Spray code identification optimal path planning method based on intelligent optimization algorithm

Abstract

The invention discloses a code spraying identification optimal path planning method based on an intelligent optimization algorithm, and relates to the technical field of code spraying control, and the method comprises the steps: dispersing a code spraying region into a three-dimensional space, and constructing a dynamic environment model of marking parameter constraint; based on the code spraying task priority queue, a hybrid optimization algorithm is adopted to carry out collaborative optimization of marking parameters and path nodes, and an optimal path candidate set is generated; an obstacle avoidance model is constructed based on the optimal path candidate set, a collision-free path is generated, the pose of a nozzle is adjusted in combination with visual feedback, and an anti-interference code spraying action sequence is output; monitoring the operation feedback of the code spraying equipment, and carrying out the iterative updating of the marking parameter library and the algorithm rule. According to the method, the core advantages of multiple intelligent optimization algorithms are fused, a mixed collaborative optimization architecture is constructed, the limitation of a single algorithm in a complex scene is effectively broken through, and a traditional path planning algorithm is easily troubled by a local optimal solution and is difficult to meet multi-dimensional constraint conditions at the same time.

Description

Technical Field

[0001] The present invention relates to the field of inkjet coding control technology, and in particular to a method for planning an optimal path for inkjet coding based on an intelligent optimization algorithm. Background Art

[0002] As a key link in modern industrial production, inkjet coding and marking technology achieves multiple functions such as product traceability, anti-counterfeiting, and compliance management through the dynamic printing of text, graphics, and coded information. Its core operation relies on the coordinated control of the coding stroke and idle stroke, which together constitute the operating logic of the coding system. The coding stroke refers to the movement path of the printhead during the printing phase. It must be precisely matched with the surface characteristics of the product, the material type, and the printed content to ensure that the information is firmly attached and accurately positioned. The idle stroke is the movement trajectory of the printhead when it is not printing. It is used to connect different printing tasks or adjust the operating position. Its efficiency directly affects the overall production rhythm.

[0003] The connection defects between the coding stroke and the idle stroke have multiple impacts on the production system: first, efficiency bottlenecks. A high proportion of idle strokes will significantly slow down the production line speed, especially in high-speed sorting scenarios such as beverages and cigarettes. Millisecond-level delays may cause production capacity loss; second, quality risks. If mechanical vibrations or path deviations in the idle stroke are not corrected in time, the coding may be blurred, overlapped or misaligned, affecting the readability of the product traceability code; third, operating and maintenance costs rise. Frequent path adjustments accelerate nozzle wear and tear, and additional manual intervention is required to calibrate the equipment, weakening the advantages of automation.

[0004] In current technology, the optimization of the coding stroke mainly focuses on path planning algorithms and hardware response speed. Traditional path planning mostly uses static strategies, which makes it difficult to dynamically adapt to speed fluctuations or sudden shutdowns of the production line, resulting in the printhead frequently starting and stopping or circumventing during the idle stroke. This not only increases mechanical wear, but may also cause coding position errors due to inertial offset. For example, on a food packaging line, the inkjet printer needs to quickly switch the printing content of products of different specifications. If the idle stroke connection is not smooth, it is easy to cause information misalignment or batch confusion. At the hardware level, although high-precision motion control systems can reduce positioning deviations at the end of the idle stroke, the complex mechanical calibration process and high maintenance costs limit its popularity.

[0005] While existing technologies attempt to mitigate this issue through intelligent algorithms and sensor feedback, limitations remain. Path planning algorithms often overlook environmental variables (such as the impact of temperature and humidity on material deformation), leading to a disconnect between the intended trajectory and actual requirements. Real-time monitoring systems are limited by data processing latency, making it difficult to achieve instantaneous corrections on high-speed production lines. Furthermore, protocol barriers across equipment brands make it difficult for the inkjet printer to seamlessly integrate with other production line modules (such as sorters and labelers), further amplifying the negative effects of idle travel. Summary of the Invention

[0006] Based on this, it is necessary to provide an optimal path planning method for inkjet marking based on an intelligent optimization algorithm to address the above technical problems.

[0007] The present invention provides a method for planning an optimal path for inkjet marking based on an intelligent optimization algorithm, comprising: S1. Discretize the coding area into three-dimensional space and construct a dynamic environment model constrained by marking parameters; S2. Based on the priority queue of coding tasks, a hybrid optimization algorithm is used to coordinate the optimization of marking parameters and path nodes to generate the optimal path candidate set; S3: Build an obstacle avoidance model based on the optimal path candidate set, generate a collision-free path, adjust the printhead posture based on visual feedback, and output an anti-interference coding action sequence; S4. Monitor the operation feedback of the inkjet coding equipment and iteratively update the marking parameter library and algorithm rules.

[0008] Furthermore, the coding area is discretized into three-dimensional space, and a dynamic environment model with marking parameter constraints is constructed, including: S11, obtaining the surface parameters of the material to be coded, the kinematic parameters of the coding equipment and sensor data, and generating an original data set of marking parameters; S12. Based on the original data set of marking parameters, the scanning rate interval, the inflection point jump delay and the line spacing control coefficient are calculated to construct the marking parameter constraint relationship matrix; S13, obtaining the real-time sensor data stream collected by the inkjet printer and establishing a dynamic constraint rule library; S14. Divide the coding area into several three-dimensional grids, mark the coding attributes of each grid, and introduce the nozzle response delay to expand the time dimension to generate a four-dimensional space-time dynamic environment model.

[0009] Furthermore, the real-time sensor data stream collected by the inkjet printer is obtained, and a dynamic constraint rule library is established, including: S131, using a Kalman filter to fuse visual positioning data, vibration signals, and temperature data in the real-time sensor data stream to eliminate temporal and spatial deviations between sensors; S132. Based on the marking parameter constraint relationship matrix, a marking parameter association model is constructed using a Bayesian network to convert the physical constraints into path rule constraints. The path rule constraints include surface roughness constraints, vibration amplitude constraints, and idle stroke scanning rate constraints. S133. Classify the complexity level according to the type of coding content, and dynamically adjust the path rule constraint weight according to the complexity level of the current coding content; S134. Integrate and record various path rule constraints and corresponding weights to build a dynamic rule library.

[0010] Furthermore, based on the priority queue of the coding task, a hybrid optimization algorithm is used to coordinate the optimization of the marking parameters and the path nodes, and the optimal path candidate set is generated, including: S21. Generate a mixed coding population of weeds and gray wolves through dual population initialization and parameter coupling; S22, based on the iteration of individuals within the hybrid coding population, cross-population information interaction is achieved to generate a hybrid search strategy that integrates global exploration and local exploitation; S23. Generate a path node sequence based on the individuals in the population, perform optimal solution according to the hybrid search strategy through multi-objective fitness modeling and constraint integration, and generate fitness scores and constraint tags; S24. Output a dynamically optimized optimal path candidate set based on the iterative results of the hybrid optimization algorithm.

[0011] Furthermore, through dual population initialization and parameter coupling, the weed-gray wolf mixed coding population is generated, including: S211. Initialize the weed population size and the wolf population size. The two populations share the same solution space. S212, encode each seed individual in the weed population as X IWO =( x 1,…, x n ;σ), where x i represents the path node coordinates, and i =1, 2, ..., n ;σ represents the diffusion radius; S213, encode the gray wolf individuals in the gray wolf population as X GWO =( x 1,…, x n , t 1,…, t m ),in, t j represents the marking parameters, and j =1, 2, ..., m ; S214, associating the linear decline parameter of the gray wolf population with the diffusion radius of the weed population, calculating a nonlinear attenuation factor, and when the nonlinear attenuation factor decays to a preset attenuation threshold, switching from global exploration of the weed population to local exploration of the gray wolf population; S215. When the population diversity index is less than a preset index threshold, the Gaussian perturbation mechanism of the weed population is activated to generate mutant individuals to fill the blind area of ​​the solution space.

[0012] Furthermore, based on the iteration of individuals within the hybrid coding population, cross-population information interaction is achieved, and a hybrid search strategy that integrates global exploration and local development is generated, including: S221. Set the iteration cycle. In the global exploration phase, generate the number of seeds based on the fitness value. Generate new seed coordinates through normal distribution. Keep the top 10% of the weed population in each generation and directly inject the α, β, and δ wolf candidate values ​​of the gray wolf population to achieve path-parameter coordinated optimization. S222. When the nonlinear attenuation factor reaches a preset attenuation threshold, the global exploration is switched to the local development phase. The positions of individual gray wolves are updated according to their levels. In combination with lens imaging, the mirror solution is learned in reverse, and the high-quality solutions are retained through Pareto front screening. S223. Monitor the population convergence speed during the global exploration phase in real time. When the convergence speed is less than the set convergence threshold, trigger the phase switch in advance and execute local development of the gray wolf population.

[0013] Furthermore, a path node sequence is generated based on the individuals in the population. Through multi-objective fitness modeling and constraint integration, the optimal solution is obtained according to the hybrid search strategy, and the fitness scores and constraint tags are generated, including: S231. Generate a path node sequence based on the iterative solution of the hybrid optimization algorithm, obtain marking parameter values, establish a composite objective function integrating economic objectives, quality objectives, and stability objectives, and assign optimization weights to each objective; S232. Mark solutions that violate path rule constraints, impose exponential penalties on them, and gradually tighten constraint tolerances to eliminate infeasible solutions. S233. Normalize the solution calculated by the composite objective function, calculate the comprehensive fitness score according to the optimization weight of each objective, eliminate all solutions marked as violating the constraints, and sort the remaining feasible solutions in descending order according to the comprehensive fitness score to generate a priority queue; if the priority queue is empty, select the solution with the least constraint violations as the temporary candidate.

[0014] Furthermore, the formula of the composite objective function is: ; ; ; Where, F econ Indicates the economic target score; ( x i , y i ) represents the coordinates of the path node; n Indicates the number of path nodes; t jump，j Indicates the jTheoretical time required for the secondary empty stroke movement; t delay，j Indicates the j Real-time delay compensation of the secondary inflection point; λ represents the penalty coefficient; m Indicates the number of marking parameters; F qual Indicates the quality target score; σ d Indicates the standard deviation of line distance; Δ d Indicates the line distance tolerance threshold; F stab represents the stability target score; v i Indicates the i The nozzle speed at the path node; Δ t Indicates the time interval between adjacent path nodes; a max Indicates the maximum acceleration of the nozzle.

[0015] Furthermore, an obstacle avoidance model is constructed based on the optimal path candidate set to generate a collision-free path. The printhead position is adjusted in combination with visual feedback, and the output anti-interference coding action sequence includes: S31. Generate a dynamic obstacle heat map using the laser radar point cloud data collected by the inkjet printer and evaluate the path feasibility of the optimal path candidate set; S32. Perform a random perturbation test on each feasible candidate path, calculate the actual execution time of the path, and evaluate the path risk including collision probability, execution deviation, and ink dot quality degradation; S33. Obtain the high-risk paths generated by the assessment, construct an adaptive sampling domain, generate constraint-aware paths, and form a coding table identification path through dynamic pruning and fusion; S34. Inversely infer the marking parameters based on the geometric characteristics of the inkjet marking path, output the action sequence of the inkjet marking device, and monitor the line distance deviation in real time. When the line distance deviation exceeds the preset threshold, parameter compensation is triggered.

[0016] Furthermore, the laser radar point cloud data collected by the inkjet printer is used to generate a dynamic obstacle heat map, and the feasibility of the optimal path candidate set is evaluated, including: S311. Overlaying the laser radar point cloud data with the dynamic environment model, incrementally updating the obstacle topology map, and generating a dynamic obstacle heat map; S312. Path features are extracted from the optimal path candidate set, and collision prediction is performed with the newly added obstacle position. If the minimum distance between the candidate path and the obstacle is less than the safety radius, it is marked as a high-risk path, and a path feasibility label is output based on the prediction result.

[0017] Furthermore, the high-risk paths generated by the evaluation are obtained, an adaptive sampling domain is constructed, a constraint-aware path is generated, and a coding table identification path is formed through dynamic pruning and fusion, including: S331. With the highest risk path as the center, an elliptical sampling domain is constructed, and an artificial potential field guidance factor is introduced to make the sampling points biased towards the direction of the high-quality path in the optimal path candidate set; S332. When the node is expanded, the feasibility of the marking parameters is simultaneously verified. If the curvature of the new path segment is greater than the preset curvature threshold, the path node is rejected and the scanning rate is degraded. A bidirectional path planning strategy is adopted to simultaneously expand the tree structure from the starting point and the end point of the risk segment. S333: splice the local corrected path segment with the non-risk segment in the optimal path candidate set, and use a B-spline curve for smooth transition to obtain a corrected inkjet marking path.

[0018] The beneficial effects of the present invention are: 1. By integrating the core strengths of multiple intelligent optimization algorithms, a hybrid collaborative optimization architecture was constructed, effectively overcoming the limitations of single algorithms in complex scenarios. Traditional path planning algorithms are susceptible to local optimal solutions and struggle to simultaneously meet multi-dimensional constraints (such as kinematic restrictions and environmental dynamics). By dynamically coupling the global exploration of Invasive Weed Optimization (IWO) with the local development strategy of Gray Wolf Optimization (GWO), joint optimization of path nodes and marking parameters is achieved. This collaborative mechanism not only enhances the algorithm's coverage of high-dimensional solution spaces but also avoids the risk of premature convergence through dynamic switching strategies, significantly improving the overall robustness of path planning. Furthermore, the algorithm supports flexible adjustment of multi-objective weights to adapt to the priority requirements of different scenarios. For example, in precision manufacturing scenarios, it prioritizes coding quality, while in logistics scenarios, it prioritizes efficiency, fully demonstrating the versatility and scalability of the method.

[0019] 2. In response to common dynamic disturbances in inkjet printing operations (such as mechanical vibration, changes in ink characteristics, and sudden obstacles), the present invention designs a real-time correction mechanism based on incremental perception; by fusing real-time sensor data, it can dynamically update the environmental model and quickly screen a set of candidate paths with high anti-interference properties.

[0020] 3. The optimization model incorporates inkjet process constraints (such as line spacing uniformity and ink dot coverage) and physical limitations of the equipment (such as servo motor response delay and printhead motion curvature) into a framework for jointly processing multi-dimensional constraints. Using a dynamic penalty function and a feasible solution prioritization mechanism, the algorithm iteratively eliminates inferior solutions that violate constraints. It also utilizes a Pareto frontier screening strategy to retain non-dominated solutions, ensuring that the final set of candidate paths achieves both engineering feasibility and multi-objective optimality. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 The present invention is a flowchart of an optimal path planning method for inkjet marking based on an intelligent optimization algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0023] See also Figure 1 , provides an optimal path planning method for inkjet marking based on an intelligent optimization algorithm, including: S1. Discretize the coding area into three-dimensional space and construct a dynamic environment model constrained by marking parameters.

[0024] In the description of the present invention, the coding area is discretized into three-dimensional space, and a dynamic environment model with marking parameter constraints is constructed, which includes: S11. Obtain the surface parameters (roughness / adhesion) of the material to be coded, the kinematic parameters (maximum acceleration / air scan rate) and sensor data (vibration / temperature / humidity) of the coding equipment, and generate the original data set of marking parameters.

[0025] Specifically, the device parameter collection reads the maximum scanning rate of the printhead servo motor, the inflection point jump delay, and the minimum interval time of the inkjet trigger signal through the device protocol (such as Modbus-TCP).

[0026] Material property testing uses a laser confocal microscope to obtain material surface roughness (Ra value ≤ 0.8μm) and adhesion data. Combined with the ink viscosity-pressure relationship curve (for example, the line width of UV ink at a pressure of 0.25MPa is 0.1mm), the minimum feasible line spacing is derived (for example, the line spacing on the aluminum alloy surface is ≥ 0.08mm).

[0027] Sensor data synchronization requires the integration of a vibration sensor (range ±5g) and an infrared temperature probe (accuracy ±0.5°C) to monitor the printhead jitter amplitude and ambient temperature and humidity in real time, and to establish an initial library of compensation parameters (for example, for every 10°C increase in temperature, the ink diffusion rate increases by 12%).

[0028] S12. Based on the original data set of marking parameters, the scanning rate interval, inflection point jump delay and line spacing control coefficient are calculated to construct the marking parameter constraint relationship matrix.

[0029] Specifically, the introduction and calculation methods of the scanning rate interval, inflection point jump delay and line spacing control coefficient include the following aspects: 1. Scan rate interval: Dynamically set the nozzle movement rate threshold per unit area based on material adhesion and ink viscosity to avoid ink dot diffusion or uneven line spacing due to excessive speed.

[0030] According to the ink jet dynamics model, the scanning rate ( v ) and line spacing ( d ), the calculation formula is: ; Where, Q is the ink flow rate (ml / s); w is the nozzle width (mm), and the optimal rate range of different materials (such as plastic packaging) is calibrated through experiments. v opt =800mm / s).

[0031] 2. Inflection point jump delay: Based on the nozzle servo motor response curve, the mechanical delay at the path inflection point is quantified and a jump time compensation model is established.

[0032] Based on the S-shaped acceleration and deceleration curve of the servo motor, the delay of direction switching at the inflection point is quantified ( t d ) and the path curvature radius ( r ), and its calculation formula is: ; Where, a max is the maximum acceleration; t response is the servo system response time (e.g. 1.5ms).

[0033] 3. Line spacing control coefficient: The minimum feasible line spacing is derived through the material surface energy parameters (such as the adhesion data of aluminum alloy LD10CZ) to ensure the clarity and wear resistance of the marking.

[0034] Combining the material surface energy and ink adhesion, a line spacing tolerance model is constructed, and its calculation formula is: ; Where, σ ink is the surface tension of the ink (e.g. 72mN / m); σ material is the surface energy of the material (such as 450mJ / m 2 ); k is the process correction factor (calibrated experimentally to be 0.12).

[0035] This step transforms marking parameters into physical constraints for path planning through multi-source data fusion. For example, in high-speed printing scenarios, the idle scanning rate needs to be adjusted synchronously with the conveyor belt speed to prevent code overlap or breakage caused by speed mismatch.

[0036] S13. Obtain the real-time sensor data stream collected by the inkjet printer and establish a dynamic constraint rule library.

[0037] In the description of the present invention, obtaining the real-time sensor data stream collected by the inkjet printer and establishing a dynamic constraint rule base include: S131. Use Kalman filter to fuse visual positioning data, vibration signals and temperature data in real-time sensor data stream to eliminate the temporal and spatial deviation between sensors.

[0038] Specifically, a Kalman filter is used to fuse visual positioning data (accuracy ±0.05mm), vibration signals (frequency 0-500Hz), and temperature data to eliminate temporal and spatial deviations between sensors. For example, in high temperature and high humidity environments, the scanning rate interval is dynamically adjusted based on the temperature and humidity compensation coefficient (for example, the rate decreases by 8% for every 10°C increase in temperature).

[0039] S132. Based on the marking parameter constraint relationship matrix, a marking parameter association model is constructed using a Bayesian network to convert physical constraints into path rule constraints. The path rule constraints include surface roughness constraints, vibration amplitude constraints, and idle stroke scanning rate constraints.

[0040] Specifically, for constraint rule generation, it is necessary to build a marking parameter association model based on the Bayesian network to transform physical constraints into path planning rules. For example, the following types of path rule constraints are established: Rule 1: If the surface roughness of the material Ra >1.6μm, the lower limit of line spacing is increased to 0.12mm (to avoid uneven ink penetration); Rule 2: If the vibration amplitude is detected to be greater than 0.3g, the inflection point jump delay compensation value is increased by 20% (to reduce the inkjet printing misalignment caused by mechanical shock); Rule 3: The idle scan rate must not exceed 80% of the device's maximum rate.

[0041] S133. Classify the complexity level according to the type of the coding content, and dynamically adjust the path rule constraint weight according to the complexity level of the current coding content.

[0042] Specifically, complexity levels are determined based on the type of coding content (e.g., simple characters, QR codes, high-precision graphics), and path rule constraint weights are dynamically adjusted. The following grading is recommended: 1. Low complexity (e.g., numbers / letters): efficiency is prioritized over quality, with 70% weighting for idle speed and 30% for line spacing uniformity. 2. High complexity (e.g., micron-level QR codes): quality is weighted to 80% and speed is reduced to 20%. A fuzzy logic controller is used to convert the complexity score (based on character density and number of corners) into weight coefficients for a multi-objective optimization function.

[0043] For example, in pharmaceutical packaging coding, QR codes must meet the ISO / IEC 15438 standard (line spacing tolerance of ±3μm). When the system detects a QR code complexity level of 5 (the highest), it automatically sets the line spacing uniformity weight to 0.8 and enables piezoelectric ceramic fine-tuning mode to ensure barcode readability.

[0044] S134. Integrate and record various path rule constraints and corresponding weights to build a dynamic rule library.

[0045] S14. Divide the coding area into several three-dimensional grids, mark the coding attributes of each grid, and introduce the nozzle response delay to expand the time dimension to generate a four-dimensional space-time dynamic environment model.

[0046] Specifically, the coding area is discretized into regular three-dimensional grid cells (usually 0.1mm×0.1mm×0.1mm cubes), and each grid is annotated with the following core coding attributes: 1. Feasibility of inkjet printing: According to the surface roughness of the material ( Ra Value), ink adhesion and other parameters dynamically determine whether printing is allowed. For example, the area with too high roughness ( Ra >1.6μm) is prohibited from printing to avoid uneven ink penetration.

[0047] 2. Dynamic obstacles: The ToF sensor detects workpiece deviation or foreign objects on the conveyor belt in real time and marks them as inaccessible areas.

[0048] 3. Marking parameter constraints: Inherit process parameters such as scanning rate and line spacing threshold to generate a rate-accuracy mapping table. For example, the allowed scanning rate for plastic surfaces is 800mm / s, while the metal surface needs to be reduced to 500mm / s to ensure ink dot uniformity.

[0049] Using voxel modeling, the continuous space is divided into discrete units, each storing material, geometric features, and dynamic state data. Three-dimensional spatial features are extracted through point cloud annotation technology, and real-time spatial topology is obtained using LiDAR or high-precision scanning equipment.

[0050] At the same time, the nozzle response delay parameter is introduced to expand the three-dimensional space model into a four-dimensional space-time topological structure. It is necessary to calculate the inflection point jump delay based on the S-shaped acceleration and deceleration curve of the nozzle servo motor, construct a time axis, and predict the workpiece position within the next 0.5 seconds (combined with the conveyor belt speed and acceleration model). The space-time occupancy probability distribution map is generated to avoid collisions between the nozzle and moving objects.

[0051] Finally, the time axis is bound to the spatial coordinates to generate a four-dimensional grid matrix containing position (X, Y, Z) and time (T). Each cell records the inkjet trigger status, delay compensation value, and environmental interference coefficient. The four-dimensional model is rendered on the digital twin platform and compared with the actual inkjet results (such as ink dot coverage and line spacing uniformity). If the line spacing deviation exceeds 5%, an iterative update of the rule base is triggered, such as optimizing the rate-line spacing relationship coefficient or adjusting the delay compensation weight.

[0052] S2. Based on the priority queue of coding tasks, a hybrid optimization algorithm is used to coordinate the optimization of marking parameters and path nodes to generate the optimal path candidate set.

[0053] It should be noted that Invasive Weed Optimization (IWO) is a nature-inspired metaheuristic method. Inspired by the dispersal strategies of weeds, it simulates the fundamental processes of weed seed dispersal, growth, reproduction, and competitive extinction in nature. It has the advantages of being easy to implement, having a simple structure, and having outstanding global search capabilities.

[0054] The Gray Wolf Optimization (GWO) algorithm, inspired by the natural hunting behavior of gray wolf packs, is a swarm intelligence heuristic. This algorithm has a simple mathematical model and a strong balance between exploration and exploitation, making it adaptable to complex problems in diverse fields and amenable to various improvements. In this model, the α wolf is the optimal solution, followed by the β wolf, then the δ wolf, and finally the ω wolf. Gray wolves use hunting techniques based on the α, β, and δ wolves, with the ω wolf acting as a follower. The main hunting phases of gray wolves are: stalking and pursuing prey, encircling and approaching prey, and attacking and apprehending prey.

[0055] In step S2, IWO and GWO collaborate in the following ways: 1. Dynamic coupling strategy: In the initial stage, IWO is mainly used to explore the solution space by spreading seeds over a large area; in the later stage, it switches to GWO and uses the three-wolf collaboration model to fine-tune the path nodes and marking parameters.

[0056] 2. Parameter complementation: IWO's random diffusion mechanism makes up for GWO's defect of easily falling into local optimality, while GWO's hierarchical collaboration mechanism improves IWO's convergence speed.

[0057] 3. Constraint integration: The hybrid algorithm uses marking parameters (such as scanning rate and line spacing tolerance) as constraints of the fitness function and eliminates illegal solutions through a dynamic penalty function.

[0058] Combining the two to form a hybrid optimization algorithm enhances global-local collaborative search capabilities. During the global exploration phase, IWO's seed diffusion mechanism rapidly locates potential high-quality areas within a complex path solution space. For example, in dynamic obstacle scenarios, IWO can generate a diverse set of candidate paths to avoid obstacles. During the local exploration phase, GWO refines IWO's initial solution through the collaboration of α, β, and δ wolves. Experiments have shown that GWO can shorten path lengths by 10%-15% while also reducing marking errors caused by mechanical vibration at inflection points.

[0059] This system also improves interference resistance and real-time performance. Sensor data is used to update the environmental model in real time (e.g., temperature changes cause fluctuations in ink diffusion rates). The hybrid algorithm triggers marking parameter compensation rules (e.g., scan rate degradation) to ensure path feasibility. The combination of IWO's parallel seed generation and GWO's rapid convergence enables the algorithm to converge within 200 iterations, meeting the real-time requirements of production lines (response time ≤ 50ms).

[0060] In the description of the present invention, based on the inkjet task priority queue, a hybrid optimization algorithm is used to perform collaborative optimization of marking parameters and path nodes, and the optimal path candidate set is generated, including: S21. Generate a weed-gray wolf mixed coding population through dual population initialization and parameter coupling.

[0061] In the description of the present invention, generating a weed-grey wolf mixed coding population through dual population initialization and parameter coupling includes: S211. Initialize the weed population size and the gray wolf population size. The two populations share the same solution space.

[0062] S212, encode each seed individual in the weed population as X IWO =( x 1,…, x n ;σ); where, x i represents the path node coordinates, and i =1, 2, ..., n ; σ represents the diffusion radius.

[0063] S213, encode the gray wolf individuals in the gray wolf population as X GWO =( x 1,…, x n , t 1,…, t m );in, t jrepresents the marking parameters, and j =1, 2, ..., m .

[0064] S214. Correlate the linear decline parameter of the gray wolf population with the diffusion radius of the weed population, calculate the nonlinear attenuation factor, and when the nonlinear attenuation factor decays to a preset attenuation threshold, switch from global exploration of the weed population to local development of the gray wolf population.

[0065] Among them, the nonlinear attenuation factor μ The switching threshold used to control global search (IWO) and local development (GWO) is calculated as follows: ; Where, a It represents the convergence factor of GWO, with an initial value of 2 and linearly decreasing to 0 with the number of iterations. It is used to adjust the search range. σ Indicates the IWO diffusion radius (initial value 5mm), which is dynamically adjusted according to the material (e.g. 3mm for plastic surface); k Indicates the current iteration number; k max Indicates the total number of iterations.

[0066] S215. When the population diversity index is less than a preset index threshold, the Gaussian perturbation mechanism of the weed population is activated to generate mutant individuals to fill the blind area of ​​the solution space.

[0067] Specifically, the Shannon entropy or Hamming distance can be calculated as the diversity index. <H threshold The perturbation mechanism is triggered when the path is too dense, applying Gaussian mutation to low-fitness weed individuals. For example, in complex marking path optimization, when the path node clustering is too high, the perturbation mechanism generates obstacle-avoiding path candidates to avoid local optimality.

[0068] S22. Based on the iteration of individuals within the hybrid coding population, cross-population information interaction is achieved to generate a hybrid search strategy that integrates global exploration and local development.

[0069] In the description of the present invention, based on the iteration of individuals within the hybrid coding population, cross-population information interaction is achieved, and a hybrid search strategy that combines global exploration and local development is generated, including: S221. Set the iteration cycle. In the global exploration phase, generate the number of seeds according to the fitness value. The new seed coordinates are generated through normal distribution. The top 10% individuals of the weed population are retained in each generation. The α, β, and δ wolf candidate values ​​of the gray wolf population are directly injected to achieve path-parameter coordinated optimization.

[0070] S222. When the nonlinear attenuation factor reaches the preset attenuation threshold, the global exploration is switched to the local development stage, the individual positions of the gray wolves are updated according to their levels, and the mirror solution is learned in reverse by combining lens imaging, and the high-quality solutions are retained through Pareto front screening.

[0071] S223. Monitor the population convergence speed during the global exploration phase in real time. When the convergence speed is less than the set convergence threshold, trigger the phase switch in advance and execute local development of the gray wolf population.

[0072] S23. Generate a path node sequence based on the individuals in the population, perform optimal solution according to the hybrid search strategy through multi-objective fitness modeling and constraint integration, and generate fitness scores and constraint tags.

[0073] In the description of the present invention, a path node sequence is generated based on population individuals, and the optimal solution is obtained according to a hybrid search strategy through multi-objective fitness modeling and constraint integration. The fitness scores and constraint tags are generated, including: S231. Generate a path node sequence based on the iterative solution of the hybrid optimization algorithm, obtain the marking parameter values, establish a composite objective function integrating the economic goal, quality goal and stability goal, and assign optimization weights to each goal.

[0074] In the description of the present invention, the formula of the composite objective function is: ; ; ; Where, F econ Indicates the economic target score; ( x i , y i ) represents the coordinates of the path node; n Indicates the number of path nodes; t jump，j Indicates the j Theoretical time required for the secondary empty stroke movement; t delay，j Indicates the j Real-time delay compensation of the secondary inflection point; λ represents the penalty coefficient; m Indicates the number of marking parameters; F qual Indicates the quality target score; σ d Indicates the standard deviation of line distance; Δ d Indicates the line distance tolerance threshold; F stab represents the stability target score; vi Indicates the i The nozzle speed at the path node; Δ t Indicates the time interval between adjacent path nodes; a max Indicates the maximum acceleration of the nozzle.

[0075] S232. Mark solutions that violate path rule constraints, impose exponential penalties on the solutions, and gradually tighten the constraint tolerance to eliminate infeasible solutions.

[0076] Specifically, a dynamic rule base (such as surface roughness constraints and vibration amplitude thresholds) is used to mark illegal solutions. For example, when the path curvature exceeds the maximum steering capability of the nozzle, the flag is triggered. An exponential penalty function is used, and its calculation formula is: ; Where, c i is the constraint violation amount; τ It is a dynamically tightened tolerance threshold, initially set to ±5% and gradually reduced to ±1% in the later stages.

[0077] After every five iterations, if the cumulative penalty value of a solution exceeds a threshold, it is removed from the population to avoid wasting resources. For example, in a logistics scenario, solutions with path lengths exceeding the limit by more than 10% are directly eliminated.

[0078] S233: Normalize the solutions calculated for the composite objective function, calculate a comprehensive fitness score based on the optimization weights of each objective, eliminate all solutions marked as violating constraints, and sort the remaining feasible solutions in descending order of their comprehensive fitness scores to create a priority queue. If the priority queue is empty, select the solution with the fewest constraint violations as a temporary candidate.

[0079] Specifically, each objective function value is mapped to the interval [0, 1], for example, through Min-Max normalization or Z-Score normalization. After removing all solutions marked as violating constraints, the remaining solutions are sorted in descending order of overall fitness to create a priority queue. For example, in drone path planning, the top 20% of solutions are considered high-priority candidates. A temporary candidate mechanism: If the queue is empty, the solution with the fewest constraint violations (e.g., the fewest violations or the lowest penalty) is selected. A risk assessment function (e.g., a weighted safety distance score) is used to determine the optimal compromise solution.

[0080] S24. Based on the iterative results of the hybrid optimization algorithm, a dynamically optimized optimal path candidate set (a candidate set consisting of various solutions in the priority queue) is output.

[0081] S3. Build an obstacle avoidance model based on the optimal path candidate set to generate a collision-free path, adjust the nozzle position based on visual feedback, and output an anti-interference coding action sequence.

[0082] In the description of the present invention, an obstacle avoidance model is constructed based on the optimal path candidate set to generate a collision-free path, and the printhead posture is adjusted in combination with visual feedback to output an anti-interference coding action sequence, including: S31. Generate a dynamic obstacle heat map using the laser radar point cloud data collected by the inkjet printer, and evaluate the path feasibility of the optimal path candidate set.

[0083] In the description of the present invention, the laser radar point cloud data collected by the inkjet printer is used to generate a dynamic obstacle heat map, and the path feasibility of the optimal path candidate set is evaluated, including: S311. Superimpose the lidar point cloud data with the dynamic environment model, incrementally update the obstacle topology map, and generate a dynamic obstacle heat map.

[0084] Specifically, the real-time point cloud is overlaid with a historical dynamic environment model (such as a 3D occupancy grid) to update obstacle occupancy status based on Bayesian probabilities. For example, if a voxel is occupied by three consecutive frames of point clouds, its occupancy probability increases from 0.5 to 0.9. The voxel occupancy probabilities are mapped to a color gradient (e.g., red indicates high-density obstacle areas, blue indicates safe zones) to generate a dynamic heat map.

[0085] The potential obstacle distribution in the occluded area is predicted in real time through a sparse convolutional neural network (such as the Minkowski Engine).

[0086] S312. Path features are extracted from the optimal path candidate set, and collision prediction is performed with the newly added obstacle position. If the minimum distance between the candidate path and the obstacle is less than the safety radius, it is marked as a high-risk path, and a path feasibility label is output based on the prediction result.

[0087] Specifically, we extract features such as curvature, acceleration limit, and node spacing from the optimal path candidate set, and build a path constraint rule library based on the kinematic model of the inkjet printer (such as maximum steering angle and scanning rate). We set the dynamic safety radius R according to the physical size of the printhead (such as 5mm diameter) and the mechanical vibration amplitude. safe (e.g. 10mm). When calculating the distance field, the candidate path needs to be discretized and sampled (e.g. one sampling point every 2mm) and the Euclidean distance between each point and the obstacle in the heat map needs to be calculated. If the minimum distance d min <R safe , marked as high-risk paths.

[0088] Finally, combined with obstacle motion prediction (such as Kalman filtering or optical flow method), the temporal overlap probability of path nodes and obstacle trajectories is evaluated, and a conflict probability label is output (such as low risk: <5%, medium risk: 5%-20%, high risk: >20%).

[0089] S32. Perform a random perturbation test on each feasible candidate path, calculate the actual execution time of the path, and evaluate the path risk including collision probability, execution deviation, and ink dot quality attenuation.

[0090] Specifically, random perturbation testing of candidate paths requires combining dynamic environmental models with process parameters, generating diverse path variants through Monte Carlo simulation or Gaussian noise injection, and quantifying risk indicators in actual execution. Perturbation types include: 1. Geometric perturbation: Gaussian noise (standard deviation of 10%-20% of the safety radius) is applied to path node coordinates to simulate mechanical vibration or positioning error. 2. Parameter perturbation: Random fluctuations of ±5% are applied to marking parameters (such as scan rate and line spacing) to simulate temperature changes or ink viscosity fluctuations.

[0091] The candidate paths are discretized into equally spaced execution points (e.g., one point every 2 mm), and perturbations are injected to generate a set of variant paths. The printhead's motion trajectory is simulated on the digital twin platform, and the deviation between the actual execution time and the theoretical value is calculated (e.g., a delay of ±3ms due to vibration).

[0092] Based on the lidar point cloud, a three-dimensional occupancy grid is constructed, and the overlapping probability of the perturbation path and the voxels occupied by the obstacle is counted. The Euclidean distance between each sampling point on the path and the nearest obstacle is calculated. If the minimum distance d min <R safe , trigger the collision mark and count the total collision probability.

[0093] Measure the node position offset of the perturbed path relative to the original path (such as root mean square error RMSE), and set the threshold to the device positioning accuracy (such as ±0.1mm). Compare the actual execution time T actual With the theoretical time T ideal , define the deviation coefficient. Calibrate the scan rate through experiments v The relationship between the ink dot diameter D (e.g., D = k / v + C, where k is the material coefficient) and the ink dot uniformity deviation is calculated based on the disturbed parameters. For example, if the line spacing tolerance exceeds ±5μm, it is determined to be quality degradation.

[0094] S33. Obtain the high-risk paths generated by the assessment, construct an adaptive sampling domain, generate constraint-aware paths, and form the coding table identification path through dynamic pruning and fusion.

[0095] In the description of the present invention, obtaining the high-risk path generated by the evaluation, constructing the adaptive sampling domain, generating the constraint-aware path, and forming the coding table identification path through dynamic pruning and fusion include: S331. With the highest risk path as the center, an elliptical sampling domain is constructed, and an artificial potential field guidance factor is introduced to make the sampling points biased towards the direction of the high-quality path in the optimal path candidate set.

[0096] Specifically, a dynamic elliptical sampling domain is constructed with the highest-risk path as the central axis. The length of the major axis is positively correlated with the path risk value (for every 0.1 increase in the risk factor, the major axis radius increases by 15%), and the minor axis radius is determined by the device's kinematic constraints (such as the nozzle's swing amplitude). An artificial potential field guidance factor is introduced, superimposing an attractive field (pointing toward the low-risk direction of the optimal path candidate set) and a repulsive field (away from the red high-density areas in the obstacle heat map) when generating sampling points.

[0097] A regional adaptive sampling (RAS) mechanism is employed to increase the sampling density in high-curvature risk segments (curvature > 0.2 rad / mm) to three times that of conventional segments, while reducing it to 50% in smooth segments to conserve computational resources. A Monte Carlo perturbation test is used to generate candidate path variants, assessing collision probability and ink quality degradation. The top 20% low-risk paths are retained as high-quality directional guidance benchmarks.

[0098] S332. When the node is expanded, the feasibility of the marking parameters is synchronously checked. If the curvature of the new path segment is greater than the preset curvature threshold, the path node is rejected and the scanning rate is downgraded. A bidirectional path planning strategy is adopted to simultaneously expand the tree structure from the starting point and the end point of the risk segment.

[0099] Specifically, a scanning rate-curvature correlation model is established: when the curvature of a path segment exceeds a threshold (e.g., 0.25 rad / mm), the scanning rate is automatically reduced (rate = base rate / (1 + 5 * curvature)) to prevent excessive ink dot spread due to mechanical vibration. A constraint-aware decoder is introduced to verify the compatibility of coding parameters (power, line spacing) with material properties in real time. If a parameter violation is detected (e.g., line spacing <0.1mm when metal surface roughness >3μm), dynamic parameter compensation rules are triggered.

[0100] The bidirectional path planning strategy simultaneously expands the RRT* tree structure from the start and end points of the risky segment. The node expansion step size is adaptively adjusted based on the local obstacle density (reduced to 1mm when the density exceeds 30%). The integrated AMCL adaptive particle filter mechanism dynamically increases the number of sampled particles at path bifurcations (from the default 500 to 1200), improving the success rate of path exploration in complex areas. The cumulative risk value of the path is evaluated after every five nodes are expanded, and branches with risk increases exceeding 20% ​​are pruned.

[0101] S333: splice the local corrected path segment with the non-risk segment in the optimal path candidate set, and use a B-spline curve for smooth transition to obtain a corrected inkjet marking path.

[0102] Specifically, the hybrid decoder of the Diff-RNTraj model is used to topologically match the corrected high-risk segments with the original low-risk segments, and the timing of the path nodes is aligned using the dynamic time warping (DTW) algorithm. Diff-RNTraj is a diffusion model for generating road network-constrained trajectories (RNTraj). It aims to address two core issues with traditional trajectory generation methods: 1. Lack of geographic validity: Existing methods directly generate GPS points, causing trajectories to deviate from the actual road network and be unsuitable for downstream tasks (such as navigation and traffic analysis). 2. Information loss: The generated trajectories lack segment-level semantic information (such as the vehicle's segment and mobility rate), requiring additional road matching and introducing errors. This model generates mixed-data trajectories (including discrete segment IDs and continuous mobility rates) directly on the road network in an end-to-end manner. Using a diffusion model framework, it improves the spatial rationality and privacy security of the trajectories.

[0103] After the above topological matching, a transition buffer (length = 2 times the nozzle diameter) is introduced at the splicing point, and a cubic B-spline curve is used for smooth connection. The curvature change rate is limited to 0.05 rad / mm. 2 To avoid mechanical shock caused by sharp turns, a spatial validity loss function is applied, illegal paths are eliminated through physical rule verification (such as minimum turning radius ≥ 15mm), and the remaining paths are optimized for energy minimization.

[0104] Finally, the Pareto front solution set is constructed, and different types of optimal paths can be selected according to the real-time working conditions: 1. High-precision mode: quality weighting 70% (ink dot uniformity), economy weighting 20% ​​(path length), stability weighting 10% (acceleration); 2. High-efficiency mode: Economy weighting 60%, quality weighting 30%, and stability weighting 10%.

[0105] S34. Inversely infer the marking parameters based on the geometric characteristics of the inkjet marking path, output the action sequence of the inkjet marking device, and monitor the line distance deviation in real time. When the line distance deviation exceeds the preset threshold, parameter compensation is triggered.

[0106] Specifically, the image processing module identifies the geometric features of the coding path, such as line segments, curvature, and node spacing. For example, the QR code coverage is determined based on the parallel relationship between line segments and closed quadrilateral areas, and the path curvature and the kinematic constraints of the printhead (such as the maximum acceleration of 0.5 m / s) are combined to determine the coverage of the QR code. 2 ) Reverse scan rate v and ink jetting frequency f The association model.

[0107] Using a Bayesian optimization framework, the system takes path geometry (such as curvature radius R and node density) as input and outputs the optimal combination of marking parameters (scan rate, ink drop volume, and nozzle height). For example, for high-curvature path segments (R < 10mm), the scan rate is automatically reduced to 60% of the baseline value to prevent ink dot spread.

[0108] Convert the path node into a sequence of sprinkler motion instructions, including: 1. Spatial coordinate instruction: Generate XYZ axis movement trajectory based on grid solution space (resolution 0.1mm); 2. Timing synchronization command: Set the nozzle trigger timing (±0.1ms accuracy) according to the matching relationship between the scanning rate v and the ink drop frequency f; 3. Auxiliary action instructions: such as nozzle lifting (height compensation ±0.5mm), ink circuit pressure adjustment (range 0.2-0.8MPa).

[0109] Finally, the PID controller adjusts parameters in real time. For example, if changes in the material surface roughness are detected, the ink pressure is dynamically increased by 10% to maintain the ink penetration depth.

[0110] Regarding parameter compensation, a dual-frequency laser sensor is used to measure the actual line distance d in real time. actual , and the target value d target Comparison. The ink dot morphology is captured by a stroboscopically synchronized industrial camera (frame rate 1000fps), and the line distance uniformity index is calculated based on the edge detection algorithm. actual −d target When the error is greater than 5%, the scan rate is fine-tuned (in steps of ±0.1 m / s). A closed-loop feedback system (delay <10 ms) evaluates the compensated line distance in real time. If the error still exceeds the limit, the system switches to a redundant printhead or triggers a shutdown protection.

[0111] S4. Monitor the operation feedback of the inkjet coding equipment and iteratively update the marking parameter library and algorithm rules.

[0112] Specifically, incremental knowledge base management can be used to establish a parameter effectiveness evaluation system, including: 1. Stability index: standard deviation of line distance after 1000 consecutive prints; 2. Economic index: ink consumption per unit area; 3. Compatibility index: supported material types / surface roughness range.

[0113] The top 10% high-quality parameter combinations are solidified and stored every month to eliminate outdated rules.

[0114] In summary, leveraging the technical solutions of the present invention, this method integrates the core advantages of multiple intelligent optimization algorithms to construct a hybrid collaborative optimization architecture, effectively overcoming the limitations of single algorithms in complex scenarios. Traditional path planning algorithms are susceptible to local optimal solutions and struggle to simultaneously meet multi-dimensional constraints (such as kinematic constraints and environmental dynamics). By dynamically coupling the global exploration of Invasive Weed Optimization (IWO) with the local development strategy of Gray Wolf Optimization (GWO), joint optimization of path nodes and marking parameters is achieved. This collaborative mechanism not only enhances the algorithm's coverage of the high-dimensional solution space but also mitigates the risk of premature convergence through dynamic strategy switching, significantly improving the overall robustness of path planning. Furthermore, the algorithm supports flexible adjustment of multi-objective weights, adapting to the priorities of different scenarios. For example, focusing on coding quality in precision manufacturing scenarios and prioritizing efficiency in logistics scenarios fully demonstrates the method's versatility and scalability. To address dynamic disturbances common in coding operations (such as mechanical vibration, ink property changes, and unexpected obstacles), this method designs a real-time correction mechanism based on incremental perception. By integrating real-time sensor data with a digital twin verification platform, the system dynamically updates the environmental model and rapidly screens for highly interference-resistant path candidate sets. For example, when detecting changes in the line spacing tolerance threshold due to temperature fluctuations in ink viscosity, the algorithm automatically triggers marking parameter compensation rules, synchronously adjusting the scanning rate and inflection point delay to prevent degradation in inkjet quality. Furthermore, the improved RRT* algorithm introduces an artificial potential field guidance factor in local path replanning, combined with kinematic constraint verification, to ensure that the corrected path meets both obstacle avoidance requirements and printhead acceleration limits, thus maintaining stable execution even under sudden interference. This invention embeds inkjet process constraints (such as line spacing uniformity and ink dot coverage) and device physical constraints (such as servo motor response delay and printhead motion curvature) into the optimization model, constructing a multi-dimensional constraint joint processing framework. Using a dynamic penalty function and a feasible solution prioritization mechanism, the algorithm iteratively eliminates inferior solutions that violate constraints. A Pareto front screening strategy is employed to retain non-dominated solutions, ensuring that the final set of path candidate sets is both engineering feasible and multi-objective optimal. For example, when the path curvature exceeds the maximum steering capability of the printhead, the system automatically triggers a scan rate degradation strategy to simultaneously optimize path smoothness and execution stability. This constraint integration capability enables the method to effectively balance conflicting indicators such as path length, energy consumption, and quality, meeting the stringent requirements of refined management in industrial scenarios.

[0115] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

Claims

1. A method for optimal path planning for inkjet marking based on an intelligent optimization algorithm, characterized in that: include: S1. Discretize the coding area into three-dimensional space and construct a dynamic environment model constrained by marking parameters; S2. Based on the priority queue of coding tasks, a hybrid optimization algorithm is used to coordinate the optimization of marking parameters and path nodes to generate the optimal path candidate set; S3: Build an obstacle avoidance model based on the optimal path candidate set, generate a collision-free path, adjust the printhead posture based on visual feedback, and output an anti-interference coding action sequence; S4. Monitor the operation feedback of the coding equipment and iteratively update the marking parameter library and algorithm rules; Wherein, the S2 includes: S21. Generate a mixed coding population of weeds and gray wolves through dual population initialization and parameter coupling; S22, based on the iteration of individuals within the hybrid coding population, cross-population information interaction is achieved to generate a hybrid search strategy that integrates global exploration and local exploitation; S23. Generate a path node sequence based on the individuals in the population, perform optimal solution according to the hybrid search strategy through multi-objective fitness modeling and constraint integration, and generate fitness scores and constraint tags; S24. Output a dynamically optimized optimal path candidate set based on the iterative results of the hybrid optimization algorithm.

2. The optimal path planning method for inkjet marking based on intelligent optimization algorithm according to claim 1 is characterized in that: The method of discretizing the coding area into three-dimensional space and constructing a dynamic environment model with marking parameter constraints includes: S11, obtaining the surface parameters of the material to be coded, the kinematic parameters of the coding equipment and sensor data, and generating an original data set of marking parameters; S12. Based on the original data set of marking parameters, the scanning rate interval, the inflection point jump delay and the line spacing control coefficient are calculated to construct the marking parameter constraint relationship matrix; S13, obtaining the real-time sensor data stream collected by the inkjet printer and establishing a dynamic constraint rule library; S14. Divide the coding area into several three-dimensional grids, mark the coding attributes of each grid, and introduce the nozzle response delay to expand the time dimension to generate a four-dimensional space-time dynamic environment model.

3. The optimal path planning method for inkjet marking based on intelligent optimization algorithm according to claim 2 is characterized in that: The acquisition of real-time sensor data stream collected by the inkjet printer and establishment of a dynamic constraint rule base include: S131, using a Kalman filter to fuse visual positioning data, vibration signals, and temperature data in the real-time sensor data stream to eliminate temporal and spatial deviations between sensors; S132. Based on the marking parameter constraint relationship matrix, a marking parameter association model is constructed using a Bayesian network to convert the physical constraints into path rule constraints, wherein the path rule constraints include surface roughness constraints, vibration amplitude constraints, and idle stroke scanning rate constraints; S133. Classify the complexity level according to the type of coding content, and dynamically adjust the path rule constraint weight according to the complexity level of the current coding content; S134. Integrate and record various path rule constraints and corresponding weights to build a dynamic rule library.

4. The optimal path planning method for inkjet marking based on an intelligent optimization algorithm according to claim 1 is characterized in that: The generation of a weed-gray wolf mixed coding population by dual population initialization and parameter coupling includes: S211. Initialize the weed population size and the wolf population size. The two populations share the same solution space. S212, encode each seed individual in the weed population as X IWO =( x 1,…, x n ;σ), where x i represents the path node coordinates, and i =1, 2, ..., n ;σ represents the diffusion radius; S213, encode the gray wolf individuals in the gray wolf population as X GWO =( x 1,…, x n , t 1,…, t m ),in, t j represents the marking parameters, and j =1, 2, ..., m ; S214, associating the linear decline parameter of the gray wolf population with the diffusion radius of the weed population, calculating a nonlinear attenuation factor, and when the nonlinear attenuation factor decays to a preset attenuation threshold, switching from global exploration of the weed population to local exploration of the gray wolf population; S215. When the population diversity index is less than a preset index threshold, the Gaussian perturbation mechanism of the weed population is activated to generate mutant individuals to fill the blind area of ​​the solution space.

5. The optimal path planning method for inkjet marking based on intelligent optimization algorithm according to claim 4 is characterized in that: The hybrid search strategy based on the internal iteration of individuals in the hybrid coding population, realizing cross-population information interaction, and generating the fusion of global exploration and local development includes: S221. Set the iteration cycle. In the global exploration phase, generate the number of seeds based on the fitness value. Generate new seed coordinates through normal distribution. Keep the top 10% of the weed population in each generation and directly inject the α, β, and δ wolf candidate values ​​of the gray wolf population to achieve path-parameter coordinated optimization. S222. When the nonlinear attenuation factor reaches a preset attenuation threshold, the global exploration is switched to the local development phase. The positions of individual gray wolves are updated according to their levels. In combination with lens imaging, the mirror solution is learned in reverse, and the high-quality solutions are retained through Pareto front screening. S223. Monitor the population convergence speed during the global exploration phase in real time. When the convergence speed is less than the set convergence threshold, trigger the phase switch in advance and execute local development of the gray wolf population.

6. The method for optimal path planning for inkjet marking based on an intelligent optimization algorithm according to claim 5, characterized in that: The generation of path node sequences based on population individuals, multi-objective fitness modeling and constraint integration, and optimal solution according to a hybrid search strategy to generate fitness scores and constraint tags include: S231. Generate a path node sequence based on the iterative solution of the hybrid optimization algorithm, obtain marking parameter values, establish a composite objective function integrating economic objectives, quality objectives, and stability objectives, and assign optimization weights to each objective; S232. Mark solutions that violate path rule constraints, impose exponential penalties on them, and gradually tighten constraint tolerances to eliminate infeasible solutions. S233. Normalize the solution calculated by the composite objective function, calculate the comprehensive fitness score according to the optimization weight of each objective, eliminate all solutions marked as violating the constraints, and sort the remaining feasible solutions in descending order according to the comprehensive fitness score to generate a priority queue; if the priority queue is empty, select the solution with the least constraint violations as the temporary candidate.

7. The method for optimal path planning for inkjet marking based on an intelligent optimization algorithm according to claim 6, characterized in that: The formula of the composite objective function is: ； ； ； Where, F econ Indicates the economic objective score; ( x i , y i ) represents the coordinates of the path node; n Indicates the number of path nodes; t jump，j Indicates the j Theoretical time required for the secondary empty stroke movement; t delay，j Indicates the j Real-time delay compensation of the secondary inflection point; λ represents the penalty coefficient; m Indicates the number of marking parameters; F qual Indicates the quality target score; σ d Indicates the standard deviation of line distance; Δ d Indicates the line distance tolerance threshold; F stab represents the stability target score; v i Indicates the i The nozzle speed at each path node; Δ t Indicates the time interval between adjacent path nodes; a max Indicates the maximum acceleration of the nozzle.

8. The method for optimal path planning for inkjet marking based on an intelligent optimization algorithm according to claim 1, characterized in that: The obstacle avoidance model is constructed based on the optimal path candidate set to generate a collision-free path, and the print head posture is adjusted in combination with visual feedback to output an anti-interference coding action sequence, including: S31. Generate a dynamic obstacle heat map using the laser radar point cloud data collected by the inkjet printer and evaluate the path feasibility of the optimal path candidate set; S32. Perform a random perturbation test on each feasible candidate path, calculate the actual execution time of the path, and evaluate the path risk including collision probability, execution deviation, and ink dot quality degradation; S33. Obtain the high-risk paths generated by the assessment, construct an adaptive sampling domain, generate constraint-aware paths, and form a coding table identification path through dynamic pruning and fusion; S34. Inversely infer the marking parameters based on the geometric characteristics of the inkjet marking path, output the action sequence of the inkjet marking device, and monitor the line distance deviation in real time. When the line distance deviation exceeds the preset threshold, parameter compensation is triggered.

9. The method for optimal path planning for inkjet marking based on an intelligent optimization algorithm according to claim 8, characterized in that: The method of generating a dynamic obstacle heat map using the laser radar point cloud data collected by the inkjet printer and evaluating the path feasibility of the optimal path candidate set includes: S311. Overlaying the laser radar point cloud data with the dynamic environment model, incrementally updating the obstacle topology map, and generating a dynamic obstacle heat map; S312. Path features are extracted from the optimal path candidate set, and collision prediction is performed with the newly added obstacle position. If the minimum distance between the candidate path and the obstacle is less than the safety radius, it is marked as a high-risk path, and a path feasibility label is output based on the prediction result.

10. The method for optimal path planning for inkjet marking based on intelligent optimization algorithm according to claim 8, characterized in that: The steps of obtaining the high-risk paths generated by the evaluation, constructing the adaptive sampling domain, generating the constraint-aware paths, and forming the coding table identification paths through dynamic pruning and fusion include: S331. With the highest risk path as the center, an elliptical sampling domain is constructed, and an artificial potential field guidance factor is introduced to make the sampling points biased towards the direction of the high-quality path in the optimal path candidate set; S332. When the node is expanded, the feasibility of the marking parameters is simultaneously verified. If the curvature of the new path segment is greater than the preset curvature threshold, the path node is rejected and the scanning rate is degraded. A bidirectional path planning strategy is adopted to simultaneously expand the tree structure from the starting point and the end point of the risk segment. S333: splice the local corrected path segment with the non-risk segment in the optimal path candidate set, and use a B-spline curve for smooth transition to obtain a corrected inkjet marking path.

Images