Workpiece running path dynamic planning and scheduling system for intelligent equipment manufacturing workshop

Abstract

The application provides a kind of intelligent equipment manufacturing workshop workpiece operation path dynamic planning and scheduling system, including the initial search ellipsoid is constructed according to workpiece start-stop coordinates;When ellipsoid exists overlap, according to priority, determine the affected source and affected workpiece, and based on the space-time occupation information of the path of the affected source, construct dynamic probability density field.Affected workpiece samples candidate nodes in its search ellipsoid, determines the acceptance probability according to the field density, screens effective nodes and connects the minimum cost parent node to generate path segment.At the same time, the system updates the density field according to the space-time conflict penalty of the path, and extracts the spatial projection of the space-time bounding box as the disturbance range after updating the path tree, triggers local re-planning for the workpieces that meet the priority conditions and intersect with the disturbance range, to reduce the risk of space-time conflict in the process of multi-workpiece transfer.

Description

Technical Field

[0001] This application belongs to the field of scheduling, and in particular relates to a dynamic planning and scheduling system for the movement path of workpieces in an intelligent equipment manufacturing workshop. Background Technology

[0002] With the continuous development of smart factories and intelligent workshops, material handling and workpiece transfer tasks are becoming increasingly frequent, parallel, and intensive. In actual production environments, multiple workpieces typically need to move continuously between different workstations, processing centers, and storage areas, and multiple intelligent handling devices often need to operate collaboratively within shared workshop spaces. Due to the compact workshop layout, limited passage space, and high task cycle requirements, concurrent operation of multiple workpieces can easily lead to path intersections, area overlaps, or passage conflicts in spatial location and time, resulting in problems such as traffic congestion, collision risks, or local deadlocks. Therefore, how to quickly plan operating paths for multiple workpieces in a complex and dynamic workshop environment that balances passage efficiency, task priority, and spatiotemporal safety has become a crucial problem that intelligent workshop logistics scheduling systems need to solve.

[0003] Existing path planning strategies typically focus on static spatial obstacle avoidance or single-workpiece path optimization, failing to adequately consider the coupling relationships between temporal, spatial, and task priorities. When multiple workpiece paths intersect, traditional methods often struggle to accurately represent the spatiotemporal occupancy status of different workpiece paths and to promptly determine the impact range of a local update on surrounding workpieces. Consequently, the system may require computationally intensive global path replanning or employ pause or waiting methods to handle conflicts, impacting overall traffic efficiency and scheduling stability. Therefore, it is necessary to propose a dynamic planning and scheduling scheme for workpiece movement paths that can assess path conflict risks within a unified spatiotemporal framework and implement targeted replanning when local disturbances occur, thereby improving the safety and operational efficiency of parallel multi-workpiece transportation processes. Summary of the Invention

[0004] To address the shortcomings of existing technologies in expressing spatiotemporal occupancy relationships during parallel multi-workpiece transfer, imperfect priority avoidance mechanisms, and high computational overhead for replanning after local path disturbances, this invention provides, in a first aspect, a dynamic planning and scheduling system for workpiece movement paths in an intelligent equipment manufacturing workshop, comprising the following modules:

[0005] The construction module is used to construct an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece. For initial search ellipsoids that overlap, the high-priority workpiece is set as the influence source and the low-priority workpiece is set as the affected body according to the priority. The spatiotemporal occupancy information of the planned path of the influence source is constructed into a dynamic probability density field in the overlapping area. The probability density value of each point in the dynamic probability density field is determined based on the shortest spatiotemporal distance from the corresponding point to the influence source path. The calculation module is used to sample and generate candidate nodes within the initial search ellipsoid of the affected body, determine the acceptance probability based on the probability density value of the candidate node in the dynamic probability density field, and when the candidate node is determined to be accepted based on the acceptance probability, the candidate node is taken as a valid node. The module searches for the minimum cost parent node in the path tree to connect and generate a new path segment, calculates the spatiotemporal conflict penalty value of the new path segment with other planned paths, and increases the probability density value of the corresponding region. The extraction module is used to calculate the spatiotemporal bounding box of the new path segment when the workpiece path tree is updated, and extract the spatial projection bounding box of the spatiotemporal bounding box in three-dimensional space as the perturbation range. If the priority of other workpieces is lower than that of the currently updated workpiece, and their paths have not been confirmed or locked, and the initial search ellipsoid of the other workpieces intersects with the perturbation range, then the local path replanning of the corresponding workpiece is triggered.

[0006] Preferably, the step of constructing an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece includes: Obtain the three-dimensional coordinates of the starting point and the target point of the workpiece to be planned; Calculate the Euclidean distance between the starting point and the target point, and use the Euclidean distance as the distance between the two foci; Obtain a preset expansion coefficient greater than 1, and use the product of the distance between the two focal points and the preset expansion coefficient as the length of the major axis of the initial search ellipsoid; Using the starting point and the target point as the foci of the ellipsoid, and combining the length of the major axis, a spatial closed surface equation containing the starting point and the target point is constructed to generate the initial search ellipsoid.

[0007] Preferably, the step of constructing a dynamic probability density field from the spatiotemporal occupancy information of the planned path of the influencing source within the overlapping region includes: Extract the three-dimensional spatial coordinates and timestamps of each node in the planned path of the influencing source, and convert the timestamps into spatial dimensions using a preset time and space conversion factor to form a four-dimensional spatiotemporal trajectory point set. The overlapping area is divided into three-dimensional spatial grids using a preset grid size to generate multiple spatial grids; Traverse the center points of each spatial grid, generate multiple time layers according to a preset discrete time step, and combine the time values ​​corresponding to each time layer with the spatial dimensions after conversion by the time and space conversion factor to calculate the minimum four-dimensional Euclidean distance from each center point to the set of four-dimensional spatiotemporal trajectory points in each time layer. Based on the negative exponential decay function, the minimum four-dimensional Euclidean distance is mapped to the probability density value of the spatial grid under the corresponding time layer, generating a dynamic probability density field containing multiple discrete time layers.

[0008] Preferably, the step of sampling and generating candidate nodes within the initial search ellipsoid of the affected body includes: Randomly generate a three-dimensional random vector that follows a uniform distribution within a unit sphere; Obtain the rotation matrix of the initial search ellipsoid of the affected body and the scaling ratio of each coordinate axis; The coordinate transformation of the three-dimensional random vector is performed using the rotation matrix and the scaling ratio to calculate the offset vector relative to the center of the ellipsoid. The center coordinates of the initial search ellipsoid are added to the offset vector to generate the three-dimensional spatial coordinates of the candidate nodes located inside the initial search ellipsoid, and a timestamp is assigned to the nodes expected to arrive.

[0009] Preferably, determining the acceptance probability based on the probability density value of the candidate node in the dynamic probability density field, and treating the candidate node as a valid node when determining acceptance based on the acceptance probability, includes: Based on the three-dimensional spatial coordinates and timestamp of the candidate node, the target grid to which the node belongs is matched and located in the dynamic probability density field, and the time layer corresponding to or closest to the timestamp is selected to extract the probability density value of the target grid in that time layer. Set a basic admission probability constant, normalize the probability density value, and subtract the normalized probability density value from the basic admission probability constant to obtain the remaining admission threshold. Generate uniformly distributed random numbers between 0 and 1; If the random number is not greater than the remaining acceptance threshold, then the candidate node is accepted and marked as a valid node to be connected.

[0010] Preferably, the step of searching for the minimum-cost parent node in the path tree to connect and generate a new path segment includes: A local neighborhood is constructed with the effective node as the center of the sphere and a preset search radius. Extract all generated nodes within the local neighborhood range from the current workpiece's path tree, and construct a set of candidate parent nodes; Calculate the connection path length and time consumption between the effective node and each node in the set of candidate parent nodes, unify the units using a unit conversion factor, and calculate the total connection cost of each candidate parent node by combining the cumulative cost of each candidate parent node in the path tree. Select the candidate parent node with the lowest total connection cost as the optimal parent node, connect the effective node to the optimal parent node, generate the new path segment, and update the path tree of the current workpiece.

[0011] On the other hand, the present invention also provides a method for dynamic planning and scheduling of workpiece movement paths in an intelligent equipment manufacturing workshop, comprising the following steps: An initial search ellipsoid is constructed based on the coordinates of the starting point and the target point of each workpiece. For initial search ellipsoids that overlap, the high-priority workpiece is set as the influence source and the low-priority workpiece is set as the affected body according to the priority. The spatiotemporal occupancy information of the planned path of the influence source is constructed into a dynamic probability density field in the overlapping area. The probability density value of each point in the dynamic probability density field is determined based on the shortest spatiotemporal distance from the corresponding point to the influence source path. Candidate nodes are generated by sampling within the initial search ellipsoid of the affected body. The acceptance probability is determined based on the probability density value of the candidate node in the dynamic probability density field. When the candidate node is determined to be accepted based on the acceptance probability, the candidate node is taken as a valid node. The minimum cost parent node is searched in the path tree to connect and generate a new path segment. The spatiotemporal conflict penalty value of the new path segment with other planned paths is calculated, and the probability density value of the corresponding region is increased. When the workpiece path tree is updated, the spatiotemporal bounding box of the new path segment is calculated, and the spatial projection bounding box of the spatiotemporal bounding box in three-dimensional space is extracted as the perturbation range. If the priority of other workpieces is lower than that of the currently updated workpiece, and their paths have not been confirmed or locked, and the initial search ellipsoid of the other workpieces intersects with the perturbation range, then the local path replanning of the corresponding workpiece is triggered.

[0012] Furthermore, the step of constructing an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece includes: Obtain the three-dimensional coordinates of the starting point and the target point of the workpiece to be planned; Calculate the Euclidean distance between the starting point and the target point, and use the Euclidean distance as the distance between the two foci; Obtain a preset expansion coefficient greater than 1, and use the product of the distance between the two focal points and the preset expansion coefficient as the length of the major axis of the initial search ellipsoid; Using the starting point and the target point as the foci of the ellipsoid, and combining the length of the major axis, a spatial closed surface equation containing the starting point and the target point is constructed to generate the initial search ellipsoid.

[0013] Furthermore, the step of constructing a dynamic probability density field in the overlapping region based on the spatiotemporal occupancy information of the planned path affecting the source includes: Extract the three-dimensional spatial coordinates and timestamps of each node in the planned path of the influencing source, and convert the timestamps into spatial dimensions using a preset time and space conversion factor to form a four-dimensional spatiotemporal trajectory point set. The overlapping area is divided into three-dimensional spatial grids using a preset grid size to generate multiple spatial grids; Traverse the center points of each spatial grid, generate multiple time layers according to a preset discrete time step, and combine the time values ​​corresponding to each time layer with the spatial dimensions after conversion by the time and space conversion factor to calculate the minimum four-dimensional Euclidean distance from each center point to the set of four-dimensional spatiotemporal trajectory points in each time layer. Based on the negative exponential decay function, the minimum four-dimensional Euclidean distance is mapped to the probability density value of the spatial grid under the corresponding time layer, generating a dynamic probability density field containing multiple discrete time layers.

[0014] Further, the step of sampling and generating candidate nodes within the initial search ellipsoid of the affected body includes: Randomly generate a three-dimensional random vector that follows a uniform distribution within a unit sphere; Obtain the rotation matrix of the initial search ellipsoid of the affected body and the scaling ratio of each coordinate axis; The coordinate transformation of the three-dimensional random vector is performed using the rotation matrix and the scaling ratio to calculate the offset vector relative to the center of the ellipsoid. The center coordinates of the initial search ellipsoid are added to the offset vector to generate the three-dimensional spatial coordinates of the candidate nodes located inside the initial search ellipsoid, and a timestamp is assigned to the nodes expected to arrive.

[0015] Further, determining the acceptance probability based on the probability density value of the candidate node in the dynamic probability density field, and treating the candidate node as a valid node when determining acceptance based on the acceptance probability, includes: Based on the three-dimensional spatial coordinates and timestamp of the candidate node, the target grid to which the node belongs is matched and located in the dynamic probability density field, and the time layer corresponding to or closest to the timestamp is selected to extract the probability density value of the target grid in that time layer. Set a basic admission probability constant, normalize the probability density value, and subtract the normalized probability density value from the basic admission probability constant to obtain the remaining admission threshold. Generate uniformly distributed random numbers between 0 and 1; If the random number is not greater than the remaining acceptance threshold, then the candidate node is accepted and marked as a valid node to be connected.

[0016] Furthermore, the step of searching for the minimum-cost parent node in the path tree to connect and generate a new path segment includes: A local neighborhood is constructed with the effective node as the center of the sphere and a preset search radius. Extract all generated nodes within the local neighborhood range from the current workpiece's path tree, and construct a set of candidate parent nodes; Calculate the connection path length and time consumption between the effective node and each node in the set of candidate parent nodes, unify the units using a unit conversion factor, and calculate the total connection cost of each candidate parent node by combining the cumulative cost of each candidate parent node in the path tree. Select the candidate parent node with the lowest total connection cost as the optimal parent node, connect the effective node to the optimal parent node, generate the new path segment, and update the path tree of the current workpiece.

[0017] This invention constructs an initial search ellipsoid to limit the workpiece path search space, reducing invalid search ranges and improving path generation efficiency. When multiple workpiece paths intersect, the influencing source and affected body are determined according to priority, and a dynamic probability density field is used to characterize the spatiotemporal occupancy risk of the path, causing affected bodies to tend to avoid high-density conflict areas during candidate node sampling. Combining spatiotemporal conflict penalties and density field negative feedback adjustment can improve the safety of subsequent path selection and reduce collision risk. After the path tree is updated, the disturbance range is determined by the spatiotemporal bounding box of the new path segment and its spatial projection. Local replanning is triggered only for workpieces that meet the priority conditions and intersect with the disturbance range, thereby reducing global recalculation overhead and improving the stability and efficiency of multi-workpiece transfer scheduling in complex workshop environments. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the initial search ellipsoid; Figure 2 This is a schematic diagram of the optimal parent node search. Detailed Implementation

[0019] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Example 1 In a first embodiment of the present invention, a dynamic planning and scheduling system for workpiece movement paths in an intelligent equipment manufacturing workshop includes the following modules: The construction module is used to construct an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece. For initial search ellipsoids that overlap, the high-priority workpiece is set as the influence source and the low-priority workpiece is set as the affected body according to priority. The spatiotemporal occupancy information of the planned path of the influence source is constructed into a dynamic probability density field in the overlapping area. The probability density value of each point in the dynamic probability density field is determined based on the shortest spatiotemporal distance from the corresponding point to the influence source path.

[0021] The system obtains the 3D spatial coordinates of the starting point and target point of each workpiece, calculates the Euclidean distance between the starting and target points as the distance between the two foci, and sets the major axis length of the ellipsoid using a preset expansion coefficient. A singular value decomposition algorithm is used to calculate the rotation matrix, transforming the ellipsoid in its standard position to its actual 3D spatial pose with the line connecting the starting and target points as its central axis, thus constructing an initial search ellipsoid. An algebraic boundary test algorithm is used to determine whether there is a volume intersection between the initial search ellipsoids of different workpieces in 3D space. If an intersection exists, the task urgency attribute and processing sequence constraint attribute of the workpieces are extracted and sorted in descending order. Workpieces ranked higher are marked as influence sources, and workpieces ranked lower are marked as affected bodies.

[0022] The influence source is extracted based on the discrete four-dimensional spatiotemporal path point set pre-calculated using the A* algorithm. A three-dimensional voxel grid is constructed with the overlapping region of the ellipsoids of the affected body and the influence source as the boundary. A nearest neighbor search algorithm is used to construct a spatial index tree from the spatiotemporal path point set of the influence source. For each central sampling point in the three-dimensional voxel grid, multiple time layers are generated according to a preset discrete time step. In each time layer, the nearest influence source path point to the sampling point is queried. The shortest spatiotemporal distance is obtained by taking the square root of the weighted sum of the squares of the spatial distance and the time difference between the sampling point and the nearest path point. A negative exponential decay function is used to calculate the probability density value, so that the closer the point is to the influence source path, the higher the mapped probability density value. All the calculated probability density values ​​are assigned to the voxel grid under the corresponding time layer to generate a dynamic probability density field containing multiple discrete time layers.

[0023] As an optional implementation, the step of constructing an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece includes: Obtain the three-dimensional coordinates of the starting point and the target point of the workpiece to be planned; Calculate the Euclidean distance between the starting point and the target point, and use the Euclidean distance as the distance between the two foci; Obtain a preset expansion coefficient greater than 1, and use the product of the distance between the two focal points and the preset expansion coefficient as the length of the major axis of the initial search ellipsoid; Using the starting point and the target point as the foci of the ellipsoid, and combining the length of the major axis, a spatial closed surface equation containing the starting point and the target point is constructed to generate the initial search ellipsoid.

[0024] Read the position information of the workpiece to be planned, such as extracting the three-dimensional coordinates of the starting point. (1.0, 2.0, 0.5)m and the three-dimensional coordinates of the target point (9.0, 8.0, 1.5)m. Calculate the Euclidean distance between the two points as the distance 2c between the two foci of the ellipsoid. By substituting the coordinates, we can calculate that 2c is 10.05m.

[0025] To ensure that the planned path has sufficient obstacle avoidance margin, a preset expansion coefficient greater than 1 is obtained. The value of this parameter ranges from 1.1 to 1.5. If the workpiece size is large or the density of obstacles in the workshop is high, a higher value should be used; in this example, it is set to [value missing]. =1.2. The distance between the two focal points, 2c, is... Multiplying these values, we obtain the major axis length of the initial search ellipsoid as 2a = 12.06 m. Based on the determined 2a and 2c, the minor axis length 2b is calculated to be approximately 6.65 m, and... and midpoint Using (5.0, 5.0, 1.0)m as the geometric center, a spatial closed surface equation is constructed, consisting of an orthogonal scaling matrix and a rotation matrix calculated based on the vectors connecting the start and end points. This equation is then used to generate the initial search ellipsoid in a three-dimensional coordinate system, defining the effective spatial boundary for subsequent node sampling. Figure 1 As shown.

[0026] As an optional implementation, the step of constructing a dynamic probability density field in the overlapping region from the spatiotemporal occupancy information of the planned path of the influencing source includes: Extract the three-dimensional spatial coordinates and timestamps of each node in the planned path of the influencing source, and convert the timestamps into spatial dimensions using a preset time and space conversion factor to form a four-dimensional spatiotemporal trajectory point set. The overlapping area is divided into three-dimensional spatial grids using a preset grid size to generate multiple spatial grids; Traverse the center points of each spatial grid, generate multiple time layers according to a preset discrete time step, and combine the time values ​​corresponding to each time layer with the spatial dimensions after conversion by the time and space conversion factor to calculate the minimum four-dimensional Euclidean distance from each center point to the set of four-dimensional spatiotemporal trajectory points in each time layer. Based on the negative exponential decay function, the minimum four-dimensional Euclidean distance is mapped to the probability density value of the spatial grid under the corresponding time layer, generating a dynamic probability density field containing multiple discrete time layers.

[0027] The planned trajectory point sequence of high-priority workpieces affecting the source is analyzed, and the three-dimensional spatial coordinates (x, y, z) and corresponding timestamps t of each trajectory point in the sequence are extracted. To handle spatiotemporal constraints within the same mathematical dimension, preset time and space conversion factors are input. This factor typically represents the average safe operating speed of the workpiece, and its value can range from 0.5 m / s to 1.5 m / s. Here, it is set... =1.0m / s. This is achieved by multiplying the timestamp t by... Transformed into spatial dimensions Combining and constructing such as (x,y,z, A four-dimensional spatiotemporal trajectory point set in the form of ) represents the spatiotemporal occupancy trajectory of the influencing source.

[0028] When constructing the density field, a preset mesh size is used, ranging from 0.1m to 0.5m (0.2m is set here). The overlapping region of the initial search ellipsoids of the two workpieces is then divided into a voxelized 3D spatial mesh. The geometric center points of each spatial mesh are traversed, and multiple discrete-time layers are generated within a preset planning time range according to a preset discrete-time step Δt. , ... For any spatial grid center point Its three-dimensional coordinates are as follows For any time layer Using time and space conversion factors Converting this time dimension to spatial dimensions yields the corresponding time dimension coordinates. The center point of this spatial grid is represented as a four-dimensional spacetime point. Represent any trajectory point in the source path as... Calculate the four-dimensional Euclidean distance between the four-dimensional spatiotemporal point and each trajectory point in the path of the influencing source, and select the minimum value as the minimum four-dimensional Euclidean distance of the spatial grid center point at that time level. .

[0029] Using negative exponential decay function number Among them, the basic maximum density value The preferred value is 1.0, and the attenuation coefficient k is 0.8m. -1 This makes the exponent term dimensionless. =1.25m, then the probability density value of this grid is approximately 0.368, generating a dynamic probability density field. When initializing the dynamic probability density field, the larger of the base maximum density value, the maximum density value in the initial density field, and the preset minimum positive number is set as the normalized reference density value. When the density field is updated due to spatiotemporal conflict penalties, if the updated maximum density value is greater than the current normalized reference density value, then the updated maximum density value replaces the current normalized reference density value.

[0030] The calculation module is used to sample and generate candidate nodes within the initial search ellipsoid of the affected body, determine the acceptance probability based on the probability density value of the candidate node in the dynamic probability density field, and when the candidate node is determined to be accepted based on the acceptance probability, the candidate node is taken as a valid node. The module searches for the minimum cost parent node in the path tree to connect and generate a new path segment, calculates the spatiotemporal conflict penalty value of the new path segment with other planned paths, and increases the probability density value of the corresponding region.

[0031] Random coordinate points are generated within a unit sphere. These random coordinate points are multiplied by the lengths of the three semi-axes of the initial search ellipsoid of the affected body, and then translated to the center of the ellipsoid after a rotation matrix transformation, thereby generating three-dimensional candidate nodes within the initial search ellipsoid of the affected body. The probability density value of the candidate node in the corresponding voxel grid in the dynamic probability density field is extracted, and the normalized probability density value is subtracted from the constant 1 using a linear inverse proportional mapping algorithm to obtain the acceptance probability. A uniformly distributed random number between 0 and 1 is generated. If the random number is less than or equal to the calculated acceptance probability, the candidate node is determined to have passed the rejection sampling verification and is included as a valid node in the RRT star algorithm node set of the affected body; otherwise, the node is discarded and resampling is performed.

[0032] In the current path tree of the affected entity, find all neighboring nodes within a defined radius centered on the effective node. Calculate the cumulative path cost connecting the effective node to each neighboring node. The cumulative path cost equals the existing root node cost of the neighboring node plus the Euclidean distance cost. Select the neighboring node with the minimum cumulative path cost as the minimum-cost parent node and connect it to generate a new path segment. Based on the timestamp of the minimum-cost parent node and the estimated time consumption of the new path segment, recalculate and assign the actual arrival timestamp of the effective node, so that the time interval of the new path segment is jointly determined by the timestamp of the minimum-cost parent node and the updated timestamp of the effective node.

[0033] After generating a new path segment, the starting coordinates, ending coordinates, starting timestamp, and ending timestamp of the new path segment are read. Combined with the workpiece's external dimensions or a preset safe expansion distance, a spatiotemporal bounding box corresponding to the new path segment during its movement is constructed. Each path segment in other planned paths is traversed, and a corresponding spatiotemporal bounding box is constructed for each. It is then determined whether the time intervals of two path segments overlap. If time overlap exists, the ratio of the overlapping volume to the union volume of the corresponding spatial bounding boxes within the overlapping time interval is calculated to obtain the spatial volume intersection-union ratio (IoU) value. This IoU value is multiplied by a preset time penalty coefficient to obtain the spatiotemporal conflict penalty value. Subsequently, using the overlapping spatial grid or the spatial grid where the new path segment is located as the center, a neighborhood voxel grid within a preset neighborhood radius is determined. Following a Gaussian weighting rule where the weight increases with proximity to the central grid, the spatiotemporal conflict penalty value is superimposed onto the probability density values ​​of the central voxel grid and the neighborhood voxel grids at the corresponding time level. By increasing the probability density value of the conflict area and its adjacent areas, the acceptance probability of subsequent candidate nodes in this area is reduced, thereby completing the negative feedback adjustment of the dynamic probability density field.

[0034] As an optional implementation, the step of sampling and generating candidate nodes within the initial search ellipsoid of the affected body includes: Randomly generate a three-dimensional random vector that follows a uniform distribution within a unit sphere; Obtain the rotation matrix of the initial search ellipsoid of the affected body and the scaling ratio of each coordinate axis; The coordinate transformation of the three-dimensional random vector is performed using the rotation matrix and the scaling ratio to calculate the offset vector relative to the center of the ellipsoid. The center coordinates of the initial search ellipsoid are added to the offset vector to generate the three-dimensional spatial coordinates of the candidate nodes located inside the initial search ellipsoid, and a timestamp is assigned to the nodes expected to arrive.

[0035] To achieve spatial node expansion, a uniformly distributed three-dimensional random vector is generated within a unit sphere of radius 1 in a standardized three-dimensional space using the rejection sampling method. For example, generating a random vector. =(0.5,-0.4,0.3). Obtain the initial search ellipsoid parameters of the affected body constructed in the previous step, extract the three-dimensional scaling diagonal matrix L calculated based on the major and minor axes of the ellipsoid, such as L=diag(6.03,3.32,3.32), and the rotation matrix R that aligns the X-axis of the standard coordinate system to the start-end direction.

[0036] The random vector is then transformed using coordinate transformation, i.e., calculated using the formula... The offset vector relative to the geometric center of the ellipsoid is calculated. The pre-calculated coordinates of the ellipsoid center are then used. Adding (5.0, 5.0, 1.0)m to the offset vector completes the back projection transformation of the candidate node coordinates, generating the candidate node coordinates in 3D space. To achieve spatiotemporal synchronous planning, based on the straight-line distance from the candidate node to the starting point and the workpiece's preset constant travel speed, for example... =0.8m / s, estimate the travel time of the workpiece, and assign the corresponding timestamp parameter to the candidate node accordingly. This ensures the integrity of spatiotemporal sampling.

[0037] As an optional implementation, determining the acceptance probability based on the probability density value of the candidate node in the dynamic probability density field, and treating the candidate node as a valid node when determining acceptance based on the acceptance probability, includes: Based on the three-dimensional spatial coordinates and timestamp of the candidate node, the target grid to which the node belongs is matched and located in the dynamic probability density field, and the time layer corresponding to or closest to the timestamp is selected to extract the probability density value of the target grid in that time layer. Set a basic admission probability constant, normalize the probability density value, and subtract the normalized probability density value from the basic admission probability constant to obtain the remaining admission threshold. Generate uniformly distributed random numbers between 0 and 1; If the random number is not greater than the remaining acceptance threshold, then the candidate node is accepted and marked as a valid node to be connected.

[0038] After obtaining candidate nodes with timestamps Then, the three-dimensional coordinates of the candidate nodes are divided by the preset grid size and rounded down to obtain the spatial index (i,j,k), and then based on the timestamp. By matching the corresponding or closest discrete-time layer index l, the four-dimensional index of the candidate node in the dynamic probability density field is determined. Then, the corresponding probability density value is extracted. If a candidate node is located outside the coverage area of ​​the dynamic probability density field, its corresponding probability density value is set to zero, and it participates in the admission decision according to the basic admission probability. A basic admission probability constant is constructed. It is usually set to 1.0 to extract the data. The value is normalized by dividing it by the normalized reference density value to obtain the normalized density, for example, 0.35. A subtraction operation is then performed to calculate the remaining acceptance threshold, i.e. =0.65, which represents the conflict-free safety margin of this space region.

[0039] Generate a uniformly distributed random number r between 0.0 and 1.0, for example, r = 0.48. For r and... Comparison: Since 0.48 is not greater than 0.65, the spatial sampling is determined to have passed the security screening, and the candidate node is marked as a valid node to be connected; if r is greater than the remaining acceptance threshold, the candidate node is discarded.

[0040] As an optional implementation, the step of searching for the minimum-cost parent node in the path tree to connect and generate a new path segment includes: A local neighborhood is constructed with the effective node as the center of the sphere and a preset search radius. Extract all generated nodes within the local neighborhood range from the current workpiece's path tree, and construct a set of candidate parent nodes; Calculate the connection path length and time consumption between the effective node and each node in the set of candidate parent nodes, unify the units using a unit conversion factor, and calculate the total connection cost of each candidate parent node by combining the cumulative cost of each candidate parent node in the path tree. Select the candidate parent node with the lowest total connection cost as the optimal parent node, connect the effective node to the optimal parent node, generate the new path segment, and update the path tree of the current workpiece.

[0041] The system uses the accepted valid nodes as the geometric center of a sphere and constructs a local spherical search neighborhood using a preset search radius. This search radius adaptively decreases based on the total number of nodes N in the current path tree, as shown in the formula: Where γ is the search radius adjustment coefficient, used to adjust the size of the local neighborhood range. γ is a constant greater than 0, preset according to the workshop aisle width, workpiece size, and path tree node density. The initial baseline value is set to 2.5m. To avoid the initial search radius of the path tree being zero or too small, the system sets a minimum search radius. When the number of nodes N in the path tree is less than a preset node number threshold, an initial baseline search radius of 2.5m is used as the search radius. When N is not less than the preset node number threshold, the search radius is calculated according to the adaptive decay formula, ensuring that the search radius is not less than [the threshold value]. .

[0042] Extract all historically generated tree nodes within the search radius and store them in a set of candidate parent nodes. For each candidate parent node in this set, traverse and calculate the three-dimensional Euclidean distance between the parent node and the valid node as the connection path length. For example, 1.5m, and the estimated time consumption obtained by dividing by the average speed. For example, 1.875s. To reduce the scale differences among multiple physical quantities, dimension conversion factors acting on length and time are introduced, for example, . =1.0, =0.8, where As a length cost weight, This is a time cost conversion factor used to convert time consumption into a cost dimension consistent with path length. The two are then linearly weighted and summed to obtain the current edge connection cost. In addition, there is the historical cumulative cost of the alternative parent nodes from the starting point to the current state. This yields the total connection cost for each node. The candidate node that minimizes the total connection cost is selected. For example, if the calculated minimum cost is 8.4, this node is chosen as the optimal parent node for connection. The timestamp of the optimal parent node is then compared with its corresponding estimated time consumption. The timestamps are summed to obtain the actual arrival timestamps after the effective node updates. After completing the anti-collision interference check, a new path segment is generated, and the current workpiece topology path tree is updated synchronously. If no candidate parent node exists within the local neighborhood, the search radius is increased to a preset upper limit, or the starting root node is used as a temporary candidate parent node in the connection judgment to ensure that the initial expansion of the path tree can be executed. A schematic diagram of the optimal parent node search is shown below. Figure 2 As shown.

[0043] The extraction module is used to calculate the spatiotemporal bounding box of the new path segment when the workpiece path tree is updated, and extract the spatial projection bounding box of the spatiotemporal bounding box in three-dimensional space as the perturbation range. If the priority of other workpieces is lower than that of the currently updated workpiece, and their paths have not been confirmed or locked, and the initial search ellipsoid of the other workpieces intersects with the perturbation range, then the local path replanning of the corresponding workpiece is triggered.

[0044] Once the RRT star algorithm of the affected object completes one iteration and adds the new path segment to the path tree, the system obtains the coordinate range of the start and end points of the new path segment on the X-axis, Y-axis, Z-axis, and time axis. It then expands the boundaries of the X-axis, Y-axis, and Z-axis in conjunction with the workpiece's external dimensions or a preset safe expansion distance to construct an axially aligned spatiotemporal bounding box (AABB). Subsequently, the time axis boundary is removed, leaving only the boundary ranges on the three spatial dimensions of the X-axis, Y-axis, and Z-axis to obtain a three-dimensional spatial projection bounding box. This three-dimensional spatial projection bounding box is then used as the disturbance range that the current new path segment may affect the surrounding workpieces.

[0045] The system employs an ellipsoid-bounding box intersection detection algorithm, or first converts the initial search ellipsoids of other workpieces into their corresponding bounding boxes, and then performs intersection detection with the disturbance range. The detection targets include other workpieces within the workshop with lower priority than the currently updated workpiece, whose paths have not yet been confirmed or locked. For workpieces with higher priority than the currently updated workpiece and whose paths have been confirmed, the system maintains their planned paths unchanged and continues to use their path spatiotemporal occupancy information as an influence source constraint in the dynamic probability density field.

[0046] If an initial search ellipsoid of a workpiece is detected to intersect with the disturbance range, the workpiece is marked as a workpiece to be replanned, and the voxel mesh within the overlapping area of ​​the disturbance range and the initial search ellipsoid of the workpiece is extracted. The probability density value of the overlapping mesh area and its neighboring voxel meshes is increased according to a preset disturbance increment, forming a high-density disturbance region for subsequent path avoidance. Subsequently, in the path tree already generated for the workpiece to be replanned, path segments that spatially or spatiotemporally overlap with the disturbance range are searched. These path segments and their subsequent connecting branches are marked as affected path segments, and unaffected preceding path nodes are retained as starting nodes for local replanning. Afterwards, within the initial search ellipsoid of the workpiece to be replanned, the system continues to re-search and update the affected path segments using candidate node sampling, acceptance probability determination, minimum cost parent node connection, effective node timestamp update, and spatiotemporal conflict penalty calculation. For other workpieces that do not intersect with the disturbance range, their already planned paths remain unchanged.

[0047] Example 2 Embodiment 2 of the present invention proposes a method for dynamic planning and scheduling of workpiece movement paths in an intelligent equipment manufacturing workshop, comprising the following steps: An initial search ellipsoid is constructed based on the coordinates of the starting point and the target point of each workpiece. For initial search ellipsoids that overlap, the high-priority workpiece is set as the influence source and the low-priority workpiece is set as the affected body according to the priority. The spatiotemporal occupancy information of the planned path of the influence source is constructed into a dynamic probability density field in the overlapping area. The probability density value of each point in the dynamic probability density field is determined based on the shortest spatiotemporal distance from the corresponding point to the influence source path. Candidate nodes are generated by sampling within the initial search ellipsoid of the affected body. The acceptance probability is determined based on the probability density value of the candidate node in the dynamic probability density field. When the candidate node is determined to be accepted based on the acceptance probability, the candidate node is taken as a valid node. The minimum cost parent node is searched in the path tree to connect and generate a new path segment. The spatiotemporal conflict penalty value of the new path segment with other planned paths is calculated, and the probability density value of the corresponding region is increased. When the workpiece path tree is updated, the spatiotemporal bounding box of the new path segment is calculated, and the spatial projection bounding box of the spatiotemporal bounding box in three-dimensional space is extracted as the perturbation range. If the priority of other workpieces is lower than that of the currently updated workpiece, and their paths have not been confirmed or locked, and the initial search ellipsoid of the other workpieces intersects with the perturbation range, then the local path replanning of the corresponding workpiece is triggered.

[0048] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A dynamic planning and scheduling system for workpiece movement paths in an intelligent equipment manufacturing workshop, characterized in that, Includes the following modules: The construction module is used to construct an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece. For initial search ellipsoids that overlap, the high-priority workpiece is set as the influence source and the low-priority workpiece is set as the affected body according to the priority. The spatiotemporal occupancy information of the planned path of the influence source is constructed into a dynamic probability density field in the overlapping area. The probability density value of each point in the dynamic probability density field is determined based on the shortest spatiotemporal distance from the corresponding point to the influence source path. The calculation module is used to sample and generate candidate nodes within the initial search ellipsoid of the affected body, determine the acceptance probability based on the probability density value of the candidate node in the dynamic probability density field, and when the candidate node is determined to be accepted based on the acceptance probability, the candidate node is taken as a valid node. The module searches for the minimum cost parent node in the path tree to connect and generate a new path segment, calculates the spatiotemporal conflict penalty value of the new path segment with other planned paths, and increases the probability density value of the corresponding region. The extraction module is used to calculate the spatiotemporal bounding box of the new path segment when the workpiece path tree is updated, and extract the spatial projection bounding box of the spatiotemporal bounding box in three-dimensional space as the perturbation range. If the priority of other workpieces is lower than that of the currently updated workpiece, and their paths have not been confirmed or locked, and the initial search ellipsoid of the other workpieces intersects with the perturbation range, then the local path replanning of the corresponding workpiece is triggered.

2. The system according to claim 1, characterized in that, The process of constructing an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece includes: Obtain the three-dimensional coordinates of the starting point and the target point of the workpiece to be planned; Calculate the Euclidean distance between the starting point and the target point, and use the Euclidean distance as the distance between the two foci; Obtain a preset expansion coefficient greater than 1, and use the product of the distance between the two focal points and the preset expansion coefficient as the length of the major axis of the initial search ellipsoid; Using the starting point and the target point as the foci of the ellipsoid, and combining the length of the major axis, a spatial closed surface equation containing the starting point and the target point is constructed to generate the initial search ellipsoid.

3. The system according to claim 1, characterized in that, The step of constructing a dynamic probability density field in the overlapping region based on the spatiotemporal occupancy information of the planned path of the influencing source includes: Extract the three-dimensional spatial coordinates and timestamps of each node in the planned path of the influencing source, and convert the timestamps into spatial dimensions using a preset time and space conversion factor to form a four-dimensional spatiotemporal trajectory point set. The overlapping area is divided into three-dimensional spatial grids using a preset grid size to generate multiple spatial grids; Traverse the center points of each spatial grid, generate multiple time layers according to a preset discrete time step, and combine the time values ​​corresponding to each time layer with the spatial dimensions after conversion by the time and space conversion factor to calculate the minimum four-dimensional Euclidean distance from each center point to the set of four-dimensional spatiotemporal trajectory points in each time layer. Based on the negative exponential decay function, the minimum four-dimensional Euclidean distance is mapped to the probability density value of the spatial grid under the corresponding time layer, generating a dynamic probability density field containing multiple discrete time layers.

4. The system according to claim 1, characterized in that, The step of generating candidate nodes by sampling within the initial search ellipsoid of the affected body includes: Randomly generate a three-dimensional random vector that follows a uniform distribution within a unit sphere; Obtain the rotation matrix of the initial search ellipsoid of the affected body and the scaling ratio of each coordinate axis; The coordinate transformation of the three-dimensional random vector is performed using the rotation matrix and the scaling ratio to calculate the offset vector relative to the center of the ellipsoid. The center coordinates of the initial search ellipsoid are added to the offset vector to generate the three-dimensional spatial coordinates of the candidate nodes located inside the initial search ellipsoid, and a timestamp is assigned to the nodes expected to arrive.

5. The system according to claim 4, characterized in that, The step of determining the acceptance probability based on the probability density value of the candidate node in the dynamic probability density field, and treating the candidate node as a valid node when determining acceptance based on the acceptance probability, includes: Based on the three-dimensional spatial coordinates and timestamp of the candidate node, the target grid to which the node belongs is matched and located in the dynamic probability density field, and the time layer corresponding to or closest to the timestamp is selected to extract the probability density value of the target grid in that time layer. Set a basic admission probability constant, normalize the probability density value, and subtract the normalized probability density value from the basic admission probability constant to obtain the remaining admission threshold. Generate uniformly distributed random numbers between 0 and 1; If the random number is not greater than the remaining acceptance threshold, then the candidate node is accepted and marked as a valid node to be connected.

6. The system according to claim 1, characterized in that, The step of searching for the minimum-cost parent node in the path tree to connect and generate a new path segment includes: A local neighborhood is constructed with the effective node as the center of the sphere and a preset search radius. Extract all generated nodes within the local neighborhood range from the current workpiece's path tree, and construct a set of candidate parent nodes; Calculate the connection path length and time consumption between the effective node and each node in the set of candidate parent nodes, unify the units using a unit conversion factor, and calculate the total connection cost of each candidate parent node by combining the cumulative cost of each candidate parent node in the path tree. Select the candidate parent node with the lowest total connection cost as the optimal parent node, connect the effective node to the optimal parent node, generate the new path segment, and update the path tree of the current workpiece.

7. A method for dynamic planning and scheduling of workpiece movement paths in an intelligent equipment manufacturing workshop, characterized in that, Includes the following steps: An initial search ellipsoid is constructed based on the coordinates of the starting point and the target point of each workpiece. For initial search ellipsoids that overlap, the high-priority workpiece is set as the influence source and the low-priority workpiece is set as the affected body according to the priority. The spatiotemporal occupancy information of the planned path of the influence source is constructed into a dynamic probability density field in the overlapping area. The probability density value of each point in the dynamic probability density field is determined based on the shortest spatiotemporal distance from the corresponding point to the influence source path. Candidate nodes are generated by sampling within the initial search ellipsoid of the affected body. The acceptance probability is determined based on the probability density value of the candidate node in the dynamic probability density field. When the candidate node is determined to be accepted based on the acceptance probability, the candidate node is taken as a valid node. The minimum cost parent node is searched in the path tree to connect and generate a new path segment. The spatiotemporal conflict penalty value of the new path segment with other planned paths is calculated, and the probability density value of the corresponding region is increased. When the workpiece path tree is updated, the spatiotemporal bounding box of the new path segment is calculated, and the spatial projection bounding box of the spatiotemporal bounding box in three-dimensional space is extracted as the perturbation range. If the priority of other workpieces is lower than that of the currently updated workpiece, and their paths have not been confirmed or locked, and the initial search ellipsoid of the other workpieces intersects with the perturbation range, then the local path replanning of the corresponding workpiece is triggered.

8. The method according to claim 7, characterized in that, The process of constructing an initial search ellipsoid based on the coordinates of the starting point and the target point of each workpiece includes: Obtain the three-dimensional coordinates of the starting point and the target point of the workpiece to be planned; Calculate the Euclidean distance between the starting point and the target point, and use the Euclidean distance as the distance between the two foci; Obtain a preset expansion coefficient greater than 1, and use the product of the distance between the two focal points and the preset expansion coefficient as the length of the major axis of the initial search ellipsoid; Using the starting point and the target point as the foci of the ellipsoid, and combining the length of the major axis, a spatial closed surface equation containing the starting point and the target point is constructed to generate the initial search ellipsoid.

9. The method according to claim 7, characterized in that, The step of constructing a dynamic probability density field in the overlapping region based on the spatiotemporal occupancy information of the planned path of the influencing source includes: Extract the three-dimensional spatial coordinates and timestamps of each node in the planned path of the influencing source, and convert the timestamps into spatial dimensions using a preset time and space conversion factor to form a four-dimensional spatiotemporal trajectory point set. The overlapping area is divided into three-dimensional spatial grids using a preset grid size to generate multiple spatial grids; Traverse the center points of each spatial grid, generate multiple time layers according to a preset discrete time step, and combine the time values ​​corresponding to each time layer with the spatial dimensions after conversion by the time and space conversion factor to calculate the minimum four-dimensional Euclidean distance from each center point to the set of four-dimensional spatiotemporal trajectory points in each time layer. Based on the negative exponential decay function, the minimum four-dimensional Euclidean distance is mapped to the probability density value of the spatial grid under the corresponding time layer, generating a dynamic probability density field containing multiple discrete time layers.

10. The method according to claim 7, characterized in that, The step of generating candidate nodes by sampling within the initial search ellipsoid of the affected body includes: Randomly generate a three-dimensional random vector that follows a uniform distribution within a unit sphere; Obtain the rotation matrix of the initial search ellipsoid of the affected body and the scaling ratio of each coordinate axis; The coordinate transformation of the three-dimensional random vector is performed using the rotation matrix and the scaling ratio to calculate the offset vector relative to the center of the ellipsoid. The center coordinates of the initial search ellipsoid are added to the offset vector to generate the three-dimensional spatial coordinates of the candidate nodes located inside the initial search ellipsoid, and a timestamp is assigned to the nodes expected to arrive.

Images