Digital twin and multi-objective optimization integrated enterprise intelligent warehouse scheduling method and system

Abstract

The present application relates to the field of information technology and intelligent warehousing technology, in particular to a digital twin and multi-objective optimization fusion enterprise intelligent warehousing scheduling method and system, the method comprising: receiving a trigger signal, collecting first monitoring data; making a target decision to obtain a plurality of pending targets; determining a monitoring group, matching the corresponding signal conditioning strategy; collecting second monitoring data, preprocessing the second monitoring data, and conditioning the preprocessed second monitoring data; making a secondary decision on the plurality of pending targets to obtain at least one execution target, and issuing the execution target and executing it. The present application effectively solves the problems of creating a high-precision digital twin model and maintaining real-time synchronization of massive data consuming huge computing resources by constructing a two-stage hierarchical decision-making framework from coarse to fine. Applied to warehouse scheduling, a high-confidence scheduling strategy can be obtained without completely replicating the reality details.

Description

Technical Field

[0001] This invention relates to the fields of information technology and intelligent warehousing technology, and in particular to an enterprise intelligent warehouse scheduling method and system that integrates digital twins and multi-objective optimization. Background Technology

[0002] The integration of digital twins and multi-objective optimization algorithms is driving a paradigm shift in intelligent warehouse scheduling from "passive response" to "proactive prediction and self-optimization." The core idea is to construct a "digital copy" that is synchronized in real-time with the physical warehouse. Through multi-objective optimization and "if-what-if" reasoning within this virtual environment, the optimal strategy is then mapped back to reality, achieving a systemic performance improvement.

[0003] At the system deployment level, the main task is to establish a standardized multi-layered architecture to achieve a precise mapping from the physical world to the digital world. A typical framework includes a perception layer (using sensors such as RFID, UWB, and LiDAR for millisecond-level data sampling), a data / twin layer (building high-fidelity digital twin models on simulation platforms such as NVIDIA Omniverse), a model / decision layer (the core of multi-objective optimization, using algorithms to deduce and generate strategies), and an application / execution layer (issuing decision instructions to execution systems such as AGVs, WMS / WCS). At the system functionality implementation level, key aspects of the scheduling problem, such as agent collaboration and path planning (handling multi-AGV / AMR collaborative operation), task allocation (rationally allocating transportation and picking tasks to multiple robots), order and storage location optimization (balancing efficiency, cost, space utilization, and human safety in complex SKU storage location allocation), production-warehousing collaboration (coordinating production and warehousing rhythms to avoid congestion), and anomaly handling (monitoring and responding to sudden anomalies and quickly generating response strategies), are defined as explicit multi-objective optimization problems, posing the main challenge to system implementation.

[0004] In existing technologies, creating high-precision models and maintaining real-time synchronization of massive amounts of data consumes huge computing resources, and completely replicating real-world details is too expensive and unnecessary. The core challenge is to find the best balance between complexity and practicality. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and system for intelligent warehouse scheduling that integrates digital twins and multi-objective optimization to address the above-mentioned problems.

[0006] This invention is implemented as follows: a digital twin and multi-objective optimization fusion enterprise intelligent warehouse scheduling method, comprising: Upon receiving the trigger signal, the first monitoring data is collected; Based on the first monitoring data, a target decision is made to obtain several undetermined targets; Based on the identified targets, monitoring groups are determined, and corresponding signal conditioning strategies are matched for each monitoring group. The monitoring group collects second monitoring data, preprocesses the second monitoring data based on the correlation between the second monitoring data and the first monitoring data, and conditions the preprocessed second monitoring data according to the matching signal conditioning strategy. Based on the adjusted second monitoring data, a secondary decision is made on several pending targets to obtain at least one execution target, which is then issued and executed.

[0007] In one embodiment, a digital twin and multi-objective optimization integrated enterprise intelligent warehouse scheduling system is provided, comprising: A sensing layer, which is used to acquire data through sensing elements; A data twin layer, which uses the data collected by the perception layer as input, is used for the digital simulation of real-world scenarios or processes; A model decision layer, which is connected to the data twin layer, is used to execute the enterprise intelligent warehouse scheduling method that integrates digital twins and multi-objective optimization as described in the embodiments of the present invention. An execution layer, which is connected to the model decision layer, is used to execute the execution objectives output by the model decision layer.

[0008] The intelligent warehouse scheduling method integrating digital twins and multi-objective optimization provided in this invention effectively solves the challenges of creating high-precision digital twin models and maintaining real-time synchronization of massive amounts of data, which consume enormous computing resources, by constructing a two-stage hierarchical decision-making framework from coarse to fine. Applied to warehouse scheduling, the first decision uses coarse-grained first monitoring data to quickly generate a diverse set of candidate targets. The second decision uses finely conditioned second monitoring data for precise selection, obtaining a high-confidence scheduling strategy without completely replicating real-world details. Simultaneously, through intelligent matching of monitoring group division and signal conditioning strategies, data dimensionality and redundancy are further reduced, significantly lowering system resource consumption and achieving an optimal balance between scheduling real-time performance and decision accuracy. Attached Figure Description

[0009] Figure 1 This is a flowchart of an enterprise intelligent warehouse scheduling method that integrates digital twins and multi-objective optimization in one embodiment; Figure 2 This is a structural block diagram of an enterprise intelligent warehouse scheduling system that integrates digital twins and multi-objective optimization in one embodiment; Figure 3This is a block diagram of the internal structure of a computer device in one embodiment. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 merely illustrative and not intended to limit the invention.

[0011] It is understood that the terms "first," "second," etc., used in this invention may be used to describe various elements herein, but unless specifically stated otherwise, these elements are not limited by these terms. These terms are used only to distinguish one element from another. For example, without departing from the scope of this invention, a first script may be referred to as a second script, and similarly, a second script may be referred to as a first script.

[0012] As attached Figure 1 As shown, this invention provides a digital twin and multi-objective optimization fusion method for intelligent warehouse scheduling in enterprises, which includes the following steps: S1. Upon receiving the trigger signal, collect the first monitoring data; S2. Based on the first monitoring data, a target decision is made to obtain several undetermined targets; S3. Based on the identified targets, determine the monitoring groups and match the corresponding signal conditioning strategies for each monitoring group; S4. Collect second monitoring data according to the monitoring group, preprocess the second monitoring data according to the correlation between the second monitoring data and the first monitoring data, and condition the preprocessed second monitoring data according to the matching signal conditioning strategy; S5. Based on the adjusted second monitoring data, make secondary decisions on several pending targets to obtain at least one execution target, issue the execution target and execute it.

[0013] In this embodiment, the above steps constitute a two-stage hierarchical optimization framework from coarse to fine. First, when the system receives a trigger signal (e.g., order arrival, inventory shortage warning, or AGV (Automated Guided Vehicle) scheduling request), it collects first monitoring data, which is typically coarse-grained, low-cost real-time data (e.g., warehouse occupancy rate, order arrival rate, AGV power distribution). Based on this data, the first objective decision is made, using a multi-objective evolutionary algorithm (e.g., NSGA-II) to quickly filter out several undetermined objectives (e.g., Pareto front solution set), each objective representing a different optimization tendency (e.g., time optimization, energy optimization, cost optimization). Next, based on these undetermined objectives, the items requiring fine-grained monitoring are determined and monitoring groups are divided, with appropriate signal conditioning strategies configured for each group. Then, high-precision second monitoring data (e.g., AGV real-time position, instantaneous motor current value) is collected based on the monitoring groups, and the first monitoring data is used as a baseline for preprocessing, eliminating invariant or redundant data, and then filtering, fusing, or feature extraction operations are performed on the data according to the matched conditioning strategy. Finally, the refined data is used to make a second decision on the target (e.g., using entropy weighting + TOPSIS), selecting a single or a small number of execution targets, which are then sent to the execution system (such as WCS, AGV scheduling module) for scheduling. The entire process avoids the high computational cost of performing complete multi-objective optimization directly on high-dimensional refined data, achieving the best balance between complexity and practicality.

[0014] The intelligent warehouse scheduling method integrating digital twins and multi-objective optimization provided in this invention effectively solves the challenges of creating high-precision digital twin models and maintaining real-time synchronization of massive amounts of data, which consume enormous computing resources, by constructing a two-stage hierarchical decision-making framework from coarse to fine. Applied to warehouse scheduling, the first decision uses coarse-grained first monitoring data to quickly generate a diverse set of candidate targets. The second decision uses finely conditioned second monitoring data for precise selection, obtaining a high-confidence scheduling strategy without completely replicating real-world details. Simultaneously, through intelligent matching of monitoring group division and signal conditioning strategies, data dimensionality and redundancy are further reduced, significantly lowering system resource consumption and achieving an optimal balance between scheduling real-time performance and decision accuracy.

[0015] In one embodiment of the present invention, receiving the trigger signal and collecting the first monitoring data includes: Upon receiving a trigger signal, determine the production stage to which the trigger signal belongs; Identify the product batches directly and indirectly related to the production process to which the trigger signal belongs; Obtain the set indicators for the identified product batch; The monitoring items corresponding to each set indicator are determined, and the first monitoring data is obtained by collecting data from the determined monitoring items.

[0016] In this embodiment, the method for collecting the first monitoring data is further defined. The aim is to achieve refined, layered data collection on demand, avoiding resource waste caused by indiscriminate global data collection. Specifically, when the system receives a trigger signal (e.g., "Material shortage alarm for production line A"), it first identifies which production stage the signal belongs to (e.g., "assembly stage," "packaging stage"). Then, it tracks all product batches directly related to this stage (e.g., the batch currently being processed) and indirectly related (e.g., the batch awaiting transfer in the previous process). Each product batch has preset production indicators (e.g., "planned completion time," "quality pass rate," "priority"). For these indicators, the system searches for the underlying physical items that need to be monitored (e.g., "AGV position," "material bin inventory," "equipment status"), and only collects the first monitoring data for these items. For example, if the trigger signal is "Batch B urgently needs to be shipped out," the associated production stage is "shipping operation," the associated batch is batch B and its preceding and following batches, the set indicators are "shipping efficiency" and "order on-time rate," and the corresponding monitoring items are "AGV queue length at the shipping outlet" and "material bin position for batch B." This approach precisely limits the collection of the first monitoring data to the production chain related to the trigger signal, effectively reducing the initial data volume and laying the foundation for rapid decision-making in the future.

[0017] In one embodiment of the present invention, the step of making a target decision based on the first monitoring data to obtain a plurality of undetermined targets includes: Based on the first monitoring data, the three key elements of the multi-objective optimization problem are determined: decision variables, objective function, and constraints. The initial parent population is obtained by initializing candidate solutions based on the determined three elements; A non-dominated sorting method is used to determine the Pareto front level of all candidate solutions in the initial parent population based on the value of each candidate solution on each objective function, and at the same time calculate the crowding distance of each candidate solution. Based on the calculated Pareto front level and crowding distance, genetic operations are performed according to the set rules to generate offspring populations. The generated offspring populations are then merged with the initial parent populations, and non-dominated sorting and crowding distance calculations are performed again. An elite retention strategy is used to select a new generation of populations. This process is repeated until the termination condition is met to obtain several undetermined targets.

[0018] This embodiment specifically describes the algorithm used for the first objective decision-making—namely, the second-generation non-dominated sorting genetic algorithm (NSGA-II). Its core advantage lies in its ability to efficiently generate a set of diverse Pareto front solutions (undetermined objectives) based on coarse initial monitoring data. The specific process is as follows: First, based on the initial monitoring data (such as the current number of AGVs, task list, and warehouse layout), decision variables (such as task allocation schemes and travel path parameters for each AGV), objective functions (such as minimizing total completion time f1, minimizing total energy consumption f2, and maximizing order fulfillment rate f3), and constraints (such as AGV load limits and time windows) are defined. Then, a set of initial candidate solutions (such as 100 different task allocation schemes) is randomly generated as the parent population. Next, non-dominated sorting is used: for each solution, its value on each objective function is calculated. If a solution is not inferior to another solution on all objectives and is better on at least one objective, then the former dominates the latter. Based on the dominance relationship, all solutions are divided into different levels (the first level is the non-dominated solution, i.e., the Pareto front). Simultaneously, the crowding distance of each solution is calculated, i.e., the density of adjacent solutions in the multi-objective space within the same front level, to maintain the diversity of the solution set. Subsequently, tournament selection is performed based on the non-dominated level (lower levels are better) and the crowding distance (larger distances are better), followed by simulated binary crossover and polynomial mutation to generate the offspring population. After merging the offspring with the parent generation, non-dominated sorting and crowding distance calculations are performed again, and an elite retention strategy is used to select the next generation. This iteration is repeated until a preset number of generations or a convergence threshold is reached. Finally, several undetermined objectives (e.g., 5–20) are selected from the optimal Pareto front, corresponding to different tendencies such as "time optimal," "energy optimal," and "balanced solution." This algorithm is robust to data noise and has controllable computational complexity, making it suitable for real-time scheduling scenarios.

[0019] In one embodiment of the present invention, determining the monitoring group based on a plurality of identified undetermined targets includes: Based on the identified several undetermined targets, obtain all the original monitoring items that are physically or logically associated with each undetermined target. Based on all the original monitoring items obtained, similar items are merged and duplicate items are removed to obtain the complete set of monitoring items; Based on the physical correlation and primary / secondary relationship between the monitoring items in the complete set of monitoring items, several monitoring groups are divided. Each monitoring group contains at least one monitoring item, and the monitoring items in the same group are coupled.

[0020] This embodiment solves the data redundancy problem caused by the simple accumulation of monitoring items associated with each target in traditional methods. For example, the first decision generates three pending targets: "time optimization," "energy optimization," and "safety optimization." The original monitoring items associated with the time optimization target include (AGV real-time location, intersection occupancy status, task queue length); energy optimization is associated with (AGV real-time location, motor current, battery voltage); and safety optimization is associated with (AGV real-time location, LiDAR obstacle distance, motor temperature). First, all items are merged and duplicates are removed (e.g., if "AGV real-time location" appears three times, only one is retained), resulting in a complete set of monitoring items. Then, they are grouped according to physical correlation and priority: "AGV real-time location" is grouped with "intersection occupancy status" and "task queue length" (reflecting time efficiency); "motor current," "battery voltage," and "motor temperature" are grouped together (reflecting energy consumption and health); and "LiDAR obstacle distance" is grouped separately (reflecting safety). Items within each group are strongly coupled (e.g., current and temperature jointly determine energy consumption status), while the groups are relatively independent. In this way, the original massive monitoring projects are compressed into a small number of highly cohesive monitoring groups, reducing the burden on subsequent adjustments and decision-making.

[0021] In one embodiment of the present invention, matching a corresponding signal conditioning strategy for each monitoring group includes: Acquire all raw data streams for each monitoring group; For each monitoring group, determine the multi-source relationship of the raw data stream, perform dynamic analysis on the raw data stream, and determine the type of interference in the raw data stream; By integrating the multi-source relationships, dynamic analysis results, and interference types obtained in the previous step, a unique type profile is obtained for each monitoring group; Based on the obtained type profile, the corresponding preset operation sequence is determined to obtain the signal conditioning strategy for the monitoring group.

[0022] In this embodiment, a signal conditioning strategy is tailored for each monitoring group. The core of this approach lies in obtaining a unique "type profile" through multi-dimensional feature integration, thereby avoiding conditioning logic conflicts. Specifically, for each monitoring group, the multi-source relationship of its internal raw data stream is first determined (single-source, redundant, complementary, or correlated); then, the dynamics are analyzed (steady-state, transient pulses, periodic, or general dynamics); finally, the interference type is evaluated (high-frequency noise, power frequency interference, impulse noise, or low-noise). Next, the results from these three dimensions are integrated into a unique type profile (one of six types) according to a predetermined priority (redundancy prioritizes complementarity / correlation, transient pulses prioritize periodicity, etc.). Based on this profile, a corresponding set of conditioning operations (e.g., median filtering, weighted averaging, PCA, etc.) is selected from a preset operation sequence library, thus obtaining the signal conditioning strategy for that monitoring group. This method ensures that the conditioning steps for each monitoring group are deterministic, consistent, and physically reasonable.

[0023] In one embodiment of the present invention, the multi-source relationship includes single-source, redundant, complementary, and correlated; the dynamics include steady-state, transient pulse, periodic, and general dynamics; the interference types include high-frequency noise, power frequency interference, pulse noise, and low noise; the type profiles from high to low include steady-state multi-source redundancy, multi-source complementarity / correlation, transient pulse master control, periodic / oscillating signal, steady-state single-source low noise, and dynamic high-noise single source. Each monitoring group performs type profile matching from high to low according to the type profile ranking, with the first match being the standard.

[0024] In this embodiment, the specific categories of the three discrimination axes and the priority order of the six types of profiles are further defined. In multi-source relationships, "single source" refers to containing only one signal; "redundancy" refers to multiple signals measuring the same physical quantity; "complementary" refers to different physical quantities jointly describing the same object (e.g., position + angle); "correlation" refers to a high linear correlation between signal waveforms. In dynamics, "steady state" refers to extremely slow changes (e.g., a change rate of less than or equal to 5% per acquisition cycle); "transient pulse" refers to the presence of spikes or steps; "periodic" refers to regular oscillations; and "general dynamics" refers to rapid and irregular changes. In interference types, "high-frequency noise," "power frequency interference," "pulse noise," and "low noise" correspond to different noise characteristics. The priority of the profile types from high to low is as follows: steady-state multi-source redundancy > multi-source complementarity / correlation > transient pulse master control > periodic / oscillating signal > steady-state single-source low noise > dynamic high-noise single source. During matching, the system sequentially determines whether the monitoring group meets the characteristics of that type. Once it does, it is classified into that type, and no further lower-priority judgments are made. The engineering advantages of this priority design are: redundancy fusion best improves reliability, complementary / correlated fusion best reduces dimensionality, transient features require special handling to avoid being smoothed out by filtering, periodic signals require frequency domain analysis, and steady-state single-source and dynamic high-noise single-source are common basic types. This rule ensures that each monitoring group has a unique and definite type profile.

[0025] In one embodiment of the present invention, for steady-state multi-source redundancy, the corresponding operation sequence is: median filtering, weighted averaging, normalization, and outputting a single value; for multi-source complementarity / correlation, the corresponding operation sequence is: lightweight pre-filtering, constructing combined features using PCA or a physical model, and outputting the first few principal components or combined indices; for transient pulse master control, the corresponding operation sequence is: median filtering, edge detection, and outputting pulse features; for periodic / oscillating signals, the corresponding operation sequence is: bandpass filtering, extracting the main frequency, amplitude, and phase, and outputting a frequency domain feature vector; for steady-state single-source low noise, the corresponding operation sequence is: moving average, normalization to [0,1], and outputting; for dynamic high-noise single source, the corresponding operation sequence is: low-pass filtering, normalization or standardization, first-order difference, and outputting.

[0026] In this embodiment, a specific treatment sequence corresponding to each type of portrait is given, illustrated with examples: Steady-state multi-source redundancy (e.g., three temperature sensors monitoring the same motor temperature): First, median filtering is used to remove outlier sensor data (such as sensors with abnormal values ​​due to faults). Then, a weighted average is calculated (the weights can be dynamically adjusted based on historical accuracy). Finally, the result is normalized (optional), and a single temperature value is output. Effect: Improves measurement reliability and suppresses random errors.

[0027] Multi-source complementary / correlated (e.g., simultaneously acquiring motor current and speed, which are highly correlated): First, perform light pre-filtering on each signal (e.g., low-pass filtering for current, moving average for speed). Then, use principal component analysis (PCA) to extract the first principal component as a comprehensive load index, or calculate power (current × voltage) based on a physical model, outputting the first 1-2 principal components or combined index. Effect: Eliminates multicollinearity and reduces feature dimensionality.

[0028] Transient pulse master control (e.g., spike signal output by collision sensor): Median filtering is used to suppress random glitches while preserving step edges; then edge detection is performed (e.g., calculating gradient threshold), the time and amplitude of pulse occurrence are extracted, and the pulse feature vector is output (note that low-pass filtering is not performed to avoid smoothing out the pulse).

[0029] For periodic / oscillatory signals (such as sinusoidal waveforms output by vibration sensors): first, a bandpass filter is used to retain components near the fundamental frequency and filter out DC and higher harmonic noise; then, a Fast Fourier Transform (FFT) or wavelet transform is performed to extract the main frequency, amplitude, and phase, and output the frequency domain feature vector.

[0030] Steady-state single-source low noise (e.g., warehouse ambient temperature signal): The residual noise is further smoothed by using a moving average, and then normalized to the [0,1] interval to output a single value.

[0031] For dynamic high-noise single sources (such as AGV real-time speed signals, which change rapidly and have Gaussian noise): first perform low-pass filtering (cutoff frequency slightly higher than the effective frequency band of the speed signal), then normalize or standardize (zero mean, unit variance), optionally calculate the first-order difference to highlight acceleration changes, and output the conditioned signal.

[0032] The above operation sequence is perfectly matched with each type of physical characteristic, ensuring the optimization and consistency of the conditioning process.

[0033] In one embodiment of the present invention, the preprocessing of the second monitoring data based on the correlation between the second monitoring data and the first monitoring data includes: Obtain the first monitoring data corresponding to the second monitoring data, and extract the baseline value of the corresponding first monitoring data; Each item in the second monitoring data is compared with the baseline value of the corresponding first monitoring data to determine the change status and obtain the processing label for each item. The second monitoring data is preprocessed according to the processing label.

[0034] In this embodiment, the core task of preprocessing is to use the first monitoring data as a baseline to filter and organize the second monitoring data, eliminating redundant data that is unchanged and has no confirmation value, thereby further reducing the input scale. The specific process is as follows: First, for each second monitoring item (such as "AGV real-time position"), find its corresponding historical value or baseline value in the first monitoring data (such as the position record at the last scheduling). Then, compare item by item: if the second data changes significantly relative to the baseline value (exceeding a preset threshold, such as a position change of more than 0.5 meters), it is marked as "changed", and the actual value or change amount is retained; if there is no significant change, it is further determined whether the item needs to be "confirmed" (such as the safety sensor needing periodic verification of normal status) or "excluded" (such as the temperature being within the allowable fluctuation range and having no impact on decision-making). For items that need to be excluded and have no change, their values ​​are replaced with 0 (i.e., change amount 0); for items that need to be confirmed, the actual value is retained. Finally, the preprocessed second monitoring data (where the excluded items have been set to 0) is sent to the subsequent monitoring group. This approach effectively reduces the interference of invalid data, while retaining key change information, making subsequent adjustments and decisions more focused.

[0035] In one embodiment of the present invention, the step of performing secondary decision-making on a plurality of pending targets based on the conditioned second monitoring data to obtain at least one execution target includes: Based on the adjusted second monitoring data, a decision matrix of size m×n is constructed, where m corresponds to m undetermined targets and n corresponds to n evaluation attributes; The decision matrix is ​​standardized to obtain the standardized matrix; Based on the degree of variation of the data under each evaluation attribute, calculate the entropy value of each evaluation attribute, and determine the objective weight of each evaluation attribute based on the entropy value; Based on the standardized matrix and the objective weights, construct a weighted standardized decision matrix and determine its positive and negative ideal solutions; Calculate the Euclidean distance between each target and the positive and negative ideal solutions; Based on the calculated Euclidean distance, the relative proximity of each target to be determined is obtained; Select at least one execution target based on the degree of relative proximity, from largest to smallest.

[0036] In this embodiment, this step specifically describes the method used for the second decision-making process—the entropy weight method combined with TOPSIS (Topology for Approximating Ideal Solutions). The aim is to objectively and robustly select the optimal execution target from the candidate set of undetermined targets obtained from the first decision. Assuming the first decision generated m=5 undetermined targets, each target has n=3 evaluation attributes (e.g., precise completion time T, precise energy consumption E, precise safety score S) under the adjusted second monitoring data. First, a decision matrix X is constructed: Because the various attributes have different units (time in seconds, energy consumption in joules, safety score 0-1), the matrix needs to be standardized to obtain a standardized matrix. A common method is vector normalization: Then, the entropy weight method is used to determine the objective weights, and the entropy value of the j-th attribute is calculated. : The higher the entropy value, the smaller the attribute variation, and the lower the weight. Weight Calculation formula: Next, a weighted standardized decision matrix is ​​constructed. ,in Determine the ideal solution A. + (Maximum values ​​of each attribute) and negative ideal solution A - (Minimum value of each attribute): Calculate the Euclidean distance between each target and the positive and negative ideal solutions: Finally, calculate the relative closeness. : The closer the value is to 1, the better the objective. All undetermined objectives are categorized as follows: Sort the data from largest to smallest and select the first one as the execution target. Alternatively, you can select the top K (e.g., backup options) as needed. This method does not require manually setting weights, is entirely data-driven, and is insensitive to outliers, making it ideal for making robust decisions from finely processed high-dimensional data.

[0037] like Figure 2 As shown, in one embodiment of the present invention, a digital twin and multi-objective optimization fusion enterprise intelligent warehouse scheduling system is proposed. The digital twin and multi-objective optimization fusion enterprise intelligent warehouse scheduling system includes: A sensing layer, which is used to acquire data through sensing elements; A data twin layer, which uses the data collected by the perception layer as input, is used for the digital simulation of real-world scenarios or processes; A model decision layer, connected to the data twin layer, is used to execute the enterprise intelligent warehouse scheduling method that integrates digital twins and multi-objective optimization as described in any one of claims 1-9; An execution layer, which is connected to the model decision layer, is used to execute the execution objectives output by the model decision layer.

[0038] In this embodiment, the system architecture organically integrates the physical world and virtual space, forming a perception-simulation-decision-execution architecture. The perception layer includes various sensors (RFID, UWB, LiDAR, current sensors, etc.) responsible for collecting first and second monitoring data. The data twin layer constructs a high-fidelity digital model based on this data, mapping warehouse layout, AGV status, inventory information, etc. in real time. The model decision layer is the core of the system, executing the intelligent warehouse scheduling method for enterprises that integrates digital twins and multi-objective optimization as provided in this embodiment of the invention. It performs deduction, optimization, and decision-making within the digital twin to generate execution objectives. The execution layer receives decision instructions and drives physical equipment such as AGVs, conveyors, and stacker cranes to complete specific tasks. Through layered design, the system utilizes the simulation capabilities of digital twins for zero-risk trial and error, and leverages multi-objective optimization algorithms to achieve intelligent scheduling. Ultimately, it achieves the best balance between computational resource consumption and decision accuracy, solving the problems of poor real-time performance and excessive overhead caused by the pursuit of holographic replication in traditional digital twin systems.

[0039] Figure 3 An internal structural diagram of a computer device in one embodiment is shown. Figure 3 As shown, the computer device includes a processor, memory, network interface, input device, and display screen connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program. When executed by the processor, this computer program enables the processor to implement the digital twin and multi-objective optimization fusion enterprise intelligent warehouse scheduling method provided in this embodiment of the invention. The internal memory may also store a computer program. When executed by the processor, this computer program enables the processor to execute the digital twin and multi-objective optimization fusion enterprise intelligent warehouse scheduling method provided in this embodiment of the invention. The display screen of the computer device can be a liquid crystal display screen or an e-ink display screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad mounted on the casing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0040] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0041] In one embodiment, a computer device is provided, the computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the following steps: Upon receiving the trigger signal, the first monitoring data is collected; Based on the first monitoring data, a target decision is made to obtain several undetermined targets; Based on the identified targets, monitoring groups are determined, and corresponding signal conditioning strategies are matched for each monitoring group. The monitoring group collects second monitoring data, preprocesses the second monitoring data based on the correlation between the second monitoring data and the first monitoring data, and conditions the preprocessed second monitoring data according to the matching signal conditioning strategy. Based on the adjusted second monitoring data, a secondary decision is made on several pending targets to obtain at least one execution target, which is then issued and executed.

[0042] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, causes the processor to perform the following steps: Upon receiving the trigger signal, the first monitoring data is collected; Based on the first monitoring data, a target decision is made to obtain several undetermined targets; Based on the identified targets, monitoring groups are determined, and corresponding signal conditioning strategies are matched for each monitoring group. The monitoring group collects second monitoring data, preprocesses the second monitoring data based on the correlation between the second monitoring data and the first monitoring data, and conditions the preprocessed second monitoring data according to the matching signal conditioning strategy. Based on the adjusted second monitoring data, a secondary decision is made on several pending targets to obtain at least one execution target, which is then issued and executed.

[0043] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated 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 steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps. Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0044] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0045] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for intelligent warehouse scheduling of enterprises by fusing digital twin and multi-objective optimization, characterized in that, The digital twin and multi-objective optimization fusion enterprise intelligent warehouse scheduling method includes: Upon receiving the trigger signal, the first monitoring data is collected; Based on the first monitoring data, a target decision is made to obtain several undetermined targets; Based on the identified targets, monitoring groups are determined, and corresponding signal conditioning strategies are matched for each monitoring group. The monitoring group collects second monitoring data, preprocesses the second monitoring data based on the correlation between the second monitoring data and the first monitoring data, and conditions the preprocessed second monitoring data according to the matching signal conditioning strategy. Based on the adjusted second monitoring data, a secondary decision is made on several pending targets to obtain at least one execution target, which is then issued and executed.

2. The digital twin and multi-objective optimization integrated enterprise intelligent warehouse scheduling method of claim 1, wherein, Upon receiving the trigger signal, the first monitoring data is collected, including: Upon receiving a trigger signal, determine the production stage to which the trigger signal belongs; Identify the product batches directly and indirectly related to the production process to which the trigger signal belongs; Obtain the set indicators for the identified product batch; The monitoring items corresponding to each set indicator are determined, and the first monitoring data is obtained by collecting data from the determined monitoring items.

3. The digital twin and multi-objective optimization integrated enterprise intelligent warehouse scheduling method of claim 1, wherein, The process of making a target decision based on the first monitoring data yields several undetermined targets, including: Based on the first monitoring data, the three key elements of the multi-objective optimization problem are determined: decision variables, objective function, and constraints. The initial parent population is obtained by initializing candidate solutions based on the determined three elements; A non-dominated sorting method is used to determine the Pareto front level of all candidate solutions in the initial parent population based on the value of each candidate solution on each objective function, and at the same time calculate the crowding distance of each candidate solution. Based on the calculated Pareto front level and crowding distance, genetic operations are performed according to the set rules to generate offspring populations. The generated offspring populations are then merged with the initial parent populations, and non-dominated sorting and crowding distance calculations are performed again. An elite retention strategy is used to select a new generation of populations. This process is repeated until the termination condition is met to obtain several undetermined targets.

4. The digital twin and multi-objective optimization integrated enterprise intelligent warehouse scheduling method of claim 1, wherein, The process of determining the monitoring group based on a number of identified undetermined targets includes: Based on the identified several undetermined targets, obtain all the original monitoring items that are physically or logically associated with each undetermined target. Based on all the original monitoring items obtained, similar items are merged and duplicate items are removed to obtain the complete set of monitoring items; Based on the physical correlation and primary / secondary relationship between the monitoring items in the complete set of monitoring items, several monitoring groups are divided. Each monitoring group contains at least one monitoring item, and the monitoring items in the same group are coupled.

5. The enterprise intelligent warehouse scheduling method integrating digital twin and multi-objective optimization according to claim 1, characterized in that, The matching of corresponding signal conditioning strategies for each monitoring group includes: Acquire all raw data streams for each monitoring group; For each monitoring group, determine the multi-source relationship of the raw data stream, perform dynamic analysis on the raw data stream, and determine the type of interference in the raw data stream; By integrating the multi-source relationships, dynamic analysis results, and interference types obtained in the previous step, a unique type profile is obtained for each monitoring group; Based on the obtained type profile, the corresponding preset operation sequence is determined to obtain the signal conditioning strategy for the monitoring group.

6. The enterprise intelligent warehouse scheduling method integrating digital twin and multi-objective optimization according to claim 5, characterized in that: The multi-source relationships include single-source, redundant, complementary, and correlated relationships; The dynamics include steady state, transient pulse, periodicity, and general dynamics; The types of interference include high-frequency noise, power frequency interference, impulse noise, and low noise; The type profiles, from high to low, include steady-state multi-source redundancy, multi-source complementarity / correlation, transient pulse master control, periodic / oscillating signals, steady-state single-source low noise, and dynamic high-noise single source. Each monitoring group matches the type profiles from high to low according to their sorting, with the first match being the most accurate.

7. The enterprise intelligent warehouse scheduling method integrating digital twin and multi-objective optimization according to claim 6, characterized in that: For steady-state multi-source redundancy, the corresponding operation sequence is: median filtering, weighted averaging, normalization, and outputting a single value. For multi-source complementarity / correlation, the corresponding operation sequence is: lightweight pre-filtering, PCA or physical model to construct combined features, and output the first few principal components or combined indices; For transient pulse master control, the corresponding operation sequence is: median filtering, edge detection, and output pulse characteristics; For periodic / oscillatory signals, the corresponding operation sequence is: bandpass filtering, extracting the main frequency, amplitude, and phase, and outputting the frequency domain feature vector; For steady-state single-source low noise, the corresponding operation sequence is: moving average, normalize to [0,1] and output; For a dynamic high-noise single source, the corresponding operation sequence is: low-pass filtering, normalization or standardization, first-order difference and output.

8. The enterprise intelligent warehouse scheduling method integrating digital twin and multi-objective optimization according to claim 1, characterized in that, The preprocessing of the second monitoring data based on the correlation between the second monitoring data and the first monitoring data includes: Obtain the first monitoring data corresponding to the second monitoring data, and extract the baseline value of the corresponding first monitoring data; Each item in the second monitoring data is compared with the baseline value of the corresponding first monitoring data to determine the change status and obtain the processing label for each item. The second monitoring data is preprocessed according to the processing label.

9. The enterprise intelligent warehouse scheduling method integrating digital twin and multi-objective optimization according to claim 1, characterized in that, The process of making secondary decisions on several pending targets based on the adjusted second monitoring data to obtain at least one execution target includes: Based on the adjusted second monitoring data, a decision matrix of size m×n is constructed, where m corresponds to m undetermined targets and n corresponds to n evaluation attributes; The decision matrix is ​​standardized to obtain the standardized matrix; Based on the degree of variation of the data under each evaluation attribute, calculate the entropy value of each evaluation attribute, and determine the objective weight of each evaluation attribute based on the entropy value; Based on the standardized matrix and the objective weights, construct a weighted standardized decision matrix and determine its positive and negative ideal solutions; Calculate the Euclidean distance between each target and the positive and negative ideal solutions; Based on the calculated Euclidean distance, the relative proximity of each target to be determined is obtained; Select at least one execution target based on the degree of relative proximity, from largest to smallest.

10. A digital twin and multi-objective optimization integrated enterprise intelligent warehouse scheduling system, characterized in that: The digital twin and multi-objective optimization integrated enterprise intelligent warehouse scheduling system includes: A sensing layer, which is used to acquire data through sensing elements; A data twin layer, which uses the data collected by the perception layer as input, is used for the digital simulation of real-world scenarios or processes; A model decision layer, connected to the data twin layer, is used to execute the enterprise intelligent warehouse scheduling method that integrates digital twins and multi-objective optimization as described in any one of claims 1-9; An execution layer, which is connected to the model decision layer, is used to execute the execution objectives output by the model decision layer.

Images