Collaborative decision determination method and device for equipment, equipment and medium

Abstract

The invention provides a collaborative decision determination method and device for equipment, equipment and a medium, and the method comprises the steps: carrying out the feature enhancement and weight fusion of a plurality of initial feature vectors of the equipment, and obtaining target fusion feature data; inputting the target fusion feature data into a pre-constructed disturbance recognition model, and determining a target disturbance type corresponding to the equipment from a plurality of preset disturbance types and a disturbance intensity value corresponding to the target disturbance type; and determining a to-be-optimized target function and a target constraint condition based on the disturbance type and the disturbance intensity numerical value, and determining a collaborative decision scheme for the equipment by using the to-be-optimized target function and the target constraint condition. Through the method and the device, the accuracy and the timeliness of equipment decision making in a real dynamic environment are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent decision-making technology for equipment and materials, and in particular to a collaborative decision-making method, device, equipment and medium for equipment and materials. Background Technology

[0002] With the increasing demand for intelligent management of equipment and materials, existing decision-making systems are facing the dual challenges of multi-source data fusion and complex disturbance response. In application scenarios such as military logistics and emergency support, the decision-making process for equipment and materials involves multi-source heterogeneous information such as material status monitoring data, environmental perception data, and business system data, exhibiting significant multimodal characteristics.

[0003] Current mainstream technical solutions mostly employ feature concatenation or linear weighting methods based on fixed weights for data fusion processing. However, these methods have significant limitations: on the one hand, they fail to fully consider the inherent feature differences between different modalities; on the other hand, they lack an effective mechanism for evaluating the dynamic importance of data, leading to problems such as information loss or feature conflicts in the fused feature space. Meanwhile, decision-making systems also need to address the challenges of various dynamic perturbations. Traditional robust optimization methods typically build models based on pre-set single perturbation assumptions. This approach struggles to accurately identify the type characteristics and intensity distribution of perturbation events in real-world application scenarios, resulting in significantly insufficient adaptability of the generated decision solutions in complex real-world environments. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method, apparatus, equipment, and medium for collaborative decision-making of equipment and materials, which realizes accurate characterization of multi-source heterogeneous data and robust decision-making under multiple types of disturbances. It effectively overcomes the shortcomings of existing technologies, such as large information loss in multimodal fusion and model construction based on only a single disturbance assumption, and improves the accuracy and timeliness of equipment and materials decision-making in real dynamic environments.

[0005] In a first aspect, embodiments of this application provide a collaborative decision-making method for equipment and materials, the collaborative decision-making method comprising: Acquire initial data of equipment and materials under multiple modal types, and extract features from multiple initial data to obtain multiple initial feature vectors; The initial feature vectors are enhanced and weighted fused to obtain the target fused feature data. The target fusion feature data is input into a pre-built disturbance identification model to determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from a variety of preset disturbance types. Based on the disturbance type and the disturbance intensity value, the objective function to be optimized and the target constraints are determined, and the collaborative decision-making scheme for the equipment is determined using the objective function to be optimized and the target constraints.

[0006] Furthermore, the step of performing feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fused feature data includes: Calculate the modal feature quality assessment factor for each initial feature vector based on the signal-to-noise ratio corresponding to each initial feature vector; The feature enhancement vector corresponding to each initial feature vector is determined based on the modal feature quality assessment factor and the feature enhancement formula corresponding to each initial feature vector. The feature enhancement vectors are weighted and fused using the feature weights corresponding to each modality type to obtain the target fused feature data; wherein the feature weights are dynamically updated based on feature relevance and decision contribution.

[0007] Furthermore, the feature weights corresponding to each modality type are calculated through the following steps: For each modality type, the modality feature correlation between that modality type and other modality types is calculated, and the feature correlation gradient is determined based on the modality feature correlation. Calculate the feature importance score corresponding to the modality type, and determine the decision sharing gradient based on the feature importance score; The feature weights corresponding to the modality type are determined based on the feature correlation gradient and the decision sharing gradient.

[0008] Furthermore, after determining the collaborative decision-making scheme, the collaborative decision-making determination method further includes: The decision execution data of the collaborative decision-making scheme during actual execution is obtained, the decision error is calculated based on the decision execution data, and the feature weights of each modality are adjusted based on the decision error.

[0009] Furthermore, the disturbance recognition model is trained through the following steps: Obtain a training dataset; wherein the training data includes fusion feature sample data corresponding to sample data under multiple modal types and perturbation sample labels corresponding to the sample data, the perturbation sample labels including perturbation type sample labels and perturbation intensity sample values; The fused feature sample data is input into the original perturbation recognition model to determine the perturbation prediction label corresponding to the sample data; The loss function is determined based on the perturbation sample label and the perturbation prediction label, and the original perturbation recognition model is iteratively trained based on the loss function until the preset training completion condition is met, thus obtaining the trained perturbation recognition model.

[0010] Secondly, embodiments of this application also provide a collaborative decision-making device for equipment and materials, the collaborative decision-making device comprising: The feature vector generation module is used to acquire initial data of equipment and materials under multiple modal types, and to extract features from multiple initial data to obtain multiple initial feature vectors; The feature data fusion determination module is used to perform feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fused feature data. The disturbance type identification module is used to input the target fusion feature data into a pre-built disturbance identification model, and determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from a variety of preset disturbance types; The decision scheme generation module is used to determine the objective function to be optimized and the objective constraints based on the disturbance type and the disturbance intensity value, and to determine the collaborative decision scheme for the equipment using the objective function to be optimized and the objective constraints.

[0011] Furthermore, when the fusion feature data determination module performs feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fusion feature data, the fusion feature data determination module is also used for: Calculate the modal feature quality assessment factor for each initial feature vector based on the signal-to-noise ratio corresponding to each initial feature vector; The feature enhancement vector corresponding to each initial feature vector is determined based on the modal feature quality assessment factor and the feature enhancement formula corresponding to each initial feature vector. The feature enhancement vectors are weighted and fused using the feature weights corresponding to each modality type to obtain the target fused feature data; wherein the feature weights are dynamically updated based on feature relevance and decision contribution.

[0012] Furthermore, the fusion feature data determination module is also used to calculate the feature weights corresponding to each modality type through the following steps: For each modality type, the modality feature correlation between that modality type and other modality types is calculated, and the feature correlation gradient is determined based on the modality feature correlation. Calculate the feature importance score corresponding to the modality type, and determine the decision sharing gradient based on the feature importance score; The feature weights corresponding to the modality type are determined based on the feature correlation gradient and the decision sharing gradient.

[0013] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the collaborative decision-making method for equipment and materials described above are performed.

[0014] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the cooperative decision-making method for equipment described above.

[0015] This application provides a method, apparatus, device, and medium for collaborative decision-making of equipment and materials. First, initial data of the equipment and materials under multiple modal types are acquired, and features are extracted from these initial data to obtain multiple initial feature vectors. Then, feature enhancement and weighted fusion are performed on the multiple initial feature vectors to obtain target fused feature data. The target fused feature data is input into a pre-constructed disturbance identification model to determine the target disturbance type corresponding to the equipment and materials from multiple preset disturbance types, as well as the disturbance intensity value corresponding to the target disturbance type. Finally, based on the disturbance type and the disturbance intensity value, an objective function to be optimized and target constraints are determined, and a collaborative decision-making scheme for the equipment and materials is determined using the objective function to be optimized and the target constraints.

[0016] This application acquires initial data of multiple modalities and extracts features, then combines feature enhancement and dynamic weighted fusion to generate target fusion feature data. A disturbance identification model accurately determines the disturbance type and corresponding intensity, and based on this, adaptively constructs an optimized objective function and constraints to generate a collaborative decision-making scheme. This achieves accurate representation of multi-source heterogeneous data and robust decision-making under multiple types of disturbances without relying on external prior assumptions or manual rules. It effectively overcomes the shortcomings of existing technologies, such as significant information loss in multi-modal fusion and model construction based solely on a single disturbance assumption, thus improving the accuracy and timeliness of equipment decision-making in real dynamic environments.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a collaborative decision-making method for equipment and materials provided in an embodiment of this application; Figure 2 A schematic diagram of the structure of a collaborative decision-making device for equipment provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0021] First, the applicable scenarios for this application will be introduced. This application can be applied to the field of intelligent decision-making technology for equipment and materials.

[0022] With the increasing demand for intelligent management of equipment and materials, existing decision-making systems are facing the dual challenges of multi-source data fusion and complex disturbance response. In application scenarios such as military logistics and emergency support, the decision-making process for equipment and materials involves multi-source heterogeneous information such as material status monitoring data, environmental perception data, and business system data, exhibiting significant multimodal characteristics.

[0023] Research has revealed that current mainstream technical solutions mostly employ feature concatenation or linear weighting methods based on fixed weights for data fusion processing. However, these methods have significant limitations: on the one hand, they fail to fully consider the inherent feature differences between different modalities; on the other hand, they lack an effective mechanism for evaluating the dynamic importance of data, leading to problems such as information loss or feature conflicts in the fused feature space. Meanwhile, decision-making systems also need to address the challenges of various dynamic perturbations. Traditional robust optimization methods typically build models based on pre-set assumptions about a single perturbation. This approach struggles to accurately identify the type characteristics and intensity distribution of perturbation events in real-world application scenarios, resulting in significantly insufficient adaptability of the generated decision solutions in complex real-world environments.

[0024] Based on this, embodiments of this application provide a collaborative decision-making method for equipment and materials to reduce packet loss rate and data transmission latency.

[0025] Please see Figure 1 , Figure 1 This is a flowchart illustrating a collaborative decision-making method for equipment and materials provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the collaborative decision-making method includes: S101: Acquire initial data of equipment and materials under multiple modal types, and extract features from multiple initial data to obtain multiple initial feature vectors.

[0026] Regarding step S101 above, in specific implementation, initial data of equipment and materials under multiple modalities are acquired, and features are extracted from multiple initial data to obtain multiple initial feature vectors. Here, the multiple modalities include four types of heterogeneous data: structured data, text data, time-series data, and image data. Multiple protocols such as HTTP / HTTPS, MQTT, and CoAP are supported, and interfaces are established with sensors, RFID readers, business systems (warehousing, scheduling), and external data platforms (logistics, policy) to achieve real-time acquisition and synchronization of the four types of heterogeneous data.

[0027] Specifically, regarding structured data Differences in the dimensions of different features can adversely affect the computational accuracy of subsequent algorithms. Standardized structured data is obtained through the standardized statistic Z-score. ,in, For the mean, is the standard deviation. Where, The original structured data matrix has rows representing samples (equipment / tasks) and columns representing structured features (inventory, cost, loss rate, etc.). For the sample size, This represents the feature dimension of structured data. For text data, it is obtained after BERT encoding. Core features were selected using TF-IDF. Among them, This refers to the feature matrix of text data encoded using BERT (such as the semantic features of maintenance logs or task instructions). For time-series data... Temporal features were extracted using CNN+LSTM. For image data, visual features are extracted using ResNet50. .

[0028] S102, perform feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fused feature data.

[0029] In specific implementation of step S102, the multiple initial feature vectors determined in step S101 are subjected to feature enhancement and weighted fusion to obtain target fused feature data.

[0030] As an optional embodiment, regarding step S102 above, the step of performing feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fused feature data includes: Step 1021: Calculate the modal feature quality evaluation factor corresponding to each initial feature vector based on the signal-to-noise ratio corresponding to each initial feature vector.

[0031] Here, the modal feature quality assessment factor is used to quantify the quality proportion of this type of modal feature in the multimodal feature set.

[0032] Regarding step 1021 above, in specific implementation, the modal feature quality assessment factor corresponding to each initial feature vector is calculated based on the signal-to-noise ratio corresponding to each initial feature vector. Specifically, the modal feature quality assessment factor is calculated using the following formula:

[0033] in, Indicates the first The modal feature quality assessment factor for class modal features has a value range of (0,1). Indicates the first The initial feature vector of the modality feature; The signal-to-noise ratio (SNR) is used to characterize feature quality.

[0034] Step 1022: Determine the feature enhancement vector corresponding to each initial feature vector based on the modal feature quality assessment factor and the feature enhancement formula corresponding to each initial feature vector.

[0035] Regarding step 1022 above, in specific implementation, after the modal feature quality assessment factor corresponding to each initial feature vector is determined, the feature enhancement vector corresponding to each initial feature vector is determined based on the modal feature quality assessment factor corresponding to each initial feature vector and the feature enhancement formula. Specifically, the feature enhancement formula is expressed by the following formula:

[0036] in, Indicates the first Feature enhancement vectors for modal features Represents the enhancement weight matrix. This indicates the bias term.

[0037] Step 1023: Use the feature weights corresponding to each modality type to perform weighted fusion of multiple feature enhancement vectors to obtain the target fused feature data.

[0038] Regarding step 1023 above, in specific implementation, multiple feature enhancement vectors are weighted and fused using the feature weights corresponding to each modality type to obtain the target fused feature data. Specifically, the target fused feature data... Calculated using the following formula:

[0039] in, Represents the feature weights of structured data. Represents the feature weights of text data. The feature weights represent the time series data. Represents the feature weights of image data. .

[0040] Here, as an optional embodiment, the feature weights for the four modal types can be pre-defined, for example, the weights can be initialized. The initial weights for the four modal types are equal.

[0041] As an alternative embodiment, the feature weights for each modality can also be dynamically updated based on feature relevance and decision contribution.

[0042] When dynamically adjusting feature weights, the feature weights for each modality type are calculated using the following steps: A: For each modality type, calculate the modality feature correlation between that modality type and other modality types, and determine the feature correlation gradient based on the modality feature correlation.

[0043] Here, the feature correlation gradient is used to characterize the degree to which the cooperative correlation strength between a certain type of feature and other types of features in a multimodal fusion system deviates from the overall average level. Specifically, the feature correlation gradient is calculated based on mutual information.

[0044] Regarding step A above, in specific implementation, for each modality type, the modality feature correlation between this modality type and other modality types is first calculated, and then the feature correlation gradient is determined based on the modality feature correlation. As an optional embodiment, the intrinsic correlation between any two modality types is first quantified, the initial feature vector corresponding to each modality type is extracted, and the modality feature correlation between the initial feature vector of this modality type and the initial feature vectors of other modality types is calculated. Then, for a modality type of interest, the average of the modality feature correlations calculated between it and all other modality types is taken as the modality correlation reference benchmark. The modality feature correlation value between this modality type and each other modality type is subtracted from the above average correlation value, and then the average of all differences is calculated to obtain the feature correlation gradient corresponding to this modality type.

[0045] B: Calculate the feature importance score corresponding to the modality type, and determine the decision sharing gradient based on the feature importance score.

[0046] Here, the decision sharing gradient is used to reflect the actual influence of a certain type of modality on the quality of the final collaborative decision-making scheme.

[0047] Regarding step B above, in specific implementation, the feature importance score corresponding to the modality type is calculated, and the decision sharing gradient of the modality type is determined based on the feature importance score. As an optional embodiment, the contribution of the modality type to the final decision is first calculated, for example, by performing performance prediction on the modality type individually, or by obtaining the feature importance score of the modality type using a feature importance method (such as SHAP). Then, the average of the feature importance scores of multiple modality types is taken, and the difference between the feature importance score of the modality type and the average is calculated; this difference is the decision sharing gradient of the modality type.

[0048] C: Determine the feature weights corresponding to the modality type based on the feature correlation gradient and the decision sharing gradient.

[0049] Regarding step C above, in practice, the feature weights corresponding to this modality type are determined based on the feature relevance gradient and the decision sharing gradient. Specifically, the feature weights are calculated using the following formula:

[0050] in, For feature correlation gradient, For the gradient of decision contribution, , To adjust the coefficient, The feature weights before adjustment. These are the adjusted feature weights.

[0051] Thus, based on steps 1021-1023 above, a modal feature quality assessment factor and a dynamic weight update mechanism are introduced. Through intramodal feature enhancement and intermodal dynamic weighting, adaptive optimization of fused features is achieved, solving the problem of feature representation failure in traditional shallow fusion.

[0052] S103, the target fusion feature data is input into a pre-built disturbance identification model to determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from a variety of preset disturbance types.

[0053] Regarding step S103 above, in specific implementation, the target feature data is fused. The data is input into a pre-built disturbance recognition model to determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from a variety of preset disturbance types. Here, the disturbance recognition model can employ a lightweight CNN recognizer; this application does not specifically limit its use. According to the embodiments provided in this application, the various preset disturbance types include... Disturbance intensity value ,in, To ensure no disturbance, This is an extreme disturbance.

[0054] As an optional embodiment, the perturbation recognition model is trained through the following steps: I: Obtain the training dataset.

[0055] Regarding step I above, in specific implementation, a training dataset is obtained. Specifically, the training data includes fused feature sample data corresponding to sample data under multiple modalities and perturbation sample labels corresponding to the sample data. The perturbation sample labels include perturbation type sample labels and perturbation intensity sample values. The method for generating fused feature sample data from the sample data is the same as the method described in steps S101-S102 above, and achieves the same technical effect, so it will not be repeated here. The perturbation type label can be any one or more of a variety of preset perturbation types.

[0056] II: Input the fused feature sample data into the original perturbation recognition model to determine the perturbation prediction label corresponding to the sample data.

[0057] Regarding step II above, in specific implementation, the fused feature sample data is input into the original disturbance recognition model to determine the disturbance prediction label corresponding to the sample data. Here, the disturbance prediction label includes a disturbance type prediction label and a disturbance intensity prediction value.

[0058] III: Determine the loss function based on the perturbation sample label and the perturbation prediction label, and iteratively train the original perturbation recognition model based on the loss function until the preset training completion condition is met, thus obtaining the trained perturbation recognition model.

[0059] Among them, the preset training completion conditions include the loss value of the loss function being less than a preset threshold or the number of iterations reaching a preset number.

[0060] Here, the loss function can perturb the quantitative difference between the sample label and the perturbation prediction label to guide the parameter optimization of the original perturbation recognition model during iterative training.

[0061] Regarding step III above, in specific implementation, a loss function is obtained based on the perturbation sample label and the perturbation prediction label to measure the difference between the two. The parameters of the original perturbation recognition model are adjusted using the loss function during training, and iterative training is performed until the preset training completion conditions are met, resulting in a trained perturbation recognition model.

[0062] S104. Based on the disturbance type and the disturbance intensity value, determine the objective function to be optimized and the target constraint conditions, and use the objective function to be optimized and the target constraint conditions to determine the collaborative decision-making scheme for the equipment.

[0063] Regarding step S104 above, in specific implementation, the objective function to be optimized and the objective constraints are determined based on the disturbance type and disturbance intensity values ​​determined in step S103 above, and the collaborative decision-making scheme for the equipment is determined using the objective function to be optimized and the objective constraints.

[0064] According to the embodiments provided in this application, a perturbation-type adaptive robust optimization algorithm is used when determining a collaborative decision-making scheme.

[0065] Specifically, the objective function to be optimized can be expressed by the following formula:

[0066] in, The total cost of inventory (holding + stockouts + losses). For scheduling costs, For the cost of disturbance loss, To improve equipment compatibility, The weight coefficients of the objective function (corresponding to respectively) , , , ).

[0067] Based on the perturbation characteristics, a dynamic weight mapping rule is customized. Specifically, when... , , , , ;when , , , , The same principle applies to the weight mapping rules for other perturbation types, ensuring that the sum of the weights is 1.

[0068] The constraints specifically include: Demand Constraints ,in, This is the demand disturbance coefficient. hour ,the remaining hour .

[0069] Inventory constraints ,in, This is the inventory disturbance coefficient. hour ,the remaining hour .

[0070] Scheduling constraints ,in, For scheduling disturbance coefficient, hour ,the remaining hour .

[0071] Once the objective function to be optimized and the objective constraints are determined, the improved alternating direction multiplier method (ADMM) is used to iterate until convergence, and the decision scheme can be output.

[0072] Here, an adaptive penalty factor is introduced into the improved ADMM algorithm. ,in, For the iterative residual, This is for adjusting the coefficient.

[0073] Thus, according to step S104 above, the output of the modality-aware dynamic weighted fusion algorithm is... It is directly used as the core input of the perturbation classification self-adaptive robust optimization algorithm, avoiding feature propagation loss, realizing seamless connection from data fusion to perturbation recognition to robust optimization, and realizing positive input linkage.

[0074] As an optional embodiment, after determining the collaborative decision-making scheme, the collaborative decision-making determination method provided in this application further includes: The decision execution data of the collaborative decision-making scheme during actual execution is obtained, the decision error is calculated based on the decision execution data, and the feature weights of each modality are adjusted based on the decision error.

[0075] Regarding the above steps, in practical implementation, after the collaborative decision-making scheme is determined, it is transformed into standardized decision instructions, such as inventory replenishment, scheduling execution, and maintenance alerts. Decision execution data is collected during the actual execution process, and the decision error is calculated based on this data. Specifically, the decision error is calculated using the following formula:

[0076] in, Let N represent the decision error, and N represent the types of decision variables. This represents the actual decision execution data for the i-th type of decision variable. This represents the model's predicted output value for the i-th type of decision variable.

[0077] Then, the decision error is fed back to the modality-aware fusion algorithm, adjusting the feature weights of each modality type based on the decision error. Specifically, the inter-modality weights are dynamically adjusted using the following formula:

[0078] in, The weights before adjustment The adjusted weights, This is for feedback on the learning rate.

[0079] In this way, based on the above steps, a dynamic closed loop of "feature fusion - robust optimization - error feedback - weight adjustment" is established, realizing deep synergy between the two core algorithms, improving decision-making accuracy, and solving the problem of algorithm silos.

[0080] Furthermore, the implementation process of this application is explained in detail below with examples of specific implementation scenarios: Taking emergency rescue equipment and material decision-making as an example, it involves rescue robots, protective equipment, communication equipment, and other materials. It requires processing sensor time-series data (operating parameters), structured data (inventory quantity, loss rate), text data (maintenance logs, task instructions), and image data (material appearance damage), as well as addressing traffic control during flood rescue operations. Disturbance scenario.

[0081] 1) Data Acquisition and Preprocessing Four types of heterogeneous data were collected through a multi-source data access module, and after cleaning by a preprocessing module, standardized initial features were obtained. , , , .

[0082] 2) Modality-aware dynamic weighted fusion Calculate the modal characteristic quality assessment factor , , , The dynamic weights W = [0.22, 0.18, 0.38, 0.22] are used to output the fused features. .

[0083] 3) Perturbation fractal adaptive robust optimization based on Identify disturbance types (Traffic control), Disturbance intensity ; , , , Construct the objective function Constraint robustness, scheduling constraints The improved ADMM algorithm converged after 15 iterations, outputting a decision scheme: replenishing 7 rescue robots, scheduling routes to avoid controlled road sections, and a transportation time of 6.2 hours. 4) Closed-loop feedback optimization Data collection and execution, decision-making error The feedback is sent to the modal perception fusion algorithm to adjust the inter-modal weights. [0.22,0.18,0.40,0.20], optimize the fusion effect in the next round.

[0084] The collaborative decision-making method for equipment provided in this application first acquires initial data of the equipment under multiple modal types and extracts features from the initial data to obtain multiple initial feature vectors; then, it performs feature enhancement and weighted fusion on the multiple initial feature vectors to obtain target fused feature data; the target fused feature data is input into a pre-constructed disturbance recognition model to determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from multiple preset disturbance types; finally, based on the disturbance type and the disturbance intensity value, it determines the objective function to be optimized and the target constraint conditions, and uses the objective function to be optimized and the target constraint conditions to determine a collaborative decision-making scheme for the equipment.

[0085] This application acquires initial data of multiple modalities and extracts features, then combines feature enhancement and dynamic weighted fusion to generate target fusion feature data. A disturbance identification model accurately determines the disturbance type and corresponding intensity, and based on this, adaptively constructs an optimized objective function and constraints to generate a collaborative decision-making scheme. This achieves accurate representation of multi-source heterogeneous data and robust decision-making under multiple types of disturbances without relying on external prior assumptions or manual rules. It effectively overcomes the shortcomings of existing technologies, such as significant information loss in multi-modal fusion and model construction based solely on a single disturbance assumption, thus improving the accuracy and timeliness of equipment decision-making in real dynamic environments.

[0086] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a collaborative decision-making device for equipment provided in an embodiment of this application. Figure 2 As shown, the collaborative decision-making device 200 includes: The feature vector generation module 201 is used to acquire initial data of equipment and materials under multiple modal types, and to extract features from multiple initial data to obtain multiple initial feature vectors. The feature data fusion determination module 202 is used to perform feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fused feature data. The disturbance type identification module 203 is used to input the target fusion feature data into a pre-built disturbance identification model, and determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from a variety of preset disturbance types; The decision scheme generation module 204 is used to determine the objective function to be optimized and the objective constraints based on the disturbance type and the disturbance intensity value, and to determine the collaborative decision scheme for the equipment using the objective function to be optimized and the objective constraints.

[0087] Furthermore, when the fusion feature data determination module 202 is used to perform feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fusion feature data, the fusion feature data determination module 202 is also used to: Calculate the modal feature quality assessment factor for each initial feature vector based on the signal-to-noise ratio corresponding to each initial feature vector; The feature enhancement vector corresponding to each initial feature vector is determined based on the modal feature quality assessment factor and the feature enhancement formula corresponding to each initial feature vector. The feature enhancement vectors are weighted and fused using the feature weights corresponding to each modality type to obtain the target fused feature data; wherein the feature weights are dynamically updated based on feature relevance and decision contribution.

[0088] Furthermore, the fusion feature data determination module 202 is also used to calculate the feature weights corresponding to each modality type through the following steps: For each modality type, the modality feature correlation between that modality type and other modality types is calculated, and the feature correlation gradient is determined based on the modality feature correlation. Calculate the feature importance score corresponding to the modality type, and determine the decision sharing gradient based on the feature importance score; The feature weights corresponding to the modality type are determined based on the feature correlation gradient and the decision sharing gradient.

[0089] Furthermore, the collaborative decision-making device 200 also includes a weight adjustment module, which, after determining the collaborative decision-making scheme, is used to: The decision execution data of the collaborative decision-making scheme during actual execution is obtained, the decision error is calculated based on the decision execution data, and the feature weights of each modality are adjusted based on the decision error.

[0090] Furthermore, the collaborative decision-making device 200 also includes a model training module, which is used to train the disturbance recognition model through the following steps: Obtain a training dataset; wherein the training data includes fusion feature sample data corresponding to sample data under multiple modal types and perturbation sample labels corresponding to the sample data, the perturbation sample labels including perturbation type sample labels and perturbation intensity sample values; The fused feature sample data is input into the original perturbation recognition model to determine the perturbation prediction label corresponding to the sample data; The loss function is determined based on the perturbation sample label and the perturbation prediction label, and the original perturbation recognition model is iteratively trained based on the loss function until the preset training completion condition is met, thus obtaining the trained perturbation recognition model.

[0091] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 300 includes a processor 310, a memory 320, and a bus 330.

[0092] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 is running, the processor 310 and the memory 320 communicate via the bus 330. When the machine-readable instructions are executed by the processor 310, they can perform the operations described above. Figure 1 The steps of the collaborative decision-making method for equipment and materials in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0093] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 The steps of the collaborative decision-making method for equipment and materials in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0094] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0096] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0097] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0098] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, 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 a portion 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 computer 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.

[0099] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, 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, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A collaborative decision-making method for equipment and materials, characterized in that, The collaborative decision-making method includes: Acquire initial data of equipment and materials under multiple modal types, and extract features from multiple initial data to obtain multiple initial feature vectors; The initial feature vectors are enhanced and weighted fused to obtain the target fused feature data. The target fusion feature data is input into a pre-built disturbance identification model to determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from a variety of preset disturbance types. Based on the disturbance type and the disturbance intensity value, the objective function to be optimized and the constraints are determined, and the collaborative decision-making scheme for the equipment is determined using the objective function to be optimized and the constraints.

2. The collaborative decision-making method according to claim 1, characterized in that, The step of performing feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fused feature data includes: Calculate the modal feature quality assessment factor for each initial feature vector based on the signal-to-noise ratio corresponding to each initial feature vector; The feature enhancement vector corresponding to each initial feature vector is determined based on the modal feature quality assessment factor and the feature enhancement formula corresponding to each initial feature vector. The feature enhancement vectors are weighted and fused using the feature weights corresponding to each modality type to obtain the target fused feature data; wherein the feature weights are dynamically updated based on feature relevance and decision contribution.

3. The collaborative decision-making method according to claim 2, characterized in that, The feature weights for each modality type are calculated using the following steps: For each modality type, the modality feature correlation between that modality type and other modality types is calculated, and the feature correlation gradient is determined based on the modality feature correlation. Calculate the feature importance score corresponding to the modality type, and determine the decision sharing gradient based on the feature importance score; The feature weights corresponding to the modality type are determined based on the feature correlation gradient and the decision sharing gradient.

4. The collaborative decision-making method according to claim 2, characterized in that, After determining the collaborative decision-making scheme, the collaborative decision-making determination method further includes: The decision execution data of the collaborative decision-making scheme during actual execution is obtained, the decision error is calculated based on the decision execution data, and the feature weights of each modality are adjusted based on the decision error.

5. The collaborative decision-making method according to claim 1, characterized in that, The perturbation recognition model is trained using the following steps: Obtain a training dataset; wherein the training data includes fused feature sample data corresponding to sample data under multiple modal types and perturbation sample labels corresponding to the sample data, the perturbation sample labels including perturbation type labels and perturbation intensity sample values; The fused feature sample data is input into the original perturbation recognition model to determine the perturbation prediction label corresponding to the sample data; The loss function is determined based on the perturbation sample label and the perturbation prediction label, and the original perturbation recognition model is iteratively trained based on the loss function until the preset training completion condition is met, thus obtaining the trained perturbation recognition model.

6. A collaborative decision-making device for equipment and materials, characterized in that, The collaborative decision-making device includes: The feature vector generation module is used to acquire initial data of equipment and materials under multiple modal types, and to extract features from multiple initial data to obtain multiple initial feature vectors; The feature data fusion determination module is used to perform feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fused feature data. The disturbance type identification module is used to input the target fusion feature data into a pre-built disturbance identification model, and determine the target disturbance type corresponding to the equipment and the disturbance intensity value corresponding to the target disturbance type from a variety of preset disturbance types; The decision scheme generation module is used to determine the objective function to be optimized and the constraints based on the disturbance type and the disturbance intensity value, and to determine the collaborative decision scheme for the equipment using the objective function to be optimized and the constraints.

7. The collaborative decision-making apparatus according to claim 6, characterized in that, When the fusion feature data determination module is used to perform feature enhancement and weighted fusion on multiple initial feature vectors to obtain target fusion feature data, the fusion feature data determination module is further used to: Calculate the modal feature quality assessment factor for each initial feature vector based on the signal-to-noise ratio corresponding to each initial feature vector; The feature enhancement vector corresponding to each initial feature vector is determined based on the modal feature quality assessment factor and the feature enhancement formula corresponding to each initial feature vector. The feature enhancement vectors are weighted and fused using the feature weights corresponding to each modality type to obtain the target fused feature data; wherein the feature weights are dynamically updated based on feature relevance and decision contribution.

8. The collaborative decision-making apparatus according to claim 7, characterized in that, The fusion feature data determination module is also used to calculate the feature weights corresponding to each modality type through the following steps: For each modality type, the modality feature correlation between that modality type and other modality types is calculated, and the feature correlation gradient is determined based on the modality feature correlation. Calculate the feature importance score corresponding to the modality type, and determine the decision sharing gradient based on the feature importance score; The feature weights corresponding to the modality type are determined based on the feature correlation gradient and the decision sharing gradient.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the collaborative decision-making method for equipment as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the collaborative decision-making method for equipment as described in any one of claims 1 to 5.

Images