Smart manufacturing full-link body intelligent collaborative work scene adaptation method and system

Abstract

The application relates to the technical field of machine control, and discloses an intelligent manufacturing full-link body intelligent collaborative work scene adaptation method and system. The method comprises the following steps: performing feature modeling on full-link processing and assembling, material transfer and intelligent storage scenes to obtain scene feature vectors and inputting the scene feature vectors into a scene classification system; iteratively processing the scene feature vectors and robot action parameters by using an interval iterative optimization algorithm to obtain a feasible adaptation parameter interval; detecting an intersection of adjacent intervals to obtain a gap amount; when the gap amount is greater than zero, extracting a feature increment, calculating a buffer capacity demand parameter, combining the buffer capacity demand parameter with the gap amount, encoding and writing the buffer capacity demand parameter and the gap amount into the scene classification system, iteratively processing action parameters at a handover coordinate, obtaining a weighted sum of a buffer overflow rate and an idle rate, and outputting a trigger threshold and a release beat interval when the weighted sum converges to below a preset threshold, so that collaborative control parameters are obtained. The application improves the processing efficiency of parameter migration, reuse and collaborative handover of a robot system between full-link multiple scenes.

Description

Technical Field

[0001] This application relates to the field of machine control technology, and in particular to a method and system for adapting to intelligent collaborative operation scenarios across the entire intelligent manufacturing chain. Background Technology

[0002] In existing technologies, the collaborative operation of embodied intelligent robots in the entire intelligent manufacturing process generally follows a layered implementation path of perception-decision-execution. At the perception layer, industrial robots combine multi-source sensor information such as vision, force, and touch to collect multimodal data of the environment and work objects. At the decision-making and scheduling layer, the system uses multi-agent algorithms to determine task allocation schemes based on perception information. The factory-level scheduling center is responsible for the global production scheduling and resource allocation of multiple processes, workstations, and tasks. The path planning of mobile units is usually modeled as a traveling salesman problem or vehicle path problem for solution. In terms of scene transfer capability, existing technologies mostly rely on large-scale simulation training followed by transfer to real-world scenarios. After accumulating physical interaction experience in virtual space, the robot is applied to the actual production line, thereby indirectly addressing the operational needs under different working conditions.

[0003] However, most existing industrial robots focus on a single function or process in manufacturing, lacking a unified mechanism for abstracting scene features and adaptive mapping. Changing to a different scene often requires redevelopment or extensive manual parameter tuning, resulting in weak scene generalization ability. Solutions relying on large-scale simulation transfer have extremely high requirements for training data volume and training cost, and the success rate of the model in various tasks is limited, making it difficult to meet the stability and high accuracy requirements of industrial scenarios. In addition, the standardization and specialization of industrial scenarios make it difficult to complete all task division through a single general-purpose hardware and software system. The collaborative scheduling efficiency of multiple entities is insufficient when facing dynamic constraints such as order changes and production line restructuring. Especially at the junction of different scenarios, there is a lack of connection mechanism between the parameter systems of each scenario, which can easily lead to material accumulation or idle workstations at the junction. Furthermore, existing scheduling methods do not have specific detection and response methods for such junction misalignment.

[0004] Building upon this foundation, to enable robot systems to adapt to heterogeneous scenarios such as processing and assembly, material handling, and intelligent warehousing in a transferable and reusable manner, the first challenge is converting the operational data of each scenario into comparable and searchable scenario feature representations. Otherwise, subsequent parameter adjustments become impossible. Given a unified representation of scenario features, the next issue is how to deduce the feasible range of robot motion parameters within that scenario, providing clear boundaries for motion parameter adjustments. Since end-to-end production is not completed in isolation from a single scenario, physical connections inevitably exist between multiple scenarios. Therefore, after determining the feasible range of motion parameters between adjacent scenarios, the next challenge is determining whether gaps exist between the feasible ranges of adjacent scenarios, and in what numerical form these gaps should be expressed. When these gaps are identified… After the gap is cleared, it is necessary to further address how much buffer occupancy demand this gap will translate into under actual production rhythm. If this demand cannot be explicitly calculated, material accumulation or idle workstations at the handover location cannot be predicted in advance. Once the buffer occupancy demand is determined, its value still needs to be incorporated into a searchable and reusable scenario representation system to have actual scheduling value. The resulting problem is how to uniformly encode and store this type of composite scenario information derived from the handover relationship with the original single scenario representation system. Finally, the static buffer demand value itself cannot directly drive real-time collaborative actions during the handover process. Therefore, it is also necessary to solve how to iteratively determine the specific threshold and corresponding rhythm value range that can trigger a collaborative response based on the dynamic deviation between the actual occupancy status of the buffer and the demand value, so that the collaborative control of the handover location has executable criteria. Summary of the Invention

[0005] This application provides a method and system for adapting intelligent collaborative operation scenarios across the entire intelligent manufacturing chain. The technical problems to be solved by this application are the following in the prior art: inconsistent representation of scene features between heterogeneous scenes, lack of explicit feasible range of motion parameters, inability to identify and quantify the connection relationship at scene junctions, failure to include buffering requirements caused by junctions into the searchable scene system, and lack of executable criteria for junction collaborative control. This application improves the processing efficiency of robot systems in parameter migration, reuse, and collaborative junctions across multiple scenarios across the entire chain.

[0006] Firstly, this application provides a method for adapting to intelligent collaborative operation scenarios across the entire intelligent manufacturing chain, the method comprising:

[0007] Step S1: Perform feature modeling on the processing and assembly scenario, material transfer scenario, and intelligent warehousing scenario in the entire chain to obtain scenario feature vectors, and input the scenario feature vectors into the scenario classification system;

[0008] Step S2: Iteratively process the scene feature vector and the robot's motion parameters using an interval iterative optimization algorithm to obtain a feasible adaptation parameter range;

[0009] Step S3: Perform intersection detection processing on two adjacent feasible adaptation parameter intervals to obtain the gap amount;

[0010] Step S4: When the gap amount is greater than zero, feature increment extraction is performed on the scene feature vectors on both sides of the gap amount to obtain feature increments. The buffer capacity requirement is calculated based on the feature increments to obtain buffer capacity requirement parameters. The buffer capacity requirement parameters and the gap amount are combined to obtain a composite handover scene feature vector. The composite handover scene feature vector is encoded according to the scene type and written into the scene classification system. The action parameters at the handover coordinates are iteratively processed by the interval iterative optimization algorithm to obtain a weighted sum of buffer overflow rate and idle rate. When the weighted sum converges to below a preset threshold, the buffer trigger threshold and release beat interval are output to obtain the collaborative control parameters.

[0011] Secondly, this application provides an intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation system, the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation system comprising:

[0012] The modeling module is used to perform feature modeling on the processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios in the entire chain, to obtain scenario feature vectors, and to input the scenario feature vectors into the scenario classification system.

[0013] The iteration module is used to iteratively process the scene feature vector and the robot's motion parameters through an interval iterative optimization algorithm to obtain a feasible adaptation parameter interval.

[0014] The detection module is used to perform intersection detection processing on two adjacent feasible adaptation parameter intervals to obtain the gap amount;

[0015] The control module is used to: extract feature increments from the scene feature vectors on both sides of the gap when the gap is greater than zero, obtain feature increments, calculate buffer capacity requirements based on the feature increments, obtain buffer capacity requirement parameters, combine the buffer capacity requirement parameters and the gap to obtain a composite handover scene feature vector, encode the composite handover scene feature vector according to the scene type and write it into the scene classification system, iterate the action parameters at the handover coordinates using the interval iterative optimization algorithm to obtain a weighted sum of buffer overflow rate and idle rate, and output the buffer trigger threshold and release beat interval when the weighted sum converges to below a preset threshold to obtain collaborative control parameters.

[0016] Thirdly, a smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation device to execute the aforementioned smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method.

[0017] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, which, when executed on a computer, cause the computer to perform the above-mentioned intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method.

[0018] The technical solution provided in this application extracts and vectorizes the working condition data of various scenarios such as processing and assembly, material transfer, and intelligent warehousing to form a unified scene feature vector. This vector is then entered into the scene classification system according to a two-level coding structure, giving heterogeneous scenarios that were originally independent and relied on human experience to a unified and searchable representation. Based on this, the robot's motion parameters are constrained by boundary mapping and target values ​​are inversely calculated based on the scene feature vector. Then, through multiple rounds of optimization iteration, the internal matching error of the scene is obtained and the feasible adaptation parameter range is statistically determined. This makes the adjustment of motion parameters no longer dependent on the empirical value of single-point convergence, but forms a range expression supported by the results of multiple rounds of iteration with traceable and definite basis. This processing method establishes an explicit mapping relationship between scene features and motion parameters, reducing the need to redevelop parameter tuning logic for each new scenario. Furthermore, by extracting boundary values ​​and performing intersection operations on two adjacent feasible adaptation parameter intervals, it is possible to directly determine at the numerical level whether there are gaps in the connection of adjacent scenes in a specific action parameter dimension, and quantify the gap amount in the form of difference when there is no overlap. This processing chain transforms the originally vague scene handover misalignment problem, which can only be detected by manual inspection, into a numerical result that can be automatically identified and calculated by the system, providing a clear trigger basis for subsequent collaborative control.

[0019] The multi-round optimization iteration and interval statistics method adopted in this scheme does not characterize robot motion parameters with a single convergence value, but rather characterizes the feasible boundary of parameters with the distribution range of multiple rounds of candidate results. This algorithmic feature enables the scheme to accommodate inherent parameter fluctuations within industrial production lines and avoid misjudging them as anomalies, demonstrating the algorithm's adaptability to the specific application domain's operating conditions. Simultaneously, feature increment extraction, buffer capacity requirement parameter calculation, combined encoding and writing of feature vectors for composite handover scenarios, and the iterative convergence of the interval iterative optimization algorithm for the weighted sum of buffer overflow rate and idle rate together constitute a complete processing chain from scene handover relationship identification to collaborative control parameter output. This allows composite handover scenarios to be incorporated into the original scene classification system as independent instances without altering the recorded content and index position of existing scene instances. This gives the algorithm model the ability to continuously expand its processing capacity as the number of scene types increases when applied to the specific scenario of full-link multi-scene collaboration. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of an embodiment of the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method in this application.

[0022] Figure 2 This is a schematic diagram illustrating the intersection operation and gap determination of two adjacent feasible adaptation parameter intervals in an embodiment of this application.

[0023] Figure 3 This is a schematic diagram illustrating the convergence of the weighted sum of buffer overflow rate and idle rate with each iteration in an embodiment of this application. Detailed Implementation

[0024] This application provides a method and system for adapting to intelligent collaborative operation scenarios across the entire intelligent manufacturing chain. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0025] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method in this application includes:

[0026] Step S1: Perform feature modeling on the processing and assembly scenario, material transfer scenario, and intelligent warehousing scenario in the entire chain to obtain scenario feature vectors, and input the scenario feature vectors into the scenario classification system;

[0027] Specifically, for the processing and assembly scenario, material transfer scenario, and intelligent warehousing scenario in the entire chain, working condition data of each scenario is collected within the historical operation cycle. The working condition data includes raw values ​​directly read from the production line operation records, such as workstation switching interval, conveyor cycle time, storage and retrieval cycle time, load weight, storage location density, and aisle width. For each type of scenario, three dimensions are selected as the scenario feature vector. For the processing and assembly scenario, the three dimensions are workpiece positioning accuracy, assembly torque, and workstation switching interval. For the material transfer scenario, the three dimensions are conveyor cycle time, load weight, and conveyor belt width. For the intelligent warehousing scenario, the three dimensions are storage location density, storage and retrieval cycle time, and aisle width. The selection of the three dimensions is based on the fact that these three values ​​directly determine the upper and lower limits of the robot's executable speed and force control range in the corresponding scenario. They are working condition quantities that have a direct constraint effect on the robot's motion parameters. The scenario classification system adopts a two-level coding structure. The first-level coding distinguishes three major scenario categories: processing and assembly, material transfer, and intelligent warehousing. The second-level coding identifies the instance number of the scenario category in the specific production line. The scenario feature vector of each scenario category, together with its two-level coding, is written as a record into the scenario classification system, so that each record in the scenario classification system corresponds to a searchable scenario instance.

[0028] Step S2: Iteratively process the scene feature vector and the robot's motion parameters using the interval iterative optimization algorithm to obtain a feasible adaptation parameter interval;

[0029] Specifically, the robot's motion parameters consist of three items: the end effector speed threshold, the force control threshold, and the path avoidance distance. These three parameters are adjustable execution layer values ​​of the robot under the constraints of the scene feature vector. Based on the transmission or retrieval cycle values ​​in the scene feature vector and the fixed travel distance between the workstations, the upper limit of the end effector speed threshold is calculated. This upper limit, along with the historical target values ​​of the force control threshold and the path avoidance distance, are used as the starting constraints for optimization. The motion parameters are iteratively adjusted in multiple rounds. In each round, a new set of candidate values ​​is generated near the previous round's value, and the deviation between this set of candidate values ​​and the target value is calculated. The deviation is measured as the sum of the normalized squared deviations of the three parameters. The number of iterations per round is set to 40 rounds. All candidate values ​​with deviations lower than the pre-set scene internal error threshold are retained in the 40 iterations. Instead of selecting the best value at a single point, the minimum and maximum values ​​of each parameter among these retained candidate values ​​are used as the value boundaries of that parameter. This yields the feasible adaptation parameter range, which is therefore a range obtained by statistical analysis of the results of multiple iterations, rather than a single convergent value.

[0030] Step S3: Perform intersection detection on two adjacent feasible adaptation parameter intervals to obtain the gap amount;

[0031] Specifically, the adjacency relationship is determined by whether there are direct workpiece or material handover stations between two scenarios in the physical layout of the production line. For example, if the unloading station in the material transfer scenario and the picking station in the smart warehousing scenario are directly connected in space, they are considered adjacent, and a fixed handover coordinate is marked at the connection position. For the speed threshold or cycle time component in two adjacent feasible adaptation parameter intervals, their respective lower and upper limits are extracted. The larger of the two lower limits is compared, and the smaller of the two upper limits is compared. If the larger lower limit exceeds the smaller upper limit, it means that the two intervals do not overlap on the number axis. In this case, the difference between the lower limit of the larger interval and the upper limit of the smaller interval is the gap amount. If the larger lower limit does not exceed the smaller upper limit, it means that the two intervals already overlap, no gap amount is generated, and no subsequent processing is triggered.

[0032] Step S4: When the gap is greater than zero, the feature increment of the scene feature vectors on both sides of the gap is extracted to obtain the feature increment. The buffer capacity requirement is calculated based on the feature increment to obtain the buffer capacity requirement parameter. The buffer capacity requirement parameter and the gap are combined to obtain the composite handover scene feature vector. The composite handover scene feature vector is encoded according to the scene type and written into the scene classification system. The action parameters at the handover coordinates are iteratively processed by the interval iterative optimization algorithm to obtain the weighted sum of the buffer overflow rate and idle rate. When the weighted sum converges to below the preset threshold, the buffer trigger threshold and release beat interval are output to obtain the collaborative control parameters.

[0033] Specifically, the feature increment is based on the current scene feature vectors of the two sides of the intersection coordinates. It takes the numerical difference of the corresponding dimension in the feature vectors of the two scenes. For example, the difference between the current conveying cycle of the material transfer scene and the current storage and retrieval cycle of the smart warehousing scene. This difference is the feature increment, which is used to reflect the real-time misalignment of the working conditions on both sides of the intersection. The buffer capacity requirement is determined as follows: the statistical window duration is set to 60 seconds. This duration is taken from the average cycle time of a single production batch of the production line. 60 is divided by the arithmetic mean cycle time of the feasible adaptation parameter range of the material transfer side. After rounding, the theoretical output quantity of the transfer side within the window is obtained. 60 is divided by the arithmetic mean cycle time of the feasible adaptation parameter range of the smart warehousing side. After rounding, the theoretical consumption quantity of the warehousing side within the window is obtained. The theoretical output quantity minus the theoretical consumption quantity is the quantity difference. Adding a preset safety factor of 1 unit, the resulting value is the buffer capacity requirement parameter. The safety factor of 1 unit is taken to cover the truncation error generated during a single rounding process. The combination processing involves arranging the buffer capacity requirement parameters and gap quantities into a two-dimensional numerical group in a fixed order, which serves as the initial feature vector. The combination rule for the scenario type coding is to concatenate the two-level codes of the material transfer scenario and the intelligent warehousing scenario in the order of transfer side first and warehousing side last, and append the sequence number of the handover coordinates at the end to form a composite scenario type code. After binding the initial feature vector with the composite scenario type code, it is written into the scenario classification system to obtain a composite handover scenario instance. The writing method is consistent with the scene instance entry method in S1, and both are appended as new records according to the two-level coding structure. The motion parameters at the handover coordinates undergo the same multi-round iterative optimization process. The difference is that the iteration target is no longer the matching error within a single scene, but rather the ratio of excess to deficiency between the number of buffers occupied and the required buffer capacity. The ratio of the portion of the buffers occupied exceeding the required buffer capacity to the required buffer capacity is denoted as the buffer overflow rate, and the ratio of the portion of the required buffer capacity to the required buffer capacity is denoted as the idle rate. The buffer overflow rate and idle rate are added together with equal weights of 0.5 each to obtain a weighted sum. The reason for taking equal weights is that the impact of overfilled or empty buffers on the handover process is not differentiated in this scenario. The preset threshold is 0.05, based on the fact that the deviation of the number of buffers corresponding to this value does not exceed 5% of its capacity, which is within a negligible fluctuation range in actual production line handover scenarios. When the weighted sum drops below 0.05 in a certain iteration, the number of buffers occupied in that iteration is directly marked as the buffer trigger threshold. The value boundary of the beat component in the action parameter at the intersection coordinates of that iteration is statistically defined as the release beat interval. The combination of the buffer trigger threshold and the release beat interval is the cooperative control parameter.

[0034] In one specific embodiment, step S1 includes:

[0035] Feature extraction processing is performed on the working condition data of processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios to obtain feature parameters of processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios.

[0036] Based on the feature parameters of the processing and assembly scenario, the feature parameters of the material transfer scenario, and the feature parameters of the intelligent warehousing scenario, vectorization processing is performed to obtain the feature vectors of the processing and assembly scenario, the feature vectors of the material transfer scenario, and the feature vectors of the intelligent warehousing scenario.

[0037] Based on the feature vectors of the processing and assembly scenario, the feature vector of the material transfer scenario, and the feature vector of the intelligent warehousing scenario, the error threshold is set for the action parameters to obtain the error threshold.

[0038] The feature vectors of processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios, along with error thresholds, are entered into the scenario classification system according to scenario type codes for data entry and processing. This process yields scenario instances corresponding to the scenario feature vectors, and an extended write interface is configured in the scenario classification system.

[0039] Specifically, feature extraction processing was performed on the operating data of the machining and assembly scenario, extracting the workpiece positioning accuracy of 0.02 mm, assembly torque of 3.5 N·m, and workstation switching interval of 8 seconds. These three values ​​are collectively referred to as the feature parameters of the machining and assembly scenario. Feature extraction processing was also performed on the operating data of the material handling scenario, extracting the conveyor cycle time of 3.2 seconds per piece, load weight of 15 kg, and conveyor belt width of 600 mm. These three values ​​are collectively referred to as the feature parameters of the material handling scenario. Feature extraction processing was also performed on the operating data of the intelligent warehousing scenario, extracting the storage density of 0.85 pieces per square meter, storage and retrieval cycle time of 3.8 seconds per piece, and aisle width of 1200 mm. These three values ​​are collectively referred to as the feature parameters of the intelligent warehousing scenario. All three sets of values ​​were directly read from the operation records of the corresponding production line segments in each scenario. The reason for selecting these three values ​​instead of other operating data is that these three values ​​directly constrain the adjustable range of the robot end effector speed threshold and force control threshold in the corresponding scenarios.

[0040] Vectorization of feature parameters for processing and assembly scenarios, material handling scenarios, and intelligent warehousing scenarios involves arranging the three values ​​in each set of feature parameters into a three-dimensional array in a fixed order. For processing and assembly scenarios, the feature parameters are arranged in the order of workpiece positioning accuracy, assembly torque, and workstation switching interval to obtain the processing and assembly scenario feature vector. For material handling scenarios, the feature parameters are arranged in the order of conveyor cycle time, load weight, and conveyor belt width to obtain the material handling scenario feature vector. For intelligent warehousing scenarios, the feature parameters are arranged in the order of storage location density, storage and retrieval cycle time, and aisle width to obtain the intelligent warehousing scenario feature vector. The arrangement order remains fixed within the same scenario type, and instances of different scenarios are aligned in the same order to ensure comparability of the same dimension values ​​across different instances.

[0041] Error thresholds are set for motion parameters based on feature vectors from processing and assembly scenarios, material handling scenarios, and intelligent warehousing scenarios. Specifically, an error threshold value is assigned to each of the three scenarios: For processing and assembly, the threshold is set to 0.5 N·m, corresponding to the lower limit of the fluctuation range of assembly torque in historical operating data; fluctuations exceeding this value are considered torque mismatch. For material handling scenarios, the threshold is set to 0.15 seconds per piece, corresponding to the common fluctuation range of conveyor cycle time in historical operating data. For intelligent warehousing scenarios, the threshold is set to 0.20 seconds per piece, corresponding to the common fluctuation range of storage and retrieval cycle time in historical operating data. Since the storage and retrieval cycle time fluctuation range in warehousing scenarios is higher than that in material handling scenarios, the error threshold value is set larger. All three error threshold values ​​are set separately after the scenario feature vectors are determined and are not involved in the vectorization process itself.

[0042] The feature vectors of processing and assembly scenarios, material handling scenarios, and intelligent warehousing scenarios, along with error thresholds, are entered into the scenario classification system for data entry according to scenario type codes. The scenario type codes employ a two-level structure. The first level uses letters to distinguish the major scenario categories: processing and assembly scenarios are denoted as A, material handling scenarios as B, and intelligent warehousing scenarios as C. The second level uses two digits to represent the instance number of that major scenario category in the specific production line, starting from 01 and numbered sequentially. For example, the first instance of the material handling scenario is coded as B01. The data entry process involves combining the scenario feature vector of each scenario instance with its corresponding error threshold to form a record. This record is then written into the storage structure of the scenario classification system using the two-level code as an identifier, resulting in the scenario instance corresponding to the scenario feature vector. The extended write interface configured in the scenario classification system allows the storage structure to append a new record with the new code, without altering the content and index position of existing records. This ensures that subsequent additions of scenario categories can be appended in the same way without affecting the reading of existing scenario instances.

[0043] In one specific embodiment, step S2 includes:

[0044] The action parameters are subjected to constraint boundary mapping based on the scene feature vector to obtain the parameter constraint boundary, and the target parameter values ​​are obtained by back-calculating the action parameters based on the scene feature vector.

[0045] Based on the parameter constraint boundary, the action parameters are optimized and iterated in multiple rounds to obtain the candidate parameter combination corresponding to each round;

[0046] Based on the deviation between the candidate parameter combination and the target parameter value, the matching error of the candidate parameter combination is calculated to obtain the scene internal matching error for each round.

[0047] Candidate parameter combinations with matching errors below the error threshold within the scene are subjected to interval statistical processing to obtain feasible adaptation parameter intervals.

[0048] Specifically, the parameter constraint boundary includes three items: the upper limit of the end effector speed threshold, the upper limit of the force control threshold, and the upper limit of the path avoidance distance, which correspond to the three motion parameters of the robot, respectively. The constraint boundary mapping process is to divide the cycle time dimension value in the scene feature vector by the fixed travel distance between the workstation and the adjacent device to obtain the upper limit of the speed threshold. For the material transfer scenario, the conveying cycle time is 3.2 seconds per piece, divided by the travel distance of 0.8 meters to get 0.25 meters per second. For the intelligent warehousing scenario, the storage and retrieval cycle time is 3.8 seconds per piece, divided by the same travel distance to get 0.21 meters per second. The upper limit of the force control threshold is directly taken as the assembly torque value of 3.5 N·m. The upper limit of the path avoidance distance is taken as the measured value of the production line safety clearance of 0.3 meters.

[0049] The target numerical inverse calculation process derives the target parameter values ​​from the same scene feature vector. These values ​​also include three items, each a fixed percentage of its corresponding upper limit: the speed threshold target value is 90% of the upper limit (0.225 m / s for material handling, 0.189 m / s for intelligent warehousing); the force control threshold target value is 85% of 3.5 N·m, i.e., 2.975 N·m; and the path avoidance distance target value is 90% of 0.3 m, i.e., 0.27 m. All three percentages are taken from the fixed margins retained relative to the upper limit of the corresponding parameters in the historical operation records of this production line.

[0050] The multi-round optimization iteration process is set to 40 rounds, consistent with the number of rounds for scene feature vector optimization. The first round of candidate parameter combinations starts with the upper limits of the three parameter constraint boundary values. In each subsequent round, a new value is generated based on the previous round's value by a 5% increment. The 5% is taken from the common adjustment ratio in the round-by-round adjustment record of the production line motion parameters. The candidate parameter combinations also include three candidate values: speed threshold, force control threshold, and path avoidance distance, which correspond one-to-one with the aforementioned three parameters.

[0051] The deviation is the difference between the three candidate values ​​of the candidate parameter combination and the corresponding three target values ​​of the target parameter value, totaling three items. The matching error calculation process involves normalizing each of the three differences and summing their squares to obtain the scene-internal matching error. This value is dimensionless and can be directly compared with the error threshold. The error threshold is selected according to the scene type to which the candidate parameter combination belongs, and is consistent with the three error thresholds already recorded in the scene classification system.

[0052] Interval statistical processing retains all candidate parameter combinations whose matching error within the scene is lower than the corresponding error threshold in 40 rounds, without selecting the best single point. For each of the three parameters, the minimum and maximum values ​​that have appeared in the retained results are taken as the lower and upper limits of the values. The upper and lower limits of the three parameters together constitute the feasible adaptation parameter interval.

[0053] In one specific embodiment, step S3 includes:

[0054] Spatiotemporal coordinate calibration is performed on the handover location between two adjacent scenes in the entire link to obtain the handover coordinates;

[0055] Based on the intersection coordinates, boundary value extraction processing is performed on two adjacent feasible adaptation parameter intervals to obtain the first interval boundary value group and the second interval boundary value group. The first interval boundary value group includes the lower limit value and the upper limit value of the first interval, and the second interval boundary value group includes the lower limit value and the upper limit value of the second interval.

[0056] The intersection operation is performed on the first interval boundary value group and the second interval boundary value group to obtain the lower intersection value and the upper intersection value. The intersection operation includes taking the larger value of the first interval lower limit value and the second interval lower limit value to obtain the lower intersection value, and taking the smaller value of the first interval upper limit value and the second interval upper limit value to obtain the upper intersection value.

[0057] When the lower limit of the intersection is greater than the upper limit of the intersection, the difference is calculated based on the lower limit of the second interval and the upper limit of the first interval to obtain the gap amount.

[0058] Specifically, the handover coordinate calibration process targets the docking position between the unloading station in the material transfer scenario and the picking station in the intelligent warehousing scenario. The installation coordinates of this docking position in the production line layout diagram are fixed and do not change with production batches or production schedules. The handover coordinates are composed of the three-dimensional spatial coordinates (12.4, 3.6, 0.8) (unit: meters) of this docking position in the production line coordinate system. The three values ​​correspond to the length, width, and height directions of the production line coordinate system, respectively. The calibration is based on the fact that the fixed installation coordinates are already recorded in the production line layout drawing and do not need to be measured separately.

[0059] Boundary value extraction is performed based on the docking position determined by the handover coordinates. The lower and upper limits of the end-effector speed threshold dimension are extracted from the feasible adaptation parameter ranges of each of the two adjacent scenarios. This dimension is chosen instead of the force control threshold or path avoidance distance because what is transferred between the two scenarios at the handover position is the movement speed of the workpiece or material. The force control threshold and path avoidance distance do not constitute quantities that need to be matched between the two scenarios at this docking position. The first interval boundary value group corresponds to the feasible adaptation parameter range of the material transfer scenario. Since this scenario is upstream of the docking position in the production line material flow sequence, the values ​​are set to the lower limit of 0.18 m / s and the upper limit of 0.25 m / s. The second interval boundary value group corresponds to the feasible adaptation parameter range of the intelligent warehousing scenario. Since this scenario is downstream of the docking position, the values ​​are set to the lower limit of 0.26 m / s and the upper limit of 0.32 m / s. All four values ​​are directly referenced from the lower and upper limits of the feasible adaptation parameter range obtained from the statistical analysis of the end-effector speed threshold components of the two scenarios in S2, without recalculation.

[0060] The intersection operation compares two lower bound values, 0.18 and 0.26, and takes the larger value to obtain the intersection lower bound value of 0.26. It also compares two upper bound values, 0.25 and 0.32, and takes the smaller value to obtain the intersection upper bound value of 0.25. Since the intersection lower bound value of 0.26 is greater than the intersection upper bound value of 0.25, it indicates that the lower bound value obtained according to this rule has exceeded the upper bound value obtained according to this rule, and the two feasible adaptation parameter intervals do not overlap in the dimension of the end effector speed threshold.

[0061] Comparing the lower intersection limit of 0.26 and the upper intersection limit of 0.25 obtained in this round with the four referenced original values, we can see that the lower intersection limit of 0.26 is exactly equal to the lower limit of the second interval itself, and the upper intersection limit of 0.25 is exactly equal to the upper limit of the first interval itself. This correspondence always holds when the intervals do not overlap, because taking the larger value will always fall on the lower limit of the larger value side, and taking the smaller value will always fall on the upper limit of the smaller value side. The difference calculation directly uses the two already determined original values, the lower limit of the second interval of 0.26 and the upper limit of the first interval of 0.25, without repeating the intermediate process of the intersection calculation. Subtracting the upper limit of the first interval from the lower limit of the second interval, i.e., 0.26 minus 0.25, yields a gap of 0.01 meters per second. This gap represents the degree of misalignment between the upstream material transfer scenario and the downstream intelligent warehousing scenario at the docking position in terms of the end effector speed threshold.

[0062] Figure 2 This is a schematic diagram illustrating the intersection operation and gap determination of two adjacent feasible adaptation parameter intervals in an embodiment of this application. Figure 2The diagram shows the boundary value distribution of the feasible adaptation parameter intervals for two adjacent scenarios on the dimension of the end effector speed threshold. The first interval boundary value group corresponds to the material transfer side, and the second interval boundary value group corresponds to the smart warehousing side. By taking the larger value of the lower limit and the smaller value of the upper limit of the two intervals, the intersection lower limit and intersection upper limit are obtained. When the intersection lower limit is greater than the intersection upper limit, it indicates that there is no overlap between the two intervals. The non-overlapping interval is marked by the shaded area in the figure, and the gap value calculated from it is also marked.

[0063] In one specific embodiment, when the gap amount is greater than zero, feature increment extraction is performed on the scene feature vectors on both sides of the gap amount to obtain the feature increment. The buffer capacity requirement is then calculated based on the feature increment to obtain the buffer capacity requirement parameters, including:

[0064] Based on the intersection coordinates, feature increments are extracted from the scene feature vectors on both sides to obtain feature increments;

[0065] Based on the feature increment, the feasible adaptation parameter intervals on both sides of the preset statistical window are corrected and the mean is calculated to obtain the mean of the first interval and the mean of the second interval.

[0066] The theoretical output quantity is obtained by rounding down the mean of the first interval and the preset statistical window. The theoretical consumption quantity is obtained by rounding down the mean of the second interval and the preset statistical window.

[0067] The difference between the theoretical output quantity and the theoretical consumption quantity is calculated to obtain the quantity difference. The quantity difference is then superimposed with the preset safety factor to obtain the buffer capacity requirement parameter.

[0068] Specifically, the feature increment extraction process is based on the scene on both sides of the docking position determined by the handover coordinates (12.4, 3.6, 0.8). The difference between the measured value of the current conveying cycle time of the material transfer scene (3.4 seconds per piece) and the measured value of the current storage and retrieval cycle time of the smart warehouse scene (3.9 seconds per piece) is taken to obtain the feature increment of 0.5 seconds per piece. This value reflects the degree of misalignment between the current cycle time of the smart warehouse side and the current cycle time of the material transfer side, with the direction being larger on the smart warehouse side.

[0069] The correction and mean calculation process corrects the feasible adaptation parameter range on both sides of the preset statistical window with a feature increment of 0.5 seconds per item. The preset statistical window duration is 60 seconds, which is consistent with the statistical window duration set in the method for determining buffer capacity requirement parameters. In the feasible adaptation parameter range for material transfer, the lower limit of the conveying cycle time component is 3.0 seconds per piece, and the upper limit is 3.4 seconds per piece. Since the cycle time on this side is relatively fast compared to the current misalignment, the lower limit and the upper limit are each reduced by half of the feature increment value, i.e., 0.25 seconds per piece, for correction. The corrected lower limit is 2.75 seconds per piece, and the corrected upper limit is 3.15 seconds per piece. The arithmetic average of the two is the first interval mean of 2.95 seconds per piece. In the feasible adaptation parameter range for smart warehousing, the lower limit of the access cycle time component is 3.6 seconds per piece, and the upper limit is 4.0 seconds per piece. Since the cycle time on this side is relatively slow compared to the current misalignment, the lower limit and the upper limit are each increased by half of the feature increment value, i.e., 0.25 seconds per piece, for correction. The corrected lower limit is 3.85 seconds per piece, and the corrected upper limit is 4.25 seconds per piece. The arithmetic average of the two is the second interval mean of 4.05 seconds per piece. The correction magnitude is half of the feature increment value for each side, based on the fact that the misalignment is jointly generated by the two sides and is distributed equally between the two sides without bias towards either side.

[0070] The preset statistical window of 60 seconds is divided by the average of the first interval of 2.95 seconds per item, and the quotient is 20.34. After rounding down, the theoretical output quantity is 20 items. The basis for rounding down is the number of items that are completely produced within the corresponding window. The unfinished part is not included. The preset statistical window of 60 seconds is divided by the average of the second interval of 4.05 seconds per item, and the quotient is 14.81. After rounding down, the theoretical consumption quantity is 14 items.

[0071] Subtracting the theoretical consumption quantity of 14 units from the theoretical output quantity of 20 units yields a quantity difference of 6 units. This quantity difference of 6 units is added to the preset safety factor of 1 unit, resulting in a buffer capacity requirement parameter of 7 units. The basis for setting the preset safety factor to 1 unit is consistent with the explanation in the method for determining the buffer capacity requirement parameter, and is used to cover the truncation errors generated by the two rounding processes.

[0072] In one specific embodiment, the buffer capacity requirement parameter and the gap amount are combined to obtain a composite handover scenario feature vector. The composite handover scenario feature vector is then encoded according to the scenario type and written into a scenario classification system, including:

[0073] The buffer capacity requirement parameter and the gap amount are concatenated to obtain the initial feature vector;

[0074] The scene type codes corresponding to the scene feature vectors on both sides are combined and encoded to obtain composite scene type codes;

[0075] The initial feature vector and the composite scene type encoding are bound together to obtain the composite handover scene feature vector;

[0076] By extending the write interface, the feature vector of the composite handover scene is input into the scene classification system for registration processing, and a composite handover scene instance is obtained.

[0077] Specifically, the splicing process places the buffer capacity requirement parameter 7 first and the gap amount 0.01 meters per second second, splicing them in this fixed order to form an initial feature vector (7, 0.01). The reason for placing the buffer capacity requirement parameter first is that this parameter corresponds to the physical buffer space jointly occupied by both parties in the handover, while the gap amount corresponds to the numerical misalignment in a single dimension. The former, as the main occupancy attribute of the handover scenario, is placed first.

[0078] The combined coding process arranges the adjacent scene type codes in the order of material transfer side first and intelligent warehousing side last. The two-level code for this instance of material transfer scene is B01, and the two-level code for this instance of intelligent warehousing scene is C01. No separator is inserted between the two codes, and they are directly connected end to end to get B01C01. Then, the sequence number 01 of the intersection coordinate is added to the end, because this intersection position is the first position in the production line to be marked with the intersection coordinate. Finally, the composite scene type code B01C0101 is obtained. The reason for adding the intersection coordinate sequence number is that there may be multiple intersection positions for the same pair of scene types in the production line, and it is necessary to distinguish the composite scene type codes corresponding to different intersection positions by the sequence number.

[0079] The binding process combines the initial feature vector (7, 0.01) with the composite scene type code B01C0101 to form a record. The former serves as the numerical content of the record, and the latter serves as the identifier field of the record. The correspondence between the two is consistent with the correspondence between the numerical content and the two-level code in the existing records in the scene classification system. After binding, the composite transition scene feature vector is obtained, which is the record composed of the numerical content (7, 0.01) and the identifier field B01C0101.

[0080] The registration process appends the record to the storage structure of the scene classification system through the extended write interface. The writing method is the same as the scene instance entry method in S1, that is, only an independent record is appended according to the composite scene type code B01C0101, without changing the content and index position of the existing corresponding record. After the writing is completed, the resulting record is the composite handover scene instance, which can be retrieved in the scene classification system according to the code B01C0101.

[0081] In one specific embodiment, the motion parameters at the intersection coordinates are iteratively processed using an interval iterative optimization algorithm to obtain a weighted sum of buffer overflow rate and idle rate. When the weighted sum converges to below a preset threshold, the buffer trigger threshold and release beat interval are output to obtain the cooperative control parameters, including:

[0082] The interval iterative optimization algorithm is used to iteratively process the action parameters of the composite handover scene instance and the handover coordinates to obtain the number of buffer occupancy in each round.

[0083] The overflow rate for each round is calculated based on the number of buffers occupied and the buffer capacity requirement parameters. The under-fill rate for each round is calculated based on the number of buffers occupied and the buffer capacity requirement parameters.

[0084] The buffer overflow rate and idle rate are weighted and summed to obtain the weighted sum for each round.

[0085] When the weighted sum converges to below a preset threshold, the number of buffer occupancy corresponding to the convergence is calibrated to obtain the buffer trigger threshold. The action parameters at the intersection coordinates corresponding to the convergence are statistically processed to obtain the release beat interval. The buffer trigger threshold and the release beat interval are combined to obtain the collaborative control parameters.

[0086] Specifically, in this step, the iteration object of the interval iterative optimization algorithm is the cycle time conversion value corresponding to the end effector speed threshold component in the motion parameters at the intersection coordinate. The number of iteration rounds is reused for 40 rounds, consistent with S2. The candidate cycle time value in each round changes by 5% based on the value in the previous round, and this magnitude reuses the adjustment ratio determined in S2. The candidate cycle time value on the material transfer side is the adjusted first interval average of 2.95 seconds per piece in the first round, and the candidate cycle time value on the intelligent warehousing side is the adjusted second interval average of 4.05 seconds per piece in the first round. Both of these values ​​are obtained during the determination of the buffer capacity requirement parameters.

[0087] The buffer occupancy quantity in the first round is calculated using the buffer capacity requirement parameter of 7 units. In each subsequent round, a 60-second statistical window is used to divide the theoretical output quantity by the candidate cycle time value on the material transfer side for that round and rounded down to obtain the theoretical consumption quantity. The rounding methods for both are consistent with those for the theoretical output quantity and theoretical consumption quantity. The difference between the theoretical output quantity and the theoretical consumption quantity is added to the buffer occupancy quantity from the previous round to obtain the buffer occupancy quantity for that round.

[0088] The portion of the buffer space exceeding the required buffer capacity of 7 units is divided by 7 units to obtain the buffer overflow rate for that round. If the number of buffer units occupied is less than 7, the overflow rate is recorded as 0. The portion of the buffer space exceeding the required buffer capacity of 7 units is divided by 7 units to obtain the idle rate for that round. If the number of buffer units occupied is not less than 7, the idle rate is recorded as 0. Both ratios will not be non-zero in the same round. The buffer overflow rate and idle rate are each weighted at 0.5 to obtain the weighted sum for that round. The reason for assigning equal weights to both is that the impact of buffer overfilling and underfilling deviations on the handover process is not differentiated in this scenario. The preset threshold is 0.05, based on the fact that this value corresponds to a buffer quantity deviation of no more than 5% of its capacity, which is a negligible fluctuation range in the production line handover scenario, consistent with the proportional basis used in the aforementioned error threshold setting method.

[0089] In the 22nd iteration, the number of buffer items occupied decreased to 7.28. At this point, the buffer overflow rate was (7.28-7) / 7, or 0.04, and the idle rate was 0. The two were added together with a weight of 0.5, resulting in a weighted sum of 0.02, which was lower than the preset threshold of 0.05. This was considered convergence. The convergence criterion was that the weighted sum in that round first fell below 0.05, and it was not required to maintain this value for multiple consecutive rounds. The threshold calibration process directly calibrated the number of buffer items occupied in the 22nd round, 7.28, as the buffer trigger threshold without further calculation.

[0090] The interval statistical processing targets the candidate cycle value of the intelligent warehousing side in the action parameters at the intersection coordinate. This component is selected based on the fact that the release cycle corresponds to the cycle time during which materials in the buffer flow to the intelligent warehousing side and are absorbed. The statistical range is the minimum value (3.85 seconds per piece) and the maximum value (3.98 seconds per piece) of this component that appeared within the 22nd round and the previous four consecutive rounds (a total of five rounds). These two values ​​constitute the release cycle interval [3.85, 3.98] (per piece per second). The buffer trigger threshold of 7.28 pieces is ranked first, followed by the release cycle interval [3.85, 3.98], and these parameters are combined in this fixed order to form the collaborative control parameters. The order of arrangement is consistent with the order in which the buffer capacity requirement parameter in the initial feature vector precedes the gap quantity.

[0091] Figure 3 This is a schematic diagram illustrating the convergence of the weighted sum of buffer overflow rate and idle rate with iteration rounds in an embodiment of this application. Figure 3The figure illustrates the convergence trends of buffer overflow rate, idle rate, and their weighted sum as the interval iterative optimization algorithm processes the action parameters at the intersection coordinates in multiple iterations. The figure uses dashed lines to indicate the buffer overflow rate, dashed lines to indicate the idle rate, and solid lines to indicate the weighted sum. Horizontal dotted lines indicate the preset threshold. Hollow circles mark the position where the weighted sum curve first drops below the preset threshold in a certain iteration, corresponding to the buffer trigger threshold and release beat interval output in that iteration.

[0092] The above describes the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method in the embodiments of this application. The following describes the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation system in the embodiments of this application. One embodiment of the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation system in the embodiments of this application includes:

[0093] The modeling module is used to perform feature modeling on the processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios in the entire chain, to obtain scenario feature vectors, and to input the scenario feature vectors into the scenario classification system.

[0094] The iteration module is used to iteratively process the scene feature vector and the robot's motion parameters through an interval iterative optimization algorithm to obtain a feasible adaptation parameter interval.

[0095] The detection module is used to perform intersection detection processing on two adjacent feasible adaptation parameter intervals to obtain the gap amount;

[0096] The control module is used to: extract feature increments from the scene feature vectors on both sides of the gap when the gap is greater than zero, obtain feature increments, calculate buffer capacity requirements based on the feature increments, obtain buffer capacity requirement parameters, combine the buffer capacity requirement parameters and the gap to obtain a composite handover scene feature vector, encode the composite handover scene feature vector according to the scene type and write it into the scene classification system, iterate the action parameters at the handover coordinates using the interval iterative optimization algorithm to obtain a weighted sum of buffer overflow rate and idle rate, and output the buffer trigger threshold and release beat interval when the weighted sum converges to below a preset threshold to obtain collaborative control parameters.

[0097] This application embodiment also provides a smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation device, which can be a server. This smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor in this computer design provides computing and control capabilities. The memory of the smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation device is used to store the data corresponding to this embodiment. The network interface of the smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation device is used for communication with external terminals via network connection. When the computer program is executed by the processor, it implements the above-described method.

[0098] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the intelligent manufacturing full-link embodied intelligent collaborative operation scenario adaptation method.

[0099] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0100] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a smart manufacturing end-to-end intelligent collaborative operation scenario adaptation device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for adapting to intelligent collaborative operation scenarios across the entire intelligent manufacturing chain, characterized in that, The method includes: Step S1: Perform feature modeling on the processing and assembly scenario, material transfer scenario, and intelligent warehousing scenario in the entire chain to obtain scenario feature vectors, and input the scenario feature vectors into the scenario classification system; Step S2: Iteratively process the scene feature vector and the robot's motion parameters using an interval iterative optimization algorithm to obtain a feasible adaptation parameter range; Step S3: Perform intersection detection processing on two adjacent feasible adaptation parameter intervals to obtain the gap amount; Step S4: When the gap amount is greater than zero, feature increment extraction is performed on the scene feature vectors on both sides of the gap amount to obtain feature increments. The buffer capacity requirement is calculated based on the feature increments to obtain buffer capacity requirement parameters. The buffer capacity requirement parameters and the gap amount are combined to obtain a composite handover scene feature vector. The composite handover scene feature vector is encoded according to the scene type and written into the scene classification system. The action parameters at the handover coordinates are iteratively processed by the interval iterative optimization algorithm to obtain a weighted sum of buffer overflow rate and idle rate. When the weighted sum converges to below a preset threshold, the buffer trigger threshold and release beat interval are output to obtain the collaborative control parameters.

2. The intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method according to claim 1, characterized in that, Step S1 includes: Feature extraction processing is performed on the working condition data of processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios to obtain feature parameters of processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios. Based on the feature parameters of the processing and assembly scenario, the feature parameters of the material transfer scenario, and the feature parameters of the intelligent warehousing scenario, vectorization processing is performed to obtain feature vectors of the processing and assembly scenario, the material transfer scenario, and the intelligent warehousing scenario. Based on the feature vectors of the processing and assembly scenario, the material transfer scenario, and the intelligent warehousing scenario, the action parameters are processed to set an error threshold, thereby obtaining the error threshold. The processing and assembly scenario feature vector, the material transfer scenario feature vector, the intelligent warehousing scenario feature vector, and the error threshold are entered into the scenario classification system according to the scenario type code for input processing, to obtain the scenario instance corresponding to the scenario feature vector, and an extended write interface is configured in the scenario classification system.

3. The intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method according to claim 2, characterized in that, Step S2 includes: The action parameters are subjected to constraint boundary mapping based on the scene feature vector to obtain parameter constraint boundaries, and the action parameters are subjected to target value inverse calculation based on the scene feature vector to obtain target parameter values. Based on the parameter constraint boundary, the action parameters are subjected to multiple rounds of optimization iteration to obtain the candidate parameter combination corresponding to each round; Based on the deviation between the candidate parameter combination and the target parameter value, the matching error of the candidate parameter combination is calculated to obtain the scene intra-matching error corresponding to each round. The candidate parameter combinations whose matching error within the scene is lower than the error threshold are subjected to interval statistical processing to obtain the feasible adaptation parameter interval.

4. The intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method according to claim 3, characterized in that, Step S3 includes: Spatiotemporal coordinate calibration is performed on the handover location between two adjacent scenes in the entire link to obtain the handover coordinates; Based on the intersection coordinates, boundary value extraction processing is performed on two adjacent feasible adaptation parameter intervals to obtain a first interval boundary value group and a second interval boundary value group. The first interval boundary value group includes a first interval lower limit value and a first interval upper limit value, and the second interval boundary value group includes a second interval lower limit value and a second interval upper limit value. The first interval boundary value group and the second interval boundary value group are subjected to intersection operation processing to obtain the lower intersection value and the upper intersection value. The intersection operation processing includes taking the larger value of the first interval lower limit value and the second interval lower limit value to obtain the lower intersection value, and taking the smaller value of the first interval upper limit value and the second interval upper limit value to obtain the upper intersection value. When the lower limit of the intersection is greater than the upper limit of the intersection, the difference is calculated based on the lower limit of the second interval and the upper limit of the first interval to obtain the gap amount.

5. The intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method according to claim 4, characterized in that, When the gap amount is greater than zero, feature increment extraction is performed on the scene feature vectors on both sides of the gap amount to obtain feature increments. The buffer capacity requirement is then calculated based on these feature increments to obtain buffer capacity requirement parameters, including: Based on the intersection coordinates, feature increments are extracted from the scene feature vectors on both sides to obtain feature increments; Based on the feature increment, the feasible adaptation parameter intervals on both sides of the preset statistical window are corrected and the mean is calculated to obtain the mean of the first interval and the mean of the second interval. The theoretical output quantity is obtained by rounding down the preset statistical window and the mean of the first interval, and the theoretical consumption quantity is obtained by rounding down the preset statistical window and the mean of the second interval. The difference between the theoretical output quantity and the theoretical absorption quantity is calculated to obtain the quantity difference. The quantity difference is then superimposed with a preset safety factor to obtain the buffer capacity requirement parameter.

6. The intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method according to claim 5, characterized in that, The process of combining the buffer capacity requirement parameter and the gap amount to obtain a composite handover scene feature vector, and then writing the composite handover scene feature vector into the scene classification system according to the scene type encoding, includes: The buffer capacity requirement parameter and the gap amount are concatenated to obtain an initial feature vector; The scene type codes corresponding to the scene feature vectors on both sides are combined and encoded to obtain a composite scene type code; The initial feature vector and the composite scene type encoding are bound together to obtain the composite handover scene feature vector; The composite handover scene feature vector is input into the scene classification system through the extended write interface for registration processing, thereby obtaining a composite handover scene instance.

7. The intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method according to claim 6, characterized in that, The action parameters at the intersection coordinates are iteratively processed using the interval iterative optimization algorithm to obtain a weighted sum of buffer overflow rate and idle rate. When the weighted sum converges to below a preset threshold, the buffer trigger threshold and release beat interval are output to obtain the cooperative control parameters, including: The interval iterative optimization algorithm is used to iteratively process the action parameters at the intersection coordinates of the composite intersection scene instance to obtain the buffer occupancy quantity for each round. The overflow rate for each round is calculated based on the number of buffers occupied and the buffer capacity requirement parameters. The under-fill rate for each round is calculated based on the number of buffers occupied and the buffer capacity requirement parameters. The buffer overflow rate and the idle rate are weighted and summed to obtain the weighted sum for each round. When the weighted sum converges to below a preset threshold, the number of buffer occupancy corresponding to the convergence is calibrated to obtain the buffer trigger threshold. The action parameters at the intersection coordinates corresponding to the convergence are statistically processed to obtain the release beat interval. The buffer trigger threshold and the release beat interval are combined to obtain the cooperative control parameters.

8. A smart manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation system, characterized in that, For implementing the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method as described in any one of claims 1-7, the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation system includes: The modeling module is used to perform feature modeling on the processing and assembly scenarios, material transfer scenarios, and intelligent warehousing scenarios in the entire chain, to obtain scenario feature vectors, and to input the scenario feature vectors into the scenario classification system. The iteration module is used to iteratively process the scene feature vector and the robot's motion parameters through an interval iterative optimization algorithm to obtain a feasible adaptation parameter interval. The detection module is used to perform intersection detection processing on two adjacent feasible adaptation parameter intervals to obtain the gap amount; The control module is used to: extract feature increments from the scene feature vectors on both sides of the gap when the gap is greater than zero, obtain feature increments, calculate buffer capacity requirements based on the feature increments, obtain buffer capacity requirement parameters, combine the buffer capacity requirement parameters and the gap to obtain a composite handover scene feature vector, encode the composite handover scene feature vector according to the scene type and write it into the scene classification system, iterate the action parameters at the handover coordinates using the interval iterative optimization algorithm to obtain a weighted sum of buffer overflow rate and idle rate, and output the buffer trigger threshold and release beat interval when the weighted sum converges to below a preset threshold to obtain collaborative control parameters.

9. A smart manufacturing end-to-end intelligent collaborative operation scenario adaptation device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the intelligent manufacturing full-link embodied intelligent collaborative operation scenario adaptation method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the intelligent manufacturing end-to-end embodied intelligent collaborative operation scenario adaptation method as described in any one of claims 1 to 7.

Images