Warehouse management system based on artificial intelligence

Abstract

The invention relates to the technical field of intelligent warehousing, and discloses an artificial intelligence-based warehouse management system, which comprises a cargo information acquisition module, an intelligent warehousing planning module, an automatic carrying scheduling module, a dynamic inventory management module and an abnormity early warning module, the cargo information acquisition module is used for acquiring basic attribute information, position information and state information of cargoes; and the intelligent storage planning module is electrically connected with the cargo information acquisition module, and is used for constructing a warehouse layout optimization model and planning a cargo storage position based on the cargo attribute information and the warehouse space data. Through the intelligent warehouse planning module, a reinforcement learning algorithm is adopted, the cargo turnover rate, the storage safety and the space utilization rate are taken as optimization targets, different storage strategies are simulated, iterative training is carried out, and an optimal warehouse layout scheme is generated, so that compared with a traditional fixed rule planning mode, cargo characteristics and warehouse demand changes can be dynamically adapted; the warehouse space utilization rate is effectively improved, and storage resource waste is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent warehousing, in particular to a warehouse management system based on artificial intelligence. BACKGROUND

[0002] With the continuous innovation of Internet of Things and artificial intelligence technology, warehouse management systems are gradually transforming towards intelligentization. Modern warehouses realize preliminary collection of cargo information and basic monitoring of storage environment by deploying sensors, automated equipment and information management platforms, which to some extent improves the efficiency of warehousing operations. However, in actual application, the existing warehouse management system still has some problems.

[0003] In terms of warehouse planning, the traditional system is mostly based on fixed rules or manual experience to divide storage areas, which is difficult to adapt to the dynamic changes of cargo turnover rate, storage safety and space utilization rate in real time, resulting in that the warehouse space resources cannot be optimally configured. In the cargo handling link, the path planning and scheduling of some automated handling equipment lack comprehensive consideration of real-time environmental data and equipment status, which is prone to path conflicts, equipment idling or excessive load, etc., affecting the overall handling efficiency. At the inventory management level, most systems rely on simple data statistical methods to predict cargo in-out quantity, which cannot accurately capture the complex rules in historical inventory data and sales data, resulting in frequent inventory accumulation or shortage, increasing warehousing cost and reducing order fulfillment rate. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a warehouse management system based on artificial intelligence, which solves the problems mentioned in the background.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: a warehouse management system based on artificial intelligence, comprising: a cargo information acquisition module, an intelligent warehousing planning module, an automated handling scheduling module, an inventory dynamic management module and an abnormality early warning module;

[0006] The cargo information acquisition module is used to acquire the basic attribute information, position information and state information of the cargo.

[0007] The intelligent warehousing planning module is electrically connected with the cargo information acquisition module, and is used to construct a warehouse layout optimization model based on cargo attribute information and warehouse space data, and plan the storage position of the cargo.

[0008] The automated handling scheduling module is used to receive cargo handling instructions, combine real-time environmental data and equipment status of the warehouse, and generate optimal handling paths and scheduling schemes.

[0009] The inventory dynamic management module is electrically connected with the goods information acquisition module and the automatic handling scheduling module, and is used for updating inventory data in real time, analyzing inventory change trend, and predicting replenishment demand.

[0010] The abnormality early warning module is used for monitoring abnormal conditions in the warehouse operation process, including goods abnormality, equipment failure, and environmental abnormality, generating early warning information and pushing.

[0011] Preferably, the goods information acquisition module comprises an RFID identification unit, an image acquisition unit and a sensor group.

[0012] The RFID identification unit is used for identifying the electronic tags of goods and acquiring basic attribute information of the goods.

[0013] The image acquisition unit adopts a high-definition camera to acquire images of the appearance of goods, and identify the packaging state and damage condition of the goods.

[0014] The sensor group comprises a temperature and humidity sensor and a weight sensor, and is used for monitoring environmental parameters and real-time weight information of goods storage.

[0015] Preferably, the intelligent warehouse planning module adopts a reinforcement learning algorithm, takes the goods turnover rate, storage safety and space utilization rate as optimization targets, simulates different storage strategies, iteratively trains, and generates an optimal warehouse layout scheme.

[0016] The warehouse layout optimization model partitions the warehouse space according to the size, weight and storage requirement of goods, divides different storage areas, and determines the storage priority of each area.

[0017] Preferably, the reinforcement learning algorithm of the intelligent warehouse planning module is specifically as follows:

[0018] The warehouse is divided into grid units to construct a state space, and each unit state comprises idle / occupied, goods type, storage time and turnover rate level information.

[0019] The action space is set as a combination of goods placement and transfer operations.

[0020] The reward function is composed of a turnover rate difference comparison reward, a safety reward based on the goods danger level and safety distance, a space utilization rate reward according to the proportion of idle areas in the warehouse, and a supplementary reward related to the length of the handling path.

[0021] A deep Q network architecture is adopted for policy optimization, an experience replay buffer is used to reduce sample correlation, and a double network structure is used for stable training.

[0022] The random warehouse layout is taken as an initial state, and the optimal layout scheme is generated through multiple rounds of action interaction and network parameter updating.

[0023] Preferably, the automated handling scheduling module comprises a path planning unit and a device scheduling unit.

[0024] The path planning unit adopts an improved Dijkstra algorithm combined with real-time environmental data to plan a collision-free and efficient handling path for the automated handling device.

[0025] The device scheduling unit cooperatively schedules multiple automated handling devices based on the running state, load condition and task priority of the devices to achieve reasonable allocation of tasks.

[0026] Preferably, the improved Dijkstra algorithm introduces a heuristic function to preferentially explore nodes closer to the target by evaluating the estimated distance from the node to the target node, thereby reducing the search range.

[0027] Meanwhile, according to the real-time obtained position and size information of obstacles in the warehouse, the path weight between nodes is dynamically adjusted, and for nodes close to obstacles, the path weight between adjacent nodes is significantly increased, so that the algorithm actively avoids obstacles during the search process.

[0028] Preferably, the inventory dynamic management module constructs an inventory prediction model, which is based on historical inventory data and sales data and uses a time series analysis algorithm to predict the amount of goods in and out of the warehouse in the future period.

[0029] The inventory dynamic management module automatically generates a replenishment application form according to the prediction result and the preset inventory threshold, and sends it to the procurement department.

[0030] Preferably, the construction process of the inventory prediction model is as follows:

[0031] The historical inventory data and sales data are divided into training set and test set according to time dimension;

[0032] An autoregressive integrated moving average model is used to analyze the training set data, identify the stationarity of the data, determine the difference order of the model, and use the partial autocorrelation function and autocorrelation function to determine the autoregressive order and moving average order;

[0033] In the model training process, the model parameters are optimized by minimizing the mean square error;

[0034] Finally, the test set is used to verify the trained model, evaluate the prediction accuracy of the model, and adjust the model parameters according to the verification result.

[0035] Preferably, the abnormality early warning module is provided with a multi-level early warning mechanism, including general abnormality, serious abnormality and emergency abnormality;

[0036] For general abnormality, such as slight damage of goods packaging, the system sends prompt information to the warehouse manager;

[0037] For serious abnormality, such as the running parameters of the equipment exceeding the normal range, the system generates an alarm information and suspends the operation of the related equipment;

[0038] For emergency abnormality, such as warehouse fire and flood, the system immediately starts the emergency handling program and notifies the relevant emergency department.

[0039] Preferably, it also includes a data management module for storing, cleaning and analyzing various types of collected data; the data management module adopts distributed storage technology, establishes data index and improves data query and retrieval efficiency;

[0040] The data management module analyzes the circulation law of goods and the use efficiency of equipment through data mining algorithm, and provides decision support for warehouse management.

[0041] The present application provides a warehouse management system based on artificial intelligence. It has the following beneficial effects:

[0042] The intelligent warehouse planning module adopts reinforcement learning algorithm, takes the turnover rate of goods, storage safety and space utilization rate as the optimization target, simulates different storage strategies and iteratively trains to generate the optimal warehouse layout scheme. Compared with the traditional fixed rule planning mode, the system can dynamically adapt to the changes of goods characteristics and warehouse demand, effectively improve the warehouse space utilization rate and reduce the waste of warehouse resources.

[0043] The path planning unit improves the Dijkstra algorithm, combines real-time environmental data, introduces heuristic function and dynamic weight adjustment mechanism, and plans collision-free and efficient carrying path for the automated handling equipment; the device scheduling unit cooperatively schedules according to the device state, load and task priority, improves the operation efficiency of the handling equipment, reduces the idle running and path conflict of the equipment, and greatly improves the overall efficiency of the warehouse goods handling.

[0044] The inventory dynamic management module constructs an inventory prediction model based on historical inventory data and sales data, uses time series analysis algorithm to deeply process and feature extract the data, accurately predicts the future goods in and out warehouse quantity, and compared with the traditional simple statistical prediction method, improves the prediction accuracy, automatically generates a replenishment application form combined with the preset inventory threshold, effectively reduces the inventory backlog cost, and reduces the order delay phenomenon caused by lack of goods. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The system architecture and workflow of the present application are shown in the figure. DETAILED DESCRIPTION

[0046] The technical solutions of the present application will be described clearly and completely below in combination with the drawings of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0047] Embodiment:

[0048] Please refer to the accompanying Figure 1 The warehouse management system based on artificial intelligence provided by the embodiment of the present application comprises a cargo information acquisition module, an intelligent warehouse planning module, an automatic carrying scheduling module, an inventory dynamic management module and an abnormality early warning module.

[0049] The cargo information acquisition module is used to acquire basic attribute information, position information and state information of the cargo.

[0050] The intelligent warehouse planning module is electrically connected with the cargo information acquisition module, and is used to construct a warehouse layout optimization model based on cargo attribute information and warehouse space data, and plan a cargo storage position.

[0051] The automatic carrying scheduling module is used to receive a cargo carrying instruction, combine real-time environment data and equipment states of the warehouse, and generate an optimal carrying path and a scheduling scheme.

[0052] The inventory dynamic management module is electrically connected with the cargo information acquisition module and the automatic carrying scheduling module, and is used to update inventory data in real time, analyze inventory change trends and predict replenishment demand.

[0053] The abnormality early warning module is used to monitor abnormal conditions in the warehouse running process, including cargo abnormality, equipment failure and environmental abnormality, generate early warning information and push the early warning information.

[0054] Specifically, the cargo information acquisition module collects various basic information of the cargo as a data source; the intelligent warehouse planning module constructs a layout optimization model according to the collected data and warehouse space data; the automatic carrying scheduling module generates a carrying scheme in combination with environment and equipment states; the inventory dynamic management module updates and analyzes inventory data in real time; and the abnormality early warning module monitors abnormal conditions and pushes early warning. The modules form a complete warehouse management closed loop through data interaction and algorithm driving.

[0055] Compared with the traditional warehouse management mode, the whole process of data collection to exception handling is realized intelligent, avoiding the inefficiency and high failure rate of manual intervention, and through the cooperation of multiple modules, it can quickly respond to various changes in the warehouse, significantly improve the overall operation efficiency and management level of the warehouse, reduce the operation cost, and enhance the competitiveness of enterprises in the field of logistics and warehousing.

[0056] The cargo information collection module includes an RFID identification unit, an image collection unit and a sensor group;

[0057] The RFID identification unit is used for electronic tag identification of the cargo to obtain the basic attribute information of the cargo;

[0058] The image collection unit uses a high-definition camera to collect images of the appearance of the cargo, and identifies the packaging state and damage of the cargo;

[0059] The sensor group includes a temperature and humidity sensor and a weight sensor, which are used to monitor the storage environment parameters and real-time weight information of the cargo.

[0060] Specifically, the RFID identification unit uses radio frequency identification technology to non-contact read the basic attribute information of the cargo through an electronic tag; the image collection unit uses a high-definition camera to collect and analyze images of the appearance of the cargo based on computer vision technology; the sensor group monitors the storage environment parameters and weight information of the cargo in real time. Multiple data collection methods complement each other to ensure the comprehensiveness and accuracy of the collected data.

[0061] Compared with a single collection method, the multi-dimensional data collection method can more accurately and comprehensively obtain cargo information and timely discover potential problems of the cargo, such as packaging damage and abnormal environmental parameters, providing a reliable data basis for subsequent warehouse planning and inventory management, avoiding cargo management errors caused by missing or inaccurate information, and improving the refinement of warehouse cargo management.

[0062] The intelligent warehouse planning module uses a reinforcement learning algorithm, takes the cargo turnover rate, storage safety and space utilization rate as the optimization target, simulates different storage strategies, iteratively trains, and generates the optimal warehouse layout scheme;

[0063] The warehouse layout optimization model partitions the warehouse space according to the size, weight and storage requirements of the cargo, divides different storage areas, and determines the storage priority of each area.

[0064] Specifically, taking the reinforcement learning algorithm as the core, this module takes the cargo turnover rate, storage safety and space utilization rate as the optimization target, simulates different storage strategies, iteratively trains, makes the algorithm learn the optimal strategy in a large amount of data interaction, partitions the warehouse space according to the characteristics of the cargo, determines the storage priority, and realizes the reasonable planning of the warehouse space.

[0065] The traditional fixed rule or manual experience warehouse planning mode is changed, the layout can be optimized in real time according to the dynamic change of the demand of goods, the warehouse space utilization rate is improved, the goods storage confusion is reduced, the handling cost and goods loss caused by unreasonable layout are reduced, and the warehouse resources are more efficiently configured.

[0066] The reinforcement learning algorithm of the intelligent warehouse planning module is specifically:

[0067] The warehouse is divided into grid units to construct a state space, and each unit state contains idle / occupied, goods type, storage time and turnover rate level information;

[0068] The action space is set as the operation combination of goods placement and transfer;

[0069] The reward function is composed of a turnover rate difference comparison reward, a safety reward based on the danger level of goods and a safe distance, a space utilization rate reward according to the proportion of idle areas in the warehouse, and a supplementary reward related to the length of the handling path;

[0070] A deep Q network architecture is used for policy optimization, an experience replay buffer is used to reduce sample correlation, and a double network structure is used for stable training;

[0071] Taking a random warehouse layout as the initial state, after multiple rounds of action interaction and network parameter update, the optimal layout scheme is generated, and the weights of each reward term are dynamically adjusted according to the real-time warehouse goods type, so as to realize the adaptive change of the optimization objective function.

[0072] Specifically, compared with the traditional warehouse planning algorithm, the reinforcement learning algorithm can more comprehensively consider various factors in warehouse management, dynamically adapt to different goods and warehouse scenarios, realize precise optimization of warehouse layout, further improve the turnover rate of goods, enhance the storage safety, reduce the operation cost, and make the warehouse always maintain efficient operation in the complex and changeable business environment.

[0073] The automatic handling scheduling module includes a path planning unit and a device scheduling unit;

[0074] The path planning unit uses an improved Dijkstra algorithm combined with real-time environmental data to plan a collision-free and efficient handling path for the automatic handling device;

[0075] The device scheduling unit cooperatively schedules multiple automatic handling devices based on the running state, load condition and task priority of the device, so as to realize reasonable allocation of tasks.

[0076] Specifically, the path planning unit adopts an improved Dijkstra algorithm to plan the carrying path in combination with real-time environmental data; and the device scheduling unit schedules multiple automated carrying devices in coordination according to the device operating state, load condition and task priority, to ensure efficient completion of the carrying task.

[0077] The problem of unreasonable path planning and scheduling of traditional automated carrying devices is solved, path conflicts, device idling or excessive loading and the like are avoided, the efficiency of goods carrying is improved, the carrying time is shortened, the energy consumption of the devices is reduced, the automation operation level of the warehouse is improved, and the warehouse logistics is more smooth and efficient.

[0078] The improved Dijkstra algorithm introduces a heuristic function to evaluate the estimated distance from a node to a target node, preferentially explore nodes closer to the target, and reduce the search range.

[0079] Meanwhile, according to the real-time acquired position and size information of obstacles in the warehouse, the path weight between nodes is dynamically adjusted, and for nodes close to obstacles, the path weight between the node and adjacent nodes is significantly increased, so that the algorithm actively avoids obstacles in the search process.

[0080] Specifically, the heuristic function is introduced to evaluate the estimated distance from a node to a target node, preferentially explore nodes closer to the target, and reduce the search range; and according to the obstacle information in real-time environmental data, the path weight between nodes is dynamically adjusted, which can actively avoid obstacles, thereby generating a collision-free and efficient carrying path.

[0081] Compared with the traditional Dijkstra algorithm, the improved algorithm can plan the optimal carrying path more quickly and accurately in a complex warehouse environment, reduce the time consumption of carrying devices in path planning, improve the operating efficiency of the devices, reduce the collision risk in the goods carrying process, ensure the safety of goods and devices, and further improve the safety and efficiency of warehouse carrying operations.

[0082] The inventory dynamic management module constructs an inventory prediction model, which is based on historical inventory data and sales data, uses a time series analysis algorithm to predict the amount of goods in and out of the warehouse in a future period of time; and the inventory dynamic management module automatically generates a replenishment application form according to the prediction result and a preset inventory threshold, and sends it to the procurement department.

[0083] Specifically, by constructing an inventory prediction model, based on historical inventory data and sales data, using a time series analysis algorithm, the data is deeply processed and feature extracted to predict the amount of goods in and out of the warehouse in the future. Then, in combination with the preset inventory threshold, a replenishment application form is automatically generated to realize dynamic management of inventory.

[0084] It can accurately capture the complex rules in inventory data and sales data, effectively avoid inventory accumulation or shortage, reduce warehouse costs, improve order fulfillment rate, and realize fine management of inventory.

[0085] The construction process of the inventory prediction model is as follows:

[0086] The historical inventory data and sales data are divided into training set and test set according to time dimension;

[0087] The autoregressive integrated moving average model is used to analyze the training set data, the stationarity of the data is identified, the difference order of the model is determined, and the autoregressive order and moving average order are determined by using the partial autocorrelation function and autocorrelation function;

[0088] In the model training process, the model parameters are optimized by minimizing the mean square error;

[0089] Finally, the test set is used to verify the trained model, evaluate the prediction accuracy of the model, and adjust the model parameters according to the verification results.

[0090] Specifically, first, the historical inventory data and sales data are divided into training set and test set according to time dimension; then the autoregressive integrated moving average model is used to analyze the training set data, the difference order is determined by identifying the stationarity of the data, and the autoregressive order and moving average order are determined by using the partial autocorrelation function and autocorrelation function; in the training process, the model parameters are optimized by minimizing the mean square error; finally, the test set is used to verify the accuracy of the model, and the parameters are adjusted according to the results to ensure the reliability of the model.

[0091] Compared with traditional simple prediction methods, this model can more accurately predict the amount of goods in and out of the warehouse, provide strong support for enterprise inventory decision-making, reduce inventory management errors caused by inaccurate prediction, reduce enterprise operation risk, and improve the intelligent level of enterprise inventory management.

[0092] The abnormality early warning module has a multi-level early warning mechanism, including general abnormality, serious abnormality and emergency abnormality;

[0093] For general abnormalities, such as slight damage to goods packaging, the system sends a prompt message to the warehouse manager;

[0094] For serious abnormalities, such as equipment operating parameters exceeding the normal range, the system generates an alarm message and suspends the operation of the related equipment;

[0095] For emergency abnormalities, such as warehouse fire, flood, the system immediately starts the emergency handling program and notifies the relevant emergency departments.

[0096] Specifically, by setting a multi-level early warning mechanism, abnormal conditions are divided into general abnormalities, serious abnormalities and emergency abnormalities, and different handling measures are taken for different levels of abnormalities, such as sending prompt information, generating alarms and suspending equipment, starting emergency handling programs, etc., to realize hierarchical management of warehouse abnormal conditions. Compared with a single abnormality monitoring method, the multi-level early warning mechanism can more comprehensively and timely discover various abnormal conditions in the warehouse operation process.

[0097] It also includes a data management module for storing, cleaning and analyzing various types of collected data; the data management module uses distributed storage technology to establish data indexes and improve data query and retrieval efficiency;

[0098] The data management module analyzes the circulation rules of goods and the use efficiency of equipment through data mining algorithms to provide decision support for warehouse management.

[0099] Specifically, distributed storage technology is used to store various types of collected data, and data indexes are established to improve query and retrieval efficiency; data mining algorithms are used to analyze the data to mine information such as the circulation rules of goods and the use efficiency of equipment, and to provide decision support for warehouse management. Efficient management and deep utilization of data help enterprise managers better understand the warehouse operation status and discover potential problems and optimization space.

[0100] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made to these 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. Warehouse management system based on artificial intelligence, characterized in that, The application relates to a warehouse management system. The system comprises a cargo information acquisition module, an intelligent warehouse planning module, an automated handling and scheduling module, an inventory dynamic management module and an abnormality early warning module. The cargo information acquisition module is used for acquiring basic attribute information, position information and state information of cargos. The intelligent warehouse planning module is electrically connected with the cargo information acquisition module and is used for constructing a warehouse layout optimization model based on cargo attribute information and warehouse space data, and planning a cargo storage position. The automated handling and scheduling module is used for receiving a cargo handling instruction, combining real-time environment data and equipment states, and generating an optimal handling path and a scheduling scheme. The inventory dynamic management module is electrically connected with the cargo information acquisition module and the automated handling and scheduling module, and is used for updating inventory data in real time, analyzing inventory change trends and predicting replenishment requirements. The abnormality early warning module is used for monitoring abnormal conditions in a warehouse operation process, including cargo abnormality, equipment failure and environmental abnormality, generating early warning information and pushing the early warning information.

2. The warehouse management system based on artificial intelligence according to claim 1, characterized in that, The cargo information acquisition module comprises an RFID identification unit, an image acquisition unit and a sensor group. The RFID identification unit is used for identifying an electronic tag of a cargo and acquiring basic attribute information of the cargo. The image acquisition unit adopts a high-definition camera to acquire images of a cargo appearance, identify a packaging state and damage condition of the cargo. The sensor group comprises a temperature and humidity sensor and a weight sensor, and is used for monitoring cargo storage environment parameters and real-time weight information. 3.The artificial intelligence-based warehouse management system of claim 1, wherein, The intelligent warehouse planning module adopts a reinforcement learning algorithm, takes a cargo turnover rate, storage safety and space utilization rate as optimization targets, generates an optimal warehouse layout scheme through simulation of different storage strategies and continuous iterative training. The warehouse layout optimization model partitions warehouse space according to sizes, weights and storage requirements of cargos, divides different storage areas and determines a cargo storage priority of each area.

4. The warehouse management system based on artificial intelligence according to claim 3, characterized in that, The reinforcement learning algorithm of the intelligent warehouse planning module specifically comprises the following steps: A warehouse is divided into grid units to construct a state space, and state of each unit comprises idle / occupied, cargo type, storage time and turnover rate grade information. An action space is set as an operation combination of cargo placement and transfer. A reward function is composed of a turnover rate difference comparison reward, a safety reward based on a cargo danger grade and a safety distance, a space utilization rate reward according to an idle area proportion of the warehouse and a supplementary reward related to a handling path length. A deep Q network architecture is adopted for policy optimization, an experience replay buffer is used to reduce sample correlation, and a double-network structure is used for stable training. A random warehouse layout is taken as an initial state, an optimal layout scheme is generated through multiple rounds of action interaction and network parameter updating, and each reward item weight is dynamically adjusted according to real-time incoming cargo types, so that adaptive changes of an optimization target function are realized. 5.The artificial intelligence-based warehouse management system according to claim 1, wherein, The automated handling and scheduling module comprises a path planning unit and a device scheduling unit. The path planning unit adopts an improved Dijkstra algorithm combined with real-time environment data to plan a collision-free and efficient handling path for an automated handling device. The device scheduling unit cooperatively schedules multiple automated handling devices based on the running state, load condition and task priority of the devices, so as to realize reasonable allocation of tasks.

6. The warehouse management system based on artificial intelligence according to claim 5, characterized in that, The improved Dijkstra algorithm introduces a heuristic function, which evaluates the estimated distance from a node to the target node, preferentially explores nodes closer to the target, and reduces the search range. Meanwhile, according to the real-time acquired position and size information of the obstacles in the warehouse, the path weight between nodes is dynamically adjusted, and for the nodes close to the obstacles, the path weight between the nodes and adjacent nodes is significantly increased, so that the algorithm actively avoids obstacles in the search process. 7.The artificial intelligence-based warehouse management system of claim 1, wherein, The inventory dynamic management module constructs an inventory prediction model, which is based on historical inventory data and sales data, uses a time series analysis algorithm, and predicts the amount of goods in and out of the warehouse in the future period; The inventory dynamic management module automatically generates a replenishment application form according to the prediction result and the preset inventory threshold, and sends it to the procurement department.

8. The artificial intelligence-based warehouse management system according to claim 7, characterized by, The construction process of the inventory prediction model is as follows: The historical inventory data and sales data are divided into training set and test set according to time dimension; An autoregressive integrated moving average model is used to analyze the training set data, identify the stationarity of the data, determine the difference order of the model, and use the partial autocorrelation function and autocorrelation function to determine the autoregressive order and moving average order; In the model training process, the model parameters are optimized by minimizing the mean square error; Finally, the test set is used to verify the trained model, evaluate the prediction accuracy of the model, and adjust the model parameters according to the verification result. 9.The artificial intelligence-based warehouse management system of claim 1, wherein, The abnormal early warning module is provided with a multi-level early warning mechanism, including general abnormality, serious abnormality and emergency abnormality; For general abnormalities, such as slight damage to goods packaging, the system sends a prompt message to the warehouse manager; For serious abnormalities, such as device operating parameters exceeding the normal range, the system generates an alarm and suspends the operation of the related device; For emergency abnormalities, such as warehouse fire and flood, the system immediately starts the emergency handling program and notifies the relevant emergency departments. 10.The warehouse management system based on artificial intelligence according to claim 1, wherein, It also includes a data management module for storing, cleaning and analyzing various collected data; the data management module uses distributed storage technology to establish data index and improve data query and retrieval efficiency; The data management module analyzes the circulation rules of goods and the use efficiency of equipment through data mining algorithm, and provides decision support for warehouse management.

Images