Warehouse inventory management and optimization method based on artificial intelligence

Abstract

The invention relates to a warehouse inventory management and optimization method based on artificial intelligence, and the method comprises the steps: constructing a novel mathematical model through comprehensive consideration of the storage position of a commodity, tray partition, order priority, system state and other factors, solving through employing a deep reinforcement learning DQN network, achieving the intelligent decision of a commodity pickup position, and generating an optimal picking path; according to the method provided by the invention, the order processing time can be effectively shortened, aiming at the requirements of small-batch and multi-category orders, the storage efficiency is improved, the movement distance of a stacker and the number of goods taking times are reduced, the operation cost is reduced, and flexible scheduling can be performed according to the priority to adapt to the real-time state change of the system; compared with a traditional method, after the optimization strategy is adopted, the number of times of taking out the trays is remarkably reduced, and the method has the remarkable beneficial effects, can be widely applied to the field of intelligent warehousing, improves the automation, informatization and intelligentization levels of warehousing management and meets the requirements of the modern logistics industry for an efficient warehousing system.

Description

Technical Field

[0001] The present invention relates to the field of intelligent warehousing technology, and in particular to a warehouse inventory management and optimization method based on artificial intelligence. Background Art

[0002] With the rapid development of information technology and the booming e-commerce, the logistics industry is undergoing unprecedented transformation. Against this backdrop, intelligent warehousing systems have attracted widespread attention for their advantages in improving logistics efficiency and reducing operating costs. By integrating advanced automation equipment, Internet of Things (IoT) technology, big data analysis, and artificial intelligence (AI) algorithms, intelligent warehousing systems achieve automation, informatization, and intelligence in warehouse management, thereby improving the speed and accuracy of logistics distribution. In many warehousing application scenarios, small-batch, multi-category orders typically involve multiple commodities, each with low demand but a wide variety of categories. The complexity and variability of these commodities place higher demands on the picking, sorting, and distribution of warehousing systems. Traditional warehousing systems focus more on the storage and distribution of large-scale, single-category goods and often struggle to efficiently cope with the demands of small-batch, multi-category orders. Furthermore, irrational storage location allocation leads to lengthy product search times and complex picking routes, reducing overall efficiency. Furthermore, traditional warehousing systems lack the flexibility to schedule orders based on priority and have poor adaptability to real-time system status changes. Summary of the Invention

[0003] In view of this, the present invention provides an artificial intelligence-based warehouse inventory management and optimization method, which uses artificial intelligence algorithms to solve the optimal picking plan for the total number of goods required by users, effectively solving the problem that existing warehouse systems are difficult to efficiently respond to small-batch, multi-category order demands, shortening order processing time, improving warehouse efficiency, and reducing operating costs.

[0004] To achieve the above objectives, the present invention provides an artificial intelligence-based warehouse inventory management and optimization method, comprising the following steps: S1. Build a warehouse inventory grid model using a grid, including picking stations, shelves, and lanes. Each shelf contains several pallets, each pallet contains several partitions, and each partition stores a specific SKU. Mark the picking station locations, shelf locations, and the locations and partitions within each pallet on the shelf in the grid model. S2, obtain state space and action space; S3. Determine the objective function and set constraints to establish an order picking optimization model; S301, minimizing the number of pallets required and maximizing the clearance of the entire pallet as the objective function; S302. Setting constraints to meet order requirements, including order requirement constraints and pallet inventory limit constraints; S303: Establish an order picking plan optimization model, and the expression is: ; in, represents the objective function of the number of pallets required, represents the objective function of clearing the entire plate, represents the weight coefficient; S4. Use the deep Q network DQN to train the order picking solution optimization model to generate a picking strategy; S5. Evaluate and optimize the picking strategy, and output the optimal picking strategy.

[0005] Preferably, the state space includes order information, inventory information, location and status information of the stacker, and information of currently picked goods. The information contained in the state space is quantified and combined to generate a state vector for describing the current state of the warehousing system. The action space is the picking behavior that the stacker crane can perform in the current state.

[0006] Preferably, the objective function of the number of pallets required is The expression is: ; in, Indicates order From the tray Take away the goods the number of Indicates the quantity of inventory in the warehouse. Represents the current order task pool collection; Clearing the entire plate objective function The expression is: ; in, Indicates pallet Medium Commodity of inventory.

[0007] Preferably, the order requirement constraint expression is: ; The pallet inventory limit constraint expression is: ; in, Indicates order Required goods The number of

[0008] Preferably, the deep Q network DQN training the order picking solution optimization model includes the following steps: S401, initialize the network weights and set the hyperparameter learning rate, experience pool size, and randomly select the initial state of the state space and action space , based on the initial state Select the action to perform; S402: The stacker interacts with the warehouse environment and performs a picking operation, and the warehouse environment returns to the next state. and the reward function , that is, the new storage system state after executing the action, the current state, execution action, reward, and experience of the next state quadruple Deposit into the experience pool; S403, the next state Consider it as the current state, and repeat steps S401 and S402 until the number of experience pools reaches the preset threshold; S404, update DQN network parameters; Randomly sample the reward function and next state from the full experience pool, calculate the predicted Q value of the current DQN network for the next state and action, calculate the Q value of the next state of the target network, input the Q value of the next state into the Q-Network, minimize the mean square error loss between the predicted Q value and the Q value target through the optimization algorithm, update the weight and bias parameters of the Q-Network, interact the updated Q-Network with the environment, generate experience quadruple and store it in the experience pool S405: Take the next state as the current state and repeat steps S403 and S404 until the Q-Network converges. The converged Q-Network is the picking strategy.

[0009] Preferably, evaluating and optimizing the generated picking strategy includes the following steps: S501: Input the test order into the intelligent warehousing system, and the stacker crane performs the picking task according to the picking strategy generated by the DQN network; S502. Calculate various performance indicators during the test, including the number of pallets used, order processing time, picking path length, and pallet clearing rate, to determine whether the preset performance requirements are met. If not, repeat step S404. S503: Output a picking strategy that meets the preset performance requirements, which is the optimal picking strategy.

[0010] Compared with the prior art, the present invention has the following beneficial effects: The present invention establishes a mathematical model combined with a deep reinforcement learning (DQN) algorithm to solve the problem, intelligently decide the location for picking up goods and generate the optimal picking path. For multi-batch and multi-category inventory, it comprehensively considers the storage location of goods, pallet partitioning, order priority and system status, reduces the moving distance and number of pickup times of the stacker, thereby shortening the order processing time and improving warehousing efficiency. At the same time, the method provided by the present invention can also realize the clearing of the entire pallet, freeing up storage space, improving the utilization rate of storage space, and further reducing storage costs. In addition, the intelligent decision-making capability helps to improve the accuracy and flexibility of inventory management, and better meet customers' requirements for order processing speed and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a process framework diagram of the present invention; Figure 2 This is a schematic diagram of the core process of the DQN network of the present invention; Figure 3 This is a graph showing how the reward function of the present invention changes with iteration; Figure 4 This is a schematic diagram of the intelligent warehouse inventory grid model of the present invention; Figure 5 This is a comparison chart of the number of tray removals using the optimization strategy of the present invention and the traditional strategy. DETAILED DESCRIPTION

[0012] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0013] Example 1 In this embodiment, a solution is proposed for the automatic dispatch optimization of small-batch, multi-category warehouse orders. The solution can intelligently determine which aisle and pallet each item for each order should be picked up from, based on the specific needs of the order. This decision-making process needs to consider multiple factors, including but not limited to the storage location of the item, the partitioning of the pallet, the priority of the order, and the real-time status of the warehouse system. In recent years, deep reinforcement learning algorithms have demonstrated powerful intelligent decision-making capabilities in multiple fields. Through interactive learning between intelligent agents and the environment, they can automatically explore optimal strategies in complex environments, providing new ideas for solving warehouse order dispatch scheduling. At the same time, deep reinforcement learning algorithms can handle high-dimensional state spaces and action spaces and adapt to dynamically changing environments, making deep reinforcement learning have potential application advantages in complex and changing scenarios such as warehousing and logistics. The stacker crane is a key automated device in the warehouse system. It is responsible for moving within the aisles. Through the navigation system, it can automatically identify and reach the aisle and shelf location where the goods required for the order are located. Once at the designated location, the stacker crane will perform the picking operation, remove the required goods from the pallet and transport them to the next processing link. When a batch of orders arrives, the warehousing system needs to calculate the optimal picking plan based on the demand for each item in the order, ensuring the minimum number of pick-ups while using the least number of pallets to meet the order demand, thereby reducing the stacker crane's movement time and energy consumption. The core goal is to optimize the picking plan to reduce the number of pallets required, thereby improving overall warehousing efficiency.

[0014] Therefore, this embodiment provides a warehouse inventory management and optimization method based on artificial intelligence, which specifically includes the following steps: S1. Build a warehouse inventory grid model using a grid, including picking stations, shelves, and aisles. Each aisle is lined with multi-layer shelves. Each shelf contains several pallets, each pallet has several partitions, and each partition stores a specific SKU. This design allows a single palletizer to store a variety of goods, increasing storage flexibility and space utilization. The grid model is designed to label the picking station locations, shelf locations, and the locations and partitions within each pallet on each shelf, clarifying the spatial relationships and distances between each location. S2, obtain state space and action space; The state space includes order information (such as the demand for each item in the order), inventory information (the inventory of each item on each pallet), the location and status of the stacker crane, and information about the currently picked items. This information is quantified and combined to generate a state vector that describes the current state of the warehouse system. The action space is the picking behavior that the stacker can perform in its current state (e.g., selecting a certain product on a specific pallet for picking, where each action in the action space corresponds to a different pallet and product combination); S3. Determine the objective function and set constraints to establish an order picking plan optimization model.

[0015] S301, minimizing the number of pallets required and maximizing the clearance of the entire pallet as the objective function; Minimizing the number of pallets required means reducing the number of shelves and aisles visited by the stacker crane, thereby reducing movement time and energy consumption. The objective function for the number of pallets required is: The expression is: ; in, Indicates order From the tray Take away the goods the number of Indicates the quantity of inventory in the warehouse. Represents the current order task pool collection; Clearing the entire pallet means clearing the entire pallet as much as possible to improve storage space utilization and reduce storage costs. The objective function of clearing the entire pallet is to The expression is: ; in, Indicates pallet Medium Commodity of inventory.

[0016] S302. Setting constraints to meet order requirements, including order requirement constraints and pallet inventory limit constraints; Ensure that for each item in each order, the total amount of the item picked from all related pallets is at least equal to the order demand for the item. This can be achieved by mathematically modeling the order demand and the picking quantity to ensure the completeness and accuracy of the order. The order demand constraint expression is: ; Ensure that the total amount of a certain product taken from each pallet does not exceed the inventory of the product on the pallet during the picking process. This can avoid insufficient inventory or data inconsistency caused by excessive picking. The pallet inventory limit constraint expression is: ; in, Indicates order Required goods The number of

[0017] S303: Establish an order picking plan optimization model, and the expression is: ; in, represents the objective function of the number of pallets required, represents the objective function of clearing the entire plate, Represents the weight coefficient, by adjusting the weight coefficient To balance the importance of these two optimization objectives.

[0018] S4. Use the deep Q network DQN to train the order picking solution optimization model and generate the picking strategy; In the constructed order picking solution optimization model, the objective function is non-convex, non-smooth, and non-coercive, and it is difficult to solve using traditional methods. However, with the development of artificial intelligence, DeepQNetwork (DQN), as an algorithm that combines deep learning and reinforcement learning, uses deep neural networks to approximate the Q-value function, which can effectively handle complex state spaces. The core idea of DQN is to use deep neural networks to approximate the Q-value function. The Q-value function is used in reinforcement learning to evaluate the expected reward of taking a certain action in a certain state. Traditional Q-learning methods perform well when the state space is small, but in complex environments, the dimension of the state space is often very high, making the storage and update of the Q-value table impractical. DQN introduces deep learning technology and uses neural networks to approximate the Q-value function, so that it can handle high-dimensional state spaces. Therefore, in this embodiment, the order picking solution optimization model generated by DQN training is used, and the core steps are as follows: Figure 2 As shown; Construct a deep Q network (DQN), consisting of an input layer, hidden layers, and an output layer. The number of neurons in the input layer should match the dimensions of the state vector to receive and process state information. The hidden layer can use multiple fully connected layers or convolutional layers (if the state information has a spatial structure). The number of neurons in the output layer is consistent with the size of the action space, which is used to output the Q value corresponding to each possible action. S401, initialize the network weights and set the hyperparameter learning rate and experience pool size to ensure that the network has good initial performance at the beginning of training, and randomly select the initial states of the state space and action space , initial state It can be an empty order pool or a scenario containing a small number of initial orders. At the same time, the experience pool is empty and ready to collect experience samples. Based on the initial state Select the action to perform; S402: The stacker (agent) interacts with the warehouse environment (Environment) and performs the picking operation. The warehouse environment (Environment) returns to the next state. and the reward function , that is, the new storage system state after executing the action, the current state, execution action, reward, and experience of the next state quadruple Deposit into the experience pool; S403, the next state Consider it as the current state, and repeat steps S401 and S402 until the number of experience pools reaches the preset threshold; S404, update DQN network parameters; Randomly sample the reward function from the full experience pool and the next state , calculate the current DQN network for the next state The predicted Q value of the action is used to calculate the Q value of the next state of the target network, and the Q value of the next state is input into the Q-Network to minimize the mean square error loss between the predicted Q value and the Q value target through the optimization algorithm, update the weight and bias parameters of the Q-Network, and interact the updated Q-Network with the environment to generate an experience quadruple. And store it in the experience pool S405: Set the next state as the current state and repeat steps S403 and S404 until the Q-Network converges. The converged Q-Network is the picking strategy. S5. Evaluate and optimize the picking strategy, and output the optimal picking strategy.

[0019] S501: Input the test order into the intelligent warehousing system, and the stacker crane performs the picking task according to the picking strategy generated by the DQN network; S502. Calculate various performance indicators during the test, including the number of pallets used, order processing time, picking path length, and pallet clearance rate, to determine whether the preset performance requirements are met. If the preset performance requirements are not met, repeat step S404 or adjust the DQN network structure and hyperparameters. Repeating the training, evaluation, and optimization process continuously improves the performance of the strategy until a satisfactory optimization effect is achieved. This effectively solves the outbound scheduling optimization problem for small-batch, multi-category orders in a real warehousing environment and meets the company's warehousing efficiency and cost control requirements. S503: Output a picking strategy that meets the preset performance requirements, which is the optimal picking strategy.

[0020] Example 2 In this embodiment, the intelligent warehouse map model is constructed by grid modeling. Figure 3 As shown, it includes three picking stations on the far left, seven rows of shelves in the middle area, and six aisles. Each shelf has several pallets, and each pallet has several partitions. Each partition can only store one type of stock keeping unit (SKU) goods. We randomly selected 20 orders from real order data. To simplify the calculation process, we excluded pallets that only contained materials not needed for this batch of orders. This batch of orders required 24 types of materials, and there were 162 pallets containing these materials. The number of pallets was reduced from the initial 1779 to 162, a reduction of approximately 10 times. Using this simplified method, the inventory matrix was reduced from 1799 × 920 to 162 × 24, and the order matrix was reduced to 20 × 24. The calculation process is optimized to only choose whether to take a pallet among the pallets storing the corresponding materials. That is, when it is determined that the material is not stored in the pallet, the loop will not be entered. This greatly reduces the number of calculations from 3240 cycles to about 500 times, and the amount of calculation is reduced by 6 times.

[0021] During training, the reward function The value of Figure 4 As shown, by comparing the types and quantities of materials that need to be taken out with the types and quantities of materials actually taken out, it can be concluded that the types and quantities of materials taken out meet the requirements of the actual order.

[0022] Among the 20 orders tested, there were 27 requirements for different materials for different orders. According to the previous rules, it would take 27 times to take the pallet. After optimization, the pallet was taken out 24 times in total, reducing the number of pallets required. As the number of orders increases, the overall warehousing efficiency will be greatly improved.

[0023] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A warehouse inventory management and optimization method based on artificial intelligence, characterized in that: The following steps are involved: S1. Build a warehouse inventory grid model using a grid, including picking stations, shelves, and lanes. Each shelf contains several pallets, each pallet contains several partitions, and each partition stores a specific SKU. Mark the picking station locations, shelf locations, and the locations and partitions within each pallet on the shelf in the grid model. S2, obtain state space and action space; S3. Determine the objective function and set constraints to establish an order picking optimization model; S301, minimizing the number of pallets required and maximizing the clearance of the entire pallet as the objective function; S302. Setting constraints to meet order requirements, including order requirement constraints and pallet inventory limit constraints; S303: Establish an order picking plan optimization model, and the expression is: ； in, represents the objective function of the number of pallets required, represents the objective function of clearing the entire plate, represents the weight coefficient; S4. Use the deep Q network DQN to train the order picking solution optimization model to generate a picking strategy; S5. Evaluate and optimize the picking strategy, and output the optimal picking strategy.

2. The method for warehouse inventory management and optimization based on artificial intelligence according to claim 1, characterized in that: The state space includes order information, inventory information, the location and status information of the stacker, and information about the currently picked goods. The information contained in the state space is quantified and combined to generate a state vector for describing the current state of the warehousing system. The action space is the picking behavior that the stacker crane can perform in the current state.

3. The method for warehouse inventory management and optimization based on artificial intelligence according to claim 1, characterized in that: Objective function of the number of pallets required The expression is: ； in, Indicates order From the tray Take away the goods the number of Indicates the quantity of inventory in the warehouse. Represents the current order task pool collection; Clearing the entire plate objective function The expression is: ； in, Indicates pallet Medium Commodity of inventory.

4. The method for warehouse inventory management and optimization based on artificial intelligence according to claim 3, characterized in that: The order requirement constraint expression is: ； The pallet inventory limit constraint expression is: ； in, Indicates order Required goods The number of 5. The method for warehouse inventory management and optimization based on artificial intelligence according to claim 1, characterized in that: The deep Q network DQN training the order picking solution optimization model includes the following steps: S401, initialize the network weights and set the hyperparameter learning rate, experience pool size, and randomly select the initial state of the state space and action space , based on the initial state Select the action to perform; S402: The stacker interacts with the warehouse environment and performs a picking operation, and the warehouse environment returns to the next state. and the reward function , that is, the new storage system state after executing the action, the current state, execution action, reward, and experience of the next state quadruple Deposit into the experience pool; S403, the next state Consider it as the current state, and repeat steps S401 and S402 until the number of experience pools reaches the preset threshold; S404, update DQN network parameters; Randomly sample the reward function from the full experience pool and the next state , calculate the current DQN network for the next state The predicted Q value of the action is used to calculate the Q value of the next state of the target network, and the Q value of the next state is input into the Q-Network to minimize the mean square error loss between the predicted Q value and the Q value target through the optimization algorithm, update the weight and bias parameters of the Q-Network, and interact the updated Q-Network with the environment to generate an experience quadruple. And store it in the experience pool S405: Take the next state as the current state and repeat steps S403 and S404 until the Q-Network converges. The converged Q-Network is the picking strategy.

6. The method for warehouse inventory management and optimization based on artificial intelligence according to claim 1, characterized in that: Evaluating and optimizing the generated picking strategy involves the following steps: S501: Input the test order into the intelligent warehousing system, and the stacker crane performs the picking task according to the picking strategy generated by the DQN network; S502. Calculate various performance indicators during the test, including the number of pallets used, order processing time, picking path length, and pallet clearing rate, to determine whether the preset performance requirements are met. If not, repeat step S404. S503: Output a picking strategy that meets the preset performance requirements, which is the optimal picking strategy.

Images