A method and system for intelligent interpolation of path point data in robot trajectory planning

Abstract

The application provides a kind of robot trajectory planning path point data intelligent interpolation processing method and system, it is related to robot control technical field, the method includes: collection discrete path point data, wherein discrete path point data includes chip pickup point, vision alignment point and substrate placement point;Wherein, chip pickup point is the core position of chip adsorption positioning, it corresponds the preset space coordinate interval of each chip center in chip supply area;The function of vision alignment point is mechanical arm posture calibration reference, corresponding the coordinate coverage range of fixed reference mark point in camera calibration area, the function of substrate placement point is chip final mounting positioning, corresponding the preset mounting coordinate interval of each pad on substrate;Discrete path point data is constructed as dynamic regular polygon topological network model;Based on dynamic regular polygon topological network model.The application can solve the problem of insufficient trajectory continuity, low mounting precision and consistency in the prior art.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method and system for intelligent interpolation processing of path point data in robot trajectory planning. Background Technology

[0002] Existing methods for robot trajectory planning generally employ fixed-parameter interpolation kernel functions for trajectory reconstruction. This means that the range of the basis function support domain and the configuration of node vectors remain unchanged during the interpolation process, without adaptive optimization for the non-uniform curvature distribution in the actual path.

[0003] The above operating method has the following technical defects: it cannot adaptively adjust the interpolation kernel function parameters according to the path curvature distribution. This directly leads to two key problems, mainly described as follows: insufficient interpolation accuracy of high curvature feature segments, difficulty in satisfying the second-order continuous differentiability constraint of the trajectory, and sudden acceleration changes during the movement of the robotic arm, causing vibration of the end effector. Taking the chip-to-substrate bonding application as an example, the local features of key path points such as chip pickup points and vision alignment points are not specifically enhanced, and positional deviations are easily accumulated during the interpolation process, reducing the consistency of batch bonding operations. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent interpolation processing method for path point data in robot trajectory planning, which can solve the problems of insufficient trajectory continuity and low mounting accuracy and consistency in the prior art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: In a first aspect, a method for intelligent interpolation processing of path point data in robot trajectory planning is provided, the method comprising: [1] Collect discrete path point data, which includes chip pick-up points, vision alignment points, and substrate placement points. Among them, the chip pick-up point is the core position for chip adsorption and positioning, which corresponds to the preset spatial coordinate range of the center of each chip in the chip feeding area. The vision alignment point serves as a reference for robotic arm posture calibration, corresponding to the coordinate coverage range of fixed reference marker points within the camera calibration area. The substrate placement point serves as the final chip mounting and positioning, corresponding to the preset mounting coordinate range of each pad on the substrate. Discrete path point data are constructed into a dynamic regular polygon topology network model; based on the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons, and the chip picking point, vision alignment point and substrate placement point are respectively equivalent to local feature circles. Feature enhancement parameters are generated by analyzing dynamic regular polygons and local feature circles. Based on the feature enhancement parameters, spatial location correction is performed on discrete path point data to generate an optimized path point sequence; The optimized path point sequence is processed to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained. Based on curvature distribution information, the curvature and interpolation kernel function parameters are adaptively mapped. By dynamically adjusting the basis function support domain and node vector configuration of the interpolation kernel function, a dynamically adjusted interpolation kernel function is generated. The optimized path point sequence is interpolated using a dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint. The continuous trajectory that satisfies the second-order continuous differentiability constraint is then output to the robotic arm control terminal to drive the robotic arm to perform a high-precision mounting operation.

[0006] Furthermore, the discrete path point data is constructed into a dynamic regular polygon topology network model; based on the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons, and the chip pickup point, vision alignment point, and substrate placement point are respectively equivalent to local feature circles, including: The discrete path point data is processed to extract and separate critical path points; Based on critical path points, an initial framework for constructing a dynamic regular polygon topology network model is established. Within the framework of the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons; Within the framework of the same dynamic regular polygon topology network model, the chip pickup point, vision alignment point and substrate placement point contained in the critical path points are respectively equivalent to local feature circles.

[0007] Furthermore, feature enhancement parameters are generated by analyzing the dynamic regular polygon and local feature circles, including: Based on dynamic regular polygons and local feature circles, the positional relationship of the tangent points between each local feature circle and its external dynamic regular polygon is analyzed. Based on the analytically determined tangent point position relationships, calculate the theoretical distribution positions of the vertices of the regular polygons that are externally tangent to each local feature circle; By comparing the theoretical distribution location with the actual location of the corresponding vertex in the dynamic regular polygon in the current environment, the spatial position deviation between the theoretical distribution location and the actual location can be calculated. Based on the spatial position deviation, a feature enhancement parameter is generated that includes a spatial offset vector for correcting the path point position and a curvature influence weight characterizing the importance of the path point.

[0008] Furthermore, based on the feature enhancement parameters, spatial location correction is performed on the discrete path point data to generate an optimized path point sequence, including: Based on the spatial offset vector contained in the feature enhancement parameters, the coordinate positions of the corresponding path points in the discrete path point data are translated and corrected to obtain the path points after translation correction. By using the curvature influence weight in the feature enhancement parameters, the path points after translation correction are weighted and optimized to form weighted optimized path points. Integrate all weighted and optimized path points to generate an optimized path point sequence with spatial location correction.

[0009] Furthermore, the optimized path point sequence is processed to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained, including: Based on the optimized path point sequence completed by spatial location correction, local curve segments are constructed by sequentially selecting consecutive path points. For each local curve segment, the rate of change of the tangent direction at the center point of the local curve segment is calculated by analyzing the vector relationship between the coordinates of the path points constituting the local curve segment; at the same time, the arc length of the local curve segment is calculated based on the coordinates of the path points of the local curve segment. Based on the rate of change of the tangent direction and the arc length of the local curve segment, the curvature value corresponding to the center point of each local curve segment is calculated. By integrating the curvature values ​​calculated from all path points and analyzing the distribution characteristics of the curvature values ​​along the trajectory, curvature distribution information describing the overall curvature variation of the path is finally formed.

[0010] Furthermore, based on curvature distribution information, the curvature is adaptively mapped to the interpolation kernel function parameters, and the basis function support domain and node vector configuration of the interpolation kernel function are dynamically adjusted to generate a dynamically adjusted interpolation kernel function, including: Based on curvature distribution information, by identifying regions in the path where the curvature value is higher than a preset threshold, the path is divided into high curvature feature segments and low curvature transition segments, and the division results of high curvature feature segments and low curvature transition segments are generated simultaneously. Based on the division results of high curvature feature segments and low curvature transition segments, differentiated parameter mapping strategies are configured for high curvature feature segments and low curvature transition segments respectively, in order to establish the mapping relationship between curvature values ​​and interpolation kernel function parameters; Based on the mapping relationship, the curvature values ​​of each point on the path are analyzed to dynamically calculate and determine the range of the basis function support domain and the node vector configuration corresponding to each point; By integrating the basis function support domain range and node vector configuration of each point calculated on the path, a dynamically adjusted interpolation kernel function is generated that adapts to the path curvature distribution characteristics.

[0011] Furthermore, the optimized path point sequence is interpolated using a dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint. This continuous trajectory is then output to the robotic arm control unit to drive the robotic arm to perform high-precision mounting operations, including: The interpolation kernel function is dynamically adjusted to perform interpolation calculations on the optimized path point sequence, and the interpolation results are obtained. Based on the interpolation results, a continuous trajectory point sequence that meets the position and attitude requirements is generated; The second-order continuous differentiability of the continuous trajectory point sequence is verified to ensure the continuous change of the trajectory acceleration, and a verified continuous trajectory point sequence is generated. The validated continuous trajectory point sequence is output to the robotic arm control terminal to drive the coordinated movement of each axis of the robotic arm, ultimately completing the high-precision mounting operation.

[0012] Secondly, a system for intelligent interpolation processing of path point data in robot trajectory planning, wherein the system performs the method described, including: The acquisition module is used to acquire discrete path point data, which includes chip pickup points, vision alignment points, and substrate placement points. The module is used to construct a dynamic regular polygon topology network model from discrete path point data; based on the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons, and the chip picking point, vision alignment point and substrate placement point are respectively equivalent to local feature circles. The parsing module is used to generate feature enhancement parameters by parsing dynamic regular polygons and local feature circles; The correction module is used to perform spatial location correction on discrete path point data according to feature enhancement parameters, and generate an optimized path point sequence. The calculation module is used to process the optimized path point sequence to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained. The adjustment module is used to adaptively map curvature and interpolation kernel function parameters based on curvature distribution information. It generates a dynamically adjusted interpolation kernel function by dynamically adjusting the basis function support domain and node vector configuration of the interpolation kernel function. The execution module is used to interpolate the optimized path point sequence using a dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint. The continuous trajectory that satisfies the second-order continuous differentiability constraint is then output to the robotic arm control end to drive the robotic arm to perform a high-precision mounting operation.

[0013] The above-described solution of the present invention has at least the following beneficial effects: By employing techniques such as constructing a dynamic regular polygonal topology network model, generating feature enhancement parameters to correct path points, analyzing curvature distribution characteristics, adaptively mapping curvature and interpolation kernel function parameters, and dynamically adjusting the basis function support domain and node vector configuration, this approach overcomes the technical problems of existing fixed-parameter interpolation methods being unable to adapt to non-uniform distributions of path curvature, resulting in insufficient second-order continuous differentiability of the trajectory and low mounting accuracy and consistency. This achieves the technical effect of reconstructing a trajectory that satisfies second-order continuous differentiability constraints, avoiding sudden acceleration changes and end-effector vibrations during robotic arm movement, and effectively improving chip mounting accuracy and batch operation consistency. Attached Figure Description

[0014] Figure 1 A flowchart illustrating a method for intelligent interpolation of path point data in robot trajectory planning; Figure 2 This is a schematic diagram of an intelligent interpolation processing system for path point data in robot trajectory planning. Detailed Implementation

[0015] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0016] An embodiment of the present invention proposes an intelligent interpolation processing method for path point data in robot trajectory planning, the method comprising: Step 1: Collect discrete path point data, which includes chip pickup points, vision alignment points, and substrate placement points. Step 2: Construct a dynamic regular polygon topology network model from discrete path point data; Based on the dynamic regular polygon topology network model, virtually represent all discrete path points as dynamic regular polygons, and at the same time, represent chip picking points, vision alignment points and substrate placement points as local feature circles respectively. Step 3: Generate feature enhancement parameters by analyzing the dynamic regular polygon and local feature circles; Step 4: Based on the feature enhancement parameters, perform spatial location correction on the discrete path point data to generate an optimized path point sequence; Step 5: Process the optimized path point sequence to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained. Step 6: Based on curvature distribution information, adaptively map curvature and interpolation kernel function parameters, and generate dynamically adjusted interpolation kernel function by dynamically adjusting the basis function support domain and node vector configuration of the interpolation kernel function; Step 7: The optimized path point sequence is interpolated using the dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint; the continuous trajectory that satisfies the second-order continuous differentiability constraint is output to the robotic arm control end to drive the robotic arm to perform high-precision mounting operations.

[0017] In this embodiment of the invention, by constructing a dynamic regular polygonal topology network model and equivalently representing the local feature circles of key path points, and combining feature enhancement parameters to correct path points, the accumulation of positional deviations is reduced, effectively improving the positional accuracy of path points. By analyzing and optimizing the curvature distribution of the path point sequence, and dynamically adjusting the basis function support domain and node vector configuration of the interpolation kernel function, the non-uniformity of path curvature is precisely adapted to ensure that the reconstructed trajectory strictly satisfies the second-order continuous differentiability constraint. When the robotic arm moves along this trajectory, it can avoid sudden acceleration changes and end effector vibration, improve motion stability, effectively improve the accuracy of chip mounting, and stably ensure the consistency of batch mounting operations.

[0018] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Obtain raw data containing path point location information. Specifically, this includes: activating the high-definition vision sensor mounted on the robotic arm to continuously capture images of the chip feeding area and substrate mounting area from multiple perspectives, capturing images of substrate pad markings, chip pin positions, and the spatial environment within the robotic arm's movement range; simultaneously retrieving initial path planning parameters stored in the preset mounting process file, including the path node type, coordinate accuracy requirements, and task execution sequence identifier corresponding to the mounting task; converting the image pixel information acquired by the vision sensor into physical space three-dimensional coordinates; and integrating these with the retrieved initial parameters to form a raw data set containing the three-dimensional coordinate values ​​of each potential path point, acquisition timestamps, sensor measurement confidence levels, and task association identifiers.

[0019] Step 1.2 involves processing the raw data to extract all initial path points representing the robotic arm's motion path. This includes: noise suppression of the raw data to filter out abnormal data points caused by changes in ambient light and sensor jitter; identification and removal of extreme outliers exceeding reasonable value ranges based on the 3σ principle; conversion of coordinate data from different sources into a standard format under the Cartesian coordinate system according to common industrial robot control standards, retaining six decimal places to meet high-precision mounting requirements; and data integrity verification to ensure that each data entry contains three-dimensional coordinates and task association identifiers. Finally, all initial path points that completely cover the robotic arm's motion trajectory from the chip picking start position to the substrate mounting completion position are extracted from the processed raw data.

[0020] Step 1.3: Based on the preset task roles and spatial distribution of path points, identify and filter specific path points from all initial path points to serve as chip pickup points, vision alignment points, and substrate placement points, while retaining the remaining path points after identification and filtering. Specifically, this includes: defining the function of the chip pickup point as the core position for chip adsorption and positioning, corresponding to the preset spatial coordinate range of the chip center in the chip feeding area, based on the path point role classification rules established according to the preset placement task process and combined with the functional requirements of the placement task; the function of the vision alignment point as a reference for robotic arm posture calibration, corresponding to the coordinate coverage range of fixed reference markers within the camera calibration area; and the function of the substrate placement point as the final chip placement and positioning, corresponding to the preset placement coordinate range of each pad on the substrate. Combining the spatial distribution characteristics of the path points, pair all initial path points together, calculate the Euclidean distance between each pair of path points, and use a set proximity distance... The threshold is used to determine if there are redundant path points with overlapping or excessive proximity. Specific path points are selected through dual verification of task association identifiers and spatial coordinate ranges. First, based on the task association identifier of each initial path point, candidate path points labeled as chip picking, vision alignment, and substrate placement tasks are initially selected. Then, the 3D coordinates of these candidate path points are compared point by point with the preset coordinate range of the corresponding specific path points to confirm whether the coordinates fall within the corresponding interval. Only path points that meet both conditions are determined as the three types of specific path points: chip picking point, vision alignment point, and substrate placement point. Simultaneously, for the remaining initial path points that fail the dual verification, they are traversed one by one according to the initial data collection order to check whether the spatial position of each path point is on a reasonable transition path of the robotic arm's movement. After confirming that there are no duplicates or redundancies and that the path points are necessary for trajectory connection, they are retained as valid transition path points.

[0021] Step 1.4 integrates the three specific path points—chip pick-up point, vision alignment point, and substrate placement point—with the remaining path points after identification and screening to generate discrete path point data. Specifically, this includes: starting from the chip feeding area starting position according to the logical sequence of the robotic arm placement operation, proceeding through transition path points to the chip pick-up point, vision alignment point, and substrate placement point in sequence, and returning to the initial position through the final transition path point after placement is completed. The three specific path points—chip pick-up point, vision alignment point, and substrate placement point—as well as the effective transition path points are arranged in an orderly manner. A unique sequence number is assigned to each path point and its type is labeled. The three-dimensional coordinates, sequence numbers, and type labels of all path points are integrated into a structured data format to generate discrete path point data that is complete, sequential, and clearly categorized.

[0022] In this embodiment of the invention, by acquiring the original data step by step, extracting all initial path points, and then accurately selecting specific path points such as chip pickup points and retaining the remaining path points according to the preset task roles and spatial distribution, the discrete path point data is finally integrated and generated. This not only ensures the integrity of the path point data, but also highlights the relevance of key task nodes, providing a precise and reasonable data foundation for dynamic topology modeling, feature parameter analysis and other steps, and effectively improving the reliability and accuracy of trajectory interpolation processing in subsequent steps.

[0023] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1 involves processing the discrete path point data to extract and separate critical path points. First, the complete attribute information of the discrete path point data is parsed, extracting the sequence number, type label, 3D coordinates, and functional description information for each path point. Then, a functional priority rule is preset, clearly defining that core placement actions directly related to placement accuracy have higher priority than trajectory transition actions. Specifically, chip picking, vision alignment, and substrate placement-related actions are set as the highest priority, while actions only involved in trajectory transition are set as secondary priorities. Based on this, the highest priority actions directly involved in core placement are prioritized. Path points that affect the final mounting accuracy are defined as critical path points. Subsequently, an attribute comparison and filtering mechanism is established. During operation, core keywords such as pick, alignment, and placement are first matched from the functional description of the path points. Then, it is checked whether the type label corresponds to the core mounting action. If the double matching is successful, it is determined to meet the definition of critical path points. These path points are separately classified into the critical path point set. At the same time, path points that do not match the core keywords, are labeled as transitional, and only serve as trajectory connections and do not directly participate in the core mounting action are classified into the transitional path point set. Finally, the two types of path points are clearly extracted and separated.

[0024] Step 2.2: Based on the key path points, construct the initial framework of the dynamic regular polygon topology network model. Specifically, this includes: arranging the extracted key path points in an orderly manner according to the robotic arm's motion sequence corresponding to the preset sequence number in the discrete path point data; using the Cartesian coordinate system as a unified reference; calculating the straight-line distance between adjacent key path points sequentially using a spatial distance calculation method; using the straight-line distance as the basic length parameter of the topology network edge; accurately mapping the three-dimensional coordinates of each key path point to the core node of the topology network; and establishing fixed association relationships between core nodes sequentially, following the rule that nodes are directly connected when their motion sequences are adjacent. Simultaneously, on the connecting edges of adjacent core nodes, data access ports for transition path points are evenly set according to the path point distribution density. Each port reserves an association interface matching the three-dimensional coordinates of the transition path point, thus constructing the initial framework of the dynamic regular polygon topology network model that can comprehensively cover the entire trajectory space of the robotic arm from the starting position to the placement completion position and has the ability to flexibly access transition path points.

[0025] Step 2.3: Within the framework of the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons. Specifically, this includes: within the initial framework of the dynamic regular polygon topology network model, using the vertices of the regular polygons formed by the key path points as the basic reference, refining the 3D coordinates of each transition path point; during vector projection calculation, first constructing a reference vector with two adjacent reference vertices as endpoints, subtracting the starting coordinates of the reference vector from the coordinates of the transition path points to obtain the vector to be projected, and obtaining the projection coefficient by calculating the dot product of the vector to be projected and the reference vector, divided by the square of the magnitude of the reference vector, thereby determining the relative position of the transition path points on the reference vector; using... In the topology mapping method, a coordinate association table between the baseline region and the transition path points is established. Based on the projection coefficient and spatial distance weight, each transition path point is bound to the surrounding area of ​​the corresponding regular polygon edge or vertex, forming a fixed mapping relationship. At the same time, based on the overall spatial distribution density and trajectory direction of the discrete path points, the number of sides and vertex angles of the regular polygon are dynamically adjusted. The number of sides is appropriately increased in dense path point areas to improve the representation granularity, and the vertex angles are finely adjusted in trajectory turning areas to fit the actual curvature. The consistency between the representation position and the actual spatial coordinates of each path point is verified one by one. Finally, within the framework of the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons.

[0026] Step 2.4: Within the framework of the same dynamic regular polygon topology network model, the chip pickup point, visual alignment point, and substrate placement point included in the critical path points are respectively represented as local feature circles. Specifically, this includes: within the framework of the same dynamic regular polygon topology network model, using the three-dimensional coordinates of the chip pickup point, visual alignment point, and substrate placement point as the center points, precisely setting the radius of the local feature circles step by step; for the chip pickup point, first obtain the actual side length of the chip and the effective contact range of the adsorption device, and using the chip center coordinates as a reference, initially set the radius to the minimum value that can completely cover the chip adsorption surface, and then refer to the high The error tolerance threshold for precision placement is set, with a small amount of redundancy added to ensure the stability and tolerance of the adsorption action. For the visual alignment point, the theoretical radius of the effective calibration area is calculated based on the camera's calibration field of view and working distance. Combined with the accuracy requirements of visual alignment, the radius is adjusted to cover the calibration mark point and the necessary surrounding calibration area to avoid exceeding the camera's field of view and causing calibration failure. For the substrate placement point, the actual diameter of the substrate pads and the safe distance between adjacent pads are measured. The radius is set to a value slightly larger than the pad radius without touching the safe boundary of adjacent pads, ensuring that the chip can accurately fall within the target pad during placement.

[0027] Subsequently, spatial positioning fine-tuning is performed to form a tangent relationship: First, the edges or vertices of the dynamic regular polygon corresponding to each critical path point are determined through coordinate analysis. The perpendicular distance from the center of the feature circle to the corresponding edge or the straight-line distance to the corresponding vertex are calculated. By fine-tuning the three-dimensional coordinates of the center, the distance is made exactly equal to the set radius, thereby realizing the preset tangent relationship between the feature circle and the corresponding edge or vertex of the regular polygon. During the adjustment process, the position parameters of adjacent feature circles are detected in real time to avoid overlapping or deviating from the topology framework, ensuring the structural unity of the local feature circles and the topology model, while highlighting the local spatial characteristics of the critical path points. Finally, within the framework of the same dynamic regular polygon topology network model, the chip pickup point, vision alignment point, and substrate placement point contained in the critical path points are respectively equivalent to local feature circles.

[0028] The specific construction and training process of the dynamic regular polygon topology network model is as follows: The construction of a dynamic regular polygon topology network model focuses on high-precision mounting trajectory planning for a robotic arm. It centers on learning the mapping relationship between curvature distribution and dynamic interpolation kernel function parameters, and the construction process closely aligns with the characteristics of optimized path point sequences and the requirements of mounting operations. First, the core positioning and topology design logic of the network are clarified. Based on local trajectory segments and regular polygon units, each regular polygon topology unit corresponds to a continuous path interval, i.e., a high-curvature feature segment or a low-curvature transition segment. The dynamic nature of the topology is reflected in the adaptive adjustment of the number of edges, node density, and connection method of the regular polygons according to the path curvature characteristics. In the initialization phase, based on the region division results, regular polygon units with more edges are allocated to high-curvature feature segments, as the number of edges is positively correlated with the curvature value, thus ensuring the accuracy of local feature capture. Regular polygon units with fewer edges are allocated to low-curvature transition segments to simplify the structure and improve computational efficiency. Each vertex of a regular polygon corresponds to a path point or interpolation point, and the edges correspond to the relationships between adjacent points.

[0029] Next, the input and output dimensions of the network are determined. The input layer selects key feature parameters in this invention, including the three-dimensional coordinates of optimized path points, the corresponding curvature values, the region type identifier (i.e., high-curvature feature segments or low-curvature transition segments), and the path segment length. These input features directly reflect the curvature characteristics and spatial distribution of the trajectory. The output layer focuses on the core requirements of trajectory planning, outputting the key parameters of the dynamic interpolation kernel function, including the range of the basis function support domain, the node vector density, the weight adjustment coefficient, and the calibrated three-dimensional coordinates and attitude parameters of the trajectory points, ensuring that the output results can directly serve subsequent interpolation calculations and trajectory generation.

[0030] Subsequently, the network hierarchy was constructed and inter-layer functions were defined. The network consists of a three-layer core structure: the feature extraction layer is responsible for fusing features from the input curvature values ​​and coordinate data, capturing key information such as curvature interval distribution and changes in deviation between adjacent path points, providing a basis for topology adjustment; the topology adaptive adjustment layer is the core of the model, dynamically adjusting the number of edges and node connection strength of each regular polygonal unit based on the feature extraction results. For regular polygonal units in high-curvature feature segments, the number of edges is increased and the connection density between vertices is strengthened to improve the representation ability of local features, while for low-curvature transition segments, the number of edges is reduced and connection relationships are simplified to reduce redundant calculations. At the same time, the topological continuity at the connection points of trajectory segments is ensured through the vertex sharing mechanism of adjacent regular polygonal units; the mapping prediction layer maps the adjusted topological features to the target output, achieving accurate correspondence between input features and interpolation kernel function parameters and trajectory point information through hierarchical propagation. Finally, the integrity of the network structure was verified. Verify whether the topology parameters of each regular polygonal unit match the curvature characteristics of the corresponding path segment, ensuring that the node density of units in high curvature segments is higher than that in low curvature segments; verify the inter-layer data transmission logic to ensure that the input features can accurately drive topology adjustment and that the output results meet the parameter requirements of the dynamic interpolation kernel function and trajectory generation, avoiding a disconnect between the topology structure and the actual trajectory requirements.

[0031] The model training aims to improve the accuracy and adaptability of trajectory planning. The training data and training process are consistent with the technical scenario of this invention, ensuring that the trained model can effectively support dynamic interpolation and high-precision fitting.

[0032] First, a targeted training dataset was constructed. The dataset was derived from historical trajectory data across different mounting scenarios, covering optimized path point sequences for various chip sizes and substrate types. Each sample contained complete input features, including path point coordinates, curvature values, region type, path length, and label data. It also included optimal interpolation kernel function parameters verified by second-order continuous differentiability, and qualified trajectory point coordinates and attitude parameters. During data preprocessing, invalid samples with abnormal curvature or excessive trajectory deviation were removed. The input features were normalized to ensure consistent numerical ranges across different dimensions. The training and validation sets were divided according to the proportion of high-curvature feature segments and low-curvature transition segments to ensure sufficient learning for both types of regions during training.

[0033] Next, a phased training strategy was formulated. The first phase was pre-training for feature extraction and topology adaptation. The basic structure of the topology adaptive adjustment layer was fixed, and only the feature extraction layer and the mapping prediction layer were trained. This allowed the network to initially learn the mapping rules between input features and output parameters, ensuring that the model could accurately identify the feature differences between high-curvature and low-curvature regions. The second phase was dynamic topology adjustment training. The parameter constraints of the topology adaptive adjustment layer were unlocked, allowing the network to autonomously optimize the number of sides and node connection methods of regular polygonal units based on the curvature changes in the training data. During the training process, the interpolation accuracy of trajectory points in high-curvature segments and the computational efficiency of low-curvature segments were monitored. By adjusting the training weights, the model's learning of accurate representations of high-curvature segments was strengthened. The third phase was joint fine-tuning of the entire network. Based on the training results of the first two phases, the learning rate of each layer of the network was adjusted, and the entire network was globally optimized to ensure the synergy of feature extraction, topology adjustment, and mapping prediction.

[0034] During training, evaluation metrics and adjustment mechanisms tailored to task requirements are employed. The core evaluation metrics are trajectory point coordinate prediction deviation, interpolation kernel function parameter matching degree, and second-order continuous differentiability achievement rate, with real-time monitoring of metric changes in the training and validation sets. If the trajectory point deviation in high-curvature segments exceeds the standard, the training weights of high-curvature samples are increased, and the edge number adjustment threshold of the topology adaptive adjustment layer is fine-tuned. If the computational efficiency in low-curvature segments is insufficient, the topology simplification rules for low-curvature units are optimized to reduce redundant nodes and connections. After each training round, the model's generalization ability is tested using unfamiliar mounting task data in the validation set. If insufficient adaptability is observed, training samples for the corresponding scenarios are added, and the network parameters are iteratively optimized.

[0035] Finally, the model training convergence and performance verification are completed. Training is stopped when the evaluation metrics on the training and validation sets stabilize over multiple rounds, and in the generalization test, the interpolation kernel function parameters output by the model can support dynamic interpolation to generate trajectories that meet the second-order continuous differentiability requirement, and the trajectory point deviation and attitude error are both within the allowable range of mounting accuracy. The trained model is then solidified for subsequent optimization of the dynamic calculation of interpolation kernel function parameters and continuous trajectory generation for path point sequences.

[0036] In this embodiment of the invention, discrete path point data is processed to extract and separate key path points. Based on the key path points, an initial framework of a dynamic regular polygon topology network model is constructed. Within the framework, all discrete path points are virtually represented as dynamic regular polygons, and chip pickup points, visual alignment points, and substrate placement points are equivalent to local feature circles. This not only clarifies the core key nodes in the trajectory but also constructs an ordered topology structure, enhancing the local feature recognition of key path points and providing clear structural support for the accurate analysis of feature enhancement parameters and path point position correction.

[0037] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Based on the dynamic regular polygon and local feature circles, analyze the positional relationship of the tangent points between each local feature circle and its circumscribed dynamic regular polygon. Specifically, this includes: based on the constructed dynamic regular polygon topology network model and the equivalent generated local feature circles, first extracting the three-dimensional coordinates of the center of each local feature circle. and preset radius r Next, determine the circumscribed dynamic regular polygon edge corresponding to the feature circle, and obtain the two endpoints of this edge. and ; Calculate the direction vector of the side of a regular polygon ,in , , The normal vector of the edge is obtained through the cross product of vectors. Ensure the normal vector points towards the center of the feature circle; establish the parametric equations of the edges of the regular polygon, with the equations in the form of... Where X, Y, and Z are the three-dimensional coordinates of any point on the line segment, and t is a parameter with a value ranging from 0 to 1. t =0 corresponds to the endpoint P 0 The endpoint at t=1 P 1 ;x v y v z v It is a direction vector V The three components; based on the method of calculating the distance from a point to a line, using the coordinates of the center O of the circle. and endpoints , Calculate the perpendicular distance from the center of the circle to this edge using the coordinates, and verify whether the distance is consistent with the radius. r Equal equations are used to confirm the external tangency relationship; then, the parametric equation of the perpendicular line is established, with the center O as the starting point and the normal vector N as the direction, the equation in the form of: Where s is the parameter of the perpendicular line, x n y n z n These are the three components of the normal vector N. By simultaneously solving the parametric equations of the sides of the regular polygon and the parametric equations of the perpendicular line, the specific values ​​of parameters t and s are obtained. Substituting the value of t into the parametric equations of the sides of the regular polygon, the corresponding X, Y, and Z coordinates are calculated. These coordinates are the coordinates of the tangent point between the local feature circle and the sides of the dynamic regular polygon. Record the three-dimensional coordinates of the tangent point of each local feature circle, the number of the corresponding regular polygon edge, and the number of tangent points, and complete the analysis of the positional relationship of the tangent points between each local feature circle and its externally tangent dynamic regular polygon.

[0038] Step 3.2: Based on the analyzed tangent point position relationships, calculate the theoretical distribution positions of the vertices of the regular polygons externally tangent to each local feature circle. Specifically, this includes: determining the number of sides of the externally tangent dynamic regular polygon corresponding to each local feature circle based on the analyzed tangent point position relationships. n With the center of the characteristic circle Using the reference point, set the radius of the feature circle. r Directly determine the circumradius R of the circumscribed regular polygon; calculate the central angle θ between adjacent vertices, which is 360 degrees divided by the number of sides. n This yields the specific degree measure of each central angle; from the center O to the first point of tangency... The vector is used as the starting direction vector. ,in For the initial direction vector Normalization is performed to obtain unit vectors. The normalization process is to transform the vector Divide each component by the magnitude of the vector; divide the unit vector Rotate sequentially around the central angle θ, and obtain a new unit vector after each rotation. During rotation, the magnitude of the vector remains unchanged, only the direction is changed; for each rotated unit vector... Multiply each of its three components by the radius R of the circumcircle to obtain the projection length of the vector on each axis. Then add the projection length to the coordinates of the axis corresponding to the center O of the circle, i.e. Calculate each vertex V i The three-dimensional coordinates are determined; the positional relationship between each vertex and its corresponding tangent point is verified one by one to ensure that the distance from the vertex to the center O of the circle is equal to the radius R, and that the vertex and its corresponding tangent point are on the same normal vector direction. Finally, the theoretical distribution positions of the vertices of the regular polygons externally tangent to each local feature circle are calculated. .

[0039] Step 3.3 compares the theoretical distribution positions with the actual positions of the corresponding vertices in the dynamic regular polygon in the current environment to calculate the spatial position deviation between the theoretical distribution positions and the actual positions. Specifically, this includes: first, calculating the theoretical vertices according to their sequence numbers... Actual vertices in a dynamic regular polygon topology network model Pair vertices one by one to ensure that each pair corresponds to the same position identifier; extract the 3D coordinates of each pair of vertices, where the actual vertex coordinates already include actual influencing factors such as ambient light interference, robotic arm joint transmission errors, and sensor measurement errors; calculate the coordinate differences of each pair of vertices along the three axes in the Cartesian coordinate system. X Shaft difference ; Y Shaft difference ; Z Shaft difference These three differences together constitute a three-dimensional deviation vector. ; Calculate the magnitude ΔL of the deviation vector by adding the squares of ΔX, ΔY, and ΔZ, and then taking the square root of the sum. ΔL represents the straight-line distance deviation between the theoretical distribution position and the actual position. The direction of the deviation is indicated by the sign of the difference. When the difference ΔX on the X-axis is positive, it means that the actual vertex is on the positive X-axis side of the theoretical vertex, and when it is negative, it is on the negative X-axis side. The method for determining the deviation direction of the Y-axis and Z-axis is the same as that of the X-axis. The deviation calculation of all corresponding vertices is completed to form a complete set of spatial position deviations containing the deviation vector, deviation distance, deviation direction and corresponding vertex number of each vertex.

[0040] Step 3.4: Based on the spatial position deviation, quantize and generate feature enhancement parameters that include a spatial offset vector for correcting the path point position and a curvature influence weight representing the importance of the path points. Specifically, this includes: based on the set of spatial position deviations, for each deviation vector... Quantization is performed to generate a spatial offset vector; the direction information of the offset vector is preserved, that is, the positive and negative signs of ΔX, ΔY, and ΔZ are retained, and a preset precision coefficient is set. k , where the coefficient k The accuracy is determined based on the robotic arm's mounting precision requirements; for example, based on the equipment's repeatability accuracy of ±0.01mm, it is set to 0.95; the deviation distance ΔL is multiplied by the accuracy coefficient. k The adjusted offset distance is obtained. Then, by combining the directional information, the spatial offset vector is obtained. To ensure that the offset accurately offsets the actual deviation without introducing additional errors, when generating curvature influence weights, a basic weight coefficient range is first set. The basic weight coefficient range for critical path points such as chip pickup points, vision alignment points, and substrate placement points is 0.8 to 1.0, and the basic weight coefficient range for transition path points is 0.3 to 0.7. The weights are adjusted according to the deviation distance ΔL, and a weight adjustment ratio is set. For example, for every 0.005mm increase in deviation distance, the weight coefficient increases by 0.1, and the minimum basic weight is taken when the deviation distance is zero. The deviation distance is converted into a specific weight value between 0 and 1 through linear conversion. For example, the basic weight of a critical path point is 0.8, the weight is adjusted to 0.9 when the deviation distance is 0.005mm, and the weight is adjusted to 1.0 when the deviation distance is 0.01mm. Finally, the sequence number of each path point, the corresponding spatial offset vector ΔV', and the curvature influence weight are associated and bound one by one, and integrated into a structurally standardized and data-complete feature enhancement parameter according to a unified data format.

[0041] In this embodiment of the invention, by analyzing the positional relationship between the tangent points of the local feature circle and the dynamic regular polygon, the theoretical distribution position is calculated, and the spatial position deviation is obtained by comparing with the actual position. Then, feature enhancement parameters containing spatial offset vector and curvature influence weights are quantified and generated. This not only accurately captures the positional deviation information of key path points, but also clarifies the importance level of different path points, providing accurate data basis for subsequent path point position correction. At the same time, it enhances the local features of key path points, lays the foundation for the adjustment of the adaptive interpolation kernel function, and effectively improves the pertinence and accuracy of trajectory planning.

[0042] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Based on the spatial offset vector contained in the feature enhancement parameters, perform translation correction on the coordinate positions of the corresponding path points in the discrete path point data to obtain the path points after translation correction. Specifically, this includes: first, establishing a bidirectional association between the path point sequence number and the feature enhancement parameters; then, accurately retrieving the corresponding spatial offset vector from the feature enhancement parameter set using the sequence number of each discrete path point; after retrieval, verifying the matching attributes between the path point type label and the offset vector to ensure that the chip pickup point, vision alignment point, substrate placement point, and transition path point all correspond to a unique spatial offset vector; extracting the original three-dimensional coordinate values ​​of the discrete path point, recording the original coordinate data of the X-axis, Y-axis, and Z-axis respectively, and simultaneously extracting the offset values ​​of the three axes in the spatial offset vector. The offset values ​​for each axis already contain the directional information and adjustment range required to offset the deviation. Translation correction calculations are performed on the three axes of the Cartesian coordinate system. The corrected coordinate value of the X-axis is the superposition of the original X-axis coordinate value and the X-axis offset value. If the offset value is positive, the corresponding value is increased; if it is negative, the corresponding value is decreased. The correction calculation logic for the Y-axis and Z-axis is the same as that for the X-axis, and the correction result is obtained by superimposing the original coordinate value and the corresponding axial offset value. After the correction is completed, the difference between the corrected coordinate value of the path point and the theoretical position coordinate value is calculated, and it is checked whether the difference is within the preset accuracy allowable range to ensure that the deviation has been effectively offset. Finally, the path point corresponding to each discrete path point after translation correction is obtained.

[0043] Step 4.2 involves weighting the path points after translation correction using the curvature influence weight in the feature enhancement parameters to form weighted optimized path points. Specifically, this includes: clarifying the core logic of weighted optimization: path points with higher curvature influence weights have a stronger decisive effect on trajectory accuracy, and their coordinate stability should be prioritized during optimization; path points with lower weights can be adjusted appropriately to ensure overall trajectory continuity; extracting the curvature influence weight value corresponding to each translated path point, and simultaneously retrieving the corrected coordinate data and corresponding weight value of the preceding and following path points; setting a fixed weight allocation ratio: the current path point has a weight of 50%, and the two adjacent path points each have a weight of 25%. The coordinates of the three axes are weighted and calculated separately. The optimized X-axis coordinate value is calculated by multiplying the current path point's corrected X-axis coordinate value by 50%, adding the previous path point's corrected X-axis coordinate value by 25%, and adding the next path point's corrected X-axis coordinate value by 25%. The weighted calculations for the Y and Z axes follow the same logic, obtaining the optimization result by summing the products of the corresponding axis coordinate values ​​and their respective weight percentages. After optimization, critical path points with weight values ​​greater than or equal to 0.8 are checked to ensure that their coordinate values ​​maintain the accuracy of the correction. Transitional path points with weight values ​​less than or equal to 0.7 are checked to ensure that their coordinate values ​​smoothly transition with the coordinate values ​​of adjacent path points without abrupt jumps. Finally, the weighted optimized path points corresponding to each path point that has completed translation correction are formed.

[0044] Step 4.3 integrates all weighted optimized path points to generate a sequence of optimized path points with spatial position correction. This includes: sorting all weighted optimized path points one by one according to the original sequence number of the discrete path points, strictly following the movement sequence of the robotic arm. That is, starting from the starting position in the chip feeding area, the path points sequentially pass through the transition path point, chip pickup point, vision alignment point, substrate placement point, and then return to the initial position via the final transition path point, ensuring that the arrangement order of each path point is completely consistent with the actual movement sequence of the robotic arm; and conducting a three-layer verification process. The first layer is integrity verification, which counts the total number of optimized path points and compares it with the total number of original discrete path points to confirm that no path points are missing or duplicated. The second layer is continuity verification, which calculates the difference in X-axis coordinate values ​​and Y-axis coordinate values ​​between two adjacent path points. The differences in numerical values ​​and Z-axis coordinate values ​​are used to determine whether each difference is within the effective range of a single movement of the robotic arm joint. This range is calculated based on the maximum rotation angle of the robotic arm joint. If the difference in a certain axis exceeds the range, the axial coordinate value corresponding to the next path point is finely adjusted to ensure that the trajectory has no breaks or abrupt jumps. The third layer is for accuracy verification. The optimized coordinate values ​​of key path points such as chip pickup points, vision alignment points, and substrate placement points are extracted and compared with the coordinate values ​​of the theoretical optimal positions to confirm that the deviation between the two does not exceed the preset accuracy allowable range. All path points that have passed the verification are organized by sequence number. Each path point is associated with and recorded with its three-dimensional coordinate values, curvature influence weight values, and deviation changes before and after correction. These are integrated into a structured data set, and finally, an optimized path point sequence with spatial position correction is generated.

[0045] In this embodiment of the invention, the coordinates of discrete path points are translated and corrected based on the spatial offset vector in the feature enhancement parameters to accurately offset the actual position deviation. Then, the corrected path points are weighted and optimized by the curvature influence weight to highlight the importance of key path points and weaken the influence of non-critical deviations. Finally, the optimized path point sequence with spatial position correction is formed, which not only ensures the accuracy of the path point position but also enhances the trajectory stability of the core nodes. This provides the robotic arm with a logically coherent and precision-controllable motion trajectory basis, effectively improving the accuracy of trajectory planning and the reliability of the mounting operation.

[0046] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Based on the optimized path point sequence after spatial position correction, construct local curve segments by sequentially selecting continuous path points. Specifically, this includes: Based on the optimized path point sequence after spatial position correction, continuously select path points according to their original sequence numbering. Set three fixed continuous path points per group as the basic units for constructing local curve segments. This selection rule ensures that the curve segment smoothly fits the path direction without omitting or overwriting any path points. During selection, the middle path point is designated as the center point of the local curve segment, and the two preceding and following path points are designated as the two end support points of the curve segment, forming a stable structural structure of front point, center point, and rear point. If a core is encountered... Key path points such as wafer pickup points, vision alignment points, and substrate placement points are preferentially designated as the center points of local curve segments. At this time, the preceding adjacent path point of the key path point is still selected as the front end point and the following adjacent path point as the back end point. The selection does not cross points to ensure the accuracy of curve fitting in the key area. For the first path point at the beginning of the sequence, it is selected to be combined with the next two consecutive path points, with the second path point as the center point. For the last path point at the end of the sequence, it is selected to be combined with the previous two consecutive path points, with the second to last path point as the center point. This ensures that the entire optimized path point sequence has no dead corners and finally constructs all local curve segments that correspond one-to-one with the path point sequence.

[0047] Step 5.2: For each local curve segment, the rate of change of the tangent direction at the center point of the local curve segment is calculated by analyzing the vector relationship between the coordinates of the path points constituting the local curve segment. Simultaneously, the arc length of the local curve segment is calculated based on the coordinates of the path points. Specifically, for each constructed local curve segment, a curve curvature calculation algorithm is used for specialized calculations. First, the three-dimensional coordinates of the front end, center point, and rear end point of the curve segment are fully extracted, clarifying the specific positions of the three path points in the Cartesian coordinate system. When calculating the rate of change of the tangent direction, two core vectors are obtained through coordinate operations: the coordinate values ​​of the center point are subtracted from the coordinate values ​​of the front end point to obtain the vector pointing from the front end point to the center point; the coordinate values ​​of the rear end point are subtracted from the coordinate values ​​of the center point to obtain the vector pointing from the center point. The vector pointing to the rear point is used; the angle between two vectors is measured using a vector direction detection tool, and this angle value is the change in the tangent direction. Combined with the temporal interval of the path points corresponding to the two vectors in the sequence, if the path points are generated at a uniform time, they can be equivalently replaced by the number of sequence intervals; the change in the angle is divided by the temporal interval to obtain the rate of change of the tangent direction at the center point of the local curve segment; when calculating the arc length, the straight-line distance from the front point to the center point and from the center point to the rear point are calculated separately, and the two distance values ​​are added to obtain an approximate value of the arc length; if the path points corresponding to the curve segment are densely distributed, that is, the distance between adjacent path points is less than a preset threshold, then intermediate reference points are inserted between the front point and the center point, and between the center point and the rear point, and the short distances of each segment are calculated and accumulated to further improve the accuracy of the arc length calculation.

[0048] Step 5.3: Based on the rate of change of the tangent direction and the arc length of the local curve segment, calculate the curvature value corresponding to the center point of each local curve segment. Specifically, this includes: clarifying the core definition of curvature value, namely the degree of change of the tangent direction per unit length, which is a direct quantitative indicator of the curvature degree of the local curve segment. The calculation must use the rate of change of the tangent direction and the arc length as core parameters, and strictly ensure that the parameter units are consistent; first, extract the rate of change of the tangent direction of the center point of the local curve segment from the calculation results, with the unit consistent in radians; then extract the corresponding arc length of the local curve segment, with the unit consistent in millimeters; to avoid calculation errors caused by unit mismatch; use the value of the rate of change of the tangent direction as the dividend and the value of the arc length of the local curve segment as the divisor, perform a division operation, and the result is the curvature value corresponding to the center point of that local curve segment; after the calculation is completed, perform a rationality check. If the curvature value significantly exceeds the normal range of similar curve segments, such as being much larger than the conventional curvature value in the sharp curve area of ​​the trajectory, then backtrack to check whether there are deviations in the vector angle measurement, arc length calculation, etc., correct them in time, and recalculate to finally obtain an effective curvature value that accurately reflects the curvature degree of the curve segment.

[0049] Step 5.4 integrates the calculated curvature values ​​from all path points, analyzes the distribution characteristics of curvature values ​​along the trajectory, and finally forms curvature distribution information describing the overall curvature change of the path. Specifically, this includes: first, arranging the curvature values ​​corresponding to the center points of all local curve segments according to the original numbering order of their respective center points in the optimized path point sequence, establishing a correspondence table between path point numbers and curvature values ​​to ensure that each curvature value can be accurately located at its specific position in the trajectory; then analyzing the distribution characteristics of curvature values ​​along the trajectory, first statistically analyzing key data of curvature values, including the maximum, minimum, average, and median of all curvature values, to clarify the numerical range of the overall curvature distribution; and then observing the curvature... The trend of curvature value changes is analyzed to determine whether it is a gradient increase, gradient decrease, or fluctuation. The location of the peak curvature value is marked, and it is verified whether the peak location corresponds to the turning point of the trajectory or key path point areas such as chip pickup point or vision alignment point. At the same time, the curvature distribution area is divided, defining the area with curvature value close to the minimum value and the change is gentle as the gentle area, and the area with large curvature value and drastic change is defined as the sharp curve area. Finally, all analysis results are integrated to form structured information that includes the correspondence between curvature value and path point location, key statistical data, description of change trend, and regional division results. In the end, curvature distribution information that can reasonably and accurately describe the overall curvature degree change of the path is obtained.

[0050] In this embodiment of the invention, local curve segments are constructed based on the optimized path point sequence after spatial position correction. By analyzing the coordinate vector relationship of the path points, the rate of change of the tangent direction and the arc length of the center point of the local curve segment are calculated. Then, the curvature value of each local curve segment is derived and integrated to form curvature distribution information. This not only accurately captures the details of the curvature degree in each region of the path, but also clearly presents the overall distribution characteristics of curvature along the trajectory. This provides accurate curvature data support for the planning of the robotic arm's motion speed and the adjustment of joint torque, effectively avoiding motion impact at sharp bends in the trajectory, ensuring the smoothness and stability of the robotic arm's motion, and further improving the accuracy and reliability of the mounting operation.

[0051] In a preferred embodiment of the present invention, step 6 above may include: Step 6.1: Based on curvature distribution information, by identifying regions in the path where the curvature value is higher than a preset threshold, the path is divided into high-curvature feature segments and low-curvature transition segments, and the division results of high-curvature feature segments and low-curvature transition segments are generated simultaneously. Specifically, this includes: based on curvature distribution information, first clarifying the core definition and setting logic of the preset threshold; the preset threshold is a critical curvature value that distinguishes the degree of curvature of the path, and its core function is to accurately define the high-curvature feature segments, that is, the sharp bends in the trajectory or key operation areas, and the low-curvature transition segments, that is, the smooth areas of the trajectory; its setting needs to take into account multiple factors, combined with the movement limits of the robotic arm, such as the maximum joint rotation angle, the motion acceleration limit, the mounting accuracy requirements, such as the allowable error range of ±0.01mm, and at the same time referring to the historical trajectory curvature statistics of similar mounting tasks, taking the curvature of the sharp bend area and the smooth area. The median value is used as the initial threshold, and then dynamically fine-tuned according to the specific scenario such as the current chip size and substrate type to ensure that the threshold neither misses key high curvature areas nor misclassifies smooth areas as high curvature segments. Subsequently, curvature values ​​are extracted point by point according to the path point sequence number. The curvature value of each path point is compared with the preset threshold. If the curvature value of a path point is higher than the preset threshold, the path point is marked as a high curvature point; if it is lower than or equal to the preset threshold, it is marked as a low curvature point. Consecutive high curvature points form high curvature feature segments, and consecutive low curvature points form low curvature transition segments. If a single isolated high curvature point is encountered, it is merged into an adjacent high curvature feature segment to ensure the integrity of the region. The starting path point number, ending path point number, total number of path points in the segment, and average curvature value of each segment are recorded simultaneously. Finally, a division result containing all segment classification information and key parameters is generated.

[0052] Step 6.2: Based on the division results of the high-curvature feature segment and the low-curvature transition segment, configure differentiated parameter mapping strategies for the high-curvature feature segment and the low-curvature transition segment respectively, to establish the mapping relationship between curvature values ​​and interpolation kernel function parameters. Specifically, this includes: based on the division results of the high-curvature feature segment and the low-curvature transition segment, clarifying the core interpolation requirements of the two types of regions: the high-curvature feature segment needs to ensure smooth trajectory turning through accurate interpolation to avoid impact on the robotic arm's movement; the low-curvature transition segment needs to reduce computational redundancy to improve planning efficiency while ensuring trajectory smoothness. First, carry out curvature interval subdivision work. Based on the division results of the high-curvature feature segment and the low-curvature transition segment, extract the overall curvature value range of the high-curvature feature segment and divide it evenly into three continuous sub-intervals according to the curvature magnitude, corresponding to the regions with high, medium, and low curvature in the high-curvature segment respectively; extract the overall curvature value range of the low-curvature transition segment, and similarly divide it into three continuous sub-intervals, corresponding to the regions with relatively high, medium, and low curvature in the low-curvature segment respectively, to ensure that every curvature value of the entire path can accurately fall into the corresponding subdivision interval.

[0053] For each subdivided curvature interval, the rules for determining the core parameters of the interpolation kernel function are established. The key parameters of the interpolation kernel function include the basis function support domain range, node vector density, and weight adjustment coefficient; all three are correlated with the curvature value. For each sub-interval of the high-curvature feature segment, the principle is that the larger the curvature value, the higher the parameter fitting accuracy: for high-curvature sub-intervals with high curvature, the basis function support domain range is set to a smaller value, the node vector density is set to a higher value, and the weight adjustment coefficient is set to a higher value; for high-curvature sub-intervals with moderate curvature, the basis function support domain range is moderately expanded compared to the previous interval, the node vector density is correspondingly reduced, and the weight adjustment coefficient is adjusted accordingly; for high-curvature sub-intervals with low curvature... The range of the basis function support domain is appropriately expanded again, the node vector density is further reduced, and the weight adjustment coefficient is further lowered. For each sub-interval in the low-curvature transition segment, the principle is that the smaller the curvature value, the higher the parameter adaptation efficiency: for low-curvature sub-intervals with relatively high curvature, the range of the basis function support domain is set to a medium size, the node vector density is set to a medium level, and the weight adjustment coefficient is set to a medium value; for low-curvature sub-intervals with moderate curvature, the range of the basis function support domain is appropriately expanded compared to the previous interval, the node vector density is correspondingly reduced, and the weight adjustment coefficient is lowered accordingly; for low-curvature sub-intervals with relatively low curvature, the range of the basis function support domain is appropriately expanded again, the node vector density is further reduced, and the weight adjustment coefficient is further lowered.

[0054] To verify the rationality of the parameter values, typical path segments were selected for simulation testing: Path segments corresponding to sub-intervals with higher curvature in high-curvature feature segments were selected, and interpolation calculations were performed using the corresponding parameters to check whether the trajectory met the preset smoothness and positional accuracy requirements; Path segments corresponding to sub-intervals with lower curvature in low-curvature transition segments were selected, and calculations were performed using the corresponding parameters to check whether the trajectory met the preset smoothness and computational efficiency requirements; Based on the simulation results, the parameters were iteratively adjusted. If the interpolation accuracy of high-curvature sub-intervals did not meet the standard, the range of the corresponding basis function support domain was reduced and the node vector density was increased; if the computational efficiency of low-curvature sub-intervals did not meet the standard, the range of the corresponding basis function support domain was expanded and the node vector density was reduced until the parameters of all sub-intervals could balance the accuracy and efficiency requirements.

[0055] Finally, all subdivided curvature intervals, corresponding parameter value rules, value basis, and simulation verification results are organized into a structured mapping table. The mapping table includes fields such as curvature interval range, basis function support domain range value rules, node vector density value rules, weight adjustment coefficient value rules, adaptation region type, and core performance requirements. This ensures that each curvature value can be accurately matched to a unique interpolation kernel function parameter combination through table lookup. Ultimately, a mapping relationship between curvature values ​​and interpolation kernel function parameters is established that covers all curvature cases along the entire path, is logically rigorous, and can be directly called.

[0056] Step 6.3: Based on the mapping relationship, analyze the curvature values ​​of each point on the path to dynamically calculate and determine the range of the basis function support domain and the node vector configuration corresponding to each point. Specifically, this includes: based on the structured mapping table, first construct a fast positioning mechanism for curvature values ​​and subdivided intervals, and advance the analysis work point by point according to the path point sequence number; for each path point to be processed, first extract its accurate curvature value, and compare this value with the boundary values ​​of all subdivided curvature intervals in the mapping table one by one to clarify the subdivided interval to which it belongs, that is, a sub-interval under the high curvature feature segment or the low curvature transition segment, to ensure that the interval positioning is without deviation, and to provide an accurate basis for subsequent parameter calculation; when calculating the range of the basis function support domain, first retrieve the... The value selection rules corresponding to the subdivided intervals are used to obtain the allowable fluctuation range of the support domain for that interval. Based on the interval value selection rules, dynamic fine-tuning is performed by combining the current path point curvature value with its relative position within the interval: if the curvature value is close to the upper limit of the interval, i.e., a position with a higher degree of curvature within the interval, a smaller value is taken within the allowable fluctuation range to enhance the accuracy of local representation; if the curvature value is close to the lower limit of the interval, i.e., a position with a lower degree of curvature within the interval, a larger value is taken within the allowable fluctuation range to take into account the continuity of the trajectory. During the fine-tuning process, the core logic of a narrower support domain for high curvature intervals and a wider support domain for low curvature intervals is always followed to ensure that the support domain range is adapted to the curvature characteristics of the path.

[0057] When configuring node vectors, the defined support domain of the basis functions serves as the basic framework. The node density rules corresponding to the subdivision intervals are then retrieved. Nodes are evenly distributed within the support domain according to the density rules: first, the total length of the support domain is calculated; then, the total number of nodes required within the support domain is determined according to the density rules; the total length is divided equally according to the total number of nodes to obtain the spacing between each adjacent node; finally, the specific position coordinates of each node are determined to form a complete node vector. If the interval is a high-curvature related sub-interval, the density rules tend to generate more nodes, enabling the node vectors to accurately capture curve details; if it is a low-curvature related sub-interval, the density rules reduce the number of nodes, simplifying the calculation while ensuring smoothness.

[0058] After configuring the basis function support domain range and node vectors for each path point, real-time verification is performed: This checks whether the support domain range is within the allowable fluctuation range of its respective interval, whether the node spacing of the node vectors meets the density rule requirements, and whether the total number of nodes matches the support domain length. Simultaneously, referring to the parameter results of the previous processed path point, if the current parameter differs significantly from the previous parameter, a secondary fine-tuning is performed based on the curvature change trend of the preceding and following path points to ensure continuous and smooth parameter changes between adjacent path points, avoiding abrupt parameter changes. This process of parsing, calculating, verifying, and fine-tuning is completed for all path points one by one, ultimately ensuring that each path point corresponds to a unique basis function support domain range and node vector configuration that suits its curvature characteristics.

[0059] Step 6.4 integrates the basis function support domain range and node vector configuration corresponding to each point calculated on the path, and generates a dynamically adjusted interpolation kernel function that adapts to the curvature distribution characteristics of the path. Specifically, this includes: collecting the basis function support domain range, node vector configuration, and corresponding curvature value data for all path points according to their sequence number; establishing a four-dimensional association table of path parameters, support domain range, node vectors, and curvature values ​​to ensure that each configuration parameter can be accurately mapped to the specific location and curvature characteristics on the path; focusing on the smooth transition of parameters at the boundary between high curvature feature segments and low curvature transition segments, for three path points before and after the boundary, a weighted average method is used to adjust the basis function support domain range and node vector density: the number of path points from the boundary is used as the basis for weight allocation, with the near-end path points at the boundary accounting for 60% of the weight and the far-end path points accounting for 40%. The parameter values ​​of the transition area are calculated by weight superposition to avoid the robotic arm movement jamming or trajectory distortion caused by parameter abrupt changes.

[0060] A three-layer consistency check is conducted: The first layer checks the rationality of the support domain range, verifying whether the support domain range of all path points is within the interval defined by the mapping relationship. Specifically, the support domain range of high curvature feature segments ∈ [ Oh 1, Oh 2], low curvature transition segment ∈ [ Oh3, Oh 4], of which Oh 1 represents the minimum value of the support domain for the high curvature feature segment. Oh 2 The maximum value of the support domain for the high curvature feature segment. Oh 3 represents the minimum value of the support domain for the low curvature transition section. Oh 4 The maximum value of the support domain for the low curvature transition section, and Oh 1< Oh 2< Oh 3< Oh 4; The second layer verifies the uniformity of node vectors by calculating the spacing deviation of adjacent nodes within the same support domain, ensuring that the deviation does not exceed a preset threshold, such as ±0.01mm; The third layer verifies the correlation between parameters and curvature by randomly selecting 20% ​​of the path points and checking whether their support domain range and node vector density have a preset positive or negative correlation with the curvature value. If abnormal parameters are found, backtrack to step 6.3 to re-parse the curvature value and adjust the configuration.

[0061] All parameters that passed the verification were organized hierarchically according to the structural requirements of the interpolation kernel function: First, the core framework of the function was determined to be a basis function + dynamic parameter adjustment term. The basis function was selected as a B-spline basis function adapted to the trajectory planning of the robotic arm, in order to balance smoothness and computational efficiency. The dynamic parameter adjustment term consisted of the support domain range and node vector weights, ensuring that the function could be adjusted in real time with the path curvature. The final dynamically adjusted interpolation kernel function, which is adaptive to the path curvature distribution characteristics, was generated, and its formula is as follows: ; in, K ( s This represents the dynamically adjusted interpolation kernel function, with the output value being the path parameters. s The corresponding interpolation weights; s For path parameters, arc length parameterization is used, with values ​​ranging from [0, S] (where S is the total arc length of the optimized path point sequence). Each s Each value uniquely corresponds to a spatial location on the path; This indicates that the basis function terms corresponding to all nodes are superimposed and calculated. I show Node index, value range (1, ... n ), n This represents the total number of nodes within the current basis function's support domain, which changes dynamically with the range of the support domain. w i ( s ) indicates the first i The dynamic weight coefficients of each node, and the path parameters s Corresponding curvature value k ( s They are positively correlated, and the expression is:w i ( s )= a · k ( s )+ b , a This is the weighting adjustment coefficient. b Based on the base weight values, preset by the mapping relationship, high curvature feature segments a The value is greater than that of the low curvature transition section; Indicates the first i The corresponding nodes B Spline basis functions, whose shape is determined by path parameters s and dynamic support domain range Joint decision; Represents path parameters s The corresponding dynamic basis function support domain range, and curvature value k ( s They show a negative correlation, expressed as follows: (c is the support domain adjustment coefficient,) d Minimum support domain threshold, to avoid k ( s When it is too large (Approaching 0, the c-value of the high curvature characteristic segment is smaller than that of the low curvature transition segment). n The total number of nodes within the current support domain, and the support domain range. They are positively correlated, and the expression is: ( e The node density coefficient, floor The function is a floor function, ensuring the number of nodes is a positive integer, for high curvature feature segments. e The value is greater than that of the low curvature transition section.

[0062] The core adaptive characteristic of the function is reflected in: when the path parameter s Corresponding curvature value k ( s When the curvature characteristic segment increases, the dynamic support domain range increases. Shrink, total number of nodes n Increase, weighting coefficient w i ( s As the basis functions increase, The enhanced local representation capability ensures the accuracy of trajectory interpolation in sharp curve regions; when k ( s When lowering (the low curvature transition section), expand, n reduce, w i ( sAs the base value approaches the baseline value, the global smoothing capability of the basis function is improved, balancing trajectory smoothness and computational efficiency; finally, the function expression, character meaning explanation, and parameter value range are organized into a standardized document to form a complete dynamic interpolation kernel function.

[0063] In this embodiment of the invention, high-curvature feature segments and low-curvature transition segments are identified by curvature distribution information. Differentiated parameter mapping strategies are configured for the two types of regions to establish a correspondence between curvature values ​​and interpolation kernel function parameters. Based on this relationship, the range of the basis function support domain and the configuration of node vectors at each point are dynamically determined. Finally, a dynamic interpolation kernel function that adapts to the curvature distribution characteristics of the path is generated. This achieves accurate adaptation to high-curvature sharp bends and ensures the smoothness of the trajectory in low-curvature gentle bends. It avoids the problem that fixed-parameter interpolation kernel functions cannot take into account the adaptability of different curvature regions. This allows the interpolation kernel function to dynamically fit the curvature change law of the path, providing more accurate function support for the trajectory interpolation of the robotic arm and further improving the smoothness, adaptability and motion control accuracy of trajectory planning.

[0064] In a preferred embodiment of the present invention, step 7 above may include: Step 7.1 involves using a dynamically adjusted interpolation kernel function to perform interpolation calculations on the optimized path point sequence, obtaining the interpolation results. Specifically, this includes: calling the dynamically adjusted interpolation kernel function and performing interpolation calculations segment by segment on the intervals formed by adjacent optimized path points, according to the original numbering order of the optimized path point sequence; for each path point interval, first extracting the curvature values ​​of all path points within the interval, determining the type of curvature region based on the partitioning criteria, and then matching the corresponding basis function support domain range and node vector configuration. For high-curvature feature segments, high-precision interpolation logic focusing on details is used; for low-curvature transition segments, efficient interpolation logic balancing efficiency is used; taking the optimized path points at both ends of the interval as an example... Using a fixed boundary as the coordinate system, a dynamic weight allocation mechanism is employed to calculate intermediate interpolation points between two points, adapting to the curvature variation of the path. The number of interpolation points is dynamically adjusted based on the actual length of the interval and the curvature density threshold. Interpolation points are generated at closer intervals in high curvature intervals to ensure curve fit, while those in low curvature intervals are generated at sparser intervals to reduce computation. During the calculation process, the interpolation results of adjacent intervals are retrieved in real time. By comparing the coordinate changes of the first and last interpolation points of the current interval with the connection points of adjacent intervals, the positional deviation at the connection points is ensured to be within a preset smoothing threshold, avoiding breakpoints or abrupt jumps. Ultimately, a complete interpolation calculation result covering the entire path and including the three-dimensional coordinates of all interpolation points is obtained.

[0065] Step 7.2: Based on the interpolation calculation results, generate a continuous trajectory point sequence that meets the position and attitude requirements. Specifically, this includes: Based on the interpolation calculation results, first clarify the core requirements of the mounting operation: the position accuracy must match the preset coordinate thresholds of each key link in chip picking, vision alignment, and substrate placement; the attitude must ensure that the vertical angle deviation when the chip is attached to the substrate does not exceed the allowable range; perform position calibration on the three-dimensional coordinates of each interpolation point, and make minor adjustments to the interpolation points that exceed the accuracy threshold based on the real-time feedback of the relative deviation data between the substrate and the chip, ensuring that all interpolation points fall accurately on the preset path trajectory; the attitude parameters are determined based on the tangent direction of the path, combined with the preset attitude constraints of the robotic arm's end effector, to calculate the joint angle combination corresponding to each interpolation point, and the attitude parameters of adjacent interpolation points are smoothly adjusted according to the principle of equal distribution of differences to avoid sudden angle changes that cause the chip to shift or tilt during movement; associate and bind the calibrated coordinate data with the corresponding attitude parameters one by one, arrange them sequentially according to the path time sequence, remove duplicate or invalid trajectory points that exceed the reasonable range, and finally generate a continuous trajectory point sequence that simultaneously meets the requirements of position accuracy and attitude stability.

[0066] Step 7.3 involves verifying the second-order continuity differentiability of the continuous trajectory point sequence to ensure the continuous change of trajectory acceleration, generating a verified continuous trajectory point sequence. Specifically, this includes: clarifying that the core objective of second-order continuity differentiability verification is to ensure the continuous change of trajectory acceleration, fundamentally avoiding impacts, jams, or vibrations during robotic arm movement; calculating the first derivative (i.e., velocity-related parameters) based on the coordinate differences of adjacent trajectory points according to the sequence of continuous trajectory points and using a preset fixed time step; then calculating the second derivative (i.e., acceleration-related parameters) based on the changes in adjacent velocity parameters; and determining continuity through the curve of derivative value changes. For verification, first check whether the changes in velocity parameters of adjacent trajectory points are within the smoothness threshold allowed by the robotic arm joints to ensure that the velocity transition is without sudden increases or decreases; then focus on confirming whether the changes in acceleration-related parameters are continuous without discontinuities. If a sudden change in the second derivative of a certain trajectory segment is detected, immediately locate the corresponding trajectory point interval; for the interval of sudden change, fine-tune the coordinate distribution of interpolation points within the interval, or appropriately increase the number of interpolation points to optimize the curve fitting effect. After correction, recalculate the derivative and repeat the verification until the velocity and acceleration changes of the entire path meet the second-order continuous differentiability requirements, and finally generate a continuous trajectory point sequence that has been rigorously verified.

[0067] Step 7.4 involves outputting the verified continuous trajectory point sequence to the robotic arm control terminal to drive the coordinated movement of each axis of the robotic arm, ultimately completing the high-precision mounting operation. Specifically, this includes: converting the continuous trajectory point sequence, verified for second-order continuous differentiability, into a standard motion command format supported by the robotic arm control terminal. The conversion includes key control information such as the three-dimensional coordinates, joint angle combinations, movement speed, and dwell time of each trajectory point; and stably transmitting the converted command data to the robotic arm control terminal via an industrial Ethernet communication interface. After receiving the data, the control terminal decomposes the trajectory commands and simultaneously and precisely allocates the control requirements of each trajectory point to each motion axis of the robotic arm. Each motion axis dynamically adjusts its operating status based on the received instructions and the real-time operating data fed back by its own encoder, and coordinates with each other according to the timing requirements of the trajectory point sequence to ensure that the robotic arm accurately follows the continuous trajectory point sequence. During the movement, the deviation between the actual position and the commanded position of each axis is compared in real time. If the deviation exceeds the allowable range, a correction command is immediately issued until the robotic arm completes the entire process of picking up the chip from the chip feeding area, adjusting it with vision, and placing it in the designated position on the substrate. After the operation is completed, the chip after mounting is photographed to calculate its actual position and angle deviation, and to confirm that the deviation is within the preset qualified range, thus completing the high-precision mounting operation.

[0068] In this embodiment of the invention, the optimized path point sequence is interpolated by a dynamic interpolation kernel function that is adaptive to the path curvature distribution characteristics. This accurately generates a continuous trajectory point sequence that meets the position and attitude requirements. The second-order continuous differentiability verification ensures that the trajectory acceleration changes continuously without abrupt changes. Finally, the verified trajectory point sequence is output to the robotic arm control end to drive the coordinated movement of each axis. This not only ensures the high precision and smoothness of the trajectory and effectively avoids impacts and jams during the movement, but also achieves precise coordination of the joints of the robotic arm, effectively improving the position accuracy and motion consistency of the mounting operation.

[0069] like Figure 2 As shown, an intelligent interpolation processing system for path point data in robot trajectory planning includes: The acquisition module is used to acquire discrete path point data, which includes chip pickup points, vision alignment points, and substrate placement points. The module is used to construct a dynamic regular polygon topology network model from discrete path point data; based on the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons, and the chip picking point, vision alignment point and substrate placement point are respectively equivalent to local feature circles. The parsing module is used to generate feature enhancement parameters by parsing dynamic regular polygons and local feature circles; The correction module is used to perform spatial location correction on discrete path point data according to feature enhancement parameters, and generate an optimized path point sequence. The calculation module is used to process the optimized path point sequence to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained. The adjustment module is used to adaptively map curvature and interpolation kernel function parameters based on curvature distribution information. It generates a dynamically adjusted interpolation kernel function by dynamically adjusting the basis function support domain and node vector configuration of the interpolation kernel function. The execution module is used to interpolate the optimized path point sequence using a dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint. The continuous trajectory that satisfies the second-order continuous differentiability constraint is then output to the robotic arm control end to drive the robotic arm to perform a high-precision mounting operation.

[0070] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent interpolation processing of path point data in robot trajectory planning, characterized in that, The method includes: Step 1: Collect discrete path point data, which includes chip pick-up points, vision alignment points, and substrate placement points. The chip pick-up points are the core positions for chip adsorption and positioning, corresponding to the preset spatial coordinate range of the center of each chip in the chip feeding area. The vision alignment points serve as references for robotic arm posture calibration, corresponding to the coordinate coverage of fixed reference markers within the camera calibration area. The substrate placement points serve as the final chip mounting and positioning, corresponding to the preset mounting coordinate range of each pad on the substrate. Step 2: Construct a dynamic regular polygon topology network model from discrete path point data; Based on the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons, and the chip pickup point, vision alignment point and substrate placement point are respectively equivalent to local feature circles; The dynamic nature of the topology is reflected in the adaptive adjustment of the number of sides, node density and connection method of the regular polygons according to the curvature characteristics of the path. Step 3: Generate feature enhancement parameters by analyzing the dynamic regular polygon and the local feature circle; wherein, by analyzing the positional relationship of the tangent points of the local feature circle and the dynamic regular polygon, the theoretical distribution position is calculated, the spatial position deviation is obtained by comparing with the actual position, and then the feature enhancement parameters containing the spatial offset vector and the curvature influence weight representing the importance of the path point are quantified and generated. Step 4: Based on the feature enhancement parameters, perform spatial location correction on the discrete path point data to generate an optimized path point sequence; Step 5: Process the optimized path point sequence to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained. Step 6: Based on curvature distribution information, adaptively map curvature and interpolation kernel function parameters, and generate dynamically adjusted interpolation kernel function by dynamically adjusting the basis function support domain and node vector configuration of the interpolation kernel function; Step 7: The optimized path point sequence is interpolated using the dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint; the continuous trajectory that satisfies the second-order continuous differentiability constraint is output to the robotic arm control end to drive the robotic arm to perform high-precision mounting operations.

2. The intelligent interpolation processing method for path point data in robot trajectory planning according to claim 1, characterized in that, Discrete path point data are constructed into a dynamic regular polygon topology network model; based on the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons, and the chip pickup point, vision alignment point, and substrate placement point are respectively represented as local feature circles, including: Step 2.1: Process the discrete path point data to extract and separate critical path points; Step 2.2: Based on the critical path points, construct the initial framework of the dynamic regular polygon topology network model; Step 2.3: Within the framework of the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons. Step 2.4: Within the framework of the same dynamic regular polygon topology network model, the chip picking point, vision alignment point and substrate placement point contained in the critical path points are respectively equivalent to local feature circles.

3. The intelligent interpolation processing method for path point data in robot trajectory planning according to claim 2, characterized in that, By analyzing dynamic regular polygons and local feature circles, feature enhancement parameters are generated, including: Step 3.1: Based on the dynamic regular polygon and local feature circles, analyze the positional relationship of the tangent points between each local feature circle and its circumscribed dynamic regular polygon; Step 3.2: Based on the analytically derived positional relationships of the tangent points, calculate the theoretical distribution of the vertices of the regular polygons externally tangent to each local feature circle; Step 3.3: Compare the theoretical distribution location with the actual location of the corresponding vertex in the dynamic regular polygon in the current environment to calculate the spatial position deviation between the theoretical distribution location and the actual location; Step 3.4: Based on the spatial position deviation, quantize and generate feature enhancement parameters that include a spatial offset vector for correcting the path point position and a curvature influence weight characterizing the importance of the path point.

4. The intelligent interpolation processing method for path point data in robot trajectory planning according to claim 3, characterized in that, Based on feature enhancement parameters, spatial location correction is performed on discrete path point data to generate an optimized path point sequence, including: Step 4.1: Based on the spatial offset vector contained in the feature enhancement parameters, perform translation correction on the coordinate positions of the corresponding path points in the discrete path point data to obtain the path points after translation correction. Step 4.2: By using the curvature influence weight in the feature enhancement parameters, the path points after translation correction are weighted and optimized to form weighted optimized path points; Step 4.3: Integrate all weighted and optimized path points to generate an optimized path point sequence with spatial location correction completed.

5. The intelligent interpolation processing method for path point data in robot trajectory planning according to claim 4, characterized in that, The optimized path point sequence is processed to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained, including: Step 5.1: Based on the optimized path point sequence completed by spatial location correction, select consecutive path points in sequence to construct local curve segments; Step 5.2: For each local curve segment, the rate of change of the tangent direction at the center point of the local curve segment is calculated by analyzing the vector relationship between the coordinates of the path points constituting the local curve segment; at the same time, the arc length of the local curve segment is calculated based on the coordinates of the path points of the local curve segment. Step 5.3: Calculate the curvature value corresponding to the center point of each local curve segment based on the rate of change of the tangent direction and the arc length of the local curve segment; Step 5.4: Integrate the curvature values ​​calculated from all path points, analyze the distribution characteristics of the curvature values ​​along the trajectory, and finally form curvature distribution information describing the overall curvature change of the path.

6. The intelligent interpolation processing method for path point data in robot trajectory planning according to claim 5, characterized in that, Based on curvature distribution information, the curvature and interpolation kernel function parameters are adaptively mapped, and the basis function support domain and node vector configuration of the interpolation kernel function are dynamically adjusted to generate a dynamically adjusted interpolation kernel function, including: Step 6.1: Based on curvature distribution information, by identifying regions in the path where the curvature value is higher than a preset threshold, the path is divided into high curvature feature segments and low curvature transition segments, and the division results of high curvature feature segments and low curvature transition segments are generated simultaneously. Step 6.2: Based on the division results of the high curvature feature segment and the low curvature transition segment, configure differentiated parameter mapping strategies for the high curvature feature segment and the low curvature transition segment respectively, so as to establish the mapping relationship between curvature value and interpolation kernel function parameter; Step 6.3: Based on the mapping relationship, analyze the curvature values ​​of each point on the path to dynamically calculate and determine the range of the basis function support domain and the node vector configuration corresponding to each point; Step 6.4: Integrate the basis function support domain range and node vector configuration corresponding to each point calculated on the path to generate a dynamically adjusted interpolation kernel function that adapts to the path curvature distribution characteristics.

7. The intelligent interpolation processing method for path point data in robot trajectory planning according to claim 6, characterized in that, The optimized path point sequence is interpolated using a dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint. This continuous trajectory is then output to the robotic arm control unit to drive the robotic arm to perform high-precision mounting operations, including: Step 7.1: Using the dynamically adjusted interpolation kernel function, perform interpolation calculations on the optimized path point sequence to obtain the interpolation calculation results; Step 7.2: Based on the interpolation calculation results, generate a continuous trajectory point sequence that meets the position and attitude requirements; Step 7.3: Perform second-order continuous differentiability verification on the continuous trajectory point sequence to ensure that the acceleration change of the trajectory is continuous, and generate a verified continuous trajectory point sequence. Step 7.4: Output the verified continuous trajectory point sequence to the robotic arm control terminal to drive the coordinated movement of each axis of the robotic arm, and finally complete the high-precision mounting operation.

8. A system for intelligent interpolation processing of path point data in robot trajectory planning, characterized in that, The system performs the method as described in any one of claims 1 to 7, comprising: The acquisition module is used to acquire discrete path point data, which includes chip pickup points, vision alignment points, and substrate placement points. The module is used to construct a dynamic regular polygon topology network model from discrete path point data. Based on the dynamic regular polygon topology network model, all discrete path points are virtually represented as dynamic regular polygons, and the chip picking point, vision alignment point and substrate placement point are respectively equivalent to local feature circles. The dynamic nature of the topology is reflected in the adaptive adjustment of the number of sides, node density and connection method of the regular polygons according to the curvature characteristics of the path. The parsing module is used to generate feature enhancement parameters by parsing the dynamic regular polygon and the local feature circle. Specifically, by parsing the positional relationship of the tangent points of the local feature circle and the dynamic regular polygon, the theoretical distribution position is calculated, and the spatial position deviation is obtained by comparing with the actual position. Then, the feature enhancement parameters containing the spatial offset vector and the curvature influence weight representing the importance of the path point are quantified and generated. The correction module is used to perform spatial location correction on discrete path point data according to feature enhancement parameters, and generate an optimized path point sequence. The calculation module is used to process the optimized path point sequence to calculate the curvature of each path point. By analyzing the non-uniformity of the curvature distribution, the curvature distribution information is finally obtained. The adjustment module is used to adaptively map curvature and interpolation kernel function parameters based on curvature distribution information. It generates a dynamically adjusted interpolation kernel function by dynamically adjusting the basis function support domain and node vector configuration of the interpolation kernel function. The execution module is used to interpolate the optimized path point sequence using the dynamically adjusted interpolation kernel function to reconstruct a continuous trajectory that satisfies the second-order continuous differentiability constraint. The continuous trajectory that satisfies the second-order continuous differentiability constraint is output to the robot arm control end to drive the robot arm to perform high-precision mounting operations.

Images