a time intersection index by deeply analyzing the degree of temporal intersection between the current sub-path segment and other paths through feature engineering technology are as follows: Analyze the time intersection degree between the current sub-path segment and other paths within a fixed time window through real-time path data; After identifying the time intersection interval, the time intersection index is calculated according to the length of the time intersection interval to quantify the conflict between the current sub-path segment and other paths.

8. A multimodal motion control system for a CNC machine tool loading and unloading robot, used to implement the multimodal motion control method for a CNC machine tool loading and unloading robot according to any one of claims 1 to 7, characterized in that: It includes a path generation module, a path division module, a conflict prediction module, a time window division module and a time delay module; The path generation module dynamically generates the motion path of each robot arm through real-time sensor feedback based on the task requirements of each robot arm and the current workspace state; The path division module divides the path of the robot into several sub-path segments according to a fixed time window. Each sub-path segment represents a small continuous motion area of the robot. The collision prediction module uses a pre-trained machine learning model to predict collisions when the robot generates a new sub-path segment. The time window partitioning module, when a path conflict is detected in a new sub-path segment, evaluates the severity of the conflict based on the prediction results of the machine learning model and divides the original fixed time window into several sub-time segments based on the path conflict severity assessment results; The time delay module uses time delay technology to control the execution order and timing of tasks in each sub-time period, ensuring that the motion path corresponding to each sub-time period only allows one robot arm to perform the task.

Images