An industrial stereoscopic warehouse AI intelligent management method and system

Abstract

The application relates to the technical field of industrial stereoscopic warehouse management, in particular to an AI intelligent management method and system for an industrial stereoscopic warehouse, which comprises a data acquisition module, a parameter configuration module, a path optimization module, a prediction analysis module and a task allocation module. The path optimization, the LSTM time sequence prediction model for cargo storage information prediction and the equipment task allocation are carried out, so that the whole-process intelligent management of warehouse cargo storage, sorting and transportation is completed. The application can optimize resource utilization efficiency, improve the real-time response capability and multi-task concurrent processing capability of the system, and meet the requirements of modern industrial efficient and intelligent warehouse management.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent warehousing and artificial intelligence, in particular to an AI intelligent management method and system for industrial stereoscopic warehouse. BACKGROUND

[0002] Warehouse management, also known as warehouse management, refers to the effective control of activities such as the receipt, delivery, and inventory of warehouse goods. Its purpose is to ensure the integrity of warehouse goods and the normal operation of production and business activities, and to record the activity status of various goods in a clear and concise manner to express the quantity, quality, geographical location, department, order ownership, and warehouse dispersion of warehouse goods. Warehouse management is related to logistics transportation. Generally, after sorting, various goods stored in the warehouse can be transported by local express, long-distance express, dedicated car and dedicated person delivery, global air transport, and global sea transport. Therefore, a dynamic management mechanism based on different logistics transportation methods needs to be introduced to warehouse management to ensure that limited logistics transportation resources can match the changing dynamics of goods.

[0003] The patent CN115034391B provides a one-stop model compression system for intelligent detection of welding defects. This patent realizes the lightweight compression of quality defect detection models through modules such as model warehouse management, compression algorithm configuration, and resource scheduling, thereby improving the inference efficiency of the model on the edge side. Although this technical solution is innovative in model compression and resource scheduling, its application scenarios mainly focus on welding defect detection, which has some differences from the management needs of industrial stereoscopic warehouses. Specifically, this solution does not involve the intelligent management of the entire process of goods storage, sorting, and transportation in industrial stereoscopic warehouses, nor does it provide a solution for multi-dimensional data fusion and complex task scheduling in stereoscopic warehouses. Therefore, it is difficult to directly apply to the AI intelligent management scenario of industrial stereoscopic warehouses.

[0004] However, the above-mentioned existing technology still has room for improvement in terms of overall intelligent management of industrial stereoscopic warehouses, adaptability to complex scenarios, and optimization of the entire process. For example, in dynamic inventory optimization, there is a lack of comprehensive consideration of goods distribution density, storage priority, and turnover rate. In equipment coordination scheduling, path planning algorithms and task allocation strategies are not fully combined to achieve efficient resource utilization. In multi-dimensional data analysis, the application of time series prediction models is also limited. These factors make it difficult for existing technology to fully meet the needs of modern industry for efficient and intelligent warehouse management. Therefore, an AI intelligent management method and system for industrial stereoscopic warehouses are proposed to solve the above problems. SUMMARY

[0005] In order to solve the technical problems in the related field, the present application provides an industrial stereoscopic warehouse AI intelligent management method and system, on the basis of comprehensive multi-dimensional data processing and dynamic task scheduling, through combining path planning algorithm, priority allocation strategy and time series prediction model, the whole process intelligent management of goods storage, sorting and transportation in the warehouse is carried out, so as to optimize the resource utilization efficiency and improve the real-time response ability of the system.

[0006] According to one aspect of the present application, an industrial stereoscopic warehouse AI intelligent management method is provided, the method comprising the following steps: collecting each part of the goods storage information corresponding to each historical time interval before the current time interval of the target warehouse, the current time interval taking the current time as the starting time, and the single part of goods storage information corresponding to each historical time interval being the number of various types of goods received by the target warehouse in the historical time interval, storage location distribution, turnover rate and inventory priority; obtaining various operating parameters of the target warehouse, including the maximum storage capacity of the warehouse, the number of shelves, the number of automated devices and the throughput of the entrance and exit; dynamically optimizing the storage location of the goods in the target warehouse, generating the optimal storage path, and taking the optimal storage path as the output result of the path planning module; using the LSTM time series prediction model to predict the number of various types of goods received by the target warehouse in the current time interval and their storage priority based on the each part of the goods storage information corresponding to each historical time interval before the current time interval of the target warehouse, the various operating parameters of the target warehouse and the duration of the time interval; based on the number of various types of goods and their storage priority obtained by prediction in the current time interval, performing task allocation on the automated devices in the target warehouse, and generating a device scheduling scheme; wherein, collecting each part of the goods storage information corresponding to each historical time interval before the current time interval of the target warehouse, further comprising: each historical time interval before the current time interval and the current time interval together constitute a complete time segment, and the duration of each time interval is equal.

[0007] According to another aspect of the present application, an AI intelligent management system for industrial stereoscopic warehouse is provided, which comprises the following components: a data acquisition module for acquiring respective pieces of goods storage information corresponding to each historical time interval before a current time interval of a target warehouse; a parameter configuration module for obtaining various operating parameters of the target warehouse; a path optimization module connected with the data acquisition module for dynamically optimizing the storage location of goods in the target warehouse and generating an optimal storage path; a prediction analysis module connected with the data acquisition module, the parameter configuration module and the path optimization module respectively, for predicting the number of various types of goods received by the target warehouse in the current time interval and the storage priority thereof based on the respective pieces of goods storage information corresponding to each historical time interval before the current time interval of the target warehouse, the various operating parameters of the target warehouse and the duration of the time interval by using an LSTM time series prediction model; and a task allocation module connected with the prediction analysis module for performing task allocation on the automated equipment in the target warehouse based on the number of various types of goods and the storage priority thereof obtained by prediction in the current time interval, and generating an equipment scheduling scheme; wherein the respective pieces of goods storage information corresponding to each historical time interval before the current time interval of the target warehouse further comprises that each historical time interval before the current time interval and the current time interval together constitute a complete time segment, and the duration of each time interval is equal.

[0008] It can be seen that the present application has at least the following four outstanding substantial features: the first is that the target warehouse in the current time interval just arrived each type of goods quantity and its storage priority intelligent prediction targeted screening multiple basic data, the multiple basic data includes each historical time interval respectively corresponding each part of goods storage information and each item of the target warehouse operation parameter before the current time interval, thereby providing comprehensive and effective basic data for subsequent intelligent prediction of goods storage information in the current time interval; the second is to introduce a path optimization module for executing dynamic optimization of goods storage location in the target warehouse, the path optimization module generates the optimal storage path in combination with the warehouse space layout and the goods storage priority, thereby completing the customized design of the storage path; the third is based on the predicted quantity of each type of goods in the current time interval and its storage priority, the target warehouse executes task allocation for the automation equipment, generates the equipment scheduling scheme, thereby judging the task demand of the automation equipment in advance for the target warehouse, and realizes efficient use of limited resources; the fourth is that in each training of the LSTM time sequence prediction model, the known and received quantity of each type of goods and its storage priority of the target warehouse in a certain past time interval is taken as the output content of the prediction model, and each part of the goods storage information of the target warehouse in each historical time interval respectively corresponding before the certain past time interval, each item of the target warehouse operation parameter and the duration of the time interval are taken as the input content of the prediction model, the training action is completed, thereby ensuring the effectiveness of each training.

[0009] Moreover, the present application further refines the technical solutions through the following specific embodiments: in the data acquisition module, a distributed sensor network is adopted to collect the goods storage information in the target warehouse in real time, the distributed sensor network is composed of a plurality of pressure sensors, temperature sensors and RFID tag readers installed on the shelves, the pressure sensors are used to detect the weight change of the goods, the temperature sensors are used to monitor the storage environment conditions of the goods, and the RFID tag readers are used to record the unique identification information of the goods; in the path optimization module, a two-dimensional grid map is constructed based on the spatial layout map of the target warehouse, each grid unit in the two-dimensional grid map corresponds to a storage area in the warehouse, and the optimal storage path is generated by calculating the minimum cost path from the starting point to the ending point; in the prediction analysis module, an LSTM time series prediction model is adopted to model the historical goods storage information of the target warehouse, the LSTM model contains three hidden layers, and the number of neurons in each layer is 128, 64 and 32 respectively, the input features of the model include the number of goods, the storage location distribution, the turnover rate and the inventory priority, and the output features are the predicted number of goods and the storage priority thereof; in the task allocation module, the task allocation is performed on the automated equipment in the target warehouse, the automated equipment and the to-be-executed tasks are matched by constructing a bipartite graph model, the nodes of the bipartite graph model are divided into equipment nodes and task nodes, the weight of the edge represents the cost of the equipment executing the task, and the equipment scheduling scheme is generated by finding the maximum weight matching.

[0010] To sum up, the present application realizes the intelligent management of the whole process of the industrial stereoscopic warehouse by combining multi-dimensional data processing, path planning algorithm, priority allocation strategy and time series prediction model, optimizes the resource scheduling and task allocation, improves the real-time response capability and multi-task concurrent processing capability of the system, and meets the demand of modern industry for efficient and intelligent warehouse management. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 It is a structure block diagram of the AI intelligent management system of the industrial stereoscopic warehouse of the present application.

[0012] Figure 2 It is a working process schematic diagram of the path optimization module of the present application.

[0013] Figure 3 It is an LSTM time series prediction model structure diagram of the prediction analysis module of the present application.

[0014] Figure 4 It is a task scheduling schematic diagram of the task allocation module of the present application.

[0015] The reference signs are as follows:

[0016] 1, data acquisition module; 2, parameter configuration module; 3, path optimization module; 4, prediction analysis module; 5, task allocation module; 6, distributed sensor network; 7, two-dimensional grid map; 8, LSTM model; 9, bipartite graph model; 10, automation equipment. DETAILED DESCRIPTION

[0017] The present application provides an industrial stereoscopic warehouse AI intelligent management method and system, and its specific implementation is described in detail in combination with the drawings. As shown in the figure, the system includes data acquisition module 1, parameter configuration module 2, path optimization module 3, prediction analysis module 4 and task allocation module 5. These modules are logically connected and interact with data to realize intelligent management of industrial stereoscopic warehouse. Figure 1

[0018] The data acquisition module 1 completes the real-time acquisition of the target warehouse goods storage information through the distributed sensor network 6. The distributed sensor network 6 is composed of multiple pressure sensors, temperature sensors and RFID tag readers installed on the shelves. Among them, the pressure sensor is used to detect the weight change of the goods, the temperature sensor is used to monitor the storage environment condition of the goods, and the RFID tag reader records the unique identification information of the goods. These sensors are distributed in different areas of the warehouse according to the preset spatial layout, and upload the collected data to the data acquisition module 1. The data acquisition module 1 is responsible for sorting and outputting the respective goods storage information of each historical time interval before the current time interval, each information containing the number of various goods, storage location distribution, turnover rate and inventory priority in the corresponding historical time interval. It should be noted that the current time interval and the previous historical time interval together form a complete time segment, and the duration of each time interval is equal.

[0019] The parameter configuration module 2 is used to obtain the running parameters of the target warehouse, including the maximum storage capacity of the warehouse, the number of shelves, the number of automation equipment and the throughput of the entrance and exit. The parameter configuration module 2 updates and stores the above running parameters in real time by interfacing with the warehouse management system. These parameters provide basic support for subsequent path optimization and task scheduling. The parameter configuration module 2 has a data transmission relationship with the path optimization module 3 and the prediction analysis module 4, ensuring that the running parameters can be called in time.

[0020] The path optimization module 3 dynamically optimizes the storage location of the goods in the target warehouse and generates the optimal storage path. As shown in the figure, Figure 2 ​As shown, the path optimization module 3 first constructs a two-dimensional grid map 7 according to the spatial layout of the target warehouse, where each grid cell corresponds to a storage area within the warehouse. The optimal storage path is generated by calculating the minimum cost path from the starting point to the ending point, and the cost function of the path comprehensively considers factors such as distance, storage priority, and shelf number. The path optimization module 3 transmits the generated optimal storage path as an output result to the prediction analysis module 4 for subsequent task allocation.

[0021] The prediction analysis module 4 uses an LSTM time series prediction model 8 to model the historical goods storage information of the target warehouse, thereby predicting the number of each type of goods received by the target warehouse in the current time interval and their storage priority. As shown, Figure 3 As shown, the LSTM model 8 contains three hidden layers, with 128, 64, and 32 neurons in each layer, respectively. The input features of the model include the number of goods, the distribution of storage locations, the turnover rate, and the inventory priority, and the output features are the predicted number of goods and their storage priority. In each training of the LSTM model 8, the known number of each type of goods received by the target warehouse in a past time interval and their storage priority are used as the output content of the model, and the corresponding historical goods storage information of the target warehouse in each historical time interval before the past time interval, the operating parameters of the target warehouse, and the duration of the time interval are used as the input content of the model, to complete this training action. The prediction analysis module 4 transmits the prediction results to the task allocation module 5 to provide a basis for task allocation of automated devices.

[0022] The task allocation module 5 performs task allocation for the automated devices 10 in the target warehouse and generates a device scheduling scheme. As shown, Figure 4 As shown, the automated devices 10 are matched with the tasks to be performed by constructing a bipartite graph model 9. The nodes of the bipartite graph model 9 are divided into device nodes and task nodes, and the weight of the edge represents the cost of the device performing the task. The device scheduling scheme is generated by finding the maximum weight matching, ensuring efficient use of limited resources. The task allocation module 5 combines the prediction results provided by the prediction analysis module 4 and the optimal storage path generated by the path optimization module 3 to finally develop a detailed device scheduling scheme.

[0023] In practical applications, the operation process of the system is as follows: first, the data acquisition module 1 collects the goods storage information of the target warehouse in real time through the distributed sensor network 6, and transmits the data to the prediction analysis module 4 after sorting. At the same time, the parameter configuration module 2 obtains various operating parameters of the target warehouse, and transmits them to the path optimization module 3 and the prediction analysis module 4. The path optimization module 3 generates the optimal storage path and transmits the result to the prediction analysis module 4. The prediction analysis module 4 uses the LSTM model 8 to model the historical data, predicts the number of goods and their storage priority in the current time interval, and transmits the prediction result to the task allocation module 5. The task allocation module 5 allocates tasks to the automated equipment 10, and finally generates a device scheduling scheme. The whole process realizes the intelligent management of the whole process of the industrial vertical warehouse, optimizes the resource scheduling and task allocation.

[0024] In this embodiment, the connection relationship between all modules is realized through a data interface, ensuring efficient data transmission between modules. For example, the data acquisition module 1 and the prediction analysis module 4 are connected through a data bus, and the path optimization module 3 and the prediction analysis module 4 exchange data through a special communication protocol. In addition, the specific function implementation of each module depends on the support of software programs, which run on high-performance servers to ensure the real-time response capability and multi-task concurrent processing capability of the system.

[0025] This embodiment further refines the specific implementation details of each module. For example, in the data acquisition module 1, the arrangement of the distributed sensor network 6 needs to follow certain rules, the installation position of the sensor should cover all storage areas of the warehouse, and the distance between adjacent sensors should meet the signal coverage requirements. In the path optimization module 3, the construction of the two-dimensional grid map 7 needs to be based on the actual spatial layout of the warehouse, and the size of each grid cell should match the actual storage area of the shelf. In the prediction analysis module 4, the training process of the LSTM model 8 needs to use a large-scale historical data set to improve the prediction accuracy of the model. In the task allocation module 5, the construction of the bipartite graph model 9 needs to consider the performance parameters of the automated equipment 10 and the complexity of the task to ensure the rationality of task allocation.

[0026] This embodiment realizes the intelligent management of the whole process of the industrial vertical warehouse by combining multi-dimensional data processing, path planning algorithm, priority allocation strategy and time series prediction model, optimizes resource scheduling and task allocation, improves the real-time response capability and multi-task concurrent processing capability of the system, and meets the demand of modern industry for efficient and intelligent warehouse management.

[0027] In order to better enable relevant persons in the technical field to fully understand and implement the present application, the specific implementation principles of the present application are further supplemented in the following with reference to a specific application scenario.

[0028] In a large industrial warehouse, there are various types of goods and scattered storage locations, and warehouse management faces complex demands such as dynamic inventory optimization, equipment coordination scheduling, and multi-dimensional data analysis. The AI intelligent management system of the application is used for intelligent management of the whole process in the warehouse, and the specific operation process is as follows:

[0029] First, the data acquisition module 1 collects the goods storage information of the target warehouse in real time through the distributed sensor network 6. The pressure sensor in the distributed sensor network 6 detects the weight change of the goods on the shelf, the temperature sensor monitors the storage environment condition, and the RFID tag reader records the unique identification information of the goods. These sensors are distributed in different areas of the warehouse according to the preset spatial layout, and upload the collected data to the data acquisition module 1. The data acquisition module 1 sorts and outputs the respective pieces of goods storage information corresponding to each historical time interval before the current time interval, each piece of information containing the number of each type of goods, the storage location distribution, the turnover rate and the inventory priority in the corresponding historical time interval. For example, in the last hour, the system records that A-class goods are stored on the 3rd layer shelf, B-class goods are stored on the 5th layer shelf, and calculates the respective turnover rates and priorities according to their in-out warehouse frequencies. These basic data provide comprehensive support for subsequent path optimization and task allocation.

[0030] At the same time, the parameter configuration module 2 obtains various operating parameters of the target warehouse, including the maximum storage capacity of the warehouse, the number of shelves, the number of automated devices and the throughput of the entrance and exit. These parameters are updated in real time through the interface with the warehouse management system and transmitted to the path optimization module 3 and the prediction analysis module 4. For example, the system knows that the maximum storage capacity of the warehouse is 1000 tons, the shelves are divided into 10 layers, the number of automated devices is 20, and the throughput of the entrance and exit is 50 tons per hour. These parameters provide necessary constraints for subsequent path planning and task scheduling.

[0031] The path optimization module 3 dynamically optimizes the storage location of the goods in the target warehouse and generates the optimal storage path. As shown in Figure 2 The path optimization module 3 first constructs a two-dimensional grid map 7 according to the spatial layout of the target warehouse, and each grid cell corresponds to a storage area in the warehouse. By considering factors such as distance, storage priority and shelf number, the minimum cost path from the starting point to the ending point is calculated. For example, when a batch of high-priority C-class goods needs to be transported from the entrance to the 7th layer shelf, the system calculates the optimal path as from the entrance to the 4th layer shelf and then to the 7th layer shelf, avoiding congestion caused by low-priority goods occupying the passage. The generated optimal storage path is transmitted to the prediction analysis module 4 as an output result for subsequent task allocation.

[0032] The prediction analysis module 4 uses an LSTM time series prediction model 8 to model the historical goods storage information of the target warehouse, thereby predicting the quantities of various types of goods received by the target warehouse in the current time interval and their storage priorities. As shown in Figure 3 The LSTM model 8 contains three hidden layers, with 128, 64, and 32 neurons in each layer. The input features of the model include the quantity of goods, the distribution of storage locations, the turnover rate, and the inventory priority, and the output features are the predicted quantity of goods and their storage priority. For example, based on the historical data of the past 24 hours, the system predicts that the quantity of class A goods will increase by 10 tons in the current time interval, with a high storage priority, and the quantity of class B goods will decrease by 5 tons, with a medium storage priority. In each training of the LSTM model 8, the known quantities of various types of goods received by the target warehouse in a past time interval and their storage priorities are used as the output of the model, and the corresponding historical goods storage information of the target warehouse in each historical time interval before the past time interval, the operating parameters of the target warehouse, and the duration of the time interval are used as the input of the model, thereby completing the training action. The prediction analysis module 4 transmits the prediction results to the task allocation module 5 to provide a basis for task allocation of the automated equipment.

[0033] The task allocation module 5 performs task allocation for the automated equipment 10 in the target warehouse and generates a device scheduling scheme. As shown in Figure 4 The automated equipment 10 is matched with the tasks to be performed by constructing a bipartite graph model 9. The nodes of the bipartite graph model 9 are divided into device nodes and task nodes, and the weight of the edge represents the cost of the device performing the task. For example, when class A goods need to be transported from the warehouse entrance to the 7th layer of shelves, the system calculates that the cost of device D1 performing this task is the lowest according to the device performance parameters and the complexity of the task. The device scheduling scheme is generated by finding the maximum weight matching, ensuring efficient use of limited resources. The task allocation module 5 combines the prediction results provided by the prediction analysis module 4 and the optimal storage path generated by the path optimization module 3 to finally develop a detailed device scheduling scheme. For example, the system arranges device D1 to be responsible for the warehousing task of class A goods and device D2 to be responsible for the out-of-warehouse task of class B goods, and guides the devices to complete the tasks quickly through the optimal path provided by the path optimization module 3.

[0034] In the entire process, the connection between all modules is realized through a data interface, ensuring efficient transmission of data between modules. For example, the data acquisition module 1 and the prediction analysis module 4 are connected through a data bus, and the path optimization module 3 and the prediction analysis module 4 exchange data through a special communication protocol. In addition, the specific function implementation of each module depends on the support of software programs, which run on high-performance servers to ensure the real-time response capability and multi-task concurrent processing capability of the system.

[0035] Through the above steps, the system realizes the intelligent management of the whole process of the industrial stereoscopic warehouse. For example, in a high-load scenario, the system can dynamically adjust the storage path and equipment scheduling scheme according to the real-time collected goods storage information and the predicted future demand, avoiding the problem of low efficiency caused by unreasonable resource allocation. At the same time, the system combines multi-dimensional data processing, path planning algorithm, priority allocation strategy and time series prediction model, significantly improving the efficiency of resource scheduling and task allocation, meeting the demand of modern industry for efficient and intelligent warehouse management.

[0036] In summary, the present application is realized through specific operation steps and algorithms, which solves the problems of insufficient dynamic inventory optimization, low efficiency of equipment collaborative scheduling and limited multi-dimensional data analysis capability in traditional warehouse management, and provides reliable technical support for intelligent management of industrial stereoscopic warehouse.

[0037] It should be noted that in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0038] Although the embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. An AI intelligent management method for an industrial stereoscopic warehouse, characterized in that, The method comprises the following steps: Collecting each part of the goods storage information corresponding to each historical time interval before the current time interval of the target warehouse, the current time interval taking the current time as the starting time, and the single part of the goods storage information corresponding to each historical time interval being the number of each type of goods received by the target warehouse, the storage location distribution, the turnover rate and the inventory priority in the historical time interval; Obtaining each operating parameter of the target warehouse, including the maximum storage capacity of the warehouse, the number of shelves, the number of automated equipment and the throughput of the entrance and exit; Dynamically optimizing the storage location of the goods in the target warehouse, generating an optimal storage path, and taking the optimal storage path as the output result of the path planning module; Using an LSTM time series prediction model to predict the number of each type of goods received by the target warehouse in the current time interval and its storage priority based on each part of the goods storage information corresponding to each historical time interval before the current time interval of the target warehouse, each operating parameter of the target warehouse and the duration of the time interval; Based on the number of each type of goods and its storage priority obtained by prediction in the current time interval, task allocation is performed on the automated equipment in the target warehouse to generate a device scheduling scheme. 2.The industrial stereoscopic warehouse AI intelligent management method according to claim 1, characterized in that: Using an LSTM time series prediction model to predict the number of each type of goods received by the target warehouse in the current time interval and its storage priority based on each part of the goods storage information corresponding to each historical time interval before the current time interval of the target warehouse, each operating parameter of the target warehouse and the duration of the time interval includes: taking the number of each type of goods received by the target warehouse in a certain past time interval and its storage priority as the output content of the LSTM time series prediction model, taking each part of the goods storage information corresponding to each historical time interval before the certain past time interval of the target warehouse, each operating parameter of the target warehouse and the duration of the time interval as each input content of the LSTM time series prediction model, and completing this training action. 3.The industrial stereoscopic warehouse AI intelligent management method according to claim 1, characterized in that: Based on the number of each type of goods and its storage priority obtained by prediction in the current time interval, task allocation is performed on the automated equipment in the target warehouse to generate a device scheduling scheme, which includes: constructing a bipartite graph model, the nodes of the bipartite graph model being divided into device nodes and task nodes, the weight of the edge representing the cost of the device executing the task, and the device scheduling scheme being generated by finding the maximum weight matching. 4.The industrial stereoscopic warehouse AI intelligent management method according to claim 1, characterized in that: Collecting each part of the goods storage information corresponding to each historical time interval before the current time interval of the target warehouse includes: using a distributed sensor network to collect the goods storage information in the target warehouse in real time, the distributed sensor network being composed of multiple pressure sensors, temperature sensors and RFID tag readers installed on the shelves, the pressure sensors being used to detect the weight change of the goods, the temperature sensors being used to monitor the storage environment conditions of the goods, and the RFID tag readers being used to record the unique identification information of the goods.

5. The industrial three-dimensional warehouse AI intelligent management method according to claim 3, characterized in that: The LSTM time series prediction model comprises three hidden layers, the number of neurons in each layer being 128, 64 and 32 respectively, the input features of the model including the number of goods, the storage location distribution, the turnover rate and the inventory priority, and the output features being the predicted number of goods and the storage priority thereof.

6. An industrial stereoscopic warehouse AI intelligent management system, characterized in that, The system comprises the following components: a data acquisition module configured to acquire respective pieces of goods storage information corresponding to respective historical time intervals before a current time interval of a target warehouse; a parameter configuration module configured to acquire respective operating parameters of the target warehouse; a path optimization module connected to the data acquisition module and configured to dynamically optimize the storage locations of the goods in the target warehouse and generate an optimal storage path; a prediction analysis module connected to the data acquisition module, the parameter configuration module and the path optimization module respectively and configured to predict the number of goods of various types received by the target warehouse in the current time interval and the storage priority thereof based on the respective pieces of goods storage information corresponding to respective historical time intervals before the current time interval of the target warehouse, the respective operating parameters of the target warehouse and the duration of the time interval by using an LSTM time series prediction model; a task allocation module connected to the prediction analysis module and configured to perform task allocation for the automated devices in the target warehouse based on the predicted number of goods of various types in the current time interval and the storage priority thereof and generate a device scheduling scheme. In addition, the respective pieces of goods storage information corresponding to respective historical time intervals before the current time interval of the target warehouse further comprise that the respective historical time intervals before the current time interval and the current time interval jointly constitute a complete time section, and the duration of each time interval is equal. 7.The industrial stereoscopic warehouse AI intelligent management system according to claim 6, characterized in that: The data acquisition module acquires the real-time goods storage information in the target warehouse by using a distributed sensor network, the distributed sensor network comprises a plurality of pressure sensors, temperature sensors and RFID tag readers installed on shelves, the pressure sensors are configured to detect the weight change of the goods, the temperature sensors are configured to monitor the storage environment conditions of the goods, and the RFID tag readers are configured to record the unique identification information of the goods. 8.The industrial stereoscopic warehouse AI intelligent management system according to claim 6, characterized in that: The path optimization module constructs a two-dimensional grid map based on a space layout map of the target warehouse, each grid cell in the two-dimensional grid map corresponds to a storage area in the warehouse, and the optimal storage path is generated by calculating the minimum cost path from the starting point to the ending point. 9.The industrial stereoscopic warehouse AI intelligent management system according to claim 6, characterized in that: The prediction analysis module models the historical goods storage information of the target warehouse by using an LSTM time series prediction model, the LSTM model comprises three hidden layers, the number of neurons in each layer being 128, 64 and 32 respectively, the input features of the model including the number of goods, the storage location distribution, the turnover rate and the inventory priority, and the output features being the predicted number of goods and the storage priority thereof.

Images