Intelligent Warehouse Order Picking and Handling System and Method Based on Warehouse Robots

Abstract

This application relates to the technical field of warehouse picking and handling. Specifically, it discloses an intelligent warehouse picking and handling system and method based on a warehousing robot. The warehousing robot is used to transport goods to an identification conveyor, collect images of the goods identification codes and perform image recognition to obtain goods information. At the same time, the weight data of the goods is obtained through a weight sensor. The data analysis technology based on deep learning is used to perform low-dimensional embedding coding on both of them and construct an explicit modeling feature of the goods weight. By analyzing the interaction between the explicit modeling feature of the goods weight and the embedded weight feature, the accuracy of the goods identification information is intelligently judged. And in response to the accurate goods identification information, the robot transports the goods. This application combines the dual mechanisms of image recognition (obtaining the type and quantity of goods) and weight sensor data, and can utilize more information sources to verify the correctness of the goods, so as to effectively distinguish goods with similar labels and reduce the occurrence of misidentification.

Description

Technical Field

[0001] This application relates to the technical field of warehouse picking and handling, and more specifically, to an intelligent warehouse picking and handling system and method based on a warehousing robot. Background Art

[0002] With the rapid development of e-commerce and modern logistics industry, the demand for intelligent warehousing systems is increasing day by day. As an automated device applied in the field of warehousing logistics, a warehousing robot can autonomously or semi-autonomously complete tasks such as goods handling, storage, and picking, greatly improving the efficiency and accuracy of warehousing management.

[0003] Patent CN118479184A proposes a three-dimensional warehouse picking system based on an intelligent robot and its control method. First, a handling robot delivers the goods to an identification conveyor for scanning and information entry; then, double-track and single-track mobile robots cooperate to store the goods in a warehousing shelf; when retrieving goods deep in the shelf, relevant components first perform misaligned movement preparation, and then retrieve the outer goods; then, the deep goods and the outer goods exchange positions; finally, a picking mechanism etc. deliver the goods to the handling robot via the identification conveyor, and it avoids obstacles and delivers them to the designated position.

[0004] In this process, after the goods are conveyed to the identification conveyor and the scanning operation is completed, it is necessary to verify the accuracy of the information obtained by scanning. Only in this way can the accuracy of the information be effectively guaranteed. However, in actual application scenarios, due to high label similarity or insufficient accuracy of scanning equipment, misidentification may occur, that is, one item is mistaken for another. Such misidentification not only affects the accuracy of inventory management, but may also lead to order processing errors and a decrease in customer satisfaction.

[0005] Therefore, an optimized intelligent warehouse picking and handling solution based on a warehousing robot is expected. Summary of the Invention

[0006] This application provides an intelligent warehouse picking and handling system and method based on a warehousing robot. By combining a dual mechanism of image recognition (obtaining the type and quantity of goods) and weight sensor data (obtaining the weight of goods), it can utilize more information sources to verify the correctness of goods, thereby effectively distinguishing goods with similar labels and reducing the occurrence of misidentification.

[0007] According to one aspect of the present application, there is provided a control method for an intelligent warehouse picking and handling system based on a warehousing robot, including: transporting goods to an identification conveyor by the warehousing robot; collecting an identification code image of the goods through a code scanning device of the identification conveyor; performing image recognition on the identification code image to obtain goods identification information, where the goods identification information includes goods type, goods quantity, and destination; obtaining goods weight data collected by a weight sensor; performing coding verification on the goods type, goods quantity, and the goods weight data in the goods identification information to obtain a verification result indicating whether the goods identification information is accurate, including: performing low-dimensional embedding and core feature cross-modal goods information verification on the goods type, the goods quantity, and the goods weight data to obtain the verification result; and in response to the goods identification information being accurate, transporting the goods to the destination by the warehousing robot.

[0008] According to another aspect of the present application, there is provided an intelligent warehouse picking and handling system based on a warehousing robot, including: a goods handling module for transporting goods to an identification conveyor by the warehousing robot; an identification code image collection module for collecting an identification code image of the goods through a code scanning device of the identification conveyor; a goods identification information acquisition module for performing image recognition on the identification code image to obtain goods identification information, where the goods identification information includes goods type, goods quantity, and destination; a goods weight data acquisition module for obtaining goods weight data collected by a weight sensor; a verification result determination module for performing coding verification on the goods type, goods quantity, and the goods weight data in the goods identification information to obtain a verification result indicating whether the goods identification information is accurate, including: performing low-dimensional embedding and core feature cross-modal goods information verification on the goods type, the goods quantity, and the goods weight data to obtain the verification result; and a goods transportation module for, in response to the goods identification information being accurate, transporting the goods to the destination by the warehousing robot.

[0009] An intelligent warehouse picking and handling system and method based on a warehousing robot provided by this application. Through the warehousing robot, the goods are transported to an identification conveyor to collect the identification code image of the goods, and image recognition is performed on it to obtain goods information (goods type, goods quantity, and destination). At the same time, the goods weight data collected by a weight sensor is obtained. Deep learning-based data analysis and coding technology are used to perform low-dimensional embedding coding on the goods type, goods quantity, and goods weight data in the goods identification information. Then, based on the embedded low-dimensional embedding features of the goods type and the low-dimensional embedding features of the goods quantity, a goods weight explicit modeling feature is constructed. Based on this, an intelligent judgment is made on whether the goods identification information is accurate according to the core feature response interaction representation between the goods weight explicit modeling feature and the embedded low-dimensional embedding feature of the goods weight. In response to the accuracy of the goods identification information, the robot transports the goods. By combining the dual mechanisms of image recognition (obtaining the goods type and quantity) and weight sensor data (obtaining the goods weight), this application can utilize more information sources to verify the correctness of the goods, thereby being able to effectively distinguish goods with similar labels and reduce the occurrence of misidentification. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of this application, the drawings of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of this application and do not limit this application.

[0011] Figure 1 It is a schematic flowchart of the control method of the intelligent warehouse picking and handling system based on a warehousing robot according to the embodiment of this application.

[0012] Figure 2 It is a schematic flowchart of step S5 in the control method of the intelligent warehouse picking and handling system based on a warehousing robot according to the embodiment of this application.

[0013] Figure 3 It is a schematic diagram of the data flow of step S5 in the control method of the intelligent warehouse picking and handling system based on a warehousing robot according to the embodiment of this application.

[0014] Figure 4 It is a schematic flowchart of step S54 in the control method of the intelligent warehouse picking and handling system based on a warehousing robot according to the embodiment of this application.

[0015] Figure 5 It is a schematic block diagram of the intelligent warehouse picking and handling system based on a warehousing robot according to the embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only partial embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts also belong to the scope of protection of the present application.

[0017] Based on this, the technical concept of the present application is to use a warehousing robot to transport goods to an identification conveyor to collect the identification code image of the goods, and perform image recognition on it to obtain goods information (goods type, goods quantity, and destination). At the same time, obtain the goods weight data collected by a weight sensor, and use data analysis and coding techniques based on deep learning to perform low-dimensional embedding coding on the goods type and goods quantity in the goods identification information and the goods weight data. Then, based on the embedded low-dimensional embedding features of the goods type and the low-dimensional embedding features of the goods quantity, construct the dominant modeling features of the goods weight, so as to intelligently judge whether the goods identification information is accurate according to the core feature response interaction representation between the dominant modeling features of the goods weight and the embedded low-dimensional embedding features of the goods weight. And in response to the accuracy of the goods identification information, transport the goods through the robot. By combining the dual mechanisms of image recognition (obtaining the goods type and quantity) and weight sensor data (obtaining the goods weight), the present application can utilize more information sources to verify the correctness of the goods, thereby effectively distinguishing goods with similar labels and reducing the occurrence of misidentification.

[0018] Specifically, in the technical solution of the present application, Figure 1 is a schematic flowchart of the control method for an intelligent warehouse picking and transporting system based on a warehousing robot according to an embodiment of the present application. As Figure 1 shown, the control method for the intelligent warehouse picking and transporting system based on a warehousing robot includes: S1, using a warehousing robot to transport goods to an identification conveyor; S2, collecting the identification code image of the goods through the scanning device of the identification conveyor; S3, performing image recognition on the identification code image to obtain goods identification information, where the goods identification information includes goods type, goods quantity, and destination; S4, obtaining the goods weight data collected by a weight sensor; S5, performing coding verification on the goods type and goods quantity in the goods identification information and the goods weight data to obtain a verification result indicating whether the goods identification information is accurate; S6, in response to the accuracy of the goods identification information, using a warehousing robot to transport the goods to the destination.

[0019] Exemplarily, in step S1, the storage robot transports the goods to the identification conveyor. It should be understood that the identification conveyor is equipped with dedicated scanning devices capable of centrally collecting and identifying information about the goods. Transporting the goods to the identification conveyor by the storage robot enables the collection and processing of information such as the identification code image of the goods at a centralized location, improving the efficiency and accuracy of information processing. Specifically, using the storage robot to complete this transportation task not only improves the operation efficiency but also enhances the flexibility and adaptability of the system. Specifically, the storage robot can autonomously navigate according to the instructions of the controller and avoid obstacles, safely transporting the goods from the storage location to the identification conveyor. In a specific example, the storage robot includes multiple precision components such as double-rail mobile robots, single-rail mobile robots, and various motors and belt components. These components work together to accurately position and transport the goods. The double-rail mobile robot can move horizontally on parallel ground rails, while the single-rail mobile robot can move left and right on the top rail. The combination of the two can cover most areas of the warehouse. In addition, the first belt component and the drive belt component on the guide rail drive the position adjustment component, enabling the goods transfer component and the transfer component to move along the preset path and finally placing the goods smoothly on the identification conveyor. Throughout the process, the controller plays a crucial role. It not only coordinates the actions of each mechanical component but also synchronizes and updates data through the built-in preset program and the database of the warehouse management system for subsequent inventory management, order processing, and logistics tracking. In this way, seamless docking can be achieved whether the goods are incoming or outgoing, greatly improving the operation efficiency and service quality of the entire warehousing system. Therefore, transporting the goods to the identification conveyor by the storage robot is not only a technical requirement but also an important means to improve the level of modern logistics warehousing management.

[0020] Exemplarily, in step S2, the identification code image of the goods is collected by the scanning device of the identification conveyor. It should be understood that collecting the identification code image of the goods by the scanning device of the identification conveyor can achieve efficient and accurate acquisition of goods information. When the goods are transported to the identification conveyor, the scanning device (such as a barcode scanner or an RFID reader) on the identification conveyor can automatically scan the goods, read their identification codes, and transmit this information to the controller in real time. This step ensures that key information such as the type, quantity, and destination of the goods can be quickly and accurately recorded. Compared with manual input, this method greatly reduces the possibility of human error and improves the reliability of the data. At the same time, this automated means also increases the processing speed, enabling the warehouse to process more goods in a shorter time and thus enhancing the overall operation efficiency.

[0021] In a specific example, the process of collecting the identification code image of the goods by the code scanning device of the conveyor is as follows: When the warehousing robot places the goods on the identification conveyor, the identification conveyor drives the goods on its top to move horizontally in the front-back direction. During this process, the code scanning device on the identification conveyor will be automatically activated to scan the identification code on the goods. The code scanning device is usually installed on the top or side of the conveyor to ensure that all possible positions can be covered.

[0022] Exemplarily, in step S3, image recognition is performed on the identification code image to obtain goods identification information, and the goods identification information includes the goods type, the quantity of goods, and the destination. It should be understood that considering that the identification code image contains key attribute information of the goods, such as the goods type, the quantity of goods, and the destination, this information provides basic data for subsequent judgment of the accuracy of this information. Based on this, in the technical solution of the present application, image recognition is performed on the identification code image to obtain goods identification information, and the goods identification information includes the goods type, the quantity of goods, and the destination. That is, the identification code, as an information carrier, usually integrates the key information of the goods in a coded form. By parsing the identification code image through image recognition technology, the information integrated in the identification code can be accurately extracted. In this way, by analyzing and comparing them with the actual data, the correctness of the information can be verified, and the goods can be sent to the correct destination.

[0023] In a specific example, the process of performing image recognition on the identification code image is as follows: When the goods are placed on the identification conveyor, the identification conveyor will activate the code scanning device (such as a barcode scanner or an RFID reader) installed on its top or side to scan the identification code on the goods. The identification code usually contains basic information of the goods, such as barcodes or QR codes. The code scanning device will capture the images of these identification codes and transmit them to the controller. The controller is built-in with advanced image recognition algorithms and can analyze and decode these images to extract detailed information such as the type, quantity, and destination of the goods. During the image recognition process, the controller will first preprocess the identification code image, such as denoising and enhancing the contrast, to improve the image quality. Then, using optical character recognition (OCR) technology or other specialized decoding algorithms, the specific data in the identification code is parsed.

[0024] Exemplarily, in step S4, the weight data of the goods collected by the weight sensor is obtained. It should be understood that obtaining the weight data of the goods is to further confirm the accuracy of the goods identification information. Although image recognition technology can efficiently read the barcodes or two-dimensional codes of the goods and extract information such as the type, quantity, and destination of the goods, there may still be errors in this information. For example, the label may be incorrectly pasted or damaged, resulting in inaccurate scanning results. By obtaining the actual weight data of the goods through the weight sensor and comparing it with the expected weight, the correctness of the goods identification information can be effectively verified. If the actual weight does not match the expected weight, the system can promptly detect and handle abnormal situations, avoiding inventory chaos and order processing errors caused by incorrect information. In addition, the weight data can also help optimize warehouse management and logistics operations. For example, in a three-dimensional warehouse with high-density storage, there are a large number of goods with different types and quantities, and different types of goods have different weight characteristics. Through the data provided by the weight sensor, the system can more accurately allocate storage locations, reasonably plan the handling path, and ensure the safe and efficient storage and transportation of the goods. At the same time, the weight data also provides an important reference for subsequent logistics distribution, ensuring that each piece of goods can be delivered to the designated location on time and accurately.

[0025] In a specific example, the process of obtaining the weight data of the goods collected by the weight sensor is as follows: When the goods are placed on the identification conveyor, the identification conveyor activates its built-in weight sensors. These sensors are usually installed at key positions on the conveyor, such as under the goods carriage, to ensure accurate measurement of the weight of the goods. When the goods pass through the sensors, the sensors will capture the weight data of the goods in real time and transmit this data to the controller.

[0026] Exemplarily, in step S5, encoding verification is performed on the goods type, goods quantity in the goods identification information, and the goods weight data to obtain a verification result indicating whether the goods identification information is accurate. It should be understood that, firstly, encoding verification can effectively prevent misidentification problems caused by high label similarity or insufficient scanning device accuracy. Although image recognition technology has been very mature, in actual operation, there may still be situations such as damaged, blurred, or mispasted labels, which will all affect the accuracy of the recognition result. For example, two different types of goods may have similar barcodes or QR codes, and relying solely on image recognition technology may lead to misjudgment. By combining the actual weight data obtained by the weight sensor, the system can perform comparison and verification in multiple dimensions, thereby greatly reducing the probability of misidentification. If the expected weight of a certain piece of goods does not match the actual measured weight, the system can promptly detect and correct the error to ensure the accuracy of the goods information. Secondly, encoding verification helps to improve the accuracy of inventory management. Modern warehousing systems usually need to handle a large number of goods with various types and large quantities, and accurate inventory management is crucial for the smooth operation of the supply chain. By performing encoding verification on the goods type, quantity, and weight data, the system can update and check the inventory information in real time, avoiding inventory chaos caused by inconsistent information.

[0027] In one embodiment, performing encoding verification on the goods type, goods quantity in the goods identification information, and the goods weight data to obtain a verification result indicating whether the goods identification information is accurate includes: performing low-dimensional embedding and cross-modal goods information verification of core features on the goods type, the goods quantity, and the goods weight data to obtain the verification result.

[0028] In one embodiment, Figure 2 is a schematic flowchart of step S5 in the control method of the intelligent warehouse picking and handling system based on a warehousing robot according to an embodiment of the present application. Figure 3 is a schematic diagram of data flow of step S5 in the control method of the intelligent warehouse picking and handling system based on a warehousing robot according to an embodiment of the present application. As Figure 2 and Figure 3As shown, low-dimensional embedding and cross-modal cargo information verification of the cargo type, the cargo quantity, and the cargo weight data are performed to obtain the verification result, including: S51, respectively performing low-dimensional embedding encoding on the cargo type and the cargo quantity to obtain a low-dimensional embedding encoding vector of the cargo type and a low-dimensional embedding encoding vector of the cargo quantity; S52, performing low-dimensional embedding encoding on the cargo weight data to obtain a low-dimensional embedding encoding vector of the cargo weight; S53, constructing a cargo weight explicit modeling encoding vector based on the low-dimensional embedding encoding vector of the cargo type and the low-dimensional embedding encoding vector of the cargo quantity; S54, performing cross-modal cargo information verification on the low-dimensional embedding encoding vector of the cargo weight and the cargo weight explicit modeling encoding vector to obtain a response interaction encoding vector between heterogeneous features of the cargo information; S55, obtaining the verification result based on the response interaction encoding vector between heterogeneous features of the cargo information.

[0029] Exemplarily, in step S51, respectively performing low-dimensional embedding encoding on the cargo type and the cargo quantity to obtain a low-dimensional embedding encoding vector of the cargo type and a low-dimensional embedding encoding vector of the cargo quantity. It should be understood that, considering that in order to convert the cargo type and the cargo quantity into a form that can be processed by a computer and capture the inherent features in their respective data, to provide support for subsequent analysis and processing, the present application maps these data to a low-dimensional space by respectively performing low-dimensional embedding encoding on the cargo type and the cargo quantity to obtain a low-dimensional embedding encoding vector of the cargo type and a low-dimensional embedding encoding vector of the cargo quantity. In particular, in a specific example of the present application, respectively performing low-dimensional embedding encoding on the cargo type and the cargo quantity to obtain a low-dimensional embedding encoding vector of the cargo type and a low-dimensional embedding encoding vector of the cargo quantity includes: using an embedding layer to respectively encode the cargo type and the cargo quantity to obtain the low-dimensional embedding encoding vector of the cargo type and the low-dimensional embedding encoding vector of the cargo quantity. The low-dimensional embedding encoding vector can be obtained by using an embedding layer to respectively encode the cargo type and the cargo quantity. That is, the low-dimensional vectors generated by the embedding layer can more effectively represent the features of the cargo type and quantity. Compared with the original simple encoding method, the low-dimensional embedding encoding vector can more accurately capture the essential attributes of the cargo, provide a more reliable information basis for subsequent operations such as the accuracy judgment of cargo identification information, inventory management, and path planning of warehouse robots, and improve the system's understanding and processing ability of cargo information.

[0030] Exemplarily, in step S52, the low-dimensional embedding coding is performed on the cargo weight data to obtain a low-dimensional embedded coding vector of the cargo weight. It should be understood that the low-dimensional embedding coding is performed on the cargo weight data to map the cargo weight data into a low-dimensional space, so as to obtain a low-dimensional embedded coding vector of the cargo weight. Specifically, in a specific example of the present application, performing the low-dimensional embedding coding on the cargo weight data to obtain a low-dimensional embedded coding vector of the cargo weight includes: using the embedding layer to perform embedding coding on the cargo weight data to obtain the low-dimensional embedded coding vector of the cargo weight. That is, through the embedding vector, the internal structure and pattern of the weight data can be better represented in the low-dimensional space, enabling the model to more effectively utilize this information.

[0031] Exemplarily, in step S53, based on the low-dimensional embedded coding vector of the cargo type and the low-dimensional embedded coding vector of the cargo quantity, a dominant modeling coding vector of the cargo weight is constructed. It should be understood that there is an inherent logical connection between the cargo type, the cargo quantity, and the cargo weight. Different types of cargo usually have their respective relatively fixed unit weights, and the amount of cargo directly affects the total weight. Therefore, in order to discover the associated information hidden behind the data and thus more accurately understand the formation mechanism of the cargo weight, in the present application, a dominant modeling coding vector of the cargo weight is constructed based on the low-dimensional embedded coding vector of the cargo type and the low-dimensional embedded coding vector of the cargo quantity. Specifically, in a specific example of the present application, constructing a dominant modeling coding vector of the cargo weight based on the low-dimensional embedded coding vector of the cargo type and the low-dimensional embedded coding vector of the cargo quantity includes: performing element-wise multiplication of the low-dimensional embedded coding vector of the cargo type and the low-dimensional embedded coding vector of the cargo quantity in position to obtain the dominant modeling coding vector of the cargo weight. Specifically, performing element-wise multiplication of the low-dimensional embedded coding vector of the cargo type and the low-dimensional embedded coding vector of the cargo quantity in position can reflect this inherent logical connection at the vector level. For example, the elements of the embedding vector corresponding to a certain type of fragile cargo may represent some characteristic weights of this type of cargo. After dot-multiplying with the cargo quantity vector, it can comprehensively reflect the expected weight characteristics based on the type and quantity to obtain the dominant modeling coding vector of the cargo weight. In this way, by constructing a dominant modeling coding vector of the cargo weight and comparing and verifying it with the actual weight detection result, the accuracy of the item identification information can be more effectively judged.

[0032] Exemplarily, in step S54, core feature cross-modal cargo information verification is performed on the low-dimensional embedded coding vector of the cargo weight and the explicit modeling coding vector of the cargo weight to obtain a response interaction coding vector between heterogeneous features of the cargo information. It should be understood that considering that the low-dimensional embedded coding vector of the cargo weight is obtained by directly performing low-dimensional embedding coding on the actual weight data of the cargo, reflecting the actual weight characteristics of the cargo; while the explicit modeling coding vector of the cargo weight is obtained by specific operations (such as element-wise multiplication by position) on the low-dimensional embedded coding vector based on the cargo type and quantity, containing weight-related features inferred from the cargo type and quantity. There may be potential relationships and differences between the two. Therefore, in order to verify these two different sources of feature information to more accurately judge the identification information, in this application, core feature cross-modal cargo information verification is performed on the low-dimensional embedded coding vector of the cargo weight and the explicit modeling coding vector of the cargo weight to obtain a response interaction coding vector between heterogeneous features of the cargo information. That is, by verifying these two vectors, these potential relationships can be deeply explored, for example, discovering the deviation rules between the actual weight and the expected weight under certain combinations of cargo types and quantities. This exploration helps to better understand the attributes and characteristics of the cargo, providing more valuable information for subsequent analysis and decision-making.

[0033] In one embodiment, Figure 4 This is a schematic flowchart of step S54 in the control method of the intelligent warehouse picking and handling system based on a warehousing robot according to an embodiment of the present application. As Figure 4 shown, performing core feature cross-modal cargo information verification on the low-dimensional embedded coding vector of the cargo weight and the explicit modeling coding vector of the cargo weight to obtain a response interaction coding vector between heterogeneous features of the cargo information includes: S541, extracting core features from the low-dimensional embedded coding vector of the cargo weight and the explicit modeling coding vector of the cargo weight to obtain a low-dimensional embedded feature core information anchoring coding vector of the cargo weight and an explicit modeling feature core information anchoring coding vector of the cargo weight; S542, performing cargo information feature granularity response interaction coding on the low-dimensional embedded feature core information anchoring coding vector of the cargo weight and the explicit modeling feature core information anchoring coding vector of the cargo weight to obtain a cargo information core feature granularity response interaction coding vector; S543, performing cargo information feature value granularity response interaction coding on the low-dimensional embedded feature core information anchoring coding vector of the cargo weight and the explicit modeling feature core information anchoring coding vector of the cargo weight to obtain a cargo information core feature value granularity response interaction coding vector; S544, fusing the cargo information core feature granularity response interaction coding vector and the cargo information core feature value granularity response interaction coding vector to obtain the response interaction coding vector between heterogeneous features of the cargo information.

[0034] In one embodiment, in step S541, core features are extracted from the low-dimensional embedded coding vector of the cargo weight and the explicit modeling coding vector of the cargo weight to obtain a low-dimensional embedded feature core information anchor coding vector of the cargo weight and an explicit modeling feature core information anchor coding vector of the cargo weight, including: respectively constructing autocorrelation association matrices of the low-dimensional embedded coding vector of the cargo weight and the explicit modeling coding vector of the cargo weight to obtain a low-dimensional embedded autocorrelation association matrix of the cargo weight and an explicit modeling autocorrelation association matrix of the cargo weight. Specifically, this process is represented by the formula: ; where is the low-dimensional embedded coding vector of the cargo weight, is the explicit modeling coding vector of the cargo weight, is the transpose operation, and are feature mapping functions, such as linear mapping or non-linear kernel functions, is the low-dimensional embedded autocorrelation association matrix of the cargo weight, is the explicit modeling autocorrelation association matrix of the cargo weight.

[0035] Perform cargo information autocorrelation core information anchoring on the low-dimensional embedded autocorrelation association matrix of the cargo weight and the explicit modeling autocorrelation association matrix of the cargo weight respectively to obtain the low-dimensional embedded feature core information anchor coding vector of the cargo weight and the explicit modeling feature core information anchor coding vector of the cargo weight. Specifically, this process is represented by the formula: ; where is the core information anchoring operation, is the feature decoupling operation, and are respectively the each row vector of, and are respectively corresponding low-dimensional embedded weight matrix of the cargo weight and low-dimensional embedded bias vector of the cargo weight, is matrix multiplication, is the low-dimensional embedded scoring weight vector of the cargo weight, is the th cargo weight low-dimensional embedded significant factor in the set of cargo weight low-dimensional embedded significant factors, is the normalization function, is the th normalized cargo weight low-dimensional embedded significant factor in the set of normalized cargo weight low-dimensional embedded significant factors, is the number of row vectors in, is the low-dimensional embedded feature core information anchor coding vector of the cargo weight, and are respectively the respective row vectors, and are respectively the corresponding explicit weight matrix of cargo weight and the explicit bias vector of cargo weight, is the explicit scoring weight vector of cargo weight, is the th explicit significant factor of cargo weight in the set of explicit significant factors of cargo weight, is the th normalized explicit significant factor of cargo weight in the set of normalized explicit significant factors of cargo weight, is the number of row vectors in, and and have equal numbers, is the core information anchoring coding vector of the explicit modeling features of cargo weight.

[0036] It should be understood that constructing the autocorrelation correlation matrices of the low-dimensional embedded coding vector and the explicit modeling coding vector of the cargo weight respectively and performing the core information anchoring operation on them are to ensure the high accuracy of the cargo identification information and the intelligent level of the system. Through multi-level data processing and feature extraction, this process can effectively improve the reliability and efficiency of the entire warehousing management system. First, constructing the autocorrelation correlation matrices of the low-dimensional embedded coding vector and the explicit modeling coding vector of the cargo weight is to make explicit the mutual correlation between the components within these feature vectors. Specifically, the low-dimensional embedded coding vector of the cargo weight is obtained by performing low-dimensional embedded coding on the actually measured cargo weight data, which reflects the actual weight characteristics of the cargo; while the explicit modeling coding vector of the cargo weight is generated based on the low-dimensional embedded coding vectors of the cargo type and quantity through specific operations (such as element-wise multiplication by position), reflecting the expected weight characteristics inferred from the cargo type and quantity. Through feature mapping, the system can generate the autocorrelation correlation matrices of these two feature vectors, thereby revealing the relationship between the components within them. Next, performing the cargo information autocorrelation core information anchoring operation on the generated low-dimensional autocorrelation correlation matrix of the cargo weight and the explicit modeling autocorrelation correlation matrix of the cargo weight is to further refine and purify the key information in the feature vectors. This step uses means such as feature distillation to highlight the part highly relevant to the core semantics in the feature vectors, removing redundant information and noise. Specifically, the system will use the core information anchoring network based on autocorrelation decoupling to process the internal information of the features and generate a structurally refined core anchor representation. In addition, the core information anchoring operation also has the effect of constraining the semantic consistency of the feature representation in the compressed space. By constraining the information entropy, the system can improve the generalization ability of the model and ensure its high accuracy in different situations. For example, when dealing with a large number of different types and quantities of goods, the system needs to have sufficient flexibility and adaptability. By anchoring the core information, the system can maintain stable performance in a complex and changing environment and avoid misjudgment caused by data fluctuations or noise interference.

[0037] Exemplarily, in step S542, by performing cargo information feature granularity response interaction encoding on the cargo weight low-dimensional embedded feature core information anchored encoding vector and the cargo weight explicit modeling feature core information anchored encoding vector, the system can understand various attributes of the cargo at a more detailed level. This encoding method not only focuses on the point-to-point comparison of individual feature sub-elements but also involves their overall behavior patterns in the entire feature set. For example, when processing a batch of electronic products, the system will not only analyze the specific weight differences of each product but also consider the overall behavior pattern of this batch of products, such as the total weight distribution law and its potential correlations. In this process, the system uses cargo information feature granularity response interaction encoding to capture the subtle differences and dependencies between the two feature vectors. The core of cargo information feature granularity response interaction encoding lies in revealing the deep semantic coupling within the feature vectors. In this way, the system can discover hidden patterns and associations in the cargo weight data. For example, the system can identify that there are systematic deviations between the actual weight and the expected weight of certain cargos, which may be due to changes in packaging materials, the presence of additional accessories, or weight deviations of certain products. These subtle differences are crucial for optimizing inventory management and route planning. Specifically, the calculation process of step S542 is expressed by the formula: ; where, is pointwise addition, and are the response weight matrix and the response bias vector respectively, is the hyperbolic tangent function, is the cargo information core feature granularity response interaction encoding vector.

[0038] Exemplarily, in step S543, by performing cargo information feature value granularity response interaction encoding on the cargo weight low-dimensional embedded feature core information anchored encoding vector and the cargo weight explicit modeling feature core information anchored encoding vector, the system can understand various attributes of the cargo more meticulously at the numerical level. This encoding method not only focuses on the overall weight differences but also details to the specific weight values of each individual cargo. In this process, the system uses element-wise operations to characterize the interdependencies between feature values. This approach enables the system to connect the semantic abstraction layer and the data numerical layer, thereby enriching the representational dimensions of the interaction modeling. Specifically, the system can reveal the specific numerical differences between the actual weight and the expected weight through element-wise operations and further analyze which specific cargos contribute to this deviation. Specifically, the calculation process of step S543 is expressed by the formula: ; where, is the cargo information core feature value granularity response interaction encoding vector.

[0039] Exemplarily, in step S544, during the interaction coding process of the cargo information feature granularity response, the system explores the response relationship between two feature vectors at the semantic unit level and captures the semantic coupling at the micro level. This coding method not only focuses on the point-to-point comparison of individual feature sub-elements but also involves their overall behavior patterns in the entire feature set. During the interaction coding process of the cargo information feature value granularity response, the system returns to a more fundamental numerical operation, and the element-wise operation is used to characterize the interdependence between feature values. This approach enables the system to connect the semantic abstraction layer and the data numerical layer, thus enriching the representational dimension of the interaction modeling. By the element-wise operation, the specific numerical difference between the actual weight and the expected weight is revealed, and further analysis is carried out on which specific goods have caused this deviation. Next, the system fuses these two coding vectors to generate a unified coding vector that combines multi-granularity information expression - the interaction coding vector of the response between heterogeneous features of the cargo information. This process is not just a simple concatenation operation but also includes the integration and optimization of feature information to ensure that the finally generated vector can comprehensively reflect various attributes and states of the cargo. Through this multi-level information integration, the system can make more accurate decisions in a complex warehousing environment. The fused coding vector not only contains the overall semantic information of the cargo but also contains specific numerical differences. This comprehensive information enables the system to better understand and process various information of the cargo, reduce the possibility of human errors, and improve the overall operation efficiency. Specifically, the calculation process of step S544 is expressed by the formula: ; where is the concatenation operation, is the interaction coding vector of the response between heterogeneous features of the cargo information.

[0040] Preferably, in another embodiment of the present application, in step S544, fusing the interaction coding vector of the core feature granularity response of the cargo information and the interaction coding vector of the core feature value granularity response of the cargo information to obtain the interaction coding vector of the response between heterogeneous features of the cargo information includes: performing cargo information growth distribution interaction gradient correction on the interaction coding vector of the core feature granularity response of the cargo information and the interaction coding vector of the core feature value granularity response of the cargo information to obtain a corrected interaction coding vector of the core feature granularity response of the cargo information composed of multiple corrected core feature granularity feature values of the cargo information and a corrected interaction coding vector of the core feature value granularity response of the cargo information composed of multiple corrected core feature value granularity feature values of the cargo information; fusing the corrected interaction coding vector of the core feature granularity response of the cargo information and the corrected interaction coding vector of the core feature value granularity response of the cargo information to obtain the interaction coding vector of the response between heterogeneous features of the cargo information. This process is expressed by the formula: ; where is in the The eigenvalue of a position is the eigenvalue of the position in a cosine function is the corrected eigenvalue is the corrected eigenvalue is the interaction coding vector of the core feature granularity response of the corrected goods information is the interaction coding vector of the core eigenvalue granularity response of the corrected goods information is the interaction coding vector of the response between heterogeneous features of the goods information

[0041] Here, considering the difference between the modeling representations of the interaction coding vector of the core feature granularity response of the goods information and the interaction coding vector of the core eigenvalue granularity response of the goods information may cause unstable perturbations in the feature manifold interface of the interaction coding vector of the heterogeneous features of the goods information with multi-granularity information expression after fusion. The interaction coding vector of the core eigenvalue granularity response of the goods information is used as an extended constraint representation to target the overall feature interaction distribution growth under the eigenvalue diffusion process. Based on the growth index representation under the extended constraint, the interface shape perturbation deviation of the overall distribution of the interaction coding vector of the core feature granularity response of the goods information is modeled. That is, the interaction interface gradient under each eigenvalue granularity is used as the perturbation contribution factor to determine the growth index stabilization contribution of the gradient diffusion under the eigenvalue dependence, so as to achieve the growth mode gradient correction dominated by perturbation stabilization and improve the manifold interface stability of the interaction coding vector of the heterogeneous features of the goods information

[0042] Exemplarily, in step S55, the verification result is obtained based on the response interaction coding vector between the heterogeneous features of the goods information. In one embodiment, obtaining the verification result based on the response interaction coding vector between the heterogeneous features of the goods information includes: inputting the response interaction coding vector between the heterogeneous features of the goods information into a goods information verifier based on a classifier to obtain the verification result. That is, the response interaction coding vector between the heterogeneous features of the goods information obtained by performing core feature cross-domain response interaction using the low-dimensional embedded coding vector of the goods weight and the explicit modeling coding vector of the goods weight is classified to automatically determine whether the goods identification information is accurate. In particular, the classifier is trained with a large amount of data and can learn the mapping relationship between different feature patterns and specific categories. The response interaction coding vector between the heterogeneous features of the goods information contains rich information after the interaction between the actual weight feature of the goods and the expected weight feature inferred based on the type and quantity. The goods information verifier based on the classifier can analyze and identify these complex feature patterns and determine whether they conform to the pattern of accurate goods identification information. For example, during the training process, the classifier learns that when the actual weight matches the expected weight within a certain range, the goods identification information is accurate, and then it can judge whether the current goods belong to this accurate pattern based on the interaction coding vector. In a specific example of the present application, the fused response interaction coding vector between the heterogeneous features of the goods information is input into a pre-trained Softmax classifier. The classifier calculates the unnormalized scores for each category based on the input features and converts them into a probability distribution through the Softmax function. The probability distribution output by the classifier represents the possibility of each category. The goods information verifier based on the classifier has two types of labels: "goods identification information is accurate" and "goods identification information is inaccurate". For example, the result output by the classifier may be: the probability of the "goods identification information is accurate" category is 0.95, and the probability of the "goods identification information is inaccurate" category is 0.05. Therefore, the output verification result is that the goods identification information is accurate.

[0043] Exemplarily, in step S6, in response to the goods identification information being accurate, the goods are transported to the destination by the warehousing robot. It should be understood that only when the goods identification information is accurate can it be clearly known what the specific goods to be carried are (including the goods type, quantity, etc.) and where to be transported (destination). The warehousing robot operates based on accurate information, which can avoid mistakes such as picking up the wrong goods and transporting them to the wrong place, thus ensuring the accuracy of the entire handling process. At the same time, for some special types of goods, such as fragile goods and dangerous goods, accurate identification information enables the robot to adopt appropriate handling methods and forces, ensuring the safety of the operation, preventing damage to the goods or occurrence of safety accidents, and improving the operation efficiency of the entire warehouse.

[0044] Specifically, in response to the accurate cargo identification information, the cargo is transported to the destination by the warehousing robot, including the following steps: First, the system calculates the optimal handling path based on the destination of the cargo and the current warehouse layout. This step is usually completed by the Warehouse Management System (WMS), taking into account the traffic conditions, obstacles, and priorities of other handling tasks in the warehouse to ensure that the robot can reach the target location in the shortest time and the safest way. The path planning algorithm will comprehensively consider multiple factors such as distance, time, and energy consumption to select an optimal path. Then, the system assigns the handling task to a suitable warehousing robot. Warehousing robots are usually equipped with various sensors (such as lidar, cameras, etc.) for navigation and obstacle avoidance. These sensors can sense the surrounding environment in real time to ensure that the robot will not collide or have other accidents during the movement. The robot will also follow the predetermined path to the location of the cargo according to the path planning result. After reaching the location of the cargo, the warehousing robot uses a robotic arm or other grasping devices to pick up the cargo from the shelf and transport it to the designated destination. During this process, the robot will continuously monitor its status and environmental changes to ensure the safety and accuracy of the handling operation. For example, the robot can detect the status of the cargo through sensors to ensure that the cargo is not damaged or lost. Finally, after the handling is completed, the system will record the relevant information of this handling (such as time, path, cargo status, etc.) and update the inventory information in the warehouse database. These records not only help with subsequent operations and management but also serve as the basis for data analysis to help optimize the warehouse operation strategy.

[0045] In summary, the control method of the intelligent warehouse picking and handling system based on warehousing robots according to the embodiments of the present application is clarified. It uses the warehousing robot to transport the cargo to the identification conveyor to collect the identification code image of the cargo and performs image recognition on it to obtain the cargo information (cargo type, cargo quantity, and destination). At the same time, it obtains the cargo weight data collected by the weight sensor, and uses deep learning-based data analysis and coding technology to perform low-dimensional embedding coding on the cargo type, cargo quantity, and cargo weight data in the cargo identification information. Then, based on the low-dimensional embedding features of the cargo type and cargo quantity after embedding, it constructs the cargo weight dominant modeling features, and thus intelligently judges whether the cargo identification information is accurate according to the core feature response interaction representation between the cargo weight dominant modeling features and the low-dimensional embedding features of the cargo weight after embedding. In response to the accurate cargo identification information, the robot transports the cargo. By combining the dual mechanisms of image recognition (obtaining the cargo type and quantity) and weight sensor data (obtaining the cargo weight), the present application can utilize more information sources to verify the correctness of the cargo, thereby being able to effectively distinguish cargos with similar labels and reduce the occurrence of misidentifications.

[0046] Figure 5Schematic block diagram of an intelligent warehouse picking and handling system based on a warehousing robot according to an embodiment of the present application. As Figure 5 shown, the intelligent warehouse picking and handling system 100 based on a warehousing robot includes: a goods handling module 110 for transporting goods to an identification conveyor through a warehousing robot; an identification code image acquisition module 120 for acquiring an identification code image of the goods through a code scanning device of the identification conveyor; a goods identification information acquisition module 130 for performing image recognition on the identification code image to obtain goods identification information, where the goods identification information includes goods type, goods quantity, and destination; a goods weight data acquisition module 140 for acquiring goods weight data collected by a weight sensor; a verification result determination module 150 for performing coding verification on the goods type and goods quantity in the goods identification information and the goods weight data to obtain a verification result indicating whether the goods identification information is accurate, including: performing low-dimensional embedding and cross-modal goods information verification on the goods type, the goods quantity, and the goods weight data to obtain the verification result; a goods transportation module 160 for transporting the goods to the destination through a warehousing robot in response to the accuracy of the goods identification information.

[0047] Here, those skilled in the art can understand that the specific operations of each module in the above intelligent warehouse picking and handling system based on a warehousing robot have been described in detail above with reference to Figures 1 to 4 the description of the intelligent warehouse picking and handling method based on a warehousing robot, and therefore, the repeated description thereof will be omitted.

[0048] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0049] It should be understood that the specific examples herein are only for helping those skilled in the art better understand the embodiments of the present application, rather than limiting the scope of the embodiments of the present application.

[0050] It should also be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the various processes do not mean the order of execution, and the order of execution of the various processes should be determined by their functions and internal logics, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0051] It should also be understood that the various embodiments described in this specification can be implemented either alone or in combination, and the embodiments of the present application do not limit this.

[0052] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as commonly understood by those skilled in the technical field of the present application. The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the scope of the present application. The term "and / or" used in the present application includes any and all combinations of one or more of the related listed items. The singular forms "a", "above", and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0053] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0054] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0055] In addition, the functional units in the various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0056] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this 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 for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0057] As described above, the above are only the specific implementation manners of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A control method for an intelligent warehouse picking and handling system based on a storage robot, characterized in that: include: The goods are transported to the identification conveyor by a storage robot; and the identification code image of the goods is collected by a code scanning device of the identification conveyor; Performing image recognition on the identification code image to obtain cargo identification information, the cargo identification information including cargo type, cargo quantity and destination; acquiring cargo weight data collected by a weight sensor; performing coding verification on the cargo type and cargo quantity in the cargo identification information and the cargo weight data to obtain a verification result indicating whether the cargo identification information is accurate, including: performing low-dimensional embedding and core feature cross-modal cargo information verification on the cargo type, cargo quantity and cargo weight data to obtain the verification result; in response to the cargo identification information being accurate, transporting the cargo to the destination by a warehousing robot; The cargo type, cargo quantity and cargo weight data are subjected to low-dimensional embedding and core feature cross-modal cargo information verification to obtain the verification result, including: respectively performing low-dimensional embedding coding on the cargo type and the cargo quantity to obtain a cargo type low-dimensional embedding coding vector and a cargo quantity low-dimensional embedding coding vector; performing low-dimensional embedding coding on the cargo weight data to obtain a cargo weight low-dimensional embedding coding vector; constructing a cargo weight explicit modeling coding vector based on the cargo type low-dimensional embedding coding vector and the cargo quantity low-dimensional embedding coding vector; performing core feature cross-modal cargo information verification on the cargo weight low-dimensional embedding coding vector and the cargo weight explicit modeling coding vector to obtain a cargo information heterogeneous feature response interaction coding vector; and obtaining the verification result based on the cargo information heterogeneous feature response interaction coding vector.

2. The control method of the intelligent warehouse picking and handling system based on the storage robot according to claim 1 is characterized in that: The cargo type and the cargo quantity are respectively low-dimensionally embedded and encoded to obtain a cargo type low-dimensional embedded coding vector and a cargo quantity low-dimensional embedded coding vector, including: using an embedding layer to respectively encode the cargo type and the cargo quantity to obtain a cargo type low-dimensional embedded coding vector and a cargo quantity low-dimensional embedded coding vector.

3. The control method of the intelligent warehouse picking and handling system based on the storage robot according to claim 2 is characterized in that: The cargo weight data is low-dimensionally embedded and encoded to obtain a cargo weight low-dimensional embedded coding vector, comprising: using the embedding layer to embed the cargo weight data to obtain the cargo weight low-dimensional embedded coding vector.

4. The control method of the intelligent warehouse picking and handling system based on the storage robot according to claim 3 is characterized in that: Based on the low-dimensional embedded coding vector of the cargo type and the low-dimensional embedded coding vector of the cargo quantity, a cargo weight explicit modeling coding vector is constructed, including: multiplying the low-dimensional embedded coding vector of the cargo type and the low-dimensional embedded coding vector of the cargo quantity by position points to obtain the cargo weight explicit modeling coding vector.

5. The control method of the intelligent warehouse picking and handling system based on the storage robot according to claim 4 is characterized in that: The core feature cross-modal cargo information verification is performed on the cargo weight low-dimensional embedding coding vector and the cargo weight explicit modeling coding vector to obtain a response interaction coding vector between heterogeneous features of cargo information, including: extracting core features from the cargo weight low-dimensional embedding coding vector and the cargo weight explicit modeling coding vector to obtain a cargo weight low-dimensional embedding feature core information anchor coding vector and a cargo weight explicit modeling feature core information anchor coding vector; performing cargo information feature granularity response interaction coding on the cargo weight low-dimensional embedding feature core information anchor coding vector and the cargo weight explicit modeling feature core information anchor coding vector to obtain a cargo information core feature granularity response interaction coding vector; performing cargo information feature value granularity response interaction coding on the cargo weight low-dimensional embedding feature core information anchor coding vector and the cargo weight explicit modeling feature core information anchor coding vector to obtain a cargo information core feature value granularity response interaction coding vector; fusing the cargo information core feature granularity response interaction coding vector and the cargo information core feature value granularity response interaction coding vector to obtain the cargo information heterogeneous feature response interaction coding vector.

6. The control method of the intelligent warehouse picking and handling system based on the storage robot according to claim 5 is characterized in that: Extracting core features from the cargo weight low-dimensional embedding coding vector and the cargo weight explicit modeling coding vector to obtain cargo weight low-dimensional embedding feature core information anchor coding vector and cargo weight explicit modeling feature core information anchor coding vector, including: constructing the autocorrelation association matrix of the cargo weight low-dimensional embedding coding vector and the cargo weight explicit modeling coding vector respectively to obtain cargo weight low-dimensional embedding autocorrelation association matrix and cargo weight explicit modeling autocorrelation association matrix; anchoring cargo information autocorrelation core information of the cargo weight low-dimensional embedding autocorrelation association matrix and the cargo weight explicit modeling autocorrelation association matrix respectively to obtain cargo weight low-dimensional embedding feature core information anchor coding vector and cargo weight explicit modeling feature core information anchor coding vector.

7. The control method of the intelligent warehouse picking and handling system based on the storage robot according to claim 6 is characterized in that: The cargo information core feature granularity response interactive coding vector and the cargo information core eigenvalue granularity response interactive coding vector are integrated to obtain the cargo information heterogeneous feature response interactive coding vector, including: performing cargo information growth distribution interactive gradient correction on the cargo information core feature granularity response interactive coding vector and the cargo information core eigenvalue granularity response interactive coding vector to obtain a corrected cargo information core feature granularity response interactive coding vector composed of multiple corrected cargo information core feature granularity eigenvalues ​​and a corrected cargo information core eigenvalue granularity response interactive coding vector composed of multiple corrected cargo information core eigenvalue granularity eigenvalues; and integrating the corrected cargo information core feature granularity response interactive coding vector and the corrected cargo information core eigenvalue granularity response interactive coding vector to obtain the cargo information heterogeneous feature response interactive coding vector.

8. The control method of the intelligent warehouse picking and handling system based on the storage robot according to claim 7 is characterized in that: The verification result is obtained based on the response interaction coding vector between the heterogeneous features of the cargo information, including: inputting the response interaction coding vector between the heterogeneous features of the cargo information into a cargo information verifier based on a classifier to obtain the verification result.

9. An intelligent warehouse picking and handling system based on a warehouse robot according to the control method of the intelligent warehouse picking and handling system based on a warehouse robot according to claim 1, characterized in that: include: A cargo handling module, used to carry cargo to an identification conveyor by means of a storage robot; an identification code image acquisition module, used to acquire an identification code image of the cargo by means of a code scanning device of the identification conveyor; a cargo identification information acquisition module, used to perform image recognition on the identification code image to obtain cargo identification information, wherein the cargo identification information includes cargo type, cargo quantity and destination; a cargo weight data acquisition module, used to acquire cargo weight data acquired by a weight sensor; a verification result determination module, used to perform coding verification on the cargo type and cargo quantity in the cargo identification information and the cargo weight data to obtain a verification result indicating whether the cargo identification information is accurate, including: performing low-dimensional embedding and core feature cross-modal cargo information verification on the cargo type, cargo quantity and cargo weight data to obtain the verification result; The cargo delivery module is used to deliver the cargo to the destination by a storage robot in response to the cargo identification information being accurate.

Images