A welding path planning method and system of a ship outfitting welding robot

Abstract

This invention discloses a welding path planning method and system for a welding robot used in ship outfitting. It introduces the concept of semantic modeling for welding tasks, unifying the geometric information of the weld, structural connection relationships, and welding process constraints, and decomposing the overall welding task into welding units with clear semantic attributes. A constraint relationship model between welding units is constructed, and executable welding units are dynamically determined based on on-site status information, generating a welding execution sequence that meets structural dependence, process continuity, and spatial safety requirements. During welding execution, the status of on-site obstacles, weld geometric changes, and process resource status are continuously monitored. When changes affecting the welding operation are detected, the constraint relationships are updated at the semantic welding unit level, and only the affected welding units undergo local rearrangement and local path replanning. This invention effectively improves the adaptability of welding robots to changes in the complex space of outfitting, balancing welding quality and operational efficiency.

Description

Technical Field

[0001] This invention belongs to the field of shipbuilding and welding automation technology, and particularly relates to a welding path planning method and system for a ship outfitting welding robot. Background Technology

[0002] Welding operations during ship outfitting typically occur in compartments, below deck, or in areas with dense equipment. These operations are characterized by confined spaces, complex component types, and discrete yet interconnected weld seams, making them one of the least automated and most challenging aspects of shipbuilding. With the increasing size, modularity, and functional integration of ships, outfitting welding objects exhibit characteristics such as a large number of weld seams, significant spatial overlap, and work sequences constrained by both structure and process. Traditional methods relying on manual experience for welding sequence arrangement and path planning are no longer sufficient to meet the combined requirements of efficiency, quality, and safety. Against this backdrop, utilizing welding robots to replace manual labor in outfitting welding operations has become an important direction for industry development, and welding path planning methods are one of the key foundational technologies for achieving this goal.

[0003] Current path planning technologies for marine welding robots largely rely on mature offline programming or teach-and-playback models from industrial welding scenarios. This involves pre-generating fixed welding trajectories and executing them in a predetermined sequence within a relatively stable, structurally regular environment. However, outfitting welding operations are significantly unstructured. Welds often have implicit structural connections and process dependencies. For example, some welds can only be performed after tack welding or support welding, some must be completed continuously to ensure quality, while others can be flexibly inserted based on site conditions. Existing technologies generally treat welds as independent geometric objects, focusing only on the spatial path generation of individual welds. This lack of expression of the inherent semantic relationships within the welding task leads to welding sequences relying on human experience or simple rule settings, making it difficult to systematically reflect the true operational logic of outfitting welding.

[0004] On the other hand, existing path planning methods are insufficiently adaptable to the frequently changing working environment at outfitting sites. The outfitting phase often involves multiple trades working simultaneously, and the progress of component installation, the layout of temporary tools, personnel activities, and the state of obstacles on site can all dynamically change. Traditional welding paths, once generated, are difficult to adjust flexibly during execution. When site conditions change, it is usually necessary to re-plan the entire path or even re-teach it, resulting in long planning cycles and high levels of human intervention. This not only affects the construction pace but also increases the risk of collisions, interference, or fluctuations in welding quality due to unreasonable paths. Although some existing technologies have developed path adjustment methods based on local obstacle avoidance or real-time correction, most remain at the geometric level of compensation and cannot consider the overall welding task to globally reconstruct the welding sequence and execution strategy.

[0005] Furthermore, existing welding robot path planning methods generally lack a systematic consideration of process continuity and welding quality constraints. In outfitting welding, the sequence of weld heat input, the integrity of continuous welding sections, and the rational arrangement of welding rhythm have a significant impact on structural deformation control and weld quality. However, traditional planning methods often use the shortest path or minimum movement time as the optimization objective, ignoring the constraints at the welding process level. This makes the planning results theoretically achievable, but difficult to guarantee stable welding quality in actual construction, still requiring a large amount of manual intervention and on-site adjustments. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a welding path planning method and system for a ship outfitting welding robot. Specifically, the technical solution provided by this invention is as follows: A welding path planning method for a ship outfitting welding robot includes the following steps: S1. Obtain the weld geometry information, outfitting structure spatial information, and welding process parameter information of the outfitting welding operation area, and construct the initial site state model of the welding operation under a unified coordinate system; S2. Based on the weld geometry information and welding process parameter information, the welding task to be performed is semantically decomposed, and the overall welding task is divided into multiple semantic welding units. S3. Based on the structural dependencies, welding process dependencies, and spatial interference relationships between the semantic welding units, construct a constraint relationship model between the semantic welding units to describe the constraints of each semantic welding unit on the execution order and execution conditions. S4. Combining the on-site state model and the constraint relationship model, perform an executability determination on the semantic welding unit, and determine the set of executable welding units that meet the constraint conditions and are spatially reachable at the current moment. S5. Generate a welding execution sequence based on the set of executable welding units, and generate corresponding welding paths for semantic welding units in sequence according to the welding execution sequence; S6. During the welding process, continuously monitor the on-site status. When changes are detected in the on-site obstacle status, weld geometry status, or welding process resource status that affect subsequent welding tasks, update the constraint relationship model and the set of executable welding units. S7. Under the updated constraint relationship model, the welding execution sequence is regenerated for the incomplete semantic welding units, and the welding path is re-planned only for the semantic welding units affected by the changes in the field state, thereby realizing the dynamic adjustment of the welding path planning.

[0007] Furthermore, the construction of the field state model described in step S1 includes the following steps: S101. Obtain the pose information of the welding robot base in the world coordinate system, denoted as... , This indicates the spatial pose of the welding robot base relative to the unified coordinate system; S102. Obtain the spatial distribution information of obstacles in the outfitting welding operation area, and represent the spatial distribution information of obstacles as an obstacle set O(t), where O(t) is used to describe the spatial occupancy status of outfitting structures, tooling and temporary objects in the welding operation area. S103. Obtain the geometric correction information corresponding to each weld. Based on the geometric correction information, sampling points of the weld centerline are determined. The information is updated to characterize the actual spatial position of the weld at the current moment; at the same time, the status information of welding process resources related to the welding operation is obtained, denoted as u(t), which is used to characterize the availability status of welding power source and welding process conditions; S104. The welding robot base posture, obstacle set, weld geometry correction information and welding process resource status are uniformly organized to construct an initial field state model s(t).

[0008] Furthermore, the semantic decomposition described in step S2 includes the following steps: S201. Based on the acquired weld geometry information, each weld is segmented according to the continuity of its centerline sampling points in space, dividing the weld into several geometric segments, and representing the geometric segments belonging to the same welding task as a set of weld geometric segments. ,in, j Indicates the weld number, and These represent the starting and ending sampling point indices of the weld within the geometric segment, respectively. S202. Based on the obtained set of weld geometric segments, and in combination with the connection relationship of the weld in the outfitting structure, perform structural connection semantic determination on each weld geometric segment to distinguish the connection function undertaken by different welds in the outfitting structure, and merge weld geometric segments with the same or related structural connection semantics into the same welding task unit. S203. Combining the welding process parameter information corresponding to the weld, perform welding process semantic annotation on the welding task unit to characterize whether the welding task unit has continuous welding requirements, whether interruption is allowed, and the corresponding welding process conditions during the welding execution process. A welding task unit that simultaneously possesses the structural connection semantics and welding process semantics is defined as a semantic welding unit (SWU). k .

[0009] Furthermore, the construction of the constraint relationship model described in step S3 includes the following steps: S301, using semantic welding unit SWU k The nodes are used to construct a constraint relationship graph G=(V, E), where V is the set of semantic welding unit nodes and E is the set of constraint relationship edges between semantic welding units; S302. Based on the structural connection relationship between semantic welding units, determine the semantic welding unit pairs that have sequential assembly or force dependence relationship in the outfitting structure, and establish structural dependence constraint edges between the corresponding semantic welding unit nodes in the constraint relationship diagram to limit the sequential relationship of the semantic welding units in the welding execution order. S303. Based on the welding process semantics corresponding to the semantic welding unit, determine the semantic welding unit pairs with continuous welding requirements, process sequence dependence or uninterrupted requirements, and establish welding process dependence constraint edges in the constraint relationship graph to limit the continuity and process sequence of the semantic welding unit in the welding execution process. S304. Based on the spatial distribution relationship of semantic welding units in the outfitting operation space, determine the semantic welding unit pairs that will affect the accessibility or operation space of the other semantic welding unit if one semantic welding unit is executed first, and establish spatial interference constraint edges in the constraint relationship diagram to describe the execution restrictions caused by space occupation between the semantic welding units. S305. Assign constraint attributes to each constraint edge in the constraint relationship diagram to distinguish between hard constraints that must be satisfied and soft constraints that can be adjusted under safety and process conditions. This is achieved through constraint indicator flags. h a,b Characterizing semantic welding units a With semantic welding unit b Whether the constraints between them are hard constraints, and by using constraint weights w a,b The importance of the constraint relationships between the semantic welding units is characterized, thereby constructing a semantic welding unit constraint relationship model.

[0010] Furthermore, the determination of the set of executable welding units in step S4 includes the following steps: S401. Based on the constraint relationship model, for each semantic welding unit (SWU) k Determine the set of preceding constraint units in the constraint relationship graph, and determine whether the structural dependency constraints and welding process dependency constraints related to the semantic welding unit have been satisfied, so as to filter out semantic welding units that do not meet the sequential execution conditions at the current time. S402. Under the premise of satisfying the hard constraint conditions in the constraint relationship model, and in combination with the welding robot base posture in the field state model. And the obstacle set O(t), perform spatial reachability determination on the welding path or welding area corresponding to the semantic welding unit, so as to determine whether the welding robot can reach the welding position corresponding to the semantic welding unit without spatial interference under the current site conditions; S403. For semantic welding units that simultaneously satisfy the hard constraints in the constraint relationship model and pass the spatial reachability determination, they are determined as executable semantic welding units at the current time, and all the semantic welding units determined as executable constitute the set of executable welding units E(t) at the current time.

[0011] Furthermore, the generation of the welding path in step S5 includes the following steps: S501, For the semantic welding unit (SWU) to be executed in the welding execution sequence. k Based on the set of weld geometric segments corresponding to this semantic welding unit Extracting sampling points of the weld centerline And construct the welding geometry path skeleton of the semantic welding unit according to the spatial order of the sampling points; S502. Based on the construction of the welding geometry path skeleton, and combined with the welding process semantics corresponding to the semantic welding unit, the spatial pose of the welding torch relative to the weld is determined for each sampling point in the welding geometry path skeleton, thereby obtaining the target pose sequence of the welding torch in the welding process. S503, Combining the welding robot base posture in the aforementioned field state model And the welding gun target pose sequence, and perform kinematic solution on the welding path corresponding to the semantic welding unit to obtain the feasible motion path of the welding robot when executing the semantic welding unit; S504. The obtained feasible motion path of the welding robot is used as the welding path corresponding to the semantic welding unit, and steps S501 to S503 are repeated sequentially for each semantic welding unit in the welding execution sequence to generate a complete welding path corresponding to the welding execution sequence.

[0012] Further, step S6 includes: During the welding process, based on the field state model s(t), the obstacle set O(t) and weld geometry correction information within the welding operation area are processed. And the welding process resource status u(t) is updated in real time or periodically; When a change is detected in the obstacle set, weld geometry correction information, or welding process resource status, and this change affects the semantic welding unit (SWU) that has not yet been completed, kWhen the executability is affected, the constraint relationship graph G=(V, E) is updated based on the updated field state model s(t) to correct the spatial interference constraints or welding process dependency constraints between semantic welding units. Under the updated constraint relationship model, the executability of incomplete semantic welding units is re-performed, the set of executable welding units E(t) at the current time is updated, and the updated set of executable welding units is used for the generation of subsequent welding execution sequences and the adjustment of welding paths.

[0013] Further, step S7 includes: Under the updated constraint graph G=(V, E), identify semantic welding units (SWUs) that have not yet completed welding. k Based on the updated constraint relationships, the executability of the incomplete semantic welding units is re-evaluated to determine the executable welding units that satisfy the structural dependency constraints, welding process dependency constraints, and spatial interference constraints under the current field conditions. Based on the set of executable welding units E(t) obtained by re-determination, without changing the execution results of the completed semantic welding units, the welding execution order of the incomplete semantic welding units is regenerated so that the welding execution order satisfies the sequential relationship and execution conditions defined in the updated constraint relationship model. In the regenerated welding execution sequence, the corresponding welding path generation step is re-executed only for semantic welding units whose constraint relationships or executability have changed due to changes in the field conditions. For semantic welding units that are not affected by the changes in the field conditions, their original welding paths remain unchanged, thereby realizing the local replanning and dynamic adjustment of welding path planning at the semantic welding unit level.

[0014] A welding path planning system based on the above method, the system mainly includes the following units: The on-site state modeling unit is used to acquire the weld geometry information, outfitting structure spatial information, welding robot base posture information, and welding process resource status information of the outfitting welding operation area, and to construct an on-site state model of the welding operation under a unified coordinate system to characterize the real-time spatial status of the welding robot, weld and obstacles during the welding operation. The welding task semantic decomposition unit is used to semantically decompose the welding task to be executed based on the weld geometry information and welding process parameter information, and divide the overall welding task into multiple semantic welding units that simultaneously have structural connection semantics and welding process semantics, so as to serve as the basic operation object for welding path planning and scheduling. The constraint relationship model construction unit is used to construct a constraint relationship model between the semantic welding units based on the structural connection relationship, welding process dependency relationship and spatial interference relationship between the semantic welding units, so as to describe the constraint relationship between the semantic welding units in terms of welding execution order and execution conditions. An executable welding unit determination unit is used to combine the field state model and the constraint relationship model to determine the executability of semantic welding units and identify a set of executable welding units that simultaneously meet the constraint conditions and are spatially reachable under the current field state. The welding execution sequence generation unit is used to generate a welding execution sequence based on the set of executable welding units, under the premise of satisfying the hard constraints in the constraint relationship model, so as to determine the welding execution order of the semantic welding units; The welding path generation unit is used to generate corresponding welding paths for the semantic welding unit in sequence according to the welding execution sequence, and output the welding trajectory that can be executed by the welding robot. The status monitoring and rescheduling unit is used to continuously monitor the field status during welding execution. When a change in the field status is detected that affects subsequent welding tasks, the field status model and constraint relationship model are updated, and welding execution sequences are regenerated for unfinished welding tasks at the semantic welding unit level. Welding path planning is only performed for semantic welding units affected by the change in the field status, thereby realizing dynamic adjustment of welding path planning.

[0015] Compared with the prior art, the present invention has at least the following beneficial effects: To address the problem that existing technologies generally rely on one-time offline planning and are unable to cope with changes in on-site conditions, this invention introduces an executability determination and semantic-level rescheduling mechanism based on on-site conditions. This transforms welding path planning from a static, closed pre-planning process into a rolling planning process that can dynamically adjust according to the real-time evolution of obstacle states, weld geometry changes, and process resource conditions. When on-site conditions change, this invention no longer simply overturns the original planning results entirely. Instead, it updates the constraint relationships at the semantic welding unit level. While keeping the completed welding results unaffected, it only performs local rearrangement and local replanning on the affected welding objects. This effectively avoids the problems of excessively long planning cycles, repeated interruptions to the construction cycle, and insufficient stability of automated systems caused by frequent overall path recalculation in existing technologies. It significantly improves the continuity, robustness, and engineering controllability of outfitting welding operations in complex environments.

[0016] Building upon this foundation, this invention further refines the semantic attributes of welding tasks, clearly distinguishing between interruptible and non-interruptible welding units. It also introduces freeze and safety node strategies, ensuring that during dynamic adjustments, non-interruptible welding tasks are prioritized until they reach a node that meets process and safety requirements before subsequent plan switching. This achieves dynamic scheduling and flexible response to on-site changes while effectively ensuring welding process continuity and weld quality, reducing quality risks caused by mid-process interruptions, frequent sequence changes, or forced avoidance. Compared to existing technologies that simply avoid or replan at the geometric path level, this invention provides a holistic approach to the welding process at the semantic level, making dynamic adjustment behavior itself part of the welding process logic. This significantly improves the reliability and practical value of the automated welding system for long-term stable operation in complex outfitting spaces.

[0017] Meanwhile, supported by the aforementioned dynamic scheduling mechanism, this invention extends welding path planning from a simple geometric reachability problem to a comprehensive decision-making process that integrates structural dependence, process constraints, and on-site conditions. This creates a close coupling between the welding execution sequence and the welding path generation, avoiding the problem of the welding sequence and path planning being separated in traditional methods, and reducing the reliance on human experience and on-site intervention.

[0018] In summary, this invention effectively solves the key technical bottlenecks of existing outfitting welding robots in complex and changing environments by combining semantic-level dynamic scheduling with local replanning. It is difficult to operate continuously, execute stably, and balance efficiency and quality. It demonstrates significant technological progress and comprehensive advantages in terms of welding automation adaptability, execution stability, and engineering feasibility. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0020] Figure 1 This is a flowchart illustrating the welding path planning method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the framework of the welding path planning system provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort are all within the scope of protection of the present invention.

[0022] Example 1 This embodiment provides a welding path planning method for a ship outfitting welding robot. Taking the semantic information of the welding task as the core, the welding path planning is upgraded from the traditional geometric trajectory generation to the task-level and semantic-level, realizing the collaborative planning and dynamic adjustment of the welding execution sequence and welding path.

[0023] like Figure 1 As shown, the method mainly includes the following steps: I. On-site Data Acquisition and Unified Modeling To enable welding robots to plan welding paths in complex spaces, it is necessary to acquire data and perform geometric modeling of the welding site in a unified manner, so that geometric information, sensor information and robot motion information from different sources can be calculated and accessed under the same reference system.

[0024] This embodiment first establishes a multi-coordinate system to describe the outfitting welding environment, including a world coordinate system to describe the entire compartment or work area. Sensor coordinate system used to describe sensor measurement data Robot base coordinate system used to describe the installation or movement position of a welding robot. and the robot end flange coordinate system and welding torch tool coordinate system The coordinate systems are linked through a homogeneous transformation matrix, the general form of which is:

[0025] in, Indicates from coordinate system To coordinate system pose transformation, Let be a rotation matrix. The above coordinate transformation relationship can be obtained using existing mature technologies, such as robot forward kinematics calculation, welding torch tool calibration, base installation measurement or positioning system acquisition, and sensor extrinsic parameter calibration. Since the above calibration and modeling methods are conventional technologies known in the field and do not constitute the core innovation of this invention, their specific implementation process will not be further elaborated here.

[0026] After establishing the coordinate system, the geometric model information of the outfitting and welding site is acquired. This geometric model is composed of the CAD or BIM model from the design phase and the measured point cloud data from the site. The design model includes information on the structural surfaces of the compartments, outfitting components, and theoretical welds, forming a priori geometric set; the measured point cloud is obtained through laser scanning by the welding robot, structured light, or binocular vision, and is used to reflect the actual assembly state, structural deviations, and the distribution of temporary obstacles. Arbitrary measurements in the measured point cloud... In the sensor coordinate system, it is represented as And transform it to the world coordinate system through coordinate transformation relationships:

[0027] in, , This is the transformation matrix from the sensor to the world coordinate system.

[0028] To eliminate the discrepancy between the design model and the actual assembly state, the CAD model and the measured point cloud are registered to obtain a correction transformation from the CAD coordinate system to the world coordinate system. This transformation can be solved by minimizing the distance error between the measured point cloud and the surface of the CAD model. Its objective function can be expressed as:

[0029] in, For sampling points on the surface of the CAD model, This represents the position operator that projects the transformed CAD points onto the nearest surface of the CAD model. This registration process can be implemented using mature algorithms such as point-to-surface ICP or feature-assisted registration, which are existing technologies; therefore, their specific algorithm flows will not be elaborated further. After registration, a structural geometric model consistent with the actual site is obtained, which is used for subsequent weld and obstacle environment modeling.

[0030] Based on this, the weld seams are geometrically represented and parametrically modeled. Each weld seam is represented as an ordered set of centerline points in the world coordinate system:

[0031] in, Sampling points for the centerline of the weld. This represents the number of sampling points. The weld centerline can be obtained by discretizing the weld curve in the CAD model, or by feature extraction and curve fitting of the weld area in the field point cloud. By performing a difference operation on adjacent sampling points of the centerline, the forward tangent vector of the weld at each point can be obtained:

[0032] in, For the first n The unit vector of the welding advance direction at each sampling point.

[0033] To determine the reference orientation of the welding torch, the normal information of the base material surface at the weld joint needs to be obtained. In the corrected structural mesh, neighborhood point sets are selected on both sides of the weld joint and local planes are fitted; their normal vectors can be obtained through principal component analysis. If a uniform reference normal for the welding torch needs to be constructed, the normals on both sides can be weighted, averaged, and normalized.

[0034] in, and These are the normal vectors of the base metal surfaces on both sides of the weld. and This is the weighting coefficient. The normal is used to define the welding torch orientation; its specific calculation method is a standard technique in geometric modeling and will not be elaborated further.

[0035] Based on the weld tangential and normal directions, a reference orientation for the welding torch tool coordinate system is constructed at each weld sampling point. Typically, the Z-axis of the welding torch tool coordinate system is aligned with the direction of the welding torch and the opposite direction to the weld reference normal.

[0036] Then, project the weld seam's forward direction into the normal plane to obtain the welding torch's forward direction axis:

[0037] And the third axis is obtained through cross product:

[0038] This forms the target rotation matrix of the welding torch at the sampling point. The corresponding welding torch target pose is represented as:

[0039] in, d This refers to the welding distance offset from the welding torch TCP to the weld centerline, and its value can be determined based on welding process parameters or welding torch calibration results. The above posture configuration serves only as a reference posture for welding path planning. Process details such as welding oscillation angle and torch angle correction can be further superimposed in the subsequent path generation stage, and will not be elaborated here.

[0040] Simultaneously, obstacles, temporary tools, and restricted areas in the outfitting site are calculably represented to form an obstacle set. In engineering implementation, point clouds or meshes can be used simultaneously for fine-grained collision detection, while voxel occupancy grids or range fields can be used for rapid safe distance determination. The minimum distance from any spatial point p to an obstacle can be expressed as:

[0041] And by setting a safe distance threshold d safe The minimum distance between the welding path and obstacles is constrained. This obstacle modeling and distance calculation method is a common technique in the field of robot path planning, so it is only described conceptually here.

[0042] Finally, based on the robot's kinematics model, an accessibility analysis of the weld target pose is performed. For the weld's... n The target pose of each sampling point in the base coordinate system is:

[0043] By solving the inverse kinematics, we obtain the set of joint solutions that satisfy joint constraints and collision-free constraints:

[0044] Where q is the robot joint angle vector, q min With q max These represent the upper and lower limits of the joint. The interval where feasible solutions continuously exist along the weld seam constitutes the reachable window of the weld seam at the current base position.

[0045] Through the above steps, a unified field state description vector is constructed:

[0046] in, This indicates the real-time pose of the robot base. Indicates the status of obstacles at the scene. This indicates the real-time correction results of the weld geometry. This represents the set of feasible solutions for each weld sampling point. This represents the welding resources and process status. This state vector can be updated under either periodic or event-triggered conditions.

[0047] II. Semantic Decomposition of Welding Tasks After completing the unified geometric modeling of the welding operation site, in order for the welding path planning to reflect the implicit structural and technological logic in the outfitting welding task, the original weld set needs to be further abstracted from simple geometric objects into welding task units with clear semantic attributes. This embodiment decomposes the welding task semantically, unifying the description of weld geometry, structural connection functions, and welding process constraints, thereby forming a set of semantic welding units that can be directly called by subsequent scheduling and path planning algorithms.

[0048] The input for semantic decomposition includes at least: a set of welds. Each weld The parameterization description, including the weld centerline point set, has been completed in the previous steps. ; the corresponding tangential vector Reference Normal and welding torch reference target pose In addition, it also includes a set of component and assembly information. This describes the type, material, and thickness of the components connected by the weld, as well as the process rule base R, which provides the mapping relationship between weld types and process requirements. (Field resources and process status are also mentioned.) It can be used as an optional input to help determine whether certain process semantics hold true under the current conditions.

[0049] The output of semantic decomposition, a set of semantic welding units, can be represented as:

[0050] Each semantic welding unit Each is treated as an independently schedulable and executable welding object, which contains weld geometry segments, structural connection semantics, process semantics, and corresponding execution constraint attributes.

[0051] In practice, each semantic welding unit It should include at least the following information: (1) Set of geometric segments

[0052] This indicates that the unit consists of welds. From the sampling point index arrive It consists of continuous weld segments; (2) Structural connection semantic tags

[0053] Used to characterize the functional attributes of the welding unit in the structure, such as tack weld, load-bearing weld, sealing weld or reinforcement weld; (3) Process semantic tags

[0054] Used to characterize process constraints during welding, such as whether continuous welding is required, whether intermittent execution is allowed, whether resumption after interruption is allowed, and whether it is a multi-pass, multi-layer welding process. (4) Set of constraint attributes Including consecutive group identifiers Interruption allow flag Insertion permission mark and the set of prerequisite dependencies ; (5) Process Parameter Index

[0055] Used to associate with the corresponding welding specifications or process parameter tables.

[0056] The generation process of semantic welding units generally follows the workflow of "first geometric segmentation, then semantic annotation, and then constraint generation," that is, weld set. L First, it is divided into several geometric fragment sets {Γ k Then, structural and technological semantics are assigned to each segment, and finally, corresponding execution constraint attributes are generated.

[0057] During the geometric segmentation stage, because the same weld seam may correspond to different welding postures, accessibility conditions, or process requirements at different locations in outfitting welding scenarios, it is necessary to segment the weld seam... l j The weld is divided into several independently executable segments. The segmentation rules can combine multiple criteria. First, segmentation can be based on the geometry of the weld centerline, calculated by determining the discrete curvature index.

[0058] in Sampling points for the centerline of the weld. To prevent extremely small positive numbers with a denominator of zero. When curvature Or the angle between adjacent tangential vectors When the preset threshold is exceeded, a geometric dividing point is set at the corresponding position to ensure that the welding trajectory inside each welding unit is smooth and the process is stable.

[0059] Secondly, the segmentation can be performed based on the changes in the welding torch posture. The welding torch reference posture, already constructed in the previous steps, is as follows. It can be used to calculate the attitude change angle between adjacent sampling points:

[0060] in This is the trace operation on a matrix. When... When the set threshold is exceeded, the weld seam is divided to avoid excessive changes in the welding torch posture within a single welding unit, which could affect the welding quality.

[0061] Furthermore, the accessibility analysis results obtained in the preceding steps can be used to segment the weld seam into accessibility windows. Let the feasible joint solution set at the weld seam sampling point be... Define reachability flag

[0062] Then The continuous intervals are merged to form the reachable window of the weld, and candidate welding segments are generated accordingly. This method ensures that each generated semantic welding unit has at least one feasible execution solution in the current robot base state.

[0063] For weld segments located in heat-sensitive areas or critical structural sections, the set of heat-sensitive areas can be determined based on the design model or process specifications. At the weld sampling point The boundary positions are divided so that the welding sequence and heat input strategy can be individually controlled in the subsequent scheduling phase.

[0064] After completing the geometric segmentation, structural connection semantics are assigned to each welded segment. When the functional attributes of welds are clearly marked in the design model or BIM data, the corresponding attributes can be directly read as structural semantics. When there is no explicit marking, regular inference can be made based on the type of components connected by the weld, plate thickness, and connection form. For example, pipe butt welds tend to be marked as sealing welds or critical welds, while welds on supports or positioning components tend to be marked as positioning welds or temporary welds.

[0065] During the process semantic annotation stage, process semantic tags are generated for each welding unit based on the weld type, material thickness, and process rule library R. When the weld thickness is h k And the allowable coverage thickness for a single channel is h pass At that time, it can be done through

[0066] Estimate the number of passes required for this welding unit, where Indicates rounding up. When At that time, the welding unit is marked as a multi-pass, multi-layer welding, and the sequential dependency between the layers is established in the subsequent constraint modeling.

[0067] Furthermore, based on the weld type and quality requirements, it is determined whether the welding unit requires continuous welding, and a continuous welding indicator is defined. Geometric length of the welding unit

[0068] It can be used to assist in determining continuous welding requirements. When continuous welding is required, the unit is assigned to a continuous group. This is then treated as an indivisible object during subsequent scheduling. Based on this, the insertion permission flag is further determined. and interruption permission flag The value is determined by referring to a table based on the weld type, quality grade, process requirements, and on-site resource conditions. The above-mentioned determination method belongs to a mature rule system in the field of welding technology, and therefore will not be elaborated upon here.

[0069] Finally, a set of prerequisite dependencies is generated for each welding unit based on structural and process semantics. Dependencies can originate from assembly or positioning requirements, the sequence requirements of multi-pass and multi-layer welding, and on-site resource constraints such as back-side protection or preheating. When resources are not yet ready, corresponding virtual dependency nodes can be introduced as prerequisites. The above dependency generation rules are clearly feasible in engineering and can be directly used as input for subsequent welding unit constraint diagram modeling.

[0070] Through the above semantic decomposition steps, the original weld seam set is transformed into a semantic welding unit set with clear structure, explicit semantics, and complete constraints. S This lays the foundation for the subsequent construction of constraint graphs between welding units, generation of welding execution sequences, and semantic-level dynamic rescheduling.

[0071] III. Modeling Constraint Relationships of Welding Units After obtaining the set of semantic welding units, in order to ensure that the welding execution order reflects the objectively existing structural logic, process logic, and spatial and thermal effect constraints in the outfitting welding task, it is necessary to formally model the relationships between the semantic welding units. This embodiment uses a directed constraint graph to uniformly describe the constraint relationships between welding units, thereby providing a clear computational basis for the subsequent generation and dynamic rescheduling of welding execution sequences.

[0072] The constraint relationship model is represented as a directed graph.

[0073] Among them, the vertex set Semantic welding unit index, edge set This indicates the sequence or constraint relationships between welding units. If a directed edge exists... This indicates that during the welding task scheduling and execution process, the unit... a With unit b There are constraints that need to be satisfied, and these constraints are usually embodied in "elements". a Should be in unit b "Previously completed" or "unit" a With unit b "Not suitable for execution adjacent to each other."

[0074] To improve the engineering feasibility and scheduling flexibility of the model, edge sets E Based on the different sources of constraints, it is further divided into several type subsets:

[0075] in, This represents structural dependency constraints. This indicates process dependency constraints. Indicates continuity constraints, Indicates spatial interference constraints. This represents thermal input and deformation control constraints. In addition to endpoint information, each edge can also be assigned attributes such as constraint type, hard / soft constraint flag, and weight, so as to distinguish between constraints that must be satisfied and constraints that should be satisfied as much as possible during subsequent scheduling.

[0076] (1) Structural dependency constraints

[0077] Structural dependency constraints are used to express the inherent sequential logical relationships of welding tasks in terms of assembly positioning, load-bearing connections, and sealing structures. These constraints primarily originate from the set of pre-dependencies Pre(k) determined during the semantic decomposition phase. For any semantic welding unit SWU... k When its predecessor dependency set satisfies Pre(k) ≠ When, for any k′ ∈ Pre(k), a directed edge (k′,k) ∈ Estr is introduced in the constraint graph to indicate that welding unit k′ must be executed before welding unit k.

[0078] The aforementioned prerequisite dependencies can be derived from CAD / BIM assembly relationships, process rules between tack welds and critical welds, and structural semantic tags. The structural dependencies are derived from factors such as tooling clamping or support requirements. The methods described above for generating structural dependencies from assembly or process rules are typically formalized in engineering practice as rule tables or process flow cards, representing mature process organization methods in this field. Therefore, their specific rules will not be elaborated upon here.

[0079] (2) Process Dependence Constraints

[0080] Process dependency constraints are used to express the limitations imposed on the welding execution sequence by multi-pass, multi-layer welding, interlayer sequence, and process prerequisites such as back-side protection and preheating. When the number of passes for a certain welding unit is estimated during the semantic decomposition stage... In this case, the welding unit can be divided into sub-units numbered according to the layer pass. And establish the hierarchical relationship in the constraint diagram: This is to ensure that the welding process meets the requirements of the process for the sequence of layers.

[0081] In addition, when welding unit SWU k process semantics The instruction requires certain resource conditions (such as back protection, preheating completion) to be met before its execution. When the on-site resource status u(t) indicates that the corresponding conditions are not yet ready, a virtual node r can be introduced in the constraint diagram to represent the resource ready event, and a system can be established... This transforms resource constraints into equivalent prerequisites in topology sorting. This approach has clear operability in scheduling implementation; its specific implementation relies on the status acquisition of the field I / O or MES system, and is a mature engineering method, therefore it will not be elaborated further.

[0082] (3) Continuity constraints

[0083] Continuity constraints are used to express the requirement that certain welding tasks must be performed continuously in the process and cannot be split or inserted. The semantic decomposition stage has already assigned continuous group identifiers to welding units. g k For satisfying g k = The set of welding units with g>0 needs to be guaranteed in the constraint model to appear continuously in the welding execution sequence and not be inserted by other welding units.

[0084] In engineering implementation, continuous groups can be aggregated into a supernode for processing. The internal order of the supernode can be determined according to the geometric adjacency relationship of welds or process rules, while only the overall preceding and succeeding constraints are exposed to the outside. This approach does not change the correctness of the constraint relationship and can significantly simplify the scale of the scheduling problem.

[0085] (4) Spatial interference constraints

[0086] Spatial interference constraints are used to express how the order of welding execution may affect the accessibility of the welding torch, the operating channel, or the sensor field of view in the complex space of outfitting. The core of these constraints is to determine whether executing a certain welding unit first will significantly reduce the executability of another welding unit.

[0087] Therefore, for each welding unit SWU k Construct an approximate representation of its welding torch operation channel. For its set of geometric segments... Take the TCP position of the welding torch

[0088] in, Sampling points for the centerline of the weld. The Z-axis direction represents the reference posture of the welding torch. d For weld spacing offset. Adjust each With equivalent radius By expanding and combining the results, we obtain an approximation of the tool sweep body of the welding unit:

[0089] in, Indicated by Centered on, with radius sphere, The dimensions are determined by the welding torch geometry and safety margin. Sweep body modeling and collision detection are mature technologies in robot path planning, therefore their specific algorithms will not be elaborated upon here.

[0090] Based on this, spatial interference constraints can be generated through reachability degradation determination or channel sweep volume overlap determination. For example, defining welding units. k The reachability rate

[0091] =1 indicates that a feasible joint solution set exists at this sampling point. Otherwise, it is 0, N k This represents the total number of sampling points. When executing a certain unit... a This led to another unit b When the reachability decreases significantly, a constraint diagram can be introduced ( b , a ) ∈ Espace to indicate the unit that should be executed first. b .

[0092] (5) Heat input and deformation control constraints

[0093] Thermal input constraints are used to control the welding sequence in thin plates, sealed structures, or heat-sensitive areas to prevent structural deformation caused by heat concentration. For welding unit k, from the process parameter index... Obtain welding current I k ,Voltage V k Welding speed v k and thermal efficiency or k The linear energy per unit length can be estimated:

[0094] in, For line energy, I k , V k , v k and or k All are provided by WPS or the process library. Combined with the welding unit length. L k The estimated total heat input can be obtained as follows:

[0095] Furthermore, the welding unit is defined as being located in the heat-sensitive region Ω. h The length ratio inside

[0096] in, This is an indicator function. When Larger and When the threshold is exceeded, the unit can be marked as a critical unit for thermal control, and thermal constraint edges can be introduced for adjacent units that are spatially close. E heat This limits its continuous execution. The specific cooling time or threshold is usually determined by process specifications or experience databases and is a mature process setting.

[0097] To support the dynamic generation and adjustment of subsequent welding execution sequences, each constraint edge ( a , b In addition to type, it can also include attributes such as whether it is a hard constraint, weight, and related threshold parameters. Structural dependencies, inter-layer dependencies, and continuity constraints are usually treated as hard constraints, while spatial interference and thermal input constraints can be set as soft constraints according to the risk level.

[0098] The constraint graph G can be updated as the field state s(t) changes during system operation. When obstacles, robot base pose, or resource states change, the relevant reachability indices, spatial interference relationships, or resource prerequisites are recalculated, thereby affecting E. space E proc or E heat Some edges in the graph can be added, deleted, or have their weights adjusted. Through this mechanism, the constraint graph can serve as real-time input to a dynamic scheduler, enabling semantic-level rescheduling of welding tasks.

[0099] IV. Determination of executable welding units based on site conditions During welding task execution, the outfitting site conditions are highly dynamic. Whether a welding unit is executable at any given moment depends not only on predetermined process and structural constraints, but also on various real-time factors such as robot base pose, accessibility, spatial obstructions, safety conditions, and continuous welding requirements. Therefore, before generating the welding execution sequence, it is necessary to determine the executability of semantic welding units based on the current site conditions, thereby selecting the set of welding objects that can be selected by the scheduler at the current moment.

[0100] At any moment t The system input consists of the field state vector s(t) and the welding element constraint diagram G=(V, E) constructed in the previous steps. The output of this step is the set of executable welding elements at the current moment:

[0101] Among them, if k ∈ E(t) represents the semantic welding unit SWU k If the execution prerequisites are met under the current on-site conditions, it can be considered as a candidate for the next welding scheduling step.

[0102] To determine whether the dependencies between each welding unit have been satisfied, a completion state variable is introduced:

[0103] This state variable is updated after welding is completed, quality acceptance is passed, or manual confirmation is received.

[0104] For any candidate welding unit SWU k For a unit to enter the executable set E(t), it must simultaneously meet the following four necessary conditions: First, all structural and process-related hard constraints must be satisfied; second, it must be reachable under the current base pose and environmental conditions, and there must be no risk of interruption when continuous welding is required; third, safety and obstruction conditions such as obstacles, personnel, and safe distances must be met; and fourth, when the welding unit belongs to a continuous welding group, the entire continuous group must be executable.

[0105] (1) Determination of dependency satisfaction conditions The semantic decomposition and constraint modeling stage has been completed for each welding unit SWU k The set of its preceding dependencies, Pre(k), is given. Based on this set, a dependency satisfaction indicator function is defined:

[0106] Pre(k) may include structural dependencies, process prerequisites, inter-layer dependencies, or resource virtual nodes, etc. This represents the completed state of the corresponding welding unit. When When =1, it indicates a welding unit. k The conditions for implementation have been met.

[0107] In engineering implementation, equivalence determination can also be performed directly based on the hard constraint edges in the constraint graph G. The set of all hard constraint predecessor nodes of welding element k is denoted as:

[0108] in Indicates a constraint edge. This indicates that the edge is a hard constraint. The conditions for a hard constraint to satisfy can be expressed as:

[0109] The two determination methods mentioned above are logically equivalent, and either one can be chosen in engineering implementation.

[0110] (2) Determination of reachability conditions Welding Unit SWUk The geometric extent is determined by the set of its geometric segments. Description. Sampling points for each weld within the segment. In the preceding steps, a set of feasible joint solutions has been obtained through inverse kinematics solving and collision detection. Set an reachability threshold. The reachability determination indicator function is obtained as follows:

[0111] threshold The configuration depends on whether a small number of endpoints are allowed to be unreachable, and the specific value depends on the device accuracy and process tolerance.

[0112] For welding units requiring continuous welding, merely satisfying the reachability condition is insufficient to guarantee the continuity of joint solutions during execution. To avoid mid-welding discontinuities, it is necessary to further determine whether a continuous feasible solution chain exists along the weld. For adjacent sampling points n and n+1, if... and satisfy:

[0113] Then it is assumed that the two points can be continuously connected in the joint space. Wherein, q , Joint angle vector, The maximum allowable joint jump variable between adjacent sampling points can be determined by the robot's maximum joint angular velocity. With trajectory interpolation period The calculation yielded: .

[0114] (3) Determination of safety and obstruction conditions In addition to satisfying the dependency and reachability conditions, limitations related to obstacles, personnel, and safe distances during the welding process must also be considered. The preliminary steps have defined the obstacle set O(t) and the minimum distance to each obstacle. For the welding unit SWU k Using its tools to sweep the body C k To determine security, the following definitions apply:

[0115] in d safe The safe distance threshold is determined by on-site safety regulations.

[0116] In addition, when the system maintains the restricted area set If the tool sweep of a welding unit overlaps with a no-entry zone, then that unit is determined to be unexecutable at that moment.

[0117] The construction and updating of restricted areas rely on personnel detection and security systems.

[0118] V. Dynamic Generation of Welding Execution Sequence Building upon the previous steps, this step further dynamically generates a welding execution sequence that satisfies the constraints and has the optimal execution cost, based on the constraint relationships between welding units and the robot's current execution state. This execution sequence is not generated all at once until all welding tasks are completed; instead, it uses a rolling window approach, updating in real-time according to changes in the on-site conditions during welding execution. This adapts to the complex working conditions and frequent interference within the outfitting space.

[0119] At the present moment t Input: The set of executable welding objects E(t) obtained from the previous stage; the welding unit constraint graph G=(V, E) and its edge attributes, where the edge type Used to distinguish between structural constraints, process constraints, continuity constraints, and thermal / spatial soft constraints; hard constraint markers. ∈{0,1} indicates whether the constraint must be satisfied; soft constraint weight. This indicates the degree of penalty for violating the constraint; the welding unit completion state vector. The on-site state vector s(t) includes the robot's base pose. The obstacles set O(t) and resource state u(t); and the geometric information of each welding unit. Welding torch target pose sequence Feasible joint set and process parameter index The corresponding welding current I k ,Voltage V k Welding speed v k and thermal efficiency or k .

[0120] Based on the above input, this step outputs the next welding sequence to be executed:

[0121] in, H The length of the scroll window. p i Indicates the first i The sequence P(t) consists of the welding objects to be executed. The sequence P(t) completes the first execution object at a time. p 1. It can be regenerated after a significant change in the on-site state, thereby achieving dynamic scheduling.

[0122] To ensure that the generated execution sequence meets the welding process and structural safety requirements, the executable objects are first screened using hard constraints. For any candidate object... k The set of its hard-constrained predecessor nodes is denoted as:

[0123] in This represents a directed edge in a constraint graph. This indicates that the edge is a hard constraint. (Based on the completion state of the welding unit) Define the hard constraint satisfaction indicator function:

[0124] This yields a candidate set that simultaneously satisfies both the executability criterion and the hard constraint conditions at the current moment:

[0125] In the candidate set M(t), each object is ranked and selected using a cost function. To comprehensively reflect robot motion efficiency, welding posture stability, soft constraint risk, and insertion strategy requirements, a comprehensive cost function is defined for the candidate welding object k:

[0126] Where, cur represents the robot joint state and welding torch posture corresponding to the currently executing object; l 1~ l 4 is a non-negative weighting coefficient used to adjust the relative importance of each cost item. Its specific value can be configured according to the equipment characteristics and process requirements. It is a routine parameter calibration issue, so it will not be discussed further.

[0127] Movement cost J move Used to measure the current joint state The movement cost to the starting position of the candidate welding unit. Let the starting sampling point of candidate unit k be... Its feasible joint solution set is as follows The initial target joint solution is defined as:

[0128] in, W Let be the joint weighting matrix, which can be an identity matrix. The movement cost is defined as:

[0129] This quantity can be directly calculated from the current joint state fed back by the robot controller and the results of the previous inverse kinematics solution.

[0130] attitude switching cost Jpose This is used to reflect the amplitude of changes in welding torch posture, thereby reducing the impact of frequent and large-amplitude posture changes on welding stability. Let the current welding torch rotation matrix be... The target posture of the welding torch at the starting point of the candidate unit is The attitude difference angle is defined as:

[0131] And take:

[0132] Soft constraint penalty item J soft This is used to handle constraints such as hot input coupling and spatial channel conflicts that are unsuitable for adjacent execution but are allowed to be violated. Let the previous execution object be `cur`, then:

[0133] in, This indicates the degree of soft constraint violation. For thermally constrained edges, it can be based on the thermal input of the welding unit.

[0134] and unit center distance Construct a heat penalty function; for spatially constrained edges, a tool sweep volume can be used. C k The volume overlap ratio is used to construct a penalty function. The related volume and distance calculations are mature geometric processing methods, so their specific implementations will not be elaborated upon.

[0135] Risks and costs of gap filling J risk This is used to reflect the insertion and interruptibility of welding units at the process semantic level. Let the insertion permission flag for a welding unit be... Interrupt enable flag is ,but:

[0136] in and This represents the penalty coefficient. The relevant semantic markers are obtained from the preceding semantic decomposition and process rule base determination; their construction process is part of the engineering configuration, and therefore will be omitted here.

[0137] At time t, if the candidate set M(t) is not empty, then the object with the lowest cost is selected as the next object to be executed:

[0138] To improve system real-time performance and stability, a rolling window approach is used to generate execution sequences of length H. Temporary joint states are defined. With temporary completion vector Initially equal to and .right i = 1 to H are executed sequentially: based on Choose the one with the lowest cost from the updated candidate set. p i Treat it as completed and update At the same time Update to the joint state at the end of the welding unit. The end joint state can be selected from the feasible solution set of the end sampling point, choosing the solution that is continuous with the starting solution and has the lowest cost. This selection process belongs to the conventional trajectory stitching technique, so it will not be elaborated further.

[0139] When a critical welding unit is temporarily unavailable due to resource shortages or spatial constraints, the scheduling module prioritizes selecting from the set of welding units that allow for insertion to maintain the overall job cycle time. If no object currently satisfies both the hard constraints and the executable condition, i.e., M(t) = ... If the blockage occurs, corresponding processing is triggered based on the cause, including waiting for resources to become available, adjusting the robot's base posture, or waiting for the safety prohibition to be lifted, and the execution sequence is regenerated after the state update. All of the above processing is based on mature industrial control and safety mechanisms and does not affect the implementation of this method, so it will not be elaborated further here.

[0140] During the welding process, when a welding object is completed, the site state changes significantly, or the constraint diagram structure is updated, a rolling rescheduling is triggered to recalculate E(t) and P(t), thereby realizing semantic-level dynamic programming of welding paths for complex outfitting spaces.

[0141] VI. Welding Unit-Level Path Generation Obtain the welding execution sequence This step then further targets each semantic welding unit (SWU) in the sequence. k A welding path that can be directly executed by the robot is generated. The welding path includes at least the geometric path skeleton along the weld seam, the welding torch posture sequence matching the welding process, the corresponding joint space motion trajectory, and the time parameterization result that meets the welding speed requirements, and the stability of the welding process is ensured under obstacle avoidance and motion continuity constraints. For adjacent welding units belonging to the same continuous welding group, they are further seamlessly spliced ​​at the trajectory level to avoid interruptions in the welding process.

[0142] For ease of explanation, the following descriptions are based on the field state at time t. The field state vector s(t) includes at least the robot base pose. And the obstacle set O(t). The geometric information of the weld has been obtained through the preceding steps, specifically the centerline sampling points. In addition to local tangential and normal geometric quantities; when the above geometric quantities originate from weld seam sensing correction The acquisition process is based on mature visual measurement or point cloud processing methods, so it will not be discussed further here.

[0143] 1. Construction and resampling of the geometric path skeleton of the welding unit SWU welding unit k Its corresponding geometric range is determined by the set of fragments. It means that among them j Number the weld seam. This defines the index range for the weld sampling points. The weld centerline sampling points within the segment are then sequentially stitched together to obtain the geometric path skeleton of the welding unit.

[0144] These are sampling points for the weld centerline, obtained through CAD / BIM weld centerline discretization or on-site measurement.

[0145] To ensure the stability of subsequent attitude and velocity planning, the geometric path skeleton is... C k Resampling is performed based on arc length. First, the cumulative arc length of the path points is calculated:

[0146] in c i Indicates the first [number] after splicing i Path points, N k This represents the total number of path points. Given the resampling interval. In the interval Generate parameter sequences at equal intervals. Resampled path points are obtained through linear interpolation or spline interpolation. Interpolation methods are common numerical calculation techniques, and mature interpolation algorithms can be directly used in engineering applications.

[0147] 2. Construction of the local coordinate system of the weld and establishment of welding torch posture constraints At each resampling path point At this point, the local tangential vector of the weld is calculated by the difference between adjacent points:

[0148] The endpoints can be replaced by forward or backward differential vectors. The tangential vector... Indicates the direction of welding movement.

[0149] Local normal vector of weld Normals can be obtained from the surface normals of adjacent plates, CAD weld attributes, or through local point cloud plane fitting. Normal acquisition is a mature geometric processing method, so its specific implementation will not be elaborated further.

[0150] Based on this, a local right-handed coordinate system for the weld is established:

[0151] And on Unitize, and at the same time through Orthogonal corrections are performed to reduce numerical errors.

[0152] Based on this, according to the semantic parameters of the welding process (indexed by the process parameters) (Or given by the process semantic identifier) ​​Determine the attitude constraints of the welding torch relative to the weld. Let the weld distance offset be... d k The expected TCP position for the welding torch is:

[0153] in Determined by WPS or process specifications.

[0154] Let the welding working angle be The angle of travel is The welding torch direction vector can be obtained by rotating the local coordinate system:

[0155] in , Let represent the rotation matrices about the local y-axis and z-axis, respectively. If a weld is not sensitive to the travel angle, we can let =0. Further construct the remaining axes of the welding torch coordinate system:

[0156] This yields the welding torch rotation matrix:

[0157] And construct the target pose of the welding torch:

[0158] 3. Inverse kinematics solution and continuous joint unlinking screening Transform the welding torch target pose to the robot base coordinate system:

[0159] in The robot's base pose is obtained through localization or calibration.

[0160] For each sampling point m, for Inverse kinematics solutions are performed, and feasible joint solutions are obtained by screening under joint constraints and obstacle avoidance constraints.

[0161] in For positive kinematic functions, This defines the joint's limiting range. Collision detection is implemented using existing, mature methods, so its internal details will not be elaborated upon.

[0162] To ensure the continuity of joint movement during the welding process, from each Select a continuous joint to unchain Define the joint jump variable for adjacent points:

[0163] And set the maximum allowed jump threshold:

[0164] in For the maximum joint angular velocity, The interpolation period is denoted as .

[0165] Continuous unlinking can be achieved by minimizing the following objective function:

[0166] and satisfy This problem can be solved using dynamic programming or a greedy approximation, and it falls under the category of a conventional sequence optimization problem, so the specific algorithm details will not be elaborated upon.

[0167] 4. Joint trajectory generation, time parameterization, and continuous welding unit splicing After obtaining the discrete joint sequence, continuous joint trajectories are generated through spline interpolation. And according to the welding speed v k Perform time parameterization. The relationship between arc length and time satisfies:

[0168] The discrete time increment is:

[0169] If necessary, time can be scaled up while meeting the constraints of maximum joint speed and acceleration. Time scaling is a common technology in industrial robots, so it will be briefly described here.

[0170] When a welding unit belongs to a continuous welding group, the trajectories of adjacent welding units are spliced ​​together according to the order within the group. Let the end joint state of the previous unit be... The next unit starts from the joint solution Choose the one closest to it:

[0171] And require When the conditions are met, the trajectories of each unit are connected in series on the time axis to form a group-level continuous trajectory; if the conditions are not met, a short transition segment is inserted for adjustment. The generation method of the transition segment belongs to conventional local programming, so it will not be elaborated further.

[0172] The generated welding trajectory needs to be verified before output to ensure that it meets joint limits, obstacle avoidance, and joint velocity and acceleration constraints throughout the execution process. Trajectory verification is a standard verification process before industrial robot execution and can be completed using mature implementation methods, so its specific details are briefly described here. If the verification fails, the process returns to adjust the welding torch posture parameters or path resampling density until an executable welding trajectory is obtained.

[0173] VII. Welding Execution Monitoring and Semantic-Level Rescheduling In the actual execution of ship outfitting welding, the on-site environment and process conditions have significant uncertainties. For example, the temporary entry of obstacles, changes in tooling positions, local deformation of welds due to heat input, and fluctuations in welding power output may all cause the originally generated executable set E(t), welding execution sequence P(t), or welding unit-level path to no longer meet reachability or safety requirements in subsequent execution stages. To ensure the continuity and stability of welding tasks in a dynamic environment, it is necessary to continuously monitor the on-site status during welding execution. When a change event affecting subsequent welding tasks is detected, the constraint relationship is updated at the semantic welding unit level and a rescheduling is triggered. This allows for adaptive adjustment of the remaining welding tasks without destroying the completed welding results.

[0174] During this process, completed welding units remain frozen and are no longer scheduled; welding units that are currently executing and are semantically not allowed to be interrupted are prioritized for execution to a safe node; the remaining welding units are reordered according to the updated constraint relationships, and welding paths are regenerated only for the affected welding units, so as to achieve fast, stable and computationally cost-controlled path adjustment.

[0175] After semantic-level rearrangement and local replanning are completed, the new execution sequence and updated welding path are written into the execution process, and the process enters the next monitoring cycle. If an influencing event is detected again in a subsequent cycle, the above monitoring, freezing, constraint updating, rearrangement, and replanning process is repeated to achieve continuous self-adaptation during the welding execution process.

[0176] Example 2 Based on the above method, this embodiment provides a welding path planning system for a ship outfitting welding robot. The overall goal of this system is not just to output an offline trajectory, but to organize the welding task into a set of calculable, updatable, and roll-adjustable plans and paths, under the premise of ensuring that hard constraints such as structural dependence and process continuity are strictly met. This significantly enhances the welding robot's continuous operation capability, robustness, and engineering feasibility in complex outfitting spaces, reduces reliance on human experience, and improves the consistency of welding quality.

[0177] like Figure 2 As shown, the system first includes a field state modeling unit, which is used to construct and maintain the field state model s(t), and the information it organizes includes at least the pose of the robot base in the world coordinate system. Obstacle set O(t), weld geometry correction information And the welding process resource status u(t). Among them, O(t) is used to unify the target pose of the welding torch with the kinematics solution of the robot to the same coordinate reference; O(t) is used to characterize the spatial occupancy state of outfitting structures, tooling and temporary objects and to support accessibility and spatial interference determination. Used to reflect the actual spatial position of the weld at the current moment and update the weld centerline sampling point accordingly. u(t) is used to characterize the readiness of welding resources and the availability of process conditions to ensure that process dependencies can be properly constrained during scheduling and execution.

[0178] Building upon the field condition model, the system further includes a semantic welding task decomposition unit, which transforms the input weld geometry and process parameter information into semantic welding units (SWUs). k The process uses sampling points along the weld centerline. Based on the continuity and segmentation information, weld geometric segments belonging to the same semantic welding unit are represented as follows: ,in, j Indicates the weld number, and These represent the starting and ending sampling point indices of the weld within the geometric segment, respectively. Based on these, structural connection semantics are formed by combining outfitting structure connection relationships, and process dependency semantics and continuity semantics are formed by combining welding process constraints, ensuring that each SWU... k These all become the basic operational objects for subsequent scheduling, executable determination, and path generation. Through this semantic decomposition, the system can fundamentally avoid the problems caused by existing technologies treating weld seams as independent geometric curves, which result in manual sequencing, difficulty in expressing constraints, and difficulty in adjusting changes. This makes the sequential relationship of welding tasks, continuity requirements, and interleaved execution strategies have a computable basis.

[0179] The system also includes a constraint relationship model construction unit, used to establish a constraint relationship graph G=(V, E) between semantic welding units. The node set V consists of each SWU... k The edge set E is used to express the relationships between semantic welding units, such as structural dependency constraints, welding process dependency constraints, and spatial interference constraints. Hard / soft attributes and importance representations are assigned to the edges, for example, through... h a,b Instructions to SWU a with SWU b Whether the constraints between them are hard constraints that must be satisfied is determined by... w a,b The constraint model represents the importance of constraints, thus ensuring that the scheduling process can both prevent the hard constraints from being violated and comprehensively weigh factors such as space and on-site conditions. On the one hand, this constraint relationship model supports the semantic scheduling objective of "which welds to weld first, which must be continuous, and which can be inserted", and on the other hand, it provides an "updatable structured dependency carrier" for subsequent dynamic rescheduling, so that plan adjustments no longer depend on a complete overhaul and recalculation.

[0180] To link the constraint model with the field conditions, the system further includes an executable welding unit determination unit, which determines the set of executable welding units E(t) at time t by combining s(t) and G=(V, E). This functional unit ensures, during determination, that hard constraints such as structural and process dependencies are satisfied, and also considers the current base pose. The system uses the obstacle set O(t) to assess spatial accessibility and interference risk, thus unifying the logically defined weld with the practically feasible weld in the same decision process. The resulting E(t) provides a candidate set for generating subsequent execution sequences, enabling the system to adaptively select the most suitable, safest, and most accessible welding object within the complex outfitting space, avoiding the disconnect between a feasible plan and an unexecutable site.

[0181] Based on the executable set, the system includes a welding execution sequence generation unit. This unit generates a welding execution sequence P(t) based on E(t) while satisfying the hard constraints of the constraint graph. This sequence is then used as the output of the rolling plan to drive subsequent path generation and execution. This sequence generation process is tightly coupled with the constraint graph, allowing the execution order to directly reflect the inherent constraints of the structure and process, and enabling updates as the field conditions change. Unlike traditional technologies that can only execute pre-given paths, this system outputs a set of execution sequences and paths that can be updated with time t, giving the robot the computational scheduling capability for continuous welding and gap-filling execution strategies.

[0182] To translate the plan into an executable robot trajectory, the system also includes a welding path generation unit, which processes each SWU according to the unit sequence of P(t). kGenerate the corresponding welding path. This process uses... Sampling points along the defined weld centerline A geometric path skeleton is constructed, and the target pose sequence of the welding torch on the path is determined by combining welding process semantics; then, the base pose in the field conditions is used. The target pose is transformed into a robot kinematics solution reference to obtain a feasible robot motion path and generate a welding trajectory output for execution. For unit groups with continuous welding semantics, the system can perform continuous processing on adjacent units at the trajectory level to avoid unnecessary interruptions, thereby improving execution stability and weld formation consistency while meeting process continuity requirements.

[0183] In addition, the system also includes a state monitoring and rescheduling unit, which continuously updates s(t) during the welding process, and when O(t) is detected, When changes in u(t) affect subsequent welding tasks, the constraint relationship graph G=(V, E) and the executable set E(t) are updated at the semantic welding unit level, and the subsequent welding execution sequence P(t) is regenerated accordingly. During the rescheduling process, the system keeps the results of completed welding units unchanged, only adjusting the incomplete parts, and further follows the strategy of local reordering and local replanning only for affected semantic welding units: regenerate paths for affected objects, and reuse the original paths for unaffected objects, thereby avoiding the computational overhead and job cycle interruption caused by the overall overhaul, and significantly improving the continuous operation capability and robustness under complex field conditions.

[0184] The overall workflow of this system can be summarized as follows: First, the system constructs an initial site state model based on weld geometry, outfitting structure, and process parameters. Then, it constructs a constraint relationship diagram to express the structural and process dependencies and spatial interference relationships between units. Subsequently, it combines the site state and constraint diagram to determine the executability of each unit, forming E(t), and continuously generates execution sequences. Welding paths are then generated for each unit in the sequence and output to the robot for execution. During execution, the system continuously monitors the site state. When changes occur, such as obstacle changes, weld geometry correction updates, or changes in process resource status, the system updates constraints and the executable set at the semantic layer, regenerates subsequent execution sequences, and performs local path replanning for affected units. This ultimately forms a closed-loop process of state-constraint-executable set-execution sequence-path-execution-re-monitoring. Through this closed loop, the system can not only generate executable welding paths in a static environment but also maintain the continuous feasibility of plans and paths in dynamic, congested, and frequently disturbed outfitting sites.

[0185] The above system can execute the welding path planning method of the ship outfitting welding robot described in Embodiment 1, and has the corresponding functional modules and beneficial effects of the method. For technical details not described in detail in this embodiment, please refer to the welding path planning method provided in Embodiment 1 of the present invention.

[0186] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; under the concept of the present invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the present invention as described above, which are not provided in detail for the sake of brevity; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A welding path planning method for a ship outfitting welding robot, characterized in that, Including the following steps: S1. Obtain the weld geometry information, outfitting structure spatial information, and welding process parameter information of the outfitting welding operation area, and construct the initial site state model of the welding operation under a unified coordinate system; S2. Based on the weld geometry information and welding process parameter information, the welding task to be performed is semantically decomposed, and the overall welding task is divided into multiple semantic welding units. S3. Based on the structural dependencies, welding process dependencies, and spatial interference relationships between the semantic welding units, construct a constraint relationship model between the semantic welding units to describe the constraints of each semantic welding unit on the execution order and execution conditions. S4. Combining the on-site state model and the constraint relationship model, perform an executability determination on the semantic welding unit to determine the set of executable welding units that meet the constraint conditions and are spatially reachable at the current moment. S5. Generate a welding execution sequence based on the set of executable welding units, and generate corresponding welding paths for semantic welding units in sequence according to the welding execution sequence; S6. During the welding process, continuously monitor the on-site status. When changes are detected in the on-site obstacle status, weld geometry status, or welding process resource status that affect subsequent welding tasks, update the constraint relationship model and the set of executable welding units. S7. Under the updated constraint relationship model, the welding execution sequence is regenerated for the incomplete semantic welding units, and the welding path is re-planned only for the semantic welding units affected by the changes in the field state, thereby realizing the dynamic adjustment of the welding path planning.

2. The welding path planning method as described in claim 1, characterized in that, The construction of the field state model in step S1 includes the following steps: S101. Obtain the pose information of the welding robot base in the world coordinate system, denoted as... , This indicates the spatial pose of the welding robot base relative to the unified coordinate system; S102. Obtain the spatial distribution information of obstacles in the outfitting welding operation area, and represent the spatial distribution information of obstacles as an obstacle set O(t), where O(t) is used to describe the spatial occupancy status of outfitting structures, tooling and temporary objects in the welding operation area. S103. Obtain the geometric correction information corresponding to each weld. Based on the geometric correction information, sampling points of the weld centerline are determined. The information is updated to characterize the actual spatial position of the weld at the current moment; at the same time, the status information of welding process resources related to the welding operation is obtained, denoted as u(t), which is used to characterize the availability status of welding power source and welding process conditions; S104. The welding robot base posture, obstacle set, weld geometry correction information and welding process resource status are uniformly organized to construct an initial field state model s(t).

3. The welding path planning method as described in claim 2, characterized in that, The semantic decomposition described in step S2 includes the following steps: S201. Based on the acquired weld geometry information, each weld is segmented according to the continuity of its centerline sampling points in space, dividing the weld into several geometric segments, and representing the geometric segments belonging to the same welding task as a set of weld geometric segments. ,in, j Indicates the weld number, and These represent the starting and ending sampling point indices of the weld within the geometric segment, respectively. S202. Based on the obtained set of weld geometric segments, and in combination with the connection relationship of the weld in the outfitting structure, perform structural connection semantic determination on each weld geometric segment to distinguish the connection function undertaken by different welds in the outfitting structure, and merge weld geometric segments with the same or related structural connection semantics into the same welding task unit. S203. Combining the welding process parameter information corresponding to the weld, perform welding process semantic annotation on the welding task unit to characterize whether the welding task unit has continuous welding requirements, whether interruption is allowed, and the corresponding welding process conditions during the welding execution process. A welding task unit that simultaneously possesses the structural connection semantics and welding process semantics is defined as a semantic welding unit (SWU). k .

4. The welding path planning method as described in claim 3, characterized in that, The construction of the constraint relationship model described in step S3 includes the following steps: S301, using semantic welding unit SWU k The nodes are used to construct a constraint relationship graph G=(V, E), where V is the set of semantic welding unit nodes and E is the set of constraint relationship edges between semantic welding units; S302. Based on the structural connection relationship between semantic welding units, determine the semantic welding unit pairs that have sequential assembly or force dependence relationship in the outfitting structure, and establish structural dependence constraint edges between the corresponding semantic welding unit nodes in the constraint relationship diagram to limit the sequential relationship of the semantic welding units in the welding execution order. S303. Based on the welding process semantics corresponding to the semantic welding unit, determine the semantic welding unit pairs with continuous welding requirements, process sequence dependence or uninterrupted requirements, and establish welding process dependence constraint edges in the constraint relationship graph to limit the continuity and process sequence of the semantic welding unit in the welding execution process. S304. Based on the spatial distribution relationship of semantic welding units in the outfitting operation space, determine the semantic welding unit pairs that will affect the accessibility or operation space of the other semantic welding unit if one semantic welding unit is executed first, and establish spatial interference constraint edges in the constraint relationship diagram to describe the execution restrictions caused by space occupation between the semantic welding units. S305. Assign constraint attributes to each constraint edge in the constraint relationship diagram to distinguish between hard constraints that must be satisfied and soft constraints that can be adjusted under safety and process conditions. This is achieved through constraint indicator flags. h a,b Characterizing semantic welding units a With semantic welding unit b Whether the constraints between them are hard constraints, and by using constraint weights w a,b The importance of the constraint relationships between the semantic welding units is characterized, thereby constructing a semantic welding unit constraint relationship model.

5. The welding path planning method as described in claim 4, characterized in that, The determination of the set of executable welding units in step S4 includes the following steps: S401. Based on the constraint relationship model, for each semantic welding unit (SWU) k Determine the set of preceding constraint units in the constraint relationship graph, and determine whether the structural dependency constraints and welding process dependency constraints related to the semantic welding unit have been satisfied, so as to filter out semantic welding units that do not meet the sequential execution conditions at the current time. S402. Under the premise of satisfying the hard constraint conditions in the constraint relationship model, and in combination with the welding robot base posture in the field state model. And the obstacle set O(t), perform spatial reachability determination on the welding path or welding area corresponding to the semantic welding unit, so as to determine whether the welding robot can reach the welding position corresponding to the semantic welding unit without spatial interference under the current site conditions; S403. For semantic welding units that simultaneously satisfy the hard constraints in the constraint relationship model and pass the spatial reachability determination, they are determined as executable semantic welding units at the current time, and all the semantic welding units determined as executable constitute the set of executable welding units E(t) at the current time.

6. The welding path planning method as described in claim 5, characterized in that, The generation of the welding path in step S5 includes the following steps: S501, For the semantic welding unit (SWU) to be executed in the welding execution sequence. k Based on the set of weld geometric segments corresponding to this semantic welding unit Extracting sampling points of the weld centerline And construct the welding geometry path skeleton of the semantic welding unit according to the spatial order of the sampling points; S502. Based on the construction of the welding geometry path skeleton, and combined with the welding process semantics corresponding to the semantic welding unit, the spatial pose of the welding torch relative to the weld is determined for each sampling point in the welding geometry path skeleton, thereby obtaining the target pose sequence of the welding torch in the welding process. S503, Combining the welding robot base posture in the aforementioned field state model And the welding gun target pose sequence, and perform kinematic solution on the welding path corresponding to the semantic welding unit to obtain the feasible motion path of the welding robot when executing the semantic welding unit; S504. The obtained feasible motion path of the welding robot is used as the welding path corresponding to the semantic welding unit, and steps S501 to S503 are repeated sequentially for each semantic welding unit in the welding execution sequence to generate a complete welding path corresponding to the welding execution sequence.

7. The welding path planning method as described in claim 6, characterized in that, Step S6 includes: During the welding process, based on the field state model s(t), the obstacle set O(t) and weld geometry correction information within the welding operation area are processed. And the welding process resource status u(t) is updated in real time or periodically; When a change is detected in the obstacle set, weld geometry correction information, or welding process resource status, and this change affects the semantic welding unit (SWU) that has not yet been completed, k When the executability is affected, the constraint relationship graph G=(V, E) is updated based on the updated field state model s(t) to correct the spatial interference constraints or welding process dependency constraints between semantic welding units. Under the updated constraint relationship model, the executability of incomplete semantic welding units is re-performed, the set of executable welding units E(t) at the current time is updated, and the updated set of executable welding units is used for the generation of subsequent welding execution sequences and the adjustment of welding paths.

8. The welding path planning method as described in claim 7, characterized in that, Step S7 includes: Under the updated constraint graph G=(V, E), identify semantic welding units (SWUs) that have not yet completed welding. k Based on the updated constraint relationships, the executability of the incomplete semantic welding units is re-evaluated to determine the executable welding units that satisfy the structural dependency constraints, welding process dependency constraints, and spatial interference constraints under the current field conditions. Based on the set of executable welding units E(t) obtained by re-determination, without changing the execution results of the completed semantic welding units, the welding execution order of the incomplete semantic welding units is regenerated so that the welding execution order satisfies the sequential relationship and execution conditions defined in the updated constraint relationship model. In the regenerated welding execution sequence, the corresponding welding path generation step is re-executed only for semantic welding units whose constraint relationships or executability have changed due to changes in the field conditions. For semantic welding units that are not affected by the changes in the field conditions, their original welding paths remain unchanged, thereby realizing the local replanning and dynamic adjustment of welding path planning at the semantic welding unit level.

9. A welding path planning system based on the method of any one of claims 1 to 8, characterized in that, The system includes the following units: The on-site state modeling unit is used to acquire the weld geometry information, outfitting structure spatial information, welding robot base posture information, and welding process resource status information of the outfitting welding operation area, and to construct an on-site state model of the welding operation under a unified coordinate system to characterize the real-time spatial status of the welding robot, weld and obstacles during the welding operation. The welding task semantic decomposition unit is used to semantically decompose the welding task to be executed based on the weld geometry information and welding process parameter information, and divide the overall welding task into multiple semantic welding units that simultaneously have structural connection semantics and welding process semantics, so as to serve as the basic operation object for welding path planning and scheduling. The constraint relationship model construction unit is used to construct a constraint relationship model between the semantic welding units based on the structural connection relationship, welding process dependency relationship and spatial interference relationship between the semantic welding units, so as to describe the constraint relationship between the semantic welding units in terms of welding execution order and execution conditions. An executable welding unit determination unit is used to combine the field state model and the constraint relationship model to determine the executability of semantic welding units and identify a set of executable welding units that simultaneously meet the constraint conditions and are spatially reachable under the current field state. The welding execution sequence generation unit is used to generate a welding execution sequence based on the set of executable welding units, under the premise of satisfying the hard constraints in the constraint relationship model, so as to determine the welding execution order of the semantic welding units; The welding path generation unit is used to generate corresponding welding paths for the semantic welding unit in sequence according to the welding execution sequence, and output the welding trajectory that can be executed by the welding robot. The status monitoring and rescheduling unit is used to continuously monitor the field status during welding execution. When a change in the field status is detected that affects subsequent welding tasks, the field status model and constraint relationship model are updated, and welding execution sequences are regenerated for unfinished welding tasks at the semantic welding unit level. Welding path planning is only performed for semantic welding units affected by the change in the field status, thereby realizing dynamic adjustment of welding path planning.

Images